Library of Congress Cataloging in Publication Data Main entry under title: Transcendental arguments and science. (Synthese library) Papers originally presented at a symposium held in July 1977, and sponsored by the Center for Interdisciplinary Research of the University of Bielefeld. Bibliography: p. Includes indexes. l. Knowledge, Theory of-Congresses. 2. Transcendentalism -Congresses. 3. Reasoning-Congresses. 4. Science-Philosophy -Congresses. I. Bieri, Peter, 1944II. Horstmann, RolfIII. KrUger, Lorenz. IV. Bielefeld. Universitat. Peter, 1940Zentrum fUr Interdisciplinare Forschung. BD16l.T68 121 79-4367 ISBN 90-277 -0963-7 ISBN 90-277-0964-5 pbk.
Published by D. Reidel Publishing Company, P.O. Box 17, Dordrecht, Holland Sold and distributed in the U.S.A., Canada, and Mexico by D. Reidel Publishing Company, Inc. incoln Building, 160 Old Derby Street, Hingham, Mass. 02043, U.S.A.
All Rights Reserved Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland No part of the material protected by this copyright notice may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording or by any informational storage and retrieval system, without written permission from the copyright owner Printed in The Netherlands
EDITORIAL INTRODUCTION
The goal of the present volume is to discuss the notion of a 'conceptual framework' or 'conceptual scheme', which has been dominating much work in the analysis and justification of knowledge in recent years. More specifically, this volume is designed to clarify the contrast between two competing approaches in the area of problems indicated by this notion: On the one hand, we have the conviction, underlying much present-day work in the philosophy of science, that the best we can hope for in the justification of empirical knowledge is to reconstruct the conceptual means actually employed by science, and to develop suitable models for analyzing conceptual change involved in the progress of science. This view involves the assumption that we should stop taking foundational questions of epistemology seriously and discard once and for all the quest for uncontrovertible truth. The resulting program of justifying epistemic claims by subsequently describing patterns of inferentially connected concepts as they are at work in actual science is closely connected with the idea of naturalizing epistemology, with conceptual relativism, and with a pragmatic interpretation of knowledge. On the other hand, recent epistemology tends to claim that no subsequent reconstruction of actually employed conceptual frameworks is sufficient for providing epistemic justification for our beliefs about the world. This second claim tries to resist the naturalistic and pragmatic approach to epistemology and insists on taking the epistemological sceptic seriously. Granting the widespread assumption that every attempt at establishing some sort of epistemological fundamentalism proves untenable under close scrutiny, this more traditional outlook tries to revive something like the Kantian notion of 'transcendental arguments' which are supposed to refute the sceptic by showing, as against conceptual relativism, that certain conceptual or linguistic frameworks have priodty over others, and that the application of certain concepts or linguistic structures is a necessary condition for all talk about 'knowledge' and 'experience'. Consequently, it is supposed that there is some point in forming a well-defined general concept of science that remains invariant throughout history. The volume is guided by the idea to make this contrast as explicit as possible and to probe into the general motives and implications of these different vii
viii
EDITORIAL INTRODUCTION
views. Within the range of possible topics which could be of interest in connection with this task, it concentrates on three themes. The first part of this volume concerns the structure and the function of transcendental arguments. It covers both a reconstruction of the Kantian idea of such arguments and the analysis of contemporary discussions of them by analytical philosophy in general. The second part deals with the conceptual foundations of science and tries to clarify the background assun1ptions and the methodological and epistemological consequences of different approaches within this field. The third part concerns the question whether it is possible to reconcile a transcendental approach with alternative positions. In order to discuss all these questions, which are of interest both for philosophers and scientists and stand in need of the competence of both, the editors of the present volume organized, in July 1977, a small working symposium on transcendental arguments and the conceptual foundations of science. The symposium was sponsored by the Center for Interdisciplinary Research of the University of Bielefeld and took place in Bielefeld (West Germany). The papers as well as the commenting papers in this volume have emerged from talks given at that symposium. The editors are deeply indebted to all authors for their active participation in the symposium and for their consent to the publication of their contributions in this volume. They are also indebted to the Zentrum fitr Interdisziplinare Forschung der Universitat Bielefeld for its very generous invitation and kind hospitality, as well as to the editors of this series, especially to Professor Jaakko Hintikka, for their willingness to include the present volume in the Synthese Library. PETER BIERI ROLF P. HORSTMANN LORENZ KRUGER
Bielefeld, July 1978
THE STRUCTURE AND FUNCTION OF TRANSCENDENTAL ARGUMENTS
MANFRED BAUM
TRANSCENDENT AL PROOFS IN THE
CRITIQUE OF PURE REASON
I. THE WEAKNESSES OF TRANSCENDENTAL ARGUMENTS
It is one of the effects caused by the critical philosophy, that metaphysics qua ontology and qua 'metaphysica specialis' has suffered discredit. Even the metaphysical systems of 'German Idealism' are, according to how they were understood by their authors, rooted in Kant's insight, that 'dogmatic' metaphysics is impossible. When, in Germany around the end of the 19th century, neo-Kantianism arose, the essence of Kant's critical philosophy was supposed to be its intrinsic connection with the natural sciences, especially Newtonian physics. The Critique of Pure Reason no longer found any interest as a systematic critique of all possible attempts to know the suprasensible, or as an attempt to rescue freedom of the will, which was seen as indispensible to morals. It was taken even less seriously as a destruction of a deductive ontology of the type of Christian Wolff. The effect of the first critique was so oveJWhelming that it has almost become commonplace to see the foundation of everyday or scientific experience as the proper task of theoretical philosophy. Had not Kant taught that all (theoretical) knowledge lies within the limits of actual or possible experience and that our concepts, including the mathematical ones, could not possibly have any sense and meaning, if the range of possible experience was left behind? He seemed to anticipate with this the fundamental thesis of Vienna Circle positivism, which was that every nonanalytic sentence which cannot be verified or falsified by experience is simply without significance. But, since, in this reading, Kant had linked the fate of his theoretical philosophy with the fate of Newtonian physics, it did have the merit of being scientific. But after Frege and Russell undertook to establish mathematics as a part of logic, and thereby as a purely analytic theory, and after Einstein overthrew Newtonian physics, the first Critique seemed to have become hopelessly obsolete. So to the analyst of the Critique of Pure Reason there seems to be left the unfortunate choice between an uninteresting critique of metaphysics and an interesting but out of date philosophy of science. Yet, since the appearance of Strawsons Individuals l there is, it would seem, a third possibility of understanding the Critique, namely, to read it as 3 P. Bieri, R.-P. Horstmann, and L. Kriiger (eds.), Transcendental Arguments and Science, 3-26. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
4
MANFRED BAUM
an essay in descriptive metaphysics which is intended as a description of the actual structure of our thought about the world, Le. of the conceptual scheme which is the foundation of all human experience and is not changed by history. In reading the Critique as an essay in descriptive metaphysics the danger dissolves that philosophy might become the 'owl of Minerva' of the empirical sciences in their historical development. In using this expression, Max Scheler in 1900 2 criticized the neo-Kantian way of understanding Kant even before Einstein's Theory of Relativity appeared, that is, before dusk had fallen over Newtonian mechanics. Thus, Strawson is justified against a (neo-Kantian) interpretation as given by (Collingwood and) Korner when he opposes 3 taking the principles of the pure understanding which are proven by Kant in his Critique merely as the presuppositions of the Newtonian physics of his time, which have to be replaced (according to Korner) by other principles. If Korner were right, then we would have to assume that it was Kant's ain1 only to discover the fundamental conceptual framework of his epoch, within which the scientists of that epoch had posited their problems and fonnulated their solutions. Such conceptual frameworks were never directly refuted, but rather silently abandoned as science progressed. But this would mean that Kant had not at all sought and found the universally necessary conditions of the possibility of any experience of objects. This interpretation flatly contradicts all that Kant had said about his aims and his achievements. The "merely historical view" (p. 121) is by no means an interpretation of Kant's theory of knowledge, but rather its refutation. Even if Kant had succeeded in getting to the metaphysical first principles of Newtonian natural science, his effort would have failed to lay down the principles of any future metaphysics that could be counted as science. So far Strawson is certainly right. 4 But it can be shown that an objection raised by Komer s against Kant's alleged transcendental method is, in fact, valid against transcendental arguments in Strawson's sense. There is the difficulty that Strawson never mentions 'transcendental arguments' in his book on Kant. His understanding of this phrase, therefore, must be gathered from his book Individuals. There, he gives us an example of a transcendental argument: "Given a certain general feature of the conceptual scheme of particular-identification which we have, it follows that material bodies must be the basic particulars." (p. 40) It is not quite simple to say what is transcendental in this argument. I take it as follows : You can say what the fundamental objects (of our experience) are, if they must fulfIll the conditions under which alone the identification of an object is possible for us (namely, a certain conceptual scheme). But the
TRANSCENDENTAL PROOFS IN KANT'S 'CRITIQUE'
5
decisive statement is yet to come. According to Strawson it is not the case that, on the one hand, a certain problem with the identification of objects is posited by a conceptual scheme, whereas, on the other hand, certain material objects make the solution of the problem possible (by fulfilling the conditions for the identification of possible objects). But, Strawson says, "it is only because the solution is possible that the problem exists. So with all transcendental arguments" (p. 40). This must mean that it is only because we have always been capable of identifying material objects that an investigation of the conditions under which we can do this (viz. of a certain conceptual scheme) is possible. If we keep this in mind while reading Strawson's book on Kant and look there for a parallel line of thought, we find it for instance in his description of a "transcendental investigation" (p. 18). Its object is "the conceptual structure which is presupposed in all empirical inquiry", or in other words, it is a 'theory of experience' in the sense that it discusses the conditions of any possible experience by an a priori method (p. 18). In his 'General Review' of the first en'tique, Strawson discusses 'synthesis' as the fundanlental concept of the deduction of the categories. He rejects this concept because it and its correlate, belief in disconnected data of sense as materials for the process of synthesis to work on, belong to an idealistic explanatory model of knowledge, Le. to a 'transcendental psychology', that cannot claim to be true. For if unconnected sense data and synthesis are antecedent conditions of empirical knowledge, they cannot be empirically known themselves. For a reconstruction of the deduction of the categories there remains, therefore, only one way by which it can be shown to be a viable argument; that is, by taking the deduction as an analysis of the concept of experience in general, by which it can be shown "that a certain objectivity and a certain unity are necessary conditions of the possibility of experience" (p. 31 f). This argument is a "strictly analytical argument" (p. 32, cf. p. 73).6 From these few remarks on Strawson's approach it is clear in what sense Kant's argumentation is a 'transcendental argument': (1) Strawson thinks it is necessary to reconstruct the Kantian theory of the conditions of any possible experience as an argument that is analytical, because it first assumes the concept of experience in general and then asks what makes it possible. (2) That experience is possible is something that follows from the (self-evident and therefore unmentioned) fact, that it exists, and this is seemingly something which is itself empirically known in contradistinction to the two antecedent factors mentioned above, which are supposed to render experience possible. Taken together, the two assumptions mean that the problem of
6
MANFRED BAUM
the identifIcation of certain objects of experience by means of a certain conceptual scheme is always already solved. These two presuppositions'bf Strawson's interpretation of Kant are shared by Korner. It is therefore not unfair to present his objection to transcendental arguments briefly, even without going into a further discussion of Strawson's interpretation. But Korner's objection is directed against Kant himself; therefore, he speaks of 'transcendental deductions' and not of 'transcendental arguments'. His objection is very simple: if the categories, or the principles derived from them, are not only the sufficient, but also the necessary, conditions of the possibility of experience, not only their aptness for explaining experience, but also the uniqueness of the conceptual sche.me that enjoys this aptness, has to be demonstrated. If in a transcendental deduction it is only shown that experience is possible via the categories or the principles and how it is possible, it is not at the same time demonstrated that only via these categories and principles the possibility of experience can be established. Other categories and principles could do the job as well. This objection against 'transcendental deductions' as understood by Korner? is, in fact, sound and valid, an10ng others, against Strawson's 'transcendental arguments'. For there is no difference in your method, whether you take, as a given fact that has to be explained, Euclidean geometry, Newtonian physics or everyday experience, and fmd out analytically their conditions in the categories and the principles. What n1ust be objected to in transcendental arguments in general is not that they are transcendental, but that they are analytical. And that Kant's n1ethod, at least in so far as it has a certain plausibility, is analytic is something that both Korner and Strawson believe. It is true Kant uses the analytic method in the Prolegomena. Here he asks: How is mathematics possible? How is pure natural science possible? How is nature possible? This last question is taken to be synonYlTIOUS with the question: How is experience possible? In so asking, Kant follows the analytic or rather regressive method, which is described in the same work: "Analytic method ... only means, that you proceed from that which is sought - taking it as given, and ascend to the conditions under which it is exclusively possible." (Ak. IV 276 n). All the above-mentioned questions are directed towards the conditions of the possibility of son1ething that is assumed as a given fact, namely, mathematics, physics, experience. Of course this is not to say that the Critique is not in any way a search for insight into the possibility of science and experience, but rather presupposes it as a fact. But it is only to say that the presentation of the results of the Critique in the Prolegomena,
TRANSCENDENTAL PROOFS IN KANT'S 'CRITIQUE'
7
for didactic reasons, takes its point of departure fF:om these results, which are established in the Critique by the synthetic (progressive) method. The Prolegomena proceeds this way in order to make as clear as possible the relevance of the results achieved by the synthetic method for science and experience. One of Kant's purposes in the Prolegomena was to set his theory off as clearly as possible from Berkeley's idealism. Berkeley had cast into serious doubt important fields of mathematics, as well as the application of mathematics in the cognition of nature and even the possibility of material nature itself. Therefore, Kant insists on those aspects of his own theory which are diametrically opposed to Berkeley's. And this way of presenting his philosophy can be found in the Critique itself, for Kant incorporated some passages from the Prolegomena into the introduction of the second edition of the Critique. Thus, it can be explained why this understanding of Kant's method is so widely accepted. Moreover, parts of the proof of the principles are, in fact, analytic and proceed by reasoning back to the conditions of the possibility of experience, as will be shown in Section III. This may be sufficient to explain why the characteristics of the Kantian argumentation were taken to be 'transcendental arguments'. The weaknesses of transcendental arguments are the weaknesses of the analytic method. This method, stemming from Greek mathematics, which has been discussed since the days of Plato and Aristotle, and described in detail in a much commented passage in Pappus,8 assumes as given a certain proposition and inquires into the premisses froill which it follows. Now, it is clear that by thus reasoning back to the reasons for the truth of a proposition (1) you can get to more than one reason, Le. you can arrive at more than one sufficient condition (proposition), from which the assumed proposition may follow (vide Korner). (2) None of these sufficient conditions must be true, because its corrollary is true (ex falso quodlibet). From this it follows that by the analytic method, the truth of the assumed proposition can never be demonstrated, except in those cases where the truth of the premisses has already been known. Therefore, the truth of the premisses does not depend on their establishing true propositions, but has to be presupposed for this. If the truth of the premisses has not been established elsewhere they can only be considered as hypotheses. Applied to Kant's argumentation, this would mean that, among others, the causal law had only the status of a hypothesis, which could account for the possibility of experience but need not be true for that reason. 9 (3) Finally, Kant would have formulated a mere tautologylO if the principles could claim to be valid merely as something that could explain the presupposed experience. The reason for the validity of a condition
8
MANFRED BAUM
of the possibility of experience would then be, that it was a condition of the possibility of experience. If experience is rendered possible by something that can claim to be valid only as a condition of the possibility of experience, then possible experience obviously rests ultimately on this, that it is possible. Because Kant knew this peculiarity of the analytic method he never used transcendental arguments in the Critique. Therefore, ifhe was not the author of a descriptive metaphysic or any other theory of experience, including a metaphysic of experience, the Critique of Pure Reason has to be taken the same way as its author has taken it: as a "metaphysic of metaphysics" (Ak X 269). In what follows, I try to show that there is a theory of experience inherent in the Critique ofPure Reason, but that one can only understand it adequately if it is taken as an answer to the question of the possibility of metaphysics, or more precisely of ontology. In the effort of answering this question, transcendental proofs are needed. They are meant to establish transcendental knowledge. The possibility of experience for Kant depends on such quasiontological knowledge. II. TRANSCENDENTAL PROOFS AS A TASK OF ONTOLOGY
For Kant, transcendental propositions have, or require, transcendental proofs if they are to be considered as true. Before distinguishing different sorts of transcendental argumentation and deduction, and discussing their mutual relations, we need a preliminary explanation of the concept of the 'transcendental' and the peculiarities of transcendental proofs in general. In a sense, transcendental philosophy is not different from ontology, Le. the philosophical theory of the most universal predicates of things in general. This use of the term 'transcendental philosophy', which depends on the traditional meaning of transcendental, is present when Kant says that it treats the understanding (and reason) themselves "in a system of concepts and principles which relate to objects in general without assuming objects that may be given" (B 873). Such a system is 'Ontologia' (ibid.). What is said here of transcendental philosophy - namely that it treats only of the understanding itself in its operations, i.e. independent of the objects that are given or not given to it - is a consequence of the fact that the understanding in its transcendental concepts and principles refers universally to objects in general, that is, to all possible objects. More precisely, the understanding treats these possible objects only in so far as they are possible, or in their possibility. Now, it belongs to the concept and essence and thereby to the possibility of
TRANSCENDENTAL PROOFS IN KANT'S 'CRITIQUE'
9
any object in general to be an object of kpowledge or at least of thought. Therefore, a universal theory of possible objects in general, an ontology, has to treat understanding as a faculty of knowledge and thought. For it is the understanding itself, with its operations, that is the only thing common to all possible objects whatever their conceivable differences may be. A system of the concepts and principles of the pure understanding is therefore at the same time a system of ontology, or rather: it would be, if the pure understanding alone sufficed for knowledge of an object. Since this is not so, the plan for an ontology is a mere presumption. Of things in general, without taking into account their possible givenness in a (with us necessarily sensible) intuition, nonanalytic apriori propositions such as e.g. the principle of causality cannot be shown to be true. Therefore, the 'proud name' of an ontology has to be given up in favour of the more modest name 'analytic of the pure understanding'. This analytic takes over the task of transcendental philosophy. Nevertheless, the former, but now problematic, sense of 'transcendental' (= 'ontological ') remains valid. The ontological predicates (categodes) are now concepts of the pure understanding and yield 'transcendental' pdnciples, Le. pdnciples "by which [a] general condition apdori is represented, by which it is exclusively possible that things may become objects of our knowledge in general" (Critique of Judgement B XXIX). In order to find the necessary apriori conditions, under which there can be any 'object of our knowledge' at all (besides an analysis of [sensible] intuition as a condition, under which an object can be given to us) there is needed an analysis of our understanding as a faculty of knowledge that is insufficient if taken by itself. With regard to this sort of self-knowledge, the much quoted sentence appears, according to which that knowledge is transcendental "which is occupied not so much with objects as with the mode of! our knowledge of objects in general,!! in so far as this mode of knowledge is to be possible apriori" (B 25). Pure intuitions and pure concepts as the two apriori modes of knowledge that each require the other are the themes of a transcendental philosophy. Together they could yield us the most universal knowledge not of things in general, but of things in general that can be given to us and known by us. One element of this transcendental knowledge that has to be established by transcendental proofs is the causal law. It is a peculiarity of transcendental proofs that there can be only one proof for every transcendental proposition. This follows from the fact that transcendental propositions cannot be founded on pure or empirical intuition of objects, but, if true, must be shown to be true simply by means of concepts of objects (without being analytic). A transcendental principle, such as the
10
MANFRED BAUM
causal law, is such a synthetic proposition in which a totality of subjects is represented by a concept ('everything that happens') and of which something is predicated ('has a cause'). This is a universal (affirmative, categorical) and apodeictic proposition. Since the proposition is universal, the subject-concept cannot express differences between events. There is only one concept to represent then1. And since the proposition is strictly universal, and permits no possible exception, it expresses, although it is synthetic, something that belongs to the concept of an event. Every event has, as such, a cause, or it belongs to the essence of an event to have a cause: this is what the proposition wants to say as a transcendental proposition. Since essence means nothing else but the 'internal possibility' of a thing, one can also say it belongs to the conditions which together make an event possible that it has a cause. All this follows from the proposition's being a transcendental proposition, Le. a proposition which universally asserts something essential of its object. Or rather, as Kant puts it: "In the case of transcendental propositions ... we always start from one concept only, and assert the synthetic condition of the possibility of the object in accordance with this concept" (B 815). In our example, the one concept is the concept 'event' and the object which is determined in accordance with this concept (Le. in its essence) is the event itself. In the predication 'every event is caused' an assertion is made about an object represented in the subject-concept, and the predicate-concept 'caused' or 'effected' does not contain a contingent property but something that belongs to the essence of the thing, 'event'. That is to say, the object as the thing that is represented by its concept ('event') would not be possible if it were not caused. ~ But this is not to say that it could be known by an analysis of the meaning of the concept 'event' that every event has a cause. For then the proposition would be analytic and would not need a proof, and therefore not a transcendental proof. The predicate indicates the condition of the possibility not of the concept 'event', but of events as such ('in accordance with' their concept). The predicate in the proposition is added to the subject-concept synthetically without being a contingent determination of the object 'event'. In other words, it is part not of the definition of the concept, but of the real essence of an event, to have a cause. It is included now in the concept of events to be a species of objects. If it can be shown that it belongs to the objectivity of the object 'event' to have a cause, any condition of the possibility of this object belongs to the essence of events, although it cannot be found in their specific concept. But by this consideration the proposition 'Every event has a cause' seems to become analytic again. For if it were part of the concept of the object in general to
TRANSCENDENTAL PROOFS IN KANT'S 'CRITIQUE'
11
have a cause, the causal law as a proposition about a certain species of objects would be analytic. The concept of a cause (or of being caused), however, is neither contained in the concept of an event nor in that of an object. As we shall see, it can be shown to be a condition under which alone there is something objective for our knowledge corresponding to the concept of an event. Thus, the causal law as a synthetic (nonanalytic) proposition says something about the condition of the possibility of objects that can be known by us according to a certain concept, namely ~ according to that of events. This proposition which is transcendental in the indicated sense can only be shown to be true in an adequate way if we succeed in demonstrating it as (the only) condition of the possibility of certain objects as objects that can be known by us. The events of which the causal law is universally valid are empirically given changes of states of substances which can be met with in tin1e. The perception of these events cannot by itself make legitimate the use of the category of causality, but can, at most, correspond to it. But if there is a transcendental proof for the causal law there must be a necessary connection between events as such and the concept of causality, by means of a third thing. More precisely, the being-caused of events must be the reason for the objectivity of events. But this connection can only be seen to hold by means of a third thing, because the law is a synthetic proposition. Events are, by their concept, thought of as a species of objects or as something objective. But it is not a tautology and therefore not an analytic but a synthetic necessity that they should have the objectivity for our knowledge that is contained in their concepts. It could always be the case that there were no objects at all that would correspond to the concept of an event, understood as an objective succession of states of substances. Kant always insists on the empirical character of the relation of perceptions which corresponds to the category. The principles anticipate experience, but only as far as its form is concerned. They are only the principles of investigation and not in themselves already determinate knowledge of the objects of experience. For the categories (which are the predicates of the principles) are only "indeterminate concepts of the synthesis of possible sensations" (B 751). For the causallaw 12 this means that it is a principle of the synthesis of possible empirical intuitions, under the guidance of which that which corresponds to the category of cause must be searched for. This is "the real upon which, whenever posited arbitrarily, something else always follows" (B 183). The 'something else' is the effected event, the cause of which I can determine, in the way just indicated, by an experiment.
12
MANFRED BAUM
There seems to be a contradiction between the transcendental character of the principles and the assertion that they (including the causal law) refer to possible experience in which alone that connection of perceptions can be found, which corresponds to the category in the realm of the object. Especially, the third thing that is to make the proof of a synthetic a priori proposition (e.g. the causal law) possible, seems to be quite unsuitable to its task, if it is to lie on the level of experience. This third thing must both, according to what has been said thus far, somehow lie within experience as well as be something nonempirical. It is the "possible experience" (B 794) or rather the "possibility of experience" (B 264) or even more precisely the "condition of time-determination in an experience" (B 761, cf. B 264). A proof of (among other principles) the causal law becomes possible by this third thing. Thus, the proposition "everything that happens, that is, begins to be, presupposes son1ething upon which if follows according to a rule" (A 189) can only be proven of objects of experience as such, and in their being able to be experienced. And this possibility of experience which is, among others, expressed by the causal law, is something that precedes not only every determinate experience, but also all possible objects of experience as a condition of their possibility. For objects that can only be objects for me in a possible experience necessarily underlie the conditions under which this experience is possible. Therefore, if the causal law is one of these conditions it is as an element of synthetic a priori knowledge not only possible, but even necessary (B 151). When, in fact, something as object of possible experience is presupposed, the causal law is an apodeictic proportion if only it can be shown that it is a condition of the possibility of experience. Kant once said that it is (in a way which will be made clear in Section III) a condition of the possibility of experience by stating that it makes possible the very thing by which it can be proven itself, namely experience. Kant says of the causal principle that it has the peculiar character "that it makes possible the very experience which is its own ground of proof and that in this experience it TIluSt always itself be presupposed" (B 765). This statement has long since been evaluated as the admission of a circle in the proof of this proposition (and as a document of the triumph of empiricism over apriorism). For if a proposition originally makes possible that by which it is to be proven, it seems to make itself possible or it seems necessary to presuppose it in its own proof. But the statement only says that the 'ground of proof of the causal1aw is experience in some sense, or rather that it can be established as son1ething valid of all real events, if it is established as a condition of the possibility of
TRANSCENDENTAL PROOFS IN KANT'S 'CRITIQUE'
13
the experience of these events, and that is,: not independent of the quality of being (possibly) experienced. In other wo~ds, it can be demonstrated as a presupposition of a determinate kind of experience, or it cannot be demonstrated at all. And this is to say that you have to proceed in its proof from the presupposition that experience in general is possible and then only show how a determinate experience is only possible by this principle. While this determinate experience is rendered possible by it, the proposition itself can only be proven with the presupposition that experience in general is possible. Thus, it is not true that the causal law renders experience possible while experience in its turn renders the causal law possible. Rather, it establishes the possibility of a certain experience (that of events) and thereby of its possible objects, but it is in itself only provable under the general presupposition (which must be true quite independently of this relation) that experience in general is possible. The point at issue now is, whether this latter possibility is given, and why. If the reason for experience's being possible consists in its being actual, then there is in fact a circle 13 here. For if the experience of events is something, the possibility of which I know only from its actuality or a posteriori, and if I recognize the universal causal law only as one of the presuppositions involved in it, its truth rests only on the particular cases for which it is valid. But these could only be considered as cases, if the principle could be shown to be true independent of them. Otherwise the circle would indeed be present: experience established the very thing by which it is established itself. In so arguing, the impossibility is assumed that a (according to Kant) strictly universal and apodeictic proposition is proven empirically. But,as we shall see, Kant asserts the possibility of experience quite independent of any particular actual experience. Thus, experience is not possible because we actually have experiences, but rather because it can be shown that the possibility of experience is necessary for reasons which are quite independent of the function of this presupposition in the proof of the causal law. Thus, the causal law establishes experience only in the sense that it is, in accordance with it, necessarily possible to find a cause for any given event. This cause cannot be determined by the causal law itself, but only by experience. Here the causal law establishes in a one-sided fashion, the possibility of finding out empirically certain causes, while it is in no way established or refuted by finding or not finding certain causes. III. THREE KINDS OF TRANSCENDENTAL PROOF
After the general exposition of the specific characteristics of transcendental proofs, we will now try to find out how this scheme works when applied to
14
MANFRED BAUM
the problems of the analogies of experience, the transcendental deduction of the categories and the transcendental aesthetic. I am taking them in this order because the proof of the analogies can be shown to rest on the deduction which, in turn, presupposes doctrines of the aesthetic. So not only the differences can be made clear between the three kinds of transcen<;lental proof but also their systematic interconnection. (a) As an example of the analogies I have chosen the second, in which it is proven that all events (which are only changes of states of a substance) happen according to the causal law, Le. the transition from the state A to the state B of a substance and the state B itself are always the effects of some cause. The proof runs as follows: First, it is shown by an analysis of the perception of a succession of appearances what it means to perceive an event. In this, the perceiver is conscious of the states A and B one after the other and connects the representation of A with the representation of B, which are hoth empirically given. But this combination as a connection of the contents of my consciousness depends on my in1agination and therefore is arbitrary in so far as the position of the representations in tin1e is concerned. When I represent A and when B or rather at what time I an1 conscious of these representations (that is whether A before B or B before A) is left open when they are taken merely as representations which are the content of my inner sense and its fonn, time. This follows from the nonperceptibility of time itself as an empty form of succession. It is impossible to decide empirically by a comparison of the perceptions A and B with determinate positions of time which one is the earlier one and which one is the later one. Thus it is up to me which I represent as the earlier and which as the later one. As far as the connection of two representations in empirical consciousness (necessary to every perception of events) is concerned, that is, as far as inner sense and its form, time, is concerned, there is no possibility of knowing the relation of the states which correspond to the representations empirically. Therefore, so far the objective relation of the appearances (i.e. properties of objects) of which the event consists cannot be known. By mere perception, experience of (objective) events is not possible. The second step of the proof consists in showing that empirical knowledge of an objective event is only possible, if one presupposes that the relation of the two states is so determined as to make it necessary to posit A before or after B. That is, the succession of my representations (or at least their order of dependency) is only then not left to my choice but determined by an
TRANSCENDENTAL PROOFS IN KANT'S 'CRITIQUE'
15
object (objects), if it is determined which is the preceding one and which the (necessarily) following one, Le. which one is the effect of a cause. So, only if the time-relation of the states is taken to be determined by the relation of cause and effect (i.e. by the concept of causality) can I know objective events empirically. The third part of the proof draws consequences from the unknowability of events by mere perception and fronl the requirement of a concept of the necessary connectedness of the appearances in order that there be an objective event for me. Here it is asserted that "only in so far as we subject the succession of appearances, and therefore all alteration, to the law of causality ... experience itself - in other words, empirical knowledge of appearances is ... possible" (B 234). Therefore, all these objects of experience are only then possible, if that law or the concept of causal determination is valid for them as a consequence of that act of 'subjection'. For this proof the assunlption is made that there are objective events to be empirically known. The objective necessity of the concept of causality is only then established, when it has been shown that these objective events, presupposed in this proof, cannot exist, if they are not such necessary and universal connections of temporally succeeding things as they are supposed to be according to the concept of the causal relation. If this is so, the concept of causality is not only useful for knowledge of objects (events) but even indispensable for the objects (the events) themselves which are knowable by me. Of course, the proof is given under the precondition that it is possible to know objects (including events) by experience. In this, the insights are used (a) that what is a necessary condition of experience is also a necessary condition of the objects of experience, and (b) that perception is insufficient for empirical knowledge of events. But the conclusion that it is the concept of causality which makes possible the experience which is presupposed is valid only under the two additional presuppositions (i) that we can know objects which correspond to representations given to us successively (either as simultaneous or successive objects) and (ii) that the objectivity of these objects is possible only by the categories. A justification of the categories as necessary requirements of objective knowledge and therefore of the possibility of experience in general cannot be given in the proof of the second analogy. It is here legitimately presupposed, because it has been given in the deduction of the categories. The presupposition that objects corresponding in a determinate way to our successive representations can be known by experience is equivalent to the presupposition that the (schematized) categories have been shown to be objectively valid. The only aim of the proof of the second
16
MANFRED BAUM
analogy is to show that under this general presupposition it is in this case the category of causality to which something"must correspond in experience (or rather in perception), if a certain kind of object is to be known empirically. Certainly in the deduction of the categories it has been shown that the necessary harmony of the appearances with the categories rests on an activity of the understanding in its relation to inner sense. But thereby it is only shown that it must be somehow possible to find something corresponding to the categories of the understanding among the appearances. What corresponds to the category can only be a formal characteristic of the appearances and not simply an appearance. For all empirically given representations cannot as such be anticipated or even produced by the understanding. The one accomplishment of the understanding as a faculty of knowledge can therefore only be such as to prescribe a determinate form to the representations (the content of inner sense). That is, it can only prescribe the order of connection which must be taken by those representations that are to have objective significance. Therefore in the case of causality it is not possible that there be nothing (to be found in a possible experiment) corresponding to the concept of a cause; but what corresponds to this concept is not decided by the activity of the understanding but can only be decided by perception. That there must be some possible perception corresponding to the criterion (the schema) of the category of causality can be due to the activity of the understanding and it must rest on this activity, if it is true that possible experience is itself necessary. This again is shown in the deduction. So far the proof has been analytic and proceeded by a search for the conditions of a certain experience presupposed at the beginning of it. We now discuss those passages of the proof in which after a long and detailed analysis of the experience of events the proper ground of the proof is exposed. By this discussion, the proofs dependence on the schematism chapter and on the deduction of the categories becomes obvious. This happens in the summary of the proof in the first edition (A 200ff, B 245ff "That something happens ... of such experience."). This summary follows the introduction of "a necessary law of our sensibility" (B 244) into the proof, and this law is also made use of in the above-mentioned summary. This law of our sensibility states that time, which underlies all perceptions and their connections through me, has itself a formal characteristic: In the whole of time any subsequent time is necessarily determined by the preceding time. I get to a later time only through the one preceding it, because the later time itself is exclusively possible by the preceding one by which it even becomes necessary. That is to say, the parts of time as far as they represent an ordered succession are themselves
TRANSCENDENTAL PROOFS IN KANT'S 'CRITIQUE'
17
a 'pure image' or a schema of the concept of the causality of a cause, the unschematized form of which is: something from which I can conclude something else in a determined way. The parts of time as far as they represent an ordered succession make the concept of a cause sensible and are for their part the condition for everything which appears to us as occurring in time. The "connection of times" (B 244) which consists in the fact that time" [in] a priori [fashion] determines the position of all its parts" (B 245) means for the appearances which can only be represented and known in accordance with it that among them the same order prevails t!lat exists within time itself. But since this continuity of times and the appearances in them cannot be empirically known by a comparison of the perceived appearances with the (in itself unperceivable) "absolute" time (B 245), as far as our experience and its objects are concerned it is not time that determines the position of the appearances in itself but, on the contrary, it is the appearances that detelmine not, it is true, time, but for one another their position in time: "that which follows or happens must follow in conformity with a universal rule upon that which was contained in the preceding state" (B 245). The meaning of this is: Everything that happens has a cause or is effected by something upon which it necessarily follows. This argumentation is for itself sufficient to prove the causal law (synthetically). In it no (analytic or regressive) use is made of the de facto impossibility of a reversal in the series of our perceptions which follows from the assumed objectivity of a perceived event. But, on the contrary, in the above argumentation this irreversible order of appearances is based upon a formal characteristic of time itself which prescribes a law for the appearances. But this is, of course, not to say that there can be no temporally succeeding perceptions the sequence of which depends on our choice. It means, however, that only those perceptions are perceptions of an objective event that correspond to that formal characteristic of time. All other perceptions have only a subjective significance, whether they stand in a reversible order or are de facto associations of representations or whether they are uniformly repeating combinations of facts. The perceptions selected from the set of all combinations of perceptions according to that criterion of regulated succession are obviously not given by our thought nor by time itself. They must be found empirically. In this experience only that which "can always be found in the connection of perceptions in accordance with a rule" (B 245) is regarded as an object. Now, that something of this kind, Le. an object, must be found empirically which corresponds to the concept of causal connection, this presupposition of all experimental investigation of causes and effects is not true
18
MANFRED BA UM
for the reason that otherwise I could not have an experience of events. 14 But, on the contrary, there must be experiences of events, because otherwise the formal characteristic of time corresponding to the concept of causality, its order, would have no empirical equivalent among the appearances. But this is what time must have, if it is to be the "form of inner intuition" (B 245) and thereby a fundamental condition of all perception. Under this precondition the causal connection of the appearances is a necessity holding for everything that can be empirically known as an event, although the regulated succession of the parts of time is not the empirical or perceivable condition of such an experience. The regulated succession of the parts of time is an a priori valid formal condition of all perception and of the objects that can be known by it (and certain additional means). But what occupies the a priori determined positions in time is only known by an empirical determination of the relation of the perceptions. Only that is cause or effect which can be found according to the method of performing an experiment that is contained in the causal law: "were I to posit the antecedent and the event were not to follow necessarily thereupon, I should have to regard the event 15 as merely a subjective play of my fancy" (B 247). This sentence implies two notions: (1) That I must posit something to find out whether it is the cause of an event; this follows already from the criterion (schema) of the concept of the causality of a cause. This criterion is "the real upon which, whenever posited, something else always follows" (B 183, cf. B 268). (2) The all important idea is that an event is something objective in contradistinction to the merely subjective combinations of imagination or to the association of our perceptions only by being something that can be produced arbitrarily by its cause's being posited. If I take something as an objective event I take it as something possibly producible by an experiment. This is a specific meaning of the universal proposition that reason "has insight only into that which it produces after a plan of its own" (B XIII, B XN). Only by the concept of the relation of cause and effect is it possible to make a difference between objective connections of appearances and subjective combinations of our perceptions. Moreover, this concept must have an object or it is necessarily valid, because it is a concept by which a formal characteristic of time, its regulated (irreversible) succession, is conceived which is a condition of all appearances by being the form of inner intuition of the knowing subject. (In the deduction of the categories it is shown how time as a formal intuition is itself possible.) Therefore, all objects which are events in time stand a priori in relations of cause and effect. Empirical judgments about what the events are and what is tt'1e cause of their happening can only be true, because there must
TRANSCENDENTAL PROOFS IN KANT'S 'CRITIQUE'
19
be causes and effects among the objects of empirical knowledge in general. So, the causal law is a transcendental proposition, because it predicates of one kind of objects, events, the condition not contained in their concept under which they can be obje~ts of our knowledge. The proof of the causal law is a transcendental proof, because it contains nothing but the determination of an object in general which is to be an event in accordance with this concept. In order words, the ground of the proof is that the objectivity of events in our experience is only possible by their being causally determined. The cognition of an event has possible empirical truth, only since there nlust be an object in our experience which is adequate to the concept of an event. And this is a transcendental truth. (b) The deduction of the categories, too, is a transcendental proof. What is to be demonstrated in it is that there are necessarily objects which correspond to the categories. Since it has been established in the transcendental aesthetic that every intuition which is accessible to us is sensible, those objects can only be appearances and their knowledge can only be experience. Therefore, it is true to say that it is shown in the deduction of the categories that only the objects of experience can be known to correspond to the categories. Nevertheless, the deduction of the categories is only indirectly a deduction of the possibility of experience. The categories make possible only experience and no other knowledge of objects as a consequence of their specific claim to be a priori knowledge. That is to say they are such concepts which can be shown to be true of their objects either a priori or not at all. The propositions in which the categories function as predicates have, if any, a universal and necessary validity for their objects. If the only possible candidates for these objects are the objects of experience, the deduction of the categories must prove that the objects of experience are possible, only if the categories are valid of them. The proof of the analogies of experience makes use of the concept of possible experience in quite a different manner from the deduction of the categories. In that proof it is shown that, for the empirical knowledge of certain objects, perception is insufficient and categories are therefore needed. This presupposes that such experience of certain objects is possible, and it is only asked which are the concepts (i.e. the conditions) by which it is possible, taking into account the insufficiency of perception for this task. In the deduction of the categories, on the other hand, it is shown that there are no conceivable objects to be known by us without the categories. (Categories are therefore objectively necessary for all our knowledge of objects.) That this
20
MANFRED BAUM
knowledge is knowledge of appearances and tnerefore experience follows from the apriority of the categories and the knowledge derived from them, Le. from their universal and necessary validity. By means of the categories I can anticipate experience; that is, the understanding can a priori prescribe the law to the objects of this experience. The deduction of the categories thus shows how the very possibility of experience is constituted a priori by the operations of the understanding (and of the forms of intuition, space and time). Of course, any particular experience is based on the general conditions of the possibility of experience. But this possibility of experience is, in the deduction of the categories, nothing the conditions of which are sought for. It is rather the answer to the question about the (synthetic) a priori knowledge that is possible for us. 16 After what has been said about the relation of the proofs of the principles to the deduction of the categories it must now be shown why there are no objects of perceptions without the categories, the existence of which objects is presupposed in the second analogy (except the passage B 244f). The deduction of the categories (and that of space and time) proceeds, it is true, by showing that they are the a priori conditions of the possibility of experience (B 126), but by this ambiguous formulation the deduction of the categories does not differ from the proof of the analogies. This difference becomes clear, if we realize that it is in the deduction that originally the objectivity of objects of experience in general is shown to be possible and even necessary for us. This means the same as showing why the categories are a priori valid, Le. valid of empirical objects. For this it must be shown how the understanding is itself by means of its concepts the author of experience (B 127), so that experience is nothing merely assumed in order to reason back to its conditions (which are, among other things, the categories). Rather the possibility of the objective validity of the categories is demonstrated on its own account (Ak VIII, 184). This demonstration proceeds by a syllogism from the definition of judgment and by recourse to the conditions of the formal intuitions of space and time (discussed in the 'metaphysical expositions'), I.e. by referring to intellectual and 'figurative' synthesis. By this it is shown that the categories are not the accidentally sufficient, but the only (except for space and time) necessary (a priori) conditions of experience and its objects. In this investigation obviously not certain objects (e.g. events) of a certain experience but the objectivity of objects in general must be treated. If the categories are to be the indispensable conditions of this objectivity, it must be shown how they are necessary to the notion of an object of those of our representations (perceptions) by which something is directly given to us
TRANSCENDENTAL PROOFS IN KANT'S 'CRITIQUE'
21
without the activity of the understanding, Le. 01' the (sensible) intuitions. They are representations of an object, it is true, but only in so far as they are qua representations merely determinations of the knowing subject (of its mind) and are only referred to an object by the subject. When this happens something is added to those determinations qua merely given representations by which they are originally represented as the determinations of an object. The representation of given representations qua determinations of an object is not contained in the given representations themselves. This representation implies the concept of an object, Le. that one representation which refers the given representations to that thing which is represented by them. Since the representations of the objective unity of representations is not given by the senses, this representation of the belonging together of several representations must be effected by the knowing subject itself. The representation of the combination of representations in the one representation of the object depends on an activity of combining. The representation 'combination' which is a condition of representing objects through different subjective determinations of the mind is therefore exclusively possible by that spontaneity (not productivity) of the representing subject which we call the 'thinking' performed by the 'understanding'. Thus, combination is not a given representation. Thinking as the original act of combining is a process working on intuitions or concepts (pure or empirical). The understanding is the faculty of combining (a priori) (B 135). Accordingly the concept of composition is "the only fundamental a priori concept, which originally underlies all objects of the senses in the understanding" (Ak XX 272, cf. XX 280). 'Composition' is therefore something of a super-category of which the different categories are only the species. As kinds of composition they require a given material to be combined, i.e. finally a sensible intuition (since we are not aware of having the faculty of intellectual intuition). An object thought by means of a concept, as far as its unity, i.e. the combinedness of its represented parts (or its form) is concerned, is a product of the understanding which primarily refers to the pure manifold of space and time. The understanding qua faculty of combining is the source of all synthesis. It is therefore 'original synthetic unity' and this is concretized as the unity of apperception. Self-consciousness is a precondition of all my representations of which I can make myself conscious as my representations. The consciousness of the identity of myself in all thinking is something that must be possible at any time if I am not to be divided into as many different selves as I have representations of which I am conscious. From this it follows as a law for all thinking that it stands in a necessary relation to the one self-consciousness which
22
MANFRED BAUM
is common to all representations of which I am possibly conscious. By this relationship to the one self-consciousness all my representations are a priori combined with one another, or rather: they are combined by me, because every combination is an act of the understanding. That this happens is the necessary condition for my being conscious of the identity of myself in the possible consciousness of everything of which 1 am to become conscious. And this is necessary for a being conscious of itself. By this reasoning, the highest principle of all spontaneous combining and of all conceptual or intuitive combination is arrived at: the 'original synthetic unity' as the 'transcendental unity of apperception'. Only under the precondition that I am or can be conscious of some synthesis of the possible or actual states of consciousness (synthetic unity of apperception) I can be conscious of the identity of myself (analytic unity of apperception). On these kinds of self-consciousness Kant bases the judgment and the concept. For our purpose, it is sufficient to link the definition of judgment with the synthetic unity of apperception. The nominal definition of judgment is that it is that relation of representations (concepts) which can be objectively valid (i.e. true). Its corresponding real definition is: a judgment is "an act by which given representations first become knowledge of an object" (Ak IV, 457n). By this act, the manifold of representations is combined according to that kind of connection which is necessary for the necessarily possible consciousness of the identity of myself in thinking of all the given manifold. This kind of connection is the unification of a given n1aterial of representations in the concept of an object in general which is different (not in kind but only in number) from the consciousness '1 think'. This concept serves as the rule for all determinate combination of representations which is expressed in the copula 'is' of a categorical judgment. This copula refers- to a combination of which we know that it can, as all combinations, only be brought about by an act of the subject itself. The original synthetic unity of apperception is the only conceivable principle of all combining of representations of which it can be said that it has necessity for all representing. And this necessity in the combination of given representations is exactly that which the concept of an 'object of my thinking' means. Whatever empirical intuitions may be given, they stand under that supreme law of belonging to the unity of the 'I think' and being combinable into the concept of an object which is only the notion of the synthetic unity of apperception but qua determinate notion differs from the empty notion 'I think'. That some material of representations must be combinable into the unity of possible objectivity, is necessary for the sake of the consciousness of the unity (identity) of myself. Therefore, one of
TRANSCENDENTAL PROOFS IN KANT'S 'CRITIQUE'
23
Kant's definitions of judgment reads: "A Judgm~nt is nothing but the manner in which given modes of knowledge are brought to the objective unity of apperception." (B 141) Apperception, consciousness in general (B 143) or original consciousness (B 161) fulfills the only possible sense of objectivity in the combination of given representations (concepts) as distinguished from all subjective combinations which are only the states of the perceiving minds. The necessarily possible consciousness of the identity of the thinking self is the only possible universal property (following from an internal necessity) of any understanding (not necessarily a human one). What is implied in this apperception is therefore valid for all thinking subjects. This means that the principles of the objective determination of (intuitions and) concepts, the fomls of judgment, must "all be derived from the fundamental principle of the transcendental unity of apperception" (B 142).17 If I represent the intuitions as determined according to these forms of judgment, I think an object of intuition by the categories. This thinking, therefore, happens with the same necessity that prevails in all representing in relation to the unity of selfconsciousness. Thus, everything given in a sensible (nonintellectual) intuition necessarily is subjected to the categories. This first part of the deduction is completed by the second part in which the fOffi1s of our sensibility and transcendental time-determinations (schemata, the products of 'figurative synthesis') are taken into account. It is already settled in the first part of the deduction that the objects which necessarily correspond to those categories are appearances. Only of them can the categories be shown to have objective reality, for only of these appearances can one say that their intuition must have a synthetic unity (which is necessary through the knowing subject) in order to be referable to an object. What is added in the second part are the specific human forms of intuition, namely, space and time. Therefore, all appearances must have a determinate position in space and time as formal intuitions, if they are to be taken as objects. Space and time are themselves subjected to the categories, although they are, as forms of intuition, entirely heterogeneous from the categories. Here, for the first time it is explained how and why that synthetic unity of time which was mentioned in the passage B 244f of the second analogy, namely the regulated succession of its parts, is a 'concept of time' produced by the activity of the understanding. "Thus the concept of time, although it is not the same as that of the composited, nevertheless is produced by it as far as its [time's] form is concerned." (Ak XIII, 471) The collective unity of the appearances, nature, must be in accordance with the synthetic unity of space and time. The appearances are something objective only in so far as they have a position
24
MANFRED BAUM
in space and time which is determined ~ priori. Nature as the totality of appearances is an ordered relationship of appearances in space and time made possible only through the categories. The transcendental proposition in this region was: the objects of a nonintellectual intuition or of all our intuitions which are conditioned by space and time are determined by the categories. This proposition was shown to be true in a transcendental proof. It proves of its object only that which belongs (synthetically) to it according to its concept. The object is in this case the 'object of our empirical knowledge in general' or, rather, nature as the collective unity of all objects of experience (appearances). Of this object it is shown that it has to be determined by the categories in order to answer that requirement of 'objectivity in general' which is and must be valid for any thinking understanding as such. (c) The third kind of transcendental proposition and its proof can be dealt with very briefly. Not all things in general but all objects of the senses, i.e. all appearances are in space and time. This proposition asserts something of its objects, the appearances, which is not contained in their concept, but which nevertheless belongs to them as such. As a transcendental proposition it states: it is essential to the appearances to be in space and time, because, if they were not, they could not be objects of the senses. Now the 'metaphysical exposition' of the fundamental characteristics of space and time shows that they both refer a priori to the objects of experience qua appearances, without being abstracted from these objects. It is the outcome of their 'transcendental exposition', that space and time a priori underlie all appearances of the outer senses and of inner sense since they are the forms of sensible intuition. By this reason, it is also established that the objects representable in space and time are mere appearances. The deduction of space and time, the proof of their objective validity, is therefore a transcendental proof. In it it is shown that all appearances as such necessarily are in space and time, since they could not otherwise be the objects of our senses which they are according to their concept. The three kinds of transcendental proof have thus been shown to answer the requirements of a transcendental proof in general. If the Critique ofPure Reason is to be taken as a 'metaphysic of metaphysics' it must be interpreted in an apparently rather scholastic manner. But only in reading it this way one avoids the circles and tautologies which are inherent in a theory of actual experience and which must lead to scepticism. The difficulties of Kant's solution of the problems of metaphysical, Le. a priori synthetic knowledge
TRANSCENDENTAL PROOFS IN KANT'S 'CRITIQUE'
25
are indeed great. They should not be increased by a misapprehension of his problems. Gesamthochschule Siegen NOTES In quoting from Kant's works Ak IV, 276 refers to vol. IV, p. 276 of the Prussian Academy edition. The two original editions of the Critique of Pure Reason (and of the Critique of Judgment) are referred to by A and B. I use the Kemp Smith translation of the first Critique with some modifications where it seems suitable. 1 P. F. Strawson,Individuals, London 1959. 2 M. Scheler, Die transzendentale und die psychologische Methode [The Transcendental and the Psychological Method], Leipzig 1900, pp. 56f. 3 The Bounds of Sense, London 1966, pp. 118ff. 4 But cf. Bounds pp. 23 and 28 where the intention here criticized is attributed to Kant himself. 5 'The Impossibility of Transcendental Deductions', in: Kant - Studies Today (ed. by 1. W. Beck), La Salle, Ill., 1969, pp. 230-244. 6 Strawson wants to get to the necessary conditions of experience by an analysis of the concept of experience. All definientia contained in the concept of experience are necessary to the possibility of experience. But by finding the characters of the concept of experience the possibility of experience itself can never be established. If, on the other hand, by the conditions the reasons are meant which make experience possible, these reasons can only be sufficient or insufficient. For the explanation of the possibility of experience among other things the synthesis rejected by Strawson is necessary. 7 Eva Schaper has criticized Komer convincingly. Her arguments partially agree with what is said here in Section III. Cf. E. Schaper, 'Arguing Transcendentally', Kant-Studien 63 (1972),101-116. 8 Cf. J. Hintikka and U. Remes, The Method of Analysis. Its Geometrical Origin and its General Significance, Dordrecht 1974, and the literature there indicated. 9 This objection was raised by E. Adickes in: Die deutsche Philosophie in Selbstdarstellungen [The German Philosophy in SelfDescriptions] , Leipzig 1921, vol. 2, p. 10. 10 This is F. A. Lange's objection in: Geschichte des Materialismus [History ofMaterialism], Leipzig 7 1902, vol. II, p. 131. 11 I accept Mellin's conjecture here. 12 lowe the awareness of the sigrlificance of the expelimental method for the proof of the second analogy and of the details of its structure to Prof. Klaus Reich. 13 J. Ebbinghaus was the first, to my knowledge, to have shown the untenability of the charge of a circle in Kant's proofs. Ebbinghaus' refutation of this reading of Kant's theory of knowledge is based on the relative necessity of the possibility of experience. Cf. his Gesammelte Aufstitze, Vortrtige und Reden [Collected Papers, Lectures and Addresses], Darmstadt 1968, esp. p. 103. For a recent version of the criticized reading cf. Patricia A. Crawford: 'Kant's Theory of Philosophical Proof', Kant-Studien 53 (1961/62), 257-268. On p. 262 the author says: "We know from the fact that the
26
MANFRED BAUM
experience is actual that it is possible. The experience being presupposed the transcendental principle is certain." 14 In B 244flthe proof no longer proceeds by reasoning back to a condition, but by reference to insights of the chapters on the schematism and on the deduction of the categories. 15 Here Kemp Smith's translation is incorrect. 16 Cf. Ebbinghaus I.c. 98. 1 7 Since this points to a derivation of the forms of judgment, I cannot understand why Patzig says (as against K. Reich): "ja, Kant sagt noch nicht einmal' es lasse sich eine solche Ableitung der Tafel der Urteilsformen aus diesem hochsten Punkt vornehmen" [Kant does not even say that it is possible to derive the forms of judgment from this highest point]. Cf. J. Speck (ed.), Grundprobleme der gro(3en Philosophen. Philosophie der Neuzeit II [Main Problems of Great Philosophers. Modern Philosophy II] , Gottingen 1976, p. 51.
RUDIGER BITTNER
TRANSCENDENTAL ARGUMENTS, SYNTHETIC AND ANALYTIC* Comment on Baum
How far may and should we keep to the original Kantian theory in developing a transcendental philosophy t~at is both tenable and illuminating in a contemporary setting? Right to the end, Doctor Baum replies. His account of transcendental proofs in the Cn'tique of Pure .Reason tries to defend every one of Kant's main doctrines in theoretical philosophy; or, at least, he tries to give the principal ideas by means of which the whole doctrine n1ight be defended in a more elaborate presentation. We all know how extremely hard this would be to do, if it could be done at all, and we should be grateful to Doctor Baum for pursuing this line of argument in full, so that there is a more solid basis for judging the prospects of a transcendental philosophy in the Kantian manner. Yet, considering the evidence Baum has offered, I should judge the prospects to be dim. As a defense of Kant's theory, Baum's attempt has failed. To argue for this general contention, however, a detailed examination of a large number of points in his paper would be called for, a task that could not be carried out in the allotted time. So let me concentrate on one point, Kant's doctrine of synthesis; and let me place this doctrine in the focus of some more general considerations on how a transcendental argument could proceed. That synthesis is the heart of Kant's theory does not need arguing, least of all against Baum. He rightly places the main burden of Kant's proof on the Transcendental Deduction of the Categories, and the Deduction, in turn, he finds relying heavily on synthesis. Viewed more closely, the idea of a transcendental synthesis is this. We have sensible intuitions by which an object is immediately represented. They are not produced by the understanding but given. As representations, however, they are nothing but determinations of the knowing subject; that they represent an object is not contained in the representations as given. To be a representation of an object, a representation has to be referred to an object. Referring representations to an object is not, 3;nd cannot be, given, so it must be effected by the subject. As there are many representations being related to one object, the activity of the subject may be described as a combining, and what it achieves, as a combination. But combining is not only a necessary condition of a representation's being a representation of an object; combining is also required if there is to be finite self-consciousness. For such a conscious27 P. Bieri, R.-P. Horstmann, and L. Kriiger (eds.), Transcendental Arguments and Science, 27-35. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
28
RUDIGER BITTNER
ness must be able to know, of every representaHon it has, that it has it. Let me say that a representation belongs to the domain of self-consciousness if the consciousness in question is able to know of the representation that it has it. Now it is only by my activity of combination of representations that every representation which in fact is mine can belong to the domain of selfconsciousness. And this is the same activity of combining that was required for representing an object. To quote Baum: "That some material of representations must be combinable into the unity of possible objectivity is necessary for the sake of the consciousness of the unity (identity) of myself."! This, in outline, is Baum's reconstruction of Kant's famous argument in Sections 15-17 of the second edition of the Critique. I do not accept it, for the following reasons. (1) That intuitions immediately represent an object (as indeed they must, according to their definition in A 19, B 33) is incompatible with their relation to an object being not given but due to the mind. If there is to be any use for synthesis, you cannot start from intuitions which, by definition, immediately refer to an object; you have to start from a manifold of data. Apparently, this is what Kant himself did in Sections 15 and 16 (although it is true that there are other considerations, dominant in Sections 19, 20 and footnote to 21, that would not require mere data, but intuitions). However, the assumption of sense-data is itself very questionable, for well-known reasons that I cannot discuss here. (2) On the other hand, it has not been shown that representations, even if taken as mere data, might not as well represent some relation to other representations or even some object. Kant maintains that the combination of a manifold can never come to us through the senses (da(3 die Verbindung eines Mannigfaltigen ii.berhaupt niemals durch Sinne in uns kommen kann, B 129); especially, that there is no such thing as representing something as connected in the object without a priori combination of ours (da(3 wir uns nichts, als im Objekt verbunden, vorstellen konnen, ohne es vorher selbst verbunden zu haben, B 130). But there is no argument for this. There is another reading of the Kantian sentence I quoted last. Stress the phrase "as connected in the object" (als im Objekt verbunden). Then the meaning is: A priori combination on our part is not needed for representing an object, nor for representing an object with its qualities, nor for representing different qualities connected in an object. Combination is needed if we are to represent qualities as connected in the object, that is, if we are to be aware of the connection of qualities in the object itself. Not mere representation but reconstruction of an object calls for an a priori act of combining.
COMMENT ON BAUM
29
Now this claim, sufficiently clarified, would certainly be more plausible than the former; it n1ay even prove to be self-evident. But it raises another difficulty. If synthesis is necessary, not for any representation of an object, but only for an awareness of its connectedness, general conclusions as to what it is to be an object for a finite mind shall not be forthcoming, since synthesis, in this case, seems to be a function restricted to special cases. One would have to show, then, that synthesis is actually not a special function, but somehow is involved, or presupposed, in any empirical consciousness. But this seems to be a task somewhat similar to, and certainly not easier than, the former one, viz. to prove that any representation of a connection (represented as a connection or not) is due to a combining activity of the mind. (3) It is n1aintained that every representation I have must belong to the domain of self-consciousness, and this is plausible enough. But it is further claimed that an act of synthesis of mine is needed to make my representations belong to the domain of self-consciousness, and it is not clear why this should be the case. The identity of myself and my awareness of my identity do not seem to require anything beyond the fact that the representations do belong to the domain of self-consciousness. It is not necessary that I know that they all belong to the domain, and so there is no reason for requiring my act of combination to make them belong to it. (4) Suppose it is true that synthesis is required for the unity of finite selfconsciousness. Suppose further that synthesis is required for relating representations to an object. Still it is not clear why this synthesis and that one should be one and the same. One may say with Kant that uniting representations to the concept of an object requires "unity of consciousness" in their synthesis (Einheit des Bewu~tseins in der Synthesis derselben, B 137). But of consciousness is not the same as unity of self-consciousness; some representations to the concept of this or that object is different from uniting all my representations to the domain of self-consciousness. The same argument applies if it is not identity but some weaker relation that is maintained, as e.g. objective synthesis being a special case of the synthesis of selfconsciousness, for they seem to be different not just in scope but in kind. I shall not deal with the further complications of the doctrine of threefold synthesis in Deduction A. Let me only note that Kant does not expect that doctrine to establish an identity of objective synthesis and synthesis of selfconsciousness. It is rather as a condition or foundation (Bedingung, Grund, A 106) of the former that the latter enters the stage of Deduction A, and that claim needs explanation and defense. Internal difficulties such as these tell sufficiently against the notion of
30
RUDIGER BITTNER
transcendental synthesis that Baum finds in Kant's theory. Yet there is another objection, to the analytically-minded philosopher the most important one, which rejects the notion from the outset. (5) We have no idea what is meant by a combining activity of the transcendental subject. Strawson did not put the point happily when he argued ad hominem "that we can claim no empirical knowledge" of transcendental synthesis, 2 for the defender of the notion will quite readily agree. The point is, rather, that we do not know anything, by experience or otherwise, about transcendental activities of the mind; and once we learn of it in transcendental philosophy, we cannot place it in the world as we know it. 3 The active transcendental subject is just that sort of gap in the world which it was supposed to prevent from breaking open. So we are only told a story that pretends, it is true, to uncover necessary conditions of knowledge, but fails to connect with what we know about the world. Talk of transcendental synthesis is understandable, of course, if you take it as a story, but it is incomprehensible as a philosophical theory. Remember the stranger from Elea: "It seems to me, Theaitetos, they have told us stories as if we were children ... , and when they talk about these things, do you ever understand what they mean?"4 Professor Bennett has offered a different interpretation of transcendental synthesis, rejecting any notion of a combining activity of the self. He claims that, besides explaining some passages in the en·tique, this interpretation makes of transcendental synthesis "a powerful philosophical tool". 5 It is called the 'analytic' interpretation of transcendental synthesis,6 and I shall follow Bennett by distinguishing two schools of transcendental philosophy, a synthetic and an analytic one, according to whether use is made of a combining activity or not. Whenever you know of an identity, for example, that the book here is the one you saw an hour ago, or that it is you who, having felt angry with someone yesterday, now does not care about him, you may have reasoned your way to the identification or not; in any case there are criteria of identity satisfied and you have a grasp of them. According to Bennett, this is transcendental synthesis: a conceptual complexity, viz. the criteria, which you must be able to handle for any empirical identification. In this sense, transcendental synthesis is involved in any empirical awareness of identity, and in this sense, Kant's statement is justified that "of all representations combination is the only one which cannot be given by means of objects" (da{3 unter allen Vorstellungen die Verbindungdie einzige ist, die nicht durch Objekte gegeben werden kann, B 130). To talk of transcendental synthesis in thi-s sense is not to tell a story about what has to happen in a supersensible realm in order to
COMMENT ON BAUM
31
make empirical consciousness possible, it is simply to describe what is involved in empirical awareness of identity, what it is to 1:).ave such awareness. 7 And this is the reason for calling it the analytic interpretation: transcendental synthesis, thus understood, is involved in empirical consciousness, so it only needs analysis to be brought to light. Note that transcendental synthesis on the analytic interpretation is still different from what Kant calls the analytic unity of appreception, or consciousness (B 133f.); and his claim that the principle of the necessary unity of appreception is analytic (B 135, 138) makes a different point again. I wish to argue that the clear analytic conscience, which is based on the claim to disclose only what is involved in empirical consciousness, is deceiving itself. The claim has not been justified. Nor isit self-evident. It is not the trivial one that empirical awareness of identity requires an ability for such awareness. The claim is, rather, that this ability is an ability to handle criteria of identity. We are not offered the trivial observation that sometimes we use criteria in identifying. Rather it is maintained that to know of an identity involves a grasp of identity-criteria, even if we did not in fact use criteria in coming to know the identity in question. Apparently, the procedure is that you tell what it is to have such awareness by describing how you would have come to possess it if you had done it "expressly", so to speak, for example, if identification had proved difficult, orifyouhad been defending the cqrresponding identity-statement. (One could call this a verification principle, but in a different sense than either Stroud's8 or Bennett's.9) This is a dubious procedure, I think. At any rate, Bennett's claim that grasp of the relevant criteria is involved in awareness of identity seems to be no better off than Kant's and Baum's claim that empirical knowledge involves transcendental synthesis as an activity of the knowing self. The analytic transcendental philosopher argues that empirical awareness of identity is inconceivable without transcendental synthesis, in his sense of the term, but so does the synthetic philosopher argue on behalf of his transcendental synthesis. For Baum, following Kant, referring to a synthetic activity of the mind is as little a "biographical remark" 10 as referring to a grasp of criteria is for Bennett. Both sides clai~ to discern what it means to have the awareness in question. However, both fail to substantiate the claim by producing strictly analytic statements of the unmarried bachelor type. Analytic and synthetic school, then, differ in means, not in method. There is nothing more analytic about analytic transcendental arguments, and if the word is to mean more than just "making no use of an activity of synthesis", expressions like "analytic argument" or "analytic philosophy" are misleading. I have in fact distinguished three meanings of "an~ytic" here. You may
32
RUDIGER BITTNER
call a transcendental argument analytic, if it does not refer to synthetic activities of the mind, or if it only determines what it is to have a certain piece of empirical knowledge, or if conclusions are drawn by means of analytic statements of the unmarried bachelor type. (I have been assuming that an analytic argument in the third sense is analytic in the second sense, too.) Now, there is a fourth meaning of the word, the one that Baum relies on it his paper. Here an argument is analytic if, instead of arguing from given premises to a conclusion, it finds out premises entailing a given conclusion. It is in this sense that Kant describes the Prolegomena as an analytic argument, which means in his terms that it follows the regressive method of exposition. Baum maintains that so-called analytic transcendental arguments are analytic in this sense and so are bound to fail the purpose transcendental arguments are supposed to have, at least with Kant. However, Baum's objections to that sort of analytic procedure in transcendental philosophy do not apply. Except for the ambiguous expression "analytic argument", there is no reason to suppose that Strawson, for example, is undertaking an analytic investigation of this kind. Baum. himself quotes Strawson as saying that the argument should prove "that a certain objectivity and a certain unity are necessary conditions of the possibility of experience"ll - necessary, not sufficient conditions, as Baum should have to expect. Baum's criticism does apply to the neo-Kantian inter· pretation of Hermann Cohen and his school,12 but to assimilate to it the sort of transcendental philosophy done by Strawson and other 'analytic' philosophers seems to me misguided. Of course, I do not deny that Strawson assumes a premise without argument; he could not fail to. Among his premises the conceptualizability of experience figures most prominently.13 But trivially enough, there are premises not argued for in Kant as well, the most important among them being the finiteness of self-consciousness, that is, its requiring a manifold given in intuition. 14 A close connection between these two is visible. This calls into question one of Baum's central points. He argues for a superiority of synthetic arguments in transcendental philosophy on the grounds that only such are capable of establishing an ontology, which is, at least with Kant, their real aim. However, I do not find a specific difference here that might justify an ontological preference for synthetic transcendental philosophy. It is j true, Strawson's conclusions are weaker than Kant's. But once you have, with Kant, abandoned hopes of an ontology which deals with objects as they are, irrespective of how we may experience them, once you have settled for an ontology which uncovers essential traits of objects of experience as such, you have no reason to deny ontological import to statements such as Strawson's describing "the limiting framework of ideas and
COMMENT ON BAUM
33
principles the use _and application of which are essential to empirical knowledge, and which are implicit in any coherent conception of experience which we can form". 15 There would indeed be a differenc~ in kind if Baum were right in arguing that analytic transcendental philosophy is not concerned with necessary conditions of experience and its objects. But this, as I have shown, is based on a confusion of two meanings of "analytic". Up to this point, the argument has been sceptical. Now suppose my doubts concerning synthetic as well as analytic reasoning in transcendental philosophy are justified, how are we to confront the situation? I can think of three ways. Giving up the whole investigation into foundations of knowledge is the first (and one sort of giving up of epistemology is what is called its naturalization). In recent years, there has been a growing tendency among philosophers to take the line of epistemological despair. There is, indeed, good reason for it, considering the disappointments epistemology has had to bear. Still, I do not think there is conclusive reason. The poverty of the answers we give does not tell definitively against the questions that made us embark on epistemology. However, the point will be argued pro and con in greater detail at later stages of this conference, so I shall not enter into discussion of it now. In the second place, it may be suspected that even if I am right in claiming that analytic arguments are not doing conceptual analysis in the strict sense, still, on closer inspection they will turn out to be somehow far more analytic than their synthetic rivals. Obviously, to make sense of the suggestion, one should have to explain in what sense, if not in anyone of the four I mentioned, analytic arguments are supposed to be really more analytic. I have no idea what a new sense could be. Of course, it is not just the use of the word "analytic" I am objecting to. There are quite a number of alternative terminologies available which would do about the same .~~ For example, it is said that something is "conceptually required" for the existence or function of something else; or that something "logically must" be the case if something else is; or that two or more features form a "conceptual structure". These expressions are often applied in cases where an analytic relationship in the strict sense, the third one, cannot sensibly be maintained. When they are so applied, they should be considered as metaphors, taken over from their usual field of application, strictly logical relationships, to different matters. And of course, there is nothing wrong with metaphorical talk in philosophy; you only have to explain it. So there may be good reason to talk of conceptual requirements and even whole conceptual structures when attempting to establish some sort of transcendental epistemology (or transcendental ontology, for that matter). But if there is, it should be given, and I do not think it has been.
~-.
34
RUDIGER BITTNER
But it may be that there is no such reason. So, finally, we may cease pursuing what has so far proved to be an illusion, viz. a truly analytic philosophy. There are no analytic transcendental arguments in the second sense of the term, that is, arguments determining what it is to have some sort of empirical knowledge. So we should, indeed, retum to Vienna in the 1920s: 16 we should revoke philosophy's commitment to logical analysis. In that case, I do not see any alternative to the traditional concept of philosophy as a non-empirical theory capable of yielding synthetic knowledge. Consequently, neither could "analysis" be used as a standard of philosophical method any longer. A distinction between dark and light sides of transcendental philosophy could not be drawn. Arguments would have to be assessed, not according to whether they fulfill the requirements of analytic method, but according to how coherent, how simple, how perspicuous a conception of experience they produce. How persuasive a story they tell, if you like. In that case, then, referring to something which we have never come across would not count decisively against a line of argument. After all, scientific explanations make such references too. A mental activity of synthesis, therefore, is not to be ruled out from the beginning. The fifth of the arguments advanced above (p. 30) must be rejected. It is the first four arguments that count, concerning, as they do, the coherence and usefulness of an account employing the concept of synthesis.
Universitiit Heidelberg
NOTES
* This is a revised version of the paper read at the Bielefeld conference. For critical comments on earlier drafts I am especially grateful to Peter Bieri, Dieter Henrich and David Krell. 1 ,See p. 22. 2 P. F. Strawson, The Bounds oj Sense, London 1966, p. 32. 3 See Bennett on placing transcendental activity in time: Jonathan Bennett, Kant's Analytic, Cambridge 1966, p. Ill. 4 Platon, Sophistes, 242 c, 243 b. 5 Kant's Analytic, p. Ill. 6 Kant's Analytic, p. 114. 7 Kant'sAnalytic,p.I13. 8 Barry Stroud, 'Transcendental Arguments', Journal ofPhilosophy 65 (1968), 247. 9 See p. 45 in this volume. 10 Kant's Analytic, p. 113. 11 The Bounds ofSense, p. 3lf.
COMMENT ON BAUM
35
12 Hermann Cohen, Kants Theorie der Erfahrung [Kant's Theory of Experience], Berlin 1871 (2 1885). 13 The Bounds of Sense, p. 20, 25. 14 See B 135, 145. 15 The Bounds of Sense, p. 18. 16 See Stroud, 'Transcendental Arguments', p. 256.
KONRAD CRAMER
A NOTE ON TR.ANSCENDENT AL PROPOSITIONS IN KANT'S CRITIQUE OF PURE REASON
Comment on Baum Right in the opening section of his paper on 'Transcendental Proofs in the Critique of Pure Reason'! Baum displays discontent with those interpreters who suppose that instead of asking: How are synthetic a priori judgments possible?, Kant is really asking: What are the necessary conditions of a possible experience?2 Baum is of the opinion that strategies which reduce Kant's question to that form cannot give a satisfying account of the sort of theory the 'Critique ofPure Reason has to offer. In the sanle section of his paper Baunl tries to give what he considers to be a decisive textual argument in favour of this opinion. He refers to Kant's own distinction, elaborated in the Prolegomena, between the analytic (regressive) and the synthetic (progressive) method in critical philosophy and directs our attention to the fact that Kant himself maintained that the analytic method is not only inappropriate to prove anything with regard to the possibility of experience but even insufficient to give us an adequate idea of the philosophical intention of the first Critique. 3 It is precisely by identifying Strawsonian transcendental arguments with arguments corresponding to the analytic method of the Prolegomena that Baum can make use of an objection raised by Korner against Kant's alleged transcendental deductions 4 and show that it is, in fact, valid against Strawsonian transcendental arguments, but not, as Komer thinks, against Kantian transcendental deductions. Korner's objection, Baum argues, could serve as an argumentation against Kantian transcendental deductions only if the method of the first Critique were analytic. This, however, is not the case. The weaknesses of transcendental arguments are therefore the weaknesses of the analytic method. Since transcendental arguments are purely analytical, they cannot prove anything in regard to the possibility of experience 'uberhaupt', although they may well prove something in regard to an experience the possibility of which is already granted by its factual existence. What must be objected to in transcendental arguments in general is not that they are transcendental, but that they are analytical. 5 Consequently Baum speaks with emphasis of transcendental proofs - and not arguments - as an aim of the Critique ofPure Reason. One of the issues of Rudiger Bittner's comments will be to give a critical examination of Baum's account of the structure of transcendental 'arguments and his proposal to'
37 P. Bieri, R.-P. Horstmann, and L. Kruger (eds.), Transcendental Arguments and Science, 37-43. All Rights Reserved. Copyright © 1979by D. Reidel Publishing Company, Dordrecht, Holland.
38
KONRAD CRAMER
relate them uniquely to the analytic procedure of the Prolegomena. I have some doubts of my own about the correctness of such a view. I think, however, that Baum is right in stressing the point that the first Critique offers, according to Kant's own understanding, a theory of experience in the form of a grounding of the possibility of any experience of objects in general precisely in that it answers the question: Is metaphysics as a scientific philosophy possible? Kant's question is therefore neither the question, what the necessary conditions of an experience of the sort we actually have are, nor the question, what the necessary conditions of any possible experience are. That there are necessary conditions of any possible experience of objects in general is in fact inherent in Kant's answer to the very question whether and how metaphysics as a systematic 'acroamatic' discipline (different from a 'descriptive' metaphysics as proposed by Strawson) is possible. And that question certainly is identical with the question whether there are propositions which are synthetic and a priori and yield knowledge of a certain type. Not all propositions which are qualified by Kant as synthetic and a priori and yielding knowledge need a transcendental proof. Only propositions of the type which Kant calls transcendental propositions are open to that specific sort of proof. 6 Baum has tried to reconstruct the logical structure and to defend the aim of such a proof in the second and third section of his paper. It goes without saying that his reconstruction takes into consideration from the very beginning that propositions which are to be proved transcendentally must be conceived of as both synthetic and a priori and yielding knowledge. I think it is not idle to examine a little closer than Baum's paper has done these joint properties of a transcendental proposition. If we examine what it is for a certain proposition to be both synthetic and a priori, we have to look at the concepts which serve as terms in this proposition. It is the 'sen1antics' of these terms rather than the validity of the proposition in which they appear to which our attention should be drawn in the first place if we want to understand the proof of such a proposition. According to Kant's own declarations, no concepts which contain in themselves anything empirical are allowed to enter into a proposition which is synthetic and a priori and has its place in transcendental philosophy (as a transcendental proposition certainly has). 7 The concepts which serve as terms in a transcendental proposition must therefore be expected to be a priori concepts. It is a necessary condition for a concept to count as an a priori concept that its 'content' does not admit of a derivation from experience, Le. of being 'abstracted' from specific sense-perceptions. If an a priori concept is to serve as a term in a proposition which claims to yield knowledge, it either contains,
COMMENT ON BAUM
39
according to Kant, in itself a pure intuition or it does not. If it does, the concept can be 'constructed' and thus yields mathematical knowledge. If it does not, we are confronted with a problem concerning the alleged status of the concept: being a priori, it cannot be instantiated by referring directly to specific sense-perceptions, Le. empirical intuition to which it applies. Being a concept serving as a term in a proposition which is supposed to yield knowledge, it must, according to Kant's basic doctrine that all knowledge relates to possible intuitions,8 still stand in some relation to intuition. There are, however, no intuitions available by which an a priori concept which serves as a term in a transcendental proposition could be instantiated directly. The relation of such a concept to intuition can not be conceived of as a direct instantiation of that concept. What Kant calls "objective reality", viz. reference, cannot be given to such a concept by means of exhibiting formal characteristics of, or empirical data within, intuition to which the concept should apply. It is exactly on account of this impossibility that Kant maintains that such an a priori concept "contains nothing but the synthesis of possible intuitions which are not given a priori."9 If such intuitions were given a priori, a concept containing nothing but their synthesis would again be a mathematical (or metamathematical) concept. The intuition to which our concept is related in an indirect way must therefore be empirical intuition. There is, however, no other conceivable relation of an a priori concept to empirical intuition than that fonnal relation which Kant calls synthesis of possible empirical intuitions. Such a concept relates to empirical intuition; inasmuch as it is the concept of such a kind of synthesis and nothing else. Moreover, such a concept will be a concept a priori inasmuch as it is not a concept of something representable in empirical intuition, but the concept of a synthesis of empirical data which, qua synthesis, in itself is not such a datum. If, and only if, an a priori concept contains nothing but such a synthesis we can use it in forming synthetic and a priori judgments which are philosophical in being fonned discursively, that is, according to concepts alone. A concept a priori which contains nothing but the synthesis of possible empirical intuitions is a category of relation. And it is a category of relation precisely for that very reason. For it goes without saying that empirical content proper of intuitions, their 'matter', by which things are supposed to be represented to us in space and time, can only be given to us a posteriori. The only concept, Kant argues, which represents a priori this empirical content is the concept of "a thing in general", and the synthetic and a priori knowledge of this thing in general can give us nothing more than the mere rule of the synthesis of that which sense-perception may give us a posteriori.
40
KONRAD CRAMER
"Synthetic propositions in regard to things in general, the intuition of which does not admit of being given a priori, are transcendental."lO This is Kant's definition of a transcendental propQsition. From this definition it follows that only the Analogies of Experience are members of the class of transcendental propositions." For the Analogies are the only propositions within Kant's system of propositions which fulfil all the conditions for a proposition to be transcendental in that they are (1) synthetic, (2) a priori, (3) claiming to yield knowledge and (4) have a category of relation as at least one of their terms. If this is true, we may well ask what the place of a category of relation in a transcendental proposition is. Baum has answered this question convincingly in giving an interpretation of the all-important passage which reads: "In the case of transcendental propositions, however, we start always from one concept only, and assert the synthetic condition of the possibility of the object in accordance with this concept."12 Baum has correctly pointed out that what Kant here calls "the condition of the possibility of the object" is synthetically asserted of that for which a concept stands from which the proposition starts by the very concept which serves as the predicate-term in that proposition. From this it follows that categories of relation serve as predicates in all transcendental propositions. If this is so, the next step in the analysis of transcendental propositions should be to ask: What is the content and the status of that very concept from which a transcendental proposition starts, i.e. of the concept which serves as the subject-term in a transcendental proposition? It is exactly when raising this question that a problem arises to which Baum, in his reconstruction of the logic of transcendental proofs, has 110t paid sufficient attention, if I am not mistaken. Since Baum has focused on the Second Analogy of Experience, it seems fair that I put my question as follows: What is the exact meaning and status of the concept of event (Geschehen or Ereignis) or of change (Veriinderung) which both function, as it were, interchangeable as subject-concepts of the Second Analogy.13 Change is defined by Kant as "a way of existing which follows upon another way of existing of one and the same object" .14 In the light of this definition two points should be clear: (1) It is not part of the meaning of the concept of change that all that changes is thej effect of some cause. To this point Baum has drawn the attention. From this point alone it follows that the Second Analogy - "All changes are effects" - is a synthetic proposition. (2) In order to exhibit the very meaning of the concept of change we have to make use of the concept of the relation of substance and accident. This is made clear by
COMMENT ON BAUM
41
Kant when he says that it follows from the definition of the concept of change that all that changes continues to exist during the time of its change and that only its states undergo an existence-change. 1s In order to use the concept of change meaningfully and hence be able to form a proposition of the type of the Second Analogy we must therefore be entitled to use another category of relation, different from that of cause and effect. But how can we be justified in using the category of the relation of substance and accident? It seems that we are pushed back to the First Analogy of Experience in order to be in the position to understand what is meant when we use the subjectconcept of the Second Analogy. With this I do not wish to maintain that the strong version of the First Analogy - "All existence-change is merely change in a (sempiternally existing) pennanent" - must be considered as true in order to render the concept of change the n1eaning it has. 16 A moderate version - a substance i-version in Bennett's terms 1 7 - of the First Analogy will, however, have to be accepted a.s a precondition for any meaningful use of the subject-concept of the Second Analogy. ThIS brings me to my final point. Not only is the predicate-concept of that transcendental proposition which is known as the Principle of Causality a category of relation. A category of relation is also part of the meaning of the subject-concept of this principle. We can explain this by saying that the propositional content of this principle depends on another transcendental proposition (namely, a n10derate version of a Conservation Principle), in which the category which is part of the meaning of the subject-concept of the Principle of Causality serves in turn as the predicate-concept. Such an explanation is, however, not to solve but to state a problem. The problem is that in order to accept such an explanation we must give the First Analogy an independent interpretation which shows that it itself is a transcendental proposition. The crucial point in such an interpretation is the subject-concept of the First Analogy, namely existence-change (Wechsel) or succession (Sukzession) of appearances (Erscheinungen ).18 It does not matter much whether in analyzing this concept we assign the term appearance an objective or' a purely subjective meaning. What matters is that the very concept of existence-change or succession, far from being a concept which "contains only the synthesis of possible intuitions which cannot be given a priori", simply is the concept which refers to the factum brotum that intuitions which cannot be given to us a priori follow one upon another in what Kant himself calls empirical consciousness. - TIns has a remarkable consequence: The subject-concepts of transcendental propositions cannot even be used meaningfully without recourse to the fact that we are acquainted
42
KONRAD CRAMER
with more than one sense-perception. This fact cannot, in Kant's theory of knowledge, be deduced by a transcendental proof or any other argument. It is a mere contingent fact which has to be taken up 'empirically' in reflecting on the sequence of 'empirical intuitions' in our stream of consciousness. If this is true, one of the problems with transcendental proofs in the Critique of Pure Reason is not so much that they are claimed to be proofs of a specific structure, different, as it were, from purely analytical arguments. The problem rather is how to warrant Kant's claim that those propositions which can only be prooved transcendentally fulfil the necessary condition of being possible candidates for such a proof, namely of being transcendental propositions. If the content of the subject-concept of the First Analogy were incompatible with this alleged status as an a priori concept, then the proposition in which it enters as a term could no longer be considered as a transcendental proposition. I miss a clear conception of this problem, of which Kant himself was well aware, although he did not solve it,19 in Baums paper. A satisfying exposition, if not a solution of this problem in interpretation would, however, be a precondition of his defense of transcendental proofs in the Critique of Pure Reason.
Munster NOTES This volume, p. 3ff. So T. E. Wilkerson, Kant's Critique of Pure Reason, Oxford 1976, p. 13. Wilkerson falls in with Strawson's substitution of Kant's famous question for a version which defines Kant's task as "the investigation of that limiting framework of ideas and principles the use and application of \vhich are essential to empirical knowledge, and which are implicit in any coherent conception of experience which we can form." (The Bounds of Sense, London 1966, p. 18.) For Strawson's own reservations against the synthetic a priori, cf. ibid., p. 43f. 3 Cf. Prolegomena, Academy Edition vol. IV, Vorrede, p. 263; § 4, p. 274; § 5, p. 276, 279. 4 Stephan Korner, 'The Impossibility of Transcendental Deductions', Kant-Studies Today (ed. by L. W. Beck), La Salle, Ill. 1969, pp. 230-244. For a more recent version of Komer's criticism of Kantian transcendental deductions see his ''Ober ontologische Notwendigkeit und die Begrtindung ontologischer Prinzipien' ['On Ontological Necessity and the Justification of Ontological Principles'], Neue Hefte [iir Philosophie (ed. by R. Bubner, K. Cramer, and R. Wiehl), XIV, Zur Zukunft der Transzendentalphilosophie [The Future of Transcendental Philosophy], Gottingen 1978, p. Iff., especially § § 1-3. (The volume contains, among others, articles to the topic by Roderick Chisholm, Moltke S. Gram, and Richard Rorty.) 1
2
COMMENT ON Bk'UM
43
5 Cf. Baum, p. 6. Both Komer and Strawson are spoken guilty by Baum of sharing the false belief that the method of the first Critique, "at least in so far as it has a certain plausibility" (p. 6), is analytic. 6 Cf. Critique ofPure Reason, A 782ff., B 810ff. 7 Cf. ibid., B 28. 8, Cf. ibid., A 719, B 747. 9 Ibid. (italics added). The German expression which is here translated with 'contains' is 'enthiilt,. Kemp Smith gives 'includes', which has, however, another connotation. 10 Ibid., A 720, B 748. 11 I shall not enter here into the problem what, under this proviso, the propositional status of the Principle of the Axioms of Intuition (B 202), of the Principle of the Anticipations of Sense-Perceptions (B 207), and of the Postulates of Empirical Thought in General (A 218, B 265) is. 12 Ibid., A 787, B 815. 13 Cf. ibid., A 189, B 232, and passim. 14 Ibid., A 187, B 230. 15 Ibid., 'Existence-change' is Bennett's translation of the German 'Wechsel'. Cf. Kant's Analytic, p. 187. For a more detailed analysis of the content and status of the concepts which enter as terms into the Analogies of Experience and of the problems which result from here see my articles, 'Kant's Definition of the Concept of Change and the First Analogy of Experience', Actes du Congres d'Ottawa sur Kant dans les traditions Angloamericaine et Continentale tenu du 10 au 14 Octobre 1974, Proceedings of the Ottawa Congress on Kant in the Anglo-American and Continental Traditions, held October 1014, 1974 (ed. by P. Laberge, F. Duchesneau, and B. E. Morrisey), Ottawa 1976, pp. 364-381; "Non-Pure Synthetic A Priori Judgments lin the Critique of Pure Reason', Proceedings of the Third International Kant Congress (ed. by L. W. Beck), Dordrecht 1972, pp. 247-254 (reprinted in Kant's Theory of Knowledge (ed. by L. W. Beck), Dordrecht 1974, pp. 62-70.)'. 16 I have argued in the first article mentioned in note 14 that this is not the case and that Kant never was of the opinion that it is. Cf. loco cit., p. 379, and Reflexion 6403 (1790-95), to which Dryer has drawn the attention in his excellent, though not well enough known, book on Kant, Kant's Solution for Verification in Metaphysics, London 1966,p.366. 17 Loc. cit., p. 182. 18 Cf. Critique ofPure Reason, B 224, B 233. 19 It can be shown that Kant's theory of the "predicables of the pure understanding" was to solve this problem. Cf. the second article mentioned in note 14.
JONATHAN BENNETT
ANALYTIC TRANSCENDENTAL ARGUMENTS
1. LOCKE ON THE OBJECTIVE WORLD
Someone who thinks that his own inner states are the basis for all his other knowledge and beliefs may wonder how anything can be securely built on this foundation. He need not actually doubt that his own edifice is securely founded, though he may pretend to have doubts about this in order to consider how they could be resolved if they did occur. This person is a 'Cartesian sceptic'. which implies that he is not sceptical at all. He is untouched by such crude English moves as Locke's protest that "nobody can in earnest be so sceptical", or Moore's holding up his hand as proof that there is a physical object. Such intellectual bullying is irrelevant to a serious inquiry into how epistemic foundations relate to epistemic superstructure. Locke's answer was that "There is an outer world" is a good explanatory hypothesis: various facts about my inner states, he thought, are best explained by the theory that there is an outer world. For example, from the fact that some of my 'ideas' occur without my willing them to do so, Locke infers that "it must needs be some exterior cause ... that produces those ideas in my mind" (IV.xi.5.). Let us concede for purposes of argument that if some of my 'ideas' are involuntary then something other than myself exists; still, nothing could follow about what exists other than myself. Can we strengthen the conclusion by strengthening the premises? Could a more contentful belief about an outer world be defended as explaining certain further facts about my inner states, e.g. about the order or regularity which they exhibit? Locke does argue like that, but unfortunately he pollutes all his premises - which should be purely about inner states - with an admixture of statements about the outer world; for instance, he uses the premise that men with no eyes have no visual states. But that seems to be an accidental defect in Locke's treatment. He could have cleansed his premises, as Hume nearly did, so that they spoke only of the order, coherence etc. of one's inner states; and I am sure that if he had done so he would still have argued that those facts are best explained by the hypothesis that there is a world of physical objects. Could such an argument be valid? 45 P. Bieri, R.-P. Horstmann, and L. KrUger (eds.), Transcendental Arguments and Science, 45-64. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
46
JONATHAN BENNETT
That depends upon how it handled the following problem. If one is to argue in the form: 'On the strength of data D, we are entitled to accept theory T, because T is the best explanation for D' , one must have an account of what makes this explanation better than some other. Why, for instance, is the physical-world explanation for the orderliness of my inner states better than a Cartesian-demon one? Sometimes one explanation is superior to another because there are facts which show it to be more plausible, more likely, more in tune with reality, than the other. There is a sudden loud noise, and we wonder 'Was that thunder or an explosion?'; and someone who knows the region well can tell us which explanation is more likely. But I could not have that sort of reason for preferring the outer-world theory to a rival explanation of my inner states, because there is nowhere for the required supporting facts to be found. They cannot, without begging the question, be sought in the outer world; and since they must show how my inner states relate to something else th'ey cannot be found purely within my inner states either. And the outer and inner exhaust the territory. This is quite different from the problem of explaining one particular noise, or even all noises of a given kind, or for that matter all noises whatsoever: each of those problems is limited, leaving plenty of territory not covered by the problem and thus available as a source of helpful information - we have access to the causes of noises otherwise than through hearing them. Whereas the problem of explaining the orderliness of my inner states is so broad that it sprawls over all the possibly relevant territory. Locke seems not to have noticed this difficulty about what justifies one in preferring one explanation to another. (Sometimes he writes as though the outer-world theory were the only possible explanation for the facts about inner states: but that is a rhetorical extravagance.) Nothing in his discussion rules out the idea that the outer-world theory is the best because it is the most plausible, the one with the most independent factual support; and so even if he had cleansed his premises of their question-begging elements Locke would not, I think, have been on the track of a viable solution to the outer-world problem. I now leave Locke, and stay with his problem. 2. ANOTHER MERIT IN EXPLAN ATOR Y THEORIES
One theory can he superior to another in being simpler or more powerful or both, and that basis for comparing two theories does not appeal to any further range of relevant fact. If we don't have access to any further range of relevant fact, it seems that we must base our theory-preferences upon considerations
ANALYTIC TRANSCENDENTAL ARGUMENTS
47
of power and simplicity. It also seems that in such a case the only ground we can have for accepting a given theory is just that it is superior in power and economy to any known rival explanation of the same data. That, I take it, is an ingredient in a widely accepted account of what entitles one to accept high-level theories in physics. A restricted item of highlevel theory may be defended on the ground that certain facts make it likely or plausible, but those 'facts' must involve other bits of high-level theory which are being taken for granted. If what is in question is the whole level of theory which pertains to sub-atomic items, say, the only way we can defend it is by showing how it helps us with lower-level data, and especially how it facilitates lower-level predictions. Along with the high-level theory we accept many conditionals running upwards and downwards between the two levels; this creates high-level inferential routes from some low-level statements to others, and these routes can be useful, to put it mildly. For instance, someone who judges that the H-boITlb he has made is in good working-order is moving from some observed facts to a prediction about further observed facts, and he couldn't easily do this except by going through abstract physical theory. And that, it seems, is our whole justification for accepting that level of theory namely that it organizes the lower levels for us better than does any known rival. Does that make the high-level theory genuinely explanatory? Well, it is explanatory in the sense that it confers conceptual unity on the lower level, and thus facilitates predictions, eases memory, and produces intellectual satisfaction. Apparently that is as much explanatoriness as it could possibly have. But it is natural to want something more: a genuinely explanatory theory, one is inclined to think, must be not merely useful but true. The contrast between utility and truth, echoing the quarrel between 'instrumentalism' and 'realism', needs to be discussed a little. Let T be some theory which does a good job in conceptually organizing certain data, and indeed a better job than any known rival theory. The thought that T may nevertheless not be true might be cashed out in any of three ways. (1) Perhaps logical space contains a theory which organizes past and present data even better than T does, but which we haven't thought of. (2) Perhaps there are available data - data which we would have if we acted in certain ways - which, when added to all the data we do have, would show T to be inferior to some rival in power or simplicity or both. (3) Perhaps in future we shall have data which, when added to the data we now have, will show T to be inferior to some rival. The idea that a merely useful theory may not be true could be fed by any of those sources, Le. by an awareness that a
48
JONATHAN BENNETT
superior rival might_ be found through intellectual effort, perhaps in combination with experimental effort or sheer passage of time. There can be no quarrel with a 'realisnl' which reminds us of these possibilities, and subjects every currently favoured theory to the 'sceptical' thought that it might be disloged by a superior rival. But there ren1ains the stronger kind of 'realism' which says that even if theory T is in fact superior to any rival which could be devised, in relation to any data which will or could be encountered, there is still a question as to whether the theory is true. This involves the thought: 'Perhaps the real principles which govern reality are somehow less simple and sweeping than are the regularities which appear to us or which we could encounter by letting time pass and by experimental effort'. Someone who asks us to take that thought seriously should explain what he means by a theory's being 'true', or how he takes 'the principles which really govern the universe' to be conceptually different fronl 'the regularities which are or could be encountered in our experience of the universe'. Since I don't think that he can coherently explain these things, I think that this strong 'realisnl', and its associated 'scepticism', are incoherent. But I shan't insist upon that. 3. VERIFICA TIONISM
I use 'verificationism' as a label for a certain view about my entitlement to believe that there is an objective realm. One of its components says that I anl entitled to accept that theory because the latter is superior to any known rival in its power to confer conceptual unity on my inner states. For example, the theory which says simply that my inner states are caused by a demon whose pleasure it is to give me false beliefs, though very simple, has almost no power to unify, organize, generate predictions and so on. There could be a much more powerful demon theory, because at any time when T is my actual theory of the outer world there is a rival theory which says 'My inner states are and always will be caused by an undeterrable demon whose pleasure it is to make me believe T'. That will explain everything that T explains; but it takes slightly longer to do so, and brings no compensating advantages. Of course I might come to be in states which were less satisfactorily handled by T than by its demonic cousin; and the verificationism I am presenting does not rule out that possibility, any more than it does the possibility that even now there may be some superior rival to T which I haven't thought of. The other component in verificationism is its ruling out of the sort of scepticism which is generated by what I shall call 'transcendental realism'. The
ANALYTIC TRANSCENDENTAL ARGUMENTS
49
latter is the view that the truth or falsity (if T transcends all the facts about inner states which I do, will or could have; so that even ~f we assume that nothing I can do would produce states of myself which will be better handled by some other theory than by T, the question 'But is T actually true?' remains unanswered. According to verificationism, this sort of transcendental realism involves a conceptual mistake about what it is for a theory to be true, and the associated scepticism should be condemned as the asking of a non-question. What verificationism says about the basic merits of the outer-world theory is something which Hume came close to before being swept away from it by cross-currents in his thought (pp. 195-7); and it seems to me obviously correct. But it is less obvious that verificationisrn is right in claiming that it has described all the n1erits that a theory could possibly have, i.e. its rejection of the transcendental realist thought that the truth of the outer-world theory is not secured by the theory's havmg no possible rivals which surpass it in simplicity or power. This extra element in verificationisrn is something it shares with phenomenalism-; but I am envisaging verificationism as being cautious enough to avoid the fatal defects in phenomenalism. In particular, I take it to be saying that a proper understanding of the meanings of objectivity-statements involves grasping the kinds of ways in which they can be related to statements about inner states, through conditionals running one way and conditionals running the other; but not as associating the meaning of an objectivity-statement with any definite set of conditionals, thus freezing its meaning in an objectionable manner; and still less as trying to relate the outer and the inner through biconditionals, as phenomenalism does. Those cautiously negative remarks about verificationism imply that it is, in my hands at least, a rather sketchy doctrine. Still, it has enough content to be non-trivial, while also being cautious enough to be possibly true. I believe that it is true, as far as it goes, which puts me on the same side as Kant: see, for instance, his remark about 'that unity' in" our items of knowledge 'which constitutes the concept of an object', and what he says about 'the concept of body' as "a rule for our intuitions" (A 104-6). Verificationism is not, of course, the whole content of Kant's 'transcendental idealism', but it is an important part of it. And I take it that when Kant speaks of regarding the world as 'a thing in itself, at least part of what he means is: regarding the lworld as something whose real nature is not exhausted by all the facts about lactual and achievable experience. \ But I shall not argue from verificationism taken as a premise, nor shall I attach much weight to my view that it was accepted by Kant. All I need is to
50
JONATHAN BENNETT
mention 'verificationism' and have you know wliat I mean; and even that will not come until Section 8 below. 4. TRANSCENDENTAL ARGUMENTS
The inquirer or 'sceptic' with whom I began has no problems over his knowledge of his inner states, but he does have one over his beliefs about the outer world. That was indeed how Descartes thought he was situated: he apparently couldn't grasp the idea that self-knowledge might also need explaining~ When Gassendi implicitly challenged him to explain how "you manifest yourself to yourself merely by the operation called thought", Descartes lost his temper (pp. 716, 799). Nor was the question taken any more seriously by the British empiricists; and although Spinoza and Leibniz had views about it, they are not relevant to my main topic. The first relevant contribution came from Kant, who argued for, and from, statements about what is required for awareness of one's inner states; and, most notably, presented reasons for thinking that if the sceptic knows what he thinks he does then he must also know some of the things which he thinks he doesn't. And in our own time Wittgenstein, Strawson, Shoemaker, Rorty and others have reached fairly Kantian conclusions by means of fairly Kantian arguments. The last phrase needs comment. The philosophers in question have offered what could be called analytic arguments, by which I mean ones built entirely out of the materials of conceptual analysis and conceptual inter-linkage, without taking anything from what Strawson has called 'the dark side' of Kant's thought - the side that entertains the notion of things as they are in themselves, implies that the mind atemporally creates its objects, and so on. I believe that one can indeed build a significantly Kantian argument without drawing on 'the dark side', but I shan't defend this view (e.g. against Hintikka and Rosenberg). The organizers of this conference asked me to discuss a certain kind of conceptual argument, and my concern is with that kind of argument - its powers and its limitations, but not its historical antecedents. When I call such arguments 'transcendental' that could be taken as a mere label with no Kantian implications at all. I take a 'transcendental' argument to be one which aims to rebut some fonn of scepticism by proving something about the necessary conditions for self-knowledge, self-consciousness, or the like. That is a narrower usage than some writers have adopted. Any argument which I count as 'transcendental' will imply that the scepticism which it opposes is in some way self-defeating; but I don't label as 'transcendental' every argument which seeks to show that
ANALYTIC TRANSCENDENTAL ARGUMENTS
51
some kind of scepticism is self-defeating - e.g. Malcolm's argument that scepticism about freedom is self-defeating becalJse if there is no freedom there are no intentions, and thus no assertions, so that no-one can truthfully assert 'There is no freedom'. Nor do I count as 'transcendental' every argument "to the conclusion that the truth of some principle is necessary to the possibility of the successful employment of a specified sphere of discourse" (Griffiths, p. 167) - a usage which makes the term 'transcendental argument' very broad indeed (as is pointed out by MacIntosh, pp. 185-6). An argument about what is required for the possibility of language as such is a plausible candidate for the 'transcendental' label (Kekes), and I shall son1etimes mention such arguments in passing. But I wouldn't unqualifiedly call them 'transcendental arguments' unless they included the claim (which I doubt) that self-knowledge conceptually requires linguistic capacity. On the other hand, I use 'transcendental argument' more broadly than Gram does: since "Kant has the only clear historical title to [the] notion" of a transcendental argument, Gram won't count as 'transcendental' any argument which lacks the highly 'peculiar character' demanded by some remarks of Kant's; but this leads him to conclude that there cannot be any valid transcendental arguments (p. 15 and passim). Korner also reaches this negative conclusion on the basis of a Kantian understanding of what a 'transcendental argument' is. I have no quarrel with any of these writers. I merely use 'transcendental argument' in my own way. 5. INTUITIONS, CONCEPTS, JUDGMENTS
Now, the Cartesian inquirer thinks that he knows what his inner states are while having a problem about the outer world; to which a transcendental argument replies that if he has self-knowledge then this entails some partial solution to the outer-world problem. Entails what? Well, one might try to argue that if someone has self-knowledge then:
(1) his inner states must be thus and so; or (2) he must have such and such concepts; or (3) he must employ such and such concepts. I take it that (3) involves the other two: to employ a concept you must possess it and must have something to apply it to. Can we stop short of (3)? Could it be shown, for example, that self-knowledge requires inner states to which objectivity-concepts could be applied, without requiring that they
52
JONATHAN BENNETT
actually be applied? I doubt it, and I have~never seen a transcendental argument which aimed at (1) but not (3). But some have tried to prove (2) without (3) and perhaps also without (1). That is, they have tried to prove only that self-knowledge requires a disposition to apply certain concepts if one should have the appropriate data. Strawson has an argument which is like that. (It is a 'transcendental argument' in my sense, though it concerns not the whole outer world but only other minds.) He argues that self-knowledge requires one to have a concept of 'person' which can be applied on the strength of behavioural indications, but he does not argue that the self-knower must have any actual opportunities to apply this concept to anyone other than himself (Strawson, 1959, ch. 3). This is not the place to criticize Strawson's argument in detail; but I have a general reason for preferring to argue for (3) rather than for (1) or (2) alone. The reason stems from Kant's insight that self-knowledge involves intellectual activity: to know what one's inner states are like is to make judgments of certain kinds. Now, just as one might show that to climb a mountain one must bend one's knees, or that to balance the books one must add figures, so we may be able to show that to do whatever is involved in self-knowledge one must do certain ancillary things; and since in this context the relevant 'doing' is judging, such an argument would conclude that self-knowledge requires one to make certain sorts of judgn1ents, Le. employ certain concepts. That would lead to a type-(3) conclusion. To argue only for (2), one would have to show that in order to do what is involved in self-knowledge one must be intellectually equipped to do certain other things without necessarily doing them; and I don't see how such an argun1ent could succeed. 6. A SIMPLE TRANSCENDENTAL ARGUMENT
As a basis for further discussion, I shall now present an actual transcendental argument whose conclusion is of type (3). The argument is my own, though I found it while thinking about Kant's Refutation of Idealism (Bennett §51). It concerns beliefs about one's past inner states. Starting with someone who has beliefs about his present states, I argue that he cannot add beliefs about his past unless he brings some of his inner states under objectivityconcepts, Le. unless he reasonably regards himself as inhabiting an objective realm. Prima facie, it seen1S that the protagonist's states might include recollections of his past states, thus giving him access to his past even if all of his inner states are just a chaotic jumble which cannot be treated as perceptual
ANALYTIC TRANSCENDENTAL ARGUMENTS
53
contact with an outer world. The first step in nlY argument is to show why that is not really possible. Adopt the hypothesis that the protagonist does have recollections of his past inner states. That implies that a sub-set of his inner states contain representations of other inner states: as well as pains and buzzings and glows he also has what 1 shall call K-type representations of pains and buzzings and glows; and we are supposing these to be his recollections of earlier pains etc. But this account contains no basis for connecting a K-type representation with the protagonist's past. We are pretending that when he judges 'I have a K-type representation of state S' he can infer the judgment 'I was earlier in state S'; but this supposed inference fronl the K-type representation to ajudgment about the past is simply idle for our protagonist - it is, in Wittgenstein's metaphor, a wheel which turns though nothing turns with it. I say this for two reasons. Firstly, our protagonist has no way of using any judgment about the past: since his inner states are treated by him as a chaos, in that he doesn't bring them under any general principles of order, he cannot use the judgment 'I was earlier in state S' as evidence for any other judgment. This might be turned into an attack on the assumption that he even makes judgments about his present states; but I shan't press that point, because I want to grant him his present-tense judgments in order to show that he cannot have past-tense ones. Secondly, and more important in my argument: there is a one-to-one correlation, for the protagonist, between judgments about the past and the present-tense judgments on which they are based: he is in a position to judge 'I was in state S' when and only when he is in a position to judge 'I have a K-type representation of state S'. This nleans that the supposed addition to his Weltanschauung of the supposed judgments about the past is a routine, simple, mechanical business; something which does not introduce any complexity, or any further element of structure, into the protagonist's intellectual situation. Compare our protagonist with someone who, given the very same inner states, takes his K-type states to be representations of future states. There is no real content to this supposed difference between the two: the judgments of one can be systematically paired off against those of the other; and the illusion that there is a real difference arises purely from my having chosen with no basis in the given facts - to word the judglnents differently. Then compare those two with a third person who, given the same inner states, does not regard his K-type states as representations at all. He notes that he is sometimes in pain and sometimes in a state which is R-related to
54
JONATHAN BENNETT
pain, sometunes experiences buzzing and sometimes experiences something which is R-related to buzzing, and so":on; but he does not take R to be a representing relation, let alone take it to be specifically the relation of recollecting or foreseeing. This too sounds like a totally different story, but only because 1 have chosen to word it differently: nothing in the real content of the story justifies one wording rather than another. 1 conclude that if the protagonist's inner states are treated by him as a chaos, he can have no working concept of the past, and thus no concept of the past. Now suppose that he reasonably believes that he has experience of an outer world: his inner states are suitable, and he has the appropriate concepts and applies them in ~bjectivity-judgments. Obviously, he can now draw conclusions from some of his judgments about the past; but my main concern is not with what flows from those judgments but with what flows into them, Le. with the protagonist's grounds for making judgments about the past. He must still base those judgments upon present data - his present inner states - but he is no longer restricted to one datum per judgment, for he can now bring several of his present inner states to bear on a single judgment about his past. This can happen in three ways. His judgment that he was in state S at past time t 1 maybe confi.rmed or disconfirmed (a) by his recollections of his states at other times, as when his recollection of falling at time to confirms his judgment that he was in pain at t 1; or (b) by his recollections of his other states at t 1, as when his recollection of seeing the sun at t 1 confirms his judgment that he felt wann at t 1; or (c) by data other than recollections, as when his seeing of ashes at t 2 confirms his judgment that he saw a fire at t 1. Each of these involves general judgments as well - falls are followed by pain, sunshine is accompanied by warmth, ashes are preceded by fire - and the acceptance of such 'laws' is of the essence of a belief in an outer world. My point is not that a judgment is more secure if several data confirm it. The argument does not concern the trustworthiness of judgments about the past but rather their possibility. Now that several of the protagonist's states may bear upon a single one of his judgments about the past - confirming or disconcirming it -- he has a complex relationship between 'I was ...' and 'I have a K-type representation of being ...', rather than a simple one-to-one mapping, so that his concept of how he was in the past is no longer idle. Each of his judgments about how he was at t is a summing-up of where the balance of the evidence lies - an over-all judgment based on what seems in most ways to have been the case at t. When he had no use for objectivity-concepts, on the other hand, there was at most one way in which something could seem to
ANALYTIC TRANSCENDENTAL ARGUMENTS
55
have been the case at t; and so the notion of 'what was the case' collapsed into that of 'what seems to have been the case";, which in tum degenerated into something which had nothing at all to do with the past. That completes my 'transcendental argument' for the view that anyone who has beliefs about his own past inner states must also apply objectivityconcepts. 7. THE PURSUIT OF STRENGTH
When arguing for a conditional, one wants its antecedent to be as weak, and its consequent as strong, as possible. My antecedent is 'If someone has beliefs about his own past inner states ...': I have not refuted the view that someone could at each moment know what his present inner states are, while having no beliefs about his past states or about an outer world. Still, that may be refutable by further argument (and if it isn't then my argument in Section 6 is not after all a 'transcendental' one in my sense). It is plausible to suppose (a) that beliefs are impossible unless one has standards for deciding whether they are true or false (Wittgenstein, §258; Rorty (1970), p. 222; Harrison, pp. 56-57), and (b) that I can have standards for my present application of a concept only if I have beliefs about my past applications of it (Kant, A 101;· Wittgenstein, § 260). Those two theses would jointly strengthen the conclusion of my argument by weakening its antecedent to 'If someone has beliefs about his inner states ...'. I'm not sure that this will go through, because I am not certain of (b); but all I want here is to indicate some possible argumentative strategies. I have to admit that in the conclusion of my argument, the consequent needs to be weakened. I have said that if someone makes judgments about the past he must make objectivity-judgments, but really all my argument shows is that he must bring his inner states under a system of lawlike generalizations which lets him connect several judgments about the present with a single one about the past. A system of laws which implied the existence of an outer world secure this result, but perhaps some weaker systen1 might also suffice. On this question, I am defeated. I have tried to invent laws strong enough to put the concept of the past to work yet too weak to introduce obj'ectivityconcepts; and I have tried to prove that there cannot be such laws; and I have failed in both attempts. All I can say is that a non-idle concept of .the past needs a kind of ordered complexity which I cannot see how to get without also bringing objectivity-concepts into play. Wilkerson has suggested that it is typical of transcendental arguments that
56
JONATHAN BENNETT
they don't produce necessary conditions for .self-knowledge or whatever, but only sufficient conditions together with the claim that "we are incapable, given our present conceptual resources, of thinking of any other conditions which would be ... sufficient" (Wilkerson, p. 211; see also Smith, p. 159). There are indeed interesting arguments of that fonn, though the best exanlple I know is not a transcendental argument in any accepted sense. I allude to Strawson's linking of objectivity with spatiality, by an argunlent which shows that objectivity involves certain requirements which are fulfilled by spatiality and not by anything else one can think of (Strawson, 1959, ch. 2; Bennett, p. 43). But I see no reason to think that transcendental arguments must be of that inconclusive sort. I think that my argument in Section 6 shows conclusively that someone who has a working concept of his own past nlust bring his inner states under general laws ; and I see no reason in principle why someone should not strengthen that result - still using conclusive arguments rather than Wilkersonian ones - by weakening the antecedent or strengthening the consequent or both. With any argument which is not rigorously formalized, there is admittedly a marginal possibility that some gap has been overlooked; but I take Wilkerson to be claiming a special inconclusiveness in transcendental arguments as such, and that is what I am questioning. A famous transcendental argument of Strawson's has a conclusion which is stronger than mine in both ways (Strawson, 1966, pp. 72-112). In contending that if someone has any awareness of his present states he must employ objectivity-concepts, Strawson offers a conditional with a weaker antecedent and a stronger consequent than mine. But even with the help of Rorty's patient reconstruction of that argument of Strawson's (Rorty, 1970), I am still not sure that I understand it. In contrast with that, my argument in Section 6 is relatively plain and clear; but it achieves this at the price of having a rather weak conclusion, and it may be that to strengthen it must one must forgo my kind of plainness in favor of Strawson's kind of subtlety, depth and provocative elusiveness. 8. A DILEMMA
Even if it could be proved that self-knowledge requires the use of 0 bjectivityconcepts, that is, requires the reasonable belief that there is an outer world, one might still wonder what help this gives with the problem confronting the Cartesian inquirer, the problem which Locke tried to tackle. The question could be expressed thus: even if a self-knower must believe that there is an outer world, the question still remains as to whether there is such a world. We
ANALYTIC TRANSCENDENTAL ARGUMENTS
57
may have a transcendental argument which "shows that the referents of the concepts in question are not merely useful fictigns. But, for all such an argument shows, they may still be fictions, even if indispensable ones" (Tlumak, p. 263; see also Ayer, pp. 105-9, and Smith, pp. 165-8). But 1 find that formulation hard to work with: it puts the sceptical inquirer in the position of having to say 'I believe that P, but is it really true that P?'; and that slightly peculiar question raises issues which 1 prefer not to go into here. A better formulation of essentially the same issue is provided by a dilemma. Our Cartesian inquirer is asking for the credentials of the theory that there is an outer world; but how does he construe this? Does he (a) construe the theory in a verificationist manner, so that its truth could be sufficiently assured by facts about how well it conceptually organizes his inner states? Or does he rather (b) construe it in a transcendental realist manner, so that its being more economical and powerful than any possible rival theory does not entail that it is true? The dilemma is this: if (a) he accepts verificationism, then that gives him a reasoned justification for accepting the outer-world theory, and there is no need for any transcendental argument; but if (b) he has a transcendental realist understanding of the outer-world theory then transcendental arguments will be powerless to help him with his inquiry. The case for (a) is obvious. The case for (b) depends on the fact that transcendental arguments can only prove conclusions to the effect that if someone has self-knowledge then he must satisfy certain conditions; so an outer world can enter the picture only by inference from propositions about the self-knower. There seems then to be no hope of showing that a self-knower must inhabit an outer world unless 'There is an outer world' is construed in a verificationist manner. Such a construal was clearly essential to the argument which I presented in Section 6 above. So there is the problem: it seems that (a) if a verificationist position is accepted, transcendental arguments are not needed to help the Cartesian inquirer, and (b) if it is not accepted they cannot help him. I once discussed (a), but (b) seems to have escaped my notice (Bennett, §52); and (b) but not (a) was implicitly pointed out by Williams in a remark about "Kant's insistence that his transcendental arguments gave knowledge of how things must be only because things were not things in themselves" (Williams, p. 218). It was Stroud who combined the two points to generate an argument for the conclusion that transcendental arguments cannot do any valid work in answering sceptical inquiries about the well-foundedness of the belief in an objective realm. Stroud suggests rather than asserts his conclusions; and, strictly, he is
58
JONATHAN BENNETT
addressing himself not to transcendental arguments generally but just to certain recent and prominent examples. I believe, though, that he could have cast his net wider by construing 'verificationism' more broadly than he does most of the time. I shall try to explain. The main emphasis in Stroud's paper is on a somewhat narrow kind of verificationism which involves claims of the fonn 'x makes sense only if P' , based on a verification principle which lays down necessary conditions for intelligibility. Each transcendental argument which Stroud discusses does indeed seem to use such a principle, and in each case it is arguable that the relevant kind of scepticism can be met just by the appropriate verification principle, if the latter is correct. So on this basis Stroud can indeed generate a dilemma with which to confront those transcendental arguments. But it hardly n1akes trouble for the argument which I presented in Section 6 above. That argument relies on a view about what makes a concept honest or nonidle; but it would, I suggest, be strained and inaccurate to call that a 'verification principle'. However, there is broader way of construing 'verificationism', namely as the view that there is no coherent question about a theory's truth over and above all the questions about its comparative success in conceptually unifying the data which fall under it. My argument in Section 6 is 'verificationist' in that sense: its conclusion was a proposition about how a self-knower (with a concept of his past) must conceptualize his inner states, and so it was about the outer world only on a 'verificationist' construal of statements about the outer world. Furthennore, I don't see how any transcendental argument could be even prima facie relevant to sceptical inquiries unless it was 'verificationist' in this broad sense. The narrower construal of 'verificationism' predominates in Stroud's paper and in some of the subsequent literature. It is, for instance, the only possible basis for bracketing this work of Stroud's, as Rorty does (1971, p. 4), with Judith Jarvis Thomson's discussion of a private-language argument (Thomson, p. 29). The broader construal of ,verificationism' is also present in Stroud's paper, as is noted by Goldman (pp. 106-7), but its role is a rather quiet and subdued one; and, in particular, it is not used explicitly to generate the broader dilemma which challenges the relevance to scepticism of every transcendental argument. That broader dilemma is now my sole concern. 9. TWO COUNTER-ATTACKS AGAINST THE DILEMMA
ANALYTIC TRANSCENDENTAL ARGUMENTS
59
ing the hom of it which says that such arguments are powerless unless verificationism is presupposed (Stine, pp. 49-51). :Rather than presupposing verificationism, Stine says, a good transcendental argument could constitute a defence of it. That, if it .were right, would put transcendental arguments right back into service in answering sceptical inquiries; but I don't think that it is right. Stine's argument seems to run as follows. If a transcendental argument proves that every self-knower must bring his inner states under an outerworld theory, this justifies the acceptance of the theory; but that is to justify this conceptualization of one's inner states; and isn't that all that verificationism sets out to do? Unfortunately, it is not. To answer the sceptical inquirer, the verificationist has to ll1aintain not merely that a certain way of using objectivity-concepts is all right, but also that a certain other way of using them is conceptually inadmissible - Le. that when everything has been said about the comparative utility of the outer-world theory no coherent question remains to be asked. No transcendental argument has the power to show that. Hacker has counter-attacked against the horn of the dilemma which says that if verificationism is presupposed there is nothing for transcendental arguments to do (Hacker, p. 84). He envisages a sceptic who is not my mild Cartesian inquirer but rather a radical fellow who makes claims about his inner states while denying that anything else even makes sense. "The existence of anything other than my own states," he says in effect, "is. not just dubious, not just false, but downright unintelligible." So there he sits, untouched by verificationisn1; and then a transcendental argument sets him back on his heels by showing that ifhe does not accord meaning (and indeed truth) to statements about an outer world he cannot know about his own inner states. I am not sure that exactly Hacker's sceptic is saying. Unless he is direly confused, he must agree that one can give to outer-world statements the sense conferred on them by verificationism and employed in transcendental arguments. And Hacker's sceptic does agree to this, for he sees himself as "securely in possession of his array of 'Cartesian thoughts', out of which he may, so he conjectures, build up a Humean fiction of an objective outer world"; but such a fiction would confer meaning on objectivity-statements; so it cannot be that sort of meaning which Hacker's sceptic is denying to them. The only alternative construal of him that I can find is this: Hacker's 'sceptic' is merely saying that objectivity-statements are unintelliiible if they are construed in anything but a verificationist-manner, Le. that they make no sense if the objective realm is taken to be a 'thing in itself. That, of course, isn't 'scepticism' in any reasonable sense. Also, it is not in conflict with any
60
JONATHAN BENNETT
valid transcendental argument. For a transcendental argument to have any impact on this position - namely the position of someone who says that objectivity-statements have a verificationist sense and no other - it would have to prove that self-knowledge requires that objectivity-statements be accepted both on verificationist and on thing-in-itself construals; and I don't believe for a moment that such a thing can be proved. Strawson in his 'Persons' chapter does seem to attempt something of the kind: he argues that selfknowledge requires a preparedness to apply a concept of 'person' in such a way that 'x is a person' is guaranteed by behavioural (verificationist) premises and in tum guarantees mentalistic (thing-in-itself) conclusions (Strawson, 1959, pp. 106-110). But I agree with the majority opinion that this argument of Strawson's does not succeed. 10. UNRADICAL SCEPTICISM
I see no prospect for doing any damage to the horn of the dilemma which says that transcendental arguments are powerless if verificationism is not presupposed. If someone asks 'Is the outer-world theory true?', and construes this in a non-verificationist manner as an inquiry about the world as a thing in itself, then no transcendental argument can either answer it or show that it ought not to be asked. Where the truth of verificationism is denied or questioned, transcendental arguments must stand mute in face of all sceptical inquires. However, I do think there is something wrong with the horn of the dilemma which says that if verificationism is assumed than it will answer the sceptical inquiries, so far as they can be answered, leaving transcendental arguments idle. To that extent I agree with Hacker, though not with his account of what it is that transcendental arguments can achieve which mere verificationism cannot. (In all of this, I am considering only their power to answer sceptical inquiries. The exploration of the conceptual requirements for self-knowledge, considered just as a part of the philosophy of mind, lies beyond my present scope.) Suppose we are confronted by a 'sceptical' inquirer who concedes what verificationism demands: that is, his question about the outer-world theory is only the q-q.estion as to how well that theory functions in comparison with possible rivals. Let us further suppose that he does regard the theory as doing a better job of conceptual unification of his past and present data than would be done by any rival theory of which he is aware. That still leaves him with two questions. (1) Are there things he could do - whether experimenting or
ANALYTIC TRANSCENDENTAL ARGUMENTS
61
merely waiting - which would bring him data in th~light of which the outerworld theory would be inferior to some rival theory? (2) Are his past and present data handled less well by the outer-world theory than by some other theory which he just hasn't thought of? Suppose now that there is a successful transcendental argument, A, which proves that selfknowledge requires the application of objectivity-concepts; and let us consider how far A can go towards answering those two questions. In response to the first question, A would show that I could not find myself to be possessed of inner states which did not support an outer-world theory; and that is something which mere verificationism could not show. Still, it _is not a very large gain: it does not even show that while I have self-knowledge I must inhabit this outer world, though of course a transcendental argument with a stronger conclusion than A's might show this. Similar remarks apply to the second question, about superior rival theories which might be thought up now. If A is sound, then a self-conscious creature cannot rely on a non-outer-world theory in preference to any outer-world theory as a means for handling his inner states. Once more, however, A as it stands does not rule out the possibility that a self-knower should con1e to think that he has been quite wrong about what kind of outer world he inhabits. It could be argued that A doesn't answer the second question at all: granted that a self-knower must manage his inner states with the aid of an outer-world theory, perhaps he could at the same time acknowledge that some rival theory was superior? Isn't it conceivable that I should have to do my basic thinking in terms of theory T while realizing that theory T* was more powerful or more economical? I am not sure about this. If a self-knower n1ust en1ploy T, that must be because he needs-conceptual services which T can render him; and so he cannot both need T and know a superior rival T* which renders all the main conceptual services which T renders. Objection: 'But he might know a rival T* which is over-all superior to T although it does not render all the services which T renders.' I cannot refute that, but it is so peculiar that I am prepared to walk away from it and to conclude that A does, near enough, show that we could not discover through intellectual effort that our past and present data are better handled by a non-outer-world theory. Rorty assigns to transcendental arguments this role, as a kind of bulwark against rival conceptualizations for past and present data (Rorty, 1971, pp. 10-11). But he doesn't envisage their ever proving that a certain conceptualization of one's inner states is absolutely indispensable. At best, he says, one can hope to defend a favoured conceptualization against displacement by particular rivals, proving that C cannot be dislodged by C* because any use of C*
62
JONATHAN BENNETT
would have to be parasitic on the use of C; and then for C** a fresh argument would be needed, and so on. I cannot refute this, but, as I said of a similar claim by Wilkerson, I see no reason to believe that transcendental arguments must operate in this way. Although I cannot actually produce anything fitting my specifications for 'argument A', I am not convinced that no such argument is possible. 11. RELEVANCE TO THE FOUNDATIONS OF SCIENCE
Of the two answers which transcendental arguments can give to sceptical inquiries, the one about what I might find my future states to be like has no bearing on what concepts should be used in the foundations of science now. It is addressed not the question of what I can make, intellectually, of my present data, but rather to the question of what the future may bring. So it really belongs in the province of the so-called 'problem of induction': its only relevance to the foundations of science is that may help someone in whom Hume has caused a crisis of morale, leaving him unwilling to continue at all with an activity which may at any moment collapse. There may be sound transcendental arguments which set limits to how total and sudden that collapse could be known to be - that is, arguments proving that self-awareness is not compatible with experiential chaos or with rapid and radical changes in the patterns within one's experience. But the person who is demoralized by Hume is not typical of those who theorize about the conceptual foundations of science, and the nursemaid activity of consoling him is a rather lowly form of labour. Incidentally, I don't think that any sound argument could rule out the possibility that my inner states might gradually change in such a way that although at any giventime they were fairly efficiently handled by some scientific theory, the content of my Weltanschauung slowly changed until at last it had nothing in common with the one I accept now. So if someone is demoralized by the thought: 'Why bother with science ifwe cannot get results which are certain to remain at least approximately valid?', there is no consolation for him. The second answer which transcendental arguments can give is prima facie more relevant to the conceptual foundations of science. If I cannot handle my data without the aid of concept C, not through the accidents of individual or cultural history but rather because C is conceptually required for self-knowledge, then there is no point in my looking for alternatives to C. It is a category, in roughly Kant's sense, and science is stuck with it. If Rorty is right, one can never show that C has an absolutely categorical status, but only defend it against particular rivals. That would virtually put
ANALYTIC TRANSCENDENTAL ARGUMENTS
63
transcendental arguments out of business, for they would amount merely to the case-by-case evaluations which are what a foundations-of-science theorist engages in anyway. Even if Rorty is wrong, I doubt if transcendental arguments should affect anyone's thinking about the conceptual foundations which science might have. This brings up the question of section 7 above, namely 'How much can be validly proved by a transcendental argument?' I cannot answer this, because I don't know just what is required for self-knowledge, or for any main species of it. I have argued only that self~knowledge including beliefs about the past requires the use of objectivity-concepts or something approxitnatinglto them; and although more than this can probably be shown, I don't know how much more. Still, I confidently guess that there is no concept C such that (i) C is provably required for self-knowledge or for some main species of it, and (ii) someone might seriously propose to give science new foundations which did not involve C. Kant tried to show that science must use the concepts of cause and substance in such a way as to be committed to strict determinism and to some conservation law; and those two theoretical items might (and one of them did) come under challenge in a re-thinking of the foundations of science. But since Kant failed actually to prove that self-knowledge requires an acceptance of strict determinism or of a conservation law, this example does not refute my guess. Of course it is only a guess; but I am sure enough of it so that if I were actively working on questions about the conceptual foundations of science I would not look to transcendental arguments for help. 1 University ofBritish Columbia NOTE 1 For extremely helpful comments on an earlier draft of this paper I am indebted to Michael Beebe, Judith Jarvis Thomson, and Barry Stroud.
BIBLIOGRAPHY Ayer, A. J., The Concept ofa Person, London 1963. Bennett, Jonathan, Kant's Analytic, Cambridge 1966. Descartes, Rene, Oeuvres philosophiques (ed. by F. Alquie), Paris 1967, vol. II. Goldman, Alan H., 'Can A Priori Arguments Refute the Sceptic?', Dialogue 13 (1974), 105-109. Gram, Moltke S., 'Transcendental Arguments', Nous 5 (1971), 15-26. Griffiths, A. Phillips, 'Transcendental Arguments', Proceedings of the Aristotelian Society,
64
JONATHAN BENNETT
suppl. vol 43 (1969), 165-180. Hacker, Peter, 'Are Transcendental Arguments a Version of Verificationism?', American
Philosophical Quarterly 9 (1972), 78-85. Harrison, Ross, On What There Must Be, Oxford 1974. Hintikka, Jaakko, 'Transcendental Arguments: Genuine and Spurious', Noils 6 (1972),
274-281. Hume, David, Treatise ofHuman Nature (ed. by L. A. Selby-Bigge), Oxford 1964. Kant, Immanuel, Critique ofPure Reason. Kekes, John, 'The Scandal of Philosophy', International Philosophical Quarterly 12
(1972),512-525. Korner, Stephan, 'The Impossibility of Transcendental Deductions', The Monist 51
( 1967), 317- 331. Locke, John, An Essay Concerning Human Understanding. MacIntosh, J. J., 'Transcendental Arguments', Proceedings of the Aristotelian Society 43
(1969), 181-193. Malcolm, Norman, 'The Conceivability of Mechanism', Philosophical Review 77 (1968),
45-72. Rarty, Richard, 'Strawson's Objectivity Argument', The Review of Metaphysics 24
(1970), 207 -244. Rorty, Richard, 'Verificationism and Transcendental Arguments', Nous 5 (1971), 3-14. Rosenberg, Jay, 'Transcendental Arguments Revisited', The Journal of Philosophy 72
(1975),611-624. Shoemaker, Sydney S., SelfKnowledge and SelfIdentity, Ithaca 1963. Sn1ith, G. W., 'The Concepts of the Sceptic: Transcendental Arguments and Other Minds', Philosophy 49 (1974), 149-168. Stine, William D., 'Transcendental Arguments', Metaphilosophy 3 (1972),43-52. Strawson, P. F., Individuals, London 1959. Strawson, P. F., The Bounds ofSense , London 1966. Stroud, Barry, 'Transcendental Arguments', Journal ofPhilosophy 65 (1968), 241-256. Thomson, Judith Jarvis, 'Private Languages', American Philosophical Quarterly 1 (1964),
20-31. Tlumak, Jeffrey, 'Some Defects in Strawson's Anti-Skeptical Method', Philosophical Studies 28 (1975),255-264. Wilkerson, T. E., 'Transcendental Arguments', Philosophical Quarterly 20 (1970), 200-
212. Williams, Bernard, 'Knowledge and Meaning in the Philosophy of Mind', Philosophical Review 77 (1968), 216-228. Wittgenstein, Ludwig,Philosophical Investigations, Oxford 1953.
STEPHAN KORNER
ON BENNETT'S 'ANALYTIC TRANSCENDENTAL ARGUMENTS'
Professor Bennett freely admits that his understanding of what a 'transcendental argument' is, differs from the Kantian and that he uses the term in his own - rather special - way. Although I am opposed to such terminological separatism, I shall not argue against it here. Nor shall I discuss the interesting remarks which lead up to, and which follow, what I take to be the central part of Bennett's paper, namely the characterization and exemplification of what he calls "analytic transcendental arguments". In examining them I shall be mainly concerned with their tendency to leave open deductive gaps between their premises and their conclusions, and with the difficulty of bridging these gaps in a satisfactory manner - a difficulty which they share with other transcendental arguments and which seems insuperable to me. 1. ON THE CHARACTERIZATION OF 'ANALYTIC TRANSCENDENTAL ARGUMENTS'
The general purpose of arguments of this kind is to establish that a person who knows what his inner states are, should have at least "some partial solution to the outer-world problem". In order to distinguish them from other kinds of philosophical argument, one must in addition point to some special features of their premise and conclusion and, possibly, of the implication which connects the two. What Bennett says about the premise is perfectly clear. It must express self-knowledge in a sense of 'knowledge' in which to have knowledge is ipso facto to make judgments. As may be worth emphasizing, the requirement of such self-knowledge is stronger than the requirement of mere self-awareness which, e.g. in the psychology of Descartes and Brentano, does not necessarily involve the making of judgments (iudicia, Urteile) and thus the application of concepts, but only the having of presentations (presentationes, Vorstellungen). It is less clear how Bennett understands the iInplication which connects the premise and conclusion of his analytic transcendental arguments and which he calls 'entailment'. Since this term is used in many different ways, it would have been useful if he had explained his own use by relating entailment to the deducibility-relation of some fonnal logical theory (e.g. classical 65 P. Bieri, R.-P. Horstmann, and L. KrUger (eds.), Transcendental Arguments and Science, 65-69. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
66
STEPHAN KORNER
quantification theory). As matters stand, it is not clear whether Bennett holds (a) that entailment is simply the converse of a certain fqrmal deducibility relation ; (b) that entailment is a stronger relation than formal deducibility in that certain entailments are invalid although the corresponding formal deducibility-statements are valid (as is e.g. asserted by Kant in B 817, where he rejects apagogic arguments for transcendental proofs); or (c) that 'entailment', together with terms like 'coherence' and 'logical oddity', belongs to an "informal logic" which validates some formally invalid inferences. The suspicion that Bennett may have such informal - and hence formally unassailable arguments in mind is supported e.g. by his remark that the scepticism which a transcendental argument opposes "is in some way self-defeating" (my italics). But I shall ignore this possibility here since I have elsewhere considered attempts at rehabilitating transcendental arguments by appealing to "informal logic" . 1 The conclusion of an analytic transcendental argun1ent states that a person "brings some of his inner states under objectivity concepts", Le. "reasonably regards himself as inhabiting an objective realm". Bennett does not analyse the notion of objectivity-concepts any further so that it is not clear whether or not there exists in his view a set of objectivity-concepts which does not only contain 'x is objective' and which (as Kant believed of his Categories) is and must be applied by everybody applying such concepts. 2 2. INNER STATES AND THE APPLICABILITY OF OBJECTIVITY -CONCEPTS TO THEM
An appreciation of Bennett's analytical transcendental arguments presupposes a distinction between various kinds of concepts applicable to inner states or subjective phenomena. Of particular importance is the distinction between his objectivity-concepts and what n1ay be called "cosubjectivity-concepts" . 3 One must - as is clearly shown in Bennett's sample argument distinguish between one's being aware of a subjective phenomenon per se (such as a pain) and one's judging that a subjective phenomenon (such as a recollected pain) is cosubjective in the sense that it belongs to a system of phenomena which are unified by their relation to a subject (such as oneself). For Kant the unity is a function of the 'Ich denke'; for Bennett it arises from a person's bringing "his inner states under a system of law-like generalizations" which, for example, "lets him connect several judgments about the present with a single one about the past." These remarks suggest the following defmition of cosubjectivity-concepts:
COMMENT ON BENNETT
67
A concept C(x), which is applicable to subjective phenomena, is a cosubjectivity-concept if, and only if, (a) it is the concept 'x is cosubjective' or (b) its applicability to a phenomenon logically implies the applicability to it of the concept 'x is cosubjective'. The defmition admits the possibility of mutually incompatible cosubjectivity-concepts being applied by different persons or groups of persons. It should, therefore, be noted that the inference from selfknowledge, expressed by the premise 'I judge a subjective phenomenon to be part of a system of subjective phenomena related to myself' to 'I atn applyLng some cosubjectivity-concept' is trivially valid; whereas the inference from the S3.Il1e premise to '1 am applying a certain specified cosubjectivity concept' (other than 'x is cosubjective') is a non-sequitur. Judging a subjective phenon1enon (such as a heard voice) to be cosubjective, is quite different from judging it to be objective, Le. to exist not only as a subjective phenomenon or as part of a unified system of such phenomena. By so placing subjective phenomena into an "objective realm", one leaves, of course, many questions about its nature open, especially the question how far one person's objective realm may differ from that of another without ceasing to be an objective realm. Without prejudging this issue, a definition of objectivity-concepts can now be given on the same lines as that of cosubjectivityconcepts - except that I shall, in order to simplify the discussion, restrict the applicability of these concepts to cosubjective phenomena: A concept O(x), which is applicable to cosubjective phenomena, is an objectivity-concept if, and only if, (a) it is the concept ~x is objective' or (b) its applicability to a phenomenon logically implies the applicability to\ it of the concept 'x is objective'. It is again clear that the inference from self-knowledge, expressed by the premise 'I judge a cosubjective phenomenon to be part of an objective realm' to 'I am applying some objectivity-concept' is trivially valid; whereas the inference from the same premise to 'I am applying a certain specified objectivity-concept' (other than 'x is objective') is a non-sequitur. The following brief comments on cosubjectivity and objectivity-concepts and their interrelations seem to be in order. First, it is possible and for some purposes advisable to interpolate intersubjectivity between cosubjectivity and objectivity-concepts. Second, it is possible and for some purposes advisable to distinguish a priori and a posteriori ·concepts in Kantian fashion, so that e.g., 'x is cosubjective' and 'x is objective' belong to the former class. Third, it is worth recalling that the main aim - and one of the main difficulties - of the Kantian transcendental deduction of the Categories, was to show that the possibility of objective experience depended not only on the applicability of some objectivity-concepts, but on the applicability of certain specific ones. 4
68
STEPHAN KOR'NER 3. ON THE BRIDGING OF OEDUCTIVE GAPS IN 'ANALYTIC TRANSCENDENTAL ARGUMENTS'
Bennett's sample argument consists of two steps of which the first is logically valid and the second is not. The argument is intended to lead deductively from (1) 'I judge myself to recollect a past subjective phenomenon' to (2) 'I am applying cosubjectivity concepts' and from this to (3) 'I am applying objectivity-concepts'. Yet, whereas he claims to have shown the validity of the first step, he does not also claim to have shown the validity of the second (which, according to the above definitions of cosubjectivity- and objectivityconcepts is a non-sequitur). In acknowledging his failure to bridge the deductive gap between (2) and (3), Bennett makes the significant comment that he "cannot see" (my italics) how to get "a non-idle concept of the past ... without also bringing objectivity-concepts into play". He also seems to express the hope that there are "analytic transcendental" arguments which are both logically valid and - to use a term from the language of mathematical criticism - deep enough to throw new light on some serious philosophical issue. I do not share this hope. It might, however, be justifiable if my interpretation of Bennett's somewhat imprecise requirements for the premise, conclusion and implication of analytic transcendental arguments is replaced by another. Bennett's comment draws attention to an important feature of his and other transcendental arguments, namely the acknowledgment by a thinker of a deductive gap between premise and conclusion, together with the assertion that he fmds it inconceivable (unthinkable, unimaginable etc.) for the premise to be true and the conclusion to be false. I shall call a relation between premise and conclusion which, in the manner described, bridges the deductive gap between them, "ontological implication" since it differs, at least prima facie, from logical implication and since it is mainly used in the discussion of ontological issues. It is, moreover, relative in the sense that two thinkers who agree about the principles of logic may disagree about the inconceivability that the antecedent of a certain ontological implication is true and its consequent false. Among the possible interpretations of ontological implication, the following are here of particular interest: (1) The interpretation of ontological implication as logical implication. (2) The interpretation of ontological implication as an intuitive guarantee that a deductive proof, though not yet available, will - or can in principle be provided. (3) The interpretation of ontological implication, as an absolute relation
COMMENT ON BENNETT
69
between premise and conclusion which, though different from logical implication, is no less binding on all rational beings (a view held by Kant for all rational beings, conceived in his way as perceiving and thinking). (4) The interpretation of ontological implication as what may be called a thinker's "ontologically supreme" belief. Without going into details, a thinker's ontologically supreme belief can be characterized (a) as dominating his other beliefs in the sense that he rejects as inadequate any belief which he recognizes to be inconsistent with one of his ontologically supreme beliefs; (b) as nevertheless capable of being subjected to rational argument and, thereby, to correction and replacement. Interpreted according to (1), (2) or (3) ontological implications express and reinforce opposition to ontological change and, hence, to any scientific change which - like the change from Aristotelian to classical or from classical to quantum physics - involves a change of ontology. Indeed, according to each of these interpretations the propounder of any transcendental argument would thereby exclude any possibility of its conclusion ever being abandoned for a different ontological belief. And this is a serious objection to them quite apart from their failure to close the deductive gap between premise and conclusion by some non-deductive, yet absolutely binding inference. Whereas Bennett's sample argument comes closest to the second interpretation, his general remarks also show affinities with the first and third. The fourth interpretation does not involve any attempt at replacing logical non-sequiturs by dubious, non-logical sequiturs. It not only admits ontological change, but may even be useful in understanding it. A fuller account of this interpretation presupposes an inquiry into the structure and function of ontologies and of the nature of metaphysical argument. S
University ofBristol NOTES 1 See 'Uber ontologische Notwendigkeit und die Begrtindung ontologischer Prinzipien' [On Ontological Necessity and the Justification of Ontological Principles] , Neue Hefte fiir Philosophie (ed. by R. Bubner, K. Cramer, and R. Wiehl) , XIV, Gottingen 1978. 2 For a pluralist view see Categorial Frameworks, Oxford 1971 and 1974. 3 It is ignored in Bennett's sample argument (Section 6) but acknowledged in his retros- . pective discu ssion of it (Section 7). 4 For an elaboration of these comments see the paper mentioned in note 1. S See the essays mentioned in notes 1 and 2, as well as 'On some Relations between Logic and Metaphysics' in The Logical Enterprise (ed. by A. R. Anderson, R. B. Marcus, and R. M. Martin), Yale Univ. Press, 1975.
GUNTHER PATZIG
COMMENT ON BENNETT
I shall start with some rather general, I'm afraid, remarks concerning transcendental argurnents in somewhat loose connection with what I have read in Professor Bennett's paper, and then I shall make some particular points as to some particular passages in Professor Bennett's paper. To someone who has not, as I have not, followed at all closely the discussion on transcendental arguments which started, as I think, with Strawson's book Individuals in 1959, the general drift of the argument in the substantial paper of Professor Bennett is not, from the start of it, very easy to follow. It is especially the close connection in Professor Bennett's mind between what he calls "Transcendental arguments" and the problem of possible truth about an objective world which I find puzzling. Since the very expression 'Transcendental arguments' has, and is meant to have, some definite Kantian overtones and connotations, one is immediately induced to fall back on one's own conception of Kant's general philosophy and the role of transcendental arguments inside Kant's theory, especially. And I just think that the establishment of an objective world against sceptical doubts is not high up on Kant's philosophical priority list. The establishment of objective knowledge, yes. But the transcendental deduction of the categories does not, so I think, amount to an actual proof that there is, after all, an objective world out there. It rather concentrates on the idea that the system of categories based on the table of judgments is the only possible system which allows the unification of our subjective intuitions into one consistent and coherent body of knowledge. What might be called the objective world, then, is to Kant "nothing but", to use a phrase which is very prominent in Kant, as you all know, the necessary intentional correlate to the process in combining subjective experience into intersubjective and in this sense objective knowledge. Kant has, it is true, tried to give a "Widerlegung des Idealismus"; and I think it could be discussed whether the fact that Kant added the 'Widerlegung des Idealismus' in the second edition of the Critique of Pure Reason could be used as an argument for saying that it does not belong to the main argument of the Critique of Pure Reason. And it could also, of course, because Kant thought this a necessary addition to his book, be argued that it is, for that very reason, a very important part of the overall 71 P. Bieri, R.-P. Horstmann, and L. Kruger (eds.), Transcendental Arguments and Science, 71- 75. All Righ ts Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
72
GUNTHER PATZIG
structure of Kant's undertaking. I think this is a genuinely debatable point, but I should not like to go more closely into that question here. But the 'Widerlegung des Idealismus', as I read it, amounts to showing that, to quote Kant "my empirically determined consciousness of my own existence implies the existence of objects in space outside me" (B 275), and Kant does not, of course, forget for a moment that objects in space are only appearances and, to quote Kant again, "what we call objects are nothing but mere representations of our sensibility, the fonn of which is space" (B 47). I think that it is obvious that Hume, and with very good reasons, would have rejected Kant's claim that by his transcendental considerations he had established the objectivity of the causal relation; for the kind of objectivity which Kant had established for the connection of events was certainly not of the kind Hume had in mind when he was looking for the 'hidden tie' between cause and effect. I do not want to say that transcendental arguments are not also used to avoid sceptical conclusions. They are so used. But they are basically used to justify the application of concepts in the fonnation of our experience, which had been challenged. After the, as I think, forbidding price of the restriction of all our knowledge to the field of possible experience has been paid, in some kind of philosophical devaluation, or "W'ahrungsrefonn", we may use our old concepts again, because they have been changed into the post-revolutionary currency by order of the new philosophical government, Critical Philosophy. So, all transcendental arguments, at least those in the field of theoretical philosophy, have a double function: restriction of possible claims to knowledge to the field of possible experience, on the one hand; justification of the use of concepts as having objective reality inside this sphere, on the other. I find that this double-facedness of all transcendental arguments and the conclusions established by them, is nowhere more clearly expressed in the Critique of Pure Reason than in the 'Methodenlehre' (B 787-791) where Kant uses, no doubt referring back to the image of the Copernican revolution fron1 the 'Vorrede', the picture of our planet in space as a possible illustration of the field of human knowledge. The traditional metaphysician is metaphorically described as a geographer who has not yet learned that the earth is a spheroid, but takes the naive view that the earth is a plane infinitely extended in all directions. Once you have learned that the surface of the earth is a sphere, you know that the extension of this sphere depends on the diameter of the sphere, and from measuring the curvature of even a small part of this surface you can, by simple calculation, deduce the size of the sphere. Reason is then, by Kant, compared to such a sphere; the field of knowledge is represented by
COMMENT ON BENNETT
73
our planet in this picture. The measure of curvature of the surface of reason, as one might put it pictorially, is determine'Cl by synthetic propositions a priori. This, and only this, makes it possible to determine the extension of possible empirical knowledge. The use of this picture in Kant is, I think, highly significant. The principles of experience, established by transcendental arguments, are interpreted by analogy to the surface of a sphere which contains all objects of possible knowledge. Transcendental arguments, then, are for Kant necessary for achieving two tasks at once: to determine the field of possible knowledge and to limit it off against what he would call the field of "dogmatic speculation". Well, I think that all these things cannot be seriously in doubt or even be controversial. But they should be considered, when the concept of a 'Transcendental argument' is used and some appeal to Kant's authority is explicitly or implicitly made. (I do not see what could be the point in using the term 'Transcendental argument', if not sonle appeal to Kant's authority is intended.) It is, I think, legitimate to say that transcendental arguments, in Kant's sense, can be used to establish some parts of objective knowledge against doubt. But it is equally legitimate to say that transcendental arguments are used to restrict dogmatic claims for truth. To say that some concepts or views are intimately connected with our ways of talking about objects, our methods of identifying and re-identifying things and the like, filay be an argument that we are justified in using them, a kind of deduction. But of course, it may also be used to stress the fact that our ways of looking at things and what we call the world in general are essentially influenced by some inherent presuppositions and structuring factors, which make the truth of all we might want to say dependent on these presuppositions and conceptual frames. There is, as I think, a constructive and a destructive use of transcendental arguments. And I think, in Kant, the exciting point is that you cannot have a successful constructive use of a transcendental argunlent without having at the same time a destructive use of it. I think that any theory of transcendental arguments which does not take account of both these possible (hid interrelated uses is certainly not Kantian, which might be of little import, but is, systematically speaking, something like a one-legged biped, featherless or otherwise. I now come to some individual or concrete points in Professor Bennett's paper: After having expressed my qualms about taking an argun1ent (to quote from p. 50) "which aims to rebut some form of scepticism by proving something about the necessary conditions fOf self-knowledge, self-consciousness or the like" as some paradigm case for a transcendental argument, I wish to say that I find Professor Bennett's attempts highly successful to establish, on
74
GUNTHER PATZIG
Kantian lines, that talking about the time-ordering of my inner states presupposes, if not the existence of an outer world, at least some ordered complexity, which brings objectivity-concepts into play, as he says on p. 55, and, therefore, makes the belief in a objective world at least reasonable. I see, however, no difficulty, as Professor Bennett does, in saying that - and this is, I feel, a new kind of reference to my insistence that transcendental arguments cut both ways - I see no difficulty in saying that son1e concepts are indispensable, and, at the same time, that we may be very well sceptical as to their objective validity. We just have to put our 'indispensable' beliefs and our doubts on different theoretical levels. And I think we do that in our normal daily routines without much difficulty. While I work in the university, nlY belief that the doing of philosophy and the teaching of it has some point is an indispensable belief for working efficiently, but cannot I ask myself, in a critical or depressive mood, whether there is, in fact, any point to the whole thing? I find Bennett's treatment of the verificationist versus transcendental realist theories of self-knowledge on pp. 57 to 63 highly interesting: I agree with him thaf transcendental arguments in his sense must be supported by what he calls a verificationist position. I am, however, not so sure about his claim that transcendental argun1ents could do more than verificationist ones to establish an 'outer world'-theory as the only reasonable one to accept in explaining the order of our inner states, mainly, I think, because I cannot imagine transcendental arguments which would show that self-knowledge requires the application of not just objectivity concepts (as on p. 61) but fully fledged outer world concepts. Regularity of succession of inner states would probably do the job sufficiently, as Professor Bennett himself seems to imply on pp. 54/55, if I have understood that passage correctly. Since such a necessary concept would be functionally equivalent to a category in Kant's sense (Bennett, p. 62) and since I do not believe that there are Kantian categories, because I believe there may be always alternative conceptual schemes to the ones we have accepted, I cannot fmd this train of thought convincing. But possibly Professor Bennett has more to say to support his view than he tells us in his paper. I had, fmally, some difficulties in understanding the concluding argument in Professor Bennett's paper, where he says that transcendental arguments are not very important for the problem of the foundations of empirical knowledge and science. This is a question which I may have only because I have not properly understood the way he wants to reason. The "confident guess" (p. 63) of Professor Bennett that there is no concept C, such that it is required
COMMENT ON BENNETT
75
for - I add: the expression of - self-knowledge and also redundant in the conceptual foundations of science is, it seems to me, the basis of his argument. Now I think the 'confident guess' should be equivalent to the guess that if a concept C' is fundamental for the expression of self-knowledge, it must be necessary and important for the conceptual foundation of natural science as well (I think that's mere logic). Now Professor Bennett says that Kant has not shown that cause and substance are necessary concepts for the foundation of self-knowledge and, therefore, the fact that these concepts are not, nowadays, safe candidates for the retention in all possible expositions of the foundations of natural science does, according to Professor Bennett, not invalidate that 'confident guess'. But then, why does Professor Bennett say that it is just this guess which makes hin1 think that he would not look to transcendental arguments for help in work about the conceptual foundations of natural science? Is it so, because he thinks that the activities in this field would bring out the importance of these concepts anyway, so that we do not have to fall back on any reasonings about the possible expressions of selfknowledge? And would Professor Bennett be inclined to think that the implication 'he talks about would hold also in the reverse direction such that there is no concept necessary in the explanation of the foundations of natural science which is not also necessary in the explanation of the possibility of expression of self-knowledge? I rather think that he would not want to assert this but I would be grateful to Professor Bennett if he could make his views on this point a little bit clearer. Universitiit G6ttingen
RICHARD ROR TY
TRANSCENDENT AL ARGUMENTS, SELF -REFERENCE, AND PRAGMATISM
1. INTRODUCTORY
Most transcendental arguments are anti-sceptical and anti-reductionist, claiming that the reduced world the skeptic holds out as the only legitimate option is not a genuine alternative. They have as their paradigm Kant's arguments against Hume. Such arguments fortify those philosophers who want to insist, with Kant, that there is such a thing as philosophical criticism of the rest of culture - that the philosopher can say something which science cannot about the claims to objectivity and rationality to which various parts of culture are entitled. Thought of in this way, transcendental arguments seem the only hope for philosophy as an autonomous critical discipline, the only way to say something about human knowledge which is clearly distinguishable from psychophysics on the one hand and from history and sociology of knowledge on the other. If, however, one thinks of 'transcendental argument' as a name for a certain logical structure (or, more loosely, a certain dialectical strategy) there seems no particular necessity that such arguments be used only anti-skeptically and anti-reductionistically. In this paper, I shall discuss the relation between the motive of transcendental argumentation and its form. I shall begin (in Section 2) by taking up Bubner's discussion of the dialectical pattern which the paradigm Kantian arguments exhibit. I shall be agreeing with him that their basic feature is a self-referential attack on the coherence of the skeptical alternative offered. But I shall be disagreeing with him about the necessity and the non-empirical character of the conclusions of these arguments. Then, in Section 3, I shall discuss pragmatism's claim that truth is simply the most coherent and powerful theory, and that no relation of 'correspondence to reality' need be invoked to clarify 'true' or 'knowledge'. This claim, which would put out of court the traditional notion of 'legitimizing knowledgeclaims' (and thus the traditional conception of the function of transcendental arguments), is opposed by what I shall be calling 'metaphysical realism' typified for my purposes by Sellars, Rosenberg and Putnam. The point of Section 3 is to show that the crucial ariuments against this sort of realism exhibit the features of transcendental arguments as they are analyzed by Bubner.
77 P. Bieri, R.-P. Horstmann, and L. KrUger (eds.), Transcendental Arguments and Science, 77-103. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
78 Section 4 criticizes the recent development within analytic philosophy of 'naturalized' (and, in particular, 'evolutionary') epistemology - illustrated by Rosenberg's attempt to defend realism through transcendental argumentation revolving around a sort of Tractarian picturing. I argue there that epistemology cannot be 'naturalized' without ceasing to answer Kant's quaestio juris. Section 5 presents Davidson's argument against 'the very idea of a conceptual scheme' and for the claim that 'most of our beliefs must be true'. This section tries to show that that Davidsonian argument exemplifies the same 'transcendental' strategy as Kant's anti-Humean arguments, and as Putnam's arguments against 'metaphysical realism'. Despite this similarity in strategy which seems to me sufficient to let us call Davidson's argument an instance of transcendental argumentation - the aim of his argument is to n1ake impossible the whole Cartesian and Kantian dialectic which makes skepticism and anti-skeptical transcendental argumentation possible. I construe (pace its author) Davidson's argument against the notion of 'conceptual scheme' and against the 'scheme-content distinction' as an argument for pragmatism, and thus against the possibility of epistemology. Davidson, in other words, seems to me to have found a transcendental argument to end all transcendental arguments - one which tears down the scaffolding upon which the standard paradigms of 'realistic' transcendental arguments were n10unted. In my concluding Section 6 I suggest that abandoning the schen1e-content distinction and accepting pragmatisn1 does, in a sense, mean abandoning philosophy. Philosophy does depend upon transcendental arguments for its existence, if philosophy is conceived of as a non-empirical criticism of culture. But there are other conceptions of philosophy (e .g., those of Heidegger, Dewey and the later Wittgenstein) in which philosophy is not concerned with the legitimation of knowledge-claims. 2. REALISTIC TRANSCENDENTAL ARG UMENTS
I shall call 'realistic' transcendental arguments those motivated by the need to make what Bubner calls "a transcendental presupposition". In discussing Wittgenstein's Tractatus, Bubner says: This presupposition [that 'formal logic with its pure tautologies or analytic truths does not hopelessly depart from all the factual truth of the empirical, but rather it reflects precisely those structures according to which the world is ordered'] is called transcendental, since it assumes a structural identity in the relationship between logic and reality, upon which the concept of meaning depends. A logic which did not make this presupposition would renounce all its competence with regard to the meaning of statements ...
TRANSCENDENTAL ARGUMENTATION AND ITS FORM
79
... The relationship is the sole guarantee that it is possible for a statement to reproduce a given piece of reality: to make a meaningful proposition. 1
I shall say that any transcendental argument which has as its aim to guarantee correspondence of logic, or language, or the practice of rational inquiry to the world is a 'realist' argument. Such an argument has as its paradigms the arguments (those of the Transcendental Deduction and the Refutation of Idealism) which Kant used to illustrate his claim that "only the transcendental idealist can be an empirical realist". The following are sufficient conditions for labeling an argument a 'realist' argument in my sense: (a)
(b)
(c)
A distinction between scheme and content is assumed - e.g., one between concepts and intuitions, or thought and the objects of thought, or words and the world. The internal coherence of elements of the 'scheme' side of the contrast is not thought sufficient to insure that genuine knowledge is had; some further 'legitimation' is wanted. The legitimation is thought to be something which can be achieved in an armchair, and without special consideration of the details of the knowledge-claims being legitimated.
Obviously, however, not all 'realist' arguments are transcendental, since these first three conditions could be nlet by, say, a Thomist argument for realism based on an hylomorphic analysis of substance, or Spinoza's argument that, for metaphysical reasons, "the order and connection of things is the same as the order and connection of ideas." To be transcendental, to have a distinctively Kantian flavor, a realist argument must be such that (d)
(e)
The scheme-content distinction is construed as a distinction between that which is better known to us (our subjectivity, roughly) and that which is less well known to us. our 'legitimating' transcendental knowledge of the necessary truth that content will correspond to scheme, is made possible by the fact that our subjectivity (the scheme) creates the content.
For Thomists and Spinozists, knowledge of why the nonnal procedure of inquiry achieves knowledge of what is really there is not based upon the fact that the mind notiora est quam corpus. For Kant, on the other hand, the explanation of what it is to assume the transcendental standpoint depends lessentia1ly on the claim that
80
RICHARD RORTY
If intuition must conform to the constitution of the objects, I do not see how we could know anything of the latter a priori but if the object (as object of the senses) must conform to the constitution of our faculty of intuition, I have no difficulty in conceiving such a possibility. (KdrV, B xvii)
It is only because Kant took the Cartesian doctrine that "nothing is easier for the n1ind to know than itself' to entail "we can know a priori of things only what we ourselves put into them" (B xviii) and that to entail "If we can imagine X and not Y, then this imaginability gives us knowledge of the constitution of our cognitive faculties", that he thought the Copernican revolution had made possible a special sort of argument. So once one becomes dubious about the Cartesian notion of privileged access to one's own subjectivity, the status of transcendental argumentation becomes problematic. One becomes particularly dubious about (e) above the notion that we can understand a priori knowledge by grasping 'what we ourselves put into' experience, the notion that our subjectivity somehow creates the knowable world. The first reformulation of Kant which suggests itself, once one is troubled by such doubts, is that proposed by C. I. Lewis. Here the 'mind's contribution' is identified with 'language', and that contribution is consequently relativized to a scientific epoch, as in Lewis' "pragmatic conception of the a priori." But this reformulation will not suffice to reconstruct the arguments of the Transcendental Deduction and the Refutation of Idealism unless one views the difference between Kant and Hume as analogous to the difference between Euclid and Riemann, and no one with a taste for the transcendental would do that. So the second reformulation consists in suggesting that the proper surrogate fror 'cognitive faculties' is 'meaningful use of language'. This suggestion amounts to saying that there are certain features of language which are not susceptible to change, and that these do indeed form necessary features of any possible way of speaking. This is, for example, the line which A. J. Ayer took when he claimed to give a nonpsychologistic reinterpretation of Kant's criticism of dogmatic metaphysics. On 'epistemological' interpretations of the Tractatus (including Bubner's in the article I have been discussing), it is also found in the early Wittgenstein. But familiar considerations drawn from the Philosophical Investigations, and from Quine, suggest that we no more have a privileged access to the conditions of meaningful speech than we do to the constitution of our cognitive faculties. If these considerations prevail, then if we are not to abandon the scheme-content distinction altogether, and with it all hope or even need of giving a realist argument (and a fortiori a transcendental one), we must find some new way of reconstructing Kant's arguments. This way must eschew the
TRANSCENDENTAL ARGUMENTATION AND ITS FORM
81
notion of 'privileged knowledge of our own contribution'. We must, in other words, fmd some way of saying that although it is false that only a transcendental idealist (one who claims that the world we know is structured by our contributions to it) can be an empirical realist, nevertheless Kant stumbled on a new form of realistic argumentation which can be disjoined from his transcendental idealism. So, given anti-Cartesian views of self-knowledge and anti~positivist suspicions of a 'linguistic component' in our knowledge, we can give a wider and deeper range of application to Bubner's remark that The legitimation of our forms of knowledge cannot be derived from a metaphysically dogmatic principle of higher and more indubitable insight. For having such a principle at one's disposal requires the possession of a type of knowledge other than the senseconditioned knowledge of experience which is related to objective reality and whose validity it is which requires demonstration. (p. 463)
For we now see that whether or not there is a 'contribution of our subjectivity' to the world we know, we have no special knowledge of what that contribution is. Claims to know our cognitive faculties or the nature of language are no less 'metaphysically dogmatic' than claims to know that the active intellect resembles the divine in being actus purus, or that thought and extension are parallel attributes of God. Both the transcendental and the linguistic turns thus appear as minor variations on what Heidegger called "the onto-theological tradition" and Dewey "the quest for certainty". This brings us around to the positive suggestion which Bubner makes as to the structure of transcendental arguments, in a passage which in1mediately follows the one I just quoted: The legitimation of such knowledge without the aid of absolute principles is only possible as a demonstration of the lack of alternatives to that knowledge. It can only be demonstrated that in this and in no other manner is knowledge possible, and this can only be demonstrated in that alternative forms of knowledge are ruled out. This is the way of legitimation open to Kant. (p. 463)
On this view, as Bubner says, the basic strategy of transcendental argumentation is self-referential: ... actual knowledge gains information about the general structure of itself without abandoning its own sector. Such information about the general structure of knowledge becomes at least negatively available when alternatives which are supposed to abandon the proper sector of the knowledge we have prove themselves to be unmaintainable. The impossibility of structurally different alternatives to our forms of knowledge elucidates the general structure of knowledge which we in fact have. (pp. 463-4)
82
RICHARD ROR TY
If we could make good on this notion of 'ruling Qut all alternative fOffi1S of knowledge' without appealing to truths about the 'subjective' or 'linguistic' constitution of knowledge, then indeed we should have a sort of argument which, though not transcendental in Kant's original sense, because not fulfilling conditions (d)-(e) above, would nevertheless be pretty special. The discovery of such a form of argument would gratify our feeling that there is some important insight in Kant which is relevant to the defense of realism. Unfortunately, however, I do not think that we can make good on this notion. I think that (as Bennett's and Strawson's interpretations suggest) all the arguments of the 'Refutation' and the 'Deduction' do is to rule out one alternative - the skeptical, Humean, 'sense-datum experience' alternative. We do not have the slightest idea what the other alternatives might be. To put the areas of agreement and disagreement between me and Bubner in a nutshell, Bubner thinks that "the transcendental argumentation recognizes something about understanding in general" (p. 465), whereas I think that it merely recognizes that one suggested alternative to our present understanding won't work. All that the paradigmatic Kantian arguments do is to show that a given skeptic's alternative description of experience or knowledge only seems to n1ake sense because, in Strawson's words, "the sceptic pretends to accept a conceptual scheme, but at the same time quietly rejects one of the conditions of its employment."2 To show that every alternative proposed would have the same defect would be to know in advance the range of the skeptic's imagination. To know in advance that every alternative description of the content to which we hope our scheme corresponds (or, which comes to the same thing, every alternative conceptual scheme) would have this defect would be to be able to do in philosophy what nobody dreams we can do in science predict that any new theory to come along will merely be a disguised version of our present theory. Without 'metaphysically dogrnatic principles', we cannot do in philosophy what we cannot do outside it, self-reference or no selfreference. All self-referential arguments are ad hominem arguments - arguments against a certain proposal by showing that the proposal tacitly presupposes what it purports to deny. There can be no such thing as wholesale selfreferential arguments for negative conclusions. 3 To restate once more the point at which I want to break off fronl Buhner's analysis, let me cite one further passage from his paper: ... the argument of the transcendental type makes a decisive advance over the merely factual demonstration. The advance depends upon the logical moment of selfreferentiality. If it becomes apparent that even reasoning about factual forms of knowledge and the clarification of their preconditions is not possible without making use of certain
TRANSCENDENTAL ARGUMENTATION AND ITS FORM
83
elements of a form, then it is not a merely factual state of affairs which is demonstrated. Rather, it is a logical structure that shows the validity of the form of knowledge in question. Not only does it so happen that there are no alternatives, one cannot in principle conceive of such a thing as an alternative. (pp. 464-5)
In my view, only if one has some 'metaphysically dogmatic principle' to bring to bear could one argue that one's self-referential argument has revealed a logical structure rather than a matter of fact. For only such a principle could get one from a conclusion like (A)
There cannot be a sense-datum experience
(B)
All experience must be of enduring objects in space
to
or even to (B')
All experience must be as if of enduring objects in space.
If one knew that what made it impossible for there to be a sense-datunl experience was the fact that (B) or (B') were true, then one would be all set. But all Kant did was to show (A), move dogmatically to (B'), and thence to (B) by way of the general principle of transcendental idealism that an illusion to which we cannot ilnagine an alternative counts as 'empirical reality'. But nothing in the 'Deduction' or the 'Refutation' gives us an argument that 'one cannot in principle conceive of exceptions' to (B). Pace Bubner, nothing in heaven or earth could set limits to what we can in principle conceive; the best we might do is show nobody has in fact conceived of an exception. So there can be no advance on a 'merely factual demonstration' by introducing selfreferentiality . 3. REALISM VS. PRAGMATISM
So far I have been arguing that if one preserves the usual intent of transcendental arguments the legitirnation of the claim of a scheme to apply to a content, thus giving a defense of realism - one is not going to get more than a particular negative argument against a particular suggested alternative scheme. I now want to work my way round to 'pragmatist' transcendental arguments, of which my paradignl is Davidson's argument against the notion of 'conceptual scheme'. But before discussing Davidson I need to say some general things about pragmatism, and some particular things about Peirce. When Peirce defined truth as what would be believed at the end of inquiry,
84
RICHARD RORTY
and reality as what would then be being talked about, he was making the same move as the British Idealists made when they took the common teaching of Kant and Hegel to be that truth consisted in the ultimate coherence of representations rather than in the correspondence of representations to reality. Peirce could have left out the eschatology and just said that truth consisted in that theory which best predicted and accounted for everything which one could ever wish to account for - where 'best' here means "optimizing all the usual desiderata of theories - elegance, simplicity, familiarity, and the like." All that his move amounts to is denying that we can give sense to the question 'Is our best explanation true?' As Putnam has recently said, the whole issue between realists and pragmatists comes down to the question of whether "truth is radically non-epistemic", whether "the theory that is 'ideal' from the point of view of operational utility, inner beauty and elegance ... 'conservatism', etc. might be false."4 One might put it slightly differently by saying that the whole question comes down to "Can we give a non-trivial noncircular reason for saying that we ought to adopt the theory which is 'ideal' in this sense?" If we can, then realistic projects of 'legitimation' make sense. If we cannot, then we are in a position to counter Kant by saying that 'Only the pragmatist can be an empirical realist' - that is, only if we give up the notion of legitimation can we rest content with accepting the knowledge-claims of science at face value, since no skeptical attack, and no anti-skeptical project of legitimation such as Kant's, can succeed. The only difference between pragmatism and idealism, from this point of view, is that the idealists thought that a ll1etaphysical doctrine (the mental character of the object to be represented, a character which ensured the 'union of Subject and Object') was required to get rid of the scheme-content distinction, whereas the pragmatist thinks the only argument needed is a practical one, the argument that nothing is gained by raising questions of correspondence once coherence is assured. Idealism thus was able to fend off the problem of a possible plurality of ideally predictive and elegant theories by the Spinozist device of claiming that ideal coherence somehow resulted in coincidence with the object. Peirce's own late idealistic doctrine of 'evolutionary love' was merely a Darwinized version of idealistic pantheism - an instance of what Lovejoy called the "ternporalization of the Great Chain of Being". It remained for Dewey to insist that neither science nor metaphysics was needed to guarantee that a unique best theory should be discoverable, that we could simply do without the notion of 'true' if we had that of justification, of 'warranted assertibility'. It might seem that this Deweyan rejection of the notion of 'true' is not
TRANSCENDENTAL ARGUMENTATION AND ITS FORM
85
something one could argue for. When pra.gmatism reaches this point it looks as if it were simply turning its back on a tradition (the philosophical quest for sufficient reason, roughly) rather than claiming to have arguments against the assumptions of that tradition. But this is not really the case. Pragtnatists can still bring forward arguments to- show the futility of epistemology - of projects of legitimation based on attempts to explain why our conceptual scheme has some special relation (corresponding, constituting, asymptotically approaching) to the content which fills that scheme. F or my present purposes, the interesting point is that these arguments exhibit the same sort of self-referentiality as Bubner finds in the. paradigmatic Kantian arguments. The general strategy which pragmatists us~' against realistic attempts to find some such special relation is to say that ttJ:e attempt to step outside of our current theory of the world and evaluate it by reference to its ability to 'fit' or 'cope with' the world is inevitably as self-deceptive as was Hume's attempt to escape from the Kantian categories into a world of sense-impressions. To give a better idea of what I have in mind, let me cite Sellars, Rosenberg, and Putnan1 as three philosophers who have attempted to go beyond pragmatism, by describing a tie between language and the world which will enable us to say something about why the ideally coherent theory is also one which corresponds to the real. Their projects are appropriate targets for the sort of pragmatist transcendental argument which I am attributing to Davidson. Sellars' version of the project is a convenient starting-point, since he starts with an explicit criticism of Peirce: Peirce himself fell into difficulty because, by not taking into account the dimension of 'picturing', he had no Archimedian point outside the series of actual and possible beliefs in terms of which to define the ideal or limit to which members of this series might approximate. 5
By "not taking into account the dirnension of 'picturing' ", Sellars means that Peirce did not see the need to explain what was special about the end of inquiry. He thus cOl.Hd not "make meaningful the statement that our current conceptual structure is both more adequate than its predecessors and less adequate than certain of its potential successors." (p. 138) Sellars explicates the superiority of the end of inquiry as· 'more adequate picturing', where 'picturing' is a non-intentional relation between inscriptions and other objects in the world, one belonging "in quite a different box from the concept of denotation and truth" (p. 136). (Sellars treats these two latter notions as definable only relative to a conceptual scheme, whereas 'adequacy of picturing' is a trans-scheme notion.)
86
RICHARD RORTY
Now the obvious reply to this Sellarslan strategy would seem to be that picturing cannot be, as Sellars says it is, a "matter-of-factual relation". For Gombrich-like reasons,and indeed for Peircian triadomanic reasons, we can insist that all we can have are pictures for somebody according to some convention of representation. The realization that when we reach a convention 0 representation we reach rock-bottom was, perhaps, what disgusted the Wittgenstein of the Investigations with the Tractatus. 6 So why does Sellars want to hold on to the notion of the universe getting itself pictured according to no convention? What sense can we attach to the notion that the universe is better pictured at the end of inquiry that is not parastic on the sense in which the universe is better explained at the end of inquiry, or better coped with, or better enjoyed? I do not find very clear responses to this line of questioning in Sellars, but there is a discussion of the matter by Rosenberg which suggests the line which Sellars might take. Rosenberg follows Sellars in developing a Tractarian notion of 'protocorrelational isomorphism', and he outlines this idea as follows: regularities of co-occurrence and exclusion among the empirical characteristics of the represented, objects will be mirrored by corresponding regularities of co-occurrence and exclusion among the empirical characteristics of the representing objects in virtue of which the latter are the claims which they are 7
and it is because a language is the evolutionary product of the continuing causal impact of anomolous experiential inputs to the community of representers that the structure of its extensional realization tends increasingly towards proto-correlational ismorphism with the world represented. (p. 119)
Rosenberg says that we need this notion in order to avoid Hegelianisn1, and to stick to the basically Kantian strategy which he shares with Sellars. The passage in which he seems to me to put his defense of this framework n10st clearly is one in which he is challenging an essay of mine defending Hegelian historicism against realism. He says there: While correspondence cannot be the test of truth, it is a consequence of my arguments that truth in the limit consists in an ideal ison10rphism between representing system and represented world. It is precisely by treating (ideal) truth as a limit concept that one can avoid both the Scylla of unknowable classical correspondence independent of theorizing in general, as espoused by Rorty's 'realist', and the Charybdis of viewing (ideal) truth as consilience with an ideal set of concepts deployed by us here and now, as espoused by Rorty himself. (p. 141)
IHere I take Rosenberg to be replying to the sort of rhetorical questions I
TRANSCENDENTAL ARGUMENTATION AND ITS FORM
87
addressed to Sellars by saying: there is no test for picturing, for protocorrelational isomorphism; one cannot short-circuit the usual tests of scientific theory by discovering Nature's Own Convention of Representation and then setting about to increase protocorrelational isomorphism; there is no such convention of representation; protocorrelational picturing is, in so far forth, sin1ply unlike Gombrichian picturing. The clain1 that our descendant's theory increases isomorphism is an untestable claim. But it is not put forward as an empirical thesis, it is put forward just for the purpose which Sellars said it had: to give sense to the notion of a limit, not to help us find out where we are along the series that leads to that limit. If I construe Rosenberg correctly here, the crucial question is whether we need two limit-notions or one. The limit-notion which the pragmatist has is not one of an Archimedean point. It is just that of the maximal fulfillment of the usual desiderata of theories - predictive power, simplicity, etc., etc. The limit-notion which Sellars and Rosenberg want, however, is one which answers Putnam's question about why the ideal achievement of these desiderata give us truth in son1e sense other than mere 'ideal assertibility'. The legitimation of inquiry which the realist wants will consist not in giving us a way of measuring theories against the world, but a reason for believing that continuing on as we have been doing (pursuing the same desiderata) is something which brings us into better contact with reality. Tractarian notions of picturing or ison10rphism are supposed to enable us to have a sense of 'contact with reality' which endorses no particular theory but rather the entire procedure of scientific inquiry. Thus it seems to confirm my interpretation of Rosenberg to find him saying, in a paper on 'Transcendental Argun1ents Revisited', that ... if we can, where Kant cannot, speak of temporally later conceptual schemes as increasingly adequate representations of things-in-themselves, it is not because we do, where Kant did not, know what things-in-themselves really are. Rather it is because, finding a process where Kant saw invariance, whereas he could only say that things-inthemselves are whatever ineffably lies beyond and outside of all possible representings, we can characterize them precisely as that toward representation of which this process inexorably tends. When Kant could only speak, we can both speak and gesture. 8
Putnam agrees with Sellars and Rosenberg that learning and using language, as well as the procedures of scientific inquiry, can be discussed without reference to the issue of realism-vs.-idealism, the notion of reference, or that of a non-epistemic, theory-independent, sense of 'true'. Roughly, one can say that he (like Sellars and Rosenberg) thinks that the conventionalism, relativism and pragmatism of the Investigations are sufficient to understand how inquiry
88
RICHARD RORTY
and language work. But, until very recent1y,'Putnam agreed that at the end of the day, when we ask not how inquiry and language work but rather why they succeed, we must come back to realism, reference, the Tractatus and a gesture towards an Archimedean point - the point of contact with the way the world is. As he put it, the way to use language and our other tools of inquiry can be understood without talk of reference and truth, but the success of these tools cannot. Or, using the alternative analogy of knowledge as a 'map': the picture theory was wrong as a theory of understanding, for reasons which Wittgenstein hin1self very well brought out; but not totally wrong as a theory of language functioning. It is essential to view our theories as a kind of 'map' of the world, realists contend, if we are to explain how they help us to guide our conduct as they do. But the 'use' theory is also right as an account of how language is understood. And the insights are not incompatible; a map, after all, is only a map by virtue of being employed in certain ways, but that insight does not contradict but only supplements the fact that a map is only successful if it corresponds in an appropriate way to a particular part of the earth ... 9
This analogy seems to me to express the same point which Sellars and Rosenberg want to make, and to bring the issue to a head. Pragmatists do not think it essential to view our theories as a kind of 'map' of the world. They think that if you cannot give a meaning to 'better picturing' in practice you cannot give it a meaning in theory either. But for the moment I simply want to spotlight Putnam's own recent cricicism (in 'Realism and Reason') of his earlier view, and thus, implicitly, of Sellars and Rosenberg. Putnam there distinguishes 'internal realism' - the uncontroversial claim that any sufficiently reflective scientific account of man-in-the-world will view men as constructing symbolic representations of their environment, from 'metaphysical realism', which he previously advocated, but now regards as 'incoherent'. Taking 'reference' as the notion which is to play the role of Sellars' 'picturing' and Rosenberg's 'ismorphism', Putnam argues that there will always, trivially, be a way to 'divide up theworld' so as to produce relations of reference between bits of the world and the terms of a theory which will make the ideal theory come out true. Just so, one might say with Gombrich that it is trivially true there is some convention of picturing which will make any piece of the universe a completely adequate picture of any other piece of the universe. Nor can the 'right' relation of reference (the analogue of the 'right' convention of picturing) be isolated by using a 'causal' theory of reference, since, as Putnam says, "how 'causes' can uniquely refer is as much of a puzzle as how 'cat' can, on the metaphysical realist picture." Given the psychological and physical theories which every successive scientific world-picture generates, to
TRANSCENDENTAL ARGUMENTATION AND ITS FORM
89
be sure, there will always be the ability to t~~~ce causal lines between the referent and the referring expression, but this ability is obviously theory-relative. It is thus of no help in making meaningful the notion of an Archimedean point which lies outside the series of actual and possible beliefs. It will only be part of the 'internal realism' which Putnam now says is "all the realism we want or need." The moral of this argument that every attempt to explain the success of the map will trivially succeed seems to n1e (though not, I gather, to Putnam) to be that we should give up the map analogy. But for the moment I want simply to note that Putnam's argun1ent brings out very nicely the central role of self-referential ad hominem arguments in defeating the attempt of the metaphysical realist to give content to the map analogy, when this analogy is applied to all possible theories on the one hand and 'the way the world is' on the other. Every time the realist suggests that there is a way to give content, Putnam will point out that the notion which is used to give this content will be as ad hoc as the 'unGombrichian notion of picturing' which I attributed to Rosenberg. In other words, each such notion will be like Kant's notion of 'the noumenal ground of the phenomenal', a use of an ordinary term ('ground', 'picture') in a sense which breaks away from its ordinary set of connotations and presuppositions. No phenomenal term can describe the phenomenonnoumenon relation, and we have no other tenns. As Putnam says: If concepts are particulars ('signs'), then any concept we may have of the relation between a sign and its object is another sign. But it is unintelligible, from my point of view, how the sort of relation the metaphysical realist envisages as holding between a sign and its object can be singled out either by holding up the sign itself, thus
IcoWI or by holding up yet another sign, thus
IREFERS! or perhaps
ICAUSES 1. 10 I take it that Putnam would say the same thing about holding up a sign which read
IPICTURES MORE ADEQUATELY( or one which read jPROTOCORRELATES ISOMORPHICALLYI.
90
RICHARD RORTY
This general form of argument says that,the metaphysical realist is doing just what Strawson says the skeptic did - viz., "pretending to accept a conceptual scheme" (viZ., that of relations of causality or resemblance given sense and concreteness by a scientific theory - the relations countenanced by what Putnam calls "internal" realism) while "at the same time qUietly rejecting one of the conditions of its employment" (viz., that these relations be united with the terms of the current scientific world-view and thus made 'internal' rather than remaining merely 'metaphysical'). So I think that the lesson to be learned from Putnam's criticism of his earlier self is that 'matter of factual' and 'extensional' relations (those described by Sellars and Rosenberg in their accounts of picturing or isomorphism, and those discussed by Kripke and his followers in their causal theories of reference) will prove no better than 'metaphysical' relations for purposes of giving realistic legitimations of knowledge. This seems to me to show that the metaphysical realist and the skeptic are playing the parts assigned by Kant to the dogmatist and the skeptic of his day. They both share common assumptions (roughly, those making up the 'map' analogy for empirical knowledge) which need to be transcended by transcendental argument if we are to avoid pointless and interminable Sprachstreit. I think that Putnam's argument is an instance of the sort of argument which, in Section 1, I claimed was the only sort of transcendental argument available: viz., ad hominem criticism of particular attempts to provide an interpretation of the 'map' analogy - the relation of correspondence to the way the world is - which is neither trivial nor metaphysical. 4. NATURALIZED EPISTEMOLOGY
I shall soon be saying that the presuppositions shared by the skeptics who think that knowledge cannot be legitimated and the metaphysical realists who boldly try to legitimate it are just the presuppositions which Kant foisted upon analytic philosophy, those presuppositions which make up what Davidson calls "the third and last dogma of empiricism - the distinction between scheme and content". But I want to put this off for one more section in order to take up a curious fact about recent metaphysical realism: the latest metaphysical realists pricte themselves on their naturalism, identifying the goal of realistic inquiry with the final, naturalistic, spatio-temporal, world of which empirical science is gaining knowledge. This cry of 'let's naturalize epistemology!' has become suspiciously popular within analytic philosophy in recent years. I regard it as a reactionary development, a last-ditch attempt to hold on to the atemporal, Kantian, scheme-content distinction in a desperate
TRANSCENDENTAL ARGUMENTATION AND ITS FORM
91
attempt to avoid Hegel and historicism. So I want to spend some time isolating it as a target, before saying why I think Davidson has shown how to destroy it. The obvious line for a pragmatist to take in regard to the naturalization of epistemology is to say that the sciences of nature are concerned with causal processes, whereas epistemology is concerned with social practices of giving justifications. Put in these terms, 'naturalized epistemology' seems to involve just the naturalistic fallacy of which Sellars has accused traditional empiricism: the Locke-Hume project of grounding knowledge claims on psychological mechanisms. ll Those of us who learned from Sellars to think of the Myth of the Given as a confusion of causal conditions with justifications are inclined to think that the project of finding connections between inquiry and the world needs elimination rather than naturalization. If one draws this moral one will not seek 'an Archimedean point outside the series of actual and possible beliefs'. So we pragmatists mourn Sellars as a lost leader. The pragmatist wants to define both the epistemological and the ethical ideal - the True as well as the Good - in tenns of continuing along present lines as far as we can, rather than in tenns of correspondence to something out there ('the way the world is', 'the Form of the Good'). Or, to deemphasize the eschatological once again, he wants to define these ideals in terms of the whole tangle of con1plex historically detem1ined factors which give us our day-to-day, smug, pre-philosophical sense of intellectual or moral progress. Sellars, Rosenberg and Putnam, alas, tum Kantian at the last moment, dividing the intellectual and the moral spheres from one another by thinking of the one as aiming at correspondence and the other not. The pragmatist, by contrast, breaks down the distinction between pure and practical reason, the normative and the descriptive. He avoids the Sellarsian cl~im that inquiry into empirical truth can be viewed as an attempt to picture, whereas inquiry into moral truth can only be viewed as an attempt to hold society together. Quine first gave popularity to the notion of 'naturalizing' epistemology by announcing that we should give up the notion of epistemology as 'first philosophy' and view it instead as psychological inquiry into the relations between theory and evidence. But this project transparently involves the traditional empiricist confusion between retinal irradiation, e.g., as stimulus and as premise (the confusion between causation and justification which Sellars diagnoses in Locke ).12 Despite this objection, however, the attempts by Fisk and Kripke and Boyd to revive Aristotelian notions of essence and natural necessity, as well as the popularity of 'causal theories' of knowledge
92
RICHARD RORTY
and reference, have led many philosophers to take it as ahnost self-evident that the proper upshot of Quine's destruction of the traditional dogmas of empiricism is not pragmatism, but a kind ofscientism , an attempt to become empirical about traditional epistemological problen1s. The most prevasive element in this attitude is 'evolutionary epistemology' - the notion that justification can be seen as biological rather than merely social. From this perspective, what counts as a good reason is not just a matter of what our society views as the direction of intellectual progress, but what Nature itself so views, for we are, after all, the latest of the sons of time. It seen1S reasonable to take this strategy of looking to biological evolution for legitimation of our practices of justification as central to the atternpt to naturalize epistemology. For the Quinean project of studying the relations between evidence and theory (sensory stimulation and verbal autpu t, roughly) will not be able to link up with the traditional desiderata of epistemology justifying our practices of justification - unless some such notion is supplied. Psychological study of, e.g., 'quality spaces' and inductive practices will be of no use, for example, in supplying the chief desideratum of much recent writing on the theory of knowledge - an answer to Kuhn and Feyerabend in the form of ahistorical and theory-independent definitions of 'objectivity' and 'rationality'. So it seems fair to focus upon evolutionary justification. It is also convenient, because Rosenberg is one of the clearest exponents of this notion, and because he has put it forward in the context of his reformulation of the notion of transcendental argumentation. Rosenberg follows in the steps of Lewis' pragrnatization of the a priori and Sellars' identification of the thing-in-itself with the objects of the 'scientific image'. Sellars had claimed that the true ground for the transcendental ideality of the perceptual world lies in the distinction between perceptible physical objects and the objects of theoretical science, a distinction which was blurred by Kant!3
Rosenberg generalizes this strategy by saying The categorial object-concepts with which Kant finds himself cannot be derived from experience ... Kant is therefore limited to asking how he can justify his employment of them from within. And this, in tum, constrains his understanding of the character of a transcendental deduction. We, in contrast are epistemological naturalists, and we stand in an historical setting from which we have observed theory succession in the natural sciences ... we can thus see Kant's categorial necessity as a species of natural necessity writ large. We can in other words, apply the evolutionary viewpoint to epistemology ... 14
TRANSCENDENTAL ARGUMENTATION AND ITS FORM
93
This strategy produces the following construal of transcendental argumentation: In any experience, we conceive the world we experience through the medium of a particular conceptual core. What Kant taught us to do is to be aware of this as a conceptual practice, a piece of cognitive conduct, which stands in need of legitimization. A transcendental deduction is to be an argument that legitimizes such a practice, that gives us license and title to engage in it. What the standpoint of evolu tionary naturalism implies, however, is that the argument will support this conceptual practice only when it is regarded as a reconceptualization of the experienced world under a second, successor, set of concepts, a world also conceivable (and, typically, previously conceived) under a different set of concepts constitutive of a predecessor core. 1S
Thus "the logical skeleton of a transcendental deduction," Rosenberg says, is something like this: We shall achieve the epistemic end E. The best or the only way to achieve E is to adopt or espouse or employ conceptual cores having certain relevant features F. The successor core Cs stands to its predecessor Cp in those relations in virtue of which Cs has the features F. Hence we may - indeed, we ought - adopt or espouse or employ the conceptual core C s· 16
This construal of transcendental argumentation seems to make such argumentation no different from any argument for changing from an old theory to a new and more explanatorily powerful theory. Rosenberg himself seems to take this view when he says that in the setting of evolutionary naturalism, a transcendental deduction is no different from a Peircean abduction, nor is the integrative synthesis of our experience as of a unified world anything other than what Peirce called 'the fixation of belief'. For Peirce was - in the deepest, purest, and most insightful sense - a Kantian. 17
Rosenberg thinks that the central Kantian insight which Peirce preserved was that "a transcendental deduction must be a piece of practical reasoning." I find it difficult, however, to see how Kant so constnled is different from what Rosenberg castigates as "neo-Hegelian chicanery ."18 For the self-referential feature which Rosenberg, like Bubner, fmds central to the transcendentality of transcendental arguments is now simply that "the question of justification or legitimacy can only be asked from within the standpoint of the successor core."19 But the same goes for any Kuhnian scientific revolution - we are
94
RICHARD RORTY
only going to have the facts and observations.~d experiences for which the post-revolutionary theory accounts once we are within the post-revolutionary scientific world-view. So I think Rosenberg is really saying: the self-referential aspect of transcendental arguments does not give their conclusions some sort of 'logical' (as opposed to 'factual' status) nor some 'necessary' (as opposed to 'empirical') truth; they are merely arguments for scientific revolutions writ large. The problem, of course, is that now it is not clear what epistemology adds to science. It is not clear what philosophers can do to legitimate scientific knowledge-claims which the scientists haven't already done for themselves. Rosenberg is careful to say that E - the 'overriding epistemic end-in-view' which is invoked in the first premise of transcendental deductions, is not "the attaining of a better fit between our systenl of representations and how things-in-themselves stand."20 Rather, we must, with Kant, avoid this 'innocent realism' by saying that E is 'integrative success': ... what legitimizes our adoption and employment of a successor core is that the reconceptualization of our world permits its reunification as an orderly single world in which the disconnected regularities experienced as in the world through the medium of the predecessor core re-emerge as regularities of appearance in the successor, and can be so experienced as thus not in the world but in us.... Epistemic evolution thus mirrors biological. Progress in each is measured by the successive emergence of more and more thoroughly unified, structured, and integrated systems. 21
This gives a fine description of the relation between pre- and post-revolutionary scientific theories (construed in the manner of Kuhn), but it seems entirely intelligible without the notion of 'protocorrelational isomorphisnl' or any other such attempt to make sense of 'an Archimedean point outside the series of actual and possible beliefs.' So I think that Rosenberg is confusing (A)
Using biological evolution as a model for scientific theory change
which seems a good idea, and (B)
Using the presumptive successful adaptation of our species to its environment to underwrite the direction of scientific progress since the seventeen th century.
As far as I can see, only (A) is defended in Rosenberg's discussion of transcendental argunlents, and only (B) is defended in his defense of transcendental realism against my "neo-Hegelianism", as Rosenberg calls it. Rosenberg holds that
TRANSCENDENTAL ARGUMENTATION AND ITS FORM
95
it is a consequence of n1Y argument that truth in"the limit consists in an ideal isomorphism between representing system and represented world 22
But, as I have already said, this is arrlbiguous between a limit defined internally by reference to the ideal of maximizing coherence and one defined externally ('metaphysically') in Putnam's sense - Le., one such that the ideal theory might be false. (A) is compatible with the first sort of lin1it, and (B) with the second. For (B) raises the question whether a stepmotherly Nature might not have arranged things so that the most successful species should be one which saw her not as she was but under some advantageous illusion. (B), it seems to me, both isolates what would be special and interesting about evolutionary epistemology, if such a thing were possible, and makes clear why it is not possible. To sum up, I think that any project which tries to use the results of empirical science to do what Locke tried to do with them - separate out the 'rational' and 'scientific' elements in culture - is doon1ed. My examination of Rosenberg's refoffi1ul ation of transcendental argumentation seems to me to show that the line between neo-Kantianism and neo-Hegelianism, realism and pragmatism, has becon1e so thin in the works of the naturalizing epistemologists that the difference between a naturalized Kant and a naturalized Hegel is no longer important. To put it less cryptically, once one takes away the notion of a distinction between the empirical and the a priori, the necessary and the contingent, a transcendental deduction and a Peircean abduction, the question of whether one is praising or burying transcendental arguments becomes moot. 5 . DAVIDSON'S ARG UMENT
I come, at last, to Davidson's argument against the dualism of scheme and content, and of organizing system and something waiting to be organized."23 Davidson aims his argument against the sort of scepticism (typified by Lehrer) which suggests that our descendents might use son1e conceptual scheme utterly different than our own. Lehrer takes Quine's treatment of analyticity to show that no transcendental argument can ever succeed, since even if 'the application of a concept entails the logical impossibility of misapplication', nevertheless "no matter how well entrenched a concept may be in our beliefs about the world, it remains always and constantly subject to total rejection."24 For Lehrer, any claim (such as Kant's) about the necessary structure of our experience neglects "the ubiquity of conceptual application and
96
RICHARD RORTY
conceptual change in all human thought."25 ~ehrer's position is close enough to the one often (wrongly) attributed to Kuhn to make clear the relevance of what Davidson is doing to the controversy surrounding so-called 'Kuhnian relativism.' This controversy is well described by Putnam as arising from fear of a 'meta-induction' from past instances of conceptual change to the conclusion that just as no term used in the science of more than fifty (or whatever) years ago referred, so it will turn out that no term used now (except maybe observation terms, if there are such) refers. 26
Davidson can be taken as saying that both premise and conclusion of this meta-induction are unintelligible. Since we cannot n1ake sense of the notion of a conceptual scheme made up of entirely different referring expressions fron1 our own, neither half of the sentence quoted from Putnam can be true. Most of our beliefs are true and most of the beliefs of every translatable generation of ancestors were true. Most of the tenns used in sentences expressing those beliefs referred. Translatability requires massive consensus and truth cannot outstrip translatability. I shall quote the upshot of Davidson's argument, before giving the argument itself: It would be wrong to summarize by saying that we have shown how communication is
possible between people who have different schemes, a \vay that works withou t need of what there cannot be, namely a neutral ground, or a common coordinate scheme. For we have found no intelligible basis on which it can be said that schemes are different. It would be equally wrong to announce the glorious news that all mankind - all speakers of language, at least - share a common seheme and ontology. For if we cannot intelligibly say that schemes are different, neither can we intelligibly say that they are one. In giving up dependence on the concept of an uninterpreted reality, something outside all schemes and science, we do not relinquish the notion of objective truth - quite the contrary. Given the dogma of dualism of scheme and reality, we get conceptual relativity, and truth relative to a scheme. Without the dogma, this kind of relativity goes by the board. In giving up the dualism of scheme and world, we do not give up the world, but reestablish unmediated touch with the familiar objects whose antics make our sentences and opinions true or false. 27
This amounts to a repudiation of the notion of 'a veil of representations' the Kantian counterpart to the Descartes-Locke notion of a 'veil of ideas' which makes scepticism possible and legitimation-projects desirable. Davidson's point, however, is not just one more Thomistic or Deweyan polemic against the Cartesian use of cogitatio, or the Kantian use of Vorstellung, to ,separate us from the world. It is backed up by a self-referential argument 'lagainst the coherence of this notion of a veil, when the veil is construed as a
TRANSCENDENTAL ARGUMENTATION AND ITS FORM
97
'conceptual scheme' and thus as one,~ong many 'alternative languages.' Davidson says that an alternative language, if it is to be of more philosophical interest than the mere notion of a set of beliefs so different from ours as to leave little common ground for debate (like Kuhn's revolutionary physical theories or Whorf's Amerindian languages),28 must be the expression of a set of beliefs, which are "largely true but not translatable."29 The point is to substitute the unexciting notion 'true' for the philosophically fascinating notions of 'organizing' or 'fitting' the facts or the world. Or, to put it another way, the point is to reject analyses of what it is for a sentence to be true which employ these uncashable metaphors. Davidson says that language cannot be thought of as 'organizing' experience because "the notion of language applies only to pluralities" and "whatever pluralities we take experience to consist in ... we will have to individuate according to familiar principles. A language that organizes such entities must be a language very like our own."30 If we attempt to substitute 'fitting' experience or reality for 'organizing' it, then the trouble is that the notion of fitting the totality of experience, like the notions of fitting the facts, or being true to the facts, adds nothing intelligible to the simple concept of being true. 31
Neither metaphor is any better off than Kant's notion of 'synthesis.' If we have no idea (as, ex hypothesi, in Kant, we do not) of what unsynthesized intuitions are like, we do not know what it is for concepts to synthesize them. If we do not know what an un-pluralized experience is like we do not know what it would be like to organize it. If we do not already know lots of sentences which are true of reality, we shall not gain understanding of this wordworld relation by evaluating 'fit' or 'correspondence.' Thus Davidson's self-referential arguments against the standard neo-Kantian metaphors let him boil down the issue about 'conceptual schemes' to the question "can we make sense of the notion of 'true but not translatable' without the help of such metaphors"? It is here that his own approach to semantics, the approach which has emerged in a succession of articles beginning with his 1967 paper 'Truth and Meaning'32 - comes into play. In that paper he identified giving a theory of the meaning of the expressions in a language v/ith giving a Tarskian theory of the conditions under which the sentences in the language are true. He thus attempted to undermine the distinction between a theory of meaning and a theory of reference, saying that the fonner (when made an empirical attempt to understand the functioning of the language, and fre,ed of the 'adventitious philosophical puritanism' which inspired the
98
RICHARD ROR TY
positivistic attempt to limn the ultimate structure of reality in canonical notation )33 collapsed into the latter. The result of this approach to the notion of 'meaning' is to say that the notion of a 'set of 'alternative meanings' those attaching to the expressions of an untranslatable language - is as unintelligible as the notion of 'alternative truth'. Without attempting to expound, much less to argue for, this approach, I shall simply cite the passage in which Davidson brings it to bear on the question at hand: We recognize sentences like" 'Snow is white' is true if and only if snow is white" to be trivially true. Yet the totality of such English sentences uniquely determines the extension of the concept of truth for English.... According to Tarski's convention T, a satisfactory theory of truth for a language L must entail, for every sentence of L, a theorem of the form '8 is true if and only if p' where 's' is replaced by a description of 8 and 'p' by 8 itself if L is English, and by a translation of 8 into English if L is not English ... Convention T suggests, though it cannot state, an important feature common to all the specialized concepts of truth. It succeeds in doing this by making essential use of the notion of translation into a language we know. Since Convention T embodies our best intuition as to how the concept of truth is to be used, there does not seem to be much hope for the test that a conceptual scheme is radically different from ours if that test depends on the assumption that we can divorce the notion of truth from that of translation. 34
The important premise here is that Convention T 'embodies our best intuition as to how the concept of truth is to be used'. That premise can have no defense save self-referential arguments of the sort Davidson presents seriatim against attempts to develop intuitions which take us beyond Tarski's. Tarski, on Davidson's interpretation, is telling us that we are not going to have anything more than the everyday, philosophically innocous, sense of truth. 35 The claim is essentially negative - it does not say that because Convention T gives us 'the nature of truth' we are wrong to look for Putnam's 'non-epistemic' sense of truth. It merely says that each of the various metaphors ('synthesizing', 'fitting,' 'organizing') which metaphysical realists have employed dissolve when we attempt to put them to work. So Davidson is not deducing the need to abandon the scheme-content distinction from Tarski. He is saying that when the metaphors are abandoned, we still have what we need to understand how language works - for Tarski gives us that. Given Tarski and common sense, the claim that 'n10st of our beliefs are true' seems a platitude. Given Kant, the scheme-content distinction, the fear of relativism, and other elements of a sound, professional, philosophical training, the same claim will look like dogmatism. But, as the passage I quoted from Davidson earlier says, it looks that way only if truth is thought of in the non-epistemic way which he wants to abandon. The attempt to think of truth
TRANSCENDENTAL ARGUMENT ATION AND ITS FORM
99
in this way is so much a part of what it is to be a philosopher that we feel that unless Davidson has a knockdowri~ argument based on some hitherto undreamt-of fact about the nature of language, we need not take him seriously. But all he can offer us is particular self-referential ad hominem arguments presented seriatim against particular attempts to think of truth in the desired way. There is no general argument against the scheme-content distinction. There are simply the same sort of arguments as have always been given against attempts by the sceptic to "pretend to use a conceptual scheme while quietly rejecting one of the conditions of its employment." I have presented Davidson's argument as the most effective recent argument against the possibility of transcendental philosophy. But it is equally important, for my purposes, to see it as itself a piece of transcendental argumentation. If one were to fill out the details of Davidson's arguments against glossing 'true' by 'fitting the facts' or 'organizing experience' they would, I think, be seen still more clearly as instances of the same sort of argumentation as Kant's criticisms of Hume, or Putnam's criticism's of his earlier self. The moral I want to draw is that this sort of argument works not just against 'legitimating' attempts to bring scheme and content together, but against the whole problematic of legitimation created by the scheme-content distinction. 6. CONCLUSION
Kant has often been praised for transcending the entire problematic which his predecessors - both the rationalist dogmatists and the empiricist sceptics shared. He was the first, we say, to see that epistemology can be neither Lockean psycho-physiology nor metaphysical descriptions of God's relation to Nature. His insight, we are told, lay in distinguishing the question of legitimation, the quaestio juris, from any scientific or metaphysical quaestio facti. But, as I said in Section 2, we are now inclined to say that Kant still shared too much with the Cartesian tradition he tried to overcome. Specifically, he sometimes seems to have done little more than substitute a description of a mysterious noumenal cognitive faculty for a 'physiology of the human understanding' or for a dogmatic metaphysics, thus once again reducing the quaestio juris to a quaestio facti. Even such devoted admirers of Kant as Sellars and Strawson speak of Kant's "fictitious mental machinery" and his "imaginary subject of transcendental psychology ."36 The desire to preserve Kant's notion of 'legitimation' - of the distinctiveness of a quaestio juris - leads Bubner and Rosenberg to construe transcendental arguments as 'practical' rather than 'theoretical'. They both, if I under-
100
RICHARD RORTY
stand them correctly, are trying to avoid answering a question about legitimacy with a description of a state of affairs, for any such answer would once again coalesce the quaestiones juris et facti. I think that this strategy is the only way which has any hope of giving sense to the notion of transcendental philosophy as critic of culture - i.e., as legitimator of clain1s to truth, rationality, or objectivity on the part of other disciplines. But I do not think this strategy can succeed, and thus I think we need to question the notions of 'legitin1ation', of 'transcendental philosophy,' and of Kant as paradigmatic philosopher. Kant built the Cartesian quest for certainty, for legitimation, for guarantees of rationality, into philosophy's self-image, at the decisive mon1ent when our discipline became professionalized and institutionalized. It would be tempting to call this a cultural disaster, were it not that the splendors and miseries of his project were equally necessary conditions for the emergence of Hegel and the nineteenth century. So it is fairer to say that the real disaster was the return to the spirit of Kant by Russell and by Hussed early in this century. Because of this return, the pragmatist and historicist spirit which had begun to enlerge in the nineteenth century were largely lost to philosophy. Logical enlpiricism and phenomenology between them led us to spend almost fifty years in reactionary attempts to refound the notion of transcendental philosophy, and to "put philosophy on the secure path of a science." The grip of Kant upon modern thought, and the desire to recapture the transcendentality of his argumentation, can only be broken if one ceases to ask his quaestio juris. To do this, however, one has to break the spell of the dualisn1 between scheme and content. As long as we view thought or inquiry or language as something which has to correspond to something else, realism, idealism, and scepticism will seem live issues. For the question "How can we justify our knowledge-claims without falling into infmite regress, or circularity, or relativism?" will seem urgent. 37 As long as we ask how we can break out of representations to discover how well we are representing something that is not a representation, we shall either have to revive transcendental philosophy or else resign ourselves to scepticism. But most attempts to shove the quaestio juris aside (Feyerabend's, Dewey's, Popper's) seem like simple surrenders to relativism, a kind of philosophical defeatism. What one would like is some way of having an argument for one's pragmatism - one's refusal to accept the question of legitimation. The closest thing one can get to this, however, as I have suggested in the first sections of this paper, is a series of demonstrations of the incoherence of various mechanisms proposed to accomplish legitimation. However, Davidson seems to me to have given us an argument which so to speak, functions as a
TRANSCENDENTAL ARGUMENTATION AND ITS FORM
101
recipe for constructing such particular transcendental (in the sense of selfreferential) arguments. It thereby functions as a practical argument against the utility of further attempts to revitalize the notions of 'correspondence,' 'representation,' and 'legitimation.' Suppose for the moment that Davidson is right and that philosophers gradually come to see the notion of 'legitimation' as parasitic upon that of 'correspondence' and that, in tum, as dependent on the 'dogma of the distinction between scheme and content' (the last dogma not only of empiricism but of what Heidegger calls "Western metaphysics," the tradition which identifies truth with accuracy of representation.). Does that mean that philosophy simply dries up and blows away? That we simply relax into relativism? Can pragmatism be the basis for philosophy, rather than a means of ending philosophy? I think that the proper approach to these questions is to see pragmatism as intertwined with historicism, where by 'historicism' I mean the view that philosophical problems are historically and culturally determined. On this view, the sceptical attitude towards the notion of 'correspondence to reality' which we find in Wittgenstein's Philosophical Investigations derives support from the historical account of the quest for certainty offered by Dewey, and that in turn derives support from Heidegger's more comprehensive and deeper historical study. The more we see the quest for legitimation as what Wittgenstein called "a picture which held us captive", the less significant the threat of 'relativism' will become. The more we see this picture as one imposed upon us by historically intelligible forces, the less it will seem like a permanent intuition of which philosophy must take account. The more we see the period between Kant and ourselves - the period in which philosophy seaprated from science and became an autonomous cultural institution - as an historical contingency rather than a destiny, the less we shall worry about the need to 'base' philosophy on something. We shall, perhaps, be wiser and more effective contributors to our culture the less we worry about our responsibility for determining its 'legitimacy.' Princeton University NOTES 1 Bubner, Riidiger, 'Kant, Transcendental Arguments, and the Problem of the Deduction', Review of Metaphysics 28 (l92-8 , 455 -6. All further references by page number to Bubner are to this article. 1
102
RICHARD RORTY
Strawson, P. F., Individuals, London 1959, p. 35. In this paragraph I am summarizing the argument of my 'Verificationism and Transcendental Arguments' (Nous 5 (1971),3-14), which is in part a reply to Barry Stroud's 'Transcendental Arguments' (Journal ofPhilosophy 65 (1968), 241-256). 4 This quotation is from Putnam's Presidential Address on 'Realism and Reason', reprinted in his Meaning and the Moral Sciences, London and New York 1978, p. 125. This essay offers a later, different, view than that of the rest of the book. 5 Sellars, Wilfrid, Science and Metaphysics , London and New York 1968, p. 142. 6 Cf. Wittgenstein, Ludwig, The Blue and Brown Books, Oxford 1958, pp. 24, 34. 7 Rosenberg, Jay, Linguistic Representation, Dordrecht 1974, pp. 117-118. Further references to Rosenberg, unless otherwise indicated, are to this volume. 8 ,'Transcendental Arguments Revisited', Journal ofPhilosophy 62 (1975), 621-2. 9 This quotation is from Putnam's 'Reference and Understanding' - written prior to 'Realism and Reason' - reprinted at Meaning and the Moral Sciences, p. 100. 10. 'Realism and Reason', p. 127. 11 See Sellars, Wilfrid, Science, Perception and Reality, London and New York 1963, pp. 157 -161. 12 See Quine's essays 'Epistemology Naturalized' and 'Natural Kinds' in his Ontological Relativity New York 1969. I criticize this attempt to link psychology with epistemology in 'Wittgensteinian Philosophy and Empiricial Psychology' (Philosophical Studies (1977), pp.151-172). 13 Science a~d Metaphysics, p. 56n. 14 'Transcendental Arguments Revisited', p. 618. 15 Ibid., p. 620. 16 Ibid., p. 621. 17 Ibid., pp. 623-4. 18 Cf. ibid., p. 612. 19 Ibid., p. 620. 20 Ibid., p. 621. 21 Ibid., p. 622. 22 Davidson, Donald, 'On the Very Idea of a Conceptual Scheme', Proceedings of the American Philosophical Association, 1973, p. 11. 24 Keith Lehrer, 'Skepticism and Conceptual Change' in Empirical Knowledge: Readings from Contemporary Sources, (ed. by Roderick M. Chisholm and Robert J. Swartz), Englewood Cliffs, N.J. 1973, p. 50. 25 Ibid., p. 53. 26 Hilary Putnam, 'What Is Realism?', Proceedings of the Aristotelian Society, 1976, p.184. 27 'On the Very Idea ... ', p. 20. 28 Cf. ibid., p. 6. 29 Ibid., p. 16. 30 Ibid., pp. 14-15. 31 Ibid., p. 16. 32 This article appeared in Synthese 7 (1967), 304- 23. See also Davidson's 'True to the Facts', The Journal of Philosophy 66 (1969), 158-74, and his 'In Defence of Convention T' in Truth, Syntax, and Modality (ed. by H. Leblance), North Holland, Amsterdam, 1973. For some good comment on these papers, and an attempt to fill out the larger 2
3
TRANSCENDENTAL ARGUMENTATION AND ITS FORM
103
picture which they adumbrate, see Chapter 12 of Ian Hacking's Why Does Language ",:' Matter to Philosophy?, Canlbridge 1975. 33 For this notion of the function of canonical notation, cf. Quine, Word and Object, Cambridge, Mass. 1960, p. 161. For the reference to adventitious puritanism, see Davidson, 'Truth and Meaning', p. 316. 34 'On the Very Idea ...', pp. 16-17. 3S Note Davidson's remark that although Tarski's theory is a correspondence theory nevertheless "it may be the case that no battle is won, or even joined, between correspondence theories and others" ('True to the Facts', p. 761). 36 For the first phrase, see Sellars, Wilfrid, Science and Metaphysics, London and New York 1968, p. 52n. For the second, see Strawson, P. F., The Bounds of Sense, London 1966, p. 32. 37 I borrow this formulation of the problem of 'transcendental grounding' from Apel, K.-Q., 'The Problem of Philosophical Fundamental-Grounding in light of a Transcendental Pragmatic of Language', Man and World, 1975.
WOLFGANG CARL
COMMENT ON RORTY
In my comments I shall confine myself to what Rorty says about transcendental arguments as such and to his discussion of Davidson's argument that is, according to Rorty, "a transcendental argument to end all transcendental arguments" (p. 78). So I will leave out what Rorty says about pragmatism and naturalized epistemology, partly because the critique of transcendental philosophy from these philosophical points of view is well known, partly because Rorty raises issues which will be discussed as well in other papers of the symposion. Rorty talks about transcendental arguments in general without discussing any such argument, and he gives a quite general critique of these arguments. To assess his critique one has to know what Rorty takes a transcendental argument to be. Unfortunately, however, he doesn't manage to give a clear idea of what kind of argument a transcendental argument is. He seems to vacillate between two different conceptions. On the one side he tries to characterize transcendental arguments by the purpose they are used for; such argunlents are ainled at legitimizing knowledge-claims. On the other side Rorty takes the term 'transcendental argument' as "a name for a certain logical structure" (p. 77). I think that both ways of dealing with transcendental arguments are not very convincing. Argunlents given for the purpose of legitimizing knowledge need not be transcendental. Rorty recognizes arguments with this purpose that are not transcendental - e.g. realistic arguments. So transcendental arguments form one species in a broader class of arguments defined by what they are aimed at. Now, what are they aimed at? Unfortunately, Rorty doesn't make very clear what legitimizing knowledge consists in. He mentions antisceptical and anti-reductionist arguments; but he also refers to arguments with the aim "to guarantee correspondence of logic, or language ... to the world" (p. 79) and to arguments about "the claims of objectivity and rationality which various parts of culture are entitled to make" (p. 77). All these arguments are supposed to be arguments the aim of which consists in legitimizing knowledge. But what does that conle to? According to Rorty we have a class which contains arguments e.g. of Descartes, Kant, and Moore as well as arguments of Wittgenstein's Tractatus. I have some doubts whether there is a sense of 105 P. Bieri, R.-P. Horstmann, and L. KrUger (eds.), Transcendental Arguments and Science, 105-112. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
106
WOLFGANG CARL
legitimizing knowledge which is common to all~ these arguments. And this in tum depends upon whether there is genuine agreement in these different arguments about what knowledge consists in. One might for example be very much in doubt whether the knowledge Moore tries to legitimate is the same sort of knowledge Kant argues for. It seems to me that there is a real deficiency in Rorty's paper because he doesn't even sketch some explanation of the notion of legitimation of knowledge. Different philosophers have held different views about what a legitimation of knowledge might be, if only because they have held different views about what knowledge consists in. So one doesn't get a clear idea about the broader class of arguments within which transcendental arguments form a species. But this species isn't well defined either. If one asks under what conditions arguments aimed at legitimizing knowledge become transcendental, one is rather surprised to get by way of a reference to Kant an answer which is very un-Kantian indeed. The conditions (d) and (e) on p. 000 state something about what might be called the subjective factor in knowledge, but they seem to flatly contradict Kant's own and well-known position. But even if one eliminates these conditions because of, as Rorty says, "anti-Cartesian views of self-knowledge and anti-positivist suspicions of a linguistic component in our knowledge" (p. 81), one still has a transcendental argument. So it can't be the reference to the subjective factor of knowledge which makes an argument transcendental. I do not see how Rorty comes to identify transcendental arguments as a special sort of arguments within the broader class of arguments aimed at legitimizing knowledge. Neither this class nor the sub-class of transcendental arguments seems to me to be well defined. I will now come to Rorty's criticism of the "basic strategy of transcendental arguments" (p. 81) which seems to me to be concerned with transcendental arguments as defined in a loose sense by a certain logical structure. Rorty stresses a supposed feature of these arguments, nan1ely being selfreferential - an idea he takes from a paper of Bubner and which seems to go back to Hintikka. As an example of a transcendental argument he gives the following one: he sketches a game-theoretical interpretation of first-orderquantification according to which "the meaning of quantifiers lies in their r6le in guiding the processes ('games') of verifying (in principle) the sentences of our language".1 The games of verifying sentences are called "games exploring the world". Now, if you have an argument which is supposed to establish the very possibility of such a game, and if this argument is itself of quantificational form, then we have the following situation: such an argument tries to establish the possibility of games of exploring the world and this very
COMMENT ON RORTY
107
possibility is presupposed by the logical fonn oftthe argument. As Hintikka says: "The conclusion makes possible the very argument by means of which it is established" .2 I do not want to discuss the soundness of Hintikka's consideration, but only stress the following point: if this is a case of self-reference, then the thesis that all transcendental arguments are self-referential comes down to the assertion that these arguments are some kind of presupposition argument: what the argument tries to establish is presupposed by the semantic interpretation of the logical fonn of the argument. What I am here interested in is not so much this kind of argument, but the connection Rorty tries to establish between the self-referentiality of transcendental arguments and the idea of ruling out all alternative fonns of knowledge. He incorporates such an idea in the notion of a transcendental argument and tries to criticize transcendental arguments by attacking the idea of ruling out all alternative forms of knowledge. But why must self-referential arguments as such rule out all alternative forms of knowledge? Why are we forced to connect the logical form of an argument with such a purpose? I will make a suggestion as to what the connection might be seen to consist in, but first, I will discuss the allegedly pretentious idea of ruling out all alternative forms of knowledge. What is really wrong with that idea? It is not the fact that Kant - according to Rorty - only rnanaged to lule out one alternative, it's just the programme itself which shows that anybody adhering to such an idea relies upon some "metaphysically dogIl1atic principles" (p. 82). But I think that one can give a reasonable sense to that idea. What Kant tries to show is that human knowledge is possible only given certain conditions. If such an argument is conclusive what is ruled out is the possibility that there be human knowledge that does not fulfill these conditions. If you now define forms of knowledge by a certain set of conditions, which have to obtain for there to be knowledge at all, then any argument that tries to show that knowledge has to fulfill certain conditions rules out all alternative forms of knowledge, namely those fOnTIS of knowledge which are defmed by a negation of those conditions and by what is entailed by this negation. Any successful attempt to state necessary conditions will amount to ruling out all alternative forms of knowledge - not all forms, but all alternative forms. Now Rorty's attack on the basic strategy of transcendental argumentation may be looked at in either of two ways. If may be that he doubts the very possibility of giving necessary conditions for knowledge, whatever they may be. If this is his intention, then he would have to argue for that. I think, it is not sufficient to negate such a possibility by alluding to our alleged inability
108
WOLFGANG CARL
"to do in philosophy what nobody dreams we can do in science" (p. 82). The other way to take Rorty's attack on arguing transcendentally would require us to reverse his argument. He would be focussing on the way of coming to state necessary conditions of knowledge rather than on the problem of stating such conditions. You can't do that - Rorty might be taken to argue - except by a self-referential argument. So his attack would be an attack on the way - perhaps the only way - we come to state necessary conditions for knowledge. I will try to give a rather sketchy explanation of what this way of arguing would amount to. One might ask how we can be in a position to state necessary conditions for knowledge. Any such statement will be of the form 'if something counts as knowledge, it must satisfy such-and-such conditions'. But making such a statement presupposes that one understands somehow the term 'knowledge', and you can't know the meaning of this term except by relying on some instances of the concept 'knowledge' - so-called paradigmcases. So stating necessary conditions for knowledge always presupposes that there is knowledge, and you must always single out some cases of knowledge before you can try to give necessary conditions of knowledge. I do not want to discuss this argument, but only hint that Kant's arguments in the Kritik der reinen Vernunft do not seem to rely upon given cases of knowledge, perhaps in counterdistinction to his arguments in the Prolegomena. 3 To conclude, if one sees Rorty's criticism of transcendental argumentation from this point of view, it will come down to the point that you can't state necessary conditions for knowledge without presupposing that there is some knowledge. But is tb~.s really so damaging? In any case, I conclude that his criticism is not conclusive on either account. I am now passing to the final section of Rorty's paper in which he relies upon an argument of Davidson's developed in a paper entitled 'On the very Idea of a Conceptual Scheme'.4 According to Rorty this argument "works against the whole problematic of legitimation created by the scheme-content-distinction" (p. 99). Apart from some allusion to Kantian terms like 'Vorstellung' or 'Synthesis' the main point of Rorty's fmal section is the following one: if we give up with Davidson the distinction of schemea..1"J.d content we are free to abandon all the problems of legitimizing knowledge claims and we get rid of the very possibility of arguing transcendentally. Rorty thinks of Davidson's argument "as the most effective recent argument against the possibility of transcendental philosophy" (p. 99). Whether this argument is really so effective depends upon whether Davidson's criticism of the distinction betvveen scheme and content is such that it
COMMENT ON RORTY
109
excludes the possibility of legitimizing knowly,dge claims. I want to show that this is not the case. I can do this only by giving a rough sketch of what the argument consists in that plays such a prominent role in Rorty's critique of traditional philosophy. Davidson tries to argue for two connected issues. First we cannot make sense of two radically different conceptual schemes; and second we have to give up the distinction between scheme and content, - "the third, and perhaps the last dogma of empiricism", as he calls it. 5 The connection between these issues lies in the fact that we cannot make sense of two radically different conceptual schemes, because we have to give up the distinction between scheme and content. I think that Davidson's arguments need some restrictions which make it obvious that the conclusion Rorty draws from these arguments doesn't really follow. What are radically different conceptual schemes? Granting the association of having a language and having a conceptual scheme, one may say that if conceptual schemes differ, so do languages. But we may have different languages sharing the same conceptual scheme, provided there is a way of translating one language into the other. So we have a necessary condition for two languages sharing the same conceptual scheme, and if 'having radically different schemes' just means not sharing the same scheme, we get a sufficient condition for two languages having radically different schemes, - namely that they can't be translated into each other. Davidson argues that this condition can never be fulfilled by attacking the dualism of scheme and content. The dualism of scheme and content can be fonnulated in various ways, as a distinction between something organizing or fitting, e.g. language or theory, and something waiting to be organized or being fitted, e.g. experience or sensory stimuli. Now, according to Davidson this talk about a theory or a language which has to organize or to fit or to account for experiences or sensory stimuli, really boiles down to the thesis that there is something what makes our sentences or theories true. So the fundamental relation between scheme and content is that of truth, and the different accounts of the content express only different views about the "source or nature of evidence", that is "what it takes to make our sentences and theories true". 6 The upshot of this argument seems to be that we have to reformulate the distinction between scheme and content according to the requirements of the notion of truth. Now given this account of the scheme-content-distinction, what happens to the idea of two radically different conceptual schemes? Suppose, we have two schemes or languages which may both be largely true, but not translatable into each other. Now Davidson argues that we cannot really make sense
110
WOLFGANG CARL
of such an idea, because the notion of truth~ cannot, according to the famous Convention T, be divorced from that of translation. But I think there is a flaw in this argument. What follows from the assumption that "Convention T embodies our best intuition as to how the concept of truth is used", 7 is simply that there can't be a language or scheme which is not translatable into any other language. In short, we cannot dissociate the notion of a conceptual scheme from the possibility of translation. But this doesn't exclude that there may be two languages or schemes such that they can't be translated into each other. So there is nothing wrong with the idea of two radically different conceptual schemes as long as one sticks to the possibility of a scheme or language being true. Only if one adduces some reflections about how we may be in a position to recognize or understand a given conceptual scheme, one n1ight come to the conclusion that we can only understand those schemes or languages that can be translated into our own language. That gives one the right impression of what Davidson is really arguing for. He stresses the point that we cannot dissociate the notion of understanding a conceptual scheme from that of translation into a language already understood and so he makes the rather well-known point that we cannot make sense of the idea of understanding a conceptual scheme which is radically different from ours. In his own words: "Given the underlying methodology of interpretation, we could not be in a position to judge that others had concepts or beliefs radically different from our own".8 Now let's look at the connection between Davidson's critique of the scheme-content-distinction and the question of legitimation. According to Rorty, this critique amounts to an argument for "one's refusal to accept the question of legitimation" (p. 100) and "to ask Kant's quaestio juris" (p. 100). Legitimation is thought of by Rorty as an attempt "to bring scheme and content together", to bring us from our veil of representation to the world represented by our' Vorstellungen'. That suggests that the notion of legitimation is in some way dependent on the assumption that there is this famous dualism between language, theories or ideas on the one side and an ineffable world, the Given, on the other. But I can't see that there is a necessary connection between the notion of legitimation and this very dubious dualism. And I want to argue that Davidson's argument gives us all we need for making sense of the quaestio juris. Davidson's critique of the scheme-content-distinction consists in two sharply distinguished attacks on the members of this distinction. The notion of a conceptual scheme has to be abandoned because there is no way of
COMMENT ON ROR TY
111
recognizing a language with a radically different conceptual scheme from the one we use in our own language. What Davidson's critique of the notion of a scheme really amounts to is that we can't understand a given language or linguistic behavior which cannot be translated into our language and so doesn't share our conceptual scheme. This is a thesis about understanding a language or a scheme and is of no concern regarding the legitimacy of a Kantian quaestio juris. Davidson's attack on the notion of content is quite different. He doesn't really argue for the elimination of this notion, but rather for a better explanation in tenns of truth. The metaphors of a conceptual scheme fitting or organizing some content come down to "the simple thought that something is an acceptable conceptual scheme or theory if it is true".9 As the relation of scheme and content is given by the notion of truth, so the idea of a content is to be redefmed in terms of "what it takes to make our sentences and theories true". That we have to take account of something that makes our sentences true, whatever it may be, is a consequence of, as Davidson says, "the notion of objective truth" .10 As Aristotle once remarked, a sentence is true because of the fact it states. ll We cannot make our statements true just by saying so: TIleir being true depends upon something other than their being stated. 12 So given the notion of objective truth we can make sense of the notion of content as what makes our sentences or theories true. And given this notion of content we have all we need for making sense of the quaestio juris and of the business of legitimizing knowledge claims. The necessity of asking this question arises precisely because we cannot make our statements true just by saying so, because we have, with Davidson and the Convention T, the notion of an objective truth. To answer the quaestio juris, to legitimize one's knowledge clain1s, is nothing else but to give an account of or to show what makes a staten1ent true. So, contrary to Rarty, the very idea of legitimation may be founded on Davidson's argument and especially on his understanding of Convention T. 1
Universitiit Gdttingen NOTES 1 'Quantifiers, Language-Games, and Transcendental Arguments, in Logic, LanguageGames and Information, Oxford 1973, p. 103. 2 'Transcendental Arguments: Genuine and Spurious', Nous 6 (1972), 278.
112
WOLFGANG CARL
3 AA IV, 274/5.
Proc. and Addres. of the Am. Phil. Assoc. 47 (1973/4). sOp. cit., p. 11. 6 OPt cit., p. 16. 70 p. cit.,p.17. 8 Ope cit., p. 20. 9 Ope cit., p. 16. lOOp. cit., p. 20. 11 Cat. 14b 20. 12 This dependence ren1ains preserved, even if one, as Davidson does, gives up the notion of fact. Following Tarski he explains the notion of truth by the concept of satisfaction, and satisfaction is "a relation between language and the world"CTrue to the Facts', J Ph 66 (1969), 758). 4
DIETER HENRICH
CHALLENGER OR COMPETITOR?
On Rorty 's Account of Transcendental Strategies
To begin I must apologize for not being able to present carefully worked out comnlents on Richard Rorty's thoughtful and stimulating paper. I explained to the organizers of the colloquium in advance that, for reasons beyond my control, I would be unable to prepare extensive comnlents unless the paper were available to me before the middle of May. Considering the richness and complexity of Professor Rorty's argumentation, it is certainly no surprise that the paper arrived much later. So, as is usually the case, richness on one side means poverty on the other: I will have to confme myself to a few remarks. They can be taken as a list of theses, questions and objections that, although in need of further support, raise issues which may be of interest at some later point in our discussion. Rorty's paper is fully integrated into a broader strategy that he began to conceive in the late sixties and has since developed in various publications. This paper is also in itself strategic to a considerable extent. In it Rorty does not produce but rather reports arguments against features of a kind of philosophical reasoning that can be described as 'transcendental'. In a way, it is necessary to go back to these arguments in previous articles if we are to come to tenns with what Rorty says. Nontheless, I feel we should not do that - but instead begin with an examination of the new contributions to and extensions of Rorty's strategy that are the core of the paper presently under discussion. This is the way in which his argument proceeds: (a) Transcendental arguments cannot be built into a self-contained theory. They are ad hominem 9 which means they are rejections of certain particular alternatives to ,a given conceptual framework. (b) As such, they have been used traditionally as means for justifying claims to the effect that a certain fonn of knowledge is indispensable or warranted. And in this context they have a Kantian ring. (c) But they can be disconnected fronl this factual, historical origin. They then can be used against Kantian projects - even against any Kantian attempt to justify knowledge-clainls in a way that differs from the justification of ordinary scientific or nonnative theories. (d) There are two such reversed transcendental arguments in recent literature: 113 P. Bieri, R.-P. Horstmann, and L. KrUger (eds.), Transcendental Arguments and Science, 113-120. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Compan!,-!!~~t!!~_~~!,-!ioJJfl-n,(JL
_
114
DIETER HENRICH
One is Putnam's argument against metaphysical realism. It destroys one of the premises of any transcendental strategy in the Kantian style: the distinction between a world and the conditions of our knowledge of it. The other is Davidson's argument from the minimal condition for translatability. It destroys the idea of a difference between what there is and a system or competing systems of concepts by which it is organized, structured, 'combined' . (e) Once this twofold destruction is completed, the very notion of 'justifying' a conceptual scheme becomes obsolete. The Quaestio juris collapses into the quaestio facti, the question of the de facto optimal success of a given scientific theory at a given time. I want now to formulate eight theses with regard to this argument. (1) Rorty's attempt to view Putnam's and Davidson's arguments as parallel, in the sense that both are instances of transcendental argument with a negative effect, I find artful but unconvincing. Davidson's argument, by his own admission, does have the foml that is described as 'transcendental' in postStrawsonian literature; Putnam's argument, however, lacks every recognizable feature of this kind. It charges metaphysical realisnl with both a simple non sequitur (viz., that fronl internal assumptions about objects no conclusion with regard to objects in sensu absoluto can be drawn) and an ordinary inconsistency in the theory of meaning (viz., that the nleaning of 'object' in the absolute sense cannot be accounted for within the same meaning-theory that applies to internal objectivity). (2) The problem of Kant's transcendental deduction has nothing to do with the problem of nletaphysical realism. Although Kant is, as a matter of fact, a metaphysical realist, and although the paradigm case of a so-called 'transcendental argument' (probably the only case in Kant's works) can be found in his 'refutation of Idealism', his epistemological project is quite independent of these features. Regardless of whether we can think meaningfully about a world that is independent from our discourse in the external (metaphysical) sense, we can wonder what the status of the notion of an object is within this discourse, how we employ this notion, and in what way we can understand the claim that our discourse about objects yields justified knowledge. (3) Whereas the preceding thesis seems to me to be fairly obvious, this one might have a more paradoxical look. But I am prepared to defend it: It is well known that Kant believed in some variety of the theory of simple, unconnected sense data. (It affects the fonn Kant gave to his transcenIdental reasoning.) But this reasoning as such does not hinge upon it. (And
COMMENT ON RORTY
115
Sellars, for one, is very well aware of this). Let";us assume that what is given to us does not require combination because Direct Realism is true. On this assumption, we are capable of encountering nicely-structured things in the world that in no way depend upon a contribution from our side. We can still wonder in what our having knowledge of them consists. The main point Kant makes in section 15 of the second edition of the first Critique is not that a combination in the object cannot be conceived except as the product of a combinatory activity of the mind. That would be an entirely unjustified bare claim and a highly implausible one indeed. Rather, the point is that we cannot understand what a combination in an object is, in what it consists (we cannot conceive combination "as such"), apart from an activity of combining we are capable of performing. If we don't understand what it means to combine, we cannot understand explicitly what it is to be combined. And since the notion of an object is the notion of a manifold that exhibits a certain order, we then would not be in possession of the notion of an object as such. But without it 'objectivity' would also be a meaningless tenn. This is, of course, a disputable argument (its status is, by the way, not that of a transcendental argument). It is an argument, however, and an ingenious one at that. Moreover, it is strong enough to provide a foundation for a transcendental deduction if a further premise can be defended: namely, that the knowledge of what it is to be an object is a priori. I shall return, momentarily, to the question of whether one can dispense with a non-relative notion of aprioricity. At present, I think it is obvious that the Kantian program does not simply presuppose a concept/world or a schema/content distinction in the form Rorty attacks. And I doubt that Mr. Rorty would want to deny any distinction between our knowledge and discourse on the one hand, and what they are of, on the other. (4) It is a mistake to describe the transcendental strategy as the defense of one conceptual fratnework against another. The idea of competing conceptual frameworks has a later and different origin. Although this idea can be used in a reconstruction of Kant's Critique, it does not capture its original intention which on its own merits remains of philosophical importance. It was Strawson who established the unfortunate association (upon which Rorty relies) between the problem of transcendental reasoning and the possibility of competing frameworks. Getting rid of frameworks, however, does not make transcendentalism evaporate. To justify a knowledge-claim i~ not necessarily to justify it against a competitor. Because of the lack of a paradigm of his new method within philosophy Kant modelled his transcendental strategy upon certain juridical cases in
116
DIETER HENRICH
civi1law, and he obtained his tenn 'deductiOn' in its peculiar sense from the juridical literature. Such a 'deduction' consists in the proof that a certain legal claim originated under proper circumstances and is therefore justified. Now there are in fact certain court cases (disputes about property) where such a deduction is given only when a conflicting claim has been made. But there are many others where the question is simply whether a right exists or not (for instance, the claim or right to receive a pension, to impose a demand, to get a promotion, to use a title, etc.) In these cases, a challenger need not also be a competitor. It is enough that there is someone to whom the claim is addressed or to whom it applies and who is in the position to raise questions with regard to its legitimacy. Kant is of the opinion that all cases of philosophical interest are of the latter kind. It is the sceptic who rejects claims the tradition or some philosophy imposes upon him. And he can do so without dependence upon a conceptual framework of a more modest and still consistent structure: he need only challenge the knowledge-claims of the framework in question. Theories are justified by deductive and inductive arguments and it is easy to see in what way they can be challenged. But, since basic frameworks have a different status, one might object that they cannot be challenged at all, unless one poses an alternative, and hence, competitive framework. This objection is i1lfounded. A sceptic can detect within a framework reasons or motivatiol1s, which can be expressed as reasons, for adopting certain features of the framework, and he then can point out that these reasons are not valid. In this way the alleged ultimacy of the framework and the irreducibility of its components become dubious. Such a scepticism commences with the feeling that some of the basic notions we employ in our description of the world and that seenl to be constitutive for it are obscure and possibly obsolete. When it discovers that the origins of these "indispensable" notions be in false belief, this scepticism can also be led to the conclusion that the coherence and systematic unity an established ontology claims to have is fictitious; and it might be led further to the suspicion that the very notion of a foundation for knowledge should be abandoned. Ordinary doubts with regard to all kinds of very general ClainlS originate in this way. There is no reason for expecting that the roots of philosophical scepticism should be different and that the ingenuity of the sceptic depends totally upon his capacity for producing alternatives. His primary strength lies in the identification of groundless claims. - Even Hume's attack against causality can be understood in this way. His construction of the origin of our belief in causation follows upon his criticism of the rationality of this notion, which is totally independent of any sketch of a
COMMENT ON ROR TY
11
competing framework. The attack can be read more easily as the destruction of a claim that is crucial for an existing framework within itself. Another way of objecting to such a claim depends indeed upon ontological imagination: It consists in designing an alternative that is equally plausible and powerful although nobody (for obvious reasons) is prepared to adopt it. But such an alternative can still not be regarded as a competitor. For the sceptic claims no original right for it and is not willing to support it for its own sake. It is used only as a device for showing that the challenged claim is without sufficient warrant. These are only two of the various stratagems the sceptic has at his disposal. He is unlikely to be convinced by the holism characteristic of the frameworks the Strawsonian Kantian and the Rortyan pragmatist still share. Therefore he will be inclined to challenge all frameworks from within in exactly the same way - namely by showing that they are built up through a particular process that is constitutive for and similar to ordinary reasoning, although that reasoning ~s invalid as soon as it is connected with the claim to be irresistable and to lead to a result that in all respects is consistent. Such a criticism, which is not in principle different from the internal criticism to which any theory is subject, can and must be independent of the requirement of an available alternative worthy of serious consideration because it seems to be n10re plausible and/or rigid than the challenged theory. The power of the challenge does not rest on the stiffness of the competition. A sceptic, owing to the generality of his basic point, would be very unwise to make his challenge dependent upon a claim that is equally strong as the one he challenges, regardless of whether it is in favour of a more modest or a more extensive conceptual framework. For that reason it was philosophically right that Kant was not so much concerned with the possibility that a particular conceptual scheme would tum out not to be ultimate but instead with the possibility that no knowledge claim could be justified at all, - regardless of its competitive status. The impact of the true sceptic upon him and therefore the source of his fear, was not the relativity and arbitrariness of frameworks (and consequently also of theories) but rather the groundlessness of what we take to be our knowledge. This makes very understandable his dissatisfaction with a form of justification which shows that some suggested alternative a semi-sceptic might produce in fact presupposes our ordinary conceptual scheme. Kant would have looked anyway for a possibility of making the notion of a person who can raise doubts and conceive of schemas (the 'transcendental unity of apperception') dependent upon the validity of a particular set of categories whose justification would then be in such a connection. This brings us, of course, to \
118
DIETER HENRICH
the tricky problem of how such a justification carl be carried through. Rorty, who in this regard adopts the same sceptical role as Strawson believes it cannot, although his reasons are to be found in other papers. So we should possibly not go into the issue today. But I want to say at least that one has provided justification if one can show that a certain knowledge or normative claim does have a meaningful position within the context of all our warranted beliefs and the principles on which they rest. Kant's justification of the Categorical Imperative is structured in this way. We should always keep in mind Kant's practical philosophy when we try to detennine the structure of his arguments. And we should not with ease accept a theory of or against conceptual frameworks that by definition already makes the justification of nonnative claims in a broader context to be an enterprise that is meaningless.! (5) No Transcendentality without aprioricity! I know little about the philosophy of logic and mathematics that John Dewey holds to be true. And Dewey's associate in Rorty's spirit, Heidegger, certainly has none. Rorty has long aimed at a new pragmatism. But I think that we had to learn from Quine that any solid pragmatism must be able to acconlmodate a philosophy of logic that is based upon pragmatist principles. And a pragmatist philosophy of logic has to oppose strong aprioricity. The very heart of Quine's indeterminacy thesis is that there is a sense in which logic is context-dependent. Putnam has stressed this on several occasions. His rejection of Metaphysical Realism would not have come about without a relativistic philosophy of logic. One can, to be sure, reject metaphysical realism without adopting any relativism in matters of logic but not if the main reason for the rejection is the context-dependence of all discourse. It is an open question to what extent one can envisage logic and arithmetic as being accomnl0dated by a pragmatist philosophy. There are principles that seem to be indispensable in order to make sense of the very notion of logic as such - some form of the principle of non-contradiction, and similarly nowadays also something like the successor function. It should be noted that Kant's transcendental deduction would accompllsh little without the way in which it involves logic and mathematics. It should be noted, too, that it was Kant who made the philosophy of logic interesting and important for philosophy to an extent that can be compared only to Plato and to Frege - although he personally did not have such an interest. And above all it should be clear that although he connected logic with the basic notion of 'understanding a combination', he could not possibly think of logic as being relativized or even naturalized. Is there any alternative to a Kantian philosophy of logic that believes in the (strong) aprioricity of logic or at least
COMMENT ON RO'RTY
119
some basic principles of logic but stops short of Platonism? If such an alternative is lacking and if Platonism is unappealing to us, we have a choice only between the relativity of pragmatism and some close or remote relative of Kantian views. What Rorty's choice has to be might seem to be clear. (6) But not quite so (although he makes the choice against aprioricity on p. 80). For his most powerful weapon in defense of pragmatism is Davidson's argument. Davidson, however, for the very reason that he defines meaning in terms of truth, is bound to accept a non-relativized account of logic. He lacks the most plausible devices for introducing alternative logics: the manyvalued forms of which Quine (and Putnam) are constantly making use. And he himself says so explicitly in a note to his 'Radical Interpretation'. One cannot forge iron and water together. It seems therefore that Rorty's Davidsonian transcendental strategy turns out to be not only ad hoc but also contra hominem (Ricardum). (7) In case both Platonism and Pragmatism in the philosophy of logic founder, and Kantianism becomes appealing again, it is not too hard to foresee ways in which propositions about the person as the knower might again be brought onto the philosophical stage with a prospect of receiving applause. The analysis of propositions with truth-values assigned will not be complete without referring to epistemic states (as Davidson keeps saying, if I understand hin1 correctly), and most of the arguments that are both sound and 'transcendental' appeal to self-ascription of states or even epistemic states (Shoemaker's 1968 argun1ent). But to establish some plausible connection between Logic, Epistemology and the core of the notion of a person within the domain of aprioricity is the root from which any Kantianism has to grow. (8) I admire the way in which Rorty (contrary to the received image of style among analytic philosophers) aims at a philosophy of history and civilization. But I don't understand in what way he can say that the philosopher who turned to wisdom by abandoning frameworks can still contribute to the culture qua philosopher. If there were frameworks, he could possible criticize old ones and advocate (or support) new ones - as Dewey believed, although it should be noted that Quinean indeterminacy is a matter totally different from Deweyan historicity and that Wittgenstein was probably mislead when he himself used his theory of language-gan1es in an accounting for differences between cultures and ages. But for Rorty frameworks are surely lost to science anyway. The philosophers' contribution also could not possibly be a Heideggerian negative prophecy. For this would depend on the very idea of conceptual frameworks, although on those that are anonymously received rather than imposed. So will his contribution be a Wittgensteinian therapy?
120
DIETER HENRICH
Or will philosophy happily dissolve itself,into Literature and the Arts, more specifically into the Art of Living and Loving?
Universitiit Heidelberg NOTE 1 The preceding section includes also a sketch of a position I defended in a controversy with J. Bennett during the conference. See also: 'Die Deduktion des Sittengesetzes' ['The Deduction of the Moral Law'] , in Denken im Schatten des Nihilismus (ed. by A. Schwan), Darmstadt 1975.
II
THE CONCEPTUAL FOUNDATIONS OF SCIENCE
C. F. VON WEIZSACKER
THE PRECONDITIONS OF EXPERIENCE AND THE UNITY OF PHYSICS
Table of Contents 1. Introduction
123
A. Three Preliminary Chapters
126
2. Kant· 3. Science as an Element in Human History 4. Cognition as an Element in Organic Evolution
126 129 133
B. Three Systematic Steps
138
5. Time 6. Probability and Quantum Theory 7. The Space-Time Continuum
138 143 147
C. A Philosophical Reflection
152
8. The Union of Science and the Philosophy of Science
152
1. INTRODUCTION
The table of contents shows that this is a complex paper. It is something like an abridged version of a book still to be written.! Perhaps the reader may find it more easily accessible if the author briefly introduces him into the personal, autobiographical preconditions of the views he is going to present. This is the paper of a physicist who studied and practised theoretical physics because of its relevance for philosophy, and who found himself disappointed by all philosophies which pretended to interpret physics. My first direct, personal contact with physics happened fifty years ago. I met Heisenberg in Copenhagen in December 1926, and in Aprl11927 he told me the uncertainty principle, which was still unpublished, during a tax-ride in Berlin. Theoretical physicists of that time lived under the impression of two or three fairly recent scientific revolutions: Special and General Relativity, and Quantum Theory. Our great n1asters were Einstein and Bohr. Under this impregnation I first conceived the philosophical problem of physics partly in the methodological disguise: "how are scientific revolutions possible?", and partly in its substantial specialisation: "what is the meaning of quantum theory?". 123 P. Bieri, R.-P. Horstmann, and L. KrUger (eds.), Transcendental Arguments and Science, 123-158. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
124
C. F. VON WEIZSACKER
Scientific revolutions had presented themselves to our perception as transitions from one very simple theory (e.g. classical mechanics) towards another, seemingly even simpler theory. The simplicity of successful theories was the main mystery: "why can we explain the basic assumptions of Newtonian mechanics or of Quantum theory to a mathematically trained mind on a few pages of print, while there are literally speaking billions of single empirical data which are correctly predicted by the theory?" The interpretation of quantum theory evidently required the abandonment of classical ontology and the understanding of the role of the observer for the very meaning of the terms of physics. This meant a revolution of substance, not only in physics but apparently in philosophy. The physicist who asked for a philosophical interpretation of these revolutionary events was confronted with three contending schools of philosophical thought: realism, positivism, and Kantianism. None of them offered a satisfactory answer. Realism, most definitely in its special form of materialism, seemed to offer no more than a dogmatisation of the world-view of Newtonian mechanics. It 'explained' the simplicity of classical physics, if at all, by dogmatically repeating its more general principles. Precisely therefore it was (and still is) entirely helpless in trying to understand the simplicity of quantum theory. And it was forced straightforward to deny the philosophical relevance of the role of the observer in quantum theory. This inability of realism to cope with the real progress of physics in our time I took to be a consequence of its basic philosophical weakness. It starts from an undefined and unexplained concept of reality. Realism presupposes what might well be reduced to a simpler concept and thereby limited in its applicability: the idea of things in space and time which 'exist objectively', as the tautological expression says. Positivism, then in the heyday of the Vienna School, recommended itself to the physicists by not objecting to their new theoretical ideas. It had even exerted a positive influence on the creation of these ideas. Both Einstein and Heisenberg were influenced by Mach in their early days. This influence had profoundly contributed to their analysis of the role of observation in the definition of basic concepts like simultaneity and position resp. momentum. Down to our days, the Copenhagen interpretation of quantum theory has been (mis-) interpreted as positivistic. Yet positivism, while not objecting to scientific revolutions, did not contribute to their understanding. In positivism, the simplicity of basic theories stays an unexplained empirical fact; a strict empiricist would even tend to talk us out of the sense of wonder at this fact,
THE PRECONDITIONS OF EXPERIENCE
125
and to teach us modestly to accept it like' any other fact. But as a reserch physicist, I had been taught that any strikingly simple empirical fact is a challenge for the theorist. Thus I interpreted the simplicity of basic theories as a challenge for the philosopher. The weakness of empiricism could be seen by radicalising the question. Hume's problem teaches us that cumulative experience is unable to justify not only the very simple laws found by modern physics but any universal laws. The concept of 'learning from experience' gave me the impression of being as little analysed in empiricism as the concept of 'reality' was in realism. Kantianism promised not only to ask but even to answer the two fundamental questions which were forgotten by its two opponents. It did not just accept experience but it asked what are its preconditions. And it found among these preconditions the synthesis by which we, the subjects of knowledge, arrange our sense-impression in the frame of objects in space and time. Yet Kantianism had been profoundly discredited in the eyes of physicists by ascribing a priori validity to precisely those views of earlier science which were overthrown by the scientific revolutions of our century. I could not accept the defence that these were only factual errors, understandable in the 18th-century science in which Kant had been trained. They seemed to be errors which discredited the belief that we can know, by evidence or by transcendental arguments, what statements are a priori valid; they invalidated the very concept of Apriori as Kant had put it. I remained with the impression that Kant had understood the problem of a philosophy of science more profoundly than any of our contemporaries, but that his attempted solution, as it stood, was no solution to our problem. These considerations may serve as an introduction to my paper since I shall present a philosophy of science which tries to answer precisely those questions which in my opinion were left open by the three competing philosophies; an answer which at the same time applies precisely to the state of physics produced by the scientific revolutions in the beginning of our century. According to this attempted philosophy Kant was at least right in his central idea that the fundamentallaws of physics formulate the preconditions of experience. If this is true, philosophy of science and fundamental science cannot be treated separately. Part B of this paper shows the interlinkage of philosophy, logic, and physics ensuing from this idea. Part A seems necessary in order to clarify the rather complicated relationship of this philosophy to three existing sets of ideas: first, Kant's own philosophy; second, the more recent developments in the philosophy of science; third, the innatism of some modern biologists.
C. F. VON WEIZSACKER
126
A. THREE PRELIMINARY",CHAPTERS
2. KANT This section is by far too brief for an exposition of Kant's philosophy of science. It might be read as an explanation why I shall not make a terminological use of most of the current Kantian terms. In reading modern authors I am sometimes puzzled by the question which meaning they attack to tenns like a priori and a posteriori, analytic and synthetic, or to the key-word of this conference, the tenn transcendental. These words have fairly clear meanings within the full context of Kant's philosophy, but these meanings cannot be taken over into a context different from his philosophy. Since we cannot completely evade them it will be useful briefly to remind us of their meaning in his context. 2.1. What is a pn'ori? Kant introduces us into the problem of his theoretical philosophy by stating that both mathematics and physics rest on synthetic judgments a priori. What does he mean by a priori? What does he mean by synthetic? 'A priori' is a predicate which Kant applies to all kinds of ideas and mental acts (Vorstellungen und Verstandshandlungen), 1ik~ intuitions (Anschauungen), concepts (Begriffe), judgments (Urteile). He defmes a cognition (Erkenntnis) to be a priori if it takes place (stattfindet) independently of all experience (Critique of Pure Reason, B 3). There are two connected criteria for a cognition a priori. Experience only teaches us facts but not necessities. A judgment which is thought together with its necessity (zugleich mit seiner Notwendigkeit gedacht wird) and which is thought in strict universality such a judgment is a judgment a priori (B 3). Kant explicitly denies that ideas a priori are innate (Streitschrlft gegen Eberhard, Ak. Ausg. VIII, p. 221). They are made (gemacht), but our ability to n1ake them is innate. Ideas a priori are not earlier in time than ideas a posteriori, but they are earlier in validity. Philosophical reflection shows them to be preconditions for the truth of judgments a posteriori. How much of this view can we take over into our own philosophy? I first accept the statement that philosophy begins with a reflection on cognitive acts which we were able to perfonn long before we started to do philosophy. This implies that philosophical terms must be introduced by reference to prephilosophical knowledge; hence they will be anlbiguollS from the outset. In order to make them less ambiguous, it is useful, as we do just now, to
THE PRECONDITIONS OF EXPERIENCE
127
remind ourselves of the problem which the inventor of a philosophical term intended to solve by its introduction. In Kant's case this is Hume's problem which I shall further on use in two versions. Temporal version: How can we logically deduce predictions for the future from observations made in the past? Logical version: How can we infer universal judgments from particular experience? Hume and Kant were aware of the insolubility of this problem: we cannot correctly draw anyone of these two inferences. Hume's reaction is a 'scepticism' which accepts a psychological explanation but no justification for our belief in natural laws. Kant was convinced that these laws must admit a justification. Hence he had to believe that there is cognition (true cognition, of course) which does not rely on experience. For him, ideas a priori are epistemologically necessary. But how are they possible? 2.2. What is synthetic?
According to Kant, our ideas a priori are made by ourselves. The mental act by which they are made is an example of what Kant calls synthesis. We consider his main example: synthetic judgments. For these he gives an explicit definition which is unfortunately restricted to categoric judgments. Let the categoric judgment be 'A is B'. If the predicate B is contained in the subject A ('versteckterweise': in a hidden manner), the judgment is analytic; else it is synthetic. In Kant's language B is contained in A precisely if the concept A is a subconcept of the concept B; we might be tempted to say, if the class A is contained in the class B (remark that Kant's language is not extensional!). True analytic judgments are true a priori. Judgments a posteriori, on the other hand, are synthetic: they add something to the concept A which was not 'contained in it' a priori. But Hume's problem cannot, in Kant's view, be resolved unless there are synthetic judgments a priori. The modern ways of using the term 'analytic' are, so I feel, ambiguous, and, in addition, different from Kant's own ambiguous use. Here I only mention one of the modern uses of the term since it will become relevant in the sequel (Section 5). Logicians call a proposition analytically true if it is true by its logical form alone. There is a wisespread view that this kind of analytic truth is the only truth we can know a priori. This view, however, gives me the impression of being inconsistent. Let p be a proposition; e.g. p = \;f x (x v 7 x). Whether the statement P = Df. 'p is analytically true' is true itself, might be doubted by somebody; in this case the question will be decided by a recourse to the laws of logic. If the laws of logic were not true a priori this recourse would not guarantee the a priori-truth of analytic propositions. But if the laws of logic are a priori true, but not due to their own logical form, they
128
C. F. VON WEIZsAcKER
would be the simplest example of synthetic judgments a priori. I leave this question unresolved for the moment. 2.3. Who is the Transcendental Subject? I refer to the Transcendental Deduction of the categories in the first edition of the Critique ('Deduction A', A 95-130). The deduction of a concept, in Kant's language, is the justification of its use. The deduction (similarly the exposition) of a concept a priori is called metaphysical if it exhibits the concept as being a concept a priori. It is called transcendental if it shows the concept to be a precondition of experience. Our experience - so Deduction A runs - is dispersed over time. Yet in order to be cognition at all it must be united into one consciousness (synthetic unity of apperception). This is possible because our understanding proceeds by concepts. A concept, according to Kant, is an universal idea which can be contained in many singular ideas such as single acts of experience. Thus a concept may represent many perceptions dispersed over time. The fundamental concepts, the categories, are valid in experience because they nlake the experience of a unique consciousness, of a subject, po~sible. This subject is called the transcendental (or logical) subject as distinct fronl the empirical subject, since it is defmed not as the object of self-knowledge but as the subject of all knowledge, Le. as the ultimate precondition of knowledge. The empirical subject is known as an appearance only, like all other objects of knowledge; the transcendental subject is not known at all, it is presupposed by the fact that there is knowledge, hence a knower. The transcendental subject is, so to speak, the Platonic Form of subjectivity; it is that structure of reality which makes knowledge possible. 2.4. How Can the Laws ofPhysics be Justified? I have enumerated some of Kant's conceptual tools. What does he achieve with these tools in the interpretation of science? Kant takes his start with the fact that science exists. He knows two sciences which he accepts as strictly justifiable: mathematics and physics. His problem is: how can the existence of justifiable science be explained? Mathematics, Kant contends, rests on synthetic judgments a priori. The necessary synthesis is perfonned by mental construction in pure intuition. The objects of geometry are produced by construction in space, those of arithmetic by constructing time (A 143, B 182). This theory fmds its modern counterpart in intuitionism. As to physics, the first question is why mathematics applies to experience in the form of mathematical laws of nature. Kant's answer is that the pure
THE PRECONDITIONS OF EXPERIENCE
129
intuition - space and time - in which mathematics constructs its objects is uno actu the form of all emprirical intuition, i.e. the fOffi1 in which all senseperceptions are perceived. Hence mathematics must necessarily apply to the ordering of sense-perceptions (axioms of intuition, B 292f.). There are other principles of pure understanding (Grundsiitze des reinen Verstandes) like those of the existence and conservation of substance, and of causality, which formulate the conditions under which concepts can apply to the sensations dispersed in time such as to unite them in one consciousness. If these principles were already the laws of physics, Kant might consider his problem as solved, since they fOffi1ulate nothing but the conditions under which experience, that is the conceptual ordering of perceptions in time, is possible at all. Hume's problem would then be solved not in an absolute but in a relative sense. It would not be proved that experience must be possible (nobody can prove that) but it would be shown that precisely these laws must hold, if experience is to be possible. However, there is a long way to go from the principles of pure understanding to the special laws of physics, a way which Kant, I think unsuccessfully, tried to pave in his last twenty years. For us, there are two objections to Kant's theory. The first one I have already mentioned in the first section of this paper. Kant stated several views as a priori certain which were abondoned by later mathematics and physics; this fact casts a fundamental doubt on the idea of a priori knowledge. The second objection is more subtle, but it will be fundan1ental for my own analysis. The way is long from the Principles of Pure Understanding to the special laws of physics. Even if we accept the Principles, they will not solve Hume's problem for the special1aws. The justification of special laws by experience is as impossible in Kant's philosophy as it was in Hume:ls. A1llaws that can be justified by Kant's approach must themselves be preconditions of experience. Hence I shall not draw on Kant's philosophy in this paper except as a model which presents us with a view on the magnitude and the inner coherence of our problems. 3. SCIENCE AS AN ELEMENT IN HUMAN HISTORY
In this section I mention those steps in the philosophy of science as developed in the last five decades of which I shall make argumentative use in my own presentation. 3.1. Falsification Modem philosophy of science has taken its starting point within the frame of
130
C. F. VON WEIZsAcKER
empIrIcism. In one form or another it believed that there are single data of experience, and that natural laws are universal propositions about such data. The most famous of those authors who were aware of the problems into which this view must necessarily run, is Popper. He emphasized the logical version of Hume's problem. Universal propositions cannot be verified by collecting all the single data to which they refer. Popper's answer to the problem is that they can be falsified by single counter-instances. Hence natural laws can be proposed as hypotheses, and used as long as they are not empirically falsified. Already in his early writings Popper compared this struggle for survival of the scientific hypotheses with natural selection in the sense of Darwin. In a recent paper he stresses the point. 2 Gene-mutations, patterns of behaviour, and scientific hypotheses are not produced by 'instruction from without'. They are produced by internal processes in the learning system - in an individual, in a social group, in the scientific community -, they are inherited by 'instruction from within the system' and they survive according to their success. This view is a good first approximation to a description of the role of science as an element in human history; I an1 going to use it in Section 4. Taken as stated, Popper's view would, however, be insufficient for solving the philosophical puzzle of the possibility of empirical science. Within science we can say that mutations and patterns of behaviour succeed if they are adpted to the conditions of the real world; here we describe both, the mutations and behavioural patterns on the one side, and the real world in which they succeed on the other side, in the same terms provided by the present status of natural science. We can test the adaptation of behaviour to the facts (adaequatio actionis ad rem) because existing science offers a common conceptual frame for describing both sides of the comparison. But it would be a petitio principii if we were to apply the same argument to the success of science itself; if we were to compare the hypotheses of science with 'the real world itself' for which we have no description except precisely by the hypotheses of science. It is this petitio principii to which Popper sometime refers as to his "robust realism". The 'real world' to which this realism refers is not a concept of a scientific theory. This philosophy of science cannot be scientific itself (a fact of which Popper is aware), but I maintain that it is not even a philosophical explanation for the success of science; it is an unexplained faith. Is 'real' itself a meaningful concept? The methodological problem becomes more evident when we understand that even the empirical falsification of an hypothesis is possible only if we can apply concepts which in their turn are meaningful only if certain othe~__
THE PRECONDITIONS OF EXPERIENCE
131
hypotheses are true. Popper is aware of thi,s fact, too. But this fact greatly diminishes the difference in validity between verification and falsification. It permits us to restate our philosophical problem by asking which set of hypotheses must be true if or as far as the term 'real world' is to be meaningful. 3.2. Paradigms
Suppose for the sake of argument that science is empirical, and that philosophy of science is to be scientific itself. A central concept of the philosophy of science is experience. It follows from these premisses that philosophy of science must find out empirically what experience means. This is what Kuhn has set out to do. The available experience about science is in the history of science. A scientific empiristic philosophy of science must be a set of hypotheses on the history of science. Using excellent historical evidence, Kuhn distinguishes normal science under a successful paradigm from scientific revolutions which replace one paradigm by another. This is Darwinism again, but the competing genes or species are here no longer replaced by hypotheses, that is by propositions, but by paradigms, that is by patterns of behaviour. Kuhn is fully aware that his description does not solve the basic philosophical problem why paradigms are at all successful. 3 And he rightly asks the question what the concept of truth can mean if applied to a successful paradigm. This question will either lead to the 'gnoseomorphy' of behaviour (see Section 4 of this paper) or back to the concept of theory. 3.3. Closed Theories
Fifteen years before Kuhn, Heisenberg,4 with far less historical material but with the experience of a great creative scientist, described the phenomenon, called a revolution by Kuhn, as the transition from one closed theory (abgeschlossene Theorie) to another. A theory is closed not by being final (Heisenberg did not believe in final theories) but by closing itself into the form of a simple, consistent mathematical formalism. Closure is Heisenberg's more detailed description of what I called simplicity in the Introduction. Any successful closed theory has a field of (approximate) empirical applicability. Its successor will be a theory with a broader field of applicability. Only the successor to a closed theory can precisely describe the degree of precision with which its predecessor could be applied to the empirical facts. The earlier theory turns out to be a special case or a limiting case of the later one. I have tried to indicate a connection between Heisenberg's and Kuhn's descriptions. s A mathematical formalism is not yet a part of physics unless its
132
C. F. VON WEIZSACKER
terms are empirically interpreted. This phy§Jcal semantic to a new theory can in the beginning only be given in terms of its predecessor which is already interpreted, and originally in everyday language. This interpretation is not fully theory-guided; it is done as a mode of behaviour under certain paradigms. This duality of prerequistes for any empirical theory, viz. a mathematical formalism on the one hand, a pattern of behavior on the other, has been investigated with the full arsenal of modern logic and epistemology by Sneed 6 and Stegmtiller. 7 Unfortunately I have not studied these works in any detail, due to other preoccupations but I was struck by the similarity of their original problem with the problem of this subsection.
3.4. Semantical Consistency The existence of at least two essential ingredients in any theory of physics opens up a horizon of new philosophical questions. A theory in physics contains a mathematical structure on the one hand, semantic-giving patterns of behaviour and language on the other. Since both ingredients are essential, their relationship cannot be fortuitous; we must try to explain their necessity, and the consequences of their symbiosis. The first arising questions will lead us into two apparently opposite directions. We may try to give anyone of the two ingredients a genetic or systematic priority, and to explain the other one as its necessary complement. Under the view-point of a genetic priority' of behavioural patterns we will look at the problem in the ensuing section. In the present subsection, however, we shall tentatively stick to the historically established prejudice in favour of theory; we shall accept the mathematical structure as the 'hard core' of a theory and ask on the role of its physical semantic. Relativity and Quantum Theory have treated this problem under the title of the theory of measurement. Within which linguistic or theoretical frame are the concepts defined which we use in describing a n1easurement by which a quantity occurring in the theory is assigned its empirical value in a given experimental situation? If the frame of everyday language were sufficient, there would seem to be no philosophical problem in the theory of measurement. But this is evidently not the case. Everyday language is full of ambiguities. We need the theory in order to give our concepts sufficient precision. The full awareness of this problem is reached in Einstein's dictum: "Only the theory decides what can be observed" (Erst die Theorie entscheidet, was man beobachten kann).8 But which theory in a sequence of successive 'closed' theories will decide what can be measured in one of them? In the historical
THE PRECONDITIONS OF EXPERIENCE
133
development the supplier of the concepts for discussing the problem of measuren1ent in a new theory which is just in process of formation, will be its predecessor. Thus one tried to define the empirical meaning of concepts in the electrodynamics of moving bodies as well as in the quantum theory of the aton1 by applying classical methanics and electrodynamics to the measuring process. But it is Einstein's point that this is not suffieicnt. Only Special Relativity in the first, Quantum Mechanics in the second case could decide what was actually measurable. This is how Einstein's concept of relative simultaneity and Heisenberg's principle of indeterminacy were established. No philosophy of science can be adequate to modern physics which would not be able to interpret the absolutely central role of this structure in the theories of physics. One may suspect a vicious circle implied in Einstein's dictum. The theory is given its empirical meaning by the prescriptions for the measurement of its quantities; how can it define itself the meaning of these prescriptions? Bohr's remark that in quantum theory every measurement must be described in classical terms, is a contribution to the analysis of this very intricate problem. In fact, the attribution of meaning to the concepts of a new theory is a historical process which has never so far become fully transparent. But Einstein's dictum can at least be interpreted as a regulative postulate. I propose to call it the postulate of semantical consistency: The description of the measurement of a quantity occurring in any physical theory should not be in contradiction with the application of this theory to the measuring process. In this version the principle appears trivial and hence harmless. Its acuity lies in the double fact that (a) concepts of a theory are not adequately defined without a theory of measurement, (b) they are historically introduced before the theory is sufficiently developed for admitting a test of its semantical consistency. I doubt whether anyone of the 'closed' theories of physics that have so far appeared would stand a thorough scrutiny of its semantical consistency. Tlus doubt is one of the main motives in the systen1atic attempt presented in part B. 4. COGNITION AS AN ELEMENT IN ORGANIC EVOLUTION 9
4.1. The Gnoseomorphy ofEvolution Science is an element in human history. Human history is a late product of organic evolution. Evolution itself is gnoseon1orpluc. It has a form or structure in common with cognition. Translating evolution into Greek as physis (which literally means growth) we may as well say that cognition is physio-
134
C. F. VON WEIZsAcKER
morphic. We try to learn something on the structure of both cognition an evolution by comparing them. In this subsection I interpret the ideas presented by Popper in the passage quoted in subsection 3.1, and the organic theory of cognition of K. Lorenz, as given in his book Die Ruekseite des Spiegels. 10 What is the common structure of cognition and evolution? Cognition accumulates information; so does evolution. A cognitive act is true, according to the classical definition, if it is adequate to the fact (adaequatio intel/eetus ad rem). An evolutive step is successful if it is adequate to the ecological niche. Just as the cognitive act is said to mirror the fact to which it is adequate, the species mirrors its ecological niche. Lorenz compares not only evolution but all of life with cognition. In this sense I described in 3.1 the adaptation of any behaviour as an adequation of the action to the fact, thereby implying a pragmatic definition of truth which would apply to cognition because it applies to any behaviour. For the present paper it is not the biological fruitfulness of this compari· son that n1atters, but the philosophical problem implied in it. First: What is the concept of information in which we can say that both cognition and evolution accumulate information? Information is originally defined as information for human subjects. Shannon defined information as the entropy of a telegraphic message: the expectation value of the number of binary alternatives that will be decided by receiving the message. Take the printed telegram. Its (syntactic!) information content is defined only under the presupposition of an educated reader of latin letters who is able to recognize that two letters 'e' in different places and consisting of slightly differently shaped clusters of black color are the san1e letter. A fly crawling over the sheet of paper would decide differently on its information content. The modern concept of information tends, however, towards objectivation. It is the means of this process of objectivation that deserve our philosophical interest. First, information does not refer to a lonely subject, but to communicating subjects; the information content of a message is measured such as to be identical for all subjects participating in the con1n1unication. Second, the human partners can, for the purpose of exchanging n1easurable information, be replaced by man-made sending and receiving instruments. Third, we think of the organs of living organisms which are not even man-made, as sending and receiving instruments as well. Thus we speak of information for an instrument or an organ. This is strict· ly speaking the information which is ascribed to the instrument by a human
THE PRECONDITIONS OF EXPERIENCE
135
judge who understands the functioning of the instrument or organ. I express this by saying, information is only defined under a concept. The same chromosome contains a different amount of information for the biologist under the concept of its being a genetic message and for the chemist under the concept of its being some assembly of marcomolecules. When we have defined what we will call the function of an organ, we will be able to say how much information a certain message contains under the objective concept of this function. Hence the philosophical problenl is what we mean objectively by a concept, e.g. the concept of a function. This question is still ambiguous. The line of traditional Kantianism would be to understand objectivation as an act of the thinking subject, and the concept as the tool of objectivation; the function then is thought of or represented by means of the concept. But Kant himself describes concepts as functions of thought. Then the concept is a function. This idea leads us into the biological or behavioural preliminaries of logic. Only in pursuing both paths will we understand the problem. 4.2. Biological Preliminaries to Logic It was a logical error of empiricism to think of a singular judgment as sinlpler than a universal judgment, to consider the single case as given and the general proposition as the result of a generalisation out of or beyond what is given. Correspondingly it was a logical error to think of proper names or singular terms as simple and of predicates as something to be defined with respect to singulars, e.g. as classes of individuals. In a series of analytic steps these errors have been dismounted, beginning with Frege's theory of the concept as a function (in the mathematical sense of 'function') and Russell's theory of definite descriptions, and proceeding through the development of linguistic philosophy.ll I shall not delve into these problems, but I wish to emphasize here that the behavioural background of a linguistic or pragmatic approach to logic points in the same direction. The more primitive an animal, the larger is the contribution of inborn patterns in its behaviour, the smaller the contribution of individual learning. This is exactly what a cybernetic model of animal behaviour would teach us to expect. A sin1ple computer can only have 'inborn' behaviour, Le. fixed programs; a new program must be written for the computer, not by the computer. A program for learning from experience, Le. for program change by the computer as a reaction of the computer to outside reactions, needs a fairly large memory. And even in higher animals where the contribution of individuallearning is considerable, the ability to learn is still inborn, just as the
136
C.F. VON WEIZSACKER
computer. This is so even in man; our language is learnt, but our ability to learn a language is inborn. Hence the simplest analysis of behaviour will refer to inborn behaviour. The simplest behavioural pattern is a motion (or a series of coordinated motions) which is triggered by an outside influence. Let us call this an atomic action. In describing it we may already borrow certain terms from logic. The action is two-valued; a yes-no-mechanism decides whether it is started or not. It reacts to a situation which we can characterize by a predicate or a concept. Thus it is the behavioural forerunner to a predicative ('categoric') judgment. But it is not comparable to a full categoric judgment, since it contains only a predicate, not a subject ( a 'singular term'). In this brief analysis we discover a number of logically relevant facts. The 'judgment' is gentically prior to the 'concept'; in fact, the original 'concept' is a 'judgment'. The 'concept' or 'predicate' is genetically prior to the proper name. The universal is genetically prior to the singular, precisely because it is simpler. The full structure of the Categoric judgment is probably only reached on the level of human language, in the form of the predicative sentence. The predicative sentence is independent of the situation in which it is spoken. This it achieves by attributing the predicate not to the present situation but to the singular term. 12 The singular is an invention made for this purpose. A proper name can only be defined by a conceptual description, combined with a demonstrative reference to unique spatio-temporal events. The ability to use sentences is not yet the level of logic. Logic means reflection on such abilities, and most examples used in logic are sentences, judgments, inferences etc. which are used by persons who are able so to reflect on what they themselves are doing. Only in this level of reflection can the difference between the success ofan action and the truth of a statement be defined.
4.3 Comparison with Plato and Kant A clear perception of these structures can be found in the ideas of some non· empiricist philosophers. Their perception is, of course, direct, one might say, phenomenological. It is not induced by a theory of evolution; on the other hand, it is embedded in a metaphysic. One may hence suspect that those philosphers saw more and explained less than we can see or explain. Plato's phenomenological attitude is close to the most radical possible interpretation of the preceding subsection. In modern logic, Plato's own view was, however, never even considered as a possibility. Hence the term 'Platonisnl' was used in a sense which is as far froill Plato as North from South. In Plato's view, the only things that deserved the predicate of 'being' were the forms
THE PRECONDITIONS OF EXPERIENCE
137
(ideai). In modern philosophy, the Platonic fonn was replaced by its subjective counterpart, the concept. In Aristotle's syllogistic, both subject and predicate are forms. 13 But Aristotle introduced the concept of substances which can# only be subjects, not predicates of true categoric judgments. In this tradition modern logic assumed that individuals (corresponding to Aristotle's substances) are designated by proper names, and forms (or universals, as they then were called) by predicates. In modern logic, 'Platonism' is the name of the view that even predicates can be treated like substances; when Platonism was criticized, it was said that it interpreted predicates like substances. Plato, however, believed, that there are no individual substances at all. Forms, as the only being, are the only things that can be understood; a categoric judgn1ent is by necessity the predication of a form on a form. Individuals of which we can n1eaningfully speak are sufficiently detailed forms. It seems that, epistemologically speaking, Plato's view is better conlpatible with our behavioural preliminaries to logics than any other philosophy. But this ren1ark is so far only an aphorisnl, not a theory; especially it leaves open what would be an appropriate ontology. Kant's philosophy of the A priori has been compared with biological innatism by Lorenz. According to Lorenz. Kant has correctly seen that we know many structures of reality without having learnt them from individual experience. Lorenz criticizes Kant's view that since we did not learn these structures in individual experience we did not learn them from the things at all, and that hence they apply only to appearances, not to the things in themselves. Lorenz says that the species has learnt them objectively by adaptation, by the gnoseomorphic process of evolution. Hence, Lorenz maintains, we are fully justified in attributing them to what he calls "the real world". A Kantian will of course answer that 'a priori' is not 'innate', since 'innate' is an empirical term, 'a priori' a transcendental one. Omitting the questionable tenn 'transcendental' one might say that cog..llition a priori is by definition true, innate cognition may be - and often is - inaccurate or bluntly mistaken. These arguments prepare a nest of possible fallacies. I should try to disentangle them as follows. Kant's observation that we have an understanding of structures not learnt from individual experience seems to be correct; our spatial intuition which was so important for the survival of our ancestors, the apes in the trees, is Lorenz' most striking example, but the structures of logic may equally be mentioned. I unhesitatingly follow Lorenz' genetic explanation of this fact, and hence Lorenz' remark that Kant's argument on things in thenlselves and appearances here leaves a loophole. On the other hand I have already criticized Lorenz' (and Popper's) use of the word 'real'. His
138
C. F. VON WEIZsAcKER
argument stays within science, more precisely within the world-view of classical physics; it cannot give a nletaphysical justification to the world-voew of a historical subculture like the pre-quantum-theoretical scientific comnlunity. In the title of his book (l.c. 4.1) Lorenz calls the physiological structure of our cognitive apparatus "the back of the mirror", but he does not ernphasize that the back of the mirror is described by him in the way he sees it in his own InirrOL On the other hand I would not follow those philosophers who strictly separate a priori and innate. I refer on the one hand to Kant's own, Inore precise description (1.1), on the other hand to the inability of apriorists in establishing any positive a priori knowledge beyond doubt.
4.4 A Remark on Reductionism Some readers may feel that my use of biological arguments on cognition is ruthless reductionism. I confess that I am not afraid of reductionism, if we know what we mean by reducing one structure to another. I used the term gnoseomorphy in order to point out that our interpretation of cognition is a prerequisite of our interpretation of evolution. This is not yet a philosophy of either evolution or cognition; it is propaedeutic to such a philosophy. I only stress the unity of the phenomenon known under the two titles of evolution and of cognition. The real problem of reductionism is whether these biological structures can be reduced to physics. Again I would not object to this program. But this leads us back to our central problem: What is physics?
B. THREE SYSTEMA TIC STEPS 5. TIME
5.1. The Program The problem of an empirical justification of special laws of nature appears insoluble in all three philosophical approaches. In the ensuing systematic attempt I shall assunle that this problem is insoluble because it does not exist. It does not exist because there are no ultimate specia11aws of nature. History of science shows how special laws are introduced as hypotheses, faithfully used, and finally reduced to more fundamental and hence more general laws. Heisenberg's description of a sequence of closed theories is a good approximation to the historical process in the field of fundamental laws. To presentday physics, nearly all the special laws of physics and chenlistry are logical consequences of its present basic theory, of quantum theory. They are logical
THE PRECONDITIONS OF EXPERIENCE
139
consequences in a double sense. More"'superficially: they can (in principle, disregarding mathematical complications) be deduced from the fundamental laws as their special cases, if applied to special situations or problems. More profoundly: the basic theory can also prove (in principle) that precisely these special situations and problems themselves are mathematically possible consequences from the fundamental laws. In other words, the theory itself defines the possible cases to which it can be applied. The hypothesis of biological reductionism, or physicalism, says that these cases to which the fundamental theory applies as described, include the phenomena of organic life. If these assumptions are correct, the only problem of justification left to science is the justification of the fundamental laws of physics themselves. Hence I shall concentrate on this single task. On the other hand, the present theoretical situation in physics is far from satisfactory from the point of view of this task, in three aspects. First, obviously the problem of the justification of the fundamental laws is not solved, it is not even clearly formulated. This fonnulation is the first task in our program. Second, quantum theory itself is not satisfactorily interpreted. Personally I stick to the Copenhagen Interpretation, and I would offer a bet that when the theory will be satisfactorily interpreted, practically all Bohr's sentences on this problem written since his 1927 paper on complementarily will turn out to be correct. But the interpretation would presuppose a clear understanding of the semantical consistency of the theory (cf. subsection 3.4), an understanding which is not now available. Third, quantum theory is not yet an all-embracing fundamental theory. Its connection with Special and General Relativity and hence with Cosmology is notadequately understood. The theory of elementary particles is not yet achieved. The 'three systematic steps' of this part of my paper are loosely correlated to these three problems. It is the guiding, ambitious hypothesis of this systematic attempt that the fundamental laws of physics can indeed by fully justified as preconditions of all possible objectivating experience. The terms in which this sentence is formulated will need further commentary. Yet I begin by the remark that, methodologically, this hypothesis may well have overstated our case, without serious detriment to the enterprise. At least we shall try to understand as much as possible of those laws on the indicated line, leaving the rest to a better philosophy of the future. But I do not wish to loose the sting of an ambitious program. And my personal philosophy would at least be well pleased if this program were strictly possible; I shall resume this question in Section 8. But how to evade the two difficulties, mentioned at the end of 1.4, which invalidated Kant's own solution? The answer is implied in the way I stated
140
C. F. VON WEllSAcKER
the program, but it may be made explicit. I take the second question first. The problem of special laws does; no longer exist, if the program is well formulated. This means, on the other hand, that we can no longer be satisfied, as Kant still seemed to be in the Critique of Pure Reason, by justifying the principles of pure understanding. Like the later Kant we must rather try to justify the fundamental laws of physics. Hence the distinctive line between philosophy of science and fundamental physics must be eradicated; retaining it would be an absolute impediment to the intended progress. The problem of the unreliability of presumed a priori knowledge will be treated by two interrelated attitudes which might be called strict transcendentalism and historicism. They are essential to our procedure and deserve full explanation. Strict transcendentalism means that, using Kant's vocabulary, we will not use metaphysical but only transcendental arguments. I.e. we will never maintain that any judgment is a priori true; in this sense we will eliminate the word 'a priori' from our vocabulary. We will only argue that, given an explicitly stated concept of experience, certain other concepts or judgments seem necessary for establishing a theoretical frame within which this given concept of experience can meaningfully be used. Historicism means that we will never crave timeless validity for anyone of our statements. I wish to emphasize that I see this principle not as an expression of scepticism or of irrationalism but as an awareness of the only possible meaning of statements made in history. We do our thinking now - not in the past, not in the future, not in eternity. It is true, as Kant presupposes in Deduction A, that concepts have a time;.bridging validity. But the structure of this time-bridging can and hence must be subject to investigation. This will be done in the temporal logic which is part of our program. The investigation will be part of a feedback which itself is inevitable in a consciously historical philosophy: temporal logic, part of the project, is itself needed to clearify the meaning of the project. Every step of our analysis is consciously open to criticism by its own future continuation. S.2. Experience and Time
Our first methodological reflection thus indicates the importance of an analysis of time. But this reflection is not our systen1atic starting point. Our starting point is the search for preconditions of experience. Therefore we need at least a preliminary definition of experience. I propose the formula: Experience means to learn from the past for the future. With purpose this formula is expressed in 'trivial' everyday language. Philosophy, when it has
THE PRECONDITIONS OF EXPERIENCE
141
reached the level of reflection at which it can embark on a systematic procedure, must then start with an explicit reflection ~n the structures implied in our ability to use everyday language. The formula is further meant to imply that the test of scientific experience is in its prognostic power. Thus it explicitly states that function of experience which in the original, temporal version of Hume's problem is the core of the seeming paradox: the relevance of the past for the future. This relevance, as we know, is not a logical implication of the future by the past. We rather expect to find it more fundamental than logic, and hence to fmd it a precondition for the meaning of logic, too. But we shall proceed more slowly. Whoever understands the meaning of experience, as exposed in our formula or similarly, evidently possesses some understanding of time. He reacts understandingly to terms like 'future' and 'past'. The tenses, or modes of time, are within his horizon of awareness. My second step is the expectation that a full explication of the structures already implied in our awareness of the modes of time might be sufficient for formulating the preconditions of experience, i.e. the fundamental laws of physics. When I shall speak of 'time' in the following without further qualification I shall always mean this structure of the modes of time - present, past, future.
5.3. The Role a/Time in Fundamental Science Before embarking on the next systematic step I shall recall the role of the modes of time in interpreting statistical thermodynamics and evolution. I may be permitted to do this in a narrative, slightly autobiographical form. When I was a young theoretical physicist I tried to understand how the Second Law of Thermodynamics can be derived from reversible fundamental theories. 14 This is done by statistical consideration. Given a state of nonmaxinlal entropy of an isolated system at one moment of time, to, it is probable that its entropy will be higher at a later time. This statement expresses a prediction which will be correctly made when to will be the present time. But the sheer mathematical concept of probability does not yet define a 'direction of time', or a 'time-arrow', as physicists sometimes metaphorically say. If nothing were known on the time before to, exactly the same consideration would imply that entropy will be higher at an earlier time, too,15 Only since 'earlier time' means a time which is past when to is present this consequence is excluded because the past can be known in principle and the predictive sense of probability does not apply to known facts, only to possibilities. It is essential to see that it would be self-contradictory to apply the concept of probability for retrodiction as well as for prediction. Any moment in the past
142
C. F. VON WEIZSACKER
was once present, Then the probabilis!ic prediction was justified, that means we know by experience that the conclusion drawn from the application of
probabilities to what was then (say at t l < to) future leads to the correct statement of the Second Law from that moment on down to our present time. But applying the same probabilistic arguments for retrodiction in the same system at any later time (t 2 :t, < t 2 < to) would lead to the inverse of the Second Law (for t < t 2 ) which is not only not borne out by our experience but is in straightforward contradiction with the prediction made from t I. This means that the Second Law in its statistical interpretation presupposes the modes of time in the sense that the past is factual, the future is possible, and probability is a quantification of possibility. In a phenomenology of time these are very simple statements with a high degree of evidence. But I have always encountered great difficulties in convincing physicists that this structure is necessary for understanding the Second Law. Physicists seem to think that the existence of a present moment and the distinction of past and future are only 'subjective'. Yet I know by experience that in long arguments I can convince a sufficiently patient partner, even a physicist, that no less than this 'tense structure' of time is a precondition of the Second Law. For me, it was an important moment when I realized that the same time structure is thus presupposed in statistical mechanics and in formulating Hume's problem. This was the first place where I understood that the separation of a philosophy of science from fundamental science may be artificial. The same temporal use of probability is made in quantum theory. This can be seen in a discussion of prediction and retrodiction in the theory of measurement. I shall not enter into this problem here. It is important to see taht the irreversibility of evolution follows from precisely the same time structure as the irreversibility of thermodynamics. 1 6 In systems sufficiently far from thermodynamical equilibrium there can be a statistical tendency of growing differentiation, and the Darwinian analysis of selection is exactly an argument which makes predictive use of probability. This implies that all the evolutionary arguments of the first part of this paper did tacitly presuppose the structure of the time modes.
5.4 Temporal Logic We resume the systematic procedure by sketching the idea of a temporal logic. In a linguistic approach to logic, temporal logic would have to formalize the way in which everyday language refers to time. This amounts more or less to a formalization of the use of verbsP Mathematical logic has taken no account of this problem since mathematical propositions are considered as
THE PRECONDITIONS OF EXPERIENCE
143
timelessly valid. This leads to the paradoxical fact that mathematical language can easily express nonsensical ideas which any competent speaker of everyday language will instinctively avoid. Thus in verbally st'\ting the paradox of timereversal with respect to the Second Law, I had to use the awkward phrase that 'entropy will be higher at an earlier time'. Had I used everyday language I would have had to say that 'entropy was higher at an earlier time', and this use of the past tense would already have implied the correction of the fallacy, in making it understood that past events are facts, not possibilities. The central idea of temporal logic is that the rules of classical two-valued logic apply to statements on the past but not to statements on the future. 'Napoleon was born in 1769' is either true or false, for he was either born in 1769 or not (some historians think he was born in 1767 and falsified his birthdate in order to be officially born under French rule and hence to be . admitted to the French army). Already Aristotle (De Interpretatione, Chapter 9) doubted whether the law of the excluded middle applies' to statements on the future. Is the statement 'there will be a naval battle tomorrow' true or false today? My proposal is to define statements on the future strictly as neither true nor false but as possible, necessary, impossible etc., Le. to ascdbe to them (temporal) modalities and no truth-values at all. I shall not go into the details of such a logic here. One systematic question is, however, in1portant. Is classical timeless logic systematically prior to temporal logic or viceversa, or is there no hierarchy in this respect? I am inclined to take an operationalist view in the foundations of logic and mathematics. Operations are done in time. Hence I would expect temporal logic to be the logic in which we describe those operations which define the concepts of classical logic. This is what I meant when in connection with Hume's problem I said that the structure of time on which experience rests may be more fundamental than (classical) logic. Under this supposition, the inability of logic to justify empirical inference is not surprising: logic formalizes less than is contained in the structure of time which gives an operational sense to logic and thus, in fact, establishes logic. If we wish to use Kantian language (cf. 2.3) this would be a way of describing the sense in which the laws of logic are synthetic judgments a priori. 6. PROBABILITY AND QUANTUM THEORY
This section can only give an outline of considerations more amply described elsewhere. The en1phasis here is not on their technical details but on their philosophical me aning.
144
C. F. VON WEIZsAcKER
6.1. Classical Probability Besides the considerations in Die Einheit der Natur, Section II 5, I refer to a paper on the enlpirical meaning of probability.18 Instead of ascribing the qualitative modality of possibility to futuric sentences, with its two limiting cases of necessity and impossibility, we can try to n1easure the degree of possibility. The measurement is done by counting cases. The prediction of the relative number of cases (of a relative frequency) is what we call probability. More strictly vie can define probability as the expectation value of a relative frequency. For the argument why this is not a circle in definition I refer to my paper. Assuming that the set of possible events (or the set of sentences expressing them) forms always a Boolean lattice, Kolrnogorov's axioms can be justified. The Boolean lattice means classical logic.
6.2. Abstract Quantum Theory Here I refer the reader to two papers by Drieschner and myself. 19 The idea is to justify a set of axioms, sufficient for a full definition of abstract quantum theory, by considerations of ten1porallogic only. Abstract quantum theory here means the existence of a Hilbert space for any physical object whose metric admits the well-known probability interpretation, the composition rule for such objects and their spaces, and the idea of the change of states with time as described by unitary transformations. It excludes any ascription of speific empirical meanings to specific linear operators in this space; this problem will be treated in Section 7 of the present paper. The considerations of temporal logic consist in an interpretation of so-called quantum logic as a futuric nl0dal logic. The ideal goal which so far has not been fully achieved would be to introduce one additional axiom into the foundations of probability which would express the idea that futuric logic is not two-valued, that means not deterministic, and which would replace the Boolean lattice of futuric sentences by the correct quantum-theoretica1lattice. This lattice is a projective geometry which can be interpreted as the lattice of sub spaces of a Hilbert space. Leaving aside the technicalities I express a few philosophical ID1plications of the very idea of such a program. The universal validity of such a simple mathematical frmne as abstract quantum theory which impressed us from the outset (see the Introduction), which has stayed practically unchallenged now for a half century, and which has not received any philosophical explanation so far - this validity would now receive the simple explanation that abstract quantum theory is nothing but the most general franle known today for a theory of empirical probabil-
THE PRECONDITIONS OF EXPERIENCE
145
ities. Probably this quite natural interpretation has not been seriously considered so far since the lattice of events p,resupposed by the theory is nonclassical. One seems to have seen this as an odd empirical fact which made the theory not rationally understandable. Its structure was so far from being seriously considered by philosophers of probability that an immense effort was spent on understanding the empirical use of probability (with the wellknown battles between 'objective' and 'subjective' interpretations) in which the Boolean structure of the lattice of events was always taken for granted. One just did not notice that the nearly all-embracing modern theory of physics which would be supposed finally to justify the special laws of physics and hence probably also the special applications of probability to experience, is essentially non-classical. The growing tendency in recent decades to return to a classical interpretation of quantum phenomena (like the search for 'hidden parameters') shows at least the uneasiness produced by this self-contradictory status of present empirical theories on probability. As a physicist I feel, however, that this tendency has no chance of success, at least as long as it cannot explain the striking mathematical simplicitly of quantum theory. And as a philosopher I think the way into the future lies in not eliminating but understanding the non-classical logic of quantum theory as an expression of the structure of time in its modes. This idea would also solve another problem that has puzzled some logicians, and also some physicists with a highly developed sense of conceptual responsibility in mathematics. 2o The mathematical core of quantum mechanics, the theory of the Hilbert space, is a piece of traditional, classical mathematics. Classical mathematics is founded on classical logic. Non-Boolean lattices are a well-known mathematical structure, easily described by means of classical logic. What justification can defenders of the term 'quantum logic' offer for this use of the word 'logic'? I think the answer to this question has in principle been given by P. Mittelstaedt. 21 I put it in my terms here. It is not the logic used to build up the mathematical structure of the lattice of 'quantum statements' that matters but the logic implied in empirical semantic (see subsection 3.4. of this paper). This semantic is to be described by temporal logic since temporal logic formulates certain preconditions of a 11 experience. Temporal logic is, so to speak, the logic of action, and measurements are actions. In classical physics this fact was just not discovered, for two reasons. First, the problem of measurement was not seriously studied, and this for the basic second reason that classical physics, as we now see, is precisely that limiting case to quantum theory in which the consequences of the possibilistic structure of the future disappear.
146
C. F. VO N WEIZ SACKER
It is to be conceded that temporal logic itself, as far as it has been fonnulated mathematically, is a theory which has a structure analysable in terms of classical logic. Here is the problem of the self-application of logic. Will the propositions in which we formulate logic obey the laws abo u t propositions which they themselves express? Temporal logic expresses laws abo u t temporal, especially about futudc statements. Since it is a theory it consists 0 f propositions which we take to be timeless. Hence so far there is no need to apply its laws to itself. Yet there are two questions left upon by this remark. There can be futuric statements abo u t futudc statements. 22 And the classical functions of mathematics itself can be questioned by operationalism, as mentioned in subsection 5.4.; a question I am not pursuing here.
6.3. Objctivation I add a remark on the definition I gave in subsection 5.1. of the aim of the present systematic attempt. I said that we try to justify the fundamenta11aws of physics as preconditions of all 0 b j e c t i vat in g experience. The terms 'objectivating', 'objectivation' are so far undefined. In traditional transcendentalism they permit two different explanations. They can either mean that we express our experience as experience of objects, or that we produce objective experience. If objective experience is not in its tum to be defined as experience of objects it would mean non-subjective experience in the sense of being true, independent of the subject that gained it. One may either think that objective truth can only be achieved by referring it to objects (first explanation) or that it can be established independently of assumptions on objects (second explanation). My present way goes by the second explanation and poses the correctness of the first explanation as a problem. Technically speaking, I define quantumtheoretical observables as dec ida b I e a I t ern at i v e s. The simplest case is a binary alternative, a yes-no-decision. But I adn1it 'alternatives' as decidable questions with any number of possible answers. It is only required that the answers are well-defined and that agreement among the observers can be reached on the meaning of the question and of its possible answers, and on the answer which has been found when a measurement has been made. This is a merely fonnal definition; the question of what physical means we need to fulfil its conditions is part of the question of the semantical consistency of \the theory we try to build up. Only then will the validity of the first explana-
[ion be open to a test.
_ - - - - __
THE PRECONDITIONS OF EXPERIENCE
147
7. THE SPACE-TIME CONTINUUM
7.1. The Problem Even a presumed fully justification of abstract quantum theory as the general theory of an empirical use of probability would not seem to mean a justification of the fundamental laws of physics. The specific laws of physics - general, not special, but of physics, not of probability alone - would still seem to be missing. According to our present knowledge of physics, these laws might be arranged in two main groups: laws on the space-time frame, and laws defining the actually existing types of objects and forces (particles, Le. fields). The laws on the space-time frame may be subdivided into three groups: (a)
(b) (c)
the laws of physical grometry defining the three-dimensional real Euclidean space in which non-relativistic classical and quantU111 motion can be described by adding a linear time coordinate, the laws of Special Relativity, merging the space-and time-coordinates under the symmetry defined by the Poincare group, the laws of General Relativity, defining a Reinlannian four-space of which (a) and (b) are local limiting cases, explaining gravitation by the metric tensor of this geometry, and offering a frame for cosmological hypotheses.
The laws on types of objects and forces may be introduced in two steps: (d)
(e)
reduction of all existing objects of physics (and, certainly of chemistry, probably of astronomy, hopefully of biology) to a series of elementary particles with well-defined laws 9f interaction, Deduction of the elementary objects defined under (d) from a symmetry group, to which the possible states of the elementary objects are related as representation spaces.
The two lists, (a)-(c), and (d)-(e) show the common structure already preliminarily described under the title of a sequence of 'closed theories'. In both lists, the later steps will supersede the earlier ones, reducing them to the status of applications in special ('limiting') cases. Step (e) especially. gives a hope of even merging the two sequences. In (e), elementary objects of physics are mathematically defined as representation spaces of a universal symmetry group. Since Felix Klein we know that a geometry can be defmed by a symmetry group. Thus we may hope to deduce physical grometry from the symmetry group as well as the objects moving in the space of that geometry.
148
C. F. VON WEIZSACKER
One would tend to think of the physical "space-time continuum as a homogeneous space of the basic group. In fact, the choice of the group to be used in elementary particle physics may still be subject to debate, but nobody doubts that it will have to contain he Poincare group as a subgroup. This would at least merge (b) into (e). Even assun1ing that these programs will successfully be carried out, we would be left with an apparently arbitrary choice of a fundamental group which is decided by its empirical success, not by being understood as a precondition of experience. Thus we would then still have to ask the two questions:
(f) (g)
why should there be a fundamental symmetry group to physics at all? how is the choice of the actual symmetry group determined?
7.2. The Quantum Theory ofSingle Alternatives I propose an attempt at a full answer to the two questions why there should be a symmetry group and which is this group. 23 The basic idea is that this group is already implied in a meaningful ('semantically consistent') use of abstract quantum theory. Why symmetry at all? The real world is not symmetrical. If you turn this sheet of paper upside down you will not longer be able to read my text. But we assume the laws of nature to be symmetrical. A differential equation may admit a symmetry group. This does not mean that its solutions, taken separately, have the same symmetry . It n1eans that the symmetry operations transform solutions into solutions; hence that the solutions are a representation space of the group. But why should even thefundamentallaws be symmetrical? We use geometrical symmetry as the best known example. The separation of geometry from physics rests on a property of the forces actually prevailing in nature. The law of force under which a system of bodies moves is symmetrical under the Euclidean group in the approximation in which the system can be considered as dynamically independent from the rest of the world. Simply speaking: the Euclidean group is the dynamical group of free systems. This is in fact what we mean by speaking of a free body or system. The concept of objectivation in the first sense of subsection 6.3. seems to rest on the existence of such symmetries. I am not going into the intricate details of this type of considerations, since I want now to discuss quantum theory and not classical grometry. I just mention the fact that since actually a free system has not only the Euclidean group but the GaIileo- or Poincare-group as its
THE PRECONDITIONS OF EXPERIENCE
149
symmetry group the separation of geometry f:t=;om physics is not as strict as it was considered in the Greek tradition. Abstract quantum theory already possesses a symmetry group: the unitary group of its Hilbert space. But this is not what we mean by the symmetry of the laws of physics. These laws define a Hamiltonian in quantum theory, Le. a linear, self-adjoint operator in Hilbert space which is taken as the generator of the one-dimensional Lie group of the time translations, respecting the actual law of force. It is the synunetry of the equation of motion containing this operator, which we mean. The proposed attempt aims at deducing this symmetry, Le. this operator, from one additional postulate on quantum theory, and to justify this postulate by considerations of semantical consistency. The postulate says that all physical observables are composed of observables corresponding to a binary alternative. I call this alternative in German die Ur-Alternative (the original alternative), and I call the object defined by precisely one Ur-Alternative an UL The hypothesis then maintains the composition of all we can observe from urs. Semantical consistency would first seem to crave some finitism in quantum theory. In no observation can we actually distinguish more than a fmite number of different possible results. In classical logic any fmite alternative can be subdivided into a sequence of binary alternatives. We assume the same to hold in abstract quantum theory. This finitism was actually a technical means used by Drieschner (1.c.) in his axiomatic of quantum theory. But we do not need to stick to strict finitism. A denumerably infinite alternative can be described by a law of successive applications of an infinite number of binary alternatives. To be semantically consistent one would have to think of an infinite time which would elapse during these measurements: all concepts of infinity in physics or in operational mathematics indicate the open future. The actual proposal of my theory is this: Let any isolated object of physics be composed of a finite or infmite number of urs. Then their interaction should be invariant under a transformation of the state-spaces of its components in which every one of its urs is transformed by the same element of the fundamental symmetry group of the binary alternative. Let k be the running number by which we count the urs, let U' be one particular transformation of the vectorspace ('Hilbert space') of the binary alternative, and let Uk be its application to the k'th UI, then the law of motion of the whole system should be invariant under the transformation
Utot31 = U~ X U; X ... X Uk X
(1)
150
C. F. VON WEIZSACKER
7.3. First Consequences The symmetry group of the quantum theory of the binary alternative is Q = U(2) XC, where C means complex conjugation. Let us first concentrate on the subgroup SU(2). Up to an indefinite sign this group is isomorphic to the Euclidean three-dimensional real rotation group SO(3). Our dynamical hypothesis implies that the dynamics of any isolated system consisting of urs, i.e. of any isolated physical system, is invariant under the Euclidean rotation group. Hence, it must admit a representation by functions in some isotopic three-dimensional real space. We take this to be the reason why physics can be described in such a space. 24 The full group Q cannot be faithfully represented by linear operators in a two-dimensional complex vector-space. The simplest linear representation is in a four-dimensional space with (++ ---) -metric. In such a space we can as well represent the non-compact group SU(2,2) which is, up to sign, isomorphic to SO(4,2). This group, in turn, is well-known as the conformal group of special relativity. It contains the Poincare group as a subgroup. It is a very tempting idea, in this fOffi1 proposed by Castell (l.c.), that this is the reason why physics admits the symmetries provided by Special Relativity. 2S It is evident that we are now led into a complex field of mathematical investigations which cannot be described in this paper. We have not yet reached the point where we could compare our results with any details of elementary particle physics. But I may say that the interest in this line of research has internationally been growing during the last half decade.
7.4. A Possible Unification ofPhysics It may be philosophically meaningfully and is even heuristically required to ask what would be the consequences of a successful termination of the pre· sent enterprise. Such a termination would probably mean that the full sequence of seven questions (a) ... (g) in (7.1.) would be answered. (a) Physical space would be deduced as a homogeneous space of a quantum-theoretical symmetry group. This would be a typical example of the progress of theoretical physics. An element of experience which is so universal that its understanding was considered a priori or innate - and, within the scope of defensible meaning of these terms, rightly so - is reduced to some other, more abstract and more fundamental elements of our understanding. These elements - temporal logic in our case - are themselves found in a search for the most general preconditions of all experience of which we can
THE PRECONDITIONS OF EXPERIENCE
151
think in our present state of scientific consciousness. They themselves are, in a sense which is simultaneously strictly transcendental and merely historical, a priori. What we achieve is a higher unification of our understanding of nature as available in the later half of the 20th century of the Christian era. It is a new 'closed' theory. I add the technical remark that homogeneous spaces of Lie-groups in physics are parameter spaces. The existence of a position operator is not required. Continuous space in a set-theoretical sense is not introduced as an observable reality but as a mathematical tool for simply describing essentially fmite alternatives. Continuity is mathematically coupled with the open future by the mediating concept of possibility (potential infinitude). We treat possibility and probability, however, as concepts of physics from the outset, hence under the axioms of quantum theory. The quantum theory of probabilities is considered as more fundamental than the classical probability calculus. (b) Special Relativity would be similarly deduced. Measurable time is systematically introduced like measurable space, by the parametrisation of a Lie group. Nowhere do we need the assumption that time as a real coordinate can actually be measured. The modes of time are by several steps more fundamental than coordinate-time which is no more than a mathematical tool. (c) In its systematic approach this theory reaches the problem of cosmology earlier than the problem of a locally varying space-time curvature. There is a possible homogeneous space of the fundamental group which is spatially closed (e.g. a de Sitter space). The fmitude of all observables that are semantically consistently used would indicate a representation in such a space. A single binary alternative admits still of no localisation in this space. A wave parcel localized down to 10- 13 cm in a cosmic space of 10 27 cm radius would need a superposition of 1040 'urs'. The conceptual separation of elementary particle physics from cosmology needs the decision of many alternatives. Locally varying curvatures which lead to the establishment of gravitation as a local classical field will only be defmed by considerations of semantical consistency in local space-time measurements. (d) and (e) The procedure aims at a unified theory of elementary particles. The proof of the pudding would be in deducing the right symmetry group from the quantum theory of alternatives. (f) The tendency isto see the symmetry of laws as a precondition of objectivation in the first sense of (6.3.), that is of the separability of objects. This would be done in two steps: First the separability of decidable alternatives. To decide a fmite alternative means to separate it from all other possible alternatives. This is objectivation in the second sense. The theory will secondly
152
C. F. VON WEIZSACKER
have to show that its assumption lead to the existence of stable bodies in a classical limiting case. The existence of this limiting case is probably a precondition of identifiable experimental alternatives (Bohr). This circular consideration would be an example of a test on semantical consistency. Vrs are ultimate bits of infoffi1ation. What we are studying might be called the preconditions of the concept of information (see Section 4). (g) The reason for choosing the particular symmetry group lies, in this presentation, hidden in the structure of quantum theory. The ultimate understanding of the program of this section leads thus back into the program of Section 6 which in its turn tries to expand the consequences of the modal structure of time. If I am not mistaken, the modes of time, present, future, past, fact and possibility, are the clue to physics.
C. A PHILOSOPHICAL REFLECTION 8. THE UNION OF SCIENCE AND THE PHILOSOPHY OF SCIENCE
In these final remarks I shall no more argue for my philosophical views but fro m them. What can they contribute to the question of this symposium on the validity of transcendental arguments? An argument uses concepts. Concepts have a precise meaning only within a theoretical frame. For transcendental arguments this frame must be a philosophical theory. Hence transcendental arguments are not per se conclusive or inconclusive. They may be conclusive in the frame of Kant's philosophy, inconclusive in empiricism. In the philosophical frame which I propose they indicate historical steps towards greater unity. This unity is not a n1ethodological unity of different sciences but a substantial unity of science. The proposed philosophy is at the same time transcendental and historical. It is historical because it is transcendental. It is transcendental in the sense of asking for an understanding of preconditions. What is already implicitly presupposed? Presupposed by what? It is nearly irrelevant how we answer this last question in the first place; wherever we start, we will be lead to the same sources. Let us say: presupposed by science, by the historical fact that there is science. Science presupposes experience. What is presupposed by experience? At least present, past, future: at least historical time. Logic and mathematics? I think, operation, that is action. What is presupposed by action? At least time. We study the structure of time. Structure as such is usually considered as timeless. We study the structure by which time is characterised as time, which
THE PRECONDITIONS OF EXPERIENCE
153
., hence must be present at any time. But this structure is discovered in time. It was not always known, and the emphasis in its description changes through history. Where we speak of time, Greek philosophy would rather have spoken of motion (kinesis). We do our thinking now. Hence it was methodologically consistent to begin this paper by an historical description of the physics of our century. Against this background the paper proceeded in essentially two steps: an analysis of time, and a systematic scheme of the union of science. The analysis of time rest on the two concepts of fact and possibility. In slightly metaphorical terms, fact is the presence of the past, possibility is the presence of the future, and presence is the unity of time. 26 Facts and possibilities refer to each other. Facts are described by concepts, concepts are universal, Le. they are possibilities. Actual possibilities ('potentialities') as distinct from mere abstract possibilities are founded in the present facts (in evolution, e.g., apes become only possible when there are monkeys to be their ancestors; tree-climbing monkeys are only possible when there are trees to climb in). The number of facts increases, hence increases the number of possibilities. Now possibilities open up new concepts; hence they cast a new light on the present facts. This analysis of time makes use of the facts presented by the history of science and of philosophy. Our present analysis will be seen in a different light in the future - that much we may dare to predict. Certainty is not the same thing as truth. I am not going to give a formal definition of truth. At least the truth or falsehood of a sentence presupposes that the sentence is understood. Understanding depends on an awareness of the present context of the common life in which the speakers and hearers of the sentence communicate. The idea that a single proposition can per se be certain, expresses an ideal limiting case. It is apt to divert our attention from the awareness of the context. For philosophy of science today the substantial union of science - partly achieved, partly to be expected - is the relevant context. In the light of the idea of a union of science I resume the criticism of the three philosophies of realism, positivism, and transcendentalism, as given in the Introduction. I shall try to do justice to their points of strength as well as to their shortcomings. Realism has the strength of not falling a victim to the transformation of philosophy into methodology and to the resulting strict separation of the philosophy of science from fundamental science. Realism sees that any method is dependent on the structure of the truth it is supposed to fmd. But realism has not been able further philosophically to reflect on this structure.
154
C. F. VON WEIZSACKER .,
It accepted it as offered by a state of science which in 1927 already belonged to the past. Positivism has the strength in its insistence on experience as a criterion of truth. It had its weakness in not thinking empirically on experience. A Kuhnian view on the history of science and a Darwinian view on the history of gnoseomorphic behaviour make us see the historical preconditions of individual experience. Transcendentalism has the strength of directly approaching the preconditions of experience. It had the weakness of not seeing their historical nature, and precisely thereby of not getting to their heart, the structure of time. Kant himself had come closer to this central problem than anyone of his successors before Heidegger. All three philosophies are deluded by a belief in the certainty of the single elements on which they trust - specific ideas on reality, single acts of experience, single judgments a priori. None of them has seen the full contextdependence of their own concepts. Precisely thereby they miss the chance of even seeing the problem posed by the real progress of science towards a substantial unity. But what is the substance of this approaching unity? First: It is just a new closed theory, one step in a virtually infinite series of such theories? It seems difficult to imagine that there might be an infinite possible sequence of theories of increasing generality. On the other hand it seems equally difficult to imagine that there migh t be an ultimate theory. Methodology tends to evade this problem by thinking of an open field of qualitatively different kinds of experience. But this is just an evasion of hard arguments. What do we mean by saying that there will be kinds precisely of experience? What is presupposed in this idea of experience? Our present theories of physics are already far beyond a division according to the kinds of experience as given by the senses (optics, acoustics, etc.). If the considerations of this paper are of any value, quantum theory is already describing all kinds of objectivating experience in time. This description may not be the final one. But we should not expect that its level of generality will be lost again by future theories. I am inclined to think that 'theory' itself is a paradigm of which there may well be an ultimate sample. What does the idea of a theory imply? This question will not be answered without a theory on the meaning of mathematics. Here I only use traditional mathematics as a comparison. Is Euclidean geometry a closed theory? It certainly is; an adequate set of axioms of this geometry contains 'in principle' the answers to all possible questions in
THE
PRECONDITIOJ~S OF
EXPERIENCE
155
this field. But Euclidian geometry was axiomatically formulated by early Greek mathematicians, and it has been a flourishing discipline through many centuries thereafter. A more general exan1ple: practically all of modern mathematics can be ultimately reduced to statements on sets of numbers. The content of mathematics is not in this raw material but in the apparently infinite number of concepts that can be defined above this basic structure. SiInilarly, a final set of laws of physics would mean very little about the possible discoveries in future science. I could imagine that the meaningful uses of such concepts as 'concept', 'theory', 'objectivation', and 'physics' are co-extensive, and that a later stage of understanding will be able to describe the limitations imposed on our thinking by this paradigm. But these are rather apologetic remarks, trying to shield off the common prejudices of philosophical debates. The fascinating question is positive: how can the general paradigm of physics give hope to explain structures as specific as the three-dimensionality of space, even as the very existence of a common fran1e to all objective events like space, or the spectrum of elementary particles? I emphasize that this hope is in the consequence of mo dern physics, independent of my special proposals on these points. This question leads us back into the positive content of mathematics. The conceptual clues to the answer lay hidden in such technical-sounding concepts like lattice (or propositions), projective geometry, vector-space, number field, complex number, Lie-group. The positive content of fundamental physics is mathematics. But the choice between the infinitude of possible mathematical structures is made in physics under the constant guidance of experience, and the attempt to interpret the results so found is in this paper made by a transcendental argument. Thus we return to the question what a transcendental argument means. A transcendental argument is an argument on knowledge. Knowledge is knowledge 0 f some body abo u t something. This characterisation of knowledge is methodologically dualistic. It distinguishes the subject of knowledge from its object. Cartesianism hypostasizes this dualism; it thinks that what can be distinguished must be substantially different. There is no reason to believe this dualistic metaphysics. Kant's idea of the empirical subject indicates its weakness. I, ~own as known, can be distinguished from myself, presupposed as knowing; this does not mean that the concept of a subject, able to know itself, implies a duality of substances. In modern physics the situation seems even clearer. The physicist cannot define matter otherwise than as that which obeys the laws of physics. If these laws have their center in quantum theory, and if quantum theory is a theory of probabilistic predictions about my own behaviour will reveal me (the Ego) as matter in the defined
156
C. F. VON WEIZSACKER
sense. There is no need for the physicist to assume that there is any matter in the universe that would be different in its essential qualities from this matter revealed by self-knowledge; the fact of evolution, and the universal applicability of physics point towards the assumption of unity. Who, then, is the transcendental subject (cf. 2.3.)? In classical physics, finite objects could be considered to have an independent existence. This is no longer correct in quantum theory. In the Hilbert-space of a composite object, the state in which its components can be defined as independently existing, forms a subset of measure zero. Thus, independent fmite objects exist in quantum theory only in a classical limiting case. This is the limiting case in which observation is possible, as Bohr has insistently pointed out. Only in this limiting case will time become measurable and will its modal structure, distinguishing fact from possibility, become unambiguously applicable. Indeed, a fact is irreversible, and irreversibility itself is a concept blonging to a limiting case. This problem has never, to my knowledge, been satisfactorily analyzed in quantum theory. This lacuna in the theory is mirrored by the mathematical impossibility in the existing quantum theory of describing time otherwise than by a real, classical parameter. Thus we apply in fact classical ontology to our central concept, the concept of time. Technically speaking, this seems also to be the difficulty in establishing a consistent theory of measurement which would be simultaneously quantum-theoretical and relativistic. As long as this problem wit h the theory is not solved, there is no hope to answer the philosophical question on the concept of subjectivity pre s u ppo sed by the theory. Yet we may expect that finite subjects will no more be an ultimate reality in post-quantum physics than finite objects. I permit myself metaphorical language for expressing my feeling on a problem which is not yet rationally transparent. In quantum theory, a finite subject is a presupposition of a fmite measurement. I am, so to speak, defined by my possible actions. My actions separate me from other subjects; they lift me out of a sea of non-explicit subjectivity. My actions thus produce my time and hence the field of application of the concepts I share with other active subjects; our actions in communication produce our common time, our common past and future. This is the structure of subjectivity, as it were its Platonic Form. Physics itself now seems to indicate that this is only, as we say, a limiting case. Non-objectivating experience points beyond it. We may speak of nonobjectivating experience in the pre-human unity of emotion and action, in the human awareness of beauty and its contrary, in our fully affective structure, in the trans-human consciousness indicated by mystical experience.
THE PRECONDITIONS OF EXPERIENCE
157
I hope to be consistent with my descriplion of time in ending with open questions. Max-Planck-Institut, Starn berg NOTES 1 The main ideas are contained in my book Die Einheit der Natur [The Unity of Nature], Miinchen 1971, which however, is a collection of essays, not yet a systematic exposition. 2 K. R. Popper, 'The Rationality of Scientific Revolutions' in Rom Harre (ed.), Problems of Scientific Revolution: Progress and Obstacles to Progress in the Sciences, The Herbert Spencer lectures 1973. Clarendon Press, Oxford, 1975. 3 Th. S. Kuhn, The Strncture of Scientific Revolution, Univ. of Chicago Press, Chicago, 1962, last page. 4 W. Heisenberg, Dialectica 1948, reprinted in Schritte tiber Grenzen [Transgressions beyond Borders] , Miinchen 1971, p. 87. 5 Geometrie und Physik [Geometry and Physics], Section 7. in Physical Reality and Mathematical Description (ed. by Ch. P. Enz and J. Mehra), Reidel, Dordrecht, 1974. 6 J. D. Sneed, The Logical Structure ofMathematical Physics, Reidel, Dordrecht, 1971. 7 W. Stegmtiller, Theorie und Erfahrung [Theory and Experience]. Zweiter Halbband: Theoriestrukturen und Theoriedynamik [Structure and Dynamics of Theory], Berlin 1973. 8 Told by W. Heisenberg, Der Tei! und das Ganze [The Part and the Whole], Mtinchen 1969, p. 92. 9 This section is a condensed version of a chapter in nlY book: 'Der Garten des Menschlichen. Beitrage zur geschichtlichen Anthropologie' ['Mankind's Garden. Contributions to a Historical Anthropology] . Mtinchen 1977. 10 K. Lorenz, Die Ruckseite des Spiegels. Versuch einer Naturgeschichte des menschli-
chen Erkennens [The Mirror's Backside. An Essay and Natural History of Human Knowledge] ,Mtinchen 1973. 11 The best presentation of these critical results I find in E. Tugendhat, Vorlesungen zur Einfiihrung in die sprachanalytische Philosophie [Introductory Lectures on Analytic Philosophy] ,Frankfurt 1976. 12 While taking the full responsibility for my possible errors, I draw in this presentation on the last eight chapters of Tugendhat's book. 13 Cf. J. Lukasiewicz, Aristotle's Syllogistic from the Standpoint of Modem Formal Logic, Clarendon Press, Oxford, 1951. 14 Cf. 'Der zweite Hauptsatz und der Unterschied von Vergangenheit und Zukunft' ['The Second law of Thermodynamics and the Difference of Past and Future'] ,Annalen der Physik 36 (1939), 275. Reprinted in Die Einheit der Natur [The Unity ofNature]. 15 This apparent paradox was hinted at by Gibbs in his Statistical Mechanics and discussed by P. and T. Ehrenfest (Encykl. d. math. Wiss. IV2, II, Heft b, Leipzig 1912, reprinted in P. Ehrenfeld, Collected Scientific Papers, Amsterdam 1959.
158
C. F. VON WEIZSACKER
16 See P. Glansdorff and J. Prigogine, Thermod;/namic Theory of Structure, Stability and Fluctuation, New York 1971, and my contribution 'Evolution und Entropiewachstum' ['Evolution and Growth of Entropy'] in Offene Systeme I [Open Systenls I] (ed. by E. v. Weizsacker), Stuttgart 1974. 17 Valuable attempts like those of A. N. Prior: Time and Modality, Oxford 1957,Past, Present and Future, Oxford 1967, are not sufficient for our present task. 18 'Probability and Quantum Mechanics', Brit. Journal for the Philosophy of Science 24 (1973),321. 19 M. Drieschner, 'Quantum Mechanics as a General Theory of Prediction' (Thesis, Hamburg 1968, unpublished; described in Die Einheit der Natur II, 5), and my contribution 'Classical and Quantun1 Descriptions in J. Mehra (ed.), The Physicist's Conception ofNature, Reidel, Dordrecht, 1973. 20 E.g. J. M. Jauch in his book Foundations of Quantum Mechanics, Addison-Wesley, 1968; E. Scheibe, Die kontingenten Aussagen in der Physik [Contingent Propositions in Physics], Frankfurt 1964, The Logical Analysis of Quantum Mechanics, Pergamon 1973. 21 See, e.g. P. Mittelstaedt, Philosophische Probleme der modernen Physik [Philosophical Problems ofModem Physics] ,Mannheim 1966 2 , Chapter VI. 22 Cf. C. F. von Weizsacker, E. Scheibe, G. Stissmann, l.c. subsection 7.3. 23 For all technical details I must refer the reader to Die Einheit der Natur 11,5,5, and to the far more ample information in L. Castell's and my contributions to L. Castell, M. Drieschner, C. F. von Weizsacker (eds.), Quantum Theory and the Structures of Time and Space, Hanser, Mtinchen, 1975, and to a forthcoming second volume (Hanser, Mtinchen, 1977). 24 With less mathematical acriby I first expressed this idea in 'Komplementaritat und Logik' ['Complenlentarity and Logic'], Die Naturwissenschaften 42 (1955), Nr. 19/20. The same idea was independently proposed by D. Finkelstein, 'Space-Time-Code', Phys. Rev. (1968), and by R. Penrose in Quantum Theory and Beyond (ed. by T. Bastin), Cambridge, University Press, 1971. 2S See also C. F. von Weizsacker, E. Scheibe, G. Stissmann, Z. Naturforschung 13a ( 1958), 705 and D. Finkelstein 1. c. 26 This is my simplification of Picht's theory of time: G. Picht, Die Erfahrnng der Geschichte [The Experience of History] , vol. VI, 1958, reprinted in G. Picht, f1,'ahrheit, Vernunft, Verantwortung [Truth, Reason, Responsibility] ,Stuttgart 1969; and G. Picht, 'Theorie und Meditation' ['Theory and Meditation'] ,Merkur 4 (1974).
MARY HESSE
COMMENT ON VON WEIZsAcKER
In his fascinating and wide-ranging paper, von Weizsacker introduces his investigation into the subject of this Symposium by means of the transcendental question "How are scientific revolutions possible?", and its particularization "What is the nleaning of quantum theory?" Since his discussion is subtle and in some places technical, I shall first try to set out an overview of his paper in the form of what I take to be his answer to the initial question. 1. HOW ARE SCIENTIFIC REVOLUTIONS POSSIBLE?
This question arises out of two intellectual developments of the 20th century - first the actual conceptual and ontological revolutions in relativity theory and quantum theory, and second the newly sophisticated history of science in which conceptual analysis has destroyed the picture of science as a simply progressive accumulation of discovery, and replaced it by a historical sequence of fundamental paradigms, between which there seems to be no continuity of ontological framework, Le. of modes of description of the external world. Now we should note first of all that the question "How are scientific revolutions possible?" would be very simply answered if we were to content OUfselves with a purely relativistic view of the conceptual context of science - in that case revolutions are possible because different historical conditions throw up different ways of understanding the natural world. This answer assinlilates science to the same category as the cosmological and mythological views of nature held in one form or another by all peoples. The question bites, however, just because it presupposes that there is more to be said about science than this. What more von Weizsacker wants to say emerges at various places in his paper, and can be summarized like this. Historic forms of science are indeed context-dependent, and it cannot be supposed that any attempt, such as Kant's, to express the a priori will be of permanent validity. It will be valid only for the actual scientific and observational language in which it is expressed. But science does seem to be progressive in two other ways. First it has the function of '''learning from experience" (p. 140), which it performs successfully across language-contexts. Secondly, its changing context-dependent concepts seem nevertheless to pernlit progress 159 P. Bieri, R.-P. Horstmann, and L. KrUger (eds.), Transcendental Arguments and Science, 159-170. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
160
MARY HESSE
"towards a substantial unity" (p. 154). mus the original question should perhaps be formulated: "How are scientific revolutions possible, given that science does successfully learn from experience, and that its theory tends towards a substantial unity?" With von Weitzsacker's presupposition about learning from experience I have no quarrel - indeed I welcome his bold and explicit discussion of cognition, language and logic in terms of evolutionary biology. There is much more to be said about this, but I shall not say it here. I am more troubled, however, about the meaning of his other presupposition, that science progresses towards a substantial unity. If I have understood his programme correctly, he wishes to derive a fundamental mathematical structure (a symmetry group) from the basic requirements of the current quantum theory of observables, together with a postulate to the effect that the results of observation should with sufficient approximation be reducible to atomistically independent, bivalent, and compositional alternatives (equivalent to bits of information). It is then hoped to show how the laws of physical geometry and relativity, and the objects and forces of particle physics, can be deduced as specifications of this fundamental group. If such a programme (which I shall call the transcendental programme) were successful, it would indeed show that there is a structure presupposed in our present conception of science which has been and is exhibited in the mathematical structure of all fundamental theories. The answer to the transcendental question would then be: Scientific revolutions are possible because they are comparatively superficial. Underlying every conceptual revolution is a unity of structure which is dependent on our measurement theory, and so long as we maintain that theory, science will exhibit unity through revolutions. Moreover, the measurement theory is a characteristic of language and of the classical logic of propositions, and these in turn can be explained by the biological need to learn from experience and are even foreshadowed by the information-gathering process of evolution. Hence, given that we observe the function of science to be a continuation of this learning process, we can argue back to its unity of structure. This I believe to be the bare bones of the argument of the paper, insofar as it sets out to answer the original transcendental question. It is a bold programme, which, as I think von Weizsacker hints on p. 155, has its fascination for the physicist even if the philosophical debate seems to require that more be said. It is, I think, proper to ask how this programme differs from the answers given to the transcendental question in current philosophy of science. In comparison with these it seems to limit its concern in two ways. First, it applies only to fundamental laws, not to special laws: "The problem of an empirical justifica-
COMMENT ON VON WEIZSACKER
161
cation of special laws of nature ... is insoluble because it does not exist. It does not exist because there are no ultimate special laws of nature" (p. 138). Special laws in their historical and conceptual context are only hypotheses that come and go. Thus the programn1e seen1S to assimilate the relativism of scientific revolutions insofar as concerns the expression of particular theories of science, only concerning itself with the progress towards fundamental structure. But it is not clear that this would satisfy those who ask how progress through conceptual change is possible. It certainly excludes the realist answer (exemplified for instance in Putnam's "What is 'Realism'?"),that after all science does discover real entities and properties which will remain in the ontology of all future science. I have argued elsewhere that this strong realist claim is both unjustifiable and not required by an adequate conception of science. 1 But the transcendental programme so far appears to lack an explanation of how lower-level laws do persist in at least approximate form through conceptual revolutions. It is just this persistence that explains the success of science as a learning device. Although it may be possible to show how the fundamental symmetry group yields these approximate law-like structures on which the successful application of science depends, that is (to say the least) a highly ambitious programme, and nothing done so far seems to suggest even in principle how it might be carried out. The second objection likely to come from philosophers of theory-change is that the conception of a revolution is not understood radically enough in the transcendental programme. For that programme rests on the formal structure of quantum theory, and von Weizsacker appears to think it inconceivable that, though it is historically conditioned, "the present state of scientific consciousness" (p. 151) has too much distorted our knowledge of this formal structure. But if the present state of consciousness is not just to be circularly defined as a state which includes quantum theory, this seems an excessively bold claim. For compare the endeavours of Maupertuis and d'Alembert to uncover the a priori necessary structure of the Newtonian laws in the then current 'state of scientific consciousness'. It did not require abandonment of science as we know it to show that their attempt was mistaken and Newtonian mechanics required revolutionary change. Of course, we n1ay be able to see by hindsight that Newtonian mechanics was a particular specification of the same fundamental structure that we can now see has persisted in physics until the present time, but is there any case in history where a structure then picked out as a priori in von Weizacker's sense has persisted through revolutions over a long-term future? If not, there see11?-s no reason to suppose that the quantum symmetry groups may not be superceded in ways not now predictable, and that without abandoning
162
MARY HESSE
the very meaning of our science. If that were to be so, it would seem safer to rely on the persistence of the lower-level laws that are tested by applicability, than on the persistence of fundamental structures. I conclude this section, however, by saying that in terms of the trichotomy realism, positivisn1, and transcendentalism, as he describes them, I am sure von Weizsacker is right to adopt transcendentalism. The significance of science is neither the progressive discovery of more and more true statements about the ontology of the external world, nor is it concerned merely with observables. It is concerned with learning the structures of the world that lie beyond the expressions and concepts of any particular language. There are, however, some more detailed points to be considered, which I present in the form of possible difficulties in the way of the transcendental programme, drawn from some wellknown issues in the philosophy of physics. 2. ONTOLOGY
Von Weizsacker notes early in the paper that quantum theory requires the abandonment of classical ontology, and objects to most realist positions in philosophy of science that they do not take account of this. My purpose in this section will be to try to clarify these points and to examine the concept of ontology that emerges from the paper. There are two features of von Weizsacker's discussion that provide a starting point. First, he mentions his own adherence to the Copenhagen Interpretation (CI), although he does not specify in what form he understands this. Secondly, he interprets the change of ontology required by quantum theory primarily in terms of a change of logic. Now most current discussion of the ontology of quantum theory has been concerned with the problematic character of just these two theses. Let us consider therefore whether the transcendental programnie is affected by these debates. I shall deal with the thesis about quantum logic briefly. It is undoubtedly more fundamental than most discussions of theCI philosophically speaking, and will come up again in my fourth section. But its resolution does not, I think, crucially affect the following problems surrounding the CI. The point is that if quantum theory is expressed in classical logic, there are elements of the Boolean lattice of propositions with which no truth value can be associated (for example the conjoint proposition that a given particle at a given time is in a state corresponding to a certain precise position and a certain precise momentum). This state of affairs can be interpreted in at least three ways. Realists hope to be able to restore all the Boolean truth values in principle by
COMMENT ON VON WEIZsAcKER
163
asserting that all the propositions have truth value ontologically speaking, although the uncertainly principle prevents us ever knowing what these simultaneous truth values are. This interpretation tries consistently to interpret the the probabilities of states as ignorance 2 probabilities, in the sense that the probability distribution over the heads and tails exhibited by a number of coins lying on the floor is an ignorance probability. The interpretation is essentially an attempt to return to the classical ontology of particles; it leads to so-called hidden variables theories, and the present state of debate is not favourable to its truth. Without going into the technicalities here, I shall follow von Weizacker in not considering this type of realism further. The second and third interpretations are both versions of the CI, one of which retains classical logic throughout, while the other replaces the Boolean lattice of state descriptions by a non-Boolean lattice. Whether the consequence of the uncertainty principle are expressed by means of the physical or the logical postulates of the theory, however, certain problems remain about the interpretation of the CI. Various versions of the CI have little in common except that they reject the realist interpretation. This rejection is often cast in positivist language: only observables are real, hence what cannot in principle be nleasured does not exist. This, however, would be too extreme a version to be consistent with the transcendental programme, since that programme presumably also wishes to hold that there is a mathematical structure to which observables in some sense conform, and that this structure has implications at least for future observation, and perhaps for counterfactual inference, which go beyond what is now given. Here I would remark parenthetically that if this form of transcendentalism admits, as von Weizsacker does, that the a priori structure is context dependent on our current scientific language, it seems to be better described as a sophisticated form of realism. For the a priori structure is not necessary tout court, but necessary only on the assumption that current quantum theory is applicable to nature. In other words, as argued before, the transcendental component is a component of an empirical theory, and therefore it is as empirical and potentially as realistic as quantum theory is. Even von Weizsacker has not argued that quantum theory is wholly derivable by a transcendental argument from the conditions of experience as such. He rather seems to believe that quantum theory (or an approximation to it) will persist in science because it is empirical and true, and this is surely a form of realism, not of transcendentalism in the traditional sense. I shall take the CI to mean that, out of the whole classical ontology of particles and their dynamical states, only those propositions have truth value that ,ascribe to dynamical variables values that are or could be the results of
164
MARY HESSE
measurement. And as Bohr continually insisted, these measurements n1ust be specified within the classical ontology, because they are the results of nlacrosized instruments, and more importantly because what the dynamical variables are that are measured by a given apparatus (position, momentum, etc.), are specified by means of classical theory. The CI thus excludes any progran1me of realistic classical interpretation of the states of systems between measurements. This way of expressing the CI already indicates that there are problems about the possibility of reducing classical physics to quantum physics. I take it that the transcendental programme would require at least this, so that obstacles to it are obstacles to the programme. The apparent need for the catagories of classical ontology in order to specify quantum measurement, and the difficulties in the way of eliminating this requirement in favour of purely quantum categories, have given rise to the so-called problem of measurement. This, however, would not need to b~ a problem for the CI if one could regard the function of classical physics in regard to measurement as purely heuristic: it suggests what the significance of given measuring apparatus is for the interpretation of quantunl observables, but does not enter into the content of the quantum theory itself. This view underlies a remark of Hooker's to the effect that the problem of measurement is no problem for Bohr's interpretation of quantum theory. 3 However, several paradoxes of measurement discussed in the literature indicate that this simple resolution of the problem of measurement will not do. The paradoxes are usually discussed in the context of attempted realistic interpretations of quantum theory, but I shall now describe one of them in terms that I think pose difficulties also for the CI. This is the so-called paradox of the Schrodinger cat. The details are familiar: their essence is that we have an amplification to the macro-level of the uncertainty of position of a single micro-particle which is supposed to have a probability of one half of being in state A or state B respectively. Observation that the cat is dead indicates state A, that it is alive state B. Since we are presupposing the CI, a description of the situation must be given purely in terms of values of potential measurements and their probabilities, in other words, of the solutions of Schrodinger's equation with the given boundary conditions. In these terms we start with a description of the particle in a superposition of states A and B with probability of a half each, and we end with an observed state of a micro-system either 'alive' or 'dead'. Now a superposition of states must be interpreted consistently with quantum theory as asserting that the particle is both in state A and in state B, with probability of a half each,
COMMENT ON VON WEIZSACKER
165
where probability is interpreted not as a matter of ignorance but ontologically. That is, it must not be interpreted as the assertion that the particle is either (in A and not B) or (in B and not A), where the probability is a measure of our ignorance as to which state it is 'really' in. A superposition is an essentially non-classical conception. Moreover, it is known that no evolution of solutions of Schrodinger's equation can carry superpositions into mixtures of states of ignorance unless an ad hoc postulate is added variously referred to as the projection postulate, or the collapse of the wave-packet, or the realization of an eigenstate due to measurement. This postulate is required precisely where macro-states interact with classically described micro-states, that is, typically in cases of measurement. There is no purely quantum theoretical explanation of why eigenstates are realized on measurement, for the most a quantum theoretical account of the process of measurement could do without the projection postulate would be to show how the initial superposition state of the particle evolves into a superposition state of the cat, that is, a cat that is both alive and dead with an ontological probability of a half for each state. But this is a state that does not occur in nature, or at least we would not know how to recognise it if it did occur. We presuppose that we know what possible macro-states there are before adopting the projection postulate. The postulate is a bridge between quantum and classical theory without which the behaviour of classical objects cannot be explained. It seems to follow that there is no reduction in the traditional sense of classical physics to quantum physics, and a fortiori no reduction of biological systems, consciousness, and cognitive processes, etc. as the transcendental programme seems to require. Indeed if I have understood von Wiezsacker's postulate of semantic consistency correctly, these considerations seem to show that quantum theory violates it, because quantum theory does not describe measurement consistently with the application of quantum theory to the measuring process. If it did the results of our measurements should be superpositions of eigenstates, not eigenstates. Yet von Weizsacker says of the special laws of physics that they can ... be deduced from the fundamental laws as their special cases, if applied to special situations or problems. More profoundly: special situations and problems themselves are mathematically possible consequences from the fundamental laws. In other words, the theory itself defined the possible cases to which it can be applied. (p. 139)
This does not appear to be the case for the amplification of the uncertainty of a single quantum particle to the observable macro-domain. It may be replied that it does apply to the macro-consequences in the
166
MARY HESSE
limiting case of a large number of particles, sg that the normal behaviour of classical bodies that we observe is deducible from quantum theory in the statistical limit. Von Weizsacker writes: "classical physics ... is precisely that limi ting case to quantum theory in which the consequences of the possibilistic structure of the future disappear." (p. 145). However, without the projection postulate this is not true either. For all that Schrodinger's equation can tell us in the case of observable statistical distributions is that each quantum particle is spread over a superposition of states corresponding to the distribution, not that each particle has a definite position which yields the distribution overall. The observable consequences in statistical cases are identical, but the interpretation is different. That the different interpretations do correspond to real physical differences is indicated by the single particle case, which is the essence of Schrodinger's paradox. (One ought to put in a caveat here about the various suggested resolutions of Schrodinger's paradox. Some remarks of Bohr, and especially of Wigner, have indicated an alternative, idealist, interpretation, which would involve a highly non-reductive theory of consciousness. But this does not seem to be an option for von Weizsacker, unless perhaps he would be prepared to take a line on the role of the observer similar to the one I outline in my next section). 3. THE ROLE OF THE OBSERVER AND OBJECTIVATION
My next query concerns the role of the observer and the concept of 'objectivation'. Von Weizsacker explains the two senses in which this concept is used in traditional transcendentalism: "They can either mean that we express our experience as experience of objects, or that we produce objective experience" (p. 146). His transcendental programme accepts that we produce objective experience, and attempts to explain how we express that experience as experience of objects. His method involves the deduction of the fundamental laws of physical objects from our quantum theoretical language and atomic theory of measurement as I have described above. Hence von Weizsacker's answer to his question posed at the beginning of the paper: "What is the role of the observer for the meaning of the terms of physics?", seems to be that the community of observers imposes upon physics their intersubjective language and assumptions about measurement, and it is relative to these that physical concepts get their meaning. So far the general thesis is familiar to philosophers of science who follow the new 'idealists' in holding that the meaning of theoretical terms in science is contextual, and depends on the nature of the theoretical postulates as well (perhaps) as on their indirect relation to empirical data.
COMMENT ON VON WEIZsAcKER
167
This view would not, however, seen1 to be applicable to all physical theories, and does not seem to do justice to the arguhlen ts of Bohr and others to the effect that in quantum theory we are faced with a theory that shows in a peculiar way that the observer (or 'subject') cannot be divorced from the object investigated. There is a certain 'holism' in the Bohrian approach to quantum theory that does not seem consistent with the atomism of von Weizsacker's theory of observer involvement. Let us see whether there is really a connict here. Bohr's various statement on the subject-object relation, and his related concept of complementarity, are brief and obscure. 4 However, one can perhaps gloss his view as follows. There are various empirical results whose expression in quantum theoretic terms shows that neither instruments of measurement, nor observers regarded as physical recording devices, are so separate from the objects measured that the physical interaction between them can be disregarded. The uncertainty, principle in particular shows that in some circumstances this interaction is strong, and it can never in principle be allowed for and calculated away, as is the case in classical physics. These interactions are shown in the impossibility of measuring each of a pair of noncornmuting variables precisely, and are exhibited particularly strikingly in the so-called EPR experiment, where it seems that a free decision of the observer to measure one or other of a pair of such variables affects the value of another observable at a distance and without physical mediation, that is, instantaneously and without intervening passage of energy. Quantum systems are holistic in the sense that states of more than one particle are not splitable into sub states communicating only by external physical n1eans, but their parts are as it were 'instantaneously present to each other'. In particular, the observer is in principle instantaneously present to all parts of his object-systems, and this not in any dualist sense of an all-pervasive consciousness, but simply because he is, at least, a physical mechanism. If, however, we now consider that the observer is engaged in the process of trying to describe his objects without at the same time describing himself, he is bound to be describing a ruptured and to some extent distorted system. Hence, according to Bohr, arises the apparent paradox that one cannot describe reality consistently in anyone of the classical models of particles or waves, and also the Schrodinger and EPR paradoxes which both presuppose that the initial unity of observer and system is broken in order that an objective account of the system alone can be given. Bohr's insights here are perhaps the deepest of any in the literature of quatum theory, though subject to multifarious interpretation. Having tried to
168
MARY HESSE
elucidate the meaning of 'objectivation' in Bohr, I think we can conclude that von Weizsacker's programme is not inconsistent with it, for the following reason. Bohr was concerned with the particular consequences, for an atomistic classical ontology, of quantum theoretic formulations, and these consequences turn out to be startling: no classical models of external reality are possible, there is no physical determinisn1, and the relation of parts of physical systems including observers are strongly holistic. But von Weizsacker has been concerned with the structure of the theory rather than with its application to particular experin1ental situations, and (as he indicates in his penultimate paragraph) he has accepted from the beginning what Bohr regarded as the essential subject-object rupture, namely the decision to describe reality as if it were separable from the observer, and moreover, to describe it in atomistic and compositional language. (Perhaps all language is thus atomistic and compositional, but I think that still has to be· shown, and would take us too far afield here). 4. TIME
Finally I am very conscious that I have so far said nothing about what von Weizsacker may well regard as the most important element of his paper, nan1ely its conception of time. This conception can be decribed as follows. Our temporal experience is of a factual past and a future of possibilities. However, "experience means to learn from the past for the future" (p. 140), that is, the future is to some extent open but to some extent constrained by the past, and we can express what we learn about its possibilities by probability measures. This is the case in classical statistical thermodynamics, in Darwinian evolution, in quantum theory, and generally in inductive learning. Von Weizsacker wants to argue in addition that in all these cases a 'temporal' nonclassical logic is prior to the statement of these theories, because it is a transcendental precondition of experience. In particular, the second law of thermodynamics, which is a fact of experience, cannot be derived from statistical thermodynamics alone, since statistical thermodynamics is time-symmetrical about any given moment. The second law can only be derived from that theory together with the presupposition that we know that in the past the entropy of closed systems has increased in time. Similarly, Hume's problem of induction cannot be formulated without the assun1ption that the past is in principle known and the future open, hence the future cannot be deduced from the past. If I have correctly interpreted von Weizsacker's conception, the question arises as to how he wishes to relate it to quantum theory, which has· prima
COMMENT ON VON WEIZSACKER
169
facie different temporal characteristi,cs from the other applications. Here, as in statistical thermodynanlics, we have a theory that is essentially timesymmetrical. Again, as in statistical thermodynamics, we have sonle propositions about the past to which truth value can be assigned, namely past results of measurement, and we have an open future, to the possibilities of which probabilities can be assigned. However there appears to be one fundamental difference between classical statistical thermodynamics and quantum theory, namely that in the former we can conceive that all elements of the Boolean lattice of propositions about the past get, as it were, fIlled in by truth values as they become past tense propositions. That is, in all classical theory the temporality of experience is consistent with a completely factual and decidable past which is wholly independent of our actions now. This, however, is not the case in quantum theory. There we can only fill in those elements refering to the past that have actually been assigned truth values by measurement.. Many elements will remain empty because the prohibition on simultaneous nleasurement of non-commuting variables forbids decision about truth values for other elements. The CI is consistently applied to the past by asserting that there was no completely factual past, any more than there is a completely factual future, with regard to the truth values of classical logic. In this respect quantunl theory is time-symmetrical in a sense in which classical thermodynamics is not. What truth values get ascribed to the past, moreover, depends on what decisions we made about what to observe, for example a value of position coordinate existed because we then decided to measure it and not its conjugate variable. It is not just that truths and falsities were defined in the past by our actions, as in classical theory, but the very domain of possible truths and falsities was decided by them. This is just one of the indications that the quantum formalism may not be the last word. Discussions of the non-Boolean character of quantum logic are based on the presupposition that it is propositions about classically defined dynamical variables that form the universe of discourse. That not all of them have truth values may not indicate the need for a change of logic, but rather for a revolutionary break altogether with classical forms of expression. We have no idea what such a radically reformulated quantum theory would look like, but if it arrives it may become clear that the apparent transcendental continuity of fundamental mathematical structure from classical through quantum theory is an artefact of our classical ontology, not our classical logic, nor our basic view of science. In this case the transcendental programme would be of only limited applicability. University of Cambridge
170
MARY HESSE
1 Cf. H. Putnam, 'What is "Realism"?', Proc. Aris. Soc. (1975/6),177, and Mary Hesse, 'Truth and the Growth of Scientific Knowledge'. PSA 1976, (ed. by F. Suppe and Peter D. Asquith), vol. II, 1977. 2 I use the the terms 'ignorance' anq. 'ontological' to qualify two interpretations of probability, rather than the more familiar 'subjective' and 'objective', which are misleading in quantum contexts. 3 C. A. Hooker, 'The Nature of Quantum Mechanical Reality', University of Pittsburgh Series in the Philosophy of Science, vol. 5 (ed. by R. G. Colodny), 1972, p. 67. 4 See particularly N. Bohr, Atomic Theory and the Description of Nature, Cambridge 1961.
PETER MITTELSTAEDT
COMMENT ON VON WEIZsAcKER
The commentary which I will present here to C. F. von Weizsacker's approach to the philosophy of science will be divided into three parts. In the first part I will discuss methodological questions. This seems to be necessary since essential parts of this approach have not yet fully been worked out. In the second part I will discuss the relations between time, logic, and quantum logic and in the third part I will briefly consider the structure of space and time.
1. Kant has shown in the Critique ofPure Reason how the most general principles of nature science can be justified starting from the very conditions for the possibility of experience. For instance, the law of causality - "everything that happens -- presupposes something upon which it follows according to a rule" (A 189) - follows from the principle under which the temporal order of events will be established. For that reason the causal law is a synthetic judgment a priori which can, however, only be proved valid for objects of experience. C. F. von Weizsacker's program is much more ambitious. He emphasizes that he is not satisfied by justifying the principles of pure understanding - as has been done in the Critique of Pure Reason (p. 3). Instead, he tries to justify the fundamental laws of physics. As long as these fundamental laws are not yet known explicitly, - this attempt of justification means that the totality of special physical laws must be shown to be a priori valid. 2. In the Critique ofPure Reason we don't find any indication of how special physical laws can be justified. Laws which can be justified by transcendental arguments must follow from the preconditions of experience and a proof of this kind seems to be quite impossible for instance for the gravitationa11a ,v. However, Kant has actually tried to give a justification for a few special laws of classical mechanics, but without having clarified in a sufficient manner the methodological basis of this proof procedure. Even if one would accept the proof attempts in the 'metaphysische Anfangsgriinde' one could not ge"t sufficient guidance for perfonning proofs in 1
171 P. Bieri, R.-P. Horstmann, and L. KrUger (eds.), Transcendental Arguments and Science, 171-176. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
172
PETER MITTELSTAEDT
the theory under consideration since, within the framework of Kant's approach, there are still synthetic judgements a posteriori. According to von Weizsacker's program, all special laws of physics should be reduced to the fundamental laws, the a priori validity of which can then be demonstrated by transcendental arguments. Therefore in the final version of this theory there will be no special empirical laws at all. 3. Of course, von Weizsacker is fully aware of this situation and consequently he dispenses with Kant's arguments as a guiding principle: "Hence I shall not draw on Kant's philosophy in this paper ..." (p. 129). However, what I am missing here is a clear elaboration of the proof techniques that will now actually be applied in this paper in order to perform the program. Coming to the systematic steps, von Weizsacker formulates the hypothesis, "that the fundamental laws of physics can indeed be fully justified as preconditions of all possible objectivating experience". Here the difficult questions arise: first, what concept of experience is used in this hypothesis and, secondly, what is (technically) meant by the term 'fully justified'. I will come back to these two questions, which are in my opinion of great importance, in my discussion of special problems. Furthermore von Weizsacker uses the formulation "that the fundamental laws can be justified as preconditions of ...". Obviously he does not state that the fundamental laws follow from the preconditions of experience - as is the case in the Kantian demonstration. Instead, the physical laws are considered as preconditions of experience and as synthetic judgments a priori. Thus the argumentation seems to contain a petitio principii and it should be clarified further, what is really meant by this circularity. II
I will now turn to the systematic part of the paper. Since it is quite impossible to comment on all of the arguments presented there, I will confme myself to two problems, the first of which will be concerned with time, logic and quantum logic. The relations between quantum logic and time, and the attempt to justify quantum logic by a temporal logic, are of central interest for the systematics of the paper. In fact, quantum logic seems to be the first example of a fully developed structure which can be justified by transcendental arguments only and which might be considered as a fundamental law of physics. 1. Before I discuss quantum logic, I should like to comment briefly on the
COMMENT ON VON WEIZSACKER
173
situation in ordinary logic. Von Weizsacker mentions the a priori validity of the laws of logic at two places, and he gives two independent arguments for it. The first argument is with respect to Kant: Since analytical judgments a priori can be proved by a recourse to logic only, the laws of logic themselves must be synthetic judgments a priori. - The second argument recalls the operational justification of the effective logic: Since the laws of logic can be based on the general possibilities of proving propositions, this foundation obviously has some similarity to a transcendental argument. Hence, von Weizsacker claims that "the laws of logic are synthetic judgments a priori". In spite of the striking analogy between the operational foundation of logic and a transcendental argument in the spirit of Kant I would propose to use here a more careful terminology. Synthetic judgments a priori are based on those principles of pure understanding which constitute the objects of experience. Hence one could speak here of an a priori ofcategories. However, the laws of logic are based on linguistic actions and on rules which define the possibilities of proving propositions. For the validity of the laws of logic, objects of experience are not needed at all. Therefore, one should better speak of an a priori of argumentation. Both kinds of a priori judgments will be important in a scientific theory but they are different and independent components of our knowledge. 2. Next I turn to the concept of temporal logic , which seems to be of fundamental in1portance for the entire approach. Since the ten1porallogic itself is not presented in the paper, there are a few open questions which I would like to mention here briefly. The formalisation of a logic of time-dependent propositions has been carried out by N. Rescher and A. Urquhart and by A. N. Prior. The formalism is in its essential parts the modal logic S5, the semantics of which is given here by the time dependence of propositions. However, this fonnalism does not seem to be a convenient tool for justifying logic and quantum logic. This conjecture is obviously confirmed by von Weizsacker's comment that the formalism just mentioned is "not sufficient for the present task". A second point which I wish to discuss is the relation between time de..; pendence and the value defmiteness of propositions. The question whether in the framework of the object language - a proposition is value definite or not, depends on the possibilities of proving or disproving this proposition. If there exists a proof-procedure which decides between 'true' and 'false', the proposition in question will he value definite. However, if one considers in the. framework of the meta-language the modalities 'necessary' and 'possibe' a
174
PETER MITTELSTAEDT
given proposition is in general not 'necessarily true' or 'necessarily false'. This is the case for all propositions irrespective of their value-defineteness and, in particular, for statements about the futllre. For this reason I wonder whether it is really useful "to define statements on the future as neither true or false", as has beeen proposed by Weizsacker. 3. The last and most important problem concerning time and logic is the question whether "timeless logic is systematically prior to temporal logic or vice versa". It is obvious that from an operationalist point of view the starting point for establishing logic are time-dependent propositions which can be ordered in a temporal sequence. In this way a pragmatic description of dialogs can be achieved. However, in order to define the logical connectives, one has to relate two or more propositions precisely to the same time value. Starting from tin1e-dependent dialogs, this can be done by a limiting process. Further steps, which are of more formal character, then lead to the calculus of effective logic. However already as early as the definition of the logical connectives, the important difference between logic and quantum logic can be observed. For time-dependent propositions about classical systems the limit of vanishing time difference can always be performed without any difficulty. However, for time-dependent propositions about quantum systems the temporal order may be important, even in the limit of vanishing time difference. Therefore, the limiting process is only defined if the propositions can be shown to be commutable at the san1e time. In this way one arrives again at time-independent connectives which finally lead to the calculus of effective quantum logic. I have described the situation in detail in order to demonstrate that logic as well as quantum logic are structures which are based on timeless connectives, whereas the semantics of both of these structures are related to timedependent propositions. Therefore, without further explanation I cannot agree with the statements made in this paper, that quantum logic can be interpreted "as a futuric modal logic". Temporal logic is neither contained in effective logic nor in effective quantum logic. This point seen1S to be worth further discussion. I should like to add here the point that the time-independence of formal quantum logic is in fact of decisive importance for the program of an a priori justification of the fundamental laws of physics. Here the fundamental law is the algebraic structure of the Hilbert space, Le. the lattice of subspaces which is, of course, time-independent. It has been shown that the calculus of formal quantum logic, which can be justified without any recourse to empirical
COMMENT ON VON WEIZsAcKER
175
knowledge, is in fact a model of this lattice. Therefore, the algebraic structure of Hilbert space can be based on a prior:~ reasons only. III
The last part of my commentary is concerned with the structure of space and time. The attempt to justify the laws on the space time frame is perhaps the most developed problem in Weizsacker's approach, but it is only briefly summarized in Section 7 of the present paper. For all formal details the reader is referred to the literature. Therefore I will not go into the technical problems of group theory here in detail, but instead will concentrate on questions of more fundamental interest.
1. The basic idea of von Weizsacker's approach is to reduce the structure of space and time to a symmetry group and to reduce this symmetry group to the already justified abstract quantum theory. The structure of space-time, i.e. the three-dimensional geometry of the real space and the metric of the empiricaf time have also been justified by transcendental arguments in the Critique of Pure Reason. However, Kant shows that the laws of Euclidean geometry can be justified from the forms of pure intuition and from the postulates for consturcting the geometrical objects. In Kants approach the a priori validity of the geometrical axioms is a consequence of the preconditions of experience, Le. of the forms of pure intuition. Therefore, one should speak more precisely of an a priori of intuition. In contrast to this justification of the space-time structure, von Weizsacker tries to establish the symmetry group of space-time by a reduction to quantum theory and quantum logic. This means that the a priori validity of the spacetime structure which can be achieved in this approach, is a priori in the sense of the a priori of argementation. It would be interesting to better understand the relations between these two kinds of truth. 2. There is another attempt to establish the structure of the empirical space time, the con1parison of which with von Weizsacker's approach seems to be interesting. What I mean here is the operational justification of the Euclidean geometry and the metrical time, which has been partly performed by P. Lorenzen and P. Janich. This approach, which goes back to H. Dingler, has been interpreted by these authors as an elab oration and continuation of Kant's ideas. However, the new and interesting aspect of this foundation of space-time
176
PETER MITTELST AEDT
is the reduction to operations to pragmatfc actions. therefore, we are confronted here with a further kind of a priori validity which could be called pragmatic a priori. Here the space-time structure results from the pragmatic preconditions of experience, Le. from postulates for constructive actions. It would be worthwhile to further investigate the relations between the two a priori justifications of the space-time structure, both of which are based on a recourse to the preconditions of experience, and each of which is interpreted by its author as a continuation of the Kantian approach. Universitiit Kdln
KUNO LORENZ
THE CONCEPT OF SCIENCE. SOME REMARKS ON THE METHODOLOGICAL ISSUE 'CONSTRUCTION' VERSUS 'DESCRIPTION' IN THE PHILOSOPHY OF SCIENCE Rationalism and empiricism, the kindred branches of modern western philosophy since its inception with Descartes and Hobbes, have made much of the distinction between mind and body. The corresponding schism between the act and the given, reminiscent of the Aristotelian categories a'Y€tv and rraaX€tv, has consequently to be looked at differently within the philosophy of nature and the philosophy of mind. We have to consider rationalism and empiricism independently in both areas though usually they occupy attention only within the philosophy of nature. So it happens, e.g. in phenomenology, that a kind of rationalism with respect to the outer world is matched by an equally stringent kind of empiricism though hardly called this way - with respect to the inner world. Among current disputes, the well-known discussion on innateness! is another example which shows the relevance of that phenomenon, especially in connection with a widespread misconception of innateness in Kant: 2 Not ideas are innate, but our ability to make them is innate. And, certainly, the discussion is not yet closed on the issue whether dispositions - under some biological model - can be treated as facts rather than as mere possibilities of facts, Le. something like second-order-facts, if that phrase makes sense at all. Therefore, when I use the tenns 'rationalism' and 'empiricism' together 'with their cognates in the following remarks, they should be understood in that general sense, not restricted to their use in the philosophy of nature.
The two main developments originating through well-known problems of epistemology in either of these philosophical positions may now be characterised in the following way: Out of rationalism emerges transcendentalisnl to secure a unique set-up of at least the natural sciences, mathematics included. Empiricism, on the other hand, gave way to evolutionariSTI1, SOTIle kind of free choice - or necessary change - principle to be used for starting, e.g. the sciences or any other human artefact. It is common opinion to treat an epistemology of the first kind as the only 177 P. Bieri, R.-P. Horstmann, and L. KrUger (eds.), Transcendental Arguments and Science, 177-190. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
178
KUNO LORENZ
way out of epistemic scepticism taken seriously~ whereas an epistemology of the second kind bounds scepticism by some cornmon-sense-relativism which implies dropping any reliance upon science as a substitute for religion concerning matters of fundamental world view. This frame for dealing with the claims of scepticism hides a difference of presuppositions in the philosophy of science which, spelled out, may well carry a chance with it to provide for a truly unified treatment of science. I have in mind the difference of treating science as a way of presentation and as a way of research, i.e. the difference of the old person-oriented ars iudicandi on the one hand and the matter-oriented ars inveniendi on the other hand. Just as a remark I should add that this Leibnizian difference of Analysis and Synthesis, as he calls these two )artes( alternatively, is more general than Reichenbach's distinction between a context of justification and a context of discovery, both relating to propositions in their function as scientific hypotheses and thus serving a better understanding of the difference between deductive and inductive nlethods. 3 Transcendentalisnl searches for justifiable presentations (being true theories of certain domains of objects) whereas evolutionarism represents a way of adequate research (being significant encounters with certain kinds of objects), such that in the first case we get well-founded sequences of propositions, yet in the second case a well-deternlined network of mutually related objects. At once a further complication arises. To search for sequences of propositions is certainly not a purely linguistic matter. Instead of just giving a construction of certain objects on the language-level, the crucial issue is to judge upon their )relation( to the object-level with the aim of securing their truth. Analogously, to represent a network of objects cannot be done on the objectlevel alone; representation is bound to rely on linguistic means with the aim to determine the objects by precise descriptions. In order to avoid erroneous identifications, I have deliberately used here the terms 'search' and 'represent' to refer to activities within science as presentation and science as research, respectively. For, certainly, there is second-order research concerning presentations (e .g. within what is called 'science of science') and second-order presentation concerning research (e.g. within the well-known 'logic of inquiry'), and neither should be identified with what I called 'search' and 'representation', respectively. Search within presentations is search for true presentation, whereas repre.. sentation within researches is representation of significant research. What is at
THE CONCEPT OF SCIENCE
179
stake from the purely linguistic point of view is.. tlte question of how the verification (and falsification) of formulae is interrelated with the signification of terms. More generally, and in traditional terminology, we can say: both questions, the justification of propositions and the constitution of objects, belong together, yet must not be confounded. If the question of constitution is falsely treated as belonging to the problem of justification, it yields evolutionarisn1 as a brand of radical empiricism. 4 And again, if the question of justification is falsely treated as belonging to the problem of constitution, it yields transcendentalism as a brand of radical rationalism. To keep both questions within their proper bounds has a chance of success only if their mutual dependency is treated clearly and distinctly. This means especially to ask for methods to translate theories including states of theories - considering theory-change - into each other: In what sense nlay two synchronically or diachronically different theories have the same content - though saying different things of different entities. The most prominent example for the radical empiricism which arises when problems of constitution are treated as if they were problems of detennination, i.e. as if they concerned investigations into the tmth-conditions of assertions about the constituted objects, can be found in the evolutionary pragmatism of C. S. Peirce. s With slight modifications only, this is equally true of the )Analytische Wissenschaftstheorie< as it grew out of the Logical Empiricism of the Vienna Circle. For exalllpIe , the usual set-up of fonnalized theories has never been seriously questioned, Le. the start with given domains of objects on the one hand and sets of predicates together with suitably chosen axioms about those objects on the other hand. There is a freedom of choice in both respects - e.g. phenomenalistic systems may compete with physicalistic systems, and preference for some set of primitive notions and principles should always be treated as a contingent fact itself - yet there is usually no hint as to how somebody can acquire a position enabling him actually to choose among alternatives. This again remains a historical and thus contingent fact. Another example not from the sciences, but from the humanities, is the methodological position called 'hermeneutics'. Here, too, there is no serious question of how to get hold of the different domains of objects - most prominent: written texts -, they rather are viewed through the different scientific approaches to them, e.g. their interpretations, which forbid to ascribe an independent status to the objects. 6 In the converse case, the most pronlinent example for the radical rationalism
180
KUNO LORENZ
which arises when problems of justification are treated as if they were problems of constitution, Le. as if they concerned investigations into the conditions of possible experience (how objects of experience have to behave in order to be accessible to knowledge) can, of course, be found in the transcendental idealism of I. Kant. 7 It is, therefore, not accidental when proponents of the >Konstnlktive Wissenschaftstheorie< use arguments akin to Kantian ones to substantiate the claim that what is called 'protophysics' can serve as an apriori foundation for physics. In a similar vein, attempts of a )protosociology <shall serve as a clarification of the methodological position called 'dialectics' which tries to furnish a non-empirical foundation for the social sciences. 8 Due to the simplifications of this sketch, I could, of course, not yet take into account the numerous mixed cases deriving from insights into certain shortcomings of the respective positions as they grew out of recent scientific discussions. E.g. a certain transcendental framework got introduced into henneneutics;9 similarly, empirical boundary-conditions entered the argumentations in dialectics. 1o Both, the concentration on )the logic of evolution<, the rational conditions of any change in the Frankfurter Schule, and on )the evolution of logic<, the elnpirical change of any theory, in present-day analytic philosophy of science, show that dealing with problems of amalgamation between transcendentalism and evolutionarism is considered to be the essential and the most hopeful line of investigation now. I will not go into details here, but rather stress certain features of the discussion between the analytic and the constructive philosophy of science which are relevant for the epistemological issue in the sciences I am concerned with. II
For convenience of presentation, I will start with the discussion of a thesis, which Harald WoWrapp has convincingly defended recently: 11 The analytic philosophy of science on either of its three main stages, Carnap-Stegmtiller's empiricism, Popper-Lakatos' rationalism, and Kuhn-Feyerabend's historisn1, should - according to WoWrapp's claim - essentially be understood as concerned with science as a way of research, whereas the constructive philosophy of science of the Erlanger Schule, and of related positions, is basically concerned with science as a way of presentation. As an important consequence, the difference of criteria for what shall be considered as science can be stated. The criteria of science as research are
THE CONCEPT OF SCIENCE
181
essentially those of success, derived from the actual procedures of working scientists; necessary conditions are e.g. the use of well-defined predicates, the reliance on the consistency of the set of non-derived sentences, reproducibility of operations et alii. On the other hand, the criteria of science as presentation follow conditions of acceptability and are in this sense )foundational <; they derive from potential procedures of scientists and can be characterized essentially by two principles: the principle of method (Le. presentations work stepwise without )jun1ps( -~ a kind of completeness claim) and the principle of dialogue (Le. presentations can be criticized, which means it can be questioned whether they fulfil the first principle). If one proceeds this way and at the same time argues from a treatment of science in the first case as a fact and in the second case as a norm, a certain kind of simplification occurs and affects the conclusion. Therefore, it remains to be investigated whether the specific claims of failure and success Wohlrapp raised with respect to the two sets of criteria he discussed really carry conviction as they stand or whether failure and success rather depend on some further distinctions connected with the difference of research and presentation. To repeat what I said in the beginning, I would like to claim that inasmuch as questions of constitution are confounded with questions of justification or, to use the linguistic angle, questions of signification confounded with questions of verification - the criteria for science either as research or as presentation will lead to difficulties and eventually to failures. Success, therefore, in either case is dependent on a clarification of the interrelationship between answering on )what there is( and answering on )what is true(, the ontological and the episten1010gical version of the question of how the two levels, of objects and of signs Oworld ( and)language () separate within and unite into one domain of (scientific) language-games. I have tried elsewhere 12 to show that this domain has to be understood as a domain of preactions, beyond the action-act dualism as the prototype of the type-token division, and equally beyond the classical distinction of something given and something to do. The idea is simply to start with - from a later objects which do not yet bear the differentiation point of view, complex between actor and action or between action and object of an action or result of an action. From that starting point which is articulated by a very elementary and obviously fictitious language where only reference to such objects occurs 13 one has to develop both ordinary language and the language of science with the result that during this procedure, already, most of what is treated later on as logical or ontological presuppositions of a language gets decided.
182
KUNO LORENZ
The difficulty is that in giving a descripgon of this (re)construction the language of description, Le. some standard natural language in use, is far more developed syntactically and semantically than the described language during the process of its construction. Hence, in order to get an adequate account of the construction it is necessary to introduce certain devices which make sure that the description at any stage is not dependent on those features of the syntactic and semantic structure of the language of description which do not yet belong to the structure of the constructed language. For example, the difference of singular and general terms within the language of description should not be of relevance for describing the initial stage of construction, where within the fictitious elementary language only reference to preactions occurs. Rather, there should exist an explicit step of introducing that very difference within the elementary language. And this is done by proposing to introduce the singu1ar~general-distinction on the elementary level of non-analysed actions as the distinction of schema and actualisation. These descriptive terms refer to the difference of )once(, )once more(, )once more again (, ... , which is practically acquired in situations of repetitive in1itation (= imitative repetition) with respect to any preaction. 14 Here, linguistic signs are the meatls which have developed gradually through our evolution to articulate which general object belongs to which singular object (Le. under which concept a certain case falls or by which case a certain concept is fulfilled). Through language something singular becomes a symbol of something general, and, the other way round: through language something general becomes an aspect of something singular. Then, it is perhaps not any more offensive - by being liable to the pitfalls of a remake of the Cartesian dualisn1 - to say: the singular gives the empirical base, the general the rational design; both cannot be separated, and research starts with singularia, presentations with generalia. Since theories of both areas exist, hierarchies of theoreticity appear 15 and things including the empirical/rational distinction get n1ixed up. 16 As 1 already mentioned in connection with Wohlrapp's paper, it serves still further confusion, if in the first case the methodological position of the analytic philosophy of science is characterized as being descriptive, whereas in the second case the methodological position of the constructive philosophy of science bears the label of being normative. Inasmuch as questions of constitution have consequences in terms of stipulations concerning the objects of scientific discussions - the )definition ( (I prefer the more general term 'introduction') of basic predicates about them is included -, the insistence on
THE CONCEPT OF SCIENCE
183
the normative character of some fundatpental part of science - let us call it 'protoscience' - is reasonably supported. Similarly, questions of justification, concerning the context of these objects - the use of predicates, so to speak, not their introduction -, give rise to descriptive aspects of any science. Here again, it might be useful to recall that, in general, predicates within scientific languages are defined on given domains of objects, extensionally as certain classes of those objects. Hence, they cannot be treated as primary predicates, they are derivative with respect to the defining predicate for the domain of objects (Le. the >substances< defined as the instantiations of that primary predicates). We, then, speak of properties, and the usual problems concern questions of whether properties hold of objects which obey certain other descriptions, and they nev.er concern questions of elementary constitution. Those constitutional questions - unless they are non-elementary, Le. of second order, yielding don1ains of abstract objects (such processes are, of course, well known and extensively treated everywhere) - occur on a language level which is itself of a theoretical nature only. It is the elementary level I referred to earlier and which can now be characterized as the one where terms are introduced, not the one where they are used as in ordinary non-scientific and usual scientific language. E.g. when you assert 'This leaf turns yellow', the constitution of objects like leaves (Le. the introduction of the term 'leaf) is presupposed, whereas the constitution of objects like )yellows ( (norminal use of the term!) is pushed to a second-order level: Yellow is constructed as an abstract object turning the word 'yellow' from a non-primary general term - standing for a >characterizing universal ( in the sense of Strawson 17 - into a singular term, a nominator, as I suggest to call it. What remains in the case of the assertion in question is to judge upon the use of the terms 'leaf' and 'yellow' (or 'turn yellow') which by all standards is a question of true description. Constructions remain within one language- (or object-) level, descriptions concern two consecutive levels. This difference is well known, e.g. in logical theory, where formulae can either be constructed Iby formation-rules, or they can be described by means of suitably chosen predicates of a metalanguage (used for formalising the construction). The fundamental notions are partition (of a whole into parts) in the case of constructions, and attribution (of a property to an object) in the case of descriptions. Mereology and set-theory provide for the respective formalisations of these notions though their interrelation is by no means sufficiently clarified up to now. 18
184
KUNO LORENZ
I hope these remarks give sufficient Stlpport for the claim that there is no simple correlation between science as research and stating what is (being the result of research) on the one hand, and between science as presentation and issuing what shall be (being the guarantee of presentation) on the other hand. In either case, the set-up of science is not only a question of justifying a corpus of sentences, but a question of introducing meaningful terms used as constatives and/or as directives. And justification includes the search for the truth (science as theory) together with the search for the good (science as praxis) as much as meaning exhibits both, aspects of (theoretical) signification and aspects of (practical) relevance. III
The relevant difference of research and presentation as against the simplified accounts I discussed above comes in when we look for the support of a scientific theory. As far as science is treated as a way of research, this support should derive only from the descriptive power of the theory relative to the singular objects (= perceptual cores) of research, though usually, in the analytic phi1~sophy of science, due to the aforementioned lack of a clear separation between constitution and justification, this support is extended to include the explanatory power of the theory as well. It is common to use the term 'confirmation' (referring to non-elementary propositions in relation to relevant singular >data<) in this respect, and this entails a confrontation with the - I dare say unsolvable - riddles and paradoxes of induction. 19 The simple reason for the claim that nothing beyond the descriptive power of a theory can be treated within science as research only derives from the following considerations: The explanatory power of a theory refers to the kind of interrelations which obtain among the different propositions of the theory, especially to an assessment of the range of validity of fundamental principles like those of conservation in physics. Hence, an account of the explanatory power can be given only by judging upon the conceptual frame of the theory used for the argumentations in science as presentations. 2o The argument in the alternative case runs conversely: as far as science is treated as a way of presentation, the support of a scientific theory should derive only from its constructive power relative to the general objects (= conceptual frames) of presentation, though, usually, in the constructive philosophy of science, due to the same confoundation of constitution with justification, this support is extended to the regulatory (and, hence , normative) power of the theory as well. The term 'approxin1ation' (referring to elementary
THE CONCEPT OF SCIENCE
185
objects in relation to relevant general >idea [1] s<) is in this connection occasionally used, and, as an equally disturbing consequence, it becomes necessary to handle the vexations of the is-ought-gap?! Here again, it is easy to see that in science as presentations any attempt to go beyond the constructive power of a theory and to judge upon its power to issue what shall be - unless this is treated as a second-order question only, i.e. as a question of what kind of scientific activity (rather than objects of activity) should exist - will need reference to the perceptual cores of science as research, e.g. to the encounters with singular objects in experimental situations. What I should like to claim is that both the explanatory and the regulatory (or normative) power of a scientific theory can be assessed properly and without bias only if the set-up of science cuts straight through the separating line of research and presentation. This means especially that on each level within the hierarchy of theories the link between constitutional and justificational questions - and that refers to the interdependence of constructive with descriptive procedures as well - must not be lost sight of. Hence, trying to determine an adequate meaning of the two central concepts (scientific) explanation and (scientific) regulation amounts to nothing less but a reassessment of whether and how a unified treatment of science is possible. For this purpose, the concept of unified science should no longer be understood in the original historical setting along with a developing analytic philosophy of science. In the light of the considerations just offered I claim that a unified approach to science, unless it falls victim to typical >Scheinprobleme< as the one concerning the possibility of induction or the one concerning a bridge over the is-ought-gap, has to consider science with respect to activities both of research and of presentation. It has to develop a concept of science starting with a kind of unity of research and presentation 22 where the domain of (scientific) language-games uniting >world
language< in the sense I have outlined above becomes the result of the first step. These language-games of preactions together with their articulations can then be treated in their both aspects: matter-oriented (research) and person-oriented (presentation).23 For visualization of what I am driving at, I may use an example of current dispute: the different approaches to (physical) geometry. Concentrating on the research aspect of physics, the (temporal) behaviour of (physical) bodies relative to their spatial coordinates is judged with respect to quite general hypotheses concerning space-time-structure (explanation-bias!). The presentation aspect of physics, in the protophysics of the Erlanger Schule on the other hand, asks for a series of steps to introduce the funda-
186
KUNO LORENZ
mental concepts of geometry, chronometry and hylometry in that order using )idealized( operations with (physical) bodies (regulation-bias!). In the second case, what is done is to provide meaningful terms - that they can be used successfully outside presentational questions is taken for granted. It is not surprising that certain propositions come out true apriori. In the first case, something completely different happens: here, propositions about given objects are tested to secure their validity, which means to treat them as empirically based. The introduction of the tenns used is taken for granted inasmuch as presentational questions are considered to be a cura posterior. Thought theoretical activities of supplying true descriptions govern science in its research aspect or, rather, because of them, the presentational necessities like introducing meaningful terms get neglected. And, conversely, the concern with practical operations to get adequate constructions of fundamental concepts for science in its presentation aspect seduces into thinking low of problems whether those concepts can effectively be used in research situations. The real issue actually boils down to the question of whether the introduction of meaningful (geonletric) terms like 'straight', 'n-times the length of etc. can be treated as an extension of ordinary language about ordinary objects, serving better criteria of relevance according to further developed standards of significance and truth. For, if extendability fails, we are stuck in conceptual frames without prospects to satisfy them; and if presuppositions serve as substitute for explicit introductions, there is no chance to guarantee anything beyond the perceptual cores. It is easy to see that the last two conditional sentences may serve as a modern and more refined version of Kant's famous dictum that concepts without intuitions are em.pty and intuitions without concepts are blind. The refinement consists in the introduction of hierarchies of theoreticity starting with (of course not uniquely determined) common-sense experience phrased in everyday language. 24 The domains of objects of scientific discourse have to be arranged on levels of ascending and descending order without any chance to argue definitely for a universal )lowest( level - e.g. of elementary particles - sufficient for arbitrary future theories. The usual arguments between protophysicists and )deutero-physicists ( - if I may coin that term for the nloment - using coordinate-systems with a spatial or even a spatia-temporal metric are beside the point as long as the problem of introducing a )metric< is exempt from a truly mutual discussion. What
THE CONCEPT OF SCIENCE
187
can be introduced rather than merely postulated on the basis of elementary
common human experience (still ambiguous relative to the singular-.:general bifurcation) will lend itself to the discrimination of (empirical) actualisations from (rational) schemata for any preaction. And if we succeed in constituting a domain of objects - in classical mechanics it has been ideal mass-points in a homogeneous Euclidean space-time - which can serve as a common basis even for quite different physical theories, unification of science has been pushed one step further. Hence, I will fmally claim that in order to con1pare two theories T 1 and T 2 - even incommensurable ones against the claims of P. Feyerabend 25 - it is necessary and sufficient to develop a theory T which allows a faithful embedding VJt of T t into T. Apart from the fact that it is impossible to know in advance whether for two theories - e.g. for quantum-mechanics and general relativity-theory such a) supertheory ( exists, the actual task will consist in devising appropriate translations including )logical constructions( in the sense of B. Russe1l 26 for the objects of either theory by means of the prunitive terms of the supertheory. Furthermore, such embeddings will, through the larger contexts the embedded theories then have, provide conditionalisations for the validity of these theories. And claims of uniqueness (for the )defmition ( of certain tenns, e.g. for 'plane' by protophysical devices) can likewise be substantiated by explicitly referring to their dependence on boundary-conditions which do no longer lie outside the theory. Saarbri1cken
NOTES 1 Cf. the collection of essays (by Chomsky, Goodman, Putnam) presented at the Symposium on Innate Ideas in Boston, Dec. 1966, published in Synthese 17 (1967); or the paragraph on Innate Ideas in the last chapter of J. J. Katz, The Philosophy of Language, New York-London 1966. 2 C. F. v. Weizsacker in his address, p. 137f, referred to Konrad Lorenz as a representative of an erroneous adaptation of Kantian arguments to findings of a special natural science, here: ethology. 3 Cf. R. Reichenbach, Experience and Prediction, Chicago-London 1938, esp. Chapter I, and compare with the context of Leibniz's tem1S as expounded e.g. in H. Hermes, 'Die ars inveniendi und die ars iudicandi' [Ars inveniendi and ars iudicandi'] , Studia Leibnitiana Supp!. III, Wiesbaden 1969. 4 A term used by W. James for his version of pragmatism, which is exactly in line with the claim just made, cf. the collection of essays in The Philosophy of William James (ed. W. R. Corti), Hamburg 1976.
188
KUNO LORENZ
5 Of course, the radical empiricism of W. James may be included, too, since this issue can be dealt with quite independently from the dispute between J ames and Peirce on the meaning of the term 'pragmatism'. Cf. for support e.g. Peirce's argumentation against first intuitions to secure cognition in: 'Questions Concerning Certain Faculties Claimed for Man', C. S. Peirce. Collected Papers 1- VI (ed. Ch. Hartshorne and P. Weiss), Cambridge, Mass., 1931-35, 5.213ff. 6 Cf. e.g. several of the papers presented at the Loccumer Colloquium 2, from Oct. 71 (ed. U. Gerber), in: Hermeneutik als Kriterium fiir Wissenschaftlichkeit? Der Standort der Hermeneutik im gegenwiirtigen Wissenschaftskanon [Hermeneutics as a Criterion for Scientific Thought? The Position of Hermeneutics in Current Science] , Loccum 1972. 7 This derives from the fact that Kant never disputes the reality of knowledge, Le. of Newtonian physics, but tries to clarify the conditions of its possibility; cf. the relevant exposition in the last chapter (§ 15) of J . Mittelstr~, Neuzeit und Aufklarung. Studien zur Entstehung der neuzeitlichen Wissenschaft und Philosophie [Modern Age and Enlightenment. Studies in the Development of Modem Science and Philosophy] , BerlinNew York 1970. 8 Cf. the concept of a general theory of verbal communication as >protosociology< in the postscript to the new edition of J. Habermas, Erkenntnis und Interesse [Knowledge and Human Interests], Frankfurt 21973; a related though competing approach in the relevant parts of P. Lorenzen and O. Schwemmer, Konstruktive Logik, Ethik und Wissenschaftstheorie [Constructive Logic, Ethics and Philosophy of Science] , MannheimLondon-Zurich 1973, esp. III. 4 (Theory of practical knowledge). 9 Cf. K.-Q. Apel's extensive discussion on 'Das Apriori der Kommunikationsgemeinschaft' ['The Aprion of the Communicative Community'] , vol. II of his Transformation der Philosophie [Transformation of Philosophy], Frankfurt/Main 1973; further esp. R. Bubner, 'Uber die wissenschaftstheoretische Rolle der Hermeneutik' ['On the Function of Hermeneutics for the Philosophy of Science'], revised version of a paper at the Loccumer Colloquium 2 (note 6), in: Dialektik und Wissenschaft [Dialectics and Science] , Frankfurt 1973. 10 Most prominent the collection of essays by J. Habermas, Zur Rekonstrnktion des historischen Materialismus [On the Reconstrnction ofHistorical Materialism] , Frankfurt 1976. 11 H. Wohlrapp, 'Analytischer versus konstruktiver Wissenschaftsbegriff ['Analytic versus Constructive Concept of Science'], Zeitschrift fiir Allgemeine Wissenschaftstheorie VI (1975). 12 The latest account in my introduction to the reprint of R. Gatschenberger, Zeichen, die Fundamente des Wissens [Signs, the Fundaments of Knowledge], Stuttgart 1977; at the same place attempts to relate this approach with ideas of the symbolic interactionism as developped by G. H. Mead and of the genetic epistemology by J. Piaget. 13 This elementary level I understand as the same C. F. von Weizsacker refers toin his address (p. 136) where he speaks of a level of conceptual representation of atomic actions with no difference of judgment and concept yet, since no singular terms occur. As this paper tries to justify, I do not agree with the conclusion von Weizsacker draws when turning sides with the empirist's predelection of the singular over the general and declaring a >predicate feature universals< in his essay on 'Individuals' (London 1959), at least with respect to their general aspect; Straw-
THE CONCEPT OF SCIENCE
189
son forgets to include into his presentation theiX singular aspect as something on a par with the general aspect of preactions. Under the headline 'Property and Substance', being the terms for repeatable and nonrepeatable entities respectively, the same issue is at stake when R. M. Rorty in his paper on 'The Subjectivist Principle and the Linguistic Turn' (in G. L. Kline, ed., Alfred North Whitehead. Essays on his Philosophy, Englewood Cliffs, N.J. 1963) discusses and refutes - attempts of A. N. Whitehead to evade wellknown epistemological dilemmas deriving from the singular-general dichotomy, if this dichotomy is correlated in a straightforward way, i.e. without using linguistic analysis, with the body-mind dualism. 15 Representative of the treatment in the last Chapter of W. V. O. Quine, Word and Object, Cambridge, Mass. 1960, §56 (Semantic Ascent). 16 Cf. the sophisticated treatment of the empirical core (= empirical content) of a theory via Sneed's criteria of theoreticity as expounded e.g. in W. Stegmiiller, Probleme und Resultate der Wissenschaftstheorie und Analytischen Philosophie [Problems and Results of the Philosophy of Science and Analytical Philosophy] , vol. II, 2. Halbband, Chapter VIII, Heidelberg-New York 1973. 17 Cf. P. F. Strawson, op. cit. 18 Cf. K. Lorenz, 'On the Relation between the Partition of a Whole in to Parts and the Attribution of Properties to an Object', Studia logica 36 (1977). 19 Cf. the discussion of the interrelation between explanation and induction in.. C. G. Hempel, Aspekte wissenschaftlicher Erkliirung, Berlin-New York 1977 (German translation of a revised version of the last Chapter in C. G. Hempel, Aspects of Scientific Explanation and Other Essays in the Philosophy of Science , New York 1965). 20 Cf. for comparison the related remarks on the difference between descriptive and explanatory adequacy of a theory, here: of linguistics, in N. Chomsky, Aspects of the Theory ofSyntax, Cambridge, Mass. 1965, chap. I (Methodological Preliminaries). 21 On the alleged interdependence of a constructive and a normative approach to science cf. Zum normativen Fundament der Wissenschaft [On the Normative Foundation of Science] (ed. by F. Kambartel and J. Mittelstra,6), Frankfurt 1973, esp. the essays by J. Mittelstra,6, P. Janich and O. Schwemmer. The independence of 'is' and 'ought' is usually taken for granted - formalized as non-validity of ~! A
190
KUNO LORENZ
mon-sense-experience (relative to some natural language system) get discussed by W. V. O. Quine, 'Grades of Theoreticity', in: Exu:erience and Theory (L. Foster and J. W. Swanson, eds.) Cambridge, Mass. 1970. 2S Cf. his arguments in Against Method (London 1975) using examples from the history of science and of myths for the impossibility to achieve a unique (if not true) world representation. 26 Cf. Russell's Logical Atomism, ed. and with an introduction by D. Pears, London 1972.
RUDIGER BUBNER
TRANSCENDENT ALISM AND PROTOSCIENCE Comment on Lorenz
In what follows, I shall restrict myself to remarks with regard to two different, although related claims of Lorenz. I am not going to discuss all the allusions and implications of his paper. It is only the critical analysis ofKantian transcendentalism and the concept of a unified science I shall be concerned with. Equally, I am not interested right from the start to dispute the bipartite scheme of 'ars judicandi' and 'ars inveniendi' which selVes as the background for everything Lorenz has to say in the paper. In the course of my argument, however, I shall have to question the legitimacy of this scheme which echoes the old subject-object-dualism all through the terminology of its exposition (person-oriented, matter-oriented e.g.). It will turn out - at least I hope it will - that such a dualistic scheme is inadequate for rendering or criticizing the Kantian position of transcendental philosophy. Furthennore it will become clear - or, again, I hope it will - that the claim of the 'Konstrnktive Wissenschaftstheorie'. to be the successor to transcendentalism is ill-grounded. In treating this school, which Gennan provincialism has labelled 'Erlanger Schule', I shall make use of the writings l of Janich, Lorenzen and Mittelstrass besides the paper at hand. Lorenz says that Kant's transcendental idealism tends to treat "problems of justification as if they were problems of constitution, i.e. as if they concerned investigations into the conditions of possible experience" (p, 180). He criticizes this endeavour as false (p. 179), because it mingles together two different spheres: It confounds questions of constitution with questions of justification. What is needed, therefore, is "a clarifiGation of the interrelation between answering 'on what there is' and answering 'on what is true' " (p. 181). To my understanding Lorenz does not do very much to clarify the interrelation. Instead he gives us some hints about a fictional elementary language based on 'dealings' - a fairy-tale which I find hard to believe. Now it is exactly the problem of the interrelation at issue which stimulated transcendental philosophy. What do we have to know about the objects of our knowledge in order to know something definite about them before we come to know something defmite about them? This paradoxical knowledge a priori in relation to knowledge a posteriori was the offspring of Kantian philosophy. It does not, therefore, help us really to corne to grips with Kant's 191 P. Bieri, R.-P. Horstmann, and L. KrUger (eds.), Transcendental Arguments and Science, 191-195. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
192
RUDIGER BUBNER
problem, if we lay emphasis upon the strict difference between the theoretical presentation of objects of knowledge and the way of making them into objects of knowledge, which is called research. (Darstellung vs. Forschung). Even those who openly declare a disinterest in doxographical truth about an historical author are likely to miss an essential element of contemporary philosophizing along Kantian lines, by neglecting the interrelation between the two spheres. Here many divergent schools such as Popper's Criticism, Hermeneutics or the Kuhn-Feyerabend Theory of Science are on common ground. Why did Kant after all consider it a matter of utmost importance to bridge the gap between presentation and research or - to use the more modern terminology of Lorenz's - between the language and the world? The answer is well known. Kant was the first to disbelieve in the sense of a clear-cut separation of subject and object, language and world, presentation and discovery. A superhun1an being such as God is necessary to look neutrally at both spheres as separated. Hume's scepticism can be understood as the resignation vis-a.-vis the impossibility of such a standpoint for human understanding. The way out of this dilemma, which alone seemed reasonable, was to conceive of the subject as directed toward the object or vice versa. The function of presentation is to present the results of research and research is meant to result in theories about the world. The questions 'on what there is ' and 'on what is true' are no longer to be distinguished in an innocent way. What there is, is that which is possibly true and truth is that part of what there is which has been found to be the case. There is nothing new in this picture. Only one problem remains open: the transition from possible truth to actual truth. Kant's transcendentalism has delibera.tely left this to the concrete scientific activity. It is a contingent matter of procedure hic et nunc and therefore by defmition not something transcendental reflection is concerned with. Again, I cannot see how the separation of research and presentation suggested by Lorenz may bring us one step further with respect to this point. For the difference between research in general as an activity to discover truth and the truth discovered by that activity is not to be identified with the difference between 'what there is' and 'what is true'. Instead the difference lies on one of the two sides, where we have to distinguish between ars inveniendi and its concrete results. The problem that transcendental philosophy is only talking about the conditions of establishing scientific activity and not about the concrete results may be expressed in terms borrowed from Wittgenstein. Language offers us a
TRANSCENDENT ALISM AND PROTOSCIENCE
193
logical picture of the world consisting in!J1e whole of those propositions which are meaningful, i.e. possibly true. It is left to actual verification or falsification by empirical procedures of research to single out those propositions which are ture. The latter task can only be fulfilled within the realm opened up by language (logischer Raum). While this may be more or less accepted, much controversy has been raised by the basis of Kant's view. The so-called transcendental deduction of the catagories as the forms of our understanding of the world corresponds to the denial of any human knowledge of the things in themselves, i.e. the world as it is without any regard to human knowledge. The question that arises here is then, why are we obliged to look at the world only as a possible object of our knowledge? In other words, why do theoretical presentation and issues of research (to use Lorenz' terms) have to be combined? That is, finally, why is there no way of looking scientifically at the world other than through the constitution of a whole range of objects of discovery? As far as I can see, these questions can be sensibly answered only if the impossibility of an alternative can be demonstrated. This demonstration is most effectively carried out by realizing the fact that even the consideration of whether there is an alternative and what it might look like, has to make use of the very forms of understanding the legitimation of which is in question. Elsewhere I have called this structure 'self-referentiality'2. Others have put the same problem in a much more precise manner. Since this is not the topic at issue here, I do not want to go into detail. Whatever may be the device of a transcendental deduction, the job it is supposed to do is to determine the interrelation of the two spheres Lorenz takes pains to look at independently. That is to say, we can dismiss a good deal of the criticism of Kant implied in Lorenz's paper. It would have been legitimate, if not very original to cast doubt upon the basis of Kant's transcendental philosophy, but it seems to be a misunderstanding to blame him for confounding two separate spheres which he did not think to be able to be treated separately. Let us tum to the program Lorenz himself favours, since it shares important motives with Kant's own theory while avoiding the difficulties mentioned. If I am not mistaken the constructivistic approach to science tries to circumvent the difference between presentation and research by providing a rational construction of the foundations of every science. Here all the terms relevant to the special science in question are 'introduced,' as they say; that is, explained, defmed and made legitin1ate by the n1ethodical order of their introduction. One term follows the other, leaps and lacks are prohibited. It is
194
RUDIGER BUBNER
taken for granted that the methodical construction which precedes the very science is valid as the foundation needed because it (a) precedes the science and (b) is rationally introduced. The supposed rationality depends simply on the methodical order of the introduciton of relevant terms. Such a foundationlaying part is to be constructed for every science. It is called the protoscience (proto-physics, proto-sociology etc.). Now I have some doubts with regard to that program. In order to lay the ground for a given science in the sense of its proto-empirical part, there must already be some knowledge at our disposal telling us what kind of science we are dealing with, how the practical research works, which terms come in as relevant, what sort of technical operations can and must be carried out, and so on. This foregoing knowledge is by no means a priori and transcendental in the sense of holding good for all and every case at all and every time. A lot depends upon the actual status of that science, its historical place in scientific development, the pragmatic requirements of its realization etc. It may even be the case that in establishing the proto-part of a science we have to tacitly make use of some laws which can only be found by that science the procedure of which we are attempting to make possible. If this is the case, we are involved in a circle. Furthermore, one may ask, if the methodical order as such is all and everything we need in order to meet the adequate standards of rationality. We are told that we should proceed step by step without any jumps and gaps. Don't we have more specific criteria in view of any special science we wish to establish? It is not accidental and part of our pragmatic situation, but constitutive for the proto-scientific undertaking itself, that we always know more than we can possibly introduce by method. The framework for the methodical procedure of introducing the characteristic tem1S of a science must be given. All these questions are questions with regard to the conditions of constructing the proto-part which provides us with the actual conditions of a science. Kant faced a similar problem when trying to lay the groundwork of the special sciences. He was of the opinion that beyond the generality of transcendental analysis a foundation for special sciences is necessary which covers conceptually the field of just those sciences. The task of laying the foundation cannot be fulfilled by the science itself which operates on the basis of that foundation. The task is a philosophical one without being transcendental in the strict sense. Kant called upon pure reason in order to solve the problem a certain rehabilitation of metaphysics by its most severe critic. For the role of metaphysics, which is referred to here, is neither the traditional one that Kant's transcendental critique had destroyed, nor the accomplished doctrine
TRANSCENDENTALISM AND PROTOSCIENCE
195
Kant promised to build on the soil prepa~,ed by critical philosophy, once this philosophy had fulfilled its primordial job': Pure reason must play an independent role in an effort to establish the foundations of special sciences. Kant gave only one example with his 'Metaphysische AnfangsgrUnde der Naturwissenschaft', where he tried to develop concepts with en1pirical content through reason alone, such as matter, motion and force which are essential for physics. One might question, whether the traditional tenn n1etaphysics is appropriate for the non-empirical foundation of empirical sciences. It is not very clear how many empirical elements are necessarily inherent in those concepts of pure reason and if the four concepts offered by Kant are the only ones relevant. There remain intricate problems with regard to the fourth moment which Kant, following Lambert, has called 'phenomenology' (Erscheinungslehre).3 Moreover, modern philosophy of science would take more seriously th'ose contingent historical factors which characterize the institutional unity of theory and research or presentation and practice that makes up every special science. Anyway, what the constructivists have in mind corresponds to this particular part of Kantian philosophy. They want to build a whole sequence of terms relevant for a particular science starting from the most elementary level of everyday knowledge. They attribute a nonnative status to this rational construction which they consider to be an adequate reformulation of old-fashioned transcendentalism. It seems, however, revealing that what they call proto-physics used to bear the name of metaphysics. Universitat Frankfurt NOTES 1 Cf. P. Lorenzen, 'Wie ist die Objektivifcit der Physik moglich?' ['How is the Objectivity of Physics Possible?'] in Methodisches Denken [Methodological Thinking], Frankfurt 1969; P. Janich, in Zum normativen Fundament der Wissenschaft [On the Normative Foundation of Science] (ed. by F. Kambartel and J. Mittelstra.8 (eds.)), Frankfurt 1973. 2 'Kant, Transcendental Arguments and the Problem of Deduction', Rev. of Metaph. 28 (1975). 3 Cf. J. H. Larnbert, Neues Organon, 1764, vol. II; Kant's letters to Lambert froni 2.9.1770 and to Herz from 21.2.1772.
RICHARD M. BURIAN
SELLARSIAN REALISM AND CONCEPTUAL CHANGE IN SCIENCE*
In his contribution, Prof. Rorty suggested that transcendental arguments are only useful as a means of providing self-referential ad hominem arguments against particular philosophical views, showing that those views fail to make good sense of our knowledge, such as it is, and of our dealings with the world. I do not know if Rorty is correct, for I still do not understand how to delimit transcendental arguments properly. But my argument fits his description. I will offer a self-referential ad hominem argument against the theory of (scientific) truth offered by Wilfrid Sellars and largely shared by Jay Rosenberg. In spite of its specific concern with Sellars, the argument is of general interest. It has two principal strands. One, picked up by Prof. Parsons in his cornments, shows that Quine's doctrine of indeterminacy of translation is incompatible with the most interesting forms of what Sellars, Putnam and Rosenberg call Peircean realism. During the last two years I have been persuaded, rather against my will, that the indeterminacy of translation cannot be evaded. Accordingly, I hold that Peircean articulations of scientific realism cannot be adequately defended. The other major strand of the argument concerns the correction of fact-stating claims forced on us by theory replacement and by conceptual change in science. Such correction undermines the vestiges of a 'growth by accretion' model of the development of science which are still to be found in Sellars's and Rosenberg's writings. In its general thrust, my argument supports Putnam's contention that all viable articulations of scientific realism will land us in some form of what he calls "internal realism." 1
It will be useful to begin by discussing a position which nright be called transcendental or metaphysical realism. Very roughly, this is the view that we must distinguish between two worlds, an (ineffable) world of things-in-themselves, or noumena, and a world of things-as-they-appear, or phenomena; that the Truth (capital 'T') about things depends on how it stands with noumena; and that our investigations of the world can yield no more than the truth (small 't') about phenomena. A metaphysical realist holds that no matter how
197 P. Bieri, R.r;P. Horst 11U1 nn, and L. Kruger (eds.), Transcendental Arguments and Science, 197-225. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
198
RICHARD M. BURIAN
ideal an investigation we pursue, no ma~ter what epistemological and methodological perfection we allow our theories, no matter whether they save all appearances and are in all repects empirically adequate, the most they can give us is truth, not Truth. In other words borrowing an illustration from Hilary Putnam, 2 the metaphysical realist claims that the results of an absolutely ideal physics, biology, and psychology could be False because, for exanlple, in spite of everything ideal science reveals, we could all be (noumenal) brains in a (noumenal) vat rather than the more richly embodied persons we take ourselves to be. The difficulties of such a position are familiar. What is less familiar is the new twist which Putnam added to the argument - to wit that many, perhaps most contemporary accounts of our understanding of language, by inadvertently making truth and reference ineffable, make all truth and falsity as inconlprehensible as transcendental Truth and Falsity. To make this point, start by recalling a standard argument against transcendental realism. Suppose that sonle claim, say 'we in this room are not brains in a vat', survives all possible empirical tests and is embedded within a theory meeting whatever ideal epistemological and methodological standards you wish. To hold, nonetheless, that the claim might be noumenally False, we need an interpretation of our language which assigns entities to terms in such a way as to (1) preserve the empirical adequacy of our claim in the face of all tests and (2) nlake it False. Thus at least one term - I shall pick on the term 'vat' - must be satisfied by noumenal and not (or not just) phenomenal entities in order that the claim 'N is a brain in a vat' come out True for at least one of us. Now on any empirically adequate interpretation of English, the term 'vat' is satisfied by familiar entities - detectable by standard common-sense and quasi-scientific tecnhiques. These entities are the kind of thing which nlanufacturers of beer and of pickles employ in processing their products. Any empirically adequate interpretation of the English language will not allow the term 'vat' in any of its standard uses to refer to anything in this room. I should emphasize that I am not denying the possibility of error. We could be wrong about there being no vats here. What I am denying is that if all possible tests - including, for example, others' tests of the hypothesis that we are suffering mass delusion - turn out favorably, that then we know what to make of the claim under discussion. We should ask any realist who claims that even in the ideal case we could still be brains in a vat how he understands, or can bring us to understand, his use of the term 'vat'. In the end, we cannot make sense of his clainl. Now Putnam argues that anyone who accepts a Tarski-style theory of truth
SELLARSIAN REALISM AND CONCEPTUAL CHANGE
199
and uses it to construct a theory of what it is to understand language is no better off than the transcendental realist; for such a view inadvertently makes all truth and falsity as ineffable and incomprehensible as I have just argued transcendental Truth and Falsity are. As he puts it in 'Reference and Understanding.'3 [A]ssume, as is plausible, that [what it is] to understand a statement is to be explained as knowing its truth conditions. If truth is correspondence to reality, it would seem as if knowledge of what the correspondence is is presupposed by knowledge that such and such a statement stands in the relation in question to anything or does not stand in the relation in question to anything. And if understanding the statement is equated with knowing what it is for it to be the case that it stands or does not stand in the relation in question to appropriate entities, then knowledge of what the correspondence is is presupposed in the understanding of every statement. But in what could this knowledge which does not consist in the acceptance of any statements, because it is prior to the understanding of all statements - consist?
Thus a truth-conditional theory of understanding a (first) language - a theory which makes truth prior to meaning, a theory which makes truth the central c6ncept in a theory of understanding - such a theory, given a realist interpretation of truth, brings ineffability close to home. It makes knowledge of the truth conditions for non-compound sentences unintelligible. In consequence, it makes our understanding of ordinary sentences unintelligible. This, in turn, makes our understanding of terms like 'vat', as ordinarily used, as mysterious as the transcendental realists' understanding of their noumenal vat. Transcendental realism can give no clear sense to claims about the noumenal world. Truth-conditional theories of understanding can give no clear sense to ordinary claims. If the realist is to escape the absurdity of denying that we know (at least sometimes) what we are saying, he must offer an alternative theory of understanding. And if he expects truth to be of any interest to us, he must escape ineffability. Before turning to Sellars's Peircean realism, I would like to connect this discussion very briefly with the second topic of my title - conceptual change in science. It is clear that a transcendental realist, concerned with Truth, should see the conceptual changes brought about, e.g., in so-called scientific revolutions as irrational, non-progressive episodes, as 'one damn thing after another.' So, too, should anyone who makes truth conditions and reference ineffable, whether willingly or inadvertently. In either case, there is no way to decide between revolutionary competitors with respect to Truth - they are all, so far as we can tell, equally distant from the noumenal (or ineffable) Truth. There may be other reasons for preferring one theory to another -
200
RICHARD M. BURIAN
reasons of simplicity, past empirical ad~quacy, convenience or taste for example - but they cannot be judged on grounds of Truth, and there is no external viewpoint from which, or ordering principle by means of which, to judge them. I believe that a fair number of the more extreme recent treatments of revolutions as inexplicable episodes, not objectively evaluable, can be understood on the basis of this account of one way in which revolutions could seem to be thoroughly irrational. II
This brief discussion of transcendental realism will bring out the point and promise of Peircean realism. Peirce held that truth is what - given unlimited time and resources and employing ideal methodology - the community of scientists is fated to believe. This approach promises not to make truth ineffable. Yet it can clearly accommodate the difficulty of knowing the truth and it is compatible with Peirce's sound fallibalism: none of our claims is sacrosanct or uncorrectable. Furthermore, these gains are purchased while honoring the Cartesian insight that there is one world, one truth. Peircean realism thus promises to allow one to make sense of doctrines (whether they are true or false) according to which a sound methodology and a sound epistemology bring us closer to the truth and to knowledge of reality. Yet I am going to argue that, in the end, the account of truth incorporated in current versions of Peircean realism - especially in the most interesting of these, namely that of Wilfrid Sellars - is not much better off than that I attacked in discussing transcendental realism. Sellarsian realism, I will argue, comes to grief over the indetern1inacy of translation and over its inability to offer an adequate account of the correction of a theory by its more adequate successors. I believe, but I do not show in tIns paper; that the fundamental difficulty here is that Sellars's inferential sen1antics is unable to capture the correct referential core of Tarski's theory of truth. In any case, I show that although Sellarsian realism recognizes the need to make truth immanent rather than transcendent, it fails, in the end, to make it immanent enough. III
It will take a few pages to layout an account of Sellars's version of Peircean realism, for it is an extremely rich and complex doctrine. It is by no means purely Peircean - for example it incorporates a neo-Tractarian picture theory
SELLARS IAN REALISM AND CONCEPTUAL CHANGE
201
of fact-stating language and a highly distinc,tive theory of truth which, so far as I know, cannot be found in Peirce. Sellars holds that each community of language users operates within a linguistic framework which has a (developing) conceptual structure. As of a given moment, this structure may be treated, in first approximation, as a fIXed, clear, closed analyzable structure determined by (1) the linguistic roles it contains and (2) the norms governing the semantic assertibility of tokens playing those roles. It is of considerable importance to recognize that, according to Sellars, different communities - even communities speaking different languages - may employ the same conceptual structure. Conceptual structures are interlinguistic; they are not tied to the resources of a single language. Let me explain Sellars's technical term 'semantic assertibility' very briefly. Informally put, its purpose is to capture one dimension of the concept of truth, namely truth-with-respect-to-a-fran1ework. It is to be distinguished from warranted assertibility, a weaker notion. A claim is warrantedly assertible, roughly, when, according to the standards of our conceptual structure we are justified in accepting it on the basis of the available evidence. (For what Quine calls stin1ulus analytic claims, the evidence required may be nil. For everyday perceptual claims, normal perceivers and speakers are warranted in their assertions after 'just looking'. In more complex cases, speakers may not be warranted until they have done considerable calculation and testing.) A claim is, semantically assertible (hereafter 'SA') when it is not only warranted, but also correct. In general, the norms governing semantic assertibility do not suffice to determine the semantic assertibility of particular claims on particular occasions. (SM IV, 26; V, Iff.; V, 8ff. and V, 54ff.)4 For example, whether the claim 'There is a cow in the next room' is SA depends not only on the norms governing assertibility, but also on the way the world is. If one of us were to run next door and see a cow, he would probably take 'There is a cow in the next room' to be SA. (He might be warranted but wrong if he were fooled by a clever statue.) We might be warranted in accepting his claim by his seemingly honest avowal, 'My God, there is a cow next door'; we surely would be warranted in so doing if we looked for ourselves and saw what seemed, inescapably, to be a cow. Thus Sellars can use the distinction between warranted and semantic assertibility to escape Putnam's arguments against truthconditional theories of understanding. To learn a language it is enough to learn when its claims are warranted - Le., when to make and accept them and when to deny and reject them. One need not know the conditions of semantic assertibility - of truth-within-the-framework - in order tounderstand its claims.
202
RICHARD M. BURIAN
Our acceptance of a particular claim and, if tIns is different, our acceptance of the claim that it is SA - these may be correct or incorrect. But when a claim is SA, it is a truth of the framework in question. In general, the truth of the claims of the framework within which we operate is no more and no less than their semantic assertibility. When all the common-sense criteria for there being a cow next door are met, the common-sense claim, 'There is a cow in the next room', is SA, Le., is true with respect to the common-sense framework. IV
Now Sellars correctly emphasizes that it is a gross idealization to treat the norms governing semantic assertibility as closed and flXed. They change from context to context - and they change in especially interesting ways with the growth of knowledge. (For instance, should the biologist consider 'There is a cow in the next room' to be true if the entity next door meets all the common-sense criteria for being a cow but fails to contain any DNA?) As knowledge grows we alter not only our beliefs, but our concepts, our criteria, our ontologies. An enormously complex social process underlies changes in accepted scientific beliefs, changes in the concepts employed by the scientists of a certain discipline, changes in their criteria and norms governing the acceptance of claims. 5 Most philosophers who study the development of science, including Sellars, give these social processes short shrift. I believe that this is a fundamental nristake. Putnam, for one, has argued that the distinction between meaning and belief - a distinction which will playa central role in my argument and which, for as long as I can remember, Quine has argued is a distinction without a difference - can only be properly understood by recognizing its social basis. 6 In this discussion I shall have to abstract wholly from the social dimension of meaning; accordingly I shall not try to provide a developed account of the distinction between change of meaning and change of belief. Like most philosophers who deal with these topics, Sellars uses this distinction in developing his account of conceptual change. He draws on the contrast between fleshing out and correcting the claims we accept within a framework and replacing the framework with a better one. Unlike Feyerabend, 7 Sellars explicitly denies that a change of framework can be equated in any simple way with a change of theory, even when the theory in question is a 'universal, highest-level' theory. For change of framework, Sellars requires (1) that there be conceptual continuity between frameworks, and (2) that the
SELLARSIAN REALISM AND CONCEPTUAL CHANGE
203
new framework yield SA counterpart claims whenever the old one produced SA factual claims. Accordingly, he holds, (3) that until a universal theory can do these things it is, and should be, held on an as-if basis, treated in part as an instrument for calculating truths in our framework, in part as a potential successor to that framework. My immediate concern is to explore what a framework is and what the continuities between frameworks come to. Sellars's way in to these topics is unsuitable for present purposes, for it proceeds through his novel and difficult inferential (as opposed to referential) theory of semantics. A rigorous treatment of his doctrines cannot bypass foundational questions in semantics - but we can. For the central points can be made by mobilizing widely shared intuitions about a familiar example. Consider the difference between contrasting isoceles and scalene triangles on the one hand, and Euclidean and non-Euclidean (say hyperbolic) triangles on the other. In a given geometrical system, isoceles and scalene triangles both fall under a single concept of triangularity - Le. they are simply different kinds of triangles. In contrast, Euclidean and hyperbolic triangles are entities satisfying different (though related) concepts of triangulan·ty. If this difference between the two contrasts is correct, we must distinguish between different concepts of triangularity, playing similar but not identical roles in different geometrical systems - e.g., Euclidean, hyperbolic and spherical concepts of triangularity - and different ways of fulfilling one of these concepts - e.g., by being an equilateral, isoceles or scalene triangle of whichever kind. Now as I have already said, Sellars treats conceptual structures as being composed of linguistic (or, better, interlinguistic) roles plus norms governing the assertibility of claims. Thus, to say that there are different concepts or kinds of triangularity is only to say that there are conceptual structures other than ours which include assertions generically like and specifically unlike those of our assertions with the form 'x is triangular'. In our example, this claim may be supported by reference to the contrasts among triangles, rectangles, circles, and so on, in all of the geometries in question and by the specific axiomatizable similarities and differences between all Euclidean triangles, all hyperbolic triangles, and so forth. It is important to recognize that a claim like Euclidean triangularity and Riemannian triangularity are varieties of triangularity
[arrieS with it no commitment to the e~i~t~~~~_Of_~_Plato~c_a_bstract entity,
204
RICHARD M. BURIAN
triangularity, or to several such entities. (Cf. CC.) Rather, it carries with it only the commitment that terms like 'triangular', appearing in oratio recta in the context of a Riemannian geometry, and terms like 'dreieckig', appearing in a German textbook of Euclidean geometry, each playa role which is generically similar to that performed by 'triangular' in informal English, and yet that these roles are specifically different. (The easiest way to enforce recognition of the difference is this: when a single set of 'coordinating definitions' is used in a context which allows sufficiently fme-grained measurement, no entity of which the Riemannian 'triangular' is SA is one of which the Euclidean 'dreieckig' is SA.) This informal treatment of a familiar example should motivate the following claims: (1) It is often possible to recognize counterpart concepts. (2) These typically involve holistic and systematic differences between the roles played by the counterparts (or, if you prefer, between the roles played, in their separate contexts, by such linguistic tokens as ' ... is triangular'). (3) The concepts (the roles played by the tokens of a given kind) are recognized as counterparts in virtue of a large range of common (or similar) systematic relations to other counterpart concepts - cf. the relations among 'point', 'line', 'vertex', 'plane', 'closed figure', 'triangle', 'rectangle', and so on, common to all the geometries mentioned. (4) In comparing counterpart concepts, one must typically compare whole frameworks or structures of concepts. Finally, (5) These often exhibit limit relations of an interesting sort. (Spherical and hyperbolic geometries become Euclidean 'in the limit' as their intdnsic curvature goes to zero.) You should now be convinced that, whatever their precise philosophical analysis, there are rather nice cases which seem to illustrate the distinction between change ofbeliefand change ofconcept as Sellars draws it. We change our belief that a certain figure is a Euclidean triangle when we find that its 0 angle sum is 179 , and (either thereby, or independently) determine that its sides are curved. We change our concept of triangularity and our geometrical framework when, having heretofore employed Euclidean concepts, we apply the Riemannian concept of triangularity to an object with a measured angle 0 sum of 179 • In these geometrical cases, at least, no transcendental point of view is required for comparison of alternative frameworks: as axiomatic systems, our alternative geometries may (in fact) be modeled in each other, while as physical geometries they may be treated as the dialectical working our, under the pressure of empirical findings plus consideration of conceptual
SELLARSIAN REALISM AND CONCEPTUAL CHANGE
205
alternatives, of a vague generic concept which applies to any closed, threesided, approximately rectilinear object. 8
v We now have the elements of Sellars's account of conceptual change in hand. One conceptual structure or framework replaces another. The successor framework contains a large number of counterpart concepts to those of its predecessors. For all particular matter-of-fact claims of the predecessor frameworks which were SA in those frameworks - and perhaps for many claims which were not SA but which were warrantedly assertible and whose error escaped detection - there are SA counterpart claims, employing the new concepts, in the new framework. (Example: A land surveyor tells us that a plot of land is exactly rectangular. As a good naive Euclidean, I take it both that its boundaries are the shortest distances from vertex to neighboring vertex and that, extended indefinitely, the boundary lines would never meet. As more sophisticated spherical geometers, you accept the counterpart claim that the surveyor has laid out an approximate spherical rectangle, but that, depending on his techniques, either the sides, extended indefmitely, would meet, or the boundary does not provide the shortest distance from vertex to vertex. The truth of this counterpart claim entails the literal falsity of (a literal translation of?) my Euclidean claim, but it also entails its approximate truth. In detailed application to this case it even entails the impossibility of detecting any discrepancy from Euclidean expectations by n1eans of (relatively crude) surveying techniques applied in isolation to discrete plots of land and hence is compatible with the view that my Euclidean cla.im was SA. This model is intended not as an account of some unusual or isolated cases, but as the model of all conceptual change in science. Or, rather, it is the model of all philosophically justifiable conceptual change in science. It will become increasingly important for us to recognize that the model plays a normative role in Sellars's philosophy of science. VI
But before we go on to ask about the universal applicability of the model or about the desirability of enforcing its universal applicability - two tasks remain. The second will be to explore some of the strengths of Sellars's approach to conceptual change. The first, the task of the present section, is
206
RICHARD M. BURIAN
to layout three major strands of conunentary which govern the application of Sellars's model to cases. I proceed with these: (1) What are we to make of the semantic assertibility of fact-stating claims - e.g., of our surveyor's claim that a certain plot of land is rectangular? Sellars points out that a great variety of claims - e.g., ethical and semantical ones - are SA, and argues that the ground of the semantic assertibility of a claim depends on the kind of claim it is. In the case of fact-stating clain1s, Sellars employs a picture theory of language to distinguish between particular matter-of-fact claims which are correct (SA) and those which are incorrect (not SA). The key to understanding Sellars's picture theory is understanding the role of the terms which in context serve as primitive designating terms. In everyday English, such terms include names of individuals (e.g., 'Lorenz Kruger' used as the name of a particular person), but also phrases like 'that man' (used, in context, to pick out an individual of a familiar observable kind) or even such descriptive phrases as 'the woman over there with the martini'. 9 When we have straightened out all of the relevant contextual factors, a statement of fact is nothing more than a term whose job is designating an individual (e.g., the term 'Ms. Jones') appearing as qualified by a predicate (as when it appears in the sentence 'Ms. Jones is a philosopher') or appearing in relation to other terms (supposedly) designating other individuals ('Ms. Jones is next to Mr. Smith'). Terms appearing in such contexts form linguistic pictures in a manner determined by the 'rules of projection' of the language - i.e., they depict how, in the world, things might stand with individuals correlated with the designating terms. When used assertively (we cannot stop to analyze what that means) to make a claim about how things stand here and now, such pictures depict how things are supposed to stand with the entities in question. Whether or not the claim is SA depends on whether or not the linguistic picture is con-ect. Such correctness, which can only be asked about after taking into account the relevant linguistic and contextual rules of projection, is a matter of (a) whether or not the relevant criteria governing the use of the designating term are met,10 and (b) whether or not the things thus identified are qualified or related in the manner corresponding, according to the rules of projection, to the way their names are qualified or related in the sentential context. In short, Sellars treats correctness of picture in hie et nunc cases as a matter of physical correspondence between designating tokens and individuals and between sentential contexts and qualities or relations of individuals. If you will, this is the Tractarian picture theory without logical atomism or ultimacy, for what counts as a thing depends on the rules of projection, and
SELLARSIAN REALISM AND CONCEPTUAL CHANGE
207
these rules do not require things to be indivisible or the states of affairs into which they enter to be logically independent of each other. We now know, in principle, how to answer the question whether or not our surveyor's claim that a given plot of land is rectangular is true: ascertain what rules of projection he is using and check whether the appropriate individual (here, a certain demarcated section of land) stands as it is claimed it does (here, is shaped as it is claimed to be shaped). Later we will find difficulties with this answer. (2) Our model of a conceptual structure - an axiomatized or axiomatizable geometry - is, of course, far too neat. There are two principal reasons for this. (a) Terms have many uses. Correspondingly, conceptual similarity has many dimensions. In discussing conceptual similarity we have, so far, explicitly considered only similarities of formal structure. But other similarities are crucial. Consider the case (from note 9) ot the two phrases 'Ms. Jones' and 'the woman over there w'ith the nlartini'. If one were to judge from their surface structures, these phrases would have to count as yielding unrelated concepts. Yet, in context, they play the sanle, or extremely similar, roles. (This can be seen, by the way, from the fact that when I said, "The woman over there with a martini is a philosopher," the indirect discourse attribution, "He said that Ms. Jones is a philosopher," is correct.) This is so because among the many conceptual roles we require are those of standing for a (particular) individual. And that job was the essential one of my seeming descriptive phrase in this particular context. Such conceptual roles (and there are quite a nunlber of them, including, e.g., standing for an unspecified individual of a given kind) can be shared by tokens of quite unrelated surface grammars. Of particular interest among the roles terms play is that of the language-entry transition. Consider a slightly more serious example than the preceding ones, that of a well trained technician, taught to read cloud chamber photographs. Even though he has been taught little if any particle physics, even though the foundations of particle physics undergo major revolutions, his training can lead him to respond reliably to a track of a certain character with the phrase or annotation 'fast electron at (x, y)'. And the reliability of this response nleans that even though the theory of electrons were altered in the most drastic ways, so that the logical form of the term 'electron' were utterly changed, the notation 'fast electron at (x, y)' wbuld play the role of a reliable language entry transition, Le., a reliable linguistic correlate, arrived at noninferentially, of a track of a certain kind, connected (supposedly) with the effects of an electron in ways maintained by whichever theory one was working with. The reliable, intersubjective correlation with the track allows the
208
RICHARD M. BURIAN
community of physicists to come tC1,:agreement on the warranted assertibility of the technician's claim even in the face of radical disagreement over the nature and status of electrons. Because the inferential role of the term 'electron' affects the semantic assertability of the claims into which it enters, it does not follow that there is a single claim which the physicists agree to be true, but it does follow that this constant use of the term 'electron' provides a major ground for the judgment of conceptual continuity. (b) In real life, our theories and our frameworks are almost never coordinated with a rigorous calculus of the kind we philosophers have long cherished. The fact is that there are virtually no cases in which an axiomatic account of the specific differences between generically similar concepts can be constructed. This is true not only because we are ignorant, sloppy and inefficient, and not only because language en.try (and exit) roles do not correlate n~atly with the apparent logical form of the terms involved, but also because we are always in mediis rebus, working open-endedly with a dimly delimited range of alternative frameworks. Deductive closure occurs aln10st exclusively in the logicians' never-never land, yet without it there is no sure criterion for separating conceptually important similarities fron1 mere curiosities. In the end, our disparate intuitions about the conceptual affiliations of different terms, of different predicates, of different claims, rest on differences in program, in belief, in purpose, and cannot be resolved by straigl1tforward appeal to some axio"matic n1easure of similarity or to some scheme of embedding alternative systems within one another. It is only because, for many purposes, we can get by with a loosely specified framework, partially specified by a variety of analogical devices (cf. SR II, 178ff.), that we get along as well as we do. (3) This brings us to a third major group of considerations regarding change of fran1work: how is it justified, how is it constrained, how does the Peircean con1ponent enter? And here there are three major points to n1ake. The first is that Sellars considers the central purpose of science to be explanation, though not explanation as Hempel or Nagel conceive it. Ideally, to explain a range of phenomena is to describe it in a framework which, in application to those phenon1ena, yields no anomalies. 11 But we are always faced with an ocean of anon1alies, so explanation is typically redescription in such a way as to remove anomalies. Redescription may be relatively local (as when the description of hydrogen and oxygen as monatomic gases was altered in favor of a description according to which they are composed of diatomic molecules) or extremely global (as when the behavior of elementary particles was quantized). Indeed, global redescription requires revision of the laws
SELLARSIAN REALISM AND CONCEPTUAL CHANGE
209
employed, i.e., revision of the framework w,ithin which explanation occurs. 12 All such redescription guarantees that what one deduces from one's laws is not the original description of the phenomena to be explained, but a revised one. The Hempelian model of explanation is emphatically inadequate in such cases. Thus, in the most interesting cases, one deduces a counterpart description of the phenomena of concern in the new framework and not the original description, whether that belonged to an 'observation language' or to a prior theoretical language. It is in this way that the kinetic theory of heat explained the behaviors previously explained by phenomenological thermodynamics, that relativistic mechanics explained the behavior of free-falling bodies (and an immense amount more) and that transmission genetics and molecular genetics explained Mendelian heredity. Explanation is not systematization of experience, but (re)description of the world in a (more nearly) anomaly-free manner. The second point is that this account works only if one prohibits all cheapening of explanation. And this is accomplished by the following: (a) the framework within which an explanation takes place is required to be con1prehensive. That is, it must explain the success of all relevant predecessor frameworks and at least some of their failures. Or, better put, it must yield SA counterpart claims whenever its predecessors offered SA matter-of-fact claims and, in addition, some further SA claims. Thus what it explains directly is how things are; indirectly it explains how we did as well as we did while using the predecessor frameworks we used and some of the ways in which they led us astray. (b) A genuinely explanatory framework must, in the first instance, explain particular matters of fact and the behavior of individual objects. These are its primary jobs, not the derivation of the laws of an antecedent framework. A microtheoretical explanation of the empirical laws governing the behavior of observable entities is not a derivation of the old empirical laws from the theory but an explanation of "why observable things obey, to the extent that they do, these empirical laws." (LT. 121) In particular, such microexplanations of statistical laws as, e.g., an explanation of the behavior of typical samples of a certain chemically purified element in terms of the distribution of behaviors of different isotopes show that the explanation of laws is accomplished by the proper description of the behaviors of the different kinds of 'theoretical' individuals comprising a macroscopic object. In cases like this, one can show why refinement of the inductive generalizations and empirical laws of a predecessor framework was impossible within that framework - here because it was impossible to take
210
RICHARD M. BURIAN
account of differences within the nuclei of chemically identical atoms while using only chemical concepts and techniques. Such demonstrations show that the laws of the predecessor framework were, at best, approximations to the truth. With this constraint on justified conceptual change we encounter the second dimension of Sellars's concept of truth. With respect to a single framework, truth, i.e., semantic assertibllity, is absolute: its criterion is correctness of picture. But frameworks as wholes and specific counterpart claims in different frameworks can picture more or less adequately. When one framework is more adequate than another, i.e., when it contains SA counterparts for all SA claims of its predecessor and it can, in addition, show how some of the warranted claims of the predecessor fail to be SA, or when it can make picturing claims where the predecessor franlework could make none, then it is closer to the truth than its predecessor. Finally, these constraints on adequate explanation yield the regulative ideal of a perfect explanatory framework, what Sellars calls the Peircean conceptual structure, 'CSP'. In CSP warranted language entry transitions yield claims of the forms 'F(a) here, now', 'R(a,b) here, now' and so on only when properly qualified and interrelated individuals of the right categories, corresponding to 'a', 'b', ... are present. Furthermore, the rules of inferences of CSP are such that semantically assertible matter-of-fact claims, when transformed into other matter-of-fact claims, yield only semantically assertible claims. That is, CSP is an unreachable ideal franlework in which the only warranted matter-of-fact claims are SA and hence true. And conceptual changes in science are justified when they bring us closer to this ideal. VII
We have finally completed the exposition of Sellars's Peircean realism. Let us review the ways in which it promises to avoid the dilemmas of metaphysical realisnl before embarking on an argument to show that it has failed to do so. Of. central importance is the role of CSP as a means of evading ineffability while securing a unique truth. The difficulty we wish to get out of is caused by the seeming transcendence of truth. Peircean realism promises to remove the difficulty in two ways. First, it will avoid ineffability by analyzing truth as semantic assertibility. We know, so the line goes, precisely what it would take to achieve semantic assertibility in any framework, namely that its linguistic pictures tum out to be correct. If this line is to work, it is crucial that linguistic pictures be inter-
SELLARS IAN REALISM AND CONCEPTUAL CHANGE
211
preted as having a determinate composition, a composition which we know in employing our language. Second, Peircean realism replaces the ineffable realm of Dinge-an-sich with the epistemologically ideal Peircean framework - the supposedly determinate ideal end of the chain of improved explanatory frameworks - as the arbiter of truth, the arbiter of what really exists. Now if we know what it would take for our claims to be SA on the basis of our understanding of linguistic pictures, our understanding of those pictures must depend, in turn, on our understanding of the picturing correspondence. This is one point of the distinction between warranted assertibility and semantic assertibility. The former obtains, roughly, when the norms for asserting or accepting a claim have been met; it is thus a verificationist notion in the sense adumbrated by Bennett in his paper above. The latter obtains when a claim pictures correctly. Using this distinction, we can meet Putnam's strictures: language learning and the basic understanding of a language need not depend on knowledge of the substantive truth conditions of our claims: it depends only on learning to use claims in accordance with common usage. (In terms of the example of the surveyor, what we must teach our children is when to make and when to withhold the claim that a plot of land is rectangular. Only when the claim has been given a stable and definite use in this way can they - and we - intelligibly ask whether it corresponds to (pictures) a particular state of affairs or kind of state of affairs.) Warranted assertibility thus enables us to learn to use the claims of our language and to ask, intelligibly, which of our claims, thus understood, are SA. In answering this question, the Peircean framework and the doctrine of picturing are the keys to the attempt to avoid the Scylla of idealism and the Charibdis of ineffability. On the one hand, truth is not warranted assertibility: even the most warranted of claims may be wrong. Nor is truth-within-a-framework the sole arbiter of truth, for, on the other hand, the "Archimedian point outside the series of actual and possible beliefs" (SM, V, 75), required to make sense of the limit of the series of beliefs, is picturing. But picturing is not supposed to be a correspondence with transcendental entities, nor to involve transcendental entities in any way. It is the physical correspondence between those tokens which the rules of projection of our framework, in context, make into individual-designating tokens and the entities which, under the rules of projection, they correlate with. And this physical correspondence runs between the tokens as qualified in their sentential context (e.g., as placed to the left of a token of the shape 'is rectangular') and their image objects as qualified in the world (e.g., as being shaped in a specific way). Whether the correspondence obtains or not is a factual question. If it fails to obtain it
212
RICHARD M. BURIAN
should, in principle, be possible to detect the failure. And the corrections by means of which the failure is understood are embodied in improved frameworks and point toward esp. We cannot do without a conceptual structure - all entities of which we are aware and on which we (try to) operate are conceived of in some way or other. But there need be no Peircean community for us to comprehend the ideal of a esp, a conceptual structure in which all warranted language entry transitions yield semantically assertible claims and in which the inferences drawn from SA matter-of-fact statements never get 'out of synch' with the further matter·of·fact claims which the members of the Peircean community would spontaneously accept as warranted in the light of their experience. esp, by thus picturing the world, would serve as the arbiter of 'ideal truth' and of 'what really exists'. This apparatus can also account for the fact that many realists are tempted to slip into transcendentalism by claiming that an absolutely ideal theory could be false - and show the error of this claim. For whether or not a picture is correct is, in the end, a matter of brute fact, and no amount of perspicuity, perspicacity, and perseverance in following an ideal method can guarantee that one's framework will always yield correct pictures. Nonetheless, the temptation to thus downgrade ideality is, on Sellars's account, erroneous. For the standard of ideality is correctness of picture. (Here Rosenberg differs from Sellars. Recall that he holds that proto-correlation, his version of picturing, is epistemically idle.) A conceptual scheme which, in fact, yields incorrect pictures - such a conceptual scheme is not ideal and cannot count as esp. Thus although we cannot know any conceptual scheme to be ideal, we can show that all known conceptual schemes are far from it. The recognition that this is the best we can do is a part of the recognition of the difficulty of the problem of knowledge. VIII
Having shown the promise of Peircean realism, it is with a heavy heart that I set out to undermine it. But the truth is that the promise is an illusion, that it cannot be realized. For even though there is but one world, we are unable to escape ontological relativity and indeterminacy of reference in dealing with it. Even granted a Peircean framework (though this is granting too much), it cannot serve as the arbiter of truth and existence except in a Pickwickian, theory-relative sense. To suppose that the Peircean framework resolves questions of truth and existence in the full-blooded sense in which Sellars, for
SELLARSIAN REALISM AND CONCEPTUAL CHANGE
213
one, thinks it does is to make truth and existence transcendent, not immanent. I will come at these 'points from a variety of directions. IX
Consider first what is involved in change of framewor k. Our model is a change from a Euclidean to a hyperbolic geometry. Abstractly speaking, the model reflects the fact that different frameworks have different formal structures. But it is a point of model theory that, for cases as complex as those we are interested in, if there is one way of mapping the one formal structure onto (or into) the other, there are many. There is no uniquely correct way to carry out the change from one framework to another. Let us recall a familiar instance. There is no unique model or reconstruction of the natural numbers in set theory, not even within a single set theory. There is no fact of the matter, except relative to a particular reconstruction, as to whether the number two (a set!) contains the number one (another set!) or no1. 13 There is no unique set-theoretical counterpart to the number two and no fact of the matter as to which of the infinitely many systems of counterparts for the natural numbers is correct. So, too, there is no uniquely correct way of representing Euclidean claims (e.g., 'this figure is a rectangle') within hyperbolic geometry. What the Euclidean claim comes to depends on our scheme of translation in just the way in which what the number two comes to depends on which set-theoretic reconstruction of the natural numbers we employ. Sellars recognizes the difficulty in the arithmetic case. Even leaving aside set-theoretical complications, Godel's incompleteness theorem entails that there is no limit to the series of 'more adequate' axiom systems for arithmetic, Le., that there is no Peircean framework for arithmetic. Nonetheless, Sellars wishes to hold fast in the non-mathematical sciences: "in the case of factual propositions, we are haunted by the ideal of the truth about the world." (8M, V, 55) Haunted. Just so. The existence of alternative translations between frameworks - and the lack of a factual basis for deciding between them - means that alternative systems of counterparts can be devised for the claims of predecessor frameworks. But when two frameworks differ in their formal structure and one is used (or thought) to correct the other, the situation is especially interesting. We may speak either of two ways. If we require translation to preserve truth, there is no general systen1 of (truth preserving) translation between the two frameworks, while if, with Quine, we only require that all translation schemes meet a common standard of formal adequacy while preserving stimulus
214
RICHARD M. BURIAN
synonymy, stimulus analyticity, and acceptability of occasion sentences, translation may fail to preserve truth. I shall speak of translation as preserving truth. In this case, when a new framework corrects the claims of a predecessor, any consistent scheme of mapping (not translating) between the terms of the two frameworks will map some SA claims of the old framework onto counterpart claims whose negations are SA in the new. All significant change of framework must result in loss as well as gain. For example, if we interpret (Newtonian) 'mass' as rest mass, then the Einsteinian counterpart of the SA Newtonian claim (1)
Momentum
=mass
X velocity
comes out false (Le., its negation is SA), though the counterpart of the SA Newtonian claim (2)
The mass of an object is invariant
comes out true (SA). If, on the other hand, we interpret 'mass' as relativistic mass, then (1) becomes true, (2) false. 14 Even when one seeks to maximize the preservation of semantic assertibility, such interpretations yield no unique counterpart for a sentence 'p' of the old framework except, at best, relative to a particular mapping between terms of the two frameworks, and even then there is no fact of the matter about which mapping best represents the claims of the old framework within the new. There are legitimate, equally valid, alternative systems of construal for the old claims within the new framework, none ofwhich preserve the semantic assertibility of all SA factual claims of the old framework. The choice of counterpart construal, like the choice of translation, can affect the reference assigned to terms as well as the content assigned to claims. 15 Thus the Newtonian claim 'mass is invariant' is construed quite differently when we take its counterpart to be 'relativistic mass is invariant' then when we take it to be 'rest mass is invariant'. Field argues most persuasively that, in spite of the fact that each interpretation of 'mass' makes some SA Newtonian claims have false counterparts, there is no better univocal interpretation of the Newtonian term 'mass' available within relativity theory and that there is no non-arbitrary ground to prefer one of these interpretations to the other. Conclusion: we cannot be forced by all the facts to agree about the meaning of factual claims of a predecessor framework. These claims lmay legitimately be parsed in a variety of ways, and when they are; Sellars will lhave to treat alternative construals as making different linguistic pictures out lof them. In cross-framework contexts, the content of a claim - the linguistic
SELLARSIAN REALISM AND CONCEPTUAL CHANGE
215
picture it makes - is relative to an interpretation "The lore of our fathers is a fabric of sentences ... black with fact and white with convention." 16 Even if we grant (as Quine rightly insists that we should not) that within a framework we can separate some pure black and pure white strands, when we change frameworks, the blackness of fact and the whiteness of convention are changed to gray.
x Let us look more closely at the difficulties raised by the problem of 'mass'. The problem sterns from the fact that the relativistic counterpart of (1)
Momentum = mass X velocity
comes out false (Le., its negation is SA) if 'mass' is interpreted as rest mass, while the counterpart of (2)
The nlass of an object is invariant
comes out false if 'mass' is interpreted as relativistic mass. There are three ways to evade this difficulty: 17 First, to deny that (1) and (2) are SA in the Newtonian framework; second, to abandon the principle that the counterpart, in a successor framework, of an SA claim must itself be SA; and, third, to use different counterparts of the term 'nlass' in different cases. Each of these options has unpalatable consequences. Option one is unacceptable if the notion of semantic assertibility is to be of any use. The whole point of that notion is to make truth immanent, to make good sense of truth-with-respect-to-a-framework. It is therefore illegitimate to undermine the semantic assertibility of a given claim by means of distinctions (e.g., between rest mass and relativistic mass) which cannot be drawn within the framework. Relativistic considerations may not legitimately be used to undermine the semantic assertibility of (1) and (2). So the only legitimate issue is: are these Newtonian claims, despite common belief of Newtonians, not semantically, but only warrantedly assertible? Now it turns out that Newtonians subject (1) and (2) to dual treatment. Sometimes they treat them as constitutive principles of the Newtonian framework (in which case they are surely SA, but not of immediate interest to US),18 and sometimes they employ them in application to particular bodies at particular times in order to make matter-of-fact claims. So the interesting question now is whether the particular Newtonian claim, 'body b has (constant) mass m and its momentum, at different times, is given by m times its velocity' is SA. If
216
RICHARD M. BURIAN
the answer is no for any ordinary case 19 on the ground that although we never detected it the mass was changtng with the velocity (or, alternatively, that the momentunl was greater than the (rest) mass times the velocity), then
no factuq1 claim of the Newtonian framework about the mass of a body was ever SA. To take this position would be to deny the semantic assertibility of all fact-stating Newtonian claims. (For my part, I think that this argument shows that semantic assertibility, that is, truth-in-a-framework, is not truth. Very many Newtonian claims of the fornl 'the mass of body b is m' are SA, but none of them are more than approximately true. The second option frees successor frameworks from the constraint that they must include SA counterparts of all fact-stating SA claims of their predecessor frameworks. But this constraint is required to make good sense of convergence to the Peircean framework. The Peircean framework, after all, is supposed to offer us corrected versions, indeed, the correct versions, of our truths. Unless our (factual) truths are preserved in the form of counterpart truths in succeeding frameworks, truth-in-a-framework cannot count as truth about the world - there will be separate truths in separate frameworks, and scientists employing different frameworks will, in Kuhn's phrase, "live in different worlds." In these circumstances, the internal realism· within each framework secured by the semantic assertibility of its claims will be a sham, for each framework will end up constituting an independent phenomenal world. Not only will there be no non-arbitrary way of choosing between different frameworks, but our realism will threaten to land us in an unintelligible transcendental idealism. Loss of the constraint that all SA factual claims have SA counterparts in successor frameworks, while it nlay not imply Feyerabend's 'anything goes', means that we can only give a retrospective, intra-framework sense to convergence, for it means that convergence is compatible with the complete loss of some of the truths of the prior framework. 2o But such an intra-framework sense of convergence (which, incidentally, I believe Prof. Rosenberg mistakes for the stronger form of convergence sought by Prof. Sellars) makes convergence to the truth, to that which scientists are fated to believe if only they are persistent enough in their epistemologically ideal inquires, no more than an attractive fiction, a myth of a piece with Whig history of science. The third option is to shift our translation of 'mass' creatively from case to case in such a way as to allow us to rescue a SA counterpart for each SA claim of the predecessor framework(s). We could, for example, treat 'body b has constant mass' as 'the rest mass of b is invariant' while treating 'b's momenttum equals its mass times its velocity' as 'b's momentum equals its relativistic
SELLARSIAN REALISM AND CONCEPTUAL CHANGE
217
mass times its velocity'. The problem now is that we can no longer determine the content of a Newtonian claim or the proper interpretation of Newtonian terms except from outside the Newtonian framework. To put the matter in terms of Sellars's theory of linguistic picturing, if the Einsteinian is to preserve the semantic assertibility of the counterparts of Newtonian claims, he must violate the Newtonian treatment of 'mass' as playing the same pictorial role in these two sentences. The Einsteinian may either give us SA counterparts for each of these claims or he may preserve the Newtonian apparatus for determining the linguistic roles of terms (Quine would say for dividing the reference of terms), but not both. By requiring him to preserve semantic assertibility, we require him to use a framework-external principle - namely that of guaranteeing the existence of semantically assertible counterparts in the relativistic framework for certain Newtonian claims - to assign linguistic roles to Newtonian terms and criteria of satisfaction to Newtonian predicates. But this means that, once again, the Einsteinian does not determine what is purportedly pictured (i.e., claimed) by a Newtonian sentence by means of an analysis internal to the Newtonian conceptual structure. The determinateness and independence of the Newtonian picture as a precondition for resolving whether its claims are semantically assertible has once more evaporated. In Quine's language, this is just the inscrutability of reference applied to Sellars's sequence of improved frameworks - with the twist that correction is taken seriously. According to Quine in 'Ontological Relativity,' reference is doubly relativized - once to our background theory (roughly, conceptual structure) and once to the interpretation which, within that theory, we impose on the object theory. (Cf., e.g., p. 65) When our theory, in effect, globally corrects a predecessor (as relativistic mechanics corrects Newtonian mechanics - but there are many more examples of this kind), we have the choice of dividing reference our way in order to preserve truth, (e.g., by interpreting the term 'mass' differently in (1) and (2) and thus violating its supposed Newtonian univocity) or of preserving the 'picturing relations,' i.e., the division of reference internal to the predecessor framework in some regimented formulation, at the price of recognizing some of its SA claims to be false. But we cannot preserve both the old picturing apparatus and truth. Although I have couched parts of my argument in terms of, and to strike at Sellars's apparatus for analyzing conceptual change, neither the argument that there is indeterminacy from our perspective as to which of the claims of a predecessor framework are true nor the argument that there are cases in which one cannot preserve all the SA factual claims of the predecessor framework depend in any essential way on the peculiarities of Sellars's apparatus.
218
RICHARD M. BURIAN
They depend only on the facts that, across frameworks, translation is indeterminate, the linguistic roles of terms inscrutable, and fact and convention inextricable, and on the existence of cases like the relativistic one, in which global changes of structure from one framework to another result in some loss of SA claims. Our construal of a clainl determines what state of affairs it pictures; its truth and semantic assertibility can be assessed only after we have imposed this construal on the claim. Nothing is sacred about the particular construal we adopt, even when the claim is our own. Legitimate alternatives are always available, and they alter the linguistic picture (the purported reference and/or the predicates of the claim) and occasionally alter its semantic assertibility. The practical unavailability of the Peircean framework (whether or not it is available in principle) and the impossibility, in any case, of justifying a unique construal of the claims of predecessor frameworks within a (not the) epistemically ideal successor to our conceptual structure - these difficulties suffice to ensure the unavailability of a viable apparatus of comparison across frameworks except for what can be developed from within - or by extending - the very framework one is using. No single tightly unified framework can reasonably be expected to encompass all results of diverse scientific inquiries - e.g., inquiries into the physical structure of matter, into geomorphology, into secular patterns of global weather, into the genetic plasticity of various populations of organisms, into the biological functions of various hornl0nes, into the perceptual responses of humans to standardized stimuli, and into the fundamental features of communication and information exchange. Indeed, in pursuing such topics as I have just listed, there need be no uniquely correct endpoint to the revisions we should accept in our conceptual structures., for the very franlework within which we seek to resolve a problem should depend on what questions we seek to answer. No framework can be expected to allow us to ask, let alone answer, all questions bearing on a certain range of phenomena. The questions we are interested in (which will be altered, inter alia, by the growth of our knowledge) do and ought materially affect the conceptual developments which we allow. In seeking to answer new questions - and there is no 'complete' set of questions - we may find that all available conceptual structures are inadequate and thus be forced to develop new ones. Thus there is no unique Peircean framework and no unique Peircean assessment of the particular fact-stating claims which are semantically assertible (i.e., true-with-respect-to-our-framework) here and now. Sellars holds that if we are to accomplish the aims of scientific inquiry we must strive
SELLARSIAN REALISM AND CONCEPTUAL CHANGE
219
toward a unique and ideal conceptual str\!fcture within which we could detemine what 'really' exists and what 'really' is true. This commitment, I claim, is an obstacle to a proper understanding of conceptual change and to the achievement of the goals of inquiry not because we do not or should not seek truth (we do and we should), but because there is no way to enlploy truth thus conceived as a standard against which to judge the outconle of inquiry. For truth retreats into the unreachable Peircean framework. To adhere to this false standard is to succumb to the temptations of transcendentalism and to return to the pitfalls of ineffability. XI
To conclude, I would like to suggest one way in which most analyses of conceptual change in science have gone astray. We have made too much of the notion of a conceptual franlework or a conceptual structure. As soon as we reify any such notion, as soon as we make it the basis for a distinction between change of meaning and change of belief, it will land us in trouble. I suspect that the concept of meaning will not, in the end, prove useful in the analysis of science. But if it does prove useful, it will be because of the connection between meaning and the social practices underlying scientific work and communication rather than because of the connection between meaning and so-called conceptual structures. To suggest, as, each in his own way, Feyerabend, Kuhn, Rosenberg and Sellars do, that one can separate one framework from another, that within each framework some claims play a constitutive role, and that those claims determine meanings, is simply to run afoul of Quine's familiar arguments against the analytic/synthetic distinction. Framework-constituting claims are often treated as true by virtue of their meanings - cf. many treatments of the Newtonian claim 'the mass of a body is constant'. In context, when so treated, one may (if one insists on the distinction) treat such claims as analytic. But the very same claims are also used in ways which would have to be classed as synthetic. There is no adequate method for dividing constitutive clainls into analytic and synthetic and no good reason for doing so. If we deal with meanings at all, we should recognize that claims identified as framework constitutive typically play some role in fIXing meanings and some role in making synthetic claims. HaVing recognized this, we will not. be so tempted by Kuhn's claim that after a revolution scientists "live in a different world," for framework change will not be able to shift the meanings of our ternlS so drastically as to make communication impossible, it will not be responsible
220
RICHARD M. BURIAN
for bringing us up against the problems of indeterminacy of translation and inscrutability of reference. (These are problems which we must face anyhow, but they are general problems not particularly connected with the difficulties concerning scientific revolutions.) We should therefore reject any account of conceptual frameworks which makes them into arbiters of 'the' meaning of terms and claims and then goes on to use the concept of meaning as a serious tool in the analysis of theory replacement. The prospectus for our conference asked how to determine which conceptual frameworks are actually used in science. I submit that if frameworks are connected to meanings in the ways to which I have objected, there is no useful, non-arbitrary answer to this question. The complete lack of adequate criteria for discriminating one framework from another, the arbitrary expansion and contraction of frameworks to suit the needs of particular arguments, the fact that the tenn 'framework' is an accordion term, these things make the notion useless as an analytical tool. Ask yourself, for example, whether the same concept can ever appear in two distinct conceptual frameworks. (Are the concepts expressed by the tenns 'velocity', 'body', 'measuring rod', and 'clock' in relativistic and Newtonian mechanics the same or are they counterpart concepts?) If the answer is never, we are genuinely faced with the problem of radical translation, and nothing useful can be said about the preservation of middle-level laws, about correction, or about preservation of truth, of reference, of picture across frameworks. If, on the other hand, the same concepts can appear in different frameworks, we do not know what the effect of meaning-constituting claims is, nor how to decide which cases ought to count as genuine cases of framework change. The proper resolution of these difficulties consists in the recognition that for some purposes it is convenient to treat concepts appearing in different theories as if they were the same, and that for other purposes it is to treat them as if they differed - but that not much rides on which we do. For whichever we do, we must recognize that the massive interpenetration of the terminology of such competing theories as we have been examining is no accident. Correction of one theory by another makes no sense when the two can only be connected after a process of radical translation; it is only when the two can be compared within a single wider context, within a single language, that comparison and correction make sense. For this reason, progressive change of theory of the sort we all seek to understand cannot involve total change of framework. But once we dismiss total changes of framework, the obscure and difficult notion of a framework (and with it such related
SELLARSIAN REALISI\.1 AND CONCEPTUAL CHANGE
221
notions as that of a Kuhnian paradigm) no longer serves as a useful unit of analysis in dealing with theory change in science. APPENDIX
In discussing my paper at the conference, Prof. Rosenberg argued that the three ways out ot the difficulty illustrated by the relativistic revision of the concept of mass, discussed above., pp. 215ff., are not exhaustive. He agreed that the difficulty was not evaded by the proposals I examined, but claimed that a fourth way out (which he himself employed in his Linguistic Representation 21 and which he believes Prof. Sellars intended to employ) is tenable. On this account, the relevant counterparts of Newtonian claims are not relativistic claims, but claims of a model of Newtonian theory constructed within relativity theory by use of appropriate limiting conditions ((vic) -.. 0). Under these limiting conditions, one can ensure that there is no detectable difference between the magnitude of the rest mass and the relativistic mass of an object. A relativist can thus construct a model of the Newtonian situation in which a body has a mass which does not vary (detectably?) with its velocity. The statements of this model, not relativistic statements, are to be the counterparts of Newtonian statements. (It should be noted, however, that even in such a model there will still be loss of n1ass by radiation and other non-classical phenomena!) This proposal seems to me wholly unjustified as a reading of the Sellarsian texts. No matter. The exigetical question is uninteresting. The 'correct' (Le., Newtonian) behavior of the entities of the model guarantees that its claims are not semantically assertible within relativity theory. The n10del's counterparts of Newtonian claims may be used to show how a reasonable analysis of the behavior of bodies which deviate below the margin of observational error from Newtonian expectations can mislead us into accepting a false (albeit extremely useful) theory and help us to understand under what restrictions we can continue to use it. It does not show us that there are true (SA) counterparts within relativity theory to the factual claims of Newtonian mechanics, for the claims of the model (e.g., that the mass of a body does not vary with its velocity) are not SA within relativity theory. Thus the problem from which I began - which relativistic claims are SA counterparts of fact-stating claims of Newtonian mechanics? - has not been resolved. Such models explain our success with the old theory by showing that the model claims are, to the best of our knowledge, approximately true, but not that they are true. In 'good' cases, the model makes clear why we were
222
RICHARD M. BURIAN
warranted in accepting the old theory andl?erhaps even why it was reasonable for us to believe its. claims to be semantically assertible even though factstating claims couched in terms of the old conceptual apparatus or in terms of the model are not, in the light 'of our new knowledge, true. There is no reason to suppose that constraining adequate successor theories to yield such models will guarantee convergence to a unique Peircean framework. The freedom to assume that central principles of our theory are false within such models (e.g., to assume that finite velocities within a certain range do not affect the mass of an object) allows us to model any arbitrary behavior within a sufficiently general theory. Rosenberg's "retrospective justificatory arguments" are simply too easy; they are of a piece with Whig history of science. Drexel University, Philadelphia NOTES
*
I wish to thank Prof. Catherine Elgin, Simmons College, for extraordinary help with this paper. Her positive views and constructive criticisms have informed its every page. I Hilary Putnam, 'Realism and Reason', Proc. Am. Phil. Assoc. 50 (1977),483-498. I am grateful to Prof. Putnam for supplying me with advance copies of this and of the forthcoming works referred to below. 2 Ibid. 3 To be published, together with Putnam's Locke Lectures, in Meaning and the Moral Sciences, Routledge and Kegan Paul, London. The quotation is from pp. 18-19 of the typescript version of the article. 4 SM = Science and Metaphysics, Routledge and Kegan Paul, London, 1968. Citations are by chapter and section. Other references in the text to Sellars's writings will be to the works listed below, with the indicated abbreviations. CC = 'Conceptual Change,' in Glenn Pearce and Patrick Maynard, eds., Conceptual Change, Reidel, Dordrecht, 1973,pp. 77-93. LT = 'The Language of Theories', in SPR, pp. 106-126. P = 'Presupposing', Phil. Rev. 63 (1954),197-215. SPR = Science, Perception and Reality, Routledge and Kegan Paul, London, 1963. SRII = 'Scientific Realism and Irenic Instrumentalism', in R. S. Cohen and M. W. Wartofsky (eds.), Boston Studies in the Philosophy of Science, vol. 2, Humanities, New York, 1965, 171-204. 5 For the best exploration I have yet seen of the effect of.social factors on the structure and content of theoretical and factual claims in science, cf. J. Ravetz, Scientific Knowledge and its Social Problems, Oxford University Press, London, 1971. 6 H. Putnam, 'The Meaning of "Meaning" , in Putnanl, Philosophical Papers, vol. 2, Cambridge University Press, Cambridge, 1975, pp. 215-271. 7 Cf., e.g., P. K. Feyerabend, 'Explanation, Reduction and Empiricism', in H. Feigl and
SELLARS IAN REALISM AND CONCEPTUAL CHANGE
223
G. Maxwell (eds.), Minnesota Studies in the Philosophy of Science, vol. 3, University of Minnesota Press, Minneapolis, Minn., 1962, pp. 28-98; 'Problems of Empiricism', in R. Colodny (ed.), Beyond the Edge of Certainty, Prentice-Hall Englewood Cliffs, N.J., 1965, pp. 145-261, 'Reply to Criticism' in Cohen and Wartofsky (eds.), op. cit. (note 4), pp. 223-261; and Against Method, New Left Books, London; 1975. Cf. also SRII in which Sellars discusses the contrast between Feyerabendian change of theory and Sellarsian change of framework. 8 Of special interest is the fact that so-called scientific revolutions provide the best prima-facie examples of change of framework (and meaning). Such paired systems as Ptolemaic and Copernican astronomy, phenomenological and statistical thermodynamics, Mendelian and molecular genetics, and so on, fit this mold at least in first approximation. 9 Consider Ms. Jones, in the corner, drinking water from a martini glass. You and I do not know her name, but we each believe that she is drinking a martini. Or, perhaps, I believe that she is drinking a martini and you, though you know she is drinking water, recognize that that is my belief. I say, "The woman over there with the martini is a philosopher." You, having heard her lecture, assent. In this vignette, Sellars would hold, the falsity of my presupposition is of no importance because the job of the phrase is simply to correlate with an individual. (Had we known her name, 'Ms. Jones is a philosopher" would have done just as well: the logical form of the sentence is F(a).) Unlike most designating phrases, the long phrase conveys the presupposition which must be employed for the term to function unproblematically. That presupposition, however, is not asserted in the statement proper, as may be seen from the fact it need not be true but only shared or recognized as operative - for the term to do its job and the sentence to come out true. Conclusion: in this context, a descriptive phrase serves, logically, as an indivisible unit, sc. as a primitive term whose job it is to be correlated with an individual of sonte kind (here, a person). Cf. SM V, 18ff. or P. 10 Note particularly that Sellars's semantics is not referential at this point. He argues that, in spite of appearances, neither is Tarski's. Thus the sentence, 'The mirage encompasses 30° of are' may be SA without there being any entity which, in an adequate scientific account of the world, satisfies the term 'ulirage'. The (trivial) T-sentence, " 'The mirage encompasses 30°' is true if and only if the mirage encompasses 30°" is, of course true, But on Sellars's account this only COUles to "The mirage encompasses 30°" is SA if and only if an oratio recta use of the like-sounding sentence is (semantically) sanctioned in our framework (when, in context, it is faced with the relevant evidence). Thus Sellars sees the "truth move" from Sand" 's' is true if and only if p" to p as resting directly on a classification of'S' as (playing the same role as) a semantically assertible sentence and only indirectly on the way the world is. He also holds that the semantic assertibility of a sentence is as nluch a function of its place in our conceptual scheme and of the general structure of evidential support employed in that scheme as it is of the way the worId is. 11 All frameworks, of course, employ laws - this is part of what it is to be a descriptive framework. And so this account of explanation requires that we have laws which, when applied to true descriptive claims, yield further true descriptive claims. Thus the problem of induction is not how to get from fmite instances to laws, but how to get from inaccurate or inadequate laws to better ones. 12 For a closer consideration of the connection between revision of laws and replacement of frameworks, cf. Catherine Z. Elgin and Richard M. Burian, 'Wittgenstein, Sellars
224
RICHARD M. BURIAN
and Conceptual Change', forthcoming in the Proceedings of the Second International Wittgenstein Symposium. 13 Cf. Quine, 'Ontological Relativity', p. 45. This essay appears in Ontological Relativity and Other Essays, Columbia University Press, New York, 1969, pp. 26-68. Cf. also Paul Benacerraf, 'What Numbers Cannot Be', Phil. Rev. 74 (1965),47 -73. 14 A variant of this example is worked out the Hartry Field in 'Theory Change and the Indeterminacy of Reference', 1. Phil. 70 (1973), 462-480. Field's views are further developed in 'Quine and the Correspondence Theory', Phil. Rev. 83 (1974), 200-228. (I am grateful to Charles Parsons for calling this article to my attention.) Field argues against the inscrutability of reference (here~' interpretation-dependence of the reference of the Newtonian term 'mass') and for an account based on "partial reference." (The Newtonian term 'mass' refers partially to rest mass, partially to relativistic mass.) I cannot take up this matter here except to raise a worry: isn't an account in terms of partial reference useless unless, implausibly, we have access to all equally defensible alternative construals of a term like 'mass' in the new framework (here, a relativistic framework)? But how can we determine whether we have exhausted all of the legitimate construals of 'mass' in Einsteinian mechanics, of 'two' within a set theoretical framework, of 'gavagai' in "the" common-sense middle-American framework? 15 Prof. Sellars would object to this way of putting it. Within his inferential semantics, the right way to state my claim is that whether or not a term is assigned an individual sense, and which individual sense it is assigned - Le., whether or not it counts as a designating term and what it designates - is affected by counterpart construal and that no legitimate constraints on counterpart construal can eliminate conflicting and equally adequate construals. For the benefit of those of us who think referentially I shall occasionally speak referential language below, but it is always possible to recast my claims in more strictly Sellarsian style along the lines I have just indicated. 16 Quine, 'Carnap and Logical Truth', p. 132 of the reprint in Ways of Paradox, revised ed., Harvard University Press, Carrlbridge, Mass., 1976. 17 Prof. Rosenberg, in discussion, suggested a fourth way, already taken in his Linguistic Representation. See the appendix of the present paper for a brief discussion of Rosenberg's fourth way. 18 It is important to recognize that Sellars does not hold that the (SA) constitutive principles of a conceptual structure should yield SA counterpart claims in its more adequate successors. It is only SA matter-of-fact claims which have this form of inviolability. Constitutive principles of an inadequate frame work are only approximately true (if they are even that) as Prof. Parsons rightly points out in his comment. 19 There may well be extraordinary cases for which the answer is no - e.g., precisely those cases in which measurements of mass or momentum are attempted on extremely fast moving bodies. 20 I, for one, think that there is such loss in all serious cases of framework change, and that such loss undermines any version of Pericean realism which incorporates change of framework. For Sellars the denial that there ever is such loss - or, better, the insistence that adequate framework replacement must meet the norm of not allowing such loss - is of central importance. For example, he rejects contemporary' materialisms on the ground that by making both physical objects and persons into composites of colorless elementary particles they have no way to preserve the semantic assertibility of such 'ordinary' truths as 'the wall is red'. He holds that unless there are (occurently) colored scientific
SELLARSIAN REALISM AND CONCEPTUAL CHANGE
225
objects entering as constituents into states of sensory awareness of percipients when, for example, they see red walls in normal circumstances, there would be complete loss of the commonsense truth that the walls are (qualitatively) red. Science may relocate color, but it cannot legitimately eliminate it. Cf. esp. 'Science, Sense-Impressions and Sensa: A Reply to Cornaman', Rev. Metaphys. 24 (1971), 391-447. 21 Reidel, Dordrecht, 1974.
LORENZ KRUGER
SOME REMARKS ON REALISM AND SCIENTIFIC REVOLUTIONS Comment on Burian Professor Burian's paper displays and explains his tendency to move away from a scientific realism la Peirce or Sellars and to approach a more or less Quinean stance in the philosophy of science. His view appears to be characterized, among other things, by the combination of the following two claims: (1) Quine's presentation of the problem of radical translation is essentially correct, hence Sellars' or Rosenberg's versions of scientific realism developed from a Peircean metaphysical background are untenable, in particular because these versions assume that the final and ideal conceptual framework of science is uniquely determined. (2) A weaker fonn of realism of the Quinean type is still good enough as an account of science, i.e. it explains how we can successfully make use of language in finding our way about the world, why our scientific views tend to converge, l and in particular why the development of science is free from radical incornmensurabilities. I shall try to indicate my reasons for believing that both claims are false. I shall be very brief in dealing with the first claim, because its proper assessment involves too much (that is: too much for me), namely an entire philosophy of meaning and reference. At least part of this issue will, moreover, appear in the discussion of the second claim, inasmuch as it touches upon questions of meaning in science. As to the first claim, therefore, I restrict myself to a few remarks on the meta-level, Le. remarks that are not meant to contribute to the solution of the problem, but speak about it. The limitations appropriate for a comment may perhaps excuse this way of handling the point. My impression of the radical translation argument has always been somewhat similar to that of radically sceptical arguments: it appears to be logically flawless, but materially inadequate. Hence it has proved extraordinarily fruit.. ful in stimulating conceptual analyses, and it does not admit of an easy or straightforward refutation; yet, on the other hand, it is too far from the cultural experience of any member of the scientific community to ·produce lasting conviction, or at least to induce anybody to derive serious consequences from it concerning his epistemic conduct. To quote Quine: 2
a
There is an obstacle to offering an actual example of two ... rival systems of analytical
227 P. Bieri, R.-P. Horstmann, and L. Kruger (eds.), Transcendental Arguments and Science, 227-233. All Rights Resewed. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
228
LORENZ KRUGER
hypotheses. Known languages are known'through unique systems of analytical hypotheses established in tradition or painfully arrived at by unique skilled linguists. To devise a contrasting system would require an entire duplicate enterprise of translation, unaided even by the usual hints from interpreters.
My question is simply: What is, or what should be, the impact of this admission? There are certianly at least two tasks of philosophical investigation: (a) to analyse conceivabilities in order to produce conceptual clarifications and sharpening of arguments, and (b) to assess as fairly as possible our actual epistemological position in order to control our cognitive and partly also our practical activity critically. In a Kantian illustration, putting forward antinomies serves the first purpose, analysing the phenomenon and the concept of our experience the second. In my view the radical translation argument pursues only the first task; and Quine seems to accept that. But then the second task would have to be fulfilled by some other consideration. And here it is that all followers of Kant and all convergence theoreticians inspired by Peirce seem to have the chance to score a point. 3 In saying this I am implying that Quine's radical translation view is not an adequate account of our actual epistemic situation, hence in particular is not an adequate account of our actual science. I tum therefore to my criticism of the second claim I have attributed to Richard Burian. First I wish to point out that Quine's weakening of the notion of realism is not quite as innocuous as it is made by Burian. To make this plausible let me call Thomas Kuhn to witness. The hard core of his view of scientific revolutions is the thesis that pre- and post-revolutionary paradigms, or more specifically theories, are incommensurable to each other. Now, the essence of incommensurability is meaning change. Therefore Kuhn describes the revolutionary transition as a communication or translation problem, one of radical translation to be sure. 4 A few details of his reasoning are of interest here: When a (possibly partial) communication breakdown between scientists occurs, they will normally still share sensory stimuli, neural apparatus, even most of their neural programming, and - last not least - their everyday and most of their scientific language. On the basis of all this they will be able to describe concrete research results in a mutually intelligible and credible way. If now the new paradigm is able to account for significantly more of these concrete results than the old paradigm is able to do, then the adherents of the latter may become persuaded of the former's superiority. But here it is important to notice two things: First, persuasion is accomplished essentially by successful communication concerning concrete results, or- in a more Quinean language - by shared stimulus meanings and meanings
COMMENT ON BURIAN
229
of observation sentences. The attempts at translating theoretical sentences may re'main totally unsuccessful. This is the reason, why - secondly - Kuhn insists that persuasion does not necessarily imply conversion to the new paradigm. The point seems to me to be both sound and important. You can learn part of a language by learning to translate it into your home language, and you will be able to grasp quite a lot of useful and persuading information this way. But you will still not have become a convert to the new language; i.e. you will not be able to use it creatively, you will not really understand it in the sense that you have entered into the cultural community as one of its members that reproduce and develop actively the community's culture. In order to do this you will have to change your style of life, to live in the new language. The biographies of many emigrants can show how difficult this decisive step may be. Now, since Kuhn views scientific revolutions as changes of language, he is right in insisting that the explication and the explanation of scientific change must, in the last resort, be sought in psychological and sociological considerations. s These considerations are inseparable for him from epistemological considerations. 6 It is true that the radical translation position just outlined is still compatible with philosophical realism; but this realism is obviously not functioning effectively in the explanation of scientific development any more. It is therefore in danger of being dismissed as irrelevant, if not entirely empty. And if Kuhn should have explicated some of the implications of radical translation correctly, the danger for realism could legitimately be connected with Quine's philosophy of science as well. But what is more: the danger comes true. One of Kuhn's further conclusions runs like this: A scientific theory is usually felt to be better than its predecessors not only in the sense that it is a better instrument for discovering and solving puzzles but also because it is somehow a better representation of what nature is really like.
But the latter feeling, according to Kuhn, is inappropriate; he says: There is, I think, no theory-independent way to reconstruct phrases like 'really there'. . . . I do not doubt, for example, that Newton's mechanics improves on Aristotle's and that Einstein's improves on Newton's as instruments for puzzle solving. But I can see in their succession no coherent direction of ontological development. 7
What is the reason for this anti-realist and instrumentalistic conclusion? Simply the fact that the notion of truth can no longer be given meaning (let alone be 'defined') across revolutionary theory changes. And since the truth of theoretical sentences is bound up with the existence of the invisible theoretical
230
LORENZ KRUGER
entities figuring as referents of the t~nns of such sentences, all realism beyond stimuli (and perhaps immediately observable facts) is gone, too. Nor can it be claimed that Quine's position on this is any different: 8 Where it makes sense to apply 'true' is to a sentence couched in the tenus of a given theory and seen from within that theory, complete with its posited reality.
And Burian appears to subscribe to this position. He does not want to dismiss the notion of truth, only that of Truth. 9 On the other hand he seems to need a substitute for the latter; for "truth in a framework is not truth" .10 Yet what I take him to end up with is again something similar to truth in a framework: it is truth with respect to the questions at hand. For those questions detennine for him whether certain conceptual structures are adequate or whether we will be forced to develop new ones. His maxim reads: 11 ... the very framework within which we seek to resolve a problem should depend on what questions we seek to answer.
But there is no uniquely detennined or historically invariant, let alone complete, set of questions. But the relativity of the notions of truth and of reality with respect to a given theory, paradigm, framework, set of questions etc. are of course the stumbling block for all resolute realists. It is, therefore, not true - at least not without significant qualifications -, as Burian claims, that the realism which remains under radical translation assumptions is realism enough in order to account for our actual science, especially for the convergence of originally diverging opinions about the world and for the monotonous growth of knowledge. Before I conclude my paper, I should face the following rejoinder: It is simply a mistake or misunderstanding to declare the radical translation view disentitled of calling itself realist. Quine has explicitly warned us to belittle the fact that, though "everything to which we concede existence is a posit from the standpoint of the description of the theory-building process", is nevertheless "real from the standpoint of the theory that is being built" ;12 and of course "we continue to take seriously ... our own particular worldtheory" .13 This way of making the point may remind us of the Kantian language; for him all appearances are 'empirically real', but 'transcendentally ideal'. Analogously one might say that Quine presents us with theoretical entities which are 'intra-theoretically real', but 'trans-theoretically ideal'. Should we put up with this new brand of idealism? I think, we need not. And since
COMMENT ON BURIAN
231
it comes dangerously close to historical relativism, I think, we ought not. Let me elaborate on this in a few cOJ]cluding remarks. Quine's emphasis on the given conceptual frame or on our own theory14 throws into relief a trait which is con1n10n to all discontinuity theoreticians in the philosophy of science (popper, Hanson, Kuhn, Feyerabend, Quine): a strong tendency toward the historical (as opposed to the ontological) perspective as the relevant basic dimension of the explanation of science or even knowledge in general. This tendency expresses itself clearly in Quine's ready acceptance of Otto Neurath's metaphor according to which science is like a ship continuously being rebuilt while staying afloat. In Barry Stroud's interpretation 15 - approved by Quine 16 - the metaphor means that "we must rely on the on-going conc'eptual scheme we now possess" and hence adopt a "conservative attitude towards theoretical change and revision". Somewhat surprisingly, starting out with the conception of radical translations or - in Kuhn's language - revolutions has brought us back to attaching ourselves pretty finuly to tradition, to the "continuum of cultural evolution", as Quine says.17 This accords well again with Kuhn's assumption that explanation of scientific developn1ent will, in the final analysis, be sociological or historical. And Burian's relativization of truth with respect to a set of questions will commit him to essentially the same view. Now, I think that one can hardly overestimate the fruitfulness, indeed the indispensability, of the concrete historical perspective in the philosophy of science. But I do not believe that one need refrain from the attempt to explain scientific change in still different tenus, that is in tenns of the persistence of the subject matter and the ongoing discovery of its properties. This subject matter is only 'an illusion', as Quine puts it,18 if taken primarily as meaning a notoriously elusive kind of entity. But if we insist that neutrinos lack mass - Quine's example in the satne context -, we do not argue for the truth of that sentence only by embedding it in the continuum of our cultural development. Or how should we understand that physicists have not only treated the highly 'theoretical' particle, the neutrino, as a useful, perhaps - given the theory - indispensable invention, but searched for direct particle-like traces of it, until- after about two decades - these were indeed discovered in scintillation tanks. 19 The persistent subject matter does not consist of meanings, but of the referents of language and secondarily, via these, even of meanings. The last contentions obviously need proof which, obviously, cannot be offered within the frame of these remarks. Hence I can only state my present view that the required proof would have to con1prise essentially two kinds of argument: (i) Epistemological arguments that concern basic features of our recognition
232
LORENZ KRUGER
of ordinary things with their properties and relations. And one should not try to develop these arguments, I think, without including a substantial amount of factual information, which would mean to go definitely beyond what Quine calls the "philosophical point" (the analysis of conceivabilities in my phraseology) and take seriously what he cites as the viewpoint of the "actual field linguist", who is "sensible enough" to interpret 'gavagai' as 'rabbit' and not as 'rabbit stage' etc., just because there are these "enduring and relatively homogeneous objects tl10ving as a whole against a contrasting background" .20 What I want to say is that there is such an ontological-epistelnological story and that it ought to be told. (ii) Arguments from the philosophy of science. These arguments will have to implement a program of reconstructing the entire net of intertheoretic relations embodied in science. To braid that net, e.g. to relate revolutionarily different theories like Newtonian and Einsteinian mechanics to each other, has always been a main concern of scientists. It is by no means an accidental and unnecessary superadditum. Therefore, philosophers may profitably try to understand why this is SO.21 My working hypothesis would be that they can, and should, d-o this under the assumption that atoms and neutrinos are no less, but also no more 0), real than trees and tables. This rather massive brand of commonsense-plus-scientific realism would allow for a less conservative view of scientific revolutions than that which has been successfully suggested by Stroud to Quine and which is again implied in Burian's present paper. It might induce us to revert one of Kuhn's views which belongs to the core of his revolutionary view of science: instead of "no discovery without reconceptualization" we would have to say "no reconceptualization without discovery". Universitat Bielefeld NOTES 1 This convergence must not be confounded with the convergence of research toward the Truth, which Burian explicitly rejects (end of Section X). 2 Quine, W. V. 0., Word and Object, Cambridge/Massachusetts, 1960, p. 72. 3 It is true that Burian, too, offers a contribution to the second task, especially by stressing the fact that successor theories correct their predecessors, or that they show all (or most?) fact-stating claims of the latter to be at least approximately true. 4 Kuhn, T. S., Postscript to The Structure of Scientific Revolutions, 2nd ed. Chicago 1970, §5; cpo also: 'Reply to My Critics', in: Criticism and the Growth of Knowledge (ed. by I. Lakatos and A. Musgrave), London 1970, §6; and the last paragraphs of
COMMENT ON BURIAN
233
'Objectivity, Value-Judgment and Theory Choice', in: Kuhn, T. S., The Essential Tension, Chicago 1977, pp. 320-339. 5 Cf. e.g. Kuhn, T. S., 'Logic or Psychology of Research?', in I. Lakatos and A. Musgrave (eds.), Criticism and the Growth of Knowledge, London 1970, p. 21. 6 Kuhn, T. S., Preface to The Essential Tension, p. xx. 7 Kuhn, T. S., Postscript to The Structure of Scientific Revolutions, 2nd ed. Chicago 1970, p~ 206; emphasis added. 8 Quine, W. V. 0., Word and Object, p. 24; emphasis added. 9 Cpo end of his Section X. 10 Section X, end of second paragraph. 11 Section X, last paragraph but one. 12 Quine, W. V. 0., Word and Object, p. 22; cf. also his reminder of this passage in: Synthese 19 (1969), 265, in his reply to Smart. 13 Quine, W. V. 0., Word and Object, p. 24; emphasis again added. 14 Ibid., p. 24. 15 Stroud, B., 'Conventionalism and the Indeterminacy of Translation', Synthese 19 (1969), 93. 16 Quine, W. V. 0., Reply to Strou,d, ibid., p. 288. 17 Quine, W. V. 0., Word and Object, p. 76. 18 Ibid., p. 76. 19 The first successful experiment is described in: Reines, F. and Cowan, C. L., 'Detection of the Free Neutrino', Physical Review 92 (1953), 830-831. 20 Quine, W·. V. 0., Ontological Relativity, New York 1969, p. 34. 21 This question remains unanswered in the radical translation view, since cultural continuity is not specific enough to account for the peculiar continuity of science.
CHARLES PARSONS
REALISM AND UNDERDETERMINATION* Comment on Burian
Professor Burian'·s strategy is roughly to assume the indetemlinacy of translation and argue that it is incompatible with what he calls "Peircean realism" as represented particularly in the philosophy of Wilfrid Sellars. In considering whether he is right in thinking the two theses incompatible, one has to place the indeterminacy thesis in a larger setting. I shall begin by presenting an argument that would suggest that Burian's main thesis may not be very int~resting. The indeterminacy of translation is supported in the writings of Quine and his defenders by two types of arguments. The first type of argument is exemplified by the famous (or notorious) 'gavagai' example .1 It is roughly that in the situation of radical translation, where one has nothing to go on but speakers' assents and dissents to sentences in different conditions of stimulation, one will have no way of distinguishing different (in Burian's phrase) 'parsmgs' of a sentence so that different objects are picked out as subjects of predication - for example rabbits, rabbit-stages, and 'rabbithood'. The second type of argument has as a premiss the "undeterrnination of the theory of nature", that is the thesis that two theories, call them A and B, might each be compatible with all possible empirical evidence and yet be incompatible. Quine then argues that in this situation we might make our own choice between A and B and still, faced with a foreign community, be able to reconcile with all possible evidence, either the hypothesis that the foreigners hold A (and say, agree with us) or that they hold B. 2 The reason is that the data determine (roughly) only the translation of observation sentences, with respect to which A and B agree, and the assumed truth of A carries us no further. Without presenting this argument less crudely or assessing its validity, two observations are immediately relevant to Burian's project. The first is that Quine says that the second is the more fundamental argument. The second observation is that on the face of it the thesis of underdetermination seems directly to rule out Peircean realism. Burian says, "Peirce held that truth is what - given unlimited time and resources and employing ideal methodology - the community of scientists is fated to believe" (p. 200). But what reason is there to suppose there is one system of theory that "the community of scientists is fated to believe"? It 235 P. Bieri, R.-P. Horstmann, and L. Kruger (eds.), Transcendental Arguments and Science, 235-242. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
236
CHARLES PARSONS
seems that if the thesis of underdetermination were false, Peirceanism would be home free, for if two theories disagtee, at least one would in the limit be defeated by the evidence. It is still not obvious to me that this conception of truth is a form of realism. But I shall leave that question, though the discussion below may be relevant to it. It is perhaps not quite so obvious that the converse holds: if there are empirically equivalent but inconlpatible theories, then there is not one theory that the ideal scientific community is fated to believe. But it is certainly what one would expect at first sight: members of the ideal community might come to believe either one of two equivalent theories. Even if they had a drive toward agreement that rules out the possibility that some accept one and some accept another, their con1illon acceptance of one rather than another is not fated. 3 Plausible as this may be, it is not a logical consequence of underdetermination. We cannot yet rule out the possibility that acceptance of one of two 'empirically equivalent' theories is in some way determined. Quine defines two theories to be empirically equivalent if, roughly, they agree in their observational consequences. 4 This seems to me a weak definition in that inductive logic might carry one further. Whether this is so is not important for our discussion; I will assume that two theories that, given all the possible evidence, get equal marks from inductive logic are empirically equivalent. This is, of course, relative to a determination of what belongs to inductive logic. Given this understanding, it seems to me that in the abstract there might be four types of factors that could determine the choice between equivalent theories: (1) A priori principles (2) the natural constitution of man (3) theoretical intuition (4) history ('historical accident') I have not listed simplicity, which is often mentioned in this connection. I think simplicity is likely to be reflected in our inductive logic; to the extent that it is a formal notion and does not belong to inductive logic, it seems to me very doubtful that considerations of simplicity would be sufficient for this sort of decision. If it is not a formal notion (that is, if it applied intuitively in a way dependent on the subject nlatter) then I would subsume it under
(3). The order of my list is intended to be that of 'degree of determinism'. I will make some remarks about each in tum.
REALISM AND UNDERDETERMINATION
237
(1) It seems we might rule this option out of court on the ground that it is incompatible with the actual procedure ~f contemporary science. However, if such principles are taken in the fallibilist spirit in which they were taken by Professor von Weizsacker in his paper for this conference, then I think this conclusion would be too hasty. However, if a priori principles could be formulated and proved that will take up some of the slack left by empirical underdetermination, there would probably not be reason to hold Peirce's theory of truth; the connection between truth and Evidenz would be more direct. Moreover, given that the possibilities of this sort that are actually discussed base the relevant a priori on necessary or constitutive conditions for experience, it seems to follow that either underdetermination still survives when these principles have been taken into account or, since the conditions in question constrain the enlpirical evidence, there will not be empirical underdeternination in Quine's sense. (2) This seems to me to be the most interesting case, because it seems to me to show that Peircianism can be reconciled with Quine's thesis of underdetermination, but only in a somewhat paradoxical way. What we are doing is making, in a naturalistic spirit, the hypothe~is that for any growth toward the limit of available evidence and logical and mathenlatical knowledge, a convergence toward a unique ultimate theory will take place, but the reasons for this lie in the nature of nlan in the sense that they would be common to any scientific community independently of its historical origin. 5 It seems to me that something like this must be true if realism is to be viable in the context of naturalized epistemology . Now just the considerations that lead from underdetermination to indeternlinacy seem to threaten the coherence of our hypothesis. Since it is supposed to be naturalistic, we will suppose that the ultimate 'Peircean framework' contains whatever explanation for its having been the survivor in the jungle of possible theories goes with the way in which our hypothesis is realized. Let us suppose that the Pericean framework is formulated in the Peircean community's language as a certain theory A. How does A's explanation of its own acceptance work? A prior question is: what exactly does A explain? That the members of the Peircean con1ffiunity accept A. However, the statement that they believe the propositions expressed by the sentences of A is a highly theoretical statement that apparently would follow from A only if A contains some theoretical apparatus for interpreting utterances and other behavior as expressing propositional attitudes. 6 To assume that A contains such apparatus would beg the question of the relation between underdetermination and indeterminacy.
238
CHARLES PARSONS
It seems that our hypothesis sheuld be that A explains the fact that the members of the Peircean community accept or 'hold true' the sentences embodying A. This explanation is supposed to be 'naturalistic'. It might be supposed to be an explanation of the community members' disposition to give a definite sign of assent when queried about axion1s (or proximate consequences thereof) of A and to give a defmite sign of dissent when queried about the negations of these sentences. In this form, the hypothesis is not plausible. The nature of the species is not going to determine what language any conununity speaks. We may of course read the hypothesis that they possess the evidence that they do as meaning that they accept certain observation sentences under appropriate conditions. But without the assumption that A contains a theory of interpretation, it will still not carry us with any sort of necessity to a set of theoretical sentences that must be accepted. Rescue may come from a suggestion of Hilary Putnam: a realistic attribution of reference to the expressions of our language, according to which our theories are true (at least in the ideal situation), can be an explanatory model for the success of science. 7 In the actual, non-ideal situation, 'realism as a scientific hypothesis' says that with the accumulation of evidence and mathematical knowledge, the scientific conm1unity will come closer and closer to having true beliefs, and this closeness to truth will explain the success of science in achieving such aims as prediction. Now we may imagine our 'Peircean hypothesis' to extend such a realist hypothesis to explain also the community's acceptance of the theory itself. Then we will suppose A to contain an internal semantics that assigns truthconditions to the sentences of the language, which sidesteps the problem of indeterminacy by being homophonic. Of course it will not be literally true that A contains its own semantics; paradox would threaten. I would picture the matter thus: Considered as a theory of nature, A does not need to use the full mathematical power of the community's language; for example, its quantifiers do not need to include in their range absolutely all sets. For the truth-predicate, we shall have to call on more mathematics. Strictly speaking, then, what gives the explanation of A's acceptance and success is not A itself but A plus a little more mathematics; this is the 'background theory' in which the semantics of A is given. I do assume that not all of mathematics belongs to the theory of nature; if one rejects this, the kind of explanation we are concerned with can only be partial. 8 In this form, it seems that we can rule out the suggestionthat the theory B, supposed empirically equivalent to A, can accomplish a parallel feat to what
REALISM AND UNDERDETERMINATION
239
we have attributed to A. For if B implies that the natural constitution of man constrains the ultimate scientific community to accept B in the sense of assenting to those sentences that make up B, it will not be empirically equivalent to A, at least in an extended sense that implies equivalence with respect to the evidence for translation and interpretation, and indeed B will be contradicted by the observable fact that the members of the Peircean community accept A. It therefore seems that our hypothesis can be saved as an hypothesis that the Peircean community makes about itself, identified by the homophonic truth theory for its language. Can this 'provinciality' be overcome? We now have to look directly at the situation of translation. Imagine the Peircean community, holding the theory A, confronted with foreigners such that the data call for translating them in such a way that they hold either A itself or an empirical equivalent of A, say B. 9 The necessity with which speakers of the home language hold A given the evidence does not carryover in a direct way because it has to do with assenting to sentences. Given a theory of interpretation for the foreign language that takes the form of a truth theory, or which implies one,10 an inference parallel to the one the Peircean community makes about itself can be constructed: from facts about the acceptance of evidence sentences (together with other facts such as their stimulus meanings) by means of the truth theory to conclusions about the acceptance of theoretical principles. The expectation that such an inference can be made is a constraint on interpretations, since some might lead to conclusions that can be directly refuted by the foreigners' behavior. If the inference has the character that the theoretical principles are accepted because they are true, then it will rule out principles incompatible with the theory A, since clearly the truth theory when added to A cannot declare such principles true. What objective justification might there be for such a constraint? It seems on the face of it that a well-constructed theory of interpretation should allow the Peircean community just to carry over its own explanation for its acceptance of the principles of A. If this is not so, then the explanation relies on accidental features of the Peircean community's language, such as perhaps its phonology. This would contradict the basic idea with which we started, that it is necessary that a scientific community in possession of all the evidence will accept a certain theory. It would surely be permissible in such an inference to use the truth-conditions of observation sentences. It appears to be a requirement of the transfer of the inference to the foreign language that the foreigners accept observation
240
CHARLES PARSONS
sentences agreeing in truth-conditions to those accepted by the Peirceari community. However, that can be accepted as an implication of the hypothesis that the two communities possess equivalent evidence. Thus it appears that our hypothesis can be extended to a non-provincial meaning. However, the manner in which we have done this raise doubt as to its naturalistic character. The necessity with which a particular theory (or equivalence class of theories under some relation stronger than empirical equivalence) arises in the Peircean situation has to be treated as a constraint on theories of interpretation to be constructed for foreign language. It may be that the form of the Peircean community's explanation for their acceptance of the theory A will make it clear to them how to canstruct an interpretation of the foreigners that has them accepting A as well. But what grounds are there for ruling out the possibility of an alternative interpretation that will have them accepting the empirically equivalent theory B? It may even be possible to give some persuasive explanation of why the foreigners hold the queer views they do. The grounds that come to mind for preferring the A-interpretation have to do with maximizing attribution of rationality, or agreement. Realism, if that is the form the hypothesis takes, would not in this case be what we normally think of as a scientific hypothesis. The factors (3) and (4) from our above list of factors that could determine a choice between equivalent theories can be disposed of quickly. (3) By theoretical intuition I mean something like our sense of the direct intuitive plausibility of the basic principles of a theory, or our conception of a satisfying explanation where it cannot be subsumed under inductive logic and is not so strict as to count as a priori. I don't see any new any of saving Peircean realism from underdetermination in this way, since it seems to me that if it really necessarily took up all the slack, it would have to ammount to a strong a priori. (4) I think the Peircean is entitled to suppose that historically conditioned disagreements would eventually be cancelled by communication. It seems to me that this will not save the Peircean if only such contingencies can determine the choice between empirically equivalent theories, since the result would itself be contingent, or it may be the sort of dualism mentioned by Quine in 'Empirically equivalent systems of the world', that is where one holds two empirically equivalent but incompatible theories without making a choice between then1. Indeed, the possibility of just this kind of dualism makes the kind of hypothesis that we discussed under (2) somewhat doubtful as a way of saving the Peircean position.
REALISM AND UNDERDETERMIN ATION
241
I shall now return to Professor Burian's actual argument. An argumen I discern in a number of places in t]1e paper applies the doctrine of ontological relativity against Sellars' invocation of picturing as a way of gettin from warranted assertibility to truth. "It is our construal of a claim which determines what state of affairs it pictures." (p. 218). With respect to ontological relativity it should be observed that different 'parsings' that cannot be decided between on evidence acceptable from the Quinean point of view can arise where the truth of sentences is not affected, or where the sentences affected are 'don't cares' from the point of view of the purposes served by the part of language in question. The former is the case for the original 'gavagai' example and the example of Japanese classifiers. 11 The latter obtains for the different construals of mathematical objects in terms of others that arise in constructing mathematics in terms of set theory or arithmetic. In cases like these, indeterminacy of translation does not pose any threat to the Peircean conception of the truth provided that the Peircean admits that this truth can be formulated in different ways with different ontologies, at least in· the technical sense. Perhaps some would not regard this as realism, but anyone accustomed to the foundations of mathematics, where a realist conception of the relation between truth and knowledge can coexist with general acknowledgement of 'ontologically' different formulations of equivalent theories will not be so demanding in his criteria of realisn1. Burian also uses the example of mass in classical and relativistic mechanics, which is not of this form and which he may think supports a more radical critique of Peircean realism. Here I find his argument unconvincing because its force goes mainly against the specifically Sellarsian claim that a 'successor framework' must preserve all 'semantically assertible' statements of the earlier framework. The more general intent of the argument seems to be to undermine Sellars' concept of correction, and therefore to undermine the Peircean idea that successive frameworks tend toward an ideal framework that will then count as the truth. However, once we drop Sellars' claim of preservation, we can take the line, which Burian seems to accept, that from the point of view of the later framework even the most 'constitutive' principles of the earlier one are at best approximately true. But if they are approximately true, then the idea that science might tend toward an ideally adequate theory can get some support on inductive grounds, as Putnam an Boyd have argued. Though this may be only 'internal' realism, it can, as I suggested above, have the form called for by Peirce's conception of truth. The most serious threat to an ontologica11y
242
CHARLES PARSONS
relativized Peircean realism is the possioility of the sort of dualism Quine mentions in 'Empirically equivalent systems'. Columbia University NOTES
* A revised version of comments at the conference on Transcendental Arguments and the Conceptual Foundations of Science, University of Bielefeld, Germany, July 14, 1977, on Richard M. Burian, 'Sellarian Realism and Conceptual Change in Science', this volume, pp. 197-225. I am indebted to Professor Burian's helpful explanations and to discussion at the conference. The revision benefited from my seminar on Quine at Columbia University in the fall of 1977, particularly from Sidney Morgenbesser's participation. 1 Word and Object, Technology Press, Cambridge, 1960, ch. 2, esp. §12; cf. Ontological Relativity and Other Essays, Columbia University Press, New York, 1969, pp. 29-35. 2 'On the Reasons for Indeterminacy of Translation', The Journal of Philosophy 67 (1970),178-183. 3 Indeed, such a drive toward agreement could lead to acceptance of one theory as a convention in a sense like that of David Lewis, Convention, Harvard University Press, Cambridge, 1969. 4 'On Empirically Equivalent Systems of the World', Erkenntnis 9 (1975), 313-328, pp.316-3l8. 5 I shall not attempt to cope with the natural question about other possible biological species with comparable intellectual capacity. 6 Cf. recent writings of Donald Davidson, particularly 'Radical Interpretation', Dialectica 27 (1973), 313-328. My mention of propositions is not to be taken in an invidious sense, but merely as a place-holder for a description of the beliefs in question that does not assume a definite formulation in the believer's language. If A is finitely axiomatizable, say by sentences S, ... Sn, then my remark applies to the statement that the members of the community believe that S, & ... & Sn. 7 'Reference and Understanding', in his Meaning and the Moral Sciences, Routledge and Kegan Paul, London, 1978. Cf. also 'What is "realism"?', Proceedings of the Aristotelian Society 76 (1975-76), 177-194. I do not think Putnam meant to withdraw this claim after what Professor Rorty referred to as his Kehre; in any case, I am putting it in the context of the 'internal realism' he continues to defend, applied to a hypothetical Peircean framework. See 'Realism and Reason', Proceedings and Addresses of the American Philosophical Association 50 (1976-77),483-498. 8 Cf. my 'Quine on the Philosophy of Mathematics', to appear in P. A. Schilpp (ed.), The Philosophy of w. V. Quine. 9 Cf. 'On the Reasons for Indeterminacy of Translation', p. 180. 10 Cf. Davidson, 'Radical Interpretation'. 11 Ontological Relativity and Other Essays, pp. 35 -37.
III
THE TRANSCENDENTAL APPROACH AND ALTERNATIVE POSITIONS
JA Y F. ROSENBERG
TRANSCENDENTAL ARG UMENTS AND PRAGMATIC EPISTEMOLOGY
1. The leading idea of the Critique of Pure Reason is synthesis. Synthesis, too, will be the dominant theme of this essay - not only those syntheses which penneate the first Critique but a 'synthesis between what the prospectus for this conference initially posits as competing approaches to the notion of a conceptual schen1e and the justification of empirical knowledge as well. I want to begin by availing n1yself of that prospectus in settling my problematic. It provides a convenient way of introducing a variety of interlocking themes and topics, potentially bewildering in their complexities. In general terms, my project here will be to sketch a rapprochement between a pure Kantianism on the one hand and a certain species of pragmatic naturalism on the other. This is an enterprise in which I have been oc~upied for a considerable time now, and the best I will be able to do on this limited occasion is to reconstruct a rudimentary picture of some of the highlights of the synthesis which I am endeavoring to construct. As is usual, the vast bulk of the conceptual iceberg will1ie beneath the surface, but I hope that I can say enough to give you at least a reasonably clear indication of its overall shape. 2. Let me begin, then, by noting that the question focusing this symposium is posed in terms of justifying empirical knowledge. Now the theme of justification is one element of a general theory of epistemic conduct. The 'empirical knowledge' which concerns us here is embodied in a 'conceptual scheme' or 'conceptual framework' or 'representational system'! which we espouse or employ. It is such doings - such espousals or adoptions or employments which, in the first instance, are justified or unjustified, and it is with the justifiability of such epistemic conducts that we are, in consequence, primarily concerned. It is important to notice in this connection, then, that - as with any conducts - questions of justification may arise either prospectively or retrospectively. The prospective stance is appropriate to contexts of decision, occasions on which one is deliberating about what to do, while the retrospective stance applies in contexts of evaluation, when one is assessing what has been done. It is crucial to keep these two perspectives separated in our thinking, for unless we do so, we are in danger of mistaking the correct conclusion that, for example, the employment of some conceptual scheme cannot be 245 P. Bieri, R.-P. Horstmann, and L. Kruger (eds.), Transcendental Arguments and Science, 245-262. All Rights Reserved. Copyrig/1t © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
246
JAY F. ROSENBERG
justified in prospect for the stronger, but erroneous, conclusion that it cannot on that account be justified at all. 3. What is supposed to suggest that empirical knowledge stands in need of justification is the challenge of epistemological skepticism. Traditionally, this challenge has been framed in terms of certainty, but 'certainty' is an elastic term-of-art, and it is important for this reason to get clear about the relevant sense of the term and about its intended contrasts. The locus classicus for the skeptical challenge is Descartes, and it will repay us to take a quick, but careful, look at his considerations. 4. What Descartes intends by 'certainty', he tells us, is indubitability. He proposes to disqualify as certain any belief for which he can isolate even the least 'grounds for doubt'. When we look at what Descartes is prepared to entertain as constituting a ground for doubt, however, we discover that his concern is not - as one might expect - with the quality of our reasons for beliefs but with the nature of the causes of beliefs. Each of his ostensible 'grounds for doubt' takes the form of an hypothesis concerning the possible causes of belief. What troubles him most about representings issuing from the agency of a genie malign is not that they would ipso facto be false but that they would be arbitrary.2 Even if our beliefs were true, there would in such a case be no connection between the fact that they were true and the fact that they were ours, Le., that we held them. The primary Cartesian contrast, thus, is not - as it is often taken to be - one between certainty and revisability but rather one between certainty and arbitrariness. What is needed to answer this form of skepticism are not considerations which imply that our conceptual schemes are irrevisable, in whole or in part, but an argument to the effect that - although defeasible - our ways of thinking about the world are not arbitrary but determinate, and, in fact, so determined as to supply some connection between our having a certain world-picture and its being a correct worldpicture. Holding this thought for a moment, let me focus on another element of the prospectus. 5. What I have in mind here is an ostensible contrast drawn in Section (2) between views of the way in which 'competing' conceptual frames belonging to different periods of scientific development are related to one another. As the question is posed, we are to choose between the theses that (a) they can be united into one single comprehensive framework and (b) the relationships between them are essentially historical in nature.
TRANSCENDENTAL ARGUMENTS AND EPISTEMOLOGY
247
6. But putting the question in this form strikes me as rather like asking whether a dozen eggs is one thing or twelve ,things. These conceptual frameworks are related across time not simply as alternatives or competitors but as predecessors and successors. They are not a set but a sequence, and the abandonment of one for the espousal of another is itself a time-bound epistemic doing subject to questions of justification. We may recognize their relationships as essentially historical, but we are not thereby constrained to regard them as merely historical. The possibility remains that we may locate a unity for their plurality not in a synchronic summation of their contents but in the legitimacy of an epistemic process which this diachronic sequence repeatedly instantiates. The unity of science, in other words, may well be not a contentive but a methodological one.
7. The distinction between prospective and retrospective questions of justification becomes of critical importance in this connection. Even if successive passages from one system of representations to another are genuine 'scientific revolutions', incapable of being anticipated in a 'logic of discovery' and not amenable to justification in prospect, we are not precluded from seeking, and even finding, an invariant pattern of retrospective justifiabilities characterizing the epistemic process of frame-succession at all of its stages. We can see ourselves retrospectively as evolved apes even if our evolutionary ancestors could not similarly envision themselves prospectively as potential rational hominids. Adopting such an evolutionary perspective may issue in commonalities of content (as the genetic apparatus of self-replicating DNA proves to be a shared feature of all terrestrial species). This remains to be investigated and I shall, in fact, return to the question at the end of this essay. But, if successful, it must issue at least in a picture of successive conceptual frames which are not arbitrary vis-a-vis their predecessors and thus, at least potentially, in a commonality of focus or direction - that is, in the possibility of a diachronic convergence of representational systems.
8. While this retrospective epistemic determinacy is not obviously the one which exercised us in connection with skepticism a moment ago, it does show us how to effect a final reconciliation between determinateness and defeasibility, for it sensitizes us to the possibility that an absolute sense of correctness might be funded diachronically by appeal to a sequence of comparatively better approximations. If a commonality of focus or direction can indeed be
248
JA Y F. ROSENBERG
projected for conceptual schemes in their determinate diachronic succession, it becomes open to us to think of 'them as aimed at and converging upon the
correct conceptual frame. The desired antiskeptical connection between espousal and (absolute) correctness would thus be itself a connection in prospect and, for us - embedded as we are in these diachronic epistemic processes - no more than a regulative ideal. But it would not on that account be a fiction or an illusion. 9. This evolutionary stance, I would suggest, is the idea which the 20th Century needs to bring to Kant, for with it we can not only complete his case against the episten10logical skeptic but also break that last untenable dualisn1 on which he himself based that case - the dualism between the empirical and transcendental points of view for which Hegel, quite rightly I think, takes him to task (as do our emerging contemporary neo-Hegelians). But I am getting well ahead of my story now, so let me retrace my steps and elaborate my evolutionary picture a bit. 10. What is wanted of a justification in retrospect of the passage from one conceptual frame to another is an argument to the effect that the successor scheme is better than the predecessor which it supplants, where this being better is, presumably, to be a matter of its being more nearly correct. The principal diffic\llty then becomes one of locating an epistemic stance from which such an appraisal can sensibly be made. This problem, I want to propose, has an analogue in biological theorizing. The claim that the function of the heart is the circulation of the blood, for example, is intended to have more empirical content than the mere observation that the blood circulates in consequence of the action of the heart. The analogous difficulty is to supply an elucidation of this 'more' without invoking an utterly mysterious mechanism of fmal causation or some similar teological arcanum. The solution provided by the evolutionary perspective is to supply the requisite content by showing how the causal relationship between blood circulation and the action of the heart can itself be caught up in an explanatory account of how organisms so structured came to exist and persist in the present terrestrial environment. We explain the circulation of the blood by appealing to the action of the heart, but the existence of the heart is explained in turn not by a further appeal to synchronic or teleological causality but by embedding that question in a broader diachronic theoretical context. We explain the existence of the heart (the organ) by explaining the emergence and persistence of creatures with hearts (the organisms). It is this diachronic evolutionary
TRANSCENDENTAL ARGUMENTS AND EPISTEMOLOGY
249
causal account which is in fact unperspicuously encoded by the functionalists' ostensible synchronic teleological vocabulC!IY: "The heart exists in order to circulate the blood." 11. There are a number of features of this analogy which I shall want to exploit in what fo11o\'/s, but what deserves immediate attention is its strategy of contextual embedding, for that is Kant's strategy as well. Kant sees his problem as that of demonstrating the legitimacy of categorial concepts and securing epistemic authority for judgments in which such concepts are mobilized. Now according to the philosophical tradition supplied by his predecessors, there are only two modes of conceptual legitimization: Justification is either logical or evidential. Logical warrants are analytic entailments, but Hume's alarm clock had quite efficiently awakened Kant from the dogmatic dream that categorial judgments can be analytic consequences of any fmite set of experiential propositions. Yet the possibility of evidential warrants, in turn, arguably presupposes that of non-evidential - and, hence, given the tradition, logical - warrants. An appeal to ostensible evidence can only legitimize a judgment if we can warrantedly judge that the ostensible evidence genuinely is evidence. Yet something, E, can be known to be evidence for the truth of a judgment J only if we can know that J is true, always or often, when E obtains, and this requires that we have epistemic resources independent ofE for legitimizing our judgment that J is true. 12. Kant expresses these observations by noting on the one hand that analytic entailments do not yield synthetic judgments and, on the other, that evidential appeals will not support a priori (i.e., universal and necessary) conclusions, while insisting that the judgments with which he is concerned are both synthetic and a priori. They consequently require their own, unique, mode of justification, and it is this which Kant proposes to provide under the rubric of a transcendental deduction. 13. But it would be a mistake to suppose on this account that Kant intends his transcendental deduction to be a new form of argument, standing along.. side modus ponens and (perhaps) instantial induction. What he provides is not a new form of argument but rather a new form of reasoning, a new way of bringing reason to bear on questions of justification. Put crisply, his strategy is to legitimize a categorial (synthetic a priori) judgment J by embedding it in a larger judgment which is independently justifiable and from the legitimacy of which the justifiability of J itself follows. Essentially - although admittedly
250
JA Y F. ROSENBERG
telegraphically - what Kant proposes to do is to analyze the notion of a world in order to establish that the meta:;;;judgment: 'The categorial (synthetic a priori) judgment J is true of the world' is itself logically warranted.
14. Now this notion of the world has lately been the focus of significant critical scrutiny, but rather than take it on directly I want to stay with Kant's line of thought for a while longer. We should note in passing, however, that the world is also supposed to be what the conceptual frames of science more or less adequately represent, and if Kant is correct in his contention that certain categorial concepts necessarily apply to the world, it will presumably follow that there will be a categorial content necessarily shared by alilegitimate members of the evolving sequence of alternative frameworks. Whether this is in fact the case is a question which I have already once postponed until the end of this essay. Consistently, I here postpone it again. 15. Kant's notion of the world is essentially related to his analysis of experience. The world at issue is the world of possible experience and, correlatively, what sets experience - in the full Kantian sense - apart from mere representation is that experience is of a world. The central then1e fron1 which Kant develops both this notion of a world and that of experience is synthesis. The representation of a world is not a congeries of representations but a single unified synthesis of representations. Kant's claim is that it is only by being conceptually structured through the (schematized) categories that representations can be - for us -elements of such a unitary synthesis. 16. Recall our earlier discussion of skepticism. We are looking for a connection between the correctness of a conceptual scheme and our espousal of it. Kant's particular genius here lay in focusing his attention on the unasked question: Who are we? His answer has several parts. We are, to begin with, discursive intelligences. That is, we are beings who can represent a manifold only by representing its elements as in relation to one another. But beyond this, we are also apperceptive intelligences. We represent ourselves as the unitary subjects of a plurality of representings (a fact which Descartes' cogito had effectively brought to the center of the philosophical stage). This combination of discursiveness and apperception demands an internal synthesis. We can think of all our representings as ours only by thinking of them as in relations to one another. The specific relations in question for human discursive intelligences are those setting our representings in a determinate order in time, for our apperceptive unity is the unity of a continuant with a history
TRANSCENDENTAL ARGUMENTS AND EPISTEMOLOGY
251
(even if of a 'continuant' only in the limited, Humean, sense of a sequence of representings). Kant's unique insight is that the achieving of this internal synthesis requires that we also undertake an external synthesis - that is, that we represent objects as existing, persisting, and interacting in a single unitary (spatio-temporal) world, independently of our representing them. 17. This, of course, is the substance of Kant's 'Refutation of Idealism'. While his argument is fascinating - and, I think, cogent - pursuing it here would take me too far afield from my central concerns. It properly deserves an essay of its own, one at least as long as this one. So I shall simply grant it. It will be the major unargued piece of my present project. What I will do, however, is put it to work. 18. What is important about Kant's 'Refutation of Idealism' for our present purposes are the links which it forges among the espousal of realist frames (Le., conceptual schemes which posit a distinction between existence and being represented), the activity of synthesis, and our own existence as discursive apperceptive beings. Kant's key point is that the activity of synthesis is not an option for us but indeed the very activity by which we constitute ourselves. Operating epistemologically under the prime constraint of a continually developing synthesis is something which we necessarily do. This, in fact, is the fundamental locus of all necessity, but it is not a problematic apodeictic necessity which is at stake here. Operating under the constraint of synthesis is son1ething which we necessarily do only in the sense that we couldn't be discursive apperceptive intelligences unless we did it - that is, we couldn't be us. It is in engaging in a continuing project of synthesis that we consist. 19. As advertised, I shall, in what follows, presume that this Kantian point has been established. What I want to do next is to explore son1e of its consequences. The first of these concerns justification. Transposed into the idiom of practical reasoning, what Kant's argument establishes is that, as apperceptive discursive intelligences, we necessarily operate with the epistemic end of enhanced synthesis. It follows, then, that the fact that espousal or adoption of a particular conceptual frame or system of representations eventuates in such an enhanced synthesis is necessarily a reason for us to espouse or adopt it. We ought, caeteris paribus, to do so. Retrospectively, then, if our adoption of a particular conceptual frame did issue in such an enhanced synthesis, it follows that we were (at least to that extent) justified in adopting it, for to
252
JA Y F. ROSEN'BERG
have been justified in a piece of conduct precisely is for there to have been adequate reasons for our engaging in it. And this is so whether or not what was done was done for those reasons, and whether or not we were cognizant of those reasons at the time of our action. This is an important characteristic of retrospective justification. It makes perfect sense to conclude - and we can sometimes discover - that what we did was in fact justified, even though we could not have known it at the time. An action can be warranted by circun1stances which are not known to obtain at the time when the action is performed, and we can later discover that what we did was in fact so warranted. 20. There is, then, an essential connection between the self-constituting activity of synthesis and epistemic justification. The next step is to bring this picture to bear on the question of correctness. Recall that what we concluded was wanted of a just~fication in retrospect of the passage from a predecessor to a successor frame was an argument that the successor scheme is more nearly correct than the predecessor which it supplants. What we have just concluded is that such a passage which issues in an enhanced representational synthesis is ipso facto justified - Le., that the fact of enhanced synthesis necessarily is a reason for our abandoning the predecessor in favor of the successor. What is needed to bring this result into harmony with our original demands, then, is an argument to the effect that a more highly coherent system of representations is on that account n10re nearly correct. 21. For Kant, this argument logically could not be forthcoming. More precisely, what Kant could not argue what that increasing phenomenal coherence implies increasing noumenal correctness. Kant purports to operate with two perspectives, the empirical and the transcendental. From the transcendental perspective there is another world which is not the world of which we have been speaking, the world of possible experience, but rather a world of incognizable noumena or things-in-then1selves of which we can actually have no proper concept. Correctness in the fullest classical sense (an absolute or metaphysical correctness) would consist in the correspondence (or, more generally, in the adequacy) of our representings to this noumenal world. But this is an equation which Kant can never strike, for its noumenal side is always and necessarily hidden from us, both incognizable and ineffable. About things-inthemselves, absolutely nothing positive can be sensibly affirmed. Nevertheless, Kant insists that empirical reals (Le., the objective representeds of our conceptual schemes) are transcendentally ideal. From the standpoint of transcendental philosophy, their esse is concipi, and in this they contrast with things-
TRANSCENDENTAL ARGUMENTS AND EPISTEMOLOGY
253
in-themselves, with our representings, and with us ourselves - all of which are transcendental reals (although we can't really, of course, legitimately say so). 22. The perennial idealist challenge is that this biperspectival picture (essentially no different in its metaphysical content from any representative realism) is an untenable one. This is the challenge which Hegel addresses to Kant but it is equally the challenge which Berkeley addressed to Locke: How does your ostensible notion of "something, I (can) know not what" differ fronl no notion at all? What Hegel is saying to Kant is that his 'noumenon', like Locke's 'matter', is not a contentive term which can enter into an intelligible (metaphysical) hypothesis but a mere noise, empty of any sense. Transcendental biperspectivalism does not give a possible picture of what it is in which the correctness of our representings could consist - for it does not give any picture at all. 3 23. But if Kant is right about the connection between the activity of synthesis and our espousal of realist frames - and I am taking it that he is - the idealist alternative, that the esse of all objects is concipi, is also not a possible human metaphysics. For we necessarily think of ourselves as inhabitants of a world of things which exist independently of our representing them. And so we are now again face to face with the central dilemma ofsystematic ontology, the dilemma which Kant intended his biperspectivalism to resolve. We can neither be Kantian transcendentalists nor Hegelian idealists. It is the singular virtue of pragIl1atism to show how this dilemma can be genuinely
aufgehoben. 24. What the pragmatist viewpoint accomplishes is really nothing but the carrying through of Kant's 'Copernican Revolution' to its fullest implications. (Peirce, after all, was the very archetype of a Kant scholar.) The content of any conceptual frame which is possible for us is irredeemably realist. We necessarily think of ourselves as representing more or less adequately a world independent in its existence of our representing it. That, I am taking it, is one truth captured by Kant's transcendentalism. But such ontic realisn1 is epistemologically idle. Although it is an unavoidable feature of our thought, it can exercise no constraints on that thought. The presumptive adequacy of our representings to the world cannot be a criterion of correctness for those representings, for the ostensible world is accessible to us only through. the medium of a system of representations and so, necessarily, is not an object of comparison against which our conceptual schemes can be measured. And that
254
JA Y F. ROSENBERG
is the truth in all idealisms. But each classical view pays a price. We have already seen Kant's price as the price of all traditional realisms, measured in the coin of ineffability and incognizability. But the idealist pays a price as well - in the coin of arbitrariness. For if the sole constraint on conceptualization is consistency (and why even that?), then we may say and think what we like. There can be no question of correctness beyond that of the degree to which one's ways of thought and speech are shared, and the most that can be said of one who does not share them is that he is different. The idealist thus cuts himself off from all possibility of rational appraisal. Truth gives way to faith and argument to conversion. And the idealists' answer to the epistemological skeptic is: "Of course. You are absolutely right. There is no method. There is only acculturation."4 25. The pragmatist in contrast, accepts both truths - but he pays neither price. What pragmatism insists upon is the primacy of practice in the order of understanding both of ontological notions (e.g., existence independent of representings) and of epistemological notions (e.g., correctness). The 'ethics of belief' is the essential study of which both classical ontology and traditional epistemology are mislocated transpositions. The concept of correctness does not explicate justification but rather presupposes it. Correctness, in fact, is nothing but the limit of justification, the truth being (analytically) what, in the last analysis, we ought to believe. But if correctness is also (analytically) the adequacy of our systems of representations to the world which they represent, it follows that this world can be nothing but what in the last analysis a correct conceptual frame represents it as being. Neither, then, does adequacy to the world explicate correctness. On the contrary, the correctness of our representings is the criterion of their adequacy to the world, and the measure of their correctness the measure of that adequacy. Thus where the classical dialectic n10ved from formal being (as independence from representings) to correctness (as correspondence) and thence to puzzles about justification, the pragmatist reverses the order, passing from justification to correctness (as complete justification) and thence to formal being (as the representeds of justified representings in the limit). 26. It is precisely the evolutionary aspect of the pragmatist's n1etaphysical picture which rescues him both from the Scylla of ineffability (Peirce: "we have no concept of the absolutely incognizable.") and from the Charybdis of arbitrariness as well
TRANSCENDENTIAL ARGUMENTS AND EPISTEMOLOGY
255
27. It rescues him from ineffability by relocating the absolute or metaphysical correctness of our conceptual schemes in the limit. It is this which allows pragmatism to make sense of 'correspondence' or 'adequacy' without a commitment to foundationalism. For just as organisms with hearts can evolve from ancestors universally lacking them, a world can bring our representations into increasing conformity with it without producing in us any conforming representations ab initio. For a world acts on us not only productively but also selectively. Our representings are brought into increasing conformity with the world by a process of successive selection, from among the representations continuously - and even randomly - produced in us, of those representations which more nearly conform to it than their predecessors, as an environment selects from among random genetic mutations those traits conducive to the survival of the organism and the perpetuation of the species. And, as with biological evolution, the survivors of one phase of the process become elements of the environment into which the mutations driving the next phase emerge. 28. It is this last observation which enables the vindication of pragmatism from the charge of arbitrariness. What corresponds to random genetic mutation in this evolutionary analogy is the passive element of our experience. Our syntheses develop in the face of something. We find ourselves with representations (perceptions) which we then undertake to integrate into our evolving world-picture. Such representations are (necessarily) structured in terms of a particular conceptual frame (Le., all perceiving is 'perceiving as') but they do not on that account automatically cohere with the balance of the worldpicture which we entertain in terms of the descriptive parameters of that frame. They may instead be anomalous or deviant experiences. Their inferential development by means of the conceptual resources of that frame may lead to claims inconsistent with others reached in the same way. Synthesis in terms of a still-further-evolved conceptual scheme is then called for, and, in the course of it, all our representings - both old and new - may (although they need not) undergo a radical restructuring. 29. The pragmatist proposes, then, that these representations with which we continually find ourselves are the (causal) outcomes of the action of a world of reals (to which we belong) on us. And he further suggests that our being so constituted as necessarily to operate with the epistemic end of synthesis results in an increasing adequacy of our representings to that world. Call this complex view 'explanatory realism'. Why should one accept it? The pragmatist
256
JA Y F. ROSENBERG
has an answer. One should accept e~,planatory realism because it provides an account of how a continually developing synthesis in the face of experience is possible. Explanatory realism allows us to unite our picture of ourselves and our successes as cognizers with our developing picture of the 'outer' world which we both represent and inhabit. Briefly put, explanatory realism allows us to naturalize our epistemology. In this way, explanatory realism proves itself an integrative hypothesis. Adopting it enhances our developing synthesis. But that is to say, adopting explanatory realism serves to further an epistemic end which we necessarily have. It is therefore justified. Explanatory realism is thus epistemologically an explanatory hypothesis, and our reasons for accepting it are the same as our reasons for accepting any explanatory hypothesis: It diminishes the mystery - that is, it advances the synthesis. 30. From this point of view, metaphysical theorizing is epistemologically on a par with natural science. E1sewhere 5 I have distinguished two enterprises, conflated in classical ontology, which I call "negative" and "positive" ontology. Negative ontology is the philosophical explication of the categorial vocabulary, its elucidation in terms of epistemology and semantics and ultimately - in terms of families of cognitive conducts and the principles governing them, Le., an 'ethics of belief'. Positive ontology is the attempt to sketch out a picture of the world and our place in it in general terms, to answer the question "What is there?". What I have called "explanatory realism" is an especially gen.era1 answer to this exceptionally vague question. And the defense which I have offered for it identifies the proper epistemic place of such positive ontology. It also brings us conveniently back to the main theme of this conference, for what it shows us, put aphoristically, is simply that positive ontology is natural science. 31. I have been exploring a Kantian connection among our existence as apperceptive discursive intelligences, our necessary commitment to the activity of synthesis, and our espousal of realist frames. Carried through consistently, and supplemented by concepts drawn from an evolutionary analogy, this exploration resulted in our finding a defensible link between two concepts of correctness - correctness as epistemic justifiability and correctness as ontological adequacy. The topic of natural science is also usefully developed in terms of a ttip1e of concepts - explantion, redescription, and understanding - and once again synthesis will prove to be the unifying theme. What I am about to say, in fact, recapitulates in yet another idiom much of what I have already said. Furthermore, I have written fairly extensively on the subject of
TRANSCENDENTAL ARGUMENTS AND EPISTEMOLOGY
257
the epistemology of natural science elsewhere. 6 For these reasons, and for some others, I shall be very brief buf not, I hope, unconscionably so. c
32. If we think of natural science (henceforth sin1ply 'science') as a rational activity, as something which people do, it becomes apposite to inquire into the aims of this activity. What are the ends of which scientific theorizing is in service? One traditional account has it that these ends are the prediction and control of natural phenomena, but this, I think, is mistaken on a variety of counts. It errs by failing to recognize the intellectual autonomy of pure, in contrast. to applied, science. It errs by positing ends which could be adequately served by the accumulation of fragn1entary and unrelated bits of folkwisdom and technology or by the epicyclic refinements of theories deemed scientifically inadequate on what must therefore be other grounds. (That is, it fails to account for the theoretical wholism characteristic of the enterprise.) And it errs by projecting a view of inductive confirmation and falsification at odds with actual practices. Scientific theories are neither falsified by a single ostensible counter-instance nor increasingly confirmed by the mere quantitative piling up of correct predictions but rather gain and lose credibility in quantum-jumps largely as a function of the availability of alternative or con1peting theories and their relationships to those alternatives or competitors. 33. All of this becomes intelligible if explanation supplants prediction and control as the guiding aim of the scientific enterprise. But we must be careful here not to adopt uncritically the Hempelian picture of explanation as deduction, for it is part and parcel of the view of science which I wish to oppose. Instead, our picture of explanation should emerge from a recognition that it is a notion correlative to that of an anomalous or deviant phenomenon, a phenomenon which stands in need of explanation. And anomaly or deviance is a property which appertains to a phenomenon only relative to some alreadyoperative conceptual or theoretical frame. A phenomenon is anomalous or deviant, as we have seen, by reason of its failure to cohere with the balance of the world-picture which we entertain in terms of the descriptive parameters of that frame. The root notion of explanation, in consequence, turns out to be the achieving of a coherent understanding. Its method will be to offer an integrative redescription of the phenomena anomalous under their previous characterizations and, perhaps, of that world into which those phenomena must be fitted as well. The end of science, in other words, is simply the necessary end of all rational activity. It is synthesis.
258
JA Y F. ROSENBERG
34. What legitimizes our adoption and employment of a successor theory is that the reconceptualization of the world which it makes possible permits its reunification as an orderly single world. A passage from predecessor to successor theory is thus a step in the evolution of that continually developing synthesis which was the focus of our earlier concerns. The predecessor theory is the environment into which the nlutations of anomalous experience emerged and in which the new successor frame nlust survive. It survives by explaining its predecessor (s). And here especially explanation cannot consist in an Hempelian deduction, for the successor frame offers a redescription of the world in light of which the principles of its predecessors are seen as literally false. Yet to qualify as a successor, the new frame must be arguably better than the scheme which it supplants. And this 'better' must be the 'better' of an enhanced synthesis. It follows that the successor must accomplish what its predecessor accomplished in point of integration, and more. Put in terms of explanation, this inescapable epistemological constraint on theory succession requires that the successor explain, not the truth of its predecessors (for, from the standpoint of the new framework, they are false), but their descriptive successes and failures. It must explain, in other words, why the world-picture projected by its predecessors, though false, achieved as much integration as it did where it did, and why it failed and thereby engendered anomalies where it failed. 35. A justificatory argument showing the adoption of a particular theoretical framework to be epistemically warranted or legitimate on these grounds is thus necessarily retrospective. It can be constructed only when the successor frame is already in hand. Addressed to episternic conduct, it takes the form ora piece of practical reasoning citing the theory's contribution to our overriding end of synthesis as a ground for the adoption or espousal of that theory. It is just such an argument which Peirce introduce"s under the rubric 'abduction' and, as I have argued elsewhere, it is in fact the sole form underlying any justificatory reasoning, classical 'instantial induction' for one and a Kantian transcendental deduction as well. 7 36. It is the double accountability of scientific theories - to the phenomena as they are conceived in terms of the new descriptive parameters and to their conceptual predecessors and their margin of success in supplying a coherent picture of phenomena conceived under the old descriptive parameters which allows for, and indeed guarantees, the diachronic convergence of representational systems in the limit to that single (absolutely) correct con-
TRANSCENDENTAL ARGUMENTS AND
EPISTEMOLOGY
259
ceptual schen1e which stands to the process of theory succession as a regulative ideal. In the ontological idiom, the explanation by a successor frame of the descriptive successes of its predecessors comes to this: Regularities conceived or posited through the medium of predecessor-descriptions as in the world emerge as regularities of appearance when redescribed in terms of the concepts of the successor frame. (This is nothing but explanatory realism, caught in medias res.) The one question remaining on our agenda is whether every element of descriptive content can undergo this sort of ontological reappraisal or whether, as Kant contends, there are conceptual invariants taking the form of isolable descriptive contents (the schematized categories, perhaps) necessarily shared by any representational scheme which can be ours. Rather than prolong the suspense, let me give you my answer straightaway, Kant was right - almost. 37. Where he went wrong was in setting the level of generality. Of course, in one sense, this has been known for quite a while. Kant's commitment to Euclidean geometry and Newtonian mechanics has long been the subject of relatively facile critical attacks. But I do not intend to fault Kant for having been born in the 18th Century. Our syntheses evolve in the face of continuing experience, after all, and it simply took some time before enough experiences had accumulated in a rich enough conceptual environment to demand theo,; retical alternatives to Euclid and Newton for their explanatory accon1modation. What I do want to address is what sorts of invariants Kant is entitled to posit on the basis of the style of argument which he employs. For the natural response to the facile criticisms' addressed to his historicism is something like this: "Very well. Not Euclidean geometry, perhaps, but at least space. Not Newtonian mechanics, perhaps, but at least causality." The final point which I wish to secure today is that even these conclusions - permissive though they be - are too strong. 38. The operative notion of Kant's arguments is - as always - synthesis, the determinate synthesis of experience in time. We have already granted although I did not reconstruct the argument - a connection between apperceptive synthesis and the adoption of realist frames, that is, of frames projecting a distinction between an object's existence and its being represented (perceived). The location of objects in a unitary (three-dimensional) space is supposed to be' demanded in turn by this distinction, supplying a family of positions where an object can be without being perceived, grounding the distinction between certain real and apparent changes (e.g., of size - dis-
260
JA Y F. ROSENBERG
tinguishing a real change of size resulting from shrinkage of a stationary object from an apparent, perspectival, ch~nge of size resulting from a relative motion of the object away from the perceiver), and providing the basis for explanatory accounts of the failure of a particular object to be perceived by a representer at a time in terms of their positions relative to one another and to other objects. (Schematically, for example, it something visible exists which I nevertheless cannot see, it must either be too far away or there must be something opaque spatially interposed between it and me.) 39. But what this argument shows (it is Kant's First Analogy)., is only that there must be some family of relations universally defined over both objects and representers (including oneself) adequate to fulfill these demands of synthesis. Looking at the result somewhat differently, there must, in other words, be a family of principles of inference expressing relations among possible appearances of objects to representers as a function of determinate coordinates assigned to those objects and representers in a single arena, as I shall put it - but this arena need have no further resemblance to space as we customarily conceive it, an unperceivable three-dimensional continuum amenable to the mathematics of a geometry. Kant's results are thus compatible with, for example, the radical Minkowskian reconceptualization of our world's arena as a unitary four-dimensional space-time. It is one of the virtues of Strawson's construction of an 'analogue to space' for his Auditory World to have shown us dramatically just how much slack Kant's argument genuinely allows here, for his 'master-sound' is analogous to space in precisely the respects which I have canvased. It provides an arena for his auditory particulars. 8 40. The case stands similarly with respect to causation. As a unitary 'outer' arena is demanded by the distinction between existence and being perceived, some analogous grounding is needed for the distinction between (merely) being represented successively and (actually, objectively) being successive. The having of a concept of unitary space (an arena) consists in the possession of rules of inference governing the modes of sensible appearing for objects to representers in that space. Similarly, the having of a concept of determinate order in time (of an objective history) will consist in the possession of rules allowing inferences relating ostensible and actual successiveness or, equivalently, relating possible and actual orderings of appearings (perceptual representations). Kant concludes on these grounds that any world represented as having an objective history will be represented as subject to causal law, for we
TRANSCENDENTAL ARGUMENTS AND EPISTEMOLOGY
261
must posit some thing in or contemporaneous with the predecessor state of an object which is an objectively necessary condition of the existence of the successor state in order that those states be determined in the order of their existence as predecessor and successor by some thing other than the order of their apprehension. 41. But while the argument here (it is Kant's Second Analogy) is as cogent as the preceding one, it shows only that a conceptual fralne adequate to secure a determinate temporal synthesis must posit some system of necessary connections an10ng events or states of its objects assigned to successive positions in time. It must, in other words, supply for its events and the states of its objects a family of inferentially interrelated temporally conditioned descriptions governed by rules of (material) consequence - but these rules of consequence need have no further resemblance to causal laws as they have been classically conceived, reversible Laplacean state determinacies. Kant's results are thus compatible, too, with such radical reconceptualizations of the lawfulness of our word as Planck's quantum jumps, Schroedinger's psi-function, and Heisenberg's uncertainties. 42. It would not be too much of a distortion, in fact, to summarize these conclusions by saying that the level of generality at which Kant is genuinely entitled to operate yields as a conceptual invariant only that any representational scheme which can be ours must be structured by some determinate mathematics. To have established this much is no mean accomplishment. The question of the relation between mathen1atics and the world has, after all, been philosophically problematic at least since Plato produced his Theory of Forms to answer it. But it is not what Kant intended to accomplish. It was inescapable· invariants of descriptive content that he was trying for. What he got, however, was an inescapable diachronically varying mathematical form. Such are the perils of generality . 43. And with that I shall bring this essay to a close. No less general than Kant's Critique, it is on that account no less perilous. Perhaps, however, the benefits of two more centuries have enabled me to see somewhat further than Kant. If so, I deserve no special credit. My debts to Peirce are obvious; those to Wittgenstein and Sellars, although less obvious, are no less real. And, of course, it is only because Kant could see as far ashe did that I even know where to look.
University ofNorth Carolina at Chapel Hill, N. C.
262
JA Y F. ROSENBERG NOT'ES
1 I shall use all of these terms, along with a few others, more or less interchangeably. On another occasion I might want to make some distinctions, but here I am painting with a very broad brush. 2 I am indebted to my colleague Richard Smyth for calling my attention to this way of approaching the Cartesian texts. 3 Essentially this criticism of realism is popping up all over nowadays. It was expounded by Rorty in 'The World Well Lost' (Journal ofPhilosophy 69 (1972,649-65) from considerations advanced by Quine and Davidson and showed up just last year as the main thesis of Putnam's 1976 American Philosophical Association Eastern Division presidential address, argued from a modeltheoretic point of view. 4 A good picture of idealist science, in fact, is given by the emotivist ethics of logical positivism. In addition, it ought to be remarked that the idealists' views here are internally incoherent, for the question of whether my practices are idiosyncratic or cohere with those of my epistemic community itself reintroduces the demand for objective correctness. But, of course, idealists may not be particularly troubled by such incoherence. Consistency, after all, is only a demand of my epistemic subculture. Many religious systems, for example, will have no truck with it. S Especially in 'The "Given" and How to Take It - Some Reflections on Phenomenal Ontology' (Metaphilosophy 6 (1975), 3-4, pp. 303-37). 6 Especially in Chapters IV and V of Linguistic Representation, Reidel, Dordrecht, Holland, 1974. 7 For instantial induction, see Chapter V of Linguistic Representation. For transcendental deductions, see especially 'Transcendental Arguments Revisited' (Journal of Philosophy 72 (1975), 611-24). 8 I have explored Strawson's Kantianism, its strengths and its failures, in detail in 'On Strawson: Sounds, Skepticism, and Necessity', Philosophia 8 (1978).
ROLF P. HORSTMANN
CONCEPTUAL SCHEMES, JUSTIFICATION AND CONSISTENCY
Comment on Rosenberg The interesting attempt by Prof. Rosenberg to explain a whole series of very important and much discussed epistemological problems through explanatory realism calls for attention and critical examination. It is difficult to resist his plea to accept explanatory realism as a position which, by linking sound elements of a transcendental theory going back to Kant with beliefs rooted in pragmatist traditions, resolves a great number of difficulties for which there had not been very convincing solutions up to now. Another point in Rosenberg's favour is that he does not simply recommend his proposition by presenting the results to be expected from explanatory realism, but at the same time secures its validity by a prognosis on the ·price which inevitably would have to be paid in case of any alternative position. This price is supposed to consist in having to accept either arbitrariness or ineffability and incognizability which at least for philosophers is a high price to pay indeed. Whenever we are threatened with an extremely high price, however, we should ask how such a price is arrived at. Is it just fictitious or is there a real basis to it? Is it put up high only to make other things seem less costly or are these other things really less costly? I would like to examine a certain suspicion raised by the questions just asked. My remarks will be directed not so much at a critical analysis of the expected positive results of explanatory realism, but rather at an attempt to discuss Rosenberg's reasons for his gloomy prognosis relating to alternative positions. Rosenberg believes that in using conceptual schemes for the purpose of knowledge, there are at least two questions to be answered, Le. (1) how can their usage be justified and (2) how can one test their correctness. As I see it, his thesis is that a Kantian transcendental position does in fact present an appropriate antisceptical procedure to justify using conceptual schemes; but that such a position does not contribute anything worthwhile when it comes to deciding on the question of correctness of a conceptual scheme since it deals with the idea of a noumenal world which in principle is supposed to be beyond our knowledge. The inability of a transcendental theory to make a satisfactory contribution to the question of the correctness of a conceptual scheme becomes especially apparent at the very point where its considerations in relation to this question open the door to a host of idealistic approaches.
263 P. Bieri, R.-P. Horstmann, and L. Kruger (eds.), Transcendental Arguments and Science, 263-269. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
264
ROLF P. HORSTMANN
Therefore, a theory has to be found which not only does provide a tenable method of justification, but also an antiidealistic criterion for the correctness of a conceptual scheme, and this theory is to be explanatory realism. So much for an inevitably summary exposition of the thesis, I am going to discuss. The strength of the whole argument is, of course, closely linked to the validity of the criticism of the transcendental approach just outlined. I would therefore like to examine a little more closely Rosenberg's evaluation of the transcendental position. I shall begin by discussing the problem of justification. The fact that for Kant a justification for using a conceptual scheme is tied up with his theory on the necessity of synthesis and how this is done, has been clearly shown by Rosenberg: he has reconstructed the reasons for Kant's contention that we have to regard knowledge essentially as the product of a synthetic activity. He shows that this can only be achieved by us as apperceptive and discursive intelligences, if we assume the existence of a certain conceptual scheme. Rosenberg interprets Kant's idea thus reconstructed in such a way that the relationship between knowledge, conceptual scheme and synthesis as seen by Kant seems to in1ply that any conceptual scheme can be justified as long as it achieves synthesis, especially that when there are several conceptual schemes it should be considered justified to adopt the one whose synthetic power is greatest or, in Rosenberg's words, the one which allows for enhanced synthesis (cf. p. 251f.). If Rosenberg's conclusions are correct, Le. if one could reduce the transcendental approach as regards the question of justification to the proof of synthetic power as the only justifying reason, then Rosenberg ·would be right in being concerned that in this way the question of correctness of a conceptual scheme could only be decided in an idealistic manner. My opinion is that at this point Rosenberg may have fallen victim to a misunderstanding because precisely by adhering to his own reconstruction of the Kantian transcendental position, the only thing that follows from the relationship between knowledge, conceptual scheme and synthesis as seen by Kant is that we need conceptual schemes for (internal and external) synthesis. In other words, without conceptual schemes there can be no synthesis (at least by us) and also, the existence of a conceptual scheme is a precondition to the possibility of synthesis. The proof, Le. the justification of the contention that some conceptual schen1e is a necessary condition for the possibility of synthesis, does not, as I see it, say anything on the possibility of justifying any particular conceptual schen1e to become the one we should use. If this were not so, I could for exan1ple - fron1 the justification of the statement that the existence of a written text is a necessary condition for the possibility
COMMENT ON ROSENBERG
265
of reading - gain a justifying reason for reading a specific text, i.e. this paper. In my view, what is justified by the Kantian idea is not the right to use a particular conceptual scheme on account of its synthetic power; but the claim that we need a conceptual scheme for the purpose of knowledge because we cannot achieve synthesis without it because of our particular nature, i.e. being apperceptive, discursive intelligences. If what I think is correct, the problems of justification and correctness when considered in the context of a transcendental approach, appear in a different light compared with Rosenberg's presentation of them. If, from grasping the relationship between conceptual scheme and synthesis we do not gain anything which might be related to the justification of using a specific conceptual scheme, and if, at the same time, questions of justifiability of a certain conceptual scheme are in fact of crucial importance to Kant's transcendental doctrine, it is to be expected that in any discussion on justifiability the idea of the correctness of a certain scheme will have to take its place. I will briefly follow up this assumption and try to show that the Kantian position, in particular, is characterized by the fact that it does not tackle the problem of justification of a conceptual scheme separately from the problem of proving the correctness of such a scheme and that 'correctness' in this respect cannot be reduced to an idealistic concept of 'coherence' . The procedure of Kant's consideration on this point could be described as follows: a conceptual scheme can only be justified, if it is correct. It is correct, if its employment allows for an adequate knowledge of the world. In order to decide on whether it is correct or incorrect in this sense, it is sufficient to examine it with a view to fmding out whether the results of the scheme's application are in fact adequate in relation to the world as we find it. When a conceptual scheme has been found to be correct in this sense, it does not necessarily mean that it is justified. All we know is that its employment does produce correct results; but we do not know why. As long as this question remains unsettled, we cannot be sure that by using this scheme we obtain correct results under all conditions the scheme is used for. A transcendental justification of the employment of a scheme of which we know that it will produce results which cohere with the world, would consist precisely in giving those reasons for the already established correctness of the scheme which show that only this and no other scheme is correct in the sense discussed above. There may be different reasons of this kind; the reason Kant puts forward is as follows: if it can be shown that a scheme not only does fit the world, but also that without employing it, we would not be able to have
266
ROLF P. HORSTMANN
a concept of the world as it presents itse1:f~to us - then this would be the only scheme possible to us. Having put forward this idea which has been reconstructed here in a very casual manner without much attention to Kantian details, nothing has as yet been said on the correctness or incorrectness of the scheme which Kant actually preferred. The point I wanted to make was to characterize the Kantian strategy of a transcendental justification in order to show that, for a transcendental position, the idea of justification is pointless without the assumption of correctness in the sense of adequacy. Relating back to the reasons for my remarks, this result seems to me to indicate that Rosenberg's argument for the necessity of replacing a transcendental position by a pragmatist version is not fully convincing. I suppose that Rosenberg would oppose my doubts and insist that I have missed at least one point of his argumentation. With regard to the question of correctness, his thesis is, you will remember, that Kant would run into difficulties because his distinction between a phenomenal and a noumenal domain as well as the alleged limitation of the possibility of knowledge to the phenomenal domain would limit the scope of any valid discussion of correctness to this area; i.e. correctness in the fullest classical sense - meaning noumenal correctness - would by definition be beyond the scope of the transcendental approach. At this point, however, one should ask what is meant by the thesis of the impossibility of noumenal correctness. If by that is meant as I think it is - that there is no way of determining the adequacy of our ideas in relation to the noumenal world, then it is certainly correct; but in the context of Rosenberg's argument this statement can serve to describe a deficiency only on the assumption that the noumenal world is in truth the real world, Le. a world in which the objects which to us are just appearances, do in fact exist as they really are. But as you know, Kant takes the view that Noumena are the empty, though not contradictory concepts of an object which cannot have any objective reality anyway. So it is not plausible - at least within the fran1ework of the Kantian construction - to regard precisely the noumenal domain as the true reality. If on the other hand the thesis of the impossibility of noumenal correctness means that the adequacy of the appearances in relation to the object, to which I have to relate them, i.e. to the so-called transcendental object, cannot be stated, one should bear in mind that the transcendental object is not an object which is real in a specific sense, but that it merely characterizes the necessary idea of the unity of an object to which we are obliged on account of the unity of consciousness. So in both cases the contention of the impossibility of noumenal correctness is not a
COMMENT ON ROSENBERG
267
matter which does in any way tou~p upon the question of adequacy in relation to the world. I have tried to point to some difficulties which have to do with Rosenberg's explanation of the need to transform a transcendental position into a view called 'explanatory realisn1'. But even if this explanation which is based on an analysis of the deficiencies of Kant's ideas regarding justification and correctness, leaves some questions unanswered, there are plenty of other reasons available to us which in the current debate on the limitations and deficiencies of a transcendental position of a Kantian type carry some weight and may suggest the adoption of explanatory realism. Two of these reasons figure in Rosenberg's paper. The first one concerns the difficulties which would arise for Kantian theory because of its alleged radical option for Euclidean geometry and Newtonian mechanics. The other one has to do with the much regretted inability of the Kantian theory to make a useful contribution to the more recent problem of scheme-succession - be it for the simple reason that this theory cannot accomodate the idea of more than one conceptual scheme. One possibility of demonstrating the superiority of explanatory realism over a purely transcendental position n1ight well consist in examining the specific results of explanatory realism in overcoming the supposedly neuralgic points of the transcendental theory. Here I will take up only the second point, Le. scheme-succession. 1 Rosenberg's view of this question is briefly this: a conceptual scheme is correct only, if it is epistemologically justifiable and if it fulfills (at least in part) the condition of adequacy to the world. If we have two schemes, one of which is a predecessor-scheme to the other and if this is supposed to mean that the successor-scheme is n10re correct than the predecessor-scheme, then the claim to greater correctness of the successor-scheme must be argued for. This is to be done (1) by showing that the successor-scheme makes possible enhanced synthesis and (2) by showing that enhanced synthesis "results in an increasing adequacy of our representings to that world" (p. 255). The question is: what do these two demands mean and how can they be met? As regards the first, I think it cannot be answered without clarifying the second. For, if one does in fact make epistemic justifiability °a constitutive part of proving the correctness of a conceptual scheme and considers a scheme as epistemologically justified, if it can be shown that it is in fact being used, that is: permits synthesis, - then the question of whether it permits enhanced synthesis becomes one which cannot be decided without an appeal to the degree of adequacy. And it is precisely this combination of epistemic justifiability and ontological adequacy which Rosenberg considers a special
268
ROLF P. HORSTMANN
accomplishment of his position. As far as the second is concerned, Rosenberg's main idea is that ontological adequacy of a scheme is a function of the consistency of inferences drawn from the experiences themselves interpreted within a specific frame. Thus the degree of adequacy depends on the degree of consistency which can in this respect be used as a criterion for the (retrospective) evaluation of the correctness of successive schemes. Apart from some more or less technical aspects concerning the Rosenberg model, for instance how to make reliable statements on the degree of consistency of successor schemes or how to apply this model to cases of schemesuccession where the successor-scheme has been successful not because of its (proven) greater consistency, but because of the expectation of greater consistency - perhaps on the basis of some metaphysical assumptions - it remains to be clarified which are the preconditions required by this point of view. It is assumed explicitly that we live in a world which is not only responsible for our having representations, but requires us to have representations of a specific kind, i.e. which at least in principle are open to inferential development. For, if this were not so, it would be impossible for us to make plausible the use <;>f the consistency criterion which according to Rosenberg is a constitutive element when dealing with scheme-succession. But how to justify such an assumption on the relationship between world and representation? One possible answer would be to say that because consistency is in fact our criterion for the degree of adequacy of a conceptual scheme, we cannot help assuming that the possibility of inferential development of our representations is safeguarded by the nature of the world itself. Apart from the fact, however, that such a dictum raises more questions that it answers, it would in this context be question-begging. Another possibility would be to say that because of our nature as apperceptive, discursive intelligences which implies the necessity for synthesis for the purpose of knowledge, we can only know a world which is in accordance with the conditions of unity reqUired by synthesis, and these conditions are such that they should contain the postulation of consistency which itself presupposes the possibility for inferential development. If one chooses this possibility, then a justification has been given for the consistency criterion, but without any reference to the nature of the world. In other words: we are back at a transcendental standpoint which now enables us to demonstrate something which by way of explanatory realism is only presented as a fact: namely (1) that there is and why there is a connection between consistency and adequacy, and (2) why, in connection with scheme-succession, consistency can be regarded as at least one but maybe not the only criterion for the degree of correctness.
COMMENT ON ROSENBERG
269
So, in the very case of scheme-succession, too, I can't see that Rosenberg's explanatory realism does in fact irilprove on a transcendental position. It seems to be a point of view which does not diminish mysteries as it claims to, but rather s~ips a lot of questions - not so much by answering them as by not asking them. So it might be that Rosenberg sees somewhat further than Kant, but perhaps in the wrong direction. Universitiit Bielefeld NOTE 1 Concerning geometry compare J. F. Rosenberg 'Transcendental Argument Revisited' Journal of Philosophy 72 (1975), 611-624, and R. P. Horstmann, 'Space as Intuition and Geometry' Ratio 18 (1976), 17-30.
HARALD PILOT
COMMENT ON ROSENBERG
Professor Rosenberg proposes to answer the challenge of scepticism against the possibility of empirical knowledge by a construction of an epistemic process, which legitjmizes the acceptation of empirical theories. Professor Rosenberg tackles the challenge of scepticism which presents itself for him in the form of the following dilemma: (1) (2)
If empirical knowledge is not certain, but revisable, then it has to be arbitrary. If it is certain, then it has to be necessary and thus not allowing to explain the possibility of error.
Since I don't wish to argue for certainty, nlY nlain theme will be the relation between revisability and arbitrariness. In order to characterize enlpirical statements and theories as revisable without being forced to accept them arbitrarily, we will have to justify them. This justification is one of ado in g, because every statenlent is an action of representing, is the generation of the image of an object, - as in the case of the action of painting a picture. The rulegoverned behavior of uttering a statement is like the rule-governed behavior of a painter. It is purposive behavior and can be evaluated. Now we have to distinguish between the prospective and the retrospective justification of an action. The prospective justification refers to the reasons to perform an action and legitimizes it before it is done, while the retrospective justification evaluates an action after its performance. It is this distinction we have to keep in mind when we ask whether an empirical statenlent is justifiable or not. There might be empirical statenlents (regarded as actions of representing), which are not justifiable be for e they are perfonned, but which are justifiable retrospectively. Their retrospective justification may be possible, since there were circumstances we could or did not know at the time of our acting. To apply a theory is to perform a set of representing actions. To justify a theory then nleans to justify such a set of representing actions. And there may be a retrospective justification of, so to speak, theoretical acting, which was not justified at the time of its performance. But if to use a theory means to perform representing actions and if it is impossible to justify these actions prospectively, then there is no possibility of justifying the use of a theory at 271 P. Bieri, R.-P. Horstmann, and L. Kruger (eds.), Transcendental Arguments and Science, 271-275. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
272
HARALD PILOT
all. For the justification would only be possible after the performance itself. But then in performing a statement belonging to the theory, we cannot know whether it can be justified or not, since we have to wait for the successor theory to justify or reject the previous performance by retrospective evaluation. An analogous argument holds for the justification of rules of actions, since the language game of rules for a set of actions must consist of actions expressing the relevant rules. But if there is not a single action of the language game of rules being justified now, then we also have to wait for the next language game that justifies our first language ganle of rules. From this it follows, that each sentence - token we use to make a statement about some things, cannot be a justified performance of the statement. If there is only retrospective justification, there is no justification of the actual performance of a statement at all. And if there are no statements without any actual performances of them, there will be no justified statements either. Thus, if theories are systems of types of representing actions and if these actions can be justified only retrospectively, then the use of theories for representing things must be completely arbitrary, in case there are alternative theories. We may suggest that there is a t y p e of transition from a predecessor to the successor theory which legitimizes our accepting the successor and rejecting the predecessor. This is the epistemic process that Professor Rosenberg calls the unit of science. If there is such a legitimate epistenlic process, which is instantiated in every legitimate transition from a predecessor to a successor theory, then we must be able to identify those transitions which are of the relevant type and must have a set of criteria fulfilling this function. To investigate the possible criteria, we have to consider two cases: (1)
(2)
The historical order of the sequence of the predecessor and successor tlieory is a sufficient condition for the acceptance of the successor. In this case I cannot see the possibility to decide between theories which are proposed at exactly the same tinle. Their historical order is not a sufficient condition. In that case we need criteria, which can be applied to the theories themselves regarded either a~ representations of real things or as actions of a certain type. The criteria of the first type would be satisfied only, if the theories have characteristics, which are describable by another theory. The criteria of the second type must be rules of the representing actions. If we choose the first possibility, we are presupposing a universal theory of representation, which
COMMENT ON ROSENBERG
273
allows us to check every pair of theories; if we choose the second possibility, we are presupposing a universal 'ethics of belief' . The first set of criteria seems to come from a universal theory of evolution, whereas the second type is given by the necessary end of every rational agent to get perfect knowledge of the circumstances of his actions, - provided there is a theory of the evolution of knowledge. But if the justification of a theory has to be retrospective there is no general theory of evolution, which can be justified now and we have to n1ake an arbitrary decision between alternative theories of scientific revolutions. And if there is no theory of the evolution of knowledge, the search for perfect knowledge does not make sense, since there cannot be partial knowledge that might be refmed. The final possibility seems to be to construct on the one hand a regulative ideal of correctness, which is satisfied by the ideal theory, which gives us the standard for the relative evaluation of theories and on the other hand a set of rules for the actions of a purely rational agent, who is able to perforn1 his actions in the full knowledge of all circumstances. This agent is the last result of something like natural selection and therefore can only be characterized if we understand the ideal theory. Hence the rules of the actions of such an agent are given to us only if we have the ideal theory., J believe, this means that the universal theory of evolution itself must be the ideal theory which allows to justify all the other theories. Then the theory of evolution must be the theory to which all the other theories are approximations. The series of approximative theories, however, has to be infinite. For it is defmed by the postulate that every theory has a distinct follower which allows its retrospective justification. Thus the ideal theory cannot be the last successor, since there is none. We cannot even know whether one of our present theories is the ideal one or not. For we have no standard of evaluating it without knowing the ideal theory. And we have no rules guiding the actions of a purely rational agent, because these rules are detern1ined by the ideal theory which we do not know now. But the ideal theory might be something like the limiting point to which an infmite series approximates. If this is the case and theories (as representational systems) are parts of the world, then we must be able to know now that the world satisfies precisely this structure. In other words, we must be able to show that the world satisfies a certain mathematical structure om nit em p 0 r alI y. Now, if truth is explicated by correctness and correctness by justification, it follows that there is at least one set of statements
274
HARALD PILOT
about our world that must be justifiedbat all times (past, present and future): those statements that are asserting that the sequence of theories in the world satisfies the necessary conditions of a converging series. If these statements were not justified now, no retrospective justification of any empirical theory belonging to the sequence would be possible. For the justification of a predecessor of that theory we are using is legitimate, pro vi d e d that the employed theory is justifiable. This is only the case, if the used theory is a melnber of the \vhole series of theories, which approxin1ates to the ideal theory. Knowing then that a predecessor theory has been justified entails that we know that there is a converging series of representational systems in our world. But this seems to be a justifiable claim only for someone using the ideal theory itself. If we identify the ideal theory with the limiting point a series approximates to but never reaches we must use the ideal theory to assert that our world does satisfy the structure of the series. If we - as pointed out above - identify the ideal theory with the last term of an infinite discrete series, we get a contradiction. So neither way will work unless we have a framework that is independent of particular empirical theories. Does this mean that we have to choose between either (a) absolute certainty, or (b) arbitrariness, or (c) the notoriously difficult notion of the synthetic a priori? I think it doesn't. We may distinguish between two kinds of justification: a justification of empirical theories on the one hand, and a justification of the necessary conditions of representing within our world on the other hand. The second type of justification concerns the categorial framework in terms of which every representational system has to be constructed. The necessary conditions for a representational systen1 are pragmatically necessary statements, which are justified by a dialogue with the radical sceptic. This type of justification is the success of pragmatically necessary statements against the challenges of the sceptic. Professor Rosenberg presupposes that there are past, present, and future theories in order to distinguish between prospective and retrospective justifications; he further presupposes that there is a world in which our representing is taking place and that there must be representings, if we state 'something about the world (the same world, in which the representings fulfill their function of representing). At least these points must be admitted by the sceptic who is trying to show that the criteria for the acceptability of statements do apply to their negations as well, and that therefore those criteria must be rejected. But if the sceptic has to accept the necessary conditions of making any statement, there seems to be a possibility for the justification ~~~~~t~
COMMENT ON ROSENBERG
275
those necessary conditions. For if the sceptic is uttering a sentence that seems to be the negation of the necessary, conditions of stating anything, he t r i e s to make a statement, and we can show that he does not succeed. However we cannot show and should not try to show that the sceptic makes a false statement, since this entails that he makes a statement. There is no possibility of demonstrating that the claims of the radical sceptic are false. What can and must be shown is that the sceptic cannot succeed in trying to deny the necessary conditions of stating. In order to decide whether we have found necessary conditions of stating or not, we have to try to negate them. In doing so we have to be sceptical ourselves. For the challenge of the sceptic is the only possible test for the necessary conditions of representing. This test is performed by epistemologists, and it is apr e sen t undertaking. There are no past or future theories that are not representable now. The past theorizing is the now represented past theorizing, and the future theorizing is the now represented future theorizing. I think this statement to be a conceptual truth which can be justified against all challenges that the sceptic may put forward. Needless to say, this does not imply that we do now know all past or future theories. But it does imply, that there can be no future representational system, which entails that our representing of this future system was im po s sib 1e , because then it would not have been future for us. We may justify the statement 'I am representing a possible world that is accessible to a world, in which I am representing' by showing that its negation is a pragmatically contradictory statement. Then we may justify, that our representing has to satisfy some minimal spacetime conditions, if it is possible in the actual world. At the end we get a set of necessary conditions for statements. Each condition of the set must be defendable against the sceptic. There can be error, which is corrigible in the future, but there can be no error, which is corrigible by empirical tests only. It seems to me that the challenges of the sceptic are pragmatic contradictions (if they are supposed to be statements) and can be used to justify our clain1s that there are such and such necessary conditions of all representational systems. We might investigate the transcendental arguments in the Oitique ofPure Reason from this point of view and try to construct the synthetic a priori statements as pragmatically necessary statements being necessary conditions of our representational system that is part of the actual '\Todd, all other representable worlds n1ust be accessible to.
Universitat Heidelberg
BARRY STROUD
THE SIGNIFICANCE OF SCEPTICISM
One of the topics announced for this symposium is the contrast between two different approaches or tendencies in philosophical studies of the foundations of science. On the one hand there are those who would abandon the quest for a general justification of empirical knowledge in favour of a purely naturalistic study of the procedures actually employed by scientists and other knowing subjects. On the other hand there are those who take seriously the challenge of philosophical scepticism and, seeing that it cannot be met by a straightforward Cartesian or "foundationalist" theory of knowledge, resort to so-called "transcendental arguments" to show that certain concepts or principles enjoy a privileged status in our thought because without them no human knowledge or experience would be possible at all. These two alternatives are not exhaustive, and it is certainly inlpossible to say something definitive on the dispute between them, but I want to discuss one inlportant respect in which it seems to me that the second approach is superior to the first. Whatever one might think about the prospects for the success of transcendental arguments in epistemology, the need that is felt for them, and the consequent search for plausible candidates., seem to issue from a finer appreciation of the potential force of philosophical scepticism and of the considerations that have traditionally been thought to lead to it than is true of the first approach. And that is one important respect in which it seems to me that the second approach is superior, since I think we must somehow CODle to terms with the threat of philosophical scepticism if we are ever to get the kind of understanding we seek of the nature and possibility of human knowledge. In this paper I will try to make that claim as plausible as I can while remaining at the regrettably lofty level of discussing only general "strategies" or "positions" in the theory of knowledge. And even at that I will only be able to offer SOlne questions to be discussed, or some issues to be raised, and not a set of philosophical theses to challenge, or even to compete with, some of the weighty doctrines already offered us. In fact, part of what I want to talk about is the very nature or status of a pWlosophical thesis or doctrine. I think that is something that has too often been taken for granted, both in general and, in particular, in much of the discussion of this conference.
277 P. Bieri, R.~P. Horstmann, and L. Kruger (eds.), Transcendental Arguments and Science, 277-297. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
278
BARRY STROUD
A philosophical concern with the "foundations of science" is addressed to the question of how, in general, human knowledge or science is possible. The question is not simply how we know this or that sort of thing given that we know certain other things, or how we know everything of a certain sort given that we know things of another sort, but how it is possible for us to know anything at all. The problem is given its special character by two innocuousseeming facts: (1) most, if not all, of what we know is known somehow by means of the senses and (2) things can be different fron1 the ways they are perceived to be. Any realistic account of human knowledge starts fron1 these truisms; they give rise to the philosopher's preoccupation with the question of what we perceive, and with how, on that basis, we can come to know everything we ordinarily take ourselves to know. The project of Descartes' Meditations is perhaps the clearest and most famous expression of this concern, but it has been at the heart of philosophy in one form of another since then, and it dominated the subject for most of the first half of this century. One need not believe that the philosophical question ab out the possibility of human knowledge in general is perfectly clear or legitin1ate, or even fully coherent, in order to have the quite natural expectation that no satisfactory answer to it could end up implying that all or most of what we normally regard as knowledge is not knowledge at all, but at best merely strongly-held belief, guess-work, projection, or fantasy. A long tradition of failure to answer the philosophical question satisfactorily in this respect has led in recent years to a livelier awareness of the special and problematic character of the whole enterprise of giving a philosophical account of human knowledge than was true forty years ago when most philosophers were busily engaged in it. It is now widely believed that the traditional concern with human knowledge and its "foundations" has finally been superseded. The reasons for the .change have never been made very explicit, or to my mind very convincing, but the "theory of knowledge" in its more or less traditional guise - and the theory of perception in particular - has certainly not been much in fashion in Anglo-American philosophy for the last twenty years or so. Now I believe this change of fashion is not based on a proper understanding of the traditional epistemological enterprise and its inadequacies. Even if, as I believe, recent defenders of transcendental arguments have failed to meet the full force of the traditional question, they !1evertheless remain closer to the right track than their opponents who would deny the force of that question altogether or would deny the need to take it seriously.
THE SIGNIFICANCE OF SCEPTICISM
279
II
In the face of the scepticism that can seem such an inevitable outcome of the traditional epistemological enterprise a certain robust approach has always had its appeal. In 1763 Samuel Johnson thought he had refuted the philosophy of Berkeley by "striking his foot with mighty force against a large stone, till he rebounded from it" while uttering the words "I refute it thus".1 In 1939 G.E. Moore thought he had removed from philosophy the alleged "scandal" that "things outside us" had never been "proved" to exist by holding up his hands and making certain gestures while uttering the words 'Here's one hand, and here's another'. 2 I think it is difficult for anyone with some understanding of Berkeley's philosophy, or with some grasp of the philosophical demand for a proof of the existence of things outside us, to believe that these performances were philosophically successful. But I also think it is very difficult to describe accurately and illuminatingly precisely how and why they fail, and what, if anything, they do really achieve. I don't think this problem, or its difficulty, has been widely appreciated. In his "proof of an external world" Moore held up his hands and said 'Here is one hand' and 'Here is another', and he concluded that therefore there are at least two external things. And since, as he claimed, he certainly knew his "premisses" and concluded validly from then1 that external things exist, he thought he had proved that there is an external world. 3 I have said that I think it will be felt that this "proof' does not amount to a demonstration of the falsity of philosophical scepticism or of the truth of an affirmative answer to the philosophical question of whether we can ever know whether anything exists "outside us". Here I do no more than appeal to your own sense of the issue, since I confess it would be difficult to demonstrate its failure to someone who feels no deficiency in Moore's "proof'. But I ask you to consider whether your initial response is not that, considered as an answer to the philosophical question of knowledge, Moore's assertions are unjustified - in that context they are simply dogmatic and without probative force. It is known at the outset of the philosophical inquiry that people frequently think and often claim they know things in circumstances as "favourable" as those in which Moore found himself. The question is whether they do know what they think they know on those occasions, and if so, how. The considerations which have traditionally been thought to lead to a negative answer must somehow be dismissed or accounted for before the philosophical question will have been answered or exposed as i1legitin1ate. Even if Moore's
280
BARR Y STROUD
conclusion is in fact true and is incompatible with philosophical scepticism, I think we still feel that he has not actually refuted that view and shown that his own "common-sense" view is the answer. I feel, and I hope many of the rest of you feel, that scepticism might be wrong, but Moore hasn't proved that it is. Of course, this is no argument against Moore, but it does describe a natural reaction that needs accounting for. Some philosophers who found themselves with this reaction argued that Moore is not simply asserting that he knows certain things and thus begging the question against scepticisn1; he is really doing something else, such as indicating only that it is a perfectly correct use of language to assert, of certain things, that we know them, 4 or "recommending" that the word 'know' continue to be used in certain ways, S or some such thing. Moore steadfastly denied these interpretations. 6 He thought he really did know that there were human hands before him, and that he had proved the existence of the external world in the best way available for proving the existence of any "external thing". But even if these and other objections 7 could be shown to be distortions of Moore's real project, we would still not have an accurate positive description of his "proof' and what it achieves. We are still left with the original problem - how could Moore have put forward a "proof' that is so obviously inadequate as an answer to thy philosophical problem of our knowledge of the external world? We can perhaps take some steps towards an answer by distinguishing two different questions about Moore's "proof' which are not usually asked separately, probably because a certain answer to one of them is taken for granted. (i) Is what Moore says in his conclusion, viz., that there are external things, and that he knows that there are, actually inconsistent with the sceptical phiIosophical view that nobody knows whether there are any external things? (ii) Does Moore really establish his conclusion? If we feel, as we do, that Moore has not refuted philosophical scepticism, and we conclude that therefore there is something wrong with Moore's "proof' and so the answer to the second question is 'No', I think we do so only because we have already answered 'Yes' to the first question. We have taken it for granted that Moore's assertion that he knows there are external things is straightforwardly incompatible with the philosophical sceptical view that no one knows whether there are external things. But is that true? Obviously it depends on what Moore was doing in his "proof". It might well be that there is nothing at all wrong with it as a proof, and yet it does not answer the philosophical question about our knowledge of external things because its con.~
THE SIGNIFICANCE OF SCEPTICISM
281
elusion is not in fact incompatible with any philosophical answer to that question. 8 It is precisely the recognition that the philosophical question is not being settled one way or the other by claims like Moore's, made in the best possible circumstances for kn.owing things about the world around us, that leads to doubts about what the philosophical question really comes to. Moore can perhaps be seen as not understanding any question about what we know except those particular or general questions which are answered - or whose answers are asserted - in just the way he asserted that he knew there were two human hands, and therefore two "external things", before him. 9 He says 'I know - and have actually proved - that there are at least two external things', and the philosophical sceptic says 'No one knows whether there are any external things'; but must what Moore says, taken as he understands it, be incon1patible with what the sceptic says, taken as he understands it? We can perhaps appreciate that the answer to this question could be 'No' even if we do not fully understand, and are not able to state unambiguously, precisely what the philosophical sceptic is saying. III
This, in effect, is the position of Carnap and other logical positivists. Carnap thought that the philosophical question about the existence of external things is a "pseudo-question", neither an affirmative nor a negative answer is empirically confirmable, so the question is, quite literally, "meaningless".10 That is not to say that there are not perfectly meaningful questions about the existence of "external things" that can be asked and answered. rrhey are raised every day in science and in ordinary life, but they are "empirical" questions with ascertainable answers. For that very reason they are what Carnap called "internal questions", questions asked "from within" the framework of the "thing language" and answerable one way or the other by recognized empirical procedures. 11 The completely general or "external" question of the reality of the "thing world" itself, Carnap held, "is raised neither by the man in the street nor by scientists, but only by philosophers" .12 That is the question that is meaningless because all putative answers to it are empirically unverifiable. Now although Carnap regarded the philosophical question itself as meaningless, he did show enough understanding of it, or of the considerations out of which it arises, to recognize and appreciate something that Moore seemed to miss completely, viz., that the philosophical question of the external world
282
BARR Y STROUD
is not settled one way or the other by the scientist's discoveries, however legitimate they might be, or by the man in the street, however carefully he proceeds. And that is true even if that man in the street happens to be G. E. Moore holding up his hands before an audience of philosophers. Carnap could regard what Moore says and does as perfectly straightforward and legitimate and, for that very reason as irrelevant to any philosophical issue about the external world. He thought that if two geographers set off to determine whether a certain mountain they have both heard ab out is real or only legendary, they will settle the question empirically and come to agree about its existence or non-existence without thereby settling a philosophical dispute between, say, "realism" and "idealism" .13 "Realism" and "idealism" are presumably incompatible philosophical views, but the geographers will settle the "internal" question of the existence of the mountain without disagreement. Having discovered in the same way a second mountain, they could go on t6 say 'Here is one mountain' and 'Here is another', or even 'There are at least two mountains'. And I think it would also be perfectly legitimate for them, in summarizing what is now known about the earth, to say 'There are mountains', or even 'We know there are mountains'. Taken "internally", as a report of their empirical discoveries, there is nothing wrong with that. Carnap would add that there is nothing philosophical about it either - it leaves the "external" dispute between "realism" and "idealism" untouched. If that is so, there would seem to be no objection to going on one more step from what the geographers said they knew. No one would deny that a mountain is an external physical object. Perhaps sudden thoughts or natural numbers are not external physical objects, but surely mountains are, if anything is. But if we know, as we do, that there are mountains, and we know, as we do, that mountains are external physical objects, then surely we know that there are external physical objects. In making a list of the things we know to exist it would be perfectly legitimate to put external physical objects on the list. As far as I can see, there is no reason for Carnap or any other philosopher to object to any of this if it is taken (so to speak) "internally", as a report or summary of things we know about the world. 14 And that suggests that there is a way of taking Moore's "proof' in which it too is perfectly legitimate. There is no nlistake or unjustified assertion at any step of the proof, the prenusses really do imply the conclusion, and Moore really does establish that conclusion. But this way of taking it requires that we see that conclusion as "internal" in something like Carnap's sense, and so we can grant the full legitimacy of the proof without in any way abandoning our original idea that
THE SIGNIFICANCE OF SCEPTICISM
283
it is completely irrelevant to the phi1o~ophical problem of our knowledge of the external world. If Moore's interest in what we know, like that of the two geographers, was confined to those particular or general questions that are settled in the way he rightly asserted the existence of two human hands, then perhaps it is not surprising that what he said is irrelevant to any epistemic question other than those. And if it is irrelevant, then his assertion that he knows there are external things is not straightforwardly inconsistent with the sceptical philosophical view that nobody knows whether there are any external things. To suppose that it is inconsistent with that view is to suppose that there is a straightforward deduction from 'Here is a human hand' to the philosophical thesis 'There are external things'. Carnap's way of blocking that deduction was in effect to deny that there is any such proposition as the philosophical thesis 'There are external things' or 'We know there are external things'. The question whether philosophical "realism" or "idealism" is true cannot possibly be settled by any empirical verification procedures; it is a "pseudo-question", and any putative answer to it is "meaningless". Therefore one cannot infer from "There are at least two mountains",. which is settled empirically, to "There are external things", taken as an assertion of some philosophical position. I suggested that there is no reason for Carnap to deny that in certain circumstances one might well assert that there are external things, or that the existence of external things is something we know, and that one n1ight do so on the basis of having established that there are mountains, or that we know that there are. But of course in those circumstances the assertion that there are external things is not to be understood philosophically. It is just a way of stating a very general empirical fact. The very same words might seem to state some more interesting philosophical thesis as well, but according to Carnap they never do, and never could. Of course, Carnap's own positive and uncharitable account of philosophical questions and assertions rested solidly on the verifiability principle of meaning. 1S There is no "theoretical" or "cognitive" question about the "reality of the thing world" as a whole because there is no possible empirical verification of one answer rather than another. There is still a "practical" or "non-theoretical" question about whether to adopt the "thing language" at all, but that is to be settled by a decision based on the expediency and fruitfulness of adopting that set of conventions. 16 But there is no need to accept Carnap's verificationism and his positive account of philosophical questions as "practical" and not theoretical in order to acknowledge his most important point, which seems to me fundamentally correct. That is his recognition that the philosophical question of the external world and of our knowledge of it is not settled
284
BARRY STROUD
by arriving at empirical discoveries or conclusions in science or in everyday life. That is important because it involves drawing some distinction or other between ordinary or scientific assertions to the effect that there are external things and philosophical questions or assertions about external things, even if they are expressed in the very same words. Moore apparently saw only the "internal" or "empirical" half of this distinction, and so completely sidestepped the philosophical question. Of course if Carnap is right there is noth· ing meaningful that he missed, but for the same reason there is no philosophical question that he answered either. And perhaps that is what he missed. Carnap's view was that we do not have "a belief in the reality of the thing world". That implies that if we can be said to have a belief or accept a proposition to the effect that there are external things, as I suggested the geographers and therefore the rest of us can, then the proposition we are said to believe can be taken only in the ordinary way in which "There are external things" follows obviously and uncontroversially from "There are mountains" and "Mountains are external things". Carnap's reason for denying that we have the belief in ex ternal things that the philosopher is interested in was that that alleged belief could never be confirmed or disconfirmed, and so there simply is no such proposition for us to believe. Subtracting Carnap's verificationism, that is just the point the philosophical sceptic insists on. If the philosopher were right in ascribing to all of us a belief in a world of external things, and he were then to find that belief problematic and to ask whether it could ever be confirmed or disconfirmed in our experience, he would also be right in giving a negative answer. All possible experience is equally com· patible with its truth and with its falsity, so we have no reason to believe in external things. That is just Carnap's point that 'There are external things', taken philosophically, is empirically unverifiable. And that shares with scepticism the contention that if the philosophical problem of the external world were a meaningful and "theoretical" issue, it could never be settled one way or the other, and so we could never know. Since Carnap shares that conditional with the sceptic, he cannot adopt the strategy of trying to meet philosophical scepticism head-on by showing that we do know the very thing the sceptic says we do not know. In that he seems to me absolutely right. Moore did straightforwardly assert that we know that there are external things, but he did so legitimately and unproblematically only because he did not thereby manage to engage the sceptic at all. That can hardly be said to have been part of Moore's strategy; he simply did not consider it. Carnap understood the potential challenge of scepticism well enough to see that it could never be met, and for that very reason he denied
THE SIGNIFICANCE OF SCEPTICISM
285
its intelligibility. Of course, to say that 'There are external things', taken philosophy, is meaningless is an extreme view of the way it differs from its perfectly ordinary "internal" or "empirical" counterpart as Moore asserts it. But however extreme it might be, it is a view according to which there is a perceptible difference in status between the claims or questions the philosopher investigates and their ordinary or scientific "internal" counterparts. Some such distinction easily forces itself upon us when we try to take sceptical considerations seriously. That is what lies behind Kant's distinction between what he calls "empirical" and "transcendental" questions and claims,17 and it is what is responsible for our recognizing that Moore's "proof" does not settle the philosophical question of the external world. IV
What we know about the world is known somehow by means of the senses, and the possibility of knowledge of the world becomes philosophically problen1atic when we realize, with Descartes, that because of the possibility that we are dreaming or hallucinating, all of our experience could be just the way it is now even if the external world were in fact very different from the way we believe it to be. The s-ceptical challenge presents us with something that seems to be an intelligible'possibility. We seem forced to acknowledge at least the possibility of a distinction between the way things objectively are and the ways they are experienced as being, and so the challenge is to show how we could then know things are a certain way, as opposed to merely being experienced by us, and perhaps by everyone else, as being that way. 18 Carnap did not deal with the problem by acknowledging that the external world might or even could in general be different from what we experience it to be and then going on to show how we know that that possibility does not in fact obtain. He thought there was no such possibility; so it n1akes _no sense to ask whether and how we know that that possibility does not obtain. So he did not accept the sceptical challenge. If he were right, he would have stopped the sceptical threat completely, but not by showing that we really do know precisely what the philosophical sceptic says we do not know. The point is perhaps obvious, but it is important because it sometLrnes appears to be denied or ignored by many who seek a general account of human knowledge, and in the rest of this paper I want to try to fit so-called "naturalized epistemology" into the distinction I have outlined between two different ways of taking what looks like the same claim about the world or about our knowledge of it.
286
BARR Y STROUD
Quine, for example, claims not to share Carnap's conception of philosophical questions as non-cognitive or non-theoretical. He thinks that philosophy is simply more general than other "non-fiction genres" ;19 that the philosopher's concern differs from the geographer's or the physicist's "only in breadth of categories" .20 Philosophical questions are therefore as factual and empirical, and certainly as theoretical, as those of the more specialized sciences. For Quine in Word and Object: What reality is like is the business of scientists, in the broadest sense, painstakingly to surmise; and what there is, what is real, is part of that question. The question how we know what there is is simply part of the question ... of the evidence for truth about the world. The last arbiter is so-called scientific method, however amorphous. 21
And he speaks of our "hypothesis of ordinary physical objects" and of our "positing or acknowledging" such things. 22 Although the origin of our acceptance is "shrouded in prehistory",23 Quine thinks we arrive at the "hypothesis" of the existence of physical things by a "theory-building process"24 that yields "the smoothest and most adequate overall account of the world".25 The empirical study of the processes by which we arrive at our "theory of the world'" is the task of a "naturalized epistemology". 26 Carnap too thought we had all "accepted the world of things" very early in our lives "as a matter of course", but he thought that that acceptance was not to be understood as our believing or accepting as true a proposition to the effect that "the thing world as a whole" exists or is real. 27 Nor did he think we could infer from the obvious fruitfulness and efficiency of "the thing language" the conclusion that there probably is a real world of physical things. 28 Quine's view is that our "hypothesis" or "posit" of physical things is a good one, and that we thereby know (at least as well as we know anything else) that there are physical things. And our knowledge in this as in all other cases is ultimately based on sense-experience. This might look like a straightforward denial of Carnap's view, but it is not clear to what extent there really is a conflict here. That is just the question I want to raise, and it seems to me that the answer to it depends on exactly how a view like Quine's bears on the philosophical quest for an understanding of our knowledge of the external world. Quine claims to be explaining how we know there are physical objects. But is the knowledge of physical objects that he is interested in also shown to involve our knowing what the traditional epistemologist was interested in, viz., that physical objects in general really are as we experience them to be? Is it shown by Quine's naturalized epistemology, for example, how we know that we are not dreaming, or hallucinating, or are
THE SIGNIFICANCE OF SCEPTICISM
287
being victimized by an evil demon, so that we··~;know the traditional sceptical possibilities do not obtain? I think the answer must be 'No'. Certianly Quine himself never explicitly suggests that in general we do, or even ought to, consider those bizarre possibilities. And that is as it should be if, as he says, "the last arbiter is so-called scientific method, however amorphous". Those engaged in scientific "theory construction", however amorphous, do not actually proceed by considering and then legitimately eliminating all such alleged possibilities, however bizarre, and if it is true, as Quine says, that we arrive at our "theory of the world" by the "scientific method" then we do not do that either. This suggests that naturalistic views which emphasize the ways we actually come by our beliefs about the world can be accepted as perfectly legitimate even though they leave the sceptical challenge completely untouched and so provide us with no philosophical satisfaction ab out our knowledge of the world. Perhaps the staten1ent 'There are physical objects' is perfectly unassailable considered simply as a general proposition deduced unproblematically from the best scientific theories of reality, or as an expression of what we, as lay scientists, take to be the unexciting truth. And perhaps asserting, as a very general conclusion of learning theory, that we know there are physical objects because that "hypothesis" gives us "the smoothest and most adequate overall account of the world" is equally unproblematic. But I tried to indicate earlier that the general proposition 'There are external things' was easily deducible from the geographers' discovery that there are mountains, or indeed from Moore's assertion that there are at least two human hands, and in those cases that conclusion was thought to be irrelevant to the philosophical problem of the external world. It was seen as irrelevant not because it was illegitimately asserted, or because it was not true, but because it did not tend to settle the philosophical issue one way or the other. My question now is this. How does the apparently "scientific" assertion that there are physical objects, or 'that we know that there are, differ in that respect from Moore's apparently "common sense" assertion that here is one hand and here is another, or that he knows there are external things? The issue can be put this way. Is Quine's naturalized epistemology, considered in relation to the philosophical problem of our knowledge of the external world, simply a "scientific" version of Moore's so-called "common-sense" epistemology? There are strong indications that Quine intends no fundan1ental distinction in status between what he is doing and what I claim Moore was doing. He lemphasizes that the study of human knowledge as he understands it is part of psychology, and hence part of the very science whose origins it examines. 29 I
288
BAR,~
Y STROUD
He thinks it is inevitable in thinking about ourselves and our cognitive processes that we operate within the only resources we have for making sense of the world and gaining knowledge about it. And for Quine those resources can amount to nothing more than current science, broadly conceived. In more recent writings he emphasizes that even the doubts traditionally raised by philosophical sceptics are themselves "scientific doubts". 30 The epistemologist is therefore to be seen as "confronting a challenge to natural science that arises from within natural science" - the question is, "if our science were true, how could we know it?"31 - and for Quine that challenge is to be met by showing scientifically how we do in fact find out about the world, given that our sensory stimulations are as meagre as our science shows them to be. 32 The "challenge" Quine is interested in, then, is what he regards as the empirical scientific question of how we come to have our elab orate set of beliefs about the world. But I want to ask whether, when that question has been answered according to our best scientific lights, it will then have been shown that we do know precisely what the philosophical epistemologist has traditionally been concerned with when he has asked about our knowledge of the external world. If that will have been shown by Quine's project, how will it have been shown? Surely it cannot be said that in establishing that the "hypothesis" of enduring physical objects is the "theory" that provides "the sn100thest and most adequate overall account of the world" the traditional sceptical "hypotheses" have in fact been explicitly considered, carefully weighed, and eventually ruled out. And on precisely what grounds could it be maintained that the physical object "theory" is nevertheless superior in the relevant respects to the "hypothesis" that I am dreaming, or the victim of a mass hallucination or of an evil demon? Once those and other similar possibilities have been explicitly considered as alternatives, can it really be said that the physical object "hypothesis" is a simpler or better explanation or account of my experience, and hence is something that I know to be true? To give force to this rhetorical question I ask you to remember what my "experiences" would have to be like if this question had been properly raised: they would be immediate perceptions which it would seem could be equally well explained by the evil demon or the dream hypothesis as by the "hypothesis" of the external world. It is to no avoil at this point to say that the sceptical "hypotheses" are inferior and so can be disregarded because they are in fact known to be false or at best extremely improbable, or because there is in fact much less reason to believe them than to believe the physical object "hypothesis". That is precisely what is at issue once the philosophical question has.
THE SIGNIFICANCE OF SCEPTICISM
289
been raised, and the superiority 9f the physical object "hypothesis" over its competitors would have to be shown, and not just taken for granted. None of this need be taken as raising difficulties for, or even as casting aspersions on, the empirical study of the ways we in fact come to have our beliefs about the world. I mean only to insist that the results of such an investigation will not show that we acquire a general belief in physical objects by justifiably ruling out completely general sceptical possibilities according to which the world in general is different from the ways we experience it to be. We never in fact face that completely general question, except when philosophizing about human knowledge, and it seems to me that if we did have to face it and answer it in order to know anything about the world we would fail, and so scepticism would be true. So perhaps it's a good thing that we don't in fact face that question. But I am not suggesting that the results of an investigation like the one Quine recommends might not be perfectly legitimate, taken as intended, any more than I am suggesting that Moore's straightforward assertions might not be perfectly unproblematic as well. I am suggesting that the price of their legitimacy is that they do not provide an answer to the most general philosophical question ab out our knowledge of the world. That is perhaps a tendentious way of putting it, since one might find that that is a "price" one is quite willing to pay. The philosophical question that is not answered arises from the apparent realization that our experience could be just the way it is and has been even if the world we believe to exist were very different from the way we believe it to be as a result of that experience. It was because Moore appeared simply to ignore that possibility without considering it that his assertions would be dogmatic and unsupported if taken as responses to the philosophical challenge. Now it seems to me that Quine would be in an even worse position than Moore is in if his so-called "scientific" account were to be taken as a direct response to the philosophical question. Quine does not merely assert (as Moore does) that we know there are physical objects; he tries to explain, in general terms, how we know it. And if, contrary to what I have been suggesting, his answer were to be considered as relevant to the philosophical question, it seenlS to me that he would not, like Moore, simply have failed to show the falsity of scepticism; he would have accepted it. Or, perhaps a little more realistically, he would be committed to it, whether he himself would ever actually accept it or not. He sees his study of human knowledge as continuous with traditional epistemology. The new epistemology studies a physical human subject who gets a certain stimulatory input and eventually "delivers as output a description of
290
BARR Y STROUD
the three-dimensional world and its history". 33 In studying this "relation between the meager input and the torrential output"34 we are engaging in the scientific study of a species of primates and so are pursuing part of the natural science whose connections with experience we are in general trying to understand. 35 Quine puts the situation this way: We are studying how the human subject of our study posits bodies and projects his physics from his data, and we appreciate that our position in the world is just like his. Our very epistemological enterprise, therefore, and the psychology wherein it is a component chapter, and the whole of natural science wherein psychology is a component book - all this is our own construction or projection from stimulations like those we were meting out to our epistemological subject. 36
This I fmd very puzzling - not the study of how a species of primates or a particular human subject "posits bodies" or "projects his physics" from his laughably meagre data, but the claim that in studying him I appreciate that my position in the world is just like his. If I see the beliefs of a human subject as mere "posits" produced by his free creative efforts operating on his meagre stimulatory input, and so I see all of his "physics" and hence all of his beliefs about the physical world as an elaborate "construction" or "projection" from those data, then can I really avoid scepticism when I come to appreciate that my position in the world is just like his? If I ask myself the philosophical question of how I know anything at all about the world, and I find that all of my beliefs about it, including my belief that there are other physical human subjects who know the world in the ways I do, are themselves "projections" from relatively few irritations at my sensory surfaces (if any), then I easily realize that any number of other sets of beliefs would be equally worthy of "projection" and equally invulnerable to refutation, including the "hypothesis" that there is no physical world at all. I might have started my investigations of human knowledge from my own scientific view of the world, but the fact that my sceptical doubts are "scientific doubts" and arise only from within what is initially taken to be known about human beings is in itself no guarantee of the security of science, especially when I come to apply those doubts to my own case. Quine concedes that the sceptic can quite legitimately assume science in order eventually to reject it; he says "this, if carried out, would be a straightforward argument by reductio ad absurdum'',3? So he does not seem to rule out the possibility of the sceptic's arguing this way on the basis of so-called "scientific doubts". But I am claiming that the sceptical conclusion is inevitable if Quine's theory of "posits" or "projections" is taken as relevant to the philosophical question.
THE SIGNIFICANCE OF SCEPTICISM
291
Quine sees human knowledge as a complex combination of two separable factors. It seems to me that his epistemological project of distinguishing the uninterpreted contribution of the world (what he calls man's "cues") from the result of man's "free creation" (what he calls man's "world view"),38 cannot be accepted as an answer to the philosophical question about our knowledge of the external world without committing us to a total scepticism. Countless "theories" could be "projected" from those same data, and the world those "theories" purport to be about could give us no grounds for a reasonable choice among them. Even if we do in fact accept one "theory" it could not then be said that we do so because of its discoverable superiority to its competitors. On this picture each competing "theory" is based equally on the same nleagre data and, for the rest, arises only from our own subjective constitution. That is preciesely the state of affairs Descartes and other traditional epistemologists saw as threatening. It was what had to be shown not to obtain in order to explain how we can know anything at all. But Quine does not recommend naturalistic epistemology because it is the way to solve the traditional epistemological problem of the external world. That problem, he thinks, was "real but wrongly viewed", 39 so presumably he does not expect to provide satisfaction on the question as it was viewed by epistemologists of the tradition. That is perfectly consistent with the important conditional point I attributed to Carnap - that if the philosophical problem of the external world were a meaningful "theoretical" issue then scepticism would be the only answer. That is what I have been trying to insist on throughout my discussion of Quine. His naturalistic explanations of the origins of our views of the world are fully legitimate and acceptable only if they leave the sceptical challenge untouched and so provide us with no philosophical satisfaction about our knowledge of the world. They are "scientific" analogues of G. E. Moore's "common sense" assertions.
v One might well wonder at this point whether it is in any wayan objection to Quine's naturalistic epistemology to point out that it does not answer the traditional philosophical question. If we had an adequate scientific understanding of how we come to know things why should we bother at all with the antiquated sceptical problem? I think this expresses a widespread feeling anl0ng contenlporary philosophers, and by way of conclusion I would like to say something about why I do not share it. Of course many people simply find the intricacies of
292
BARRY STROUD
scepticism uninteresting, but unless traditional scepticism has been dealt with there will remain an unanswered question about human knowledge which, if legitimate, might well be described as the most fundamental question of all, viz., how is any knowledge of the world possible? If there is such a question it can hardly be said to be unimportant or not worth taking seriously in an attempt to understand human knowledge. I think it expresses our urge to see the phenomenon of human knowledge completely objectively, without a trace of anthropocentrism, and to understand only from that point of view how knowledge is possible. But if the philosophical question is illegitimate or somehow makes no sense, I think it follows that we cannot get that kind of understanding of human knowledge, and so we will have to understand the source and appeal of the illusory idea that we can. Quine's idea is that an empirical "scientific" understanding of human knowledge is all that we could hope for - that there is no way of understanding how knowledge is possible except by the straightforward investigation of its actual origins and development, and perhaps its history .40 And it seems to me that that idea must be the right one to have. But it ought to be possible to show th.at it is right - that there is in fact nothing else to understand about human knowledge and how it is possible than the facts we can discover scientifically about the actual processes of knowledge-acquisition and development. And that implies another point that seems to me not sufficiently recognized. If it were shown that that is the only kind of understanding we can have of human knowledge, then that conclusion itself would have to be taken completely seriously. If that is all that could conceivably be known or understood about human knowledge, and if, as I have been urging, no such discoveries tend to settle one way or the other the traditional question about our knowledge of the external world, and so there is no possible answer to that philosophical question, then it follows that it is an illusion to suppose that there is even anything intelligible that we still do not understand about human knowledge when we ask that question. If the study of our actual ways of knowing can give us everything there is to understand about knowledge then there is nothing left over. And since that would still not answer the philosophical question it follows that neither scepticism nor its opposite would have been established. So if we fully accept the idea that the naturalistic study of our actual ways of knowing can provide everything there is to understand about human knowledge we cannot at the same time say that it also shows that philosophical scepticism is false and that we do in fact know what traditional scepticism about the external world says we do not know. It is precisely in winning through to the realization that the naturalistic
THE SIGNIFICANCE OF SCEPTICISM
293
conception can give us everything we could possibly have that it seems to me most real epistemological progres~s is to be made. And that realisation itself a positive naturalized epistemology. We will not be the product solely must come to understand in detail just what is faulty or incoherent in the traditional question about knowledge, or in the way it arises and is given its special force. We need to understand why we have at most an illusion of raising a real question when we ask it. Those with little or no interest in scepticism may well have reached that position already, but I do not think they have done so through an adequate and convincing philosophical diagnosis of the sceptical arguments. I have argued that Carnap was on the right track at least in his realization that nothing that goes on in our ordinary or scientific epistemic practices settles the philosophical question. But he thought the question was literally meaningless, so it is no wonder that nothing could conceivably answer it. Now the problem for me, or for anyone who is impressed with the power of Descartes's First Meditation, is that the traditional epistemological question certainly does not seem straightforwardly meaningless. In fact it seems strictly analogous to ordinary questions which are fully intelligible and answerable. For example, I can ask whether I am perceiving things now as they really are or whether the distorting fluid that was put in my eyes earlier has still not worn off. I can ask whether I know that the things I perceive to be in this room are really there, or are only produced in my experience by some elaborate arrangement of wires and electrodes devised by a possibly malicious physiologist. There are empirically discoverable answers to such questions. What needs to be shown, it seems to me, is exactly how the traditional philosophical question about whether we can know that the external world really is in general as we experience it to be is not strictly analogous after all to those perfectly mundane questions: or, if it is analogous, that it is fully intelligible to us and raises a coherent possibility only because it does not have the completely general and therefore disastrous sceptical consequences it has seemed to have when asked within philosophy. 41 These are the sorts of things that would have to be shown if we are to have complete satisfaction about human knowledge, even given the truth and completeness of a fully naturalistic account.
of
VI
This is just the point at which the search for something like transcendental arguments seems most understandable and most promising. It seems as if the
294
BARRY STROUD
thing to do is to undermine traditional scepticism by demonstrating that it presupposes its own falsity, or that it could be true only if some of the conditions necessary for its own intelligibility were unfulfilled, or if it is in some other way impossible to get into the Cartesian position from which general scepticism so obviously follows, so it is no real threat. 42 Whether it will ever be possible to demonstrate such powerful conclusions seems to me to be still an open question, but if what I have been saying is right, it is essential to a satisfactory philosophical understanding of human knowledge that the appeal of traditional scepticism be shown to have some such unthreatening source. I have argued that that is required if we are ever to find ourselves in a position to accept straightforward naturalistic accounts as providing the whole truth about human knowledge. So I think that even pursuing the naturalistic programme does not exempt us from accurately accounting for the force of scepticism. But again, I want to emphasize a further consequence of the kind of diagnostic investigation I am recommending. If, for example, the possibility of the world's being different in general from the way it is experienced to be could be shown to be intelligible to us only because it lacks general sceptical implications, or because it is not a real possibility at all, then the threat of any scepticism arising from the contemplation of that alleged possibility would have been completely met. But it would be wrong to conclude that therefore we do know after all that the world is not different in general from the way it is experienced to be, and so philosophical scepticism is false. Not only does that conclusion not follow; if it had somehow been demonstrated to be true by some abstract argument we would be left once again with the original problem of how we could ever come to know such a thing - the abstract guarantee that we n1ust know it if we know anything at all about the world would not in itself be any help.43 Rather, a transcendental investigation of the presuppositions of traditional "sceptical" possibilities which exposes them as philosophically impotent would simply and completely defuse the threat of scepticism, and that is all. We would not then possess a satisfactory answer to the philosophical question, but only a sense of the loss of the original question itself, taken philosophically. Having reached that point, we could perhaps appreciate how a fully naturalistic account could give us all there is for us to understand about human knowledge, but by then the promise of naturalized epistemology completely on its own might well have lost its appeal. University of California, Berkeley
THE SIGNIFICANCE OF SCEPTICISM
295
NOTES J. Boswell, Boswell's Life of Johnson, London 1949, vol. 1, p. 292. G. E. Moore, 'Proof of an External World', in G. E. Moore, Philosophical Papers, London 1959, p. 146. 3 Ibid., p. 146. 4 See N. Malcolm, 'Moore and Ordinary Language', in P. A. Schilpp (ed.), The Philosophy ofG. E. Moore, New York 1952, pp. 356-61. 5 See A. Ambrose, 'Moore's Proof of an External World', in P. A. Schilpp, op. cit., pp. 410-17. 6 G. E. Moore, 'A Reply to My Critics', in P. A, Schilpp, op. cit., pp. 668-75. 7 See, for example, N. Malcolm, 'Defending Common Sense', Philosophical Review, 1949. (Also in E. D. Klemke (ed.), Studies in the Philosophy of G. E. Moore, Chicago 1969). There Malcolm argues that Moore, in his "proof", is actually misusing the word 'know', since its proper use requires that there be some question at issue and some doubt to be removed, that the person using it be able to give some reason for his assertion, and that there be some investigation which, if carried out, would settle the question. But these conditions, either singly or jointly, do not seem to be necessary for all correct uses of the word 'know'. For some ordinary contexts in which it is perfectly appropriate to assert that one knows things of the sort Moore claims to know in the conclusion of his "proof" or in his "defence of common sense" see T. Clarke, 'The Legacy of Skepticism', Journal of Philosophy, 1972. My whole account of Moore is heavily indebted to Clarke throughout. 8 For an elaboration of this view of the ordinary legitimacy but philosophical irrelevance of Moore's assertions, and in general of the problematic relation between philosophical questions and assertions and what goes on in ordinary life, see Clarke, op cit. 9 It is interesting that Moore goes to great lengths to point out that by 'external things' he, following Kant, means 'things to be met with in space'. That makes the question Moore is trying to answer (viz. 'Are there external things?') the same as Kant's "empirical" question about "empirically external objects". Kant held that all "outer perception . .. yields immediate proof" of the existence of such things, and that the truth of "empirical realisln" is in that sense "beyond question" (I. Kant, Critique of Pure Reasons, London 1953, A 375). So Kant would not object to Moore's proof as it stands, or even to calling it a proof, but he would not see it as removing the "scandal to philosophy and to human reason in general that the existence of things outside us ... must be accepted merely on faith, and that if anyone thinks good to doubt their existence, we are unable to counter his doubts by any satisfactory proof" (Kant, op. cit., B xl). The proof needed to remove that "scandal" is not simply the presentation in perception of an "empirically external" thing, but rather a demonstration that the apparent reasons for doubt can always be countered, and that "empirical realism" is true. For Kant, Moore's proof proves what it does only because "empirical realism" is true; it does not establish that it is true. So Kant would not see Moore in his proof as answering the philosophical question and thereby renloving the "scandal". Moore gives no sign of being aware of this fact. 10 R. Carnap, The Logical Structure of the World and Pseudoproblems in Philosophy, London 1967, pp. 332-4. Essentially the same view of the problem of "the reality of the external world" is found in M. Schlick, 'Positivism and Realism', in A. J. Ayer (ed.), Logical Positivism, Glencoe, Ill., 1959. 1
2
296
BARRY STROUD
11 R. Carnap, 'Empiricism, Semantics, and Ontology', in R. Carnap, Meaning and Necessity, Chicago 1956, pp. 206-7. 12 Ibid., p. 207. 13 R. Carnap, The Logical Structure of the World and Pseudoproblems in Philosophy, pp.333-4. 14 "Our example can easily be generalized. What is true for the mountain is true for the external world in general". Ibid., p. 334. 15 For a discussion of the verification principle of meaning and its importance for the present issue, see R. Carnap, Ibid., pp. 325-31. 16 R. Carnap, 'Empiricism, Semantics, and Ontology', PP. 208, 214. 17 Carnap's distinction between "external" and "internal" questions and (putative) statements closely parallels the Kantian distinction. Kant held, for example, that 'Things exist independently of our minds' is true when taken "empirically" but false when taken "transcendentally". "Realism" is empirically true, but transcendentally false. Carnap held that it is true when taken "internally" or "empirically" and meaningless when taken "externally" or philosophically. But both agree that there are two different ways of taking it, or two different questions which might be raised about "external things". In saying that 'Things exist independently of our minds' can be taken "transcendentally" Kant does not mean that, when it is taken that way, it asserts the existence of a realm of entities in splendid isolation beyond the reach of all possible experience. That would be a transcendent realm necessarily inaccessible to us. A question or statement is transcendental if it has a special status or role with respect to our knowledge in general; it asks for or states conditions which must be fulfilled if any knowledge is to be possible. That does not require that it describe some allegedly transcendent domain. See Kant, Ope cit., A 11-12, 296. 18 In this sketch of the source of the traditional philosophical question I am not trying to present in all its detail an argument for philosophical scepticism, much less to persuade you of its truth. I wish only to remind you of those familiar considerations about senseperception and knowledge that, rightly or wrongly, give the problem of the external world its philosophical point. I rely heavily throughout on your fan1iliarity and sympathy with the fust few pages of Descartes's First Meditation. 19 W. V. Quine, Word and Object, Cambridge, Mass. 1960, p. 275. 20 Ibid., p. 275. 21 Ibid., pp. 22-3. 22 Ibid., p. 22. 23 Ibid., p. 22. 24 Ibid., p. 22. 25 Ibid., p. 4. 26 See W. V. Quine, 'Epistemology Naturalized', in W. V. Quine, Ontological Relativity and Other Essays, New York 1969. 27 R. Carnap, 'Empiricism, Semantics, and Ontology', p. 208. 28 Ibid., p. 208. 29 W. V. Quine, 'Epistemology Naturalized', p. 83. See also W. V. Quine, 'The Nature of Natural Knowledge', in S. Guttenplan (ed.), Mind and Language, Oxford 1975, p. 68. 30 Ibid., pp. 67 -8. See also W. V. Quine, The Roots of Reference, LaSalle, Ill., 1974, pp.2-3. 31 Ibid., p. 2.
THE SIGNIFICANCE OF SCEPTICISM
297
Ibid., pp. 2-3. W. V. Quine, 'Epistemology Naturalized'~,;p. 83. 34 Ibid.,p. 83. 35 Ibid., p. 83. See also W. V. Quine, 'The Nature of Nature Knowledge', p. 68. 36 W. V. Quine, 'Epistemology Naturalized', p. 83. 37 W. V. Quine, 'The Nature of Natural Knowledge', p. 68. 38 W. V. Quine, Word and Object, p. 5. 39 W. V. Quine, The Roots ofReference, p. 2. 40 Quine himself does not emphasize the historical study of the development of human knowledge as an important part of a fully naturalized epistemology, although there is no good reason for him not to welcome it. He thinks a certain "epistemological nihilism" is reflected in the historically-oriented work of Polanyi, Kuhn, Hanson, and others, and that "cultural relativism" is thereby encouraged. Why that is more likely in the history of knowledge than in history generally Quine does not explain. His puzzling remarks perhaps indicate that there is more to naturalized epistemology as he understands it than an explanation of how we actually come to know what we do. See W. V. Quine, 'Epistemology Naturalized', pp. 87 -8. 41 For an attempt to shovl this in some detail with respect to the dream-possibility as it figures in Descartes' sceptical argument see T. Clarke, 'The Legacy of Skepticism'. My conception of the prospects of successful diagnoses of the force of scepticism is greatly indebted to that paper and even more to extensive discussions with Thompson Clarke. 42 See, for example, P. F. Strawson, Individuals, London 1959, pp. 35, 106; S. Shoemaker, SelfKnowledge and Selfldentity, Ithaca, N.Y. 1963, pp. 169-70. 43 Kant's "transcendental idealism" was an attempt to answer this question. He saw it as the only explanation of how "empirical realism" could be true, and so "empirical idealism" and scepticism could be avoided. If, as some recent commentators would argue (e.g., P. F. Strawson, The Bounds of Sense, London 1966), transcendental idealism is incoherent, it must be because the question to which it is an answer is also incoherent. It is not just that transcendental idealism is an incoherent answer to a perfectly intelligtble philosophical question. But if that is so, then one cannot reject transcendental idealism as incoherent and then go on to give a different answer to that same philosophical question. The price of adopting the "austere" interpretation of Kant, free of the "excesses of transcendental idealism", ought to be that one is then precluded from giving what seem like more palatable answers to the same problems those "excesses" were introduced to solve. And that would mean having no answers at all to those questions. That would be satisfactory only if the incoherence, unintelligibility, or illegitimacy of those putative questions had been independently established. 32
33
PETER BIERI
SCEPTICISM AND HOW TO TAKE IT
Comment on Stroud
In his paper Professor Stroud advances the thesis that the traditional approach to epistemology, which takes scepticism seriously, is superior to a currently widespread attitude of simply ignoring the sceptic and his doubts by confming the epistemological task. to some sort of naturalistic description and reconstruction of the epistemic procedures actually employed by scientists and other knowing subjects. He claims that it is only by dealing with the questions raised by the sceptic that we can arrive at a full understanding of the nature and possibility of human knowledge. What he argues for is the view that no theory of knowledge, no assessment of our epistemic situation regarding the world can ever be satisfactory unless it succeeds in giving an account of the appeal that scepticism has had in philosophy since Descartes. Even if we should be convinced that Descartes just gave us something to play with and that it is high time to realize that we should give up this admittedly suggestive paradigm of philosophy, we still ought to try to give arguments for this view. Otherwise there will ren1ain a suspicion that what we have before us is simply a change of fashion. Consequently, Stroud tries to convince us of the need for a diagnostic investigation which would expose the origins and the conceptual resources of philosophical scepticism that have been and still are responsible for its threatening force. The paper echoes Kant's methodological maxim that it will not do in philosophical contexts to ignore a position we want to discard, even if we think we have abundant reason to abandon it. Stroud insists with Kant that all philosophical criticism must fulfill a further condition: It must reveal where and why the considerations leading to its target went wrong. I am very sympathetic to this thesis in its general form, because I think, with many others, that the enterprise of a philosophical theory of knowledge, understood as the attempt to find one's way around in the cluster of concepts like knowledge, belief, evidence, truth etc. is best understood as a reaction to scepticism, and that it is, therefore, of the utmost importance to try to rnake as clear as possible what the notorious sceptical challenge really amounts to. Consequently, the aim of my comments is not to attack the general tendency of Stroud's considerations. Instead, I want to do two things: First, I want to consider the dialectics of the paper and indicate the points that seem to me to
299 P. Bieri, R.-P. Horstmann, and L. KrUger (eds.), Transcendental Arguments and Science, 299-307. All Rights Reserved. Copyright © 1979 by D. Reidel Publishing Company, Dordrecht, Holland.
300
PETER BIERI
stand in need of clarification. Second, r shall suggest a way of looking at the sceptic which, if correct and worked out in detail, would throw a somewhat different light on the role the sceptic is playing.
To begin with, the most important thing to note about Stroud's argument, taken as a whole, is that it is neither a defence nor a refutation of scepticism. Neither does Stroud attempt to show that scepticism is true and that we cannot have any knowledge of the world, nor does he intend to offer arguments which would amoung to a demonstration that the enduring persuasiveness of the sceptic's doubts rests on clearly misguided reasoning and that we do have knowledge of the world. In order to bring out the significance of scepticism he chooses a much less straightforward strategy: On the one hand, he argues that neither Moore, nor Carnap, nor Quine manage to block the sceptic's attack, because they underestimate the intricate moves available to the sceptic. On the other hand, he thinks that transcendental arguments would in principle be the adequate weapons to come to terms with scepticism, given the difficulties of foundationalist theories of knowledge. But at the same time it is obvious that he sees no more chance for that kind of argument now than he did earlier. 1 I think the most fruitful way of reacting to this dilemma is to see it as something we should not try to escape but which we should accept as inevitable. Looking at it this way, I suggest, is more helpful in understanding scepticism than an effort to avoid it. In fact, I shall later propose an interpretation of the sceptic which could make us realize that the very idea of refuting or defending him should be given up. First, however, I want to review briefly the general structure of Stroud's argument and to mark the points which, I think, are not quite clear. Stroud introduces the sceptic as asking the completely general question of how any knowledge of the world is possible. His sceptic arrives at this question by a quick and seemingly innocuous line of thought. He starts by referring us to the various cases of distorted sense-experience and summarizes these facts by saying that sometimes things look different from the way they are. Stroud himself shares the sceptic's belief that this is a truism and that any realistic account of human knowledge has to start from that kind of considerations. It is important to mention this fact, because it shows that the paper is basically dealing with Cartesian scepticism, which relies so heavily on the concept of illusion. The further steps of this type of scepticism are familiar enough: The
COMMENT ON STROUD
301
Cartesian sceptic goes on to the idea that, in general, our total experience could be just the way it is even it the world were complete different from what we believe it to be as a result of that experience. Particular, clearly identifiable instances of perceptual illusions are used as evidence for the general claim that all our experience is equally compatible with the truth or falsity of our beliefs about what the world is like. Assuming the validity of this generalization the Cartesian sceptic proceeds by advancing what Stroud calls 'sceptical hypotheses' to the effect that we might be dreaming, be victimized by an evil demon, or that our experience might be a product of some complicated neurophysiological manipulation. Finally, the sceptic takes the offensive by claiming that we cannot justifiably talk about having any knowledge of the world until the possibility of such overall-deceptions are definitely ruled out. And the sceptic's claim is, of course, that this requirement can never be met and that, therefore, we are wrong when we think, as we do, that we are entitled to knowledge claims. If he is consequent, he will even have to claim that nothing is ever even the slightest bit more likely to be true than anything else. For all-its familiarity, I do not think that this way of describing the sceptic's reasoning is clear enough to serve as a basis for the kind of diagnostic investigation Stroud urges us to do. Before sketching an alternative description, however, I shall stay with this picture for a while and explore Stroud's further steps. He starts with a discussion of Moore's 'proof' of the external world and asks for an account of the fact that Moore, who was familiar enough with the intricacies of scepticism, could think that the sceptical challenge could be met once and for all by his well known performance. Stroud's explanation of this puzzling fact is that Moore was either dogmatic or missing the point of philosophical scepticism altogether. This interpretation relies heavily on Carnap's distinction between internal and external questions or something like this distinction. If we substract Carnap's implausible verificationism, we get a distinction between questions which can be settled by ordinary empirical means and questions which are theoretical but which concern ontological issues not accessible to empirical investigation. 2 Philosophical scepticism raises questions of the latter kind, and we may now say that if Moore's 'proof' is a reaction to an internal question, it is sound, but uninteresting, and that if it is an answer to an external, theoretical question, it is dogmatic. Assuming that the distinction between these two types of questions can be saved against Quine's attack, Stroud's way of dealing with Moore is very attractive. Still, it is hard to believe that Moore simply wanted to be dogmatic or could have been so naive as not realize the strength of traditional scepticism. It is true
302
PETER BIERI
that he emphatically asserted that he knew that there is a world outside us. But the fact that he neither engaged in a discussion with the sceptic, nor tried to analyse the concept of knowledge he uses suggests that his intention might have been to attack the whole enterprise of a philosophical theory of knowledge. Instead of looking at him as trying to refute the 'scandal' of scepticism, he could be seen as claiming that the sceptic arises because of a completely misguided interpretation of our episternic predicament. Moore would, seen from this angle, have to be understood as doubting whether the traditional epistemological theories are to be taken seriously. He might be presented as saying that the real sceptical question is how we can legitimize the traditional approaches to our epistemic situation. This, I take it, is the real meaning of the phrase that Moore was defending common sense instead of just relying on it. The next step in Stroud's argument focuses on the difference between internal and external questions in Carnap's original sense. The center of this part consists in relating this difference directly to scepticism. The important point here is that if we accept Carnap's position, we do not get a refutation of scepticism, but a new type of scepticism: The sceptic can now be described as an ontological conventionalist who need not recur to the argument from illusion described above. This sceptic cannot be taken head-on. A refutation of scepticism would now have to try to develop arguments against conventionalism, i.e., arguments to the effect that there is a core of concepts which enjoy a privileged status in every description of the world and which are necessary in order to have a language. In other words: some sort of transcendental arguments. I believe that there is another, additional reason for making the discussion of Camap's position an essential part of a diagnostic investigation of scepticism generally: This discussion shows the close interrelationship between a theory of meaning using the concept of verification and the potential force of sceptical doubts. I suspect that as long as meaning is somehow connected with justification, it will always be possible to raise sceptical questions. It is, I think, the difficulty to break that connection which explains the appeal that scepticism has had for so long. 3 I turn now to Stroud's discussion of the relationship between Quine's naturalized epistemology and scepticism. Here Stroud tries to show that even if it were true that sceptical doubts are scientific doubts we could not say that philosophical scepticism has become obsolete. His main argument runs as follows: Even if Quine were right in breaking with a long tradition by turning the traditional epistemology into an enterprise within natural science, he
COMMENT ON STROUD
303
could not rule out sceptical hypotheses by sin1ply con1paring them with the explanatory power of empirical theories. This, Stroud argues, would be begging the qwestion, because the sceptical hypotheses are not ordinary empirical hypotheses. Rather, they may be described as means for doubting the claims to knowledge of all empirical theories. Quine, of course, would reply that the very idea of such sceptical hypotheses does not make any sense. In fact, he would ask the very embarassing question: What exactly are we talking about when we refer to 'philosophical scepticism'? He would neither be impressed by Stroud's allusion to Berkeley, nor by the fact that we can ask a completely general question about knowledge which we are usually not faced with. I think that the resulting standoff is not just the game of shifting the burden of proof but should be seen, at least tentatively, as a first step in understanding all talk about scepticism along the line I suggested in connection with Moore. We should, I think, learn from it that asking for the significance of scepticism means asking how we can take the sceptic seriously without being a sceptic and without dogmatically assun1ing that the sceptical reasoning is legitimate or at least intelligible. Before reacting to the last part of Stroud's paper, I want to return to his presentation of the sceptic's considerations. It is dangerous to present the sceptic as asking the question "How is any knowledge of the world possible?" Taken literally, this question asks for an explanation of how we arrive at our knowledge claims. But then it is not a sceptical question at all, at least not in the traditional sense. It is certainly a much more general question compared with other questions asking for explanations, and it can be said that it is the most fundamental question as far as explanation of anything is concerned. But in what sense could it be said to formulate a doubt? What is the difference between asking for an explanation of knowledge and, e.g., asking "How is it possible that we can discriminate between different colors?" We may, of course, say that it is just the point to describle the sceptic as simply asking for an explanation of the fact that we claim to have knowledge of the world. We may refer to Quine's progran1 of naturalizing epistemology which reduces the traditional search for certainly to an analysis of how we come by our beliefs about the world. If we believe that the only thing left to do in epistemology is to describe and explain the 'relation between the meagre input and the torrential output' of human organisms, including their knowledge claims, there is not only nothing wrong with the above question; it is the only question about knowledge which makes sense. This way of looking at the sceptic might be right, and Stroud leaves no doubt that he thinks that an investigation along Quinean lines can give us all there is to understand about knowledge. But
304
PETER BIERI
until we have shown that Quine's view is no't only the proposal to return to psychologism in a new setting, but that he is right to draw this consequence, we cannot phrase the sceptic's question the way Stroud does, unless we want to run the risk of being accused of begging the question. A similar difficulty seems to me to lie in the fact that Stroud's sceptic relies so heavily on the concept of illusion, because this concept and the notion of deception, if they occur in a sceptical argument, are far less clear than they seem. To talk about something as an illusion, it will be conceded, makes sense only if it can be viewed against a background which we take not to be illusionary. But what is this background? Is it, as Quine claims, a system of 'posits', fabricated for coping with a wide range of stimuli, or is it the world as something independent of the way we conceptualize our experience? In other words: Is an illusion just something which is incompatible with our usual empirical beliefs, or is it something the analysis of which must refer to the world as something more than the organization of the total of our beliefs? Questions like these bear upon Stroud's point that it is an essential part of an understanding of the sceptical threat to decide whether the sceptic is entitled to generalize from particular cases of illusion to the thesis that all our experience might be illusionary. If Quine were right in claiming that talk about illusions has to be understood within a system of 'posits', and that the notion of deception is at heart a scientific notion., it would be clear that the sceptic could not use the argument from illusion for building up an argument to the effect that all of us might be deceived all of the time. Quine would then be right in saying that Descartes is not the way to begin, and we would not understand any more why Stroud's sceptic should be a challenge to Quine which, as Stroud believes, makes it plausible to look at him as, accepting scepticism. If, on the other hand, we analyse our talk about illusions as, roughly speaking, referring to son1e sort of wrongly picturing an independent world, the sceptic's generalization is perfectly intelligible, and Descartes is the right way to start. The fact that Stroud leaves this alternative concerning the notion of the world in the dark in his description of the sceptic is, it seems to me, responsible for the way he phrases the task of an investigation of scepticism which I find hard to understasnd and impossible to fulfill. In other words: Because Stroud never says precisely what the sceptical position is, the investigation he recommends is made unnecessarily difficult to understand. II
Given these difficulties I want now to sketch a sceptic who, instead of arguing
COMMENT ON STROUD
305
from illusions, focuses directly on the concept of knowledge by asking: What conditions have to be fulfilled in order to justifiably use the concept of knowledge? Can these conditions ever be fulfilled? THese questions suggest themselves when we realize that Stroud makes use of the concept of knowledge throughout his paper without ever saying what the sceptic means by this term. Descartes' argument from dreaming is, to be sure, a good illustration of sceptical reasoning. But first, we should not forget that it functions in the context of what may be called methodological scepticism as distinguished from scepticism as a philosophical position. And second, it is an argument which is far from clear, even if we detach it from tHe reasons Descartes himself brought it up for. 4 The sceptic I Have in mind starts from tHe assumption that knowledge is best analysed as justified true belief. He is well aware of tHe all too familiar problem of how to stop the notorious regress of justification. He simply leaves the question open whether tHe notion of basic statements can be defended against recent criticism, or whether we must, once and for all, yield to the considerations which are designed to make the notion of such statements seem unitelligible or even ridiculous. Instead he concentrates on the hardly contoversial point that in order to know sometmng we must have evidence for it. Given the well-known problem of bridging the gap between the truth of what we know and the evidence we need for making knowledge claims, he then develops a fairly natural line of reasoning: First, he demands that we have knowledge of the statements which fonnulate our evidence. Second, he requires that we know of our evidence that it is evidence. And third, he -is not satisfied with something like undefeated justified true belief. He thinks that the gap between truth and evidence makes it necessary to require not only that there has not been any conflicting evidence so far, but that it must be positively shown that there are and never will be such conflicting evidence. It follows that we n1ust know that there are and never will be something in the world which could possibly override all the evidence we now have. I think the natural reaction to this rigid sceptical requirement is ambiguous. On the one hand, one is inclined to simply brush it off, because its consequence would be that in order to know anything one would have to know everything, a consequence which is patently absurd. On the other hand, this radical scepticism is built up by steps which, by themselves, cannot be said to be apparently absurd. On the contrary, one might argue that it would seem arbitrary to block the underlying strategy at some point. Another dilemma, then, and again I would propose not to try to escape it, but to see it as inevitable and to ask what we can learn from it. What can we learn from it?
306
PETER BIERI
I think it may suggest an interpretati0n of the sceptic as a figure d'art as we find it in certain stage-plays - indispensible for the plot and responsible for the main conflicts, but only mistakenly taken for real by the people in the play - as the audience knows. The plot is the attempt to bridge the gap between the truth-<:onditions of a description of the world and the available evidence for that description. The main conflict, reflecting the dilemma sketched above, is given, on the one hand, by the fact that every such description may be challenged by the sceptic on the ground that even our best possible evidence for it is never conclusive, and, on the other hand, by the fact that the charges of senselessness, incoherence or emptiness brought forward against the sceptic's doubts - based on Wittgensteinian arguments, on a diagnosis like Strawson's, or on considerations like Davidson's - inevitably presuppose certain forms of verificationism regarding meaningfulness or translation which it is essential for the sceptic to reject. Stepping outside the play down to the audience would not in the least diminish the significance of scepticism. It would still mean recognizing the sceptic as an indispensible point of reference both in the analysis of epistemic justification and in the systematic and historical understanding of philosophical epistemology. But being part of the audience would make one refrain both from justifying and from dismissing philosophical views in epistemology and other areas - notably in semantics and the philosophy of mind - for reasons given by the action on the stage, and Quine's drastic change of subject from the justification to the empirical genesis of our beliefs would appear as an unnecessary move within the play. Viewed from this perspective the sceptic would appear as an artificial figure designed to bring to light the general features both of our epistemic predicament and of our theories about it, playing a role similar to that of a figure d'art designed to reveal the ethical and psychological predicament of the persons involved. Assuming that it is the second dilemma inherent in every concept of knowledge which leads to and explains the first dilemma developped by Stroud, is there anything more to understand about the sceptic in his various disguises? Universitiit Heidelberg
NOTES See his 'Transcendental Arguments'Journal of Philosophy 6S (1968), 241-256. For an analysis of this type of questions see James W. Cornman, Metaphysics, Reference, and Language, 1966, Parts II and III. 1
2
COMMENT ON STROUD
307
3 This is illustrated by Peter Unger\~~s book Ignorance, 1975, particularly by his discussion of 'absolute terms' and 'relative terms', and by Stroud's review of the book, Journal ofPhilosophy 74 (1977), 246-257. 4 See O. K. Bouwsma, 'Descartes' Skepticism of the Senses', Philosophical Essays, 1942, 51-63.
INDEX OF NAMES
Adickes, Erich 25 Alembert, Jean Ie Rond d' 161 Ambrose, A. 295 Apel, Karl Otto 103, 188f Aristotle 7,69,91,111,137,143,177, 229 Ayer, Alfred J. 57,63,80 Baum, Manfred 27f, 32, 37f, 40,42 Bennett, Jonathan 30f, 34, 41, 52, 56, 65f, 68f, 71, 73-75,82,120,211 Berkeley, George 7,253,279,303 Bittner, Rudiger 37 Bohr, Niels 123,133,139,152,156,164, 166-168 Boole, George 144f, 162f, 169f Bouwsma, O. K. 306 Boyd, Richard 241 Brentano, Franz 65 Bubner, Riidiger 77, 80-83, 85, 93, 99, 101,106,188 Burian, Richard M. 223, 227f, 230-232, 235,241f Carnap, Rudolf 180, 281-286,293, 295f, 300-302. Castell, L. 150 Chomsky, Noam 187,189 Clarke, Noam 187, 189 Clarke, Thompson 295, 297 Cohen, Hermann 32,34 Collingwood, Robin G. 4 Cornman, James W. 306 Cowan, C. L. 233 Crawford, Patricia A. 25 Darwin, Charles R. 84, 130f, 142,154, 168 Davidson, Donald 78, 83, 90f, 95-103, 105, 108-112, 114, 119, 242, 262, 306
Descartes, Rene 50, 65, 96, 105, 177, 246, 250, 278,285, 291, 293f, 296f, 299, 304f Dewey, John 78,81,84, 100f, 118f Dingler, Hugo 175 Drieschner, M. 144, 158 Ebbinghaus, Julius 25f Ehrenfest, Paul and T. 157 Einstein, Albert 3f, 123f, 132[,214,217, 224,229,232 Elgin, Catherine Z. 223 Euclid 80, 147f, 150, 155, 175, 187, 203-205,213,259,267 Feyerabend, Paul 92, 100, 180, 187, 192, 202,216,219,222,231 Field, Hartry 214, 224 Finkelstein, D. 158 Fisk, Milton 91 Frege, Gottlob 3, 118, 135 Gatschenberger, R. 188 Gibbs, Josiah W. 157 Glansdorff, P. 158 Godel, Kurt 213 Goldman Alan H. 58,63 Gombrich, Ernst H. 86-89 Goodman, Nelson 187 Gram, Moltke S. 51, 63 Griffiths, Phillips A. 51,63 Habermas, Jurgen 188 Hacker, Peter 59f, 64 Hacking, Ian 103 Hanson, Norwood R. 231, 297 Harrison, Ross 55, 64 Hegel, Gottfried W. F. 84, 86, 91, 95,100, 248,253 Heidegger, Martin 78, 81, 101, 118, 154
309
310
INDEX OF NAMES
Heisenberg, Werner 123f, 131, 133, 157, 261 Hempel, Cad G. 189, 208f, 257f Hermes, Hans 187 Hilbert, David 144f, 149, 156, 174f Hintikka, Jaakko 50,64, 106f Hobbes, Thomas 177 Hooker, C. A. 164, 170 Horstmann, Rolf P. 269 Hume, David 45,49,62, 72, 80, 85, 91, 99, 116, 125, 127, 129f, 141-143, 168,192,249,251 Hussed, Edmund 100 James, William 187f Janich, Peter 175, 189, 191, 195 Jauch, J. M. 158 Johnson, Samuel 279 Kant, Immanuel 3-8, 10-13, 23-32, 37-43, 49f, 52, 55, 62f, 65-67, 69, 71-75, 78-80, 84, 87, 89, 92, 95, 97-101, 105-108, 114-118, 124129, 135-140, 152, 155, 159, 171173, 175, 177,180,186-188,191195, 228, 230, 245, 248-253, 258261, 263-267, 269, 285, 295f, 297, 299 Kekes, John 51,64 Klein, Felix 147 Komer, Stephan 4, 6f, 25,37,42,51,64 Kolmogorov, A. N. 144 Kripke, Saul 90f Kuhn, Thomas S. 92-97, 131, 157, 180, 192, 216, 219, 221, 228f, 231-233, 297 Lakatos, Imre 180 Lambert, Johann H. 195 Lange, Friedrich A. 25 Laplace, Pierre S. de 261 Lehrer, Keith 95f, 102 Leibniz, Gottfried W. 50, 178 Lewis, Clarence I. 80, 92 Lewis, David 242 Locke, John 45f, 56, 91, 95f, 99, 253 Lorenz, Konrad 134, 137f, 157, 187
LoI:enz, Kuno 189, 191-193 Lorenzen, Paul 175, 188f, 191, 195 Lovejoy, Arthur O. 84 Lukasiewicz, Jan 157 Mach, Ernst 124 MacIntosh, J. J. 51, 64 Malcolm, Norman 64, 295 Maupertuis, Pierre Louis M. de 161 Mead, George H. 188 Mendel, Gregor 209 Minkowski, Hermann 260 Mittelstaedt, Peter 145,158 Mittelstrass, Jtirgen 188f, 191 Moore, George E. 45, 105f, 279-285, 287,289,291,295,300-302 Morgenbesser, Sidney 242 Nagel, Ernest 208 Neurath, Otto 231 Newton, Isaac 3f, 6, 124, 161, 214-217, 219-221,229,232,259,267 Pappus 7 Parsons, Charles 197, 224 Pierce, Charles S. 83-86, 179, 188, 197, 199f, 208, 210-213, 216, 218, 222, 224, 227f, 235-240, 253f, 258, 261 Penrose, R. 158 Piaget, Jean 188 Picht, Georg 158 Planck, Max 261 Plato 7, 34, 118, 136f, 261 Poincare, Jules H. 147f, 150 Polanyi, Michael 297 Popper, Karl 100, 130f, 134, 137, 157, 180,192,231 Prigogine, J. 158 Prior, A. N. 158, 173 Putnam, Hilary 77, 84f, 87-91, 95f, 98f, 102, 114, 118f, 170, 187, 197f, 201f, 211, 222, 238, 241f, 262 Quine, W. V. 80, 91f, 95, 102f, 118f, 189, 197, 201f, 213, 215, 217, 219, 224, 227-229, 230-233, 235f, 240, 242, 262, 286-292, 296f, 300-304, 306
INDEX OF NAMES Ravetz, J. 222 Reich, Klaus 25f Reichenbach, Hans 178, 187 Reines, F. 233 Rescher, Nicholas 173 Riemann, Bernhard 80,147, 203f Rorty, Richard 50, 55f, 58, 61-64, 86, 105-111, 113-115, 118f, 189, 197, 242,262 Rosenberg, Jay 50, 64, 77f, 85-88, 9095,99, 102, 197,212,216,219,221t 224, 227, 263f, 266-269, 272,274 Russell, Bertrand 3, 100, 135, 187 Schaper, Eva 25 Scheibe, Erhard 158 Scheler, Max 4, 25 Schlick, Moritz 295 Schrodinger, Erwin 164, 166, 261 Schwemmer,Oskar 188f Searle, John R. 189 Sellars, Wilfrid 77, 85-88, 90-92, 99, 102f, 115, 197, 199, 200-206, 208, 210, 212-214, 216-219, 221, 223f, 227,235,241,261 Shannon, C. E. 134 Shoemaker, Sidney 50,64, 119, 297 Smith, G~ W. 56f, 64 Smyth, Richard 262
311
Sneed, Joseph 132,157,189 Spinoz~, Benedict 50, 79, 81 Stegmtiller, Wolfgang 132, 157, 180, 189 Stine, William 58f, 64 Strawson, Peter F. 3-6, 25, 30, 32, 34, 42, 50, 52, 56, 60, 64, 71, 82f, 99, 102f, 115, 118, 183, 188f, 260,262, 297,306 Stroud, Barry 31, 34, 58, 64, 231, 233, 299-306 Tarski, Alfred 97f, 198, 200, 223 Thomson, Judith J. 58,64 Tlumak, Jeffrey 57, 64 Tugendhat, Ernst 157 Urquhart, A. 173 Weizsacker, C. F. von 159-163, 165-168, 170-174, 187f, 237 Whitehead, A. N. 189 Whorf, Benjamin L. 97 Wigner, E. P. 166 Wilkerson, T. E. 42, 55f, 62, 64 Williams, Bernard 57, 64 Wittgenstein, Ludwig 50, 53, 55, 78, 80, 86,88,101,105,119,192,261 Wohlrapp, Harald 180-182, 188 Wolff, Christian 3
INDEX OF SUBJECTS
Analogies of experience 14-24, 40-42, 260f Analytical method, Kant's 6-8, 16f, 32, 37 Assertibility, semantic/warranted 201212 Categories (Kant) 6-24, 38-42, 66, 74, 128 Causality 9-19,40f Certainty 246, 271 Change of concept (meaning) - change of belief 202-205 Change of theories 69, 160f, 179, 187, 197, 208-210, 213-221, 224, 228230,246-262,267,271-273 Concepts: - a priori 38-42,67 - objectivity- 52-63, 65-69, 74 Conceptual change 197-210, 219, 224 Conceptual scheme (framework) 4-6, 78-85,95-101,108-111,115-117, 201-205,219f Conditions of possible experience 4-25, 37f, 139f, 154, 160, 171, 237-239 Constructive philosophy of science (Erlanger Schule) 177-187, 191, 193-195 Descriptive metaphysics 4-6, 8, 38
Historicism 91, 100f, 140 Idealism 3, 84, 253 f Indeterminacy of translation/reference 212-221,224,227-232,235-251 Information 134f Inner states 45f, 50, 52-62, 66f Instrumentalism 47f, 229 Intemal-external questions (Camap) 281284,301f Intuition (Kant) 9-24,27-29, 38f Judgements about the past 52-56,66f Justification, prospective-retrospective 221f, 246-248, 252, 258, 267f, 271 Ontological relativity 212-221, 227 - 232, 241 Ontology 3, 8f, 162-166, 256 Philosophy of mathematics 6f, 118f, 128f, 154f Positivism 124, 154, 281 Pragnlatism 83-88, 91, 101, 245, 253256 Presentation-research (Darstellung-Forschung) 178-187, 192f Probability 141-145 Quantum theory 123f, 132f, 138f, 142156, 162-166, 172
Epistemology naturalized 78, 90-92, 237f, 245, 256, 277, 285-293, 302304 Evolution 92, 94, 133f, 137, 154, 177179, 247f, 254-256, 273 Explanation 185, 208f, 257f Foundationalist theories of knowledge 45, 277,300
Rationalism 177, 179f Realism: - in general 130f, 153f - explanatory 85-89, 90-95, 221f, 255-261, 263f, 267f - internal 88-90, 197, 216, 230, 238, 241 - metaphysical 77,83-90,95,114,118,
313
314
-
INDEX OF SUBJECTS
130, 153, 161, 197-200, 211, 230, 252f Peircean 83-85, 197, 199f, 210-212, 224,236-241 Sellarsian 85-90, 92, 197-221 transcendental 48f, 57,74
Scepticism: - in general 45, 48, 50-62, 77f, 82f, 90, 116f, 246, 250, 271, 274f, 277306 - Cartesian 45, 50f, 56f, 59, 78, 246, 278, 293f, 299f, 304f Scientific revolutions 69, 124, 159-162, 223, 228-231, 246f Self-consciousness (self-knowledge) 2123,27-31,50-61,65-67,73 Space-time continuum 147-152, 175f Systems of representations 245-251, 254-260 Temporallogic 140, 142-146, 173-175 Theory of measurement 132f,164f Time 14-24, 140-143, 152f, 156, 168f Transcendental apperception 22-24,66 Transcendental arguments: - in general 77-83, 89f, 93, 95, 99f, 105-107,152,155,277, 293f, 302 - analytical 30-34, 37,45,50-63, 65f, 68f - Kant's 3-24, 71-73
-
in Strawson's sense 4-6, 25, 32, 37,
50,52 Transcendental deduction 5f, 14-16, 19-24, 26, 67, 114, 116, 118, 128, 193,249,258,263 Transcendental philosophy: - in general 27, 77f, 99f, 113, 124f, 154, 177-179 Kant's 5-9, 71f, 114-118, 124f, 191f, 245, 248f, 252, 261 Transcendental principles/propositions 6, 9-24,37-42 Transcendental proofs 8-24,37 Transcendental synthesis 5, 21f, 27 -31, 39, 66f, 114, 127f, 245, 250-53, 257-259, 264 Truth: - as coherence 81, 85-90, 95 - as correctness 254-256, 263, 265267,273 - as correspondence 84-90, 95-99, 109f, 130, 198-219, 236-241 - Peirce's theory of 85-87, 93, 197, 199f, 210-212, 224, 236-241 - Sellars' theory of 85-88, 92, 200-219 - Tarski's theory of 97f, 109f, 198-200 Verificationisnl 48-50, 57f, 60, 74, 211, 283,302,306 World 250-254
