Logical Positivism [2 ed.]

Citation preview

LOGICAL POSITIVISM E dited by

A lfred

sjufes

A. J. A Y E R ,

The Free Press, New York

M/o -

Copyright © 1959 by The Free Press, a Corporation Printed in the United States o f America

All rights reserved. No part of this book may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, re­ cording, or by any information storage and retrieval system, without permission in writing from the Publisher. Collier-Macmillan Canada, Ltd., Toronto, Ontario Library of Congress Catalog Card Number: 58-6467 Second p rin tin g Ju ly 1966

Preface

Logical Positivism is the second in a series of books which will make available to the general public some of the most interesting work of philosophers of very diverse viewpoints. Each volume will deal w ith one o r, in some cases, with two, philosophical “schools” or “m ovem ents.” It is fortunate th at philosophers are rarely united by the kind of com m on purpose which inspires political o r religious “m ovem ents.” N evertheless, it is frequently helpful to consider the w ork of different writers according to the similarities in their aim and content; and this is the policy which has been adopted in de­ signing th e Library o f Philosophical M ovem ents. This volum e presents, for the first tim e in English, many of the m ost influential papers by leading members of the V ienna Circle. These and other articles contain authoritative expositions of the doctrines m ost com m only associated with logical positivism. H ow­ ever, for reasons explained by Professor A yer in his introduction, several pieces which cannot be regarded as expositions o r defenses of logical positivism have also been included. The scope of the bibliography, too, is broader than the title of the book would sug­ gest. N o volum e dealing with other forms of analytic philosophy is contem plated in this series and it was therefore thought desirable to list th e m ost im portant books and articles dealing with all types of analytic philosophy and not only with logical positivism. I wish to express my gratitude to all the translators who gen­ erously contributed their labor, to M arvin Zim m erm an, James Bayley, Irving Saltzm ann, and a num ber of my students for helping to com ­ pile the bibliography, and to Leon Satinoff, Maxwell G rober, and José H uerta-Jourda for preparing the index. Special thanks are due to Professors C arn ap and Hem pel for supplying notes which indi­ cate their present position on the issues dealt with in their papers. Professor C arnap was also kind enough to assist in the translation of his own articles. P a u l E dw ards

[v]

Contents

P

refa ce

E

d i t o r ’s

v I n t r o d u c t io n

3

Logical Atom ism 7.

B ertrand Russell “ L o g ic a l A t o m i s m ”

31

,

Philosophy Metaphysics and Meaning 2 . M oritz Schlick “ T h e T u r n in g P o i n t i n P h il o s o p h y ”

53

3 . R u d o lf Carnap “ T h e E l im in a t i o n o f M e t a p h y s ic s t h r o u g h L o g ic a l A n a l y s is o f L a n g u a g e ”

60

4. M oritz Schlick “ P o s it iv is m a n d R e a l i s m ”

5.

82

Carl G. H em pel “ T h e E m p i r i c i s t C r i t e r i o n o f M e a n in g ”

108

Logic and Mathematics 6. R u d o lf Carnap “ T h e O l d a n d t h e N e w L o g ic ”

133

7. H ans Hahn “ L o g ic , M a t h e m a t i c s a n d K now ledge of N atu re”

[vii]

147

Contents

[ viii]

Knowledge and Truth 8. R u d o lf Carnap “ P sychology

9.

in

P h y s ic a l L

anguage”

O tto Neurath 199

“P r o to c o l S e n te n c e s ”

10.

165

M oritz Schlick “T

he

F o u n d a t io n

of

K

now ledge”

209

11. A . J. A yer “ V e r if ic a t io n

and

E

x p e r ie n c e ”

228

Ethics and Sociology 12. M oritz Schlick “ W h a t I s t h e A im o f E t h ic s ?”

13.

247

C . L . Stevenson “T h e E m o t iv e M e a n in g o f E t h ic a l T e r m s ”

264

14. O tto Neurath “S o c io l o g y a n d P h y s ic a l is m ”

282

Analytical Philosophy 15. Frank D. R am sey “ P h il o s o p h y ”

16.

321

Gilbert R yle “ P h il o s o p h ic a l A r g u m e n t s ”

17. Friedrich W aismann “H o w I S e e P h il o s o p h y ”

327 345

B ib l io g r a p h y o f L o g ic a l P o s it iv is m

381

I ndex

447

LO GICAL

POSITIVISM

Editor s Introduction

I . H is t o r y o p t h e L o g ic a l P o s it iv is t M o v e m e n t T h e t e r m “ Logical Positivism ” was coined some thirty years ago

to characterize the standpoint of a group of philosophers, scientists and m athem aticians who gave themselves the nam e of the Vienna Circle. Since th at time its reference has been extended to cover other form s of analytical philosophy; so that disciples of Bertrand Russell, G. E. M oore o r Ludwig W ittgenstein at Cam bridge, or m em bers of the contem porary O xford movement of linguistic analysis m ay also find themselves described as logical positivists. This wider usage is especially favored by those who are hostile to the whole m odern developm ent of philosophy as an analytical rather than a speculative enquiry. They wish to ta r all their adversaries with a single brush. This is irritating to the analysts themselves who are rath er m ore sensitive to their differences; they would prefer that the appellation of “logical positivist” be reserved for those who share the special outlook of the V ienna Circle. In compiling this anthology, I have not been quite so strict. I have draw n mainly on the writings of the m em bers of the V ienna Circle, o r of those who stand closest to them , but I have also included several pieces which fall outside this range. They are all, in some sense, analytical but the scope of w hat I regard as analytical philosophy is wide. It allows for serious disagreem ent, not only over technical niceties, but on m ajor points of doctrine, including the m ethod and purpose of analysis itself. T he V ienna Circle cam e into being in the early 1920’s when M oritz Schlick, around whom it centered, arrived from Kiel to be­ com e professor of philosophy at the University of Vienna. O n the philosophical side its leading m em bers, besides Schlick himself, were R udolf C arnap, O tto N eurath, H erbert Feigl, Friedrich Waism ann, E dgar Zilsel and V ictor K raft; on the scientific and m athe­ m atical side, Philipp F rank, Karl M enger, Kurt Godel and H ans H ahn. A t the beginning, it was m ore of a club than an organized m ovem ent. Finding that they had a com m on interest in, and a [3]

[ 4 ]

Editor's Introduction

sim ilar approach to, a certain set of problem s, its m em bers m et regularly to discuss them . These meetings continued throughout the life of the Circle but they cam e to be supplem ented by other activi­ ties which transform ed the club into som ething m ore nearly resem ­ bling a political party. This process began in 1929 with the publi­ cation of a m anifesto entitled “W issenschaftliche W eltauffassung, D er W iener Kreis”— The V ienna Circle; Its Scientific O utlook— which gave a brief account of the philosophical position of the group and a review of the problem s in the philosophy of m athem atics and of the physical and social sciences that they were chiefly concerned to solve. This pam phlet, which was w ritten by C arnap, N eurath and H ahn, is also of interest as showing how the Circle situated itself in the history of philosophy. A fter claiming that they were developing a Viennese tradition which had flowered at the end of the nineteenth century in the work of such men as the physicists E m st M ach and Ludwig Boltzm ann, and, in spite of his theological interests, the philosopher F ranz B rentano, the authors set out a list o f those w h o -i they regarded as their main precursors. As em piricists and positivists they nam ed H um e, the philosophers of the enlightenm ent, C om te, Mill, A venarius and M ach; as philosophers of science, H elm ­ holtz, Riem ann, M ach, Poincare, Enriques, D uhem , B oltzm ann and Einstein; as pure and applied logicians, Leibniz, Peano, Frege, Schroder, Russell, W hitehead and W ittgenstein; as axiom atists, Pasch, Peano, V ailati, Pieri and H ilbert; and as m oralists and sociologists o f a positivistic tem per, Epicurus, H um e, B entham , M ill, Com te, Spencer, Feuerbach, M arx, M üller-Lyer, Popper-L ynkeus and the elder Carl M enger. This list is surprisingly com prehensive, but it m ust be rem em bered that in m ost cases it is only a question of a special aspect of the author’s works. Thus Leibniz is included for his logic, not for his m etaphysics; K arl M arx is included neither for his logic nor his metaphysics but for his scientific approach to history. If we exclude contem poraries from the list, those who stand closest to the V ienna Circle in their general outlook are H um e and M ach. It is indeed rem arkable how much of the doctrine that is now thought to be especially characteristic of logical positivism was already stated, o r at least foreshadowed, by Hum e. Am ong contem poraries, Einstein, Russell, and W ittgenstein are singled out by the authors of the pam phlet for their kinship to the V ienna Circle and the extent of their influence upon it. W ittgenstein, indeed, stood to the Vienna Circle in a special relation. H aving been a pupil of Russell’s at Cam bridge before the first world w ar he returned to V ienna and was there when his Logisch-Philosophische

Editor's Introduction

[5 ]

A bhandlung was published in 1921. This fam ous book, which is better known as Tractatus Logico-Philosophicus, the title given to its English translation, had an enorm ous effect upon the positivist m ovem ent, both in V ienna and elsewhere. It would not be quite correct to say that the V ienna Circle drew its inspiration from it. Schlick himself, in his book on the theory of knowledge, Allgem eine E rkenntnislehre, of which the first edition appeared in 1918, had independently arrived at a sim ilar conception of philosophy; and there is a hint of mysticism in the Tractatus which some members of the Circle, especially N eurath, found disquieting; but as a whole they ac­ cepted it, and it stood out as the most powerful and exciting, though not indeed the m ost lucid, exposition of their point of view. W ittgen­ stein did not officially adhere to the Circle but he m aintained close personal relations at least with Schlick and W aismann whom he con­ tinued to influence even after his departure for Cambridge in 1929. In Cam bridge, where he taught until 1947, four years before his death, he exercised an alm ost despotic sway over his pupils, and though he published nothing during these years except one short article his influence was strongly, if in m ost cases indirectly, felt by alm ost all the younger generation of British philosophers. He himself modified the rigors of his early positivism to an extent that can be m easured by com paring the Tractatus with his posthum ously pub­ lished Philosophical Investigations; and it is to his influence, com ­ bined with that of M oore, that one may largely attribute the p re­ occupation of contem porary British philosophers with the everyday uses of language, and their tendency to deal with philosophical ques­ tions in an unsystem atic, illustrative way, in contrast to the more rigorous would-be scientific m ethod which was favored by the Vienna Circle. This is one reason why they are not happy to be described as Logical Positivists. But I shall have more to say about these alter­ native conceptions of analysis later on. It was in 1929 also that the V ienna Circle organized its first in­ ternational congress. It was held at Prague and was followed at inter­ vals throughout the thirties by further congresses at Königsberg, C openhagen, Prague, Paris and Cam bridge. These meetings furthered the am bition of the Circle to develop Logical Positivism as an inter­ national m ovem ent. It had form ed an early alliance with the so-called Berlin school of which H ans R eichenbach, R ichard von Mises, Kurt G reiling and at a later date Carl Hem pel were the leading members. The congresses helped it to m ake contact also with Scandinavian philosophers such as Eino Kaila, A rne Naess, Ake Petzäll, Joergen Joergensen, and the U ppsala school of empiricists, with the Dutch

[6 ]

Editor’s Introduction

group around the philosopher M annoury who pursued w hat they called the study of Signifies, with the M ünster group of logicians under Heinrich Scholtz, with A m erican sym pathizers such as Nagel, C harles M orris and Quine, and with British analysts of various shades of opinion, such as Susan Stebbing, G ilbert Ryle, R. B. B raithw aite, John W isdom and myself. T h e brilliant C am bridge philosopher F . P. Ram sey was m arked as an adherent, bu t he died in 1930 at the early age of 26. A n alliance was also form ed with the very im portant Polish groups of philosophers and logicians, of whom Lucasiewicz, Lesnievsky, Chwistek, K otarbinski, Ajduciewicz and T arski w ere p er­ haps the m ost prom inent. T he influence of T arski’s w ork, particularly on C arnap, was noticeably strong. T he missionary spirit of the Circle found a further outlet in its publications. In 1930 it took over a journal called A nnalen der Philosophie, renam ed it E rkenntnis and m ade it, under the editorship of C arnap and R eichenbach, the principal organ o f the positivist m ovem ent. In the following years there also appeared a series of m onographs with the collective title of Einheitswissenschaft— Unified Science— and a series of books, under the general editorship of Schlick and Philipp F ran k , with the collective title of Schriften zur Wissenschaftliche W eltauffassung. Schlick him self contributed to it a book on ethics, of which the first chapter is included in this volum e, and F rank a book on the law of causality and its lim itations. A m ong the other volumes to appear in it were an im portant book by C arnap on the logical syntax of language, to which I shall have occasion to refer again, a book on sociology by N eurath with som ew hat M arxist leanings, and Karl P opper’s fam ous L o g ik der Forschung which was devoted to the philosophy of science. P o pper was not in fact a m em ­ ber of the Circle and would at no time have wished to be classed as a positivist, b u t the affinities betw een him and the positivists whom he criticized appear m ore striking than the divergencies. In any case the m em bers of the Circle did not in all points agree am ong them ­ selves. Though the logical positivist m ovem ent gathered strength through­ out the thirties, the V ienna Circle itself was in the process of dissolu­ tion. By 1933, when I attended its meetings, C arnap and F ran k had accepted chairs at the University of Prague and the discussions were chiefly carried on by Schlick, N eurath, W aism ann and H ahn. But H ahn died in 1934 and two years later Schlick was m urdered, at the age of 54, by a dem ented student who shot him as he was entering the University. T he hostile tone of the obituaries which were devoted to Schlick in the governm ental press, implying alm ost th at logical

Editor's Introduction

[7 ]

positivists deserved to be m urdered by their pupils, foreshadowed the troubles which were soon to fall upon the Circle. Except for N eurath, who had participated in the revolutionary Spartacist Gov­ ernm ent in M unich at the end of the first world war, its m embers had not been conspicuously active in politics, but their critical and scientific tem per m ade them suspect to the right-wing clerical govern­ m ents of Dolfuss and Schuschnigg and still more so to the Nazis. The m ajority of them were forced into exile. The advent of Nazism was fatal also to the Berlin school, and the Polish groups were disrupted by the war. N eurath, who had taken refuge in Holland, made a valiant attem pt to keep the m ovem ent going. T he title of Erkenntnis was changed to The Journal o f Unified Science and its place of publica­ tion to the H ague. A rrangem ents were m ade for the publication by the University of Chicago, where C arnap was established, of a series of brochures am bitiously entitled the International Encyclopedia of Unified Science. F u rth er congresses were planned. But with the out­ break o f w ar and N eurath’s death in England some years later, the m ovem ent lost its cohesion. M ost of the volumes which were designed to constitute the Encyclopedia have in fact appeared, but the Journal o f Unified Science very soon ceased publication and has not been revived. Besides C arnap, Feigl, Godel, F rank, Hem pel and Tarski are still at universities in the United States, and W aismann and Popper at universities in England. Scholtz has rem ained at M unster and K otarbinski and Ajduciewicz in Poland; and V ictor K raft resumed his chair of philosophy at the University of Vienna. But, however much influence these philosophers may exert individually, they do not constitute a school. In this sense, the logical positivist m ovem ent has been broken up. Nevertheless its tradition has been continued, especially in Eng­ land, Scandinavia and the U nited States. In Scandinavia, Kaila has been joined at Helsinki by Von W right, a pupil of W ittgenstein’s who succeeded him fo r a tim e as professor of philosophy at Cambridge, the U ppsala school still flourishes, under the direction of Hedenius, Segerstedt and M arc-W ogau, with support from the logician W edberg in Stockholm , and A rne Naess in Oslo pursues his sociologi­ cal researches into the current uses of language. Petzall continued to teach at L und until his death in 1957 and Joergensen is still teach­ ing in Copenhagen, though his positivism has been modified by an injection of M arxism. In the United States a num ber of philosophers like Q u in e , Nagel and Nelson G oodm an conduct logical analysis in a system atic scientific spirit th at is probably closer to the original ideal of the V ienna Circle than anything that is now to be m et with else-

[8 ]

Editor’s Introduction

where. In this connection Nelson G oodm an’s book The Structure o f Appearance (1 9 5 1 ) and Q uine’s collection of essays From a Logical Point o f View (1 9 5 3 ) are especially notable. T heir active interest in symbolic logic brings Quine and G oodm an also into relation with Tarski, Godel, C hurch and other mem bers of the im portant con­ tem porary group of A m erican logicians. The sam e outlook is m ain­ tained by C arnap and his pupils, notably B ar Hillel, who is now teaching at the University of Jerusalem , and by Feigl and Hem pel. O ther philosophers in the United States such as N orm an M alcolm , M ax Black, M orris Lazerow itz and C. L. Stevenson owe m ore to the influence of G. E. M oore o r the later W ittgenstein, and conse­ quently display an approach to philosophical questions which is closer to that of the contem porary British schools. In spite of the exam ple of B ertrand Russell, there is not now among British philosophers the same interest in form al logic, o r belief in the utility of symbolic techniques for clarifying philosophical issues, as is to be found in the United States. N either is there the same desire to connect philosophy with science. M y own Language, Truth and Logic, of which the first edition appeared in 1936, did something to popularize w hat may be called the classical position of the Vienna Circle; but since the w ar the prevailing tendency in Eng­ land has been to replace this uncom prom ising positivism with its blanket rejection of metaphysics, its respect for scientific m ethod, its assum ption th at in so far as philosophical problem s are genuine at all they can be definitely solved by logical analysis, by an approach to philosophy which is em pirical in the political sense, the sense in which Burke was a cham pion of empiricism. G eneralizations are dis­ trusted, particular examples are m ultiplied and carefully dissected. A n attem pt is m ade to illuminate every facet of a problem rather than to ham m er o r carve out a solution, com m on sense reigns as a con­ stitutional, if not an absolute, m onarch, philosophical theories are put to the touchstone of the way in which words are actually used. The m etaphysician is treated no longer as a crim inal but as a patient: there may be good reasons why he says the strange things th at he does. This therapeutic technique, as it has been called, is well dis­ played in the work of John W isdom, now a professor at Cam bridge, whose collected volumes of articles, O ther M inds and Philosophy and Psycho-Analysis, appeared in 1952 and 1953. A m ore robust form of therapy is practised by Gilbert Ryle, professor of metaphysics at Oxford, whose Concept o f M ind (1 9 4 9 ), with its attack on the C artesian myth of “the ghost in the m achine,” has had a very great

Editor's Introduction

[9 ]

influence. Ryle shares with W isdom a taste, and a gift, for analogy and m etaphor, and a fondness for piling up examples, but he is less afraid of a generalization, less tolerant of departures from ordinary usage, m ore direct in his m ethod than any present-day W ittgensteinian, and more ready to assume that a philosophical problem has a correct solution. W hat is now sometimes called the Oxford school, which takes its tone from J. L. A ustin more than from Ryle, carries its interest in the ordinary use of language to a point where it may be thought th at philosophical analysis has given way to the study of philology. But this tendency is not all-prevailing. The work of such philosophers as Stuart H am pshire, P. F. Strawson and David Pears shows that even within the fram ew ork of the Oxford m anner there is still room for a fairly wide latitude of approach. The charge of scholasticism which is brought against “Oxford philosophy” is not entirely baseless; but it is not a truly warranted indictment. A t the present tim e, the philosophical world is curiously divided. If positivism be taken in its widest sense, the sense in which it em­ braces all shades of analytical, linguistic, or radically empirical philosophy, it is dom inant in England and in Scandinavia, and com ­ m ands considerable allegiance in Holland and Belgium, in Australia and in the United States. Elsewhere, it makes hardly any showing at all. Theoretically, it is not in all respects at odds with M arxism: the two at least have certain enemies in com m on: but it cannot flourish under Com m unist regimes, since Lenin’s M aterialism and Em pirioCriticism, an attack on M ach and his followers which appeared in 1905, declares it to be a form of bourgeois idealism. In other coun­ tries again, one finds philosophers subscribing to neo-Thomism o r to neo-K antianism or to neo-Hegelianism or to Existentialism or w hat­ ever form of G erm an metaphysics may be in fashion. The ascend­ ancy of G erm any over France in this respect is especially remarkable. Conversely, in English-speaking countries there has been throughout the present century an almost complete disregard of the current extravagancies of G erm an speculative thought. Such national divisions are indeed regrettable. They do not occur to anything like the same extent in other branches of learning. It is especially characteristic of philosophers that they tend to disagree not merely about the solution of certain problem s but about the very nature of their subject and the m ethods by which it is to be pursued. Like others before them, the V ienna Circle believed that this could and should be remedied. They thought th at they had succeeded, where K ant had failed, in finding a way “to set philosophy upon the sure path of a science.” This end

Editor's Introduction

[ 10 ]

has not been attained: it may, indeed, be unattainable. All the same, there can be progress in philosophy and in one way and another the positivist movem ent is achieving it. II. T h e A t t a c k

on

M e t a p h y s ic s

“ When we run over libraries, persuaded of these principles, w hat havoc m ust we make? If we take in o ur hand any volum e; of divinity o r school metaphysics, for instance; let us ask, D oes it contain any abstract reasoning concerning quantity or number? N o. D oes it con­ tain any experim ental reasoning concerning m atter o f fact and exist­ ence? No. Commit it then to the flames: for it can contain nothing but sophistry and illusion.” This quotation is taken from D avid H um e’s Enquiry Concerning H um an Understanding. It is an excellent state­ m ent of the positivist’s position. In the case of the logical positivists, the epithet “logical” was added because they wished to annex the dis­ coveries of m odern logic; they believed, in particular, that the logical symbolism which had been developed by Frege, Peano and Russell would be serviceable to them . But their general outlook was very much the same as H um e’s. Like him, they divided significant proposi­ tions into two classes; form al propositions, like those of logic o r pure m athem atics, which they held to be tautological, in a sense th at I shall presently explain, and factual propositions, of which it was re­ quired that they should be em pirically verifiable. These classes were supposed to be exhaustive: so th at if a sentence succeeded neither in expressing something that was form ally true or false nor in express­ ing something that could be empirically tested, the view taken was that it did not express any proposition at all. It m ight have em otive meaning but it was literally nonsensical. A great deal of philosophical talk was held to fall into this category: talk about the absolute, or transcendent entities, o r substance, or the destiny of m an. Such utterances were said to be m etaphysical; and the conclusion draw n was th at if philosophy was to constitute a genuine branch o f know ­ ledge it m ust em ancipate itself from metaphysics. The Viennese positivists did not go so far as to say th at all m etaphysical works deserved to be com m itted to the flames: they allowed, som ew hat per­ functorily, th at such writing might have poetic m erit o r even th at it might express an exciting or interesting attitude to life. T heir point was that even so it did not state anything th at was either true o r false and consequently th at it could contribute nothing to the increase of knowledge. M etaphysical utterances were condem ned not for being emotive, which could hardly be considered as objectionable in itself, but for pretending to be cognitive, for m asquerading as som ething

Editor's Introduction

[ \\ ]

that they were not. A ttacks on m etaphysics occur fairly frequently in the history of philosophy. I have quoted Hum e and I might also have quoted K ant who m aintained that the hum an understanding lost itself in contradictions when it ventured beyond the bounds of pos­ sible experience. T he originality of the logical positivists lay in their m aking the impossibility of m etaphysics depend not upon the nature o f w hat could be known but upon the nature of what could be said. T h eir charge against the m etaphysician was that he breaks the rules which any utterance m ust satisfy if it is to be literally significant. A t the outset, their form ulation of these rules was linked with a conception of language which W ittgenstein, who inherited it from Russell, m ade fully explicit in his Tractatus. The underlying assum p­ tion is that there are statem ents which are elem entary in the sense that, if they are true, they correspond to absolutely simple facts. It m ay be that the language which we actually use does not contain the m eans of expressing these statem ents: the statem ents which it can serve to express may none of them be entirely elem entary; but these m ore com plex statem ents must still rest upon a foundation of elem entary statem ents, even if the foundation be hidden. They are significant only in so far as they say w hat would be said by affirming certain elem entary statem ents and denying certain others, that is, only in so far as they give a true or false picture of the ultimate “ atom ic” facts. T hey can, therefore, be represented as being constructed out of elem entary statem ents by the logical operations of conjunction and negation, in such a way that their truth or falsehood is entirely dependent on the truth o r falsehood of the elem entary statem ents in question. Thus, assum ing p and q to be elem entary statem ents, the “ m olecular” statem ent “p o r q” is taken to be equivalent to “not (n o t-p and n o t-# )” ; and this m eans that it is false if both p and q are false, but true in the three rem aining cases, namely that in which p and q are both true, that in which p is true and q false, and that in which p is false and q true. In general, given n elem entary statements, where n is any finite num ber, there are 2n possible distributions of truth and falsehood am ong them : and the meaning of the m ore com ­ plex statem ents which can be constructed out of them is constituted by the selection of truth distributions with which they agree or disagree. A s a rule, it will be found th at a statem ent agrees with some truth distributions and disagrees with others: am ong the possible states of affairs with which it is concerned, some would m ake it true, and others would m ake it false. T here are, however, two extrem e cases; th at in which a statem ent agrees with every truth distribution and

[ 12 ]

Editor's Introduction

that in which it agrees with none. In the form er case it is true in any circum stances whatsoever, and in the latter case false. A ccording to W ittgenstein, these two extrem es are those of tautology and contra­ diction. O n this view, all the truths of logic are tautologies; and if Russell and W hitehead succeeded in their attem pt to show that m athe­ matics is reducible to logic, so are the truths of m athem atics. W itt­ genstein himself did not allow that m athem atical statem ents were tautologies; he said that they were identities: but ap art from technical considerations, this comes to much the sam e thing. T he point is th at neither say anything about the world. T he only way in which they can add to o ur knowledge is by enabling us to derive one statem ent from another: that is, by bringing out the implications of w hat, in a sense, we know already. Tautologies say nothing because of their excessive m odesty: since they agree with every possible state of affairs, they m ake no claim upon the facts. Thus, I obtain some inform ation, w hether true o r false, about the habits of lions if I am told th at they are carnivorous, and equally if I am told th at they are not; but to tell me that they are either carnivorous o r not is to tell me nothing about them at all. Similarly, contradictions say nothing because of their excessive can ­ tankerousness; to disagree with every possible state of affairs is again to be disqualified from giving any inform ation. I learn nothing, not even anything false, about the habits of lions if I am told th a t they are and are not carnivorous. O n this interpretation, tautologies and contradictions are degenerate cases of factual statem ents. M eta­ physical assertions, on the other hand, are meaningless because they bear no relation to fact. They are not constructed out of elem entary statem ents in any way at all. Since W ittgenstein did not say w hat he took his elem entary statem ents to be, he did not m ake it quite clear at w hat point one is deem ed to enter into metaphysics. It would seem , however, th at any attem pt to characterize reality as a whole, any such assertion as that the Universe is spiritual, o r that everything happens for the best in the best of all possible worlds, m ust for him be m etaphysical; for such assertions do not discrim inate between possible states of affairs within the world— no m atter w hat happens, it is to be ch ar­ acterized as spiritual, o r regarded as happening for the best— from which it follows th at they are not factual. N either do they seem to be constructed out of factual statem ents in the way that tautologies are. And even if they were they would still say nothing. W hatever may have been W ittgenstein’s own view, his followers took it for granted th at the elem entary statem ents which yielded this

Editor’s Introduction

[ 13 ]

criterion of m eaning were reports of observations. As we shall see later on, they soon cam e to disagree about the character of these reports. T here was a dispute over the question whether they were infallible, and w hether they referred to the private sensations of the speaker, o r to public physical events. But it was agreed that, in one form o r another, they provided the touchstone by reference to which all other statem ents were empirically verified. And since, according to W ittgenstein’s theory, they alone furnished these statem ents with their factual content, they were also responsible for their meaning. This view was then sum m ed up in the fam ous slogan that the meaning of a proposition is its m ethod of verification. T he assum ption behind this slogan was that everything that could be said at all could be expressed in term s of elem entary statements. A ll statem ents of a higher order, including the most abstract scientific hypotheses, were in the end nothing more than shorthand descriptions of observable events. But this assum ption was very difficult to sus­ tain. It was particularly vulnerable when the elementary statem ents were taken to be records of the subject’s immediate experiences: for while it has sometim es been m aintained that statem ents about physical objects can be faithfully translated into statem ents about sense-data, no such translation has ever been achieved: there are, indeed, good grounds fo r supposing that it is not feasible. M oreover this choice of a basis raised the question of solipsism; the problem of m aking the transition from the subject’s private experiences to the experiences of others and to the public world. C arnap, indeed, in his Der logische A u fb a u der W elt (1 9 2 8 ) m ade a valiant attem pt to reconstruct our whole apparatus of em pirical concepts on a solipsistic foundation, taking as his starting-point the single undefined notion of remembered sim ilarity: bu t he later acknowledged that this enterprise did not succeed. T he position was easier for those who treated elementary statem ents as descriptions of physical events, though their right to do this rem ained in question: they at least were not troubled by the problem of solipsism o r by the difficulty of reducing physical objects to sense-data. But other difficulties rem ained. T he m ost serious of all, perhaps, was presented by the case of universal statem ents of law. F o r while the truth of such a statem ent may be confirmed by the accum ulation of favorable instances, it is not formally entailed by them ; the possibility th at a further instance will refute it must always rem ain open: and this m eans that statem ents of this sort are not conclusively verifiable. O n the other hand, they can be conclusively falsified in the sense that a negative instance formally contradicts them . F o r this reason Karl Popper suggested in his

[ 14 ]

Editor's Introduction

Logik der Forschung that w hat should be required of a factual statem ent was just that it be capable in theory of being falsified. A nd he argued that apart from the logical superiority of this cri­ terion it was m ore in accord with scientific practice; for scientists set up hypotheses which they test by looking for counter-exam ples: when a counter-exam ple is discovered the hypothesis is rejected o r modified; otherwise it is retained. But P opper’s criterion has dem erits of its own. F o r instance, as he him self recognizes, it allows one to deny an indefinite existential statem ent but not to affirm it. One can say that there are no abom inable snowm en, for this could be falsified by finding them , but one cannot say that there are abom inable snowmen, for this could not be falsified; the fact th at one had failed to find any would not prove conclusively th at none existed. W hat could be disproved would be that any of them existed at a particular place and time, and it is only if this further specification is given th at the statem ent becom es legitimate: otherwise it is to be counted as metaphysical. But this is to bring the frontiers of metaphysics rather close. Because of this and other difficulties the view which cam e to prevail among the logical positivists was th at the dem ands th at a statem ent be conclusively verifiable, o r that it be conclusively falsifiable, were both too stringent as criteria of m eaning. They chose instead to be satisfied with a w eaker criterion by which it was re­ quired only th at a statem ent be capable of being in some degree con­ firmed o r disconfirmed by observation; if it were not itself an elementary statem ent, it had to be such that elem entary statem ents could support it, but they did not need to entail it o r to entail its negation. U nfortunately, this notion of “ support” o r “confirm ation” has never yet been adequately form alized. V arious attem pts have been m ade to give “the verification principle,” in this w eaker form , a thoroughly precise expression, but the results have not been alto­ gether satisfactory. H owever, the em ploym ent of the principle did not wait upon its proper form ulation; its general pu rp o rt was held to be sufficiently clear. I have already given exam ples of the kind of philosophical talk th at it served to elim inate: but its destructiveness was not confined to w hat one might call the grosser form s of m eta­ physics. As employed by the Viennese positivists, it m ade short work of most of the perennial problem s of philosophy. T hus, the questions at issue between monists and pluralists, or between realists and idealists, were accounted no less spurious than questions about the limitations of Being, or a transcendent world of values. F o r w hat em pirical test could possibly go to decide w hether the w orld is one

Editor's Introduction

[ 15 ]

or m any, o r w hether the things that we perceive do or do not exist outside som eone’s mind? It is characteristic of such rival philo­ sophical theses as realism and idealism that each is consistent with all the appearances, whatever their content may happen to be. But, for the positivist, it is just this that condem ns them. A n obvious objection to the verification principle, which the positivists’ opponents were quick to seize on, is that it is not itself verifiable. I suppose th at it might be taken as an empirical hypothesis about the way in which people actually use the word “meaning,” b ut in th at case it would appear to be false; for it is not contrary to ordinary usage to say that metaphysical utterances are meaningful. N either did its sponsors put the principle forward as the result of any such em pirical investigation. But then what status did they think it had? M ight it not itself be m etaphysical? Surprisingly, W ittgenstein acceded to this charge. “ My propositions,” he said at the end of the Tractatus, “ are elucidatory in this way: he who understands me finally recognizes them as senseless, when he has climbed out through them , on them , over them . (H e must so to speak throw away the ladder, after he has climbed up on it.) H e m ust surm ount these propositions; then he sees the world rightly.” But this is a vain attem pt to have it both ways. N o doubt some pieces of nonsense are m ore suggestive than others, but this does not give them any logical force. If the verification principle really is nonsensical, it states nothing; and if one holds that it states nothing, then one can­ not also m aintain that w hat it states is true. T he V ienna Circle tended to ignore this difficulty: but it seems to me fairly clear that w hat they were in fact doing was to adopt the verification principle as a convention. They were propounding a definition of m eaning which accorded with com m on usage in the sense that it set out the conditions that are in fact satisfied by state­ m ents which are regarded as empirically informative. T heir treatm ent of a priori statem ents was also intended to provide an account of the way in which such statem ents actually function. T o this extent their w ork was descriptive; it becam e prescriptive with the suggestion th a t only statem ents of these two kinds should be regarded as either true o r false, and that only statem ents which were capable of being either true o r false should be regarded as literally meaningful. But why should this prescription be accepted? The most that has been proved is th at metaphysical statem ents do not fall into the same category as the laws of logic, or as scientific hypotheses, or as historical narratives, o r judgm ents of perception, or any other com m on sense descriptions of the “ natural” world. Surely it does not

[16]

Editor’s Introduction

follow th at they are neither true nor false, still less that they are nonsensical? No, it does not follow. O r rather, it does not follow unless one m akes it follow. The question is w hether one thinks the difference between metaphysical and com m on sense o r scientific statem ents to be sufficiently sharp for it to be useful to underline it in this way. T he defect of this procedure is that it tends to m ake one blind to the interest that metaphysical questions can have. Its m erit is that it removes the tem ptation to look upon the m etaphysician as a sort of scientific overlord. N either is this a trivial m atter. It has far too often been assumed that the m etaphysician was doing the sam e w ork as the scientist, only doing it m ore profoundly; that he was uncover­ ing a deeper layer of facts. It is therefore im portant to em phasize th at he is not in this sense describing any facts at all. B ut then what is he doing? W hat is the point of saying, like M cTaggart, that time is unreal or, like Berkeley, that physical objects are ideas in the mind of God or, like Heidegger, th at the “ nothing nihilates itse lf’? It should not be assum ed th at there is a general answer to this question, that m etaphysicians are always doing the same sort of thing. O ne must begin in any case by looking a t the context in which such pronouncem ents occur. Heidegger’s rem ark is a piece of verbiage, but it contributes in its fashion to his develop­ m ent of the theme that it is a m atter for w onder that the world exists. “Why is there anything at all,” he asks, “ and not rather nothing?” This is indeed the kind of question th at people expect philosophers to put: it has an air of profundity about it. T he trouble is th at it does not adm it of any answer. O n the face of it, M cTaggart’s con­ tention that time is unreal seems hardly more sensible. If taken literally, as implying that nothing ever happens, it is grotesquely false. A nd if it is not to be taken literally, w hat does it m ean? T he answer is to be found by looking at M cTaggart’s argum ents. H e shows himself there to be perplexed by the idea of the passage of tim e; he tries to prove that the notion of an event’s being successively future, present, and past involves a vicious infinite regress. T he proof is invalid, but we can learn som ething from it. In defending o u r use of tem poral expressions against M cTaggart’s argum ents we m ay reach a clearer understanding of all th at it implies. Berkeley, again, was concerned to discover w hat could be m eant by saying th at physical objects exist: he convinced himself by plausible argum ents th at when we speak of physical objects we can be referring only to collec­ tions of “ sensible qualities,” the existence of which consists in their being perceived; and he then brought in G od as the perm anent sen-

Editor's Introduction

[ 17 ]

sorium which was needed to keep things in being. His argum ents can be w ithstood; but they do raise im portant philosophical problems about the m eaning and justification of the statem ents that we make about the “external w orld.” T he Viennese positivists were chiefly interested in the form al and the natural sciences. They did not identify philosophy with science, b ut they believed that it ought to contribute in its own way to the advance of scientific knowledge. They therefore condem ned m eta­ physics because it failed to meet this condition. The logical analysts o f to-day are m ore indulgent. They too are opposed to metaphysics in so far as it is merely rhapsodical: even in the sphere of ethics they wish to dissociate philosophy from preaching. But they allow th at the m etaphysician may sometimes be seeing the world in a fresh and interesting way; he may have good reason for being dissatisfied with our ordinary concepts, o r for proposing to revise them . In many cases no doubt he is the victim of a logical error; but such errors m ay be instructive. If philosophical problem s arise, as W ittgenstein thought, because we are led astray by certain features of our lan­ guage, the m etaphysician, by his very extravagancies, may also contribute to their dissolution. III. L a n g u a g e

and

F act

W ith their elim ination of m etaphysics, the Viennese positivists hoped th at they had also p u t the theory of knowledge behind them, but in this they were deceived. T he first source of trouble was the notion of elem entary statem ents. Both their character and status becam e a m atter of dispute. A t the outset, as I have said, the prevailing view was that these statem ents referred to the subject’s introspectible o r sensory experi­ ences. This view was adopted because it seemed to follow from the equation of the m eaning of a statem ent with the m ethod of its veri­ fication. F o r in the last resort it is only through som eone’s having some experience th at any statem ent is actually verified. In m ost cases, the verification would consist in the perception of some physical object; b u t it was held, following Russell and ultimately Berkeley, that perceiving physical objects was to be analyzed in term s of hav­ ing sensations, o r as Russell put it, of sensing sense-data. Though physical objects might be publicly accessible, sense-data were taken to be private. T here could be no question of our literally sharing one another’s sense-data, any m ore than we can literally share one an­ o th er’s thoughts o r images o r feelings. The result was that the truth

[ 18 ]

Editor’s Introduction

of an elem entary statem ent could be directly checked only by the person to whose experience it referred. A nd not only was his judg­ ment sovereign; in the m ost favorable case, it was held to be infallible. One can indeed be m istaken about the experiences th at one is going to have in the future, or even about those that one has had in the past; it is not m aintained th at o ur m em ories cannot deceive us: but if one sets out merely to record an experience that one is actually having, then, on this view, there is no possibility o f error. Since one can lie, one’s statem ent may be false; but one cannot be in doubt o r m istaken about its truth. If it is false one knows it to be so. A way in which this point is sometimes put is by saying th at statem ents of this kind are “incorrigible.” This conception of elem entary statem ents was exposed to attack on various grounds. T here were some to whom it seem ed th a t no em pirical statem ent could be incorrigible, in the sense required. They were therefore inclined to m aintain either th a t one could be m istaken about the character of one’s present experience, so th at the statem ents which purported to record it were fallible like the rest, o r th a t these “direct records of experience” were not genuine statem ents, since they purchased their security at the expense of sacrificing all descriptive content. But the m ost serious difficulty lay in the privacy of the objects to which the elem entary statem ents were supposed to refer. If each one of us is bound to interpret any statem ent as being ultim ately a description of his own private experiences, it is hard to see how we can ever com m unicate at all. Even to speak of “each one o f u s” is to beg a question; for it would seem th at on this view the supposition that other people exist can have no m eaning for me unless I construe it as a hypothesis about my own observations of them , th at is, about the course of my own actual o r possible experiences. It was m aintained by C arnap and others th at the solipsism which seemed to be involved in this position was only m ethodological; but this was little m ore than an avowal of the purity of their intentions. It did nothing to mitigate the objections to their theory. A t first, it was thought that the difficulty about com m unication could be m et by drawing a distinction betw een the content of ex­ periences and their structure. C ontent, it was m aintained, was incom ­ m unicable. Since other people cannot sense my sense-data, o r share my thoughts o r feelings, they cannot verify the statem ents that I m ake about them ; neither can I verify the corresponding statem ents that they m ake about their experiences. A nd if I cannot verify them , I cannot understand them either. T o this extent we inhabit entirely different worlds. W hat can be verified, however, is th at these worlds

Editor's Introduction

[ 19 ]

have a sim ilar structure. I have no m eans of telling that the feeling which ano th er person records when he says that he is in pain is at all like the feeling that I call pain: I have no means of telling that the colors which he identifies by the use of certain words look at all the sam e to him as the colors for which I use these words look to me. But at least I can observe th at we apply the words on the same occa­ sions, th at his classification of objects according to their color coin­ cides with m ine; I can observe th at when he says he is in pain he displays w hat I regard as the appropriate signs. A nd this is all that is required fo r com m unication. It does not m atter to me w hat my neighbour’s experiences actually are: for all that I can ever know they are utterly different from mine. W hat m atters is that the structure of our respective worlds is sufficiently alike for me to be able to rely on the inform ation th at he gives me. A nd it is in this sense only th at we have a com m on language; we have, as it were, the same canvas which each of us paints in his ow n private fashion. It follows that if there are propositions, like the propositions of science, which have an inter-subjective m eaning, they m ust be interpreted as descriptions of structure. A s I have already rem arked, the fundam ental objection to this view is th at it inconsistently puts the “ private worlds” of other people on a level with one’s own; it results in a curious, and indeed contra­ dictory, theory of m ultiple solipsism. But, apart from this, the dis­ tinction which it tries to m ake between content and structure does not seem to be tenable. F o r w hat would be an exam ple of a statem ent which referred only to structure? There is an echo here of L ocke’s “ prim ary qualities.” B ut statem ents which refer to the “geom etrical” properties of objects, to “figure, extension, num ber and m otion” have to be interpreted in term s of content, just as m uch as statem ents which refer to colors and sounds. If I have no m eans of knowing that my neighbor m eans the same as I do by his use of color-words, I have equally no m eans of knowing th at he m eans the same by his use of w ords which refer to spatial relations or to num erical quantities. I cannot tell even that w hat I take to be the same word really is the sam e for him. All th at I am left with is the apparent harm ony of our behavior. M oreover it seems that the attem pt to draw a distinction w ithin the boundaries of descriptive language between what can and cannot be com m unicated m ust be self-defeating. It leads to the absurdity to which Ram sey draws attention in his short paper on “ Philosophy,” which is included in this volume: “ the position of the child in the following dialogue: ‘Say breakfast.’ ‘C an’t.’ ‘W hat can’t you say?’ ‘C an’t say breakfast.’ ”

[ 20 ]

Editor's Introduction

Because of such difficulties, N eurath, and subsequently C arnap, rejected this whole conception of elem entary statem ents. T hey argued that if elem entary statem ents were to serve as the basis for the intersubjective statem ents of science, they m ust themselves be intersubjective. They m ust refer, not to private incom m unicable experi­ ences, but to public physical events. M ore generally, statem ents which ostensibly refer to experiences, o r to “ m ental” states o r processes of any kind, whether one’s own or anybody else’s, m ust all be equiva­ lent to “ physical statem ents” : for it is only in this way th a t they can be publicly intelligible. This is the thesis of physicalism . I shall not dwell upon it here, as I have inserted an article by C arnap, “ Psychology in Physical Language,” which sets it out at length. T he view th at they were included in “ the physical language” took away from elem entary, or, as N eurath and C arnap called them , “ protocol,” statem ents their privileged position. They w ere no longer thought to be incorrigible. T heir truth, like th at of any other physical statem ents, was always open to question. But, m ore than this, they lost even their judicial status. If a protocol-statem ent conflicted with a statem ent of a higher order, such as a scientific hypothesis, one o r other of them would have to be abandoned, but it need not necessarily be the scientific hypothesis: in certain circum stances it m ight be m ore convenient to reject the protocol-statem ent instead. As can be seen from his paper on the foundation of knowledge ( “ O ber das F undam ent der E rkenntnis” ) Schlick found this con­ clusion unacceptable. H e argued th at to treat the reports of observa­ tion, which was w hat protocol-statem ents were supposed to be, in this cavalier fashion, was to put scientific hypotheses, and indeed all would-be em pirical statem ents, outside the control of fact. N eurath and C arnap, however, were not im pressed by this argum ent. T hey had decided by this time th at it was m etaphysical to talk o f com paring statem ents with facts. F o r w hat could this “com parison” be if not a logical relation? A nd the only thing to which a statem ent could stand in any logical relation was another statem ent. C onsequently, they were led to adopt a coherence theory of truth. T heir version of the coherence theory was in som e ways less objectionable than that which the H egelian idealists had m ade familiar. Even so, for the reasons which I set out in my paper on Verification and Experience, it seems to me quite untenable. C arnap himself abandoned it after he had been convinced by T arski of the respectability of sem antics; for sem antics provides us with the means of referring to the relationship between sentences and w hat they are used to signify. It provides, as T arsk i showed, an adequate reform u­

Editor's Introduction

[ 21 ]

lation of the correspondence theory of truth. O n the other hand C arnap has not, so far as I know, abandoned the thesis of physicalism. But, if he does still hold it, I think that he is m istaken. It now seems clear to me that statem ents about the experiences of others can not be logically equivalent to statem ents about their overt behavior; while to m aintain that the statem ents which one makes about one’s own experiences are equivalent to statem ents about the publicly observ­ able condition of one’s body is, as Ram sey put it, to feign anaesthesia. Nevertheless, the difficulties which this thesis was designed to meet rem ain. N either is it easy to see how else they can be avoided. I suggest, however, that much of the trouble may arise from the ac­ ceptance of two false assum ptions, the first being that for a language to be public it m ust refer to public objects, and the second that in m aking an em pirical statem ent one is always referring to one’s own experiences. I still think that empirical statem ents must refer to experiences, in the sense that they m ust be verifiable; but the reference need not be to the experiences of any one person, as opposed to any other. But I acknowledge that this attem pt to “neutralize” the verifi­ cation principle m eets with considerable difficulties of its own. IV. E t h ic s O ne of the attractions, especially for N eurath, of the thesis of physicalism was that it supported the doctrine of the Unity of Science. In one aspect, this was less of a doctrine than a program ; it was desired th at scientists of different disciplines should collaborate more closely with each other and with philosophers than they usually do: but it was also m aintained that they were, o r should be, speaking a com m on language, that the vocabulary of the sciences should be unified. Thus, the V ienna Circle rejected the view, which m any still hold, th at there is a radical distinction between the natural and the social sciences. T he scale and diversity of the phenom ena with which the social sciences dealt m ade them less successful in establishing scientific laws, bu t this was a difficulty of practice, not of principle: they too were concerned in the end with physical events. Even those who did not accept the thesis of physicalism agreed that there was no essential difference in aim or m ethod between the various branches of science. In the social sciences, no less than in the natural, an attem pt was m ade to form ulate hypotheses which could be tested by observation. Thus Schlick, who included ethics am ong the social sciences, denied that its results depended upon the use of any special faculty of m oral intuition. The questions which

| 22 ]

Editor’s Introduction

arise in ethics are, in his opinion, questions of fact; why people hold the principles th at they do, w hat it is that they desire, and how their desires can be fulfilled. In short, his general position is very sim ilar to th at of the Utilitarians. It has m uch the sam e m erits and m uch the same defects. The V ienna Circle as a whole was not very greatly interested in ethics; but it did not dispute Schlick’s view that if ethical statem ents were to be brought into the scientific fold, they m ust be handled in the way th at he proposed. The only question was w hether they belonged within the fold, w hether they were statem ents of fact at all. C arnap, for exam ple, m aintained th at they were not; he said th a t they were disguised im peratives. H e did not develop this suggestion, b u t it has since been given substance by R. M. H are in his book o n T he Language o f M orals (1 9 5 2 ). This im perative theory of ethics m ay be regarded as a version of the so-called Em otive T heory which, mainly through the work of English and A m erican philosophers, has come to be most closely associated with logical positivism . T he salient point is that ethical statem ents are not descriptive o f natural facts, still less of an alleged non-natural world of values: they are not descriptive of anything at all. The problem is then to determ ine how they do function. In C. L. Stevenson’s book E thics and L a n ­ guage (1 9 4 4 ), where the emotive theory was first w orked out in detail, it was argued that ethical statem ents served the dual purpose of expressing their author’s approval, o r disapproval, o f w hatever was in question and recom m ending others to share his attitude. H e laid particular em phasis upon the persuasive use of ethical term s. H is views have not passed w ithout criticism even from those who share his general standpoint; b u t the alternative accounts of ethics which these critics have p u t forw ard belong, as it were, to the sam e family. In discussions of logical positivism, this theory of ethics is apt to receive a disproportionate m easure of attention, considering th at it stands on the periphery of the system. O ne reason for this is th a t it has been thought, quite wrongly, that it was an onslaught upon morals. It has even been asserted, w ithout a shadow of em pirical evidence, that its advocates were corrupters of youth. In fact, the theory only explores the consequences of a sound and respectable point of logic which was already m ade by H um e; th at norm ative statem ents are not derivable from descriptive statem ents, or, as H um e puts it, that “ought” does not follow from “ is.” T o say th at m oral judgm ents are not fact-stating is not to say that they are unim portant, or even th at there cannot be argum ents in their favor. B ut these argum ents do not work in the way th at logical o r scientific argum ents do. It is not as if the intuitionists had discovered grounds for m oral

Editor’s Introduction

[ 23 ]

judgm ents which the em otivists tried to take away. O n the contrary, as M r. Straw son shows in his paper on “Ethical Intuitionism ,” the intuitionists themselves do not supply any foundation for moral judgm ents. It is therefore only on personal grounds that they can be entitled to p u t themselves forw ard as the guardians of virtue. V . P h il o s o p h ic a l A na ly sis

Some of the dissatisfaction th at is aroused by the emotive theory of ethics, and indeed by logical positivism in general, may be due to the fact th at people are still inclined to look to philosophy for guid­ ance as to the way they ought to live. W hen this function is denied to it, and when it is denied even the possibility of penetrating the veil o f appearance and exploring the hidden depths of reality, they feel th a t it is being trivialized. If this tim e-honored program is nonsensical, w hat rem ains? A s Ram sey says, “ philosophy m ust be of some use, and we m ust take it seriously.” But w hat function do the positivists leave it to perform ? From the point of view of W ittgenstein’s Tractatus, its function would ap p ear to be purely negative, though not for that reason un ­ im portant. “T he right m ethod of philosophy,” said W ittgenstein, “would be this. T o say nothing except w hat can be said, i.e. the propositions of natural science, i.e. som ething that has nothing to do with philosophy: and then always, when someone wished to say som ething m etaphysical, to dem onstrate to him that he had given no m eaning to certain signs in his propositions. This m ethod would be unsatisfying to the other— he would not have the feeling that we were teaching him philosophy— b u t it would be the only strictly correct m ethod.” This rather depressing view of the philosopher’s duty was not strictly m aintained by W ittgenstein himself. T he Philo­ sophical Investigations contains a great deal more than a series of proofs th at people have failed to attach any meaning to certain signs in their propositions. Nevertheless it still gives the impression that to philosophize is to get into a m uddle, or to rescue oneself or others from one. Philosophy is “ a battle against the bewitchment of o ur intelligence by m eans of language.” “ W hat is your aim in philos­ ophy? T o show the fly the way out of the fly bottle.” All the same, it is m eritorious of the fly to be there. It is the critical intelligences that get themselves bewitched. The Tractatus left no room fo r philosophical propositions. The whole field of significant discourse was covered by form al statem ents on the one hand and em pirical statem ents on the other. There re­ m ained nothing for philosophy to be about. It was for this reason

j 24 ]

Editor's Introduction

that W ittgenstein, and also Schlick, m aintained th at philosophy was not a doctrine b u t an activity. T he result of philosophizing, said Schlick, would not be to accum ulate a stock of philosophical p ropo­ sitions, but to m ake other propositions clear. B ut to m ake propositions clear it m ust be possible to talk about them. As Russell points o u t in his introduction to the Tractatus, W ittgenstein appeared not to allow fo r this, o r to allow for it only to a limited extent. H e implied th at an attem pt to describe the structure of language, as opposed to exhibiting it in use, m ust result in non­ sense. But though this conclusion may have been form ally accepted by Schlick, it was in practice disregarded by the V ienna Circle. Thus, C arnap, in his Der Logische A u fb a u der W elt, explicitly set himself to describe the structure of language by devising w hat he called a “ Konstitution-System ,” in which the various types of linguistic ex­ pressions, o r concepts, were assigned their proper places in a d e­ ductive hierarchy. If he had been questioned about the status of his own propositions, I suppose th at he would have said th at they were analytic; consisting, as they did, of definitions and their logical con­ sequences, they would belong to the realm of form al truths. H ow ever this may be, he certainly believed that these propositions were sig­ nificant; and he carried the V ienna Circle with him in holding that they were the sort of propositions th at a philosopher should be ex­ pected to put forward. The attem pt to bring philosophy within the dom ain of logic was carried further by C arnap in his book on the Logical Syntax o f Language. “ Philosophy,” he says in the foreword to this book, “ is to be replaced by the logic of science— that is to say by the logical analysis of the concepts and sentences of the sciences, for the logic of science is nothing other than the logical syntax of the language of science.” Though he speaks here of the language of science, he does not hold that there need be only one. A lternative languagesystems may be devised, and the choice between them is a m atter of convenience: this is an im portant departure from the position of W ittgenstein’s Tractatus. A ccording to C arnap, a language is char­ acterized by its form ation-rules, which specify w hat sequences of signs are to count as proper sentences of the language, and by its transform ation-rules, which lay down the conditions under which sentences are validly derivable from one another. It m ight b e thought that if the language was to have any em pirical application it m ust also contain m eaning-rules; rules which would correlate its expres­ sions with observable states of affairs: but C arnap, in this form alist stage of his philosophy, thought th at he could dispense with them. H e believed, quite m istakenly, th at statem ents of verbal equiva-

Editor’s Introduction

j 25 ]

lences could do the work not only of sem antic statem ents but even of ostensive definitions. It is in this book that C arnap m akes his fam ous distinction be­ tween the m aterial and the form al modes of speech. He distinguishes three kinds of sentences: “object-sentences,” such as “ 5 is a prim e num ber” o r “ Babylon was a big tow n,” “ pseudo-object sentences,” such as “Five is not a thing, but a num ber” “ Babylon was treated of in yesterday’s lecture,” and “ syntactical sentences” such as “ ‘Five’ is not a thing-w ord, but a num ber w ord” “T he word ‘Babylon’ occurred in yesterday’s lecture.” T he pseudo-object sentences are said to be “quasi-syntactical,” because they are syntactical sentences m asquerading as object-sentences. They are “quasi-syntactical sen­ tences o f the m aterial m ode of speech.” T ranslation from the material into the form al m ode replaces them by their syntactical equivalents. T o put it less technically, when one speaks in the form al mode one is overtly speaking about words; when one speaks in the material mode one is speaking about words while seeming to speak about things. This distinction does not of course apply to object-sentences. C arnap was not m aintaining, as some critics have supposed, that all discourse is about words. W hat he did appear to overlook, however, was the existence of a further category, that of pseudo-syntactical sentences; sentences which were about things but seemed to be about words. A s a result, he was apt to fall into the error of treating these sentences as if they were syntactical. It is with the opposite error th at he reproached most other philos­ ophers. H e m aintained that philosophical statem ents were syntactical, but that they had been treated as if they were object-statem enls, be­ cause of the fashion for expressing them in the m aterial m ode of speech. T hus, to take a selection of his examples, he argued that “T he world is the totality of facts, not of things,” the first proposition of W ittgenstein’s Tractatus, was equivalent to “Science is a system of sentences, not of nam es” : “ This circum stance is logically neces­ sary; . . . logically impossible; . . . logically possible” became “This sentence is analytic; . . . contradictory; . . . not contradictory” : K ronecker’s epigram “G od created the natural num bers; everything else in m athem atics is the work of m an” was a way of saying “The natural-num ber symbols are primitive symbols; other numerical expressions are introduced by definition.” “The only primitive data are relations between experiences” was equivalent to “ Only two-or m ore-term ed predicates whose argum ents belong to the genus oi experience-expressions occur as descriptive primitive symbols” : “Time is infinite in both directions” to “ Every positive o r negative realnum ber expression can be used as a tim e-co-ordinate.” Even the

[ 26 ]

Editor's Introduction

question of determ inism was said to “concern a syntactical difference in the system of natural laws.” In this way rival philosophical theses, if they m ade any sense at all, were represented as alternative p ro ­ posals about the way one’s language should be form ed. T hey were not true o r false, but only m ore o r less convenient. I think th at C arnap’s distinction betw een the m aterial and form al modes was fruitful, in th at it called attention to the fact th a t m any philosophical statem ents are disguised statem ents about language. W here he w ent wrong for the m ost p art was in supposing th a t they were syntactical. F o r w hat they are concerned with is not the form or order of words, but their use. T his does not com e out in C arn ap ’s examples because he illicitly smuggles semantics into syntax. T hus, “experience-expressions” is not a syntactical term . W hat m akes an expression an “experience-expression” is not its having any particular form but its being used to refer to an experience. B ut th en the question what is to count as an experience becom es im portant. N either is it to be settled by an arbitrary decision. In his m ore recent works, C arnap has recognized the legitimacy of semantics, and indeed devoted considerable attention both to the developm ent of sem antic theory and to building up sem antic sys­ tems. A n interesting effect of this has been a m arked relaxation of his philosophical austerity. H aving acquired the right to speak of the reference of words to things, he has allowed alm ost any type of w ord to denote its special sort of object, thus recreating the baroque universe which Russell had labored to depopulate. H is defense of this apparent extravagance is to be found in his p ap er on “ E m ­ piricism, Semantics and O ntology,” where he distinguishes betw een “internal” questions which arise within a given conceptual fram e­ work and “external” questions which concern the status and legiti­ macy of the fram ew ork itself. H e him self has always been chiefly interested in the external questions: he has thought it his business as a philosopher to devise linguistic systems and elaborate concepts that will be useful to the scientist. A nd no one should deny th a t this is a serious and legitimate activity. W here he is wrong, I think, is in assuming that the external questions present no serious problem : that nothing m ore is at issue than a choice of linguistic form s. It is this disregard of questions about the status of his linguistic fram ew orks that separates C arnap from the A m erican philosophers, like Quine and G oodm an, who resem ble him in their system atic approach to philosophy and in their preference for form al techniques. These philosophers are interested in w hat they call ontology, that is, in the question how far one’s choice of language com m its one to saying th at certain things exist. “T o b e,” says Q uine, “is to be

Editor's Introduction

[ 27 ]

the value of a variable” : and this m eans that the extent of what Russell called the “furniture” of the world depends upon the range of predicates that are needed to describe it. Both Quine and G ood­ m an wish this furniture to be as hard and spare as possible. They “renounce abstract entities” not just because they wish to exercise th eir logical ingenuity in seeing how well they can do w ithout them, b u t because they cannot bring themselves to believe that they exist. In the sam e spirit, G oodm an forgoes m aking any use of the notion of possible, as opposed to actual, things, o r of the distinction be­ tw een causal and accidental connections, o r of that between analytic and synthetic statem ents. “Y ou m ay,” he says, “decry some of these scruples and protest th at there are m ore things in heaven and earth than are dream t of in my philosophy. I am concerned, rather, that there should n o t be m ore things dream t of in my philosophy than there are in heaven o r earth .” It is not clear, however, either in his case o r in Q uine’s, on w hat this dem and for stringent economy is based. Q uine, indeed, allows in the end that the question of what there is m ust be settled on pragm atic grounds. A nd so he rejoins C arn ap ; b u t his pragm atism is m uch less serene. A n interest in categories, which is another way of approaching the problem of w hat there is, is characteristic also of the British philosophers who have been influenced by the later work of W itt­ genstein. But, for the m ost part, they are concerned not so much with trying to elim inate certain types of entity, o r to “ reduce” one to another, as with bringing out the resem blances and differences in the functioning of the statem ents which ostensibly refer to them. A technique which W ittgenstein himself uses for this purpose is that o f devising w hat he calls language games. T he idea is that by studying distorted o r simplified models of o ur actual language we can obtain a clearer insight into the way it really works. This is one way of protecting us against the error, into which we so easily fall, of as­ sum ing th at som ething m ust be the case, instead of looking and seeing w hat actually is the case. “W here o u r language suggests a body and there is none, there, we should like to say, is a spirit.” But this is to forsake description for bogus explanation. Very often the mental processes which we are led to postulate just do not occur. F o r in­ stance, “ it is no m ore essential to the understanding of a proposition th at we should imagine anything in connection with it than that we should m ake a sketch from it.” Such rem arks foreshadow Ryle’s attack upon the myth o f “ the ghost in the m achine.” And much as W ittgenstein disliked C arn ap ’s m ethods, there is an echo of physi­ calism in his dictum that “ an ‘inner process’ stands in need of out­ w ard criteria.”

[ 28 ]

Editor s Introduction

I suppose th at W ittgenstein is mainly responsible for the prevalent interest in the question how words are ordinarily used, though account has also to be taken of the influence of G. E . M oore. It does n o t seem to m e, however, th at M oore has ever been so greatly con­ cerned with ordinary usage as such. H e has been concerned w ith upholding the “com m onsense view” of the w orld and with analyzing the propositions which exem plify it; b u t he has not insisted th a t we lim it ourselves to ordinary usage in carrying out this analysis. W hen he does appeal to ordinary usage it is m ainly as a w eapon for dealing with other philosophers. H e shows th a t if th eir w ords are taken literally, they are using them to m ake statem ents w hich are manifestly false. It rem ains possible th a t they are saying som ething quite different from w hat they would appear to be saying, b u t then the discovery of their m eaning presents a problem . If they are not using w ords in any ordinary sense, the sense in which they are using them has to be m ade clear. T o my m ind, the m ain achievem ent of the “ordinary-language school” has been their exam ination and dissection o f the “unscien­ tific” uses of language. A good exam ple of this is J. L . A ustin’s description of w hat he calls perform ative statem ents: statem ents like “ I know . . .” o r “ I prom ise . . .” of which the purpose is n o t to assert a fact but to com m it the speaker in certain ways o r to offer some sort of guarantee. T o w hat im aginative lengths this greater flexibility in the approach to language can lead is indicated by D r. W aism ann’s paper which concludes this volum e. It shows th a t the current conception of philosophical analysis has spread fa r beyond R am sey’s idea of philosophy as sim ply issuing in definitions. But R am sey was right in saying th at it is “ all p art o f the vital w ork of clarifying and organizing o u r thought.” VI. In com piling this anthology I have tried to illustrate the historical developm ent of logical positivism, the range of its interests an d the m ain points of controversy. L ack of space has obliged m e to pass over m any pieces th at I should have liked to include. In particular, I am sorry to have had no room for Q uine’s p ap er on “T ru th by C onvention,” in which the positivists’ account o f a priori statem ents is effectively criticized, o r for C arn ap ’s influential articles on “T esta­ bility and M eaning.” It is especially to be regretted th a t the volume contains nothing of W ittgenstein. B u t neither the Tractatus LogicoPhilosophicus n o r the Philosophical Investigations, for all th eir epi­ sodic character, is a w ork to which one can do justice by selecting passages. They have to be read as a whole.

Logical Atomism

1

Logical Atomism by

B E R T R A N D RU SSELL

which I advocate is generally regarded as a species of realism , and accused of inconsistency because of the elements in it which seem contrary to that doctrine. F o r my part, I do not regard the issue between realists and their opponents as a funda­ m ental one; I could alter my view on this issue w ithout changing my m ind as to any of the doctrines upon which I wish to lay stress. I hold th at logic is w hat is fundam ental in philosophy, and that schools should be characterized rather by their logic than by their m etaphysic. M y own logic is atom ic, and it is this aspect upon which I should wish to lay stress. Therefore I prefer to describe my philosophy as “logical atom ism ,” rather than as “realism ,” w hether with o r w ithout some prefixed adjective. A few words as to historical developm ent may be useful by way o f preface. I cam e to philosophy through m athematics, o r rather through the wish to find some reason to believe in the truth of m athem atics. F rom early youth, I had an ardent desire to believe that there can be such a thing as knowledge, com bined with a great difficulty in accepting m uch th at passes as knowledge. It seemed clear th at the best chance of finding indubitable truth would be in pure m athem atics, yet some of Euclid’s axioms were obviously doubt­ ful, and the infinitesimal calculus, as I was taught it, was a mass of sophism s, which I could not bring myself to regard as anything else. I saw no reason to doubt the truth of arithm etic, but I did not then know th at arithm etic can be m ade to em brace all traditional pure m athem atics. A t the age of eighteen I read M ill’s Logic, but was profoundly dissatisfied with his reasons for accepting arithmetic and geom etry. I had not read H um e, but it seemed to me that pure em piricism (w hich I was disposed to accept) m ust lead to scepT h e p h il o s o p h y

T bis essay was R ussell’s contribution to C ontem porary British Philosophy, first series (ed. J. H . M u irh e a d ), a book published in 1924. It is here reprinted by the kind perm ission o f the a u th o r an d G eorge A llen and Unwin Ltd., London.

[31]

[ 32 ]

BERTRAND RU SSELL

ticism rather than to M ill’s support of received scientific doctrines. A t Cam bridge I read K ant and Hegel, as well as M r. B radley’s Logic, which influenced me profoundly. F o r some years I was a disciple of M r. Bradley, but about 1898 I changed my views, largely as a result of argum ents with G . E. M oore. I could no longer believe that knowing m akes any difference to w hat is known. Also I found myself driven to pluralism. Analysis of m athem atical propositions persuaded me th at they could not be explained as even partial truths unless one adm itted pluralism and the reality of relations. A n accident led me at this time to study Leibniz, and I cam e to the conclusion (subsequently confirm ed by C outurat’s m asterly re­ searches) that many of his m ost characteristic opinions w ere due to the purely logical doctrine th at every proposition has a subject and a predicate. This doctrine is one which Leibniz shares with Spinoza, Hegel, and M r. Bradley; it seemed to me th at, if it is rejected, the whole foundation fo r the m etaphysics o f all these philosophers is shattered. I therefore returned to the problem which had originally led me to philosophy, nam ely, the foundations of m athem atics, applying to it a new logic derived largely from Peano and Frege, which proved (a t least, so I believe) far m ore fruitful than that of traditional philosophy. In the first place, I found that m any of the stock philosophical argum ents about m athem atics (derived in the m ain from K ant) had been rendered invalid by the progress of m athem atics in the meanwhile. N on-Euclidean geom etry had underm ined the argum ent o f the transcendental aesthetic. W eierstrass had shown th a t the differential and integral calculus do not require the conception of the infinitesimal, and that, therefore, all th at had been said by philosophers on such subjects as the continuity of space and tim e and m otion m ust be regarded as sheer error. C antor freed the conception of infinite num ber from contradiction, and thus disposed of K ant’s antinom ies as well as m any of Hegel’s. Finally Frege showed in detail how arithm etic can be deduced from pure logic, w ithout the need of any fresh ideas o r axioms, thus disproving K ant’s assertion th at “ 7 -j- 5 = 12” is synthetic— at least in the obvious interpretation of that dictum . As all these results w ere obtained, not by any heroic m ethod, bu t by patient detailed reasoning, I began to think it probable th at philosophy had erred in adopting heroic remedies for intellectual difficulties, and th at solutions w ere to be found merely by greater care and accuracy. T his view I have come to hold more and m ore strongly as tim e w ent on, and it has led me to doubt w hether philosophy, as a study distinct from science and

Logical Atom ism

[ 33 ]

possessed of a m ethod of its own, is anything more than an un­ fortunate legacy from theology. F rege’s w ork was not final, in the first place because it applied only to arithm etic, not to other branches of m athem atics; in the second place because his prem ises did not exclude certain contra­ dictions to which all past systems of form al logic turned out to be liable. Dr. W hitehead and I in collaboration tried to remedy these two defects, in Principia M athem atica, which, however, still falls short of finality in some fundam ental points (notably the axiom of reducibility). But in spite of its shortcom ings I think that no one who reads this book will dispute its main contention, namely, that from certain ideas and axioms of formal logic, by the help of the logic of relations, all pure m athem atics can be deduced, with­ out any new undefined idea o r unproved propositions. The technical m ethods of m athem atical logic, as developed in this book, seem to me very powerful, and capable of providing a new instrum ent for the discussion of m any problem s that have hitherto rem ained sub­ ject to philosophic vagueness. Dr. W hitehead’s Concept of Nature and Principles o f N atural Know ledge may serve as an illustration of w hat I mean. W hen pure m athem atics is organized as a deductive system— i.e. as the set o f all those propositions that can be deduced from an assigned set of prem ises— it becomes obvious that, if we are to believe in the truth of pure m athem atics, it cannot be solely because we believe in the truth of the set of premises. Some of the premises are m uch less obvious than some of their consequences, and are believed chiefly because of their consequences. This will be found to be always the case when a science is arranged as a deductive system. It is not the logically simplest propositions of the system that are the m ost obvious, o r th at provide the chief part of our reasons for believing in the system. W ith the em pirical sciences this is evident. Electro-dynam ics, for exam ple, can be concentrated into M axwell’s equations, but these equations are believed because "of the observed truth of certain of their logical consequences. Exactly the sam e thing happens in the pure realm of logic; the logically first principles of logic— at least some of them — are to be believed, not on their own account, but on account of their consequences. The epistem ological question: “ Why should I believe this set of propositions?” is quite different from the logical question: “W hat is the smallest and logically simplest group of propositions from which this set of propositions can be deduced?” O ur reasons for believing logic and pure m athem atics are, in part, only inductive and

[ 34 ]

BERTRAND R U SSE L L

probable, in spite of the fact that, in their logical order, the p ropo­ sitions of logic and pure m athem atics follow from the prem ises of logic by pure deduction. I think this point im portant, since errors are liable to arise from assim ilating the logical to the epistem ological order, and also, conversely, from assim ilating the epistem ological to the logical order. T he only way in which w ork on m athem atical logic throw s light on the truth o r falsehood of m athem atics is by disproving the supposed antinom ies. This shows th at m athem atics m ay be true. But to show th at m athem atics is true w ould require other m ethods and other considerations. One very im portant heuristic maxim which D r. W hitehead and I found, by experience, to be applicable in m athem atical logic, and have since applied in various other fields, is a form o f O ckham ’s razor. W hen some set of supposed entities has neat logical p ro p er­ ties, it turns out, in a great m any instances, th at the supposed entities can be replaced by purely logical structures com posed o f entities which have not such neat properties. In th at case, in interpreting a body of propositions hitherto believed to be about the supposed entities, we can substitute the logical structures w ithout altering any of the detail of the body o f propositions in question. T his is an economy, because entities with neat logical properties are always inferred, and if the propositions in which they occur can be inter­ preted without m aking this inference, the ground for the inference fails, and o u r body of propositions is secured against the need of a doubtful step. T he principle may be stated in the form : “ W herever possible, substitute constructions o u t of known entities for inferences to unknow n entities.” T he uses of this principle are very various, but are not intelli­ gible in detail to those who do not know m athem atical logic. T he first instance I cam e across was w hat I have called “ the principle of abstraction,” o r “the principle which dispenses with abstraction.” 1 This principle is applicable in the case of any sym m etrical and transitive relation, such as equality. We are apt to infer th at such relations arise from possession of some com m on quality. T his may or may not be true; probably it is true in some cases and not in others. But all the form al purposes of a com m on quality can be served by m em bership of the group of term s having the said rela­ tion to a given term . T ake m agnitude, for exam ple. L et us suppose that we have a group of rods, all equally long. It is easy to suppose that there is a certain quality, called their length, which they all share. But all propositions in which this supposed quality occurs will 1. O ur K now ledge o f the E xternal W orld, p. 42.

Logical Atom ism

[ 35 ]

retain their truth-value unchanged if, instead of “length of the rod we take “m em bership of the group of all those rods which are as long as at.” In various special cases— e.g. the definition of real num ­ bers— a sim pler construction is possible. A very im portant exam ple of the principle is Frege’s definition of the cardinal num ber of a given set of term s as the class of all sets th at are “sim ilar” to the given set— where two sets are “sim ilar” when there is a one-one relation whose dom ain is the one set and whose converse dom ain is the other. Thus a cardinal num ber is the class of all those classes which are similar to a given class. This definition leaves unchanged the truth-values of all propositions in which cardinal num bers occur, and avoids the inference to a set of entities called “cardinal num bers,” which were never needed except for the purpose of m aking arithm etic intelligible, and are now no longer needed for th at purpose. Perhaps even m ore im portant is the fact that classes themselves can be dispensed with by sim ilar m ethods. M athem atics is full of propositions which seem to require that a class o r an aggregate should be in some sense a single entity— e.g. the proposition “the num ber of com binations of n things any num ber at a time is 2n.” Since 2" is always greater than n, this proposition leads to difficulties if classes are adm itted because the num ber of classes of entities in the universe is greater than the num ber of entities in the universe, which would be odd if classes were some among entities. Fortunately all the propositions in which classes appear to be m entioned can be interpreted w ithout supposing that there are classes. This is perhaps the m ost im portant of all the applications of our principle. (See Principia M athem atica, *20.) A nother im portant exam ple concerns what I call “definite descriptions,” i.e. such phrases as “ the even prim e,” “the present King of England,” “the present King of F rance.” There has always been a difficulty in interpreting such propositions as “the present King of France does not exist.” The difficulty arose through suppos­ ing th at “ the present King of F rance” is the subject of this proposi­ tion, which m ade it necessary to suppose that he subsists although he does not exist. But it is difficult to attribute even subsistence to “the round square” o r “ the even prim e greater than 2.” In fact, “ the round square does not subsist” is just as true as “the present King of F rance does not exist.” Thus the distinction between existence and subsistence does not help us. The fact is that, when the words “ the so-and-so” occur in a proposition, there is no corresponding single constituent of the proposition, and when the proposition is

j 36 j

BERTRAND RU SSELL

fully analyzed the words “the so-and-so” have disappeared. A n im portant consequence of the theory of descriptions is th at it is meaningless to say “A exists” unless “ A ” is (o r stands fo r) a phrase of the form “the so-and-so.” If the so-and-so exists, and x is the so-and-so, to say “jc exists” is nonsense. Existence, in the sense in which it is ascribed to single entities, is thus rem oved altogether from the list of fundam entals. T he ontological argum ent and m ost of its refutations are found to depend upon bad gram m ar. (See Principia M athematica, *14.) There are m any other exam ples of the substitution of construc­ tions for inferences in pure m athem atics, fo r exam ple, series, ordinal num bers, and real num bers. But I will pass on to the exam ples in physics. Points and instants are obvious exam ples: D r. W hitehead has shown how to construct them out of sets of events all of which have a finite extent and a finite duration. In relativity theory, it is not points or instants that we prim arily need, but event-particles, which correspond to what, in older language, m ight be described as a point at an instant, o r an instantaneous point. (In form er days, a point of space endured throughout all time, and an instant of time pervaded all space. Now the unit th at m athem atical physics wants bas neither spatial n or tem poral extension.) Event-particles are constructed by just the same logical process by which points and instants were constructed. In such constructions, however, we are on a different plane from that of constructions in pure m athe­ matics. T he possibility of constructing an event-particle depends upon the existence of sets of events with certain properties; w hether the required events exist can only be known em pirically, if at all. There is therefore no a priori reason to expect continuity (in the m athem atical sense), o r to feel confident th a t event-particles can be constructed. If the quantum theory should seem to dem and a discrete space-time, o ur logic is just as ready to m eet its requirem ents as to meet those of traditional physics, which dem ands continuity. T he question is purely em pirical, and our logic is (as it ought to b e) equally adapted to either alternative. Similar considerations apply to a particle of m atter, o r to a piece of m atter of finite size. M atter, traditionally, has tw o of those “neat” properties which are the m ark of a logical construction; first, that two pieces of m atter cannot be at the same place at the sam e tim e; secondly, th at one piece of m atter cannot be in two places at the same time. Experience in the substitution of constructions fo r infer­ ences m akes one suspicious of anything so tidy and exact. O ne

Logical Atom ism

[ 37 ]

canno t help feeling th at im penetrability is not an em pirical fact, derived from observation of billiard-balls, but is something logically necessary. This feeling is wholly justified, but it could not be so if m atter were not a logical construction. A n immense num ber of occurrences coexist in any little region of space-tim e; when we are speaking of w hat is not logical construction, we find no such prop­ erty as im penetrability, but, on the contrary, endless overlapping of the events in a part of space-tim e, however small. The reason th at m atter is im penetrable is because our definitions make it so. Speaking roughly, and merely so as to give a notion of how this happens, we may say that a piece of m atter is all that happens in a certain track in space-tim e, and that we construct the tracks called bits o f m atter in such a way that they do not intersect. M atter is im penetrable because it is easier to state the laws of physics if we m ake o u r constructions so as to secure impenetrability. Im pene­ trability is a logically necessary result of definition, though the fact th at such a definition is convenient is empirical. Bits of m atter are not am ong the bricks out of which the world is built. T he bricks are events, and bits of m atter are portions of the structure to which we find it convenient to give separate attention. In the philosophy of mental occurrences there are also oppor­ tunities fo r the application of o ur principle of constructions versus inferences. T h e subject, and the relation of a cognition to what is know n, both have th at schem atic quality that arouses our suspicions. It is clear th at the subject, if it is to be preserved at all, must be preserved as a construction, not as an inferred entity; the only question is w hether the subject is sufficiently useful to be worth constructing. T he relation of a cognition to w hat is known, again, cannot be a straightforw ard single ultim ate, as I at one time believed it to be. A lthough I do not agree with pragm atism , I think William James was right in draw ing attention to the complexity of “ knowing.” It is impossible in a general sum m ary, such as the present, to set out the reasons fo r this view. But whoever has acquiesced in our prin­ ciple will agree that here is prim a facie a case for applying it. M ost o f my A nalysis o f M in d consists of applications of this prin­ ciple. B ut as psychology is scientifically much less perfected than physics, the opportunities fo r applying the principle are not so good. T he principle depends, for its use, upon the existence of some fairly reliable body of propositions, which are to be interpreted by the logician in such a way as to preserve their truth while minimizing the elem ent of inference to unobserved entities. T he principle there­ fore presupposes a m oderately advanced science, in the absence

j 3g j

BERTRAND RU SSELL

of which the logician does not know w hat he ought to construct. U ntil recently, it would have seemed necessary to construct geo­ m etrical points; now it is event-particles that are w anted. In view of such a change in an advanced subject like physics, it is clear th at constructions in psychology m ust be purely provisional. I have been speaking hitherto of w hat it is not necessary to assume as part of the ultim ate constituents of the world. B ut logical constructions, like all other constructions, require m aterials, and it is time to turn to the positive question, as to w hat these m aterials are to be. This question, however, requires as a prelim inary a dis­ cussion of logic and language and their relation to w hat they try to represent. T he influence of language on philosophy has, I believe, been profound and almost unrecognized. If we are not to be m isled by this influence, it is necessary to becom e conscious of it, and to ask ourselves deliberately how far it is legitimate. T he subjcctpredicate logic, with the substance-attribute m etaphysic, are a case in point. It is doubtful w hether either would have been invented by people speaking a non-A ryan language; certainly they do not seem to have arisen in China, except in connection with Buddhism , which brought Indian philosophy with it. Again, it is natural, to take a different kind of instance, to suppose th at a proper nam e which can be used significantly stands for a single entity; we suppose that there is a certain m ore or less persistent being called “Socrates,” because the same nam e is applied to a series of occurrences which we are led to regard as appearances of this one being. As language grows more abstract, a new set of entities com e into philosophy, namely, those represented by abstract words— the universais. I do not wish to m aintain th at there are no universais, but certainly there are m any abstract words which do not stand for single uni­ versais— e.g. triangularity and rationality. In these respects language misleads us both by its vocabulary and by its syntax. W e must be on o ur guard in both respects if o ur logic is not to lead to a false metaphysic. Syntax and vocabulary have had different kinds of effects on philosophy. V ocabulary has m ost influence on com m on sense. It might be urged, conversely, th at com m on sense produces our vocabulary. This is only partially true. A w ord is applied at first to things which are m ore o r less sim ilar, w ithout any reflection as to w hether they have any point of identity. B ut when once usage has fixed the objects to which the w ord is to be applied, com m on sense is influenced by the existence of the w ord, and tends to sup­

Logical Atomism

[ 39 ]

pose th at one w ord m ust stand for one object, which will be a universal in the case of an adjective o r an abstract word. Thus the influence of vocabulary is tow ards a kind of platonic pluralism of things and ideas. T h e influence of syntax, in the case of the Indo-E uropean lan­ guages, is quite different. A lm ost any proposition can be put into a form in which it has a subject and a predicate, united by a copula. It is natural to infer that every fact has a corresponding form, and consists in the possession of a quality by a substance. This leads, of course, to monism, since the fact th at there were several sub­ stances (if it were a fact) would not have the requisite form. Philosophers, as a rule, believe themselves free from this sort of influence of linguistic form s, bu t m ost of them seem to me to be m istaken in this belief. In thinking about abstract m atters, the fact th at the words for abstractions are no m ore abstract than ordin­ ary words always m akes it easier to think about the words than about w hat they stand for, and it is almost impossible to resist consistently the tem ptation to think about the words. T hose who do not succum b to the subject-predicate logic are ap t to get only one step further, and adm it relations of two terms,, such as before-and-after, greater-and-less, right-and-left. Language lends itself to this extension of the subject-predicate logic, since we say “A precedes B ” “A exceeds B ,” and so on. It is easy to prove th at the fact expressed by a proposition of this sort cannot consist of the possession of a quality by a substance, or of the possession o f two o r m ore qualities by two o r m ore substances. (See Principles o f M athem atics, § 2 1 4 .) The extension of the sub­ ject-predicate logic is therefore right so far as it goes, but obviously a further extension can be proved necessary by exactly similar argum ents. How far it is necessary to go up the series of threeterm , four-term , five-term . . . relations I do not know. But it is certainly necessary to go beyond tw o-term relations. In projective geom etry, fo r exam ple, the order of points on a line o r of planes through a line requires a four-term relation. A very unfortunate effect of the peculiarities of language is in connection with adjectives and relations. All words are of the same logical type; a w ord is a class of series, of noises o r shapes according as it is heard o r read. But the meanings of words are of various different types; an attribute (expressed by an adjective) is of a different type from the objects to which it can be (w hether truly o r falsely) attributed; a relation (expressed perhaps by a preposi­ tion, perhaps by a transitive verb, perhaps in some other way) is

[ 40 ]

BERTRAND RU SSELL

of a different type from the term s between which it holds o r does not hold. The definition of a logical type is as follows: A and B are of the same logical type if, and only if, given any fact of which A is a constituent, there is a corresponding fact which has B as a constituent, which cither results by substituting B for A , o r is the negation of w hat so results. T o take an illustration, Socrates and Aristotle are of the same type, because “ Socrates was a philos­ opher” and “Aristotle was a philosopher” are both facts; Socrates and Caligula are of the same type, because “Socrates was a philos­ opher” and “Caligula was not a philosopher” are both facts. T o love and to kill are of the same type, because “ Plato loved Socrates” and “ Plato did not kill Socrates” are both facts. It follows form ally from the definition that, when two words have m eanings of differ­ ent types, the relations of the words to w hat they m ean are of different types; that is to say, there is not one relation of meaning between words and what they stand for, but as many relations of meaning, each of a different logical type, as there are logical types among the objects for which there are words. This fact is a very potent source of error and confusion in philosophy. In particular, it has m ade it extraordinarily difficult to express in words any theory of relations which is logically capable of being true, because language cannot preserve the difference of type between a relation and its terms. M ost of the argum ents for and against the reality of relations have been vitiated through this source of confusion. A t this point, I propose to digress for a m om ent, and to say, as shortly as I can, what I believe about relations. My own views on the subject of relations in the past were less clear than I thought them , but were by no m eans the views which my critics supposed them to be. Owing to lack of clearness in my own thoughts, I was unable to convey my meaning. T he subject of relations is difficult, and I am far from claiming to be now clear about it. B ut I think certain points are clear to me. A t the tim e when I w rote The Principles o f M athem atics, I had no t yet seen the necessity of logical types. The doctrine of types profoundly affects logic, and I think shows what, exactly, is the valid elem ent in the argum ents of those who oppose “external” relations. But so far from strengthening their main position, the doctrine of types leads, on the contrary, to a more com plete and radical atom ism than any th at I conceived to be possible twenty years ago. T he question of relations is one of the m ost im portant that arise in philosophy, as most other issues turn on it: monism and pluralism ; the question w hether anything is wholly true except the whole of tru th , o r wholly real except the

Logical Atom ism

[41]

whole of reality; idealism and realism, in some of their forms; perhaps the very existence of philosophy as a subject distinct from science and possessing a m ethod of its own. It will serve to make my m eaning clear if I take a passage in M r. Bradley’s Essays on T ruth and Reality, not for controversial purposes, but because it raises exactly the issues th at ought to be raised. But first of all I will try to state my own view, w ithout argum ent.2 C ertain contradictions— of which the simplest and oldest is the one about Epim enides the Cretan, who said that all Cretans were liars, which may be reduced to the m an who says “ I am lying”— convinced me, after five years devoted mainly to this one question, th at no solution is technically possible w ithout the doctrine of types. In its technical form , this doctrine states merely that a word or symbol may form part of a significant proposition, and in this sense have meaning, w ithout being always able to be substituted for another word o r symbol in the same or some other proposition w ithout producing nonsense. Stated in this way, the doctrine may seem like a truism. “ Brutus killed C aesar” is significant, but “ Killed killed C aesar” is nonsense, so that we cannot replace “B rutus” by “ killed,” although both words have meaning. This is plain common sense, but unfortunately almost all philosophy consists in an attem pt to forget it. T he following words, for example, by their very nature, sin against it: attribute, relation, complex, fact, truth, falsehood, not, liar, omniscience. T o give a meaning to these words, we have to m ake a detour by way of words or symbols and the different ways in which they may m ean; and even then, we usually arrive, not at one meaning, but at an infinite series of different meanings. W ords, as we saw, are all of the same logical type; there­ fore when the meanings of two words are of different types, the relations of the two words to w hat they stand for are also of different types. A ttribute-w ords and relation-words are of the same type, therefore we can say significantly “attribute-w ords and relationwords have different uses.” But we cannot say significantly “ attributes are not relations.” By o ur definition of types, since relations are relations, the form of words “attributes are relations” must be not false, but meaningless, and the form of words “ attributes are not relations,” similarly, must be not true, but meaningless. Nevertheless, the statem ent “attribute-w ords are not relation-words” is significant and true. 2. I am m uch indebted to my friend W ittgenstein in this m atter. See his Tractatus Logico-Philosophicus, Kegan Paul, 1922. I do not accept all his doctrines, b u t my debt to him will be obvious to those who read his book.

[ 42 ]

BERTRAND RU SSE L L

We can now tackle the question of internal and external rela­ tions, rem em bering that the usual form ulations, on both sides, are inconsistent with the doctrine of types. I will begin with attem pts to state the doctrine of external relations. It is useless to say “ terms are independent of their relations,” because “inde­ pendent” is a word which m eans nothing. Tw o events m ay be said to be causally independent when no causal chain leads from one to the other; this happens, in the special theory of relativity, when the separation between the events is space-like. Obviously this sense of “independent” is irrelevant. If, when we say “term s are independent of their relations,” we m ean “ two term s which have a given relation would be the sam e if they did not have it,” th at is obviously false; for, being w hat they are, they have the relation, and therefore whatever does not have the relation is different. If we m ean— as opponents of external relations suppose us to m ean — th at the relation is a third term which comes betw een the other two term s and is somehow hooked o n to them , th at is obviously absurd, for in th at case the relation has ceased to be a relation, and all that is truly relational is the hooking of the relation to the term s. T he conception of the relation as a third term betw een the other two sins against the doctrine of types, and m ust be avoided with the utm ost care. W hat, then, can we m ean by the doctrine of external relations? Prim arily this, th at a relational proposition is not, in general, logically equivalent form ally to one o r m ore subject-predicate propositions. Stated m ore precisely: Given a relational propositional function “x R y ,” it is not in general the case that we can find predicates «, /?, y, such that, for all values of x and y, x R y is equivalent to x a t y($y ( x , y )y (w here (x , y ) stands for the whole consisting of x and y ) , o r to any one o r two of these. This, and this only, is w hat I m ean to affirm when I assert the doctrine of external relations; and this, clearly, is at least part of w hat M r. Bradley denies when he asserts the doctrine of internal relations. In place of “ unities” o r “com plexes,” I prefer to speak of “facts.” It must be understood th at the w ord “fact” cannot occur signifi­ cantly in any position in a sentence where the w ord “sim ple” can occur significantly, n or can a fact occur where a simple can occur. We must not say “ facts are not simples.” We can say, “ T he symbol for a fact must not replace the symbol for a simple, o r vice versa, if significance is to be preserved.” But it should be observed th at, in this sentence, the w ord “ for” has different m eanings on the two occasions of its use. If we are to have a language which is to safe­

Logical Atom ism

[ 43 ]

guard us from errors as to types, the symbol for a fact m ust be a proposition, not a single w ord o r letter. Facts can be asserted or denied, bu t cannot be nam ed. (W hen I say “facts cannot be nam ed,” this is, strictly speaking, nonsense. W hat can be said without falling into nonsense is: “T he symbol for a fact is not a nam e.” ) This illustrates how m eaning is a different relation for different types. T he way to m ean a fact is to assert it; the way to m ean a simple is to nam e it. Obviously nam ing is different from asserting, and sim ilar differences exist where m ore advanced types are concerned, though language has no m eans of expressing the differences. T here are m any other m atters in M r. Bradley’s exam ination of my views which call for reply. But as my present purpose is explan­ atory rath er than controversial, I will pass them by, having, I hope, already said enough on the question of relations and complexes to m ake it clear w hat is the theory that I advocate. I will only add, as regards the doctrine of types, that m ost philosophers assume it now and then, and few would deny it, but that all (so far as I know ) avoid form ulating it precisely o r drawing from it those deductions th at are inconvenient for their systems. I com e now to some of M r. Bradley’s criticisms ( loc. cit., p. 2 8 0 ff.). H e says: “ M r. Russell’s main position has remained to myself incomprehen­ sible. On the one side I am led to think that he defends a strict pluralism, fo r which nothing is admissible beyond simple terms and external rela­ tions. On the other side M r. Russell seems to assert emphatically, and to use throughout, ideas which such a pluralism surely must repudiate. H e throughout stands upon unities which are complex and which cannot be analysed into term s and relations. These two positions to my mind are irreconcilable, since the second, as I understand it, contradicts the first flatly.”

W ith regard to external relations, my view is the one I have just stated, not the one com m only im puted by those who disagree. But with regard to unities, the question is m ore difficult. The topic is one with which language, by its very nature, is peculiarly unfitted to deal. I m ust beg the reader, therefore, to be indulgent if what I say is not exactly w hat I m ean, and to try to see what I mean in spite of unavoidable linguistic obstacles to clear expression. T o begin with, I do not believe th at there are complexes or unities in the same sense in which there are simples. I did believe this when I wrote T he Principles o f M athem atics, but, on account o f the doctrine of types, I have since abandoned this view. To

j 44 ]

BERTRAND RU SSE L L

speak loosely, I regard simples and com plexes as always of differ­ ent types. T hat is to say, the statem ents “There are simples” and “There are com plexes” use the words “ there are” in different senses. But if I use the w ords “ there are” in the sense which they have in the statem ent “there are sim ples,” then the form of w ords “ there are not com plexes” is neither true nor false, but meaningless. This shows how difficult it is to say clearly, in ordinary language, w hat I w ant to say about com plexes. In the language of m athem atical logic it is m uch easier to say w hat I w ant to say, but m uch h ard er to induce people to understand w hat I m ean when I say it. W hen I speak of “sim ples” I ought to explain th at I am speaking of something not experienced as such, but know n only inferentially as the limit of analysis. It is quite possible that, by greater logical skill, the need for assum ing them could be avoided. A logical lan­ guage will not lead to e rro r if its simple symbols (i.e. those not having any parts th at are symbols, o r any significant stru ctu re) all stand fo r objects of some one type, even if these objects are not simple. T he only draw back to such a language is that it is incapable of dealing with anything sim pler than the objects which it rep re­ sents by simple symbols. B ut I confess it seems obvious to m e (as it did to Leibniz) that w hat is com plex m ust be com posed of simples, though the num ber of constituents may be infinite. It is also obvious th at the logical uses of the old notion of substance (i.e. those uses which do not imply tem poral d u ratio n ) can only be applied, if at all, to simples; objects of oth er types do not have that kind of being which one associates with substances. T he essence of a substance, from the symbolic point of view, is th a t it can only be nam ed— in old-fashioned language, it never occurs in a proposition except as the subject o r as one of the term s of a rela­ tion. If w hat we take to be simple is really com plex, we m ay get into trouble by nam ing it, when w hat we ought to do is to assert it. F o r exam ple, if Plato loves Socrates, there is not an entity “Plato’s love for Socrates,” bu t only the fact th at P lato loves Socrates. A nd in speaking of this as “ a fact,” we are already m aking it m ore substantial and m ore of a unity than we have any right to do. A ttributes and relations, though they may be not susceptible of analysis, differ from substances by the fact th a t they suggest a structure, and th at there can be no significant symbol which symbolizes them in isolation. All propositions in which an attribute o r a relation seem s to be the subject are only significant if they can be brought into a form in which the attribute is attributed o r the relation relates. If this were not the case, there would be significant

Logical Atom ism

[ 45 ]

propositions in which an attribute o r a relation would occupy a position appropriate to a substance, which would be contrary to th e doctrine of types, and would produce contradictions. Thus the p ro p er symbol for “yellow” (assum ing for the sake of illustration th a t this is an attrib u te) is n o t the single w ord “yellow,” but the propositional function is yellow,” where the structure of the sym bol shows the position which the word “yellow” must have if it is to be significant. Similarly the relation “precedes” m ust not be represented by this one word, bu t by the symbol "x precedes y,” showing the way in which the symbol can occur significantly. (It is here assum ed th at values are not assigned to x and y when we are speaking of the attribute o r relation itself.) T h e symbol for the simplest possible kind of fact will still be of the form "x is yellow” o r “x precedes y** only that “x ” and "y" will be no longer undeterm ined variables, b u t names. In addition to the fact th at we do not experience simples as such, there is another obstacle to the actual creation of a correct logical language such as I have been trying to describe. This obstacle is vagueness. A ll o ur words are m ore o r less infected with vagueness, by which I m ean th at it is not always clear whether they apply to a given object o r not. I t is of the nature of words to be more o r less general, and not to apply only to a single particular, but that w ould not m ake them vague if the particulars to which they applied w ere a definite set. But this is never the case in practice. T he defect, however, is one which it is easy to imagine rem oved, however difficult it m ay be to rem ove it in fact. T he purpose of the foregoing discussion of an ideal logical lan­ guage (w hich would of course be wholly useless for daily life) is tw ofold: first, to prevent inferences from the nature of language to the nature o f the w orld, which are fallacious because they depend upon the logical defects of language; secondly, to suggest, by inquiring w hat logic requires of a language which is to avoid contradiction, w hat sort of a structure we may reasonably suppose the w orld to have. If I am right, there is nothing in logic that can help us to decide betw een m onism and pluralism , o r between the view th a t there are ultim ate relational facts and the view that there are none. M y own decision in favor of pluralism and relations is taken on em pirical grounds, after convincing myself that the a priori argum ents to the contrary are invalid. But I do not think these argum ents can be adequately refuted w ithout a thorough treatm ent of logical types, of which the above is a mere sketch. This brings m e, however, to a question of m ethod which I

[ 46 ]

BERTRAND RU SSE L L

believe to be very im portant. W hat are we to take as d ata in philos­ ophy? W hat shall we regard as having the greatest likelihood of being true, and w hat as proper to be rejected if it conflicts with other evidence? It seems to me th at science has a m uch greater likelihood of being tru e in the m ain than any philosophy hitherto advanced (I d o not, of course, except my o w n ). In science there are m any m atters about which people are agreed; in philosophy there are none. Therefore, although each proposition in a science may be false, and it is practically certain th a t there are som e th at are false, yet we shall be wise to build o ur philosophy upon science, because the risk of error in philosophy is pretty sure to be greater than in science. If we could hope fo r certainty in philosophy the m atter would be otherwise, b u t so far as I can see such a hope would be chimerical. O f course those philosophers whose theories, prim a facie, run counter to science always profess to be able to interpret science so th at it shall rem ain true on its ow n level, with th at m inor degree of truth which ought to content the hum ble scientist. T hose who m aintain a position of this sort are bound— so it seems to me— to show in detail how the interpretation is to be effected. In m any cases, I believe th at this w ould be quite impossible. I do not believe, for instance, th at those who disbelieve in the reality of relations (in some such sense as that explained above) can possibly interpret those num erous parts of science which em ploy asym m etrical relations. Even if I could see no way of answering the objections to relations raised (fo r exam ple) by M r. Bradley, I should still think it m ore likely than not th at some answ er was possible, because I should think an error in a very subtle and abstract argum ent m ore probable than so fundam ental a falsehood in science. A dm itting th a t every­ thing we believe ourselves to know is doubtful, it seems, nevertheless, th at what we believe ourselves to know in philosophy is m ore d oubt­ ful than the detail of science, though perhaps not m ore doubtful th an its m ost sweeping generalizations. T he question of interpretation is of im portance for alm ost every philosophy, and I am not at all inclined to deny th at m any scientific results require interpretation before they can be fitted into a co­ herent philosophy. The maxim of “constructions versus inferences” is itself a maxim of interpretation. But I think that any valid kind of interpretation ought to leave the detail unchanged, though it may give a new meaning to fundam ental ideas. In practice, this m eans th at structure m ust be preserved. A nd a test o f this is th a t all the propositions of a science should rem ain, though new m ean­

Logical Atom ism

[ 47 ]

ings m ay be found fo r their term s. A case in point, on a nonphilosophical level, is the relation of the physical theory of light to our perceptions of color. This provides different physical occur­ rences corresponding to different seen colors, and thus makes the structure of the physical spectrum the same as that of what we see w hen we look at a rainbow. Unless structure is preserved, we cannot validly speak of an interpretation. A nd structure is just w hat is destroyed by a m onistic logic. I do not m ean, of course, to suggest that, in any region of science, the structure revealed at present by observation is exactly that which actually exists. O n the contrary, it is in the highest degree probable that the actual structure is m ore fine-grained than the observed structure. This applies just as m uch to psychological as to physical m aterial. It rests upon the fact that, where we perceive a difference (e.g. between two shades of co lo r), there is a differ­ ence, but where we do not perceive a difference it does not follow th at there is not a difference. We have therefore a right, in all inter­ pretation, to dem and the preservation of observed differences, and the provision o f room fo r hitherto unobserved differences, although we cannot say in advance w hat they will be, except when they can be inferentially connected with observed differences. In science, structure is the m ain study. A large p art of the im­ portance of relativity comes from the fact that it has substituted a single four-dim ensional m anifold (space-tim e) for the two m ani­ folds, three-dim ensional space and one-dimensional time. This is a change of structure, and therefore has far-reaching consequences, but any change which does not involve a change of structure does not m ake m uch difference. The m athem atical definition and study of structure (u n d er the nam e of “relation-num bers” ) form P art IV of Principia M athem atica. T he business of philosophy, as I conceive it, is essentially that o f logical analysis, followed by logical synthesis. Philosophy is more concerned than any special science with relations of different sciences and possible conflicts between them ; in particular, it cannot acquiesce in a conflict between physics and psychology, o r between psychology and logic. Philosophy should be com prehensive, and should be bold in suggesting hypotheses as to the universe which science is not yet in a position to confirm o r confute. But these should always be presented as hypotheses, not (as is too often done) as immutable certainties like the dogmas of religion. A lthough, m oreover, com ­ prehensive construction is p art of the business of philosophy, I do not believe it is the m ost im portant part. T he m ost im portant part,

[ 48 ]

BERTRAND RU SSE L L

to my mind, consists in criticizing and clarifying notions which are apt to be regarded as fundam ental and accepted uncritically. A s instances I might m ention: m ind, m atter, consciousness, knowledge, experience, causality, will, time. I believe all these notions to be inexact and approxim ate, essentially infected with vagueness, in­ capable of form ing p art of any exact science. O ut of the original manifold of events, logical structures can be built which will have properties sufficiently like those of the above com m on notions to account for their prevalence, but sufficiently unlike to allow a great deal of error to creep in through their acceptance as fundam ental. I suggest the following as an outline of a possible structure of the world; it is no more than an outline, and is not offered as m ore than possible. The world consists of a num ber, perhaps finite, perhaps infinite, of entities which have various relations to each other, and perhaps also various qualities. Each of these entities m ay be called an “event” ; from the point of view of old-fashioned physics, an event occupies a short finite tim e and a small finite am ount of space, but as we are not going to have an old-fashioned space and an oldfashioned time, this statem ent cannot be taken at its face value. Every event has to a certain num ber of others a relation which m ay be called “com presence” ; from the point of view of physics, a col­ lection of com present events all occupy one small region in space­ time. One exam ple of a set of com present events is w hat would be called the contents of one m an’s mind at one time— i.e. all his sensations, images, mem ories, thoughts, etc., which can coexist temporally. H is visual field has, in one sense, spatial extension, but this must not be confused with the extension of physical space-tim e; every part of his visual field is com present with every oth er part, and with the rest of “the contents of his m ind” at th at tim e, and a collection of com present events occupies a minim al region in space-time. There are such collections not only w here there are brains, but everywhere. A t any point in “em pty space,” a num ber of stars could be photographed if a cam era were introduced; we believe that light travels over the regions interm ediate between its source and o ur eyes, and therefore som ething is happening in these regions. If light from a num ber of different sources reaches a certain minimal region in space-tim e, then at least one event corresponding to each of these sources exists in this m inimal region, and all these events are com present. We will define a set of com present events as a “ minimal region.” We find that minimal regions form a four-dim ensional m anifold,

Logical Atom ism

[ 49 ]

and that, by a little logical m anipulation, we can construct from them the m anifold of space-tim e th at physics requires. We find also that, from a num ber of different minimal regions, we can often pick out a set of events, one from each, which are closely similar w hen they com e from neighboring regions, and vary from one region to another according to discoverable laws. These are the laws of the propagation of light, sound, etc. We find also that certain regions in space-tim e have quite peculiar properties; these are the regions which are said to be occupied by “ m atter.” Such regions can be collected, by m eans of the laws of physics, into tracks or tubes, very m uch m ore extended in one dimension of space-time than in the other three. Such a tube constitutes the “history” of a piece of m atter; from the point of view of the piece of m atter itself, the dim ension in which it is m ost extended can be called “tim e,” but it is only the private time of that piece of m atter, because it does not correspond exactly with the dimension in which another piece of m atter is most extended. N ot only is space-tim e very peculiar within a piece of m atter, but it is also rather peculiar in its neighborhood, growing less so as the spatio-tem poral distance grows greater; the law of this peculiarity is the law of gravitation. All kinds of m atter to some extent, but some kinds of m atter (viz. nervous tissue) more particularly, are liable to form “habits,” i.e. to alter their structure in a given environment in such a way that, when they are subsequently in a similar environm ent, they react in a new way, but if sim ilar environments recur often, the reaction in the end becomes nearly uniform , while remaining different from the reaction on the first occasion. (W hen I speak of the reaction o f a piece of m atter to its environm ent, I am thinking both of the constitution of the set of com present events of which it consists, and of the nature of the track in space-time which constitutes what we should ordinarily call its m otion; these are called a “ reaction to the environm ent” in so far as there are laws correlating them with characteristics of the environm ent.) O ut of habit, the peculiari­ ties of w hat we call “m ind” can be constructed; a mind is a track of sets of com present events in a region of space-time where there is m atter which is peculiarly liable to form habits. T he greater the liability, the m ore com plex and organized the mind becomes. Thus a m ind and a brain are not really distinct, but when we speak of a m ind we are thinking chiefly of the set of com present events in the region concerned, and of their several relations to other events form­ ing parts of other periods in the history of the spatio-tem poral tube which we are considering, whereas when we speak of a brain we

[ 50 )

BERTRAND RU SSELL

are taking the set of com present events as a whole, and considering its external relations to other sets of com present events, also taken as wholes; in a word, we are considering the shape of the tube, not the events of which each cross-section of it is com posed. T he above sum m ary hypothesis would, of course, need to be amplified and refined in m any ways in order to fit in com pletely with scientific facts. It is not put forw ard as a finished theory, but merely as a suggestion of the kind of thing th at may be true. It is of course easy to imagine other hypotheses which m ay be true, for exam ple, the hypothesis th at there is nothing outside the series of sets of events constituting my history. I do not believe th at there is any m ethod of arriving at one sole possible hypothesis, and therefore certainty in m etaphysics seems to me unattainable. In this respect I m ust adm it that m any other philosophies have the advantage, since in spite of their differences inter se, each arrives a t certainty of its own exclusive truth.

Philosophy, Metaphysics and Meaning

2

The Turning Point in Philosophy by

M O R IT Z S C H L IC K

(T R A N S L A T E D BY DA VID R Y N IN )

to tim e prizes have been established for essays on the question w hat progress philosophy has m ade in a given period. The period tends to be lim ited on the one side by the nam e of some great thinker, on the other by “the present.” It was thus assumed that there is som e degree of clarity regarding the philosophic progress of m an­ kind up to the time of that thinker, bu t th at it is dubious w hat further contributions have been m ade in recent times. Such questions clearly express a certain m istrust concerning the philosophy of the period which had recently elapsed. One has the im pression of being presented only with an em barrassed form ulation of the question: H as philosophy in that period made any progress w hatever? F o r if one were sure th at contributions had been made one would also know in w hat they consisted. If the m ore rem ote p ast is regarded with less scepticism and one is rath er inclined to see in its philosophy a continuous devel­ opm ent, the explanation m ay be th at one’s attitude towards every­ thing whose place is established in history is tinged with greater re­ spect. A fu rth er point is th at the older philosophers have at least dem onstrated their historical influence. H ence in considering them one can take as one’s base their historical rather than their substantive im portance, especially since one often does not venture to distinguish betw een the two. B ut it is just the ablest thinkers who m ost rarely have believed th at the results of earlier philosophizing, including that of the classical m odels, rem ain unshakable. This is shown by the fact that basically every new system starts again from the beginning, that every thinker F

rom

t im e

" D ie W e n d e D e r P h ilo so p h ic ,” as th is piece is called in G erm a n , o p en e d th e first n u m b e r o f V o lu m e I o f Erkenntnis (1 9 3 0 /3 1 ). I t is h ere p u b lish ed w ith the k in d p erm issio n o f M rs. S ch lick a n d P ro fe sso r C a rn a p , th e co -e d ito r o f Erkenntnis.

[ 53]

[ 54 ]

MORITZ SCHLICK

seeks his own foundation and does not wish to stand on the shoul­ ders of his predecessors. D escartes (n o t w ithout reason) felt him self to be m aking a wholly new beginning; Spinoza believed th at in introducing the (to be sure quite adventitious) m athem atical form he had found the ultim ate philosophical m ethod; and K ant was convinced th at on the basis of the way taken by him philosophy would at last adopt the sure p ath of a science. F u rth er exam ples are superfluous, for practically all great thinkers have sought for a radical reform of philosophy and considered it essential. This peculiar fate of philosophy has been so often described and bem oaned that it is indeed pointless to discuss it at all. Silent scepticism and resignation seem to be the only appropriate attitudes. Tw o thousand years of experience seem to teach th a t efforts to put an end to the chaos of systems and to change the fate of philosophy can no longer be taken seriously. T o point out th a t m an has finally succeeded in solving the m ost stubborn problem s, for exam ple th a t of Daedelus, gives an inform ed person no com fort; for w hat he fears is just th at philosophy will never arrive at a genuine “ problem .” I refer to this anarchy of philosophical opinions which has so often been described, in o rd er to leave no doubt th at I am fully con­ scious of the scope and weighty significance of the conviction th at I should now like to express. F o r I am convinced th at we now find ourselves at an altogether decisive turning point in philosophy, and th at we are objectively justified in considering th at an end has com e to the fruitless conflict of systems. We are already at the present tim e, in my opinion, in possession of m ethods which m ake every such con­ flict in principle unnecessary. W hat is now required is their resolute application. These m ethods have been quietly developed, unnoticed by the m ajority of those who teach o r write philosophy; and thus a situation has been created which is not com parable to any earlier one. T h at the situation is unique and that the turning em barked upon is really decisive can be understood only by becom ing acquainted with the new paths and by looking back, from the standpoint to which they lead, upon all those efforts that have ever passed as “ philosophical.” T he paths have their origin in logic. Leibniz dimly saw their beginning. B ertrand Russell and G ottlob Frege have opened up im ­ portant stretches in the last decades, but Ludw ig W ittgenstein (in his Tractatus Logico-Philosophicus, 1922) is the first to have pushed forw ard to the decisive turning point. It is well know n th at in recent decades m athem aticians have developed new logical m ethods, a t first prim arily for the solution

The Turning Point in Philosophy

[ 55 ]

of their ow n problem s which could not be overcom e by the trad i­ tional m ethods of logic. B ut the logic thus developed has also long since show n its superiority in other ways over the old form s, and doubtless will very soon supplant them . Was I referring to this logic as the powerful m eans which is in principle capable of raising us above all philosophical conflicts? D oes it give us general rules with those help all th e traditional problem s of philosophy can at least in prin­ ciple be resolved? If this were so I should hardly have had the right to say th at a wholly new situation had been created. F o r then there would have been only a gradual, as it were, technical progress, as for example, when the invention of the internal com bustion engine finally made possible th e solution of the problem of flight. However highly the value o f the new m ethods is to be esteem ed, it is plain that nothing so fundam ental can be brought about by the m ere development of a m ethod. T h e great turning point is therefore not to be attributed to logic itself but to som ething quite different which was indeed stim ­ ulated an d m ade possible by it, but which proceeds on a much deeper level: the insight into the nature of logic itself. T h a t the logical is in some sense the purely formal has been expressed early and often; however, one was not really clear con­ cerning the nature of pure form s. T he clue to their nature is to be found in the fact that every cognition is an expression o r repre­ sentation. T h a t is, it expresses a fact which is cognized in it. This can happen in any num ber of ways, in any language, by means of any arbitrary system of signs. All these possible modes of repre­ sentation— if they otherwise actually express the same knowledge— m ust have som ething in com m on; and w hat is com m on to them is th eir logical form . So all knowledge is such only by virtue of its form. It is through its form th at it represents the fact known. But the form cannot itself in turn be represented. It alone is concerned in cognition. Everything else in the expression is inessential and accidental m aterial, not different, say, from the ink by m eans of which we write down a statem ent. T his simple insight has consequences of the very greatest im ­ portance. A bove all, it enables us to dispose of the traditional p ro b ­ lems of “ the theory of knowledge.” Investigations concerning the hum an “capacity for knowledge,” in so far as they do not become part of psychology, are replaced by considerations regarding the n ature of expression, of representation, i.e. concerning every pos­ sible “language” in the most general sense of the term . Questions

[ 56 ]

MORITZ SCHLICK

regarding the “validity and limits of knowledge” disappear. E very­ thing is knowable which can be expressed, and this is the total subject m atter concerning which meaningful questions can be raised. There are consequently no questions which are in principle u n ­ answerable, no problem s which are in principle insoluble. W hat have been considered such up to now are not genuine questions, but meaningless sequences of words. T o be sure, they look like ques­ tions from the outside, since they seem to satisfy the custom ary rules of gram m ar, but in truth they consist of em pty sounds, because they transgress the profound inner rules of logical syntax discovered by the new analysis. W herever there is a meaningful problem one can in theory always give the path that leads to its solution. F o r it becom es evident that giving this path coincides with the indication of its meaning. T he practical following out of this path m ay of course be hindered by factual circum stances— by deficient hum an capacities, for exam ple. The act of verification in which the path to the solution finally ends is always of the same sort: it is the occurrence of a definite fact th at is confirmed by observation, by m eans of im m ediate experience. In this m anner the truth (o r falsity) of every statem ent, of daily life o r science, is determ ined. T here is thus no other testing and cor­ roboration of truths except through observation and em pirical science. Every science, (in so far as we take this w ord to refer to the content and not to the hum an arrangem ents for arriving at it) is a system of cognitions, that is, of true experiential statem ents. A nd the totality of sciences, including the statem ents of daily life, is the system of cognitions. T here is, in addition to it, no dom ain of “ philosophical” truths. Philosophy is not a system of statem ents; it is not a science. But what is it then? Well, certainly not a science, but neverthe­ less som ething so significant and im portant that it may henceforth, as before, be honored as the Q ueen of the Sciences. F o r it is now here written th at the Queen of the Sciences m ust itself be a science. The great contem porary turning point is characterized by the fact that we see in philosophy not a system of cognitions, but a system of acts ; philosophy is that activity through which the meaning of statem ents is revealed o r determ ined. By m eans of philosophy statem ents are explained, by m eans of science they are verified. T he latter is con­ cerned with the truth of statem ents, the form er with w hat they actually mean. T he content, soul and spirit of science is lodged naturally in what in the last analysis its statem ents actually m ean; the philo­ sophical activity of giving meaning is therefore the A lpha and O m ega of all scientific knowledge. This was indeed correctly surm ised when

The Turning Point in Philosophy

[ 57 ]

it was said th at philosophy supplied both the foundation and the apex of the edifice of science. It was a m istake, however, to suppose that the foundation was m ade up of “ philosophical” statem ents (the statem ents of theory of know ledge), and crowned by a dom e of philosophical statem ents (called m etaphysics). It is easy to see that the task of philosophy does not consist in asserting statem ents— that bestowing meaning upon statem ents can­ not be done in turn by statem ents. F or if, say, I give the meaning of my w ords through explanatory statem ents and definitions, that is by help of other words, one m ust ask further for the meaning of these words, and so on. This process cannot proceed endlessly. It always com es to an end in actual pointings, in exhibiting what is m eant, thus in real acts; only these acts are no longer capable of, o r in need of, further explanation. T he final giving of meaning always takes place therefore, through deeds. It is these deeds or acts which constitute philosophical activity. It was one of the m ost serious errors of form er times to have believed th at the actual m eaning and ultim ate content was in turn to be form ulated in statem ents, and so was representable in cognitions. This was the erro r of “ m etaphysics.” T he efforts of metaphysicians were always directed upon the absurd end of expressing the content of pure quality (th e “essence” of things) by means of cognitions, hence of uttering the unutterable.1 Qualities cannot be “said.” They can only be shown in experience. But with this showing, cognition has nothing to do. T hus m etaphysics collapses not because the solving of its tasks is an enterprise to which the hum an reason is unequal (as for ex­ am ple K ant thought) but because there is no such task. With the disclosure of the m istaken form ulation of the problem the history of metaphysical conflict is likewise explained. If o u r conception is in general correct we must be able to establish it historically. It would have to be capable of giving some account of the change in m eaning of the word “philosophy.” Now this is actually the case. If in ancient times, and actually until recently, philosophy was simply identical with every purely theoretical scientific investigation, this points to the fact that science found itself in a state in which it saw its main task still in the clari­ fication o f its fundam ental concepts. The em ancipation of the special sciences from their com m on m other, philosophy, indicates that the m eaning of certain fundam ental concepts became clear enough to m ake successful further w ork with them possible. If, today, ethics 1. S ee m y a rtic le “ E rleb e n , E rk e n n e n , M e tap h y sik ,” Kantstudien, V ol. 31 (1 9 3 0 ).

J 5g ]

M ORITZ SCHLICK

and aesthetics, and frequently also psychology, are considered branches of philosophy, this is a sign th a t these studies do not yet possess sufficiently clear basic concepts, th at their efforts are still chiefly directed upon the meaning of their statem ents. Finally, if within a well-established science the necessity suddenly arises at som e point of reflecting anew on the true m eaning of the fundam ental concepts, and thereby a m ore profound clarification of their m ean­ ing is achieved, this will be felt at once as an em inent philosophical achievement. All are agreed that, for instance, E instein’s w ork, p ro ­ ceeding from an analysis of the m eaning of statem ents about tim e and space, was actually a philosophical achievem ent. H ere we should add that the decisive epoch-m aking forw ard steps of science are always of this character; they signify a clarification of the m eaning of the fundam ental statem ents and only those succeed in them who are endowed for philosophical activity. T h e great investigator is also always a philosopher. Frequently also the nam e of philosophy is bestow ed on m ental activities which have as their concern not pure knowledge but the con­ duct of life. This is readily understandable. F o r the wise m an rises above the uncom prehending mass just by virtue of the fact th at he can point out m ore clearly than they the meaning of statem ents and questions concerning life relationships, facts and desires. T he great turning point of philosophy signifies also a decisive turning away from certain erroneous paths which have been em barked upon since the second half of the 19th century and which m ust lead to quite a w rong assessment and evaluation of philosophy. I m ean the attem pts to claim for it an inductive character and accordingly to believe th at it consists solely of statem ents of hypothetical validity. The idea of claiming only probability for its statem ents was rem ote from earlier thinkers. They would have rejected it as incom patible with the dignity of philosophy. In this was expressed a healthy instinct for the fact th at philosophy must supply the ultim ate support of know l­ edge. The reverse side of the medal is the dogm a th at philosophy supplies unconditionally true a priori axioms, which we m ust regard as an extremely unfortunate expression of this instinct, particularly since philosophy does not consist of statem ents at all. B ut we too be­ lieve in the dignity of philosophy and deem incom patible with it the character of being uncertain and only probable; and we are happy that the decisive turning point m akes it impossible to attribute any such character to it. F or the concept of probability o r uncertainty is simply not applicable to the acts of giving m eaning which constitute philos­ ophy. It is a m atter of positing the m eaning of statem ents as som ething

The Turning Point in Philosophy

[ 59 ]

simply final. E ith er we have this meaning, and then we know w hat is m eant by the statem ent, o r we do not possess it, in which case mere em pty words confront us, and as yet no statem ent at all. There is noth­ ing in betw een and there can be no talk of the probability that the m eaning is the right one. T hus after the great turning point philosophy shows its decisive character even m ore clearly than before. It is only, indeed, because of this character that the conflict of system s can be ended. I repeat: in consequence of the insights which I have sketched we may today consider it as in principle already ended. I hope th at this may becom e increasingly clear in the pages of this journal* in the new period of its existence. C ertainly there will still be m any a rear-guard action. Certainly m any will for centuries continue to w ander further along the traditional paths. Philosophical writers will long continue to discuss the old pseudo-questions. But in the end they will no longer be listened to; they will com e to resemble actors who continue to play for some tim e before noticing th at the audience has slowly departed. Then it will no longer be necessary to speak of “ philosophical problem s” for one will speak philosophically concerning all problems, th at is: clearly and meaningfully. * Sc. Erkenntnis, E d.

3

The Elimination of Metaphysics Through Logical Analysis of Language by

RUDOLF CARNAP

(T R A N S L A T E D BY A R T H U R P A P )

1.

I n t r o d u c t io n

h e r e h a v e b e e n m any opponents of metaphysics from the G reek skeptics to the em piricists of the 19th century. Criticism s of very diverse kinds have been set forth. M any have declared th a t the doc­ trine of metaphysics is false, since it contradicts our em pirical know l­ edge. Others have believed it to be uncertain, on the ground th a t its problems transcend the limits of hum an knowledge. M any anti­ metaphysicians have declared th at occupation with m etaphysical ques­ tions is sterile. W hether or not these questions can be answ ered, it is at any rate unnecessary to worry about them ; let us devote ourselves entirely to the practical tasks which confront active m en every day of their lives! T h e developm ent of modern logic has m ade it possible to give a new and sharper answer to the question of the validity and justi­ fication of metaphysics. The researches of applied logic o r the theory of knowledge, which aim at clarifying the cognitive content o f sci­ entific statem ents and thereby the m eanings of the term s th a t occur in the statem ents, by m eans of logical analysis, lead to a positive and to a negative result. T he positive result is w orked out in the dom ain of em pirical science; the various concepts of the various branches of science are clarified; their form al-logical and epistem ological connections are m ade explicit. In the dom ain of metaphysics,

T

T h is a rtic le , originally e n title d “ Ü b erw in d u n g d e r M e ta p h y sik d u rc h L o g isch e A n aly se d e r S p ra c h e ,” a p p e a re d in Erkenntnis, V o l. II ( 1 9 3 2 ). It is p u b lis h e d h e re w ith the kind perm issio n o f P ro fesso r C a rn a p .

[ 60]

The Elimination o f Metaphysics

[ 61 ]

including all philosophy of value and normative theory, logical analysis yields the negative result that the alleged statements in this domain are entirety meaningless. Therewith a radical elimination of metaphysics is attained, which was not yet possible from the earlier ahtimetaphysical standpoints. It is true that related ideas may be found already in several earlier trains of thought, e.g. those of a nominalistic kind; but it is only now when the development of logic during recent decades provides us with a sufficiently sharp tool that the decisive step can be taken. In saying that the so-called statements of metaphysics are mean­ in g le s s we intend this word in its strictest sense. In a loose sense of the word a statement or a question is at times called meaningless if it is entirely sterile to assert or ask it. We might say this for in­ stance about the question “what is the average weight of those inhabi­ tants of Vienna whose telephone number ends with ‘S’?’* or about a statement which is quite obviously false like “in 1910 Vienna had 6 inhabitants” o r about a statement which is not just empirically. t?ut_logiQally falseT a contradictory statement such as “persons A and B are each a year older than the other.” Such sentences are really meaningful, though they are pointless or false; Jor_ji_is_ojity. meaningful sentences that are even divisible into i theoretically) fruit­ ful and sterile, true and false. In the strict sense, however, a sequence of words is meaningless if it does not, within a specified language, constitute a statement. It may happen that such_a sequence of words looks like j j statement at first glance; in that case we call it a pseudoW uemenL O ur thesis, now, is that logical analysis reveals the alleged statements of metaphysics to be pseudo-statements. A language consists of a vocabulary and a syntax, i.e. a set of words which have meanings and rules of sentence formation. These rules indicate how sentences may be formed out of the various sorts of words. Accordingly, there are two kinds of jpseudo-statements: either they contain a word which is erroneously believed to have meaning», or the constituent.words are meaningful^yej. arej?ut together m a counter-syntactical way, sa .th at they, do not yield a meaningful statement. We shall show in terms of examples that pseudo-statements of both kinds occur in metaphysics. Later we shall have to inquire into the reasons that support our contention that metaphysics in its entirety consists of such pseudo-statements. 2 . T h e S ig n if ic a n c e o f a W ord

A word which (within a definite language) has a meaning, is usually also said to designate a concept; if it only seems to have a

[ 52 ]

RUDOLF CARNAP

meaning while it really does not, we speak of a “pseudo-concept.” How is the origin of a pseudo-concept to be explained? H as not every word been introduced into the language for no other purpose than to express something or other, so that it had a definite meaning from the very beginning of its use? How, then, can a traditional lan­ guage contain meaningless words? To be sure, originally every word (excepting rare cases which we shall illustrate later) had a meaning. In the course of historical development a word frequently changes its meaning. A nd it also happens at times that a word loses its old sense without acquiring a new one. It is thus that a pseudo-concept arises. W hat, now, is the meaning of a word? W hat stipulations con­ cerning a word must be made in order for it to be significant? (It does not m atter for our investigation whether these stipulations are explicitly laid down, as in the case of some words and symbols of modern science, o r whether they have been tacitly agreed upon, as is the case for most words of traditional language.) First, the syntax o f thft wnrH milg> ^ iim d, j f , the mode of its_OCCiirrence_in the^ ^Simplest senteflCfc-form-4ft-whirh it k capable of occurring: we call this sentence form its elementary sentence . The elementaiy sentence form for the word “stone” e.g. is “x is a .stonol’; in sentences of this form some designation from the category of things occupies the place of “x,” e.g. “this diamond,” “this apple.” Secondly, fo r.a n elementary S \h$ wnrH an answer must be given to the following question, which can be formulated in various ways: (1 .) W hat sentences is S deducible from, and what sentences are deducible from S? (2 .) Under what conditions is S supposed to be true, and under what conditions false? (3 .) How is S to be verified? (4 .) W hat is the meaning of S? (1 ) is the correct formulation; formulation (2 ) accords with the phraseology of logic, (3 ) with the phraseology of the theory of knowledge, (4 ) with that of philosophy (phenom enology). W itt­ genstein has asserted that (2 ) expresses what philosophers mean by ( 4 ) : the meaning of a sentence consists in its truth-condition. ( ( 1 ) is the “metalogical” formulation; it is planned to give elsewhere a de­ tailed exposition of metalogic as the theory of syntax and meaning, i.e. relations of deducibility.) In the case of many words, specifically in the case of the over­ whelming majority of scientific words, it is possible to specify their meaning by reduction to other words ( “constitution,” definition).

The Elimination o f Metaphysics

[ 63 ]

E.g. “ ‘arthropodes’ are animals with segmented bodies and jointed legs.” Thereby the above-mentioned question for the elementary sen­ tence form of the word “arthropode,” that is for the sentence form “the thing x is an arthropode,” is answered: it has been stipulated that a sentence of this form is deducible from premises of the form “x is an animal,” “x has a segmented body,” “x has jointed legs,” and that conversely each of these sentences is deducible from the form er sentence. By means of these stipulations about deducibility (in other words: about the truth-condition. about the method of verij>ode” the meaning of the word “arthropode” is fixed. In this way every word of the language is reduced to other words and finally to the words which occur in the so-called. “.observatioa senience&ILor “protocol sentences.” It is through this reduction that the word acquires its meaning. F o r our purposes we may ignore entirely the question concerning the content and form of the primary sentences (protocol sentences) which has not yet been definitely settled. In the theory of knowledge it is customary to say that the primary sentences refer to “the given” ; but there is no unanimity on the question what it is that is given. At times the position is taken that sentences about the given speak of the simplest qualities of sense and feeling (e.g. “warm,” “blue,” “joy” and so fo rth ); others incline to the view that basic sentences refer to .total experiences and similarities between them: a still different view has it that even the basic sentences speak of things. Regardless of this diversity of opinion it is certain that a sequence of words has .a meaning only if its relations of deducibility to the protocol sentences are fixed, whatever the characteristics of the protocol sen''leuces 'may &e;""and similarly, that a word is significant only., if the sentences in which it may .ocxpc^eTx$3ucitte_tp pxp.totol sentences. Since the meaning of a word is determined by its criterion of application (in other words: by the relations of deducibility entered into by its elementary sentence-form, by its truth-conditions, by the method of its verification), the stipulation of the criterion takes away one’s freedom to decide what one wishes to “mean” by the word. If the word is to receive an exact meaning, nothing less than the criterion of application must be given; but one cannot, on the other hand, give more than the criterion of application, for the latter is a sufficient determination of meaning. The meaning is im­ plicitly contained in the criterion; all that remains to be done is to make the meaning explicit. Let us suppose, by way of illustration, that someone invented

[ 64 ]

RUDOLF CARNAP

the new word “teavy” and m aintained that there are things which are teavy and things which are not teavy. In order to learn the meaning of this word, we ask him about its criterion o f application: how is one to ascertain in a concrete case w hether a given thing is teavy o r not? L et us suppose to begin with that we get no answer from him : there are no em pirical signs of teavy ness, he says. In that case we would deny the legitimacy of using this word. If the person who uses the word says that all the same there are things which are teavy and there are things which are not teavy, only it remains for the weak, finite intellect of m an an eternal secret which things are teavy and which are not, we shall regard this as em pty verbiage. But perhaps he will assure us that he m eans, after all, som ething by the word “teavy.” But from this we only learn the psychological fact that he associates some kind of images and feel­ ings with the word. The word does not acquire a m eaning through such associations. ,If no criterion of application for the word is stipulated, then nothing is asserted by the sentences in w hich it occurs, they are but pseudo-statem ents. Secondly, take the case when we are given a criterion of appli­ cation for a new word, say “ toovy” ; in particular, let the sentence “this thing is toovy” be true if and only if the thing is quadrangular (It is irrelevant in this context w hether the criterion is explicitly stated o r whether we derive it by observing the affirmative and the nega­ tive uses of the w o rd ). Then we will say: the word “ toovy” is synon­ ymous with the word “quadrangular.” A nd we will not allow its users to tell us th at nevertheless they “intended” som ething else by it than “quadrangular” ; that though every quadrangular thing is also toovy and conversely, this is only because quadrangularity is the visible m anifestation of toovyness, b u t that the latter itself is a hidden, not itself observable property. We would reply th at after the criterion of application has been fixed, the synonymy o f “ toovy” and “quadrangular” is likewise fixed, and that we are no further at liberty to “ intend” this or th at by the word. L et us briefly summ arize the result of our analysis. L et “ a” be any word and “S ( a ) ” the elem entary sentence in which it occurs. T hen the sufficient and necessary condition for “ a” being meaninnful may be given by each of the following form ulations, which ulti­ mately say the same thing: 1. T he empirical criteria fo r a are known. 2. It has been stipulated from w hat protocol sentences “ S ( a ) ” is deducible.

The Elimination o f Metaphysics

[

55

j

3. The truth-conditions for “S ( a ) ” are fixed. 4. T he method of verification of “S(a>” is known.1 3. M e ta ph y sic a l W ords W it h o u t M ea n in g

Many words of metaphysics, now, can be shown not to fulfill the above requirement, and therefore to be devoid of meaning. Let us take as an example the metaphysical term “principle” (in the sense of principle of being, not principle of knowledge or axiom ). Various metaphysicians offer an answer to the question which is the (highest) “principle of the world” (or of “things,” of “existence,” of “being” ), e.g. water, number, form, motion, life, the spirit, the idea, the unconscious, activity, the good, and so forth. In order to discover the meaning of the word “principle” in this meta­ physical question we must ask the metaphysician under what con­ ditions a statement of the form “x is the principle of y” would be true and under what conditions it would be false. In other words: we ask for the criteria of application or for the definition of the word “principle.” The metaphysician replies approximately as follows: “x is the principle of y” is to mean “y arises out of x,” “the being of y rests on the being of x,” “y exists by virtue of x” and so forth. But these words are ambiguous and vague. Frequently they have a clear meaning; e.g., we say of a thing or process y that it “arises out of* x when we observe that things or processes of kind x are fre­ quently o r invariably followed by things or processes of kind y (causal connection in the sense of a lawful succession). But the metaphysician tells us that he does not mean this empirically-Qbser vabla relationship. F or in that case his metaphysical theses would be merely em piricaf propositions of the same kind as those of physics. T]ie_ expression “arising from ” is not to mean h ere,a relatio n of tempo ral and-eausal Sequence., which is what the word ordinarily-means. Yet, no criterion is specified for any other meaning. Consequently, the alleged “meta­ physical” meaning, which the word is supposed to have here in con­ trast to the mentioned empirical meaning, does not exist. If we reflect on the original meaning of the word “principium” (and of the corresponding Greek word apxo')> we notice the same development. The word is explicitly deprived of its original meaning “beginning”; it is not supposed to mean the temporally prior any more, but the prior in some other, specifically metaphysical, respect. The criteria for this “metaphysical respect,” however, are lacking. In both cases, 1. For the logical and epistemological conception which underlies our exposition, but can only briefly be intimated here, cf. Wittgenstein, Tractatus Logico-Philosophicus, 1922, and Carnap, Der logische Aufbau der Welt, 1928.

[

66

]

RUDOLF CARNAP

then, the w ord has been deprived of its earlier m eaning w ithout being given a new m eaning; there rem ains the w ord as an em pty shell. F rom an earlier period of significant use, it is still associatively connected with various m ental images; these in turn get associated with new m ental images and feelings in the new context of usage. B ut the w ord does not thereby become m eaningful; and j t rem ains meaningless as long as no m ethod o f verification can b e describedt A nother exam ple is the w ord “G od.” H ere w e m ust, ap art from the variations of its usage within each dom ain, distinguish the linguistic usage in three different contexts o r historical epochs, which however overlap tem porally. In its mythological use the w ord has a clear meaning. It, o r parallel words in other languages, is som e­ times used to denote physical beings which are enthroned on M ount Olympus, in H eaven o r in H ades, and which are endow ed with power, wisdom, goodness and happiness to a greater o r lesser extent. Sometimes the w ord also refers to spiritual beings which, indeed, do not have manlike bodies, yet m anifest them selves nevertheless somehow in the things o r processes of the visible w orld and are therefore em pirically verifiable. In its metaphysical use, on the other hand, the word “G od” \refers to som ething beyond experience. T he word is deliberately divested of its reference to a physical being o r to a spiritual being th at is im m anent in the physical. A nd as it is not given^a new m eaning, it becom es m eaningless. T o b e sure. It often looks as though the word “G od” had a m eaning even in metaphysics. But the definitions which are set up prove on closer inspection to be pseudo-definitions. They lead either to logically illegitimate com bina­ tions of words (o f which we shall treat later) o r to oth er m eta­ physical words (e.g. “prim ordial basis,” “ the absolute,” “the uncon­ ditioned,” “the autonom ous,” “the self-dependent” and so fo rth ), but in no case to the tru th -c o n d itio n sj)f its elem entary sentences. In the case of this word not even the first req u irem en to F lo g ic ls m e t, that is the requirem ent to specify its syntax, i.e. the form of its occurrence in elem entary sentences. A n elem entary sentence would here have to be of the form “ x is a G od” ; yet, the m etaphysician either rejects this form entirely w ithout substituting another, o r if he accepts it he neglects to indicate the syntactical category o f the variable x. (Categories are, for exam ple, m aterial things, properties of things, relations between things, num bers e tc .). T he theological usage of the word “G od” falls betw een its m ytho­ logical and its metaphysical usage. There is no distinctive m eaning here, but an oscillation from one of the m entioned two uses to the other. Several theologians have a clearly em pirical (in our term in-

The Elimination o f Metaphysics

[ 67 ]

ology, “mythological” ) concept of God. In this case there are no pseudo-statements; but the disadvantage for the theologian lies in the circumstance that according to this interpretation the statements of theology are empirical and hence are subject to the judgment of empirical science. The linguistic usage of other theologians is clearly metaphysical. Others again do not speak in any definite way, whether this is because they follow now this, now that linguistic usage, or because they express themselves in terms whose usage is not clearly classifiable since it tends towards both sides. Just like the examined examples “principle” and “God,” most of the other specifically metaphysical terms are devoid of meaning e.g. “the idea/* “the AEsoiute, “the Unconditioned,” “the Infinite,” “the being of being,” “non-being,” “thing in itself,” “absolute spirit,” “objective spirit,” “essence,” “being-in-itself,” “being-in-and-foritseif,” “emanation,” “manifestation,” “articulation,” “the Ego,” “the non-Ego,” etc. These expressions are in the same boat with “teavy,” our previously fabricated example. The metaphysician tells us that empirical truth-conditions cannot be specified; if he adds that nevertheless he “means” something, we know that this is merely an allusion to associated images and feelings which, however, do not bestow a meaning on the word. The alleged statements of meta­ physics which contain such words have no sense, assert nothing, are mere pseudo-statements. Into the explanation of their historical origin we shall inquire later. 4 . T h e S ig n if ic a n c e o f a S e n t e n c e So far we have considered only those pseudo-statements which j y n t ain a meaningless word. But there is a second kind of pseudo­ statement. J h e y consist of meaningful words, but the words are put together in such a way that nevertheless no meaning results. The syntax of a language specifies which combinations of words are admissible and which inadmissible. The grammatical syntax of natural languages, however, does not fulfill the task of elimination of senseless combinations of words in all cases. Let us take as examples the following sequences of words: 1. “Caesar is and” 2. “Caesar is a prime number” The word sequence (1 ) is formed countersyntactically; the rules of syntax require that the third position be occupied, not by a conjunction, but by a predicate, hence by a noun (with article) or by an adjective. The word sequence “Caesar is a general,” e.g., is

J

68

]

RUDOLP CARNAP

form ed in accordance with the rules of syntax. It is a meaningful w ord sequence, a genuine sentence. But, now, w ord sequence ( 2 ) is likewise syntactically correct, for it has the same gram m atical form as the sentence just m entioned. Nevertheless (2 ) is meaningless. “ Prim e num ber” is a predicate of num bers; it can be neither affirmed nor denied of a person. Since ( 2 ) looks like a statem ent yet is not a statem ent, does not assert anything, expresses neither a true n o r a false proposition, we call this word sequence a “pseudo-statem ent.” The fact that the rules of gram m atical syntax are not violated easily seduces one at first glance into the erroneous opinion that one still has to do with a statem ent, albeit a false one. B ut “ a is a prim e num ber” is false if and only if a is divisible by a natural num ber different from a and from 1; evidently it is illicit to put here “ C aesar” for “a.” This exam ple has been so chosen th a t the nonsense is easily detectable. M any so-called statem ents of m etaphysics are not so easily recognized to be pseudo-statem ents. T he fact th a t natural languages allow the form ation of meaningless sequences o f w ords without violating the rules of gram m ar, indicates th at gram m atical syntax is, from a logical point of view, inadequate. If gram m atical syntax corresponded exactly to logical syntax, pseudo-statem ents could not arise. If gram m atical syntax differentiated not only the w ordcategories of nouns, adjectives, verbs, conjunctions etc., b u t within each of these categories m ade the further distinctions th at are logically indispensable, then no pseudo-statem ents could be form ed. If, e.g., nouns were gram m atically subdivided into several kinds of words, according as they designated properties of physical objects, of num bers etc., then the words “ general” and “prim e num ber” would belong to gram m atically different word-categories, and (2 ) would be just as linguistically incorrect as ( 1 ) . In a correctly con­ structed language, therefore, all nonsensical sequences of words would be of the kind of exam ple ( 1 ) . Considerations o f gram m ar would already eliminate them as it were autom atically; i.e. in order to avoid nonsense, it would be unnecessary to pay attention to the meanings of the individual words over and above their syn­ tactical type (th eir “ syntactical category,” e.g. thing, property of things, relation between things, num ber, property of num bers, relation between num bers, and so fo rth ). It follows th at if our thesis that the statem ents of m etaphysics are pseudo-statem ents is justifiable, then m etaphysics-cm ilri not, ftyftn Sn--a logically constructed language. This is the great philosophical im­ portance of the task, which at present occupies the logicians, of building a logical syntax.

[69]

The Elimination o f Metaphysics 5. M

e t a p h y s ic a l

P s e u d o -s t a t e m e n t s

L et us now take a look at some exam ples of metaphysical pseudo­ statem ents of a kind where the violation of logical syntax is especially obvious, though they accord with historical-gram m atical syntax. We select a few sentences from that m etaphysical school which at present exerts the strongest influence in G erm any.2 “W hat is to be investigated is being only and— nothing else; being alone and further— nothing; solely being, and beyond being— nothing. What about this Nothing? . . . Does the Nothing exist only because the Not, i.e. the Negation, exists? O r is it the other way around? Does Negation and the N ot exist only because the Nothing exists? . . . We assert: the Nothing is prior to the N ot and the Nega­ tion. . . . W here do we seek the Nothing? H ow do we find the Nothing. . . . We know the Nothing. . . . Anxiety reveals the Nothing. . . . T h a t fo r which and because of which we were anxious, was ‘really’— nothing. Indeed: the N othing itself— as such— was present. . . . What about this Nothing?— The Nothing itself nothings.” In o rd er to show that the possibility of forming pseudo-statem ents is based on j i logical defect of language, we set up the schema below. T he sentences under I are gram m atically as well as logically im­ peccable, hence meaningful. The sentences under II (excepting B 3) are in gram m atical respects perfectly analogous to those under I. Sentence form IIA (as question and answ er) does not, indeed, satisfy the requirem ents to be imposed on a logically correct language. But it is nevertheless meaningful, because it is translatable into correct language. T his is shown by sentence IIIA , which has the same m ean­ ing as IIA . Sentence form IIA then proves to be undesirable because we can be led from it, by m eans of gram matically faultless operations, to the meaningless sentence form s IIB , which are taken from the above quotation. These form s cannot even be constructed in the correct language of Colum n III. Nonetheless, their nonsensicality is not obvious at first glance, because one is easily deceived by the analogy with the meaningful sentences IB. The fault of our language identified here lies, therefore, in the circum stance that, in contrast to a logically correct language, it admits of the same grammatical form for meaningful and meaningless word sequences. To each sentence in w ords we have added a corresponding form ula in the 2.

The

follow ing q u o ta tio n s

(o rig in al ita lic s)

a re ta k en

fro m

M . H eidegger,

Was 1st M etaphysik? 1929. W e co u ld ju st as well have selected passages fro m any o th e r o f th e n u m e ro u s m e tap h y sicia n s o f th e present o r o f th e p a s t; yet th e selected p assag es seem to us to illu stra te o u r thesis especially well.

RUDOLF CARNAP

[ 70 1

notation of symbolic logic; these form ulae facilitate recognition of the undesirable analogy between IA and IIA and therew ith o f the origin of the meaningless constructions IIB. /.

II. Transition from Sense to Nonsense in Ordinary Language

Meaningful Sentences of Ordinary Language

A . W hat is outside? O u (? ) R ain is outside O u (r)

III. Logically Correct Language

A. T h ere is n othing A. W h at is outside? (does not exist a n y ­ O u (? ) thing) w hich is N o th in g is outside outside. O u (n o ) ~ (3 x ).O u ( x )

B. W hat about this rain? B. “W h at about this N othing?” ? (n o ) (i.e. w hat does the rain do? o r: w hat else can be said about this rain?

B. N o n e o f these form s can even be co n stru cted .

?(r) 1. W e know the rain K (r)

1. “W e seek the N o th in g ” “W e find th e N o th in g ” “W e know the N o th in g ” K (n o )

2. T he rain rains R (r)

2. “T h e N othing nothings” N o (n o ) 3. “T h e N othing exists only because . . .” E x (n o )

O n closer inspection of the pseudo-statem ents under IIB , we also find some differences. T he construction of sentence ( 1 ) is simply based on the m istake of em ploying the w ord “ nothing” as a noun, because it is custom ary in ordinary language to use it in this form in order to construct a negative existential statem ent (see IIA ). In a correct language, on the other hand, it is not a particular name, but a certain logical form of the sentence th at serves this purpose

The Elimination o f Metaphysics

( 71 ]

(see III A ). Sentence IIB2 adds something new, viz. the fabrication of the meaningless word “to nothing.” This sentence, therefore, is senseless for a twofold reason. We pointed out before that the mean­ ingless words_,Qf,metaphysics, usually o w e jh e ir origin tp the fact th at^x u ean in g fu 1 w o rd js depnye_d. pf its m egping through its meta­ phorical use in mptapFiycirc Rut here we confront one’ of those rare cases where a new word is introduced which never had a meaning to begin with. Likewise sentence IIB3 must be rejected for two rea­ sons. In respect of the error of using the word “nothing” as a noun, it is like the previous sentences. But in addition it involves a contra­ diction. For even if it were admissible to introduce “nothing” as a name o r description of an entity, still the existence of this entity would be denied in its very definition, whereas sentence (3 ) goes on to affirm its existence. This sentence, therefore, would be contradic­ tory, hence absurd, even if it were not already meaningless. In view of the gross logical errors which we find in sentences IIB, we might be led to conjecture that perhaps the word “nothing” has in Heidegger’s treatise a meaning entirely different from the customary one. A nd this presumption is further strengthened as we go on to read there that anxiety reveals the Nothing, that the Nothing itself is present as such in anxiety. For here the word “nothing” seems to refer to a certain emotional constitution, possibly of a religious sort, o r something or other that underlies such emotions. If such were the case, then the mentioned logical errors in sentences IIB would not be committed. But the first sentence of the quotation at the beginning of this section proves that this interpretation is not possible. The combination of “only” and “nothing else” shows un­ mistakably that the word “nothing” here has the usual meaning of a logical particle that serves for the formulation of a negative exis­ tential statement. This introduction of the word “nothing” is then immediately followed by the leading question of the treatise: “What about this Nothing?” . But our doubts as to a possible misinterpretation get completely dissolved as we note that the author of the treatise is clearly aware of the conflict between his questions and statements, and logic. "Question and answer in regard to the Nothing are equally absurd in themselves. . . . The fundamental rule of thinking commonly appealed to, the law of prohibited contradiction, general 'logic/ destroys this question.” All the worse for logic! We must abolish its sovereignty: “If thus the power of the understanding in the field of questions concerning Nothing and Being is broken, then the fate of the sovereignty of ‘logic* within philosophy is thereby decided as

£ 72 ]

RUDOLF CARNAP

well. The very idea of ‘logic’ dissolves in the whirl of a m ore basic questioning.” But will sober science condone the whirl of counterlogical questioning? T o this question too there is a ready answ er: “T he alleged sobriety and superiority of science becom es ridiculous if it does not take the N othing seriously.” Thus we find here a good confirmation of o u r thesis; a m etaphysician himself here states that his questions and answers are irreconcilable with logic and the sci­ entific way of thinking. The difference between our thesis and that of the earlier anti­ metaphysicians should now be clear. We do not regard metaphysics as “ mere speculation” o r “fairy tales.” T he statem ents of a fairy tale do not conflict with logic, but only with experience; they are perfectly meaningful, although false. M etaphysics is not “supersti­ tion”; it is possible to believe true and false propositions, but not to believe meaningless sequences of words. M etaphysical statem ents are not even acceptable as “working hypotheses”; for an hypothesis m ust be capable of entering into relations o f deducibility with (tru e o r false) em pirical statem ents, which is just w hat pseudo-statem ents cannot do. W ith reference to the so-called limitation of human knowledge an attem pt is sometimes m ade to save m etaphysics by raising the following objection: m etaphysical statem ents are not, indeed, veri­ fiable by m an nor by any other finite being; nevertheless they might be construed as conjectures about the answers which a being with higher o r even perfect powers of knowledge would m ake to our questions, and as such conjectures they would, after all, be m ean­ ingful. T o counter this objection, let us consider the following. If the meaning of a word cannot be specified, o r if the sequence of words does not accord with the rules of syntax, then one has not even asked a question. (Ju st think of the pesudo-questions: “ Is this table teavy?” , “ is the num ber 7 holy?” , “ which num bers are darker, the even or the odd ones?” ). W here there is no question, not even an omniscient being can give an answer. Now the objector may say: just as one who can see may com m unicate new knowledge to the blind, so a higher being might perhaps com m unicate to us m eta­ physical knowledge, e.g. w hether the visible world is the m anifesta­ tion of a spirit. H ere we m ust reflect on the m eaning of “ new knowl­ edge.” It is, indeed, conceivable that we might encounter anim als who tell us about a new sense. If these beings were to prove to us F erm at’s theorem o r were to invent a new physical instrum ent or were to establish a hitherto unknow n law of nature, then our knowl­ edge would be increased with their help. F o r this sort of thing we

The Elimination o f Metaphysics

[ 73 ]

can test, just the way even a blind m an can understand and test the whole of physics (and therewith any statem ent m ade by those who can see). But if those hypothetical beings tell us something which we cannot verify, then we cannot understand it either; in that case no inform ation has been com m unicated to us, but mere verbal sounds devoid of meaning though possibly associated with images. It fol­ lows that our knowledge can only be quantitatively enlarged by other beings, no m atter w hether they know more or less or every­ thing, but no knowledge of an essentially different kind can be added. W hat we do not know for certain, we may come to know with greater certainty through the assistance of other beings; but what is unintelligible, meaningless for us, cannot become meaningful through someone else’s assistance, however vast his knowledge might be. Therefore no god and no devil can give us metaphysical knowledge. 6. M

e a n in g l e s s n e s s

o f

all

M

e t a p h y s ic s

T he exam ples of metaphysical statem ents which we have anal­ yzed were all taken from just one treatise. But our results apply with equal validity, in part even in verbally identical ways, to other m eta­ physical systems. T hat treatise is completely in the right in citing approvingly a statem ent by Hegel ( “pure Being and pure Nothing, therefore, are one and the sam e” ). T he metaphysics of Hegel has exactly the same logical character as this m odern system of m eta­ physics. A nd the same holds for the rest of the metaphysical systems, though the kind of phraseology and therewith the kind of logical errors that occur in them deviate more o r less from the kind that occurs in the exam ples we discussed. It should not be necessary here to adduce further examples of specific metaphysical sentences in diverse systems and submit them to analysis. We confine ourselves to an indication of the most fre­ quent kinds of errors. Perhaps the m ajority of the logical mistakes that are committed when pseudo-statem ents are made, are based on the logical faults infecting the use of the word “to be” in our language (and of the corresponding words in other languages, at least in most European languages). T he first fault is the ambiguity of the word “to be.” It is sometimes used as copula prefixed to a predicate ( “ I am hungry” ), sometimes to designate existence ( “ I am ” ). This mistake is ag­ gravated by the fact that metaphysicians often are not clear about this ambiguity. The second fault lies in the form of the verb in its second meaning, the m eaning of existence. The verbal form feigns

|

74

j

RUDOLF CARNAP

a predicate where there is none. T o be sure, it has been known for a long time that existence is not a property (cf. K ant’s refutation of the ontological proof of the existence of G o d ). B ut it was not until the advent of m odem logic that full consistency on this point was reached: the syntactical form in which m odem logic introduces the sign for existence is such th at it cannot, like a predicate, be ap­ plied to signs for objects, bu t only to predicates (cf. e.g. sentence IIIA in the above tab le ). M ost m etaphysicians since antiquity have allowed themselves to be seduced into pseudo-statem ents by the verbal, and therewith the predicative form of the w ord “to be,” e.g. “ I am ,” “G od is.” W e meet an illustration of this erro r in D escartes’ “cogito, ergo sum .” L et us disregard here the m aterial objections th at have been raised against the premise— viz. w hether the sentence “ I think” ade­ quately expresses the intended state of affairs o r contains perhaps an hypostasis— and consider the two sentences only from the form allogical point of view. We notice at once two essential logical mis­ takes. T he first lies in the conclusion “ I am .” T he verb “to be” is undoubtedly m eant in the sense of existence here; for a copula can­ not be used w ithout predicate; indeed, D escartes’ “ I am ” has always been interpreted in this sense. But in th at case this sentence violates the above-m entioned logical rule th at existence can be predicated only in conjunction with a predicate, not in conjunction with a nam e (subject, proper n am e). A n existential statem ent does not have the form “a exists” (as in “ I am ,” i.e. “ I exist” ) , but “there exists something of such and such a kind.” T he second e rro r lies in the transition from “I think” to “ I exist.” If from the statem ent “P ( a ) ” ( “a has the property P ” ) an existential statem ent is to be deduced, then the latter can assert existence only with respect to the predicate P, not with respect to the subject a of the prem ise. W hat follows from “ I am a E uropean” is not “ I exist,” but “ a E uropean exists.” W hat follows from “ I think” is not “ I am ” but “ there exists som e­ thing that thinks.” T he circum stance that o ur languages express existence by a verb ( ‘to be” o r “to exist” ) is not in itself a logical fault; it is only inap­ propriate, dangerous. T he verbal form easily misleads us into the m isconception that existence is a predicate. O ne then arrives at such logically incorrect and hence senseless modes of expression as were just exam ined. Likewise such form s as “ Being” o r “ Not-Bcing,” which from time im m em orial have played a great role in m etaphysics, have the same origin. In a logically correct language such form s can­ not even be constructed. It appears th at in the L atin and the G erm an

The Elimination of Metaphysics

[ 75 ]

languages the form s “ens” o r “das Seiende” were, perhaps under the seductive influence of the G reek exam ple, introduced specifically for use by m etaphysicians; in this way the language deteriorated logically w hereas the addition was believed to represent an improvement. A nother very frequent violation of logical syntax is the so-called “type contusion” of concepts. W hile the previously mentioned mistake consists in the predicative use of a symbol with non-predicative m eaning, in this case a predicate is, indeed, used as predicate yet as predicate of a different type. We have here a violation of the rules o f the so-called theory of types. A n artificial example is the sentence we discussed earlier: “C aesar is a prim e num ber.” Names of persons and nam es of num bers belong to different logical types, and so do accordingly predicates of persons (e.g. “general” ) and predicates of num bers ( “prim e num ber” ). The erro r of type confusion is, unlike the previously discussed usage of the verb “to be,” not the preroga­ tive of m etaphysics but already occurs very often in conversational language also. But here it rarely leads to nonsense. The typical am biguity o f words is here of such a kind th at it can be easily removed. Exam ple: 1. ‘T h is table is larger th an th a t.” 2. “T h e height o f this table is larg er than th e height o f th a t table.” H ere th e w ord “larger” is used in ( 1 ) fo r a relation betw een objects, in (2 ) fo r a relation between num ­ bers, hence fo r tw o distinct syntactical categories. T h e m istake is h ere u n im p o rtan t; it could, e.g., be elim inated by w riting “la rg e rl” and “larg er2 ” ; “la rg e r l” is then defined in term s o f “larger2” by declaring statem ent form (1 ) to be synonym ous w ith (2 ) (an d others o f a sim ilar k in d ).

Since the confusion of types causes no harm in conversational language, it is usually ignored entirely. This is, indeed, expedient for the ordinary use of language, but has had unfortunate consequences in metaphysics. H ere the conditioning by everyday language has led to confusions of types which, unlike those in everyday language, are no longer translatable into logically correct form. Pseudo-statements of this kind are encountered in especially large quantity, e.g., in the writings of Hegel and Heidegger. The latter has adopted many p e­ culiarities of the H egelian idiom along with their logical faults (e.g. predicates which should be applied to objects of a certain sort are instead applied to predicates of these objects or to “being” o r to “existence” o r to a relation betw een these objects). H aving found that m any m etaphysical statem ents are meaning­ less, we confront the question w hether there is not perhaps a core of m eaningful statem ents in m etaphysics which would rem ain after elim ination of all the m eaningless ones.

j

j

RUDOLF CARNAP

Indeed, the results we have obtained so far might give rise to the view that there are many dangers of falling into nonsense in metaphysics, and that one must accordingly endeavor to avoid these traps with great care if one wants to do metaphysics. But actually the situation is that meaningful metaphysical statement? are im pos.sible. This follows from the task which metaphysics sets itself :_to discover and formulate a k ind.of knowledge which is not accessible, to empirical science. JVe have seen earlier that die m eaning of a statement lies jn_the method of Its verification.“A statement„ asserts only so In u ch as is verifiable with respect tc> it. Therefore a sentence can be used Uidy to asseTTan empirical proposition, if indeed it is used to assert any­ thing at all. If something were to lie, in principle, beyond possible experiencer -*t“'eould be neither said nor thought nor asked. (Meaningful) statements are divided into the following kinds. First there statements which are true solely by virtue of their form ( ‘tautologies” according to Wittgenstein; they correspond ap­ proximately 1o Kant’s “analytic judgments” ). They say nothing about reality. Tfre formulae oM ogic and mathematics are of this kind. They are not themselves factual statements, but serve for the transformation of jsuch sta te m e n t- Secondly there are the negations of such statements ( “contradictions”). They are self-contradictory, hence false bv virtue of their form. With respect to all other state­ ments TKe decision about truth or falsehood lies in the protocol sen­ tences. They are therefore (true or false) empirical statements and belong to the domain of empirical science. Any statement one desires _to construct which does not fall within these categories becomesautomatically meaningless. Sincejnetaphysics does not want to assert analytic propositions, nor l o fall within the domain of empirical science, i t is compelled to employ words for which no criteria of ^application are specified and which are therefore devoid o f-sen se^ or else to combine meaningful words in such a wav that neither an analytic ( or contradictory) statement nor an e m p iri^ i is produced? In either case pseudo-statements are the inevitable product. Logical analysis, then, pronounces the verdict of meaningless­ ness on any alleged knowledge that pretends to reach above o r behind experience. This verdict hits, in the first place, any speculative m eta­ physics, any alleged knowledge by pure thinking or by pure intuition that pretends to be able to do without experience. But the verdict equally applies to the kind of metaphysics which, starting from ex­ perience, wants to acquire knowledge about that which transcends experience by means of special inferences (e.g. the neo-vitalist thesis

The Elimination of Metaphysics

[ 77 ]

o f the directive presence of an “entelechy” in organic processes, which supposedly cannot be understood in term s of physics; the question concerning the “essence of causality,” transcending the ascertain­ m ent of certain regularities of succession; the talk about the “thing in itself” ). F urther, the sam e judgm ent m ust be passed on all philos­ ophy o f norms, o r philosophy of value, on any ethics or esthetics as a norm ative discipline. F o r the objective validity of a value o r norm is (even on the view of the philosophers of value) not empirically, verifiable n or deducible from em pirical statem ents: Eence“TT cannot txTasserted (in a m eaningful statem ent) at'alT In other words: Either em pirical criteria are indicated for the use of “good” and “ beauti­ ful” and the rest of the predicates th at are employed in the normative sciences, o r they are not. In the first case, a statem ent containing such a predicate turns into a factual judgm ent, but not a value judg­ m ent; in the second case, it becomes a pseudo-statem ent. It is alto­ gether im possible to m ake a statem ent th at expresses a value judgment. Finally, the verdict of meaninglessness also hits those m eta­ physical m ovem ents which are usually called, im properly, epistemological movem ents, th at is realism (insofar as it claims to say more th an the em pirical fact th at the sequence of events exhibits a certain regularity, which m akes the application of the inductive m ethod pos­ sible) and its opponents: subjective idealism, solipsism, phenom enal­ ism , and positivism (in the earlier sense). B ut what, then, is left over for philosophy, if all statem ents w hat­ ever that assert som ething are of an em pirical nature and belong to factual science? W hat rem ains is not statements, nor a theory, nor a system, but only a method: the m ethod of logical analysis. The foregoing discussion has illustrated the negative application of this m ethod: in th at context it serves to eliminate meaningless words, m eaningless pseudo-statem ents. In its positive use it serves to clarify m eaningful concepts and propositions, to lay logical foundations for factual science and for m athem atics. The negative application of the m ethod is necessary and im portant in the present historical situation. But even in its present practice, the positive application is more fertile. W e cannot here discuss it in greater detail. It is the indicated task of logical analysis, inquiry into logical foundations, that is meant by “scientific philosophy ” in contrast to metaphysics. T he question regarding the logical character of the statements which we obtain as the result of a logical analysis, e.g. the state­ m ents occurring in this and other logical papers, can here be an­ swered only tentatively: such statem ents are partly analytic, partly em pirical. F o r these statem ents about statem ents and parts of state­

| 7g j

RUDO LF CARNAP

ments belong in part to pure metalogic (e.g. “ a sequence consisting of the existence-symbol and a noun, is not a sentence ) , in p a rt to descriptive metalogic (e.g. “the word sequence at such and such a place in such and such a book is m eaningless” ). M etalogic will be discussed elsewhere. It will also be shown there th a t the metalogic which speaks about the sentences of a given language can be for­ m ulated in th at very language itself. 7. M e t a p h y s i c s a s E x p r e s s i o n an A t t it u d e t o w a r d L if e

of

O u r claim that the statem ents of m etaphysics are entirely m ean­ ingless, that they do not assert anything, will leave even those who agree intellectually with o ur results with a painful feeling of strange­ ness: how could it be explained th at so m any m en in all ages and nations, am ong them em inent minds, spent so m uch energy, nay veritable fervor, on m etaphysics if the latter consisted of nothing but mere words, nonsensically juxtaposed? A nd how could one ac­ count fo r the fact th at m etaphysical books have exerted such a strong influence on readers up to the present day, if they contained not even errors, but nothing at all? These doubts are justified since m etaphysics does indeed have a content; only it is not theoretical content. T he (pseudo)statem ents of m etaphysics do not serve for the description of states o f affairs, neither existing ones (in th at case they would be true statem ents) nor non-existing ones (in th at case they would be at least false statem ents). They serve for the expression of the gen­ eral attitude of a person towards life ( “ Lebenseinstellung, Lebensgefiihl” ). Perhaps we m ay assume that m etaphysics originated from m yth­ ology. The child is angry at the “ wicked table” which h u rt him. Primitive m an endeavors to conciliate the threatening dem on of earthquakes, o r he worships the deity of the fertile rains in gratitude. Here we confront personifications of natural phenom ena, which are the quasi-poetic expression of m an’s em otional relationship to his environm ent. T he heritage of mythology is bequeathed on the one hand to poetry, which produces and intensifies the effects of m yth­ ology on life in a deliberate way; on the other hand, it is handed dow n to theology, which develops mythology into a system. W hich, now, is the historical role of m etaphysics? Perhaps we may regard it as a substitute for theology on the level of system atic, conceptual th in k ­ ing. T he (supposedly) transcendent sources of knowledge of theol­ ogy are here replaced by natural, yet supposedly trans-em pirical

The Elimination o f Metaphysics

[ 79 ]

sources of knowledge. O n closer inspection the same content as that o f mythology is here still recognizable behind the repeatedly varied dressing: we find th at m etaphysics also arises from the need to give expression to a m an’s attitude in life, his em otional and volitional reaction to the environm ent, to society, to the tasks to which he devotes himself, to the m isfortunes th at befall him. This attitude m anifests itself, unconsciously as a rule, in everything a m an does o r says. It also im presses itself o n his facial features, perhaps even on the character of his gait. M any people, now, feel a desire to create over and above these m anifestations a special expression of th eir attitude, through which it might becom e visible in a m ore suc­ cinct and penetrating way. If they have artistic talent they are' able to express themselves by producing a w ork of art. M any writers have already clarified the way in which the basic attitude is m ani­ fested through the style and m anner of a work of art (e.g. Dilthey and his stu d en ts). [In this connection the term “ world view” ( “Welt­ anschauung” ) is often used; we prefer to avoid it because of its ambiguity, which blurs the difference between attitude and theory, a difference which is of decisive im portance for our analysis.] W hat is here essential for o ur considerations is only the fact that art is an adequate, m etaphysics an inadequate m eans for the expression of the basic attitude. O f course, there need be no intrinsic objection to one’s using any m eans of expression one likes. But in the case of m eta­ physics we find this situation: through the form of its works it pre­ tends to be som ething that it is not. T he form in question is that o f a system of statem ents which are apparently related as premises and conclusions, th at is, the form of a theory. In this way the fiction of theoretical content is generated, whereas, as we have seen, there is no such content. It is not only the reader, but the metaphysician himself who suffers from the illusion that the m etaphysical state­ m ents say something, describe states of affairs. The metaphysician believes that he travels in territory in which truth and falsehood are a t stake. In reality, however, he has not asserted anything, but only expressed something, like an artist. T h at the metaphysician is thus deluding him self cannot be inferred from the fact that he selects language as the m edium of expression and declarative sentences as the form of expression; for lyrical poets do the same w ithout suc­ cum bing to self-delusion. But the m etaphysician supports his state­ m ents by argum ents, he claims assent to their content, he polemicizes against m etaphysicians of divergent persuasion by attem pting to refute their assertions in his treatise. Lyrical poets, on the other hand, do not try to refute in their poem the statem ents in a poem by some

j gQ j

RUDOLF CARNAP

other lyrical poet; for they know they are in the dom ain o f a rt and not in the dom ain of theory. Perhaps music is the purest m eans of expression of the basic attitude because it is entirely free from any reference to objects. T he harm onious feeling or attitude, which the m etaphysician tries to ex­ press in a monistic system, is more clearly expressed in the music of M ozart. A nd when a m etaphysician gives verbal expression to his dualistic-heroic attitude tow ards life in a dualistic system , is it not perhaps because he lacks the ability of a Beethoven to express this attitude in an adequate medium? M etaphysicians are m usicians w ith­ out musical ability. Instead they have a strong inclination to work within the medium of the theoretical, to connect concepts and thoughts. Now, instead of activating, on the one hand, this inclina­ tion in the dom ain of science, and satisfying, on the other hand, the need for expression in art, the m etaphysician confuses the tw o and produces a structure which achieves nothing for knowledge and som e­ thing inadequate for the expression of attitude. O ur conjecture that m etaphysics is a substitute, albeit an inade­ quate one, for art, seems to be further confirm ed by the fact th at the m etaphysician who perhaps had artistic talent to the highest degree, viz. Nietzsche, alm ost entirely avoided the e rro r of th at con­ fusion. A large part of his work has predom inantly em pirical con­ tent. We find there, for instance, historical analyses of specific artistic phenom ena, o r an historical-psychological analysis of morals. In the work, however, in which he expresses m ost strongly that which others express through metaphysics or ethics, in Thus Spake Zarathustra, he does not choose the misleading theoretical form , but openly the form of art, of poetry.

R E M A R K S BY T H E A U T H O R (1 9 5 7 ) To section 1, " m e t a p h y s i c s T his term is used in this p ap er, as usually in E urope, fo r the field o f alleged know ledge o f th e essence o f things w hich transcends the realm o f em pirically fo unded, inductive science. M etaphysics in this sense includes system s like those o f F ichte, Schelling, H egel, Bergson, H eidegger. But it does n o t include endeavors tow ards a synthesis and generalization o f th e results o f th e various sciences. To section 1, “meaning.” T o d ay we distinguish various k in d s o f m eaning, in particu lar cognitive (designative, referen tial) m eaning on the one h and, and non-cognitive (expressive) m eaning com ponents, e.g. em otive and m otivative, o n th e o ther. In the presen t p ap er, th e w ord

The Elimination o f Metaphysics

[ 81 ]

“m eaning” is alw ays understood in the sense o f “cognitive m eaning.” T h e thesis th at the sentences o f m etaphysics are meaningless, is thus to be understood in the sense that they have no cognitive m eaning, no asser­ tive content. T he obvious psychological fact that they have expressive m eaning is thereby not denied; this is explicitly stated in Section 7. To section 6, "m etalogic" This term refers to the theory o f expres­ sions o f a language and, in particular, o f their logical relations. T oday we w ould distinguish between logical syntax as the theory o f purely form al relations and sem antics as the theory o f m eaning and truthconditions. To section 6, realism and idealism. T h at both the affirmative and the negative theses concerning the reality o f the external world are pseudo­ statem ents, I have tried to show in the m onograph Scheinprobleme in der Philosophic: Das Fremdpsychische und der Realismusstreit, Berlin, 1928. T he sim ilar natu re o f the ontological theses about the reality o r un reality o f abstract entities, e.g., properties, relations, propositions, is discussed in “Em piricism , Sem antics, and O ntology,” Revue Intern, de Philos. 4, 1950, 20-40, reprinted in: Meaning and Necessity, second edi­ tion, C hicago, 1956.

4

Positivism and Realism by

M O R IT Z S C H L IC K

(T R A N S L A T E D BY DAVID R Y N IN )

1. P

r e l im in a r y

Q

u e s t io n s

is defined by the principles which it considers fundam ental and to which it constantly recurs in argument. But in the course of the historical developm ent of such a view these principles tend to change— w hether they be reform ulated, extended, o r restricted, o r even gradually altered in meaning. A t some time then the question arises w hether we should still speak of the developm ent of the one point of view at all, and retain its old nam e; o r whether a new viewpoint has arisen. If, along with the developed view there also exists an “orthodox” viewpoint which retains the first principles in their original form and meaning, sooner o r later some term inological distinction betw een the old and the new will arise autom atically. But where this is not clearly the case, where rather the different exponents of a “viewpoint” em ­ ploy extremely different, even contradictory, form ulations and m ean­ ings of the principles, confusion arises; the adherents and opponents of the view talk at cross purposes; each selects those statem ents which can be used in defense of his own opinions, and all ends in fatal m isunderstanding and obscurity. These confusions disappear only when the different principles are distinguished, and each is tested sep ­ arately for its meaning and truth. In such an exam ination of p rin ­ ciples one quite ignores, for the time, the question of the historical contexts of their origins, and of their names. I should like to apply these considerations to the m odes o f thinkE

very

p h il o s o p h ic a l

p o in t

of

v ie w

T h is article first a p p e a re d in Erkenntnis, V o lu m e I I I ( 1 9 3 2 /3 3 ) a n d is h e re re­ p u b lish ed w ith th e k in d p erm issio n o f M rs. S ch lic k a n d P ro fe sso r C a rn a p . I t w as w ritte n in resp o n se to criticism s o f positivism in a le c tu re by M a x P la n c k e n title d “ P o sitivism us u n d R e alc A u ssen w elt,” p u b lish e d in 1931 by th e A k a d e m isc h e V erlag sgcsellschaft, Leipzig.

[ 82]

Positivism and Realism

[ 83 ]

ing grouped together under the nam e “ positivism.” They have, from the rime A ugust Com te invented the term until the present, under­ gone a developm ent which furnishes a good exam ple of w hat has just been said. But I do no t do this with the historical aim, say, of determ ining a strict concept of positivism as it has appeared in his­ tory, but rath er in order to contribute something to a positive settle­ m ent of the dispute carried on nowadays concerning certain prin­ ciples which pass as fundam ental to positivism. Such a settlement concerns me the m ore because I myself advocate some of these principles. I am concerned here only to m ake their meaning as clear as possible; w hether o r not one will, after this clarification, attribute them to “positivism ” is a question of very little im portance. If one wishes to characterize every view which denies the possi­ bility of m etaphysics as positivistic this is quite unobjectionable, as a m ere definition; and I should in this sense call myself a strict positivist. But this holds, of course, only under the presupposition o f a special definition of “ m etaphysics.” W hat the definition of m eta­ physics is which m ust be adopted here need not interest us at the m om ent; b u t it hardly agrees with the form ulations usual in philo­ sophic literature, and fu rth er determ inations of positivism which refer to such form ulations lead at once into confusions and difficulties. If we say, as frequently has been said, that metaphysics is the the­ ory of “true being,” of “ reality in itself,” of “transcendent being” this obviously im plies a (contradictory) spurious, lesser, apparent being; as has indeed been assum ed by all m etaphysicians since the tim e of P lato and the Eleatics. This apparent being is the realm of “appear­ ances,” and while the true transcendent reality is to be reached only w ith difficulty, by the efforts of the m etaphysician, the special sciences have to do exclusively with appearances which are perfectly accessible to them . T he difference between the ways in which these tw o “ m odes of being” are to be known, is then explained by the fact th at the appearances are immediately present, “given,” to us, while m etaphysical reality m ust be inferred from them in some roundabout m anner. A nd thus we seem to arrive at a fundam ental concept of the positivists, for they always speak of the “given,” and usually form ulate their fundam ental principle in the proposition that the philosopher as well as the scientist m ust always rem ain within the given, th at to go beyond it, as the m etaphysician attem pts, is im possible o r senseless. T hus it am ounts to identifying the “given” of the positivist with the “ appearances” of m etaphysics, and to believing that positivism is at bottom a m etaphysics, from which one has left, o r stricken,

[ 84 ]

MORITZ SCHLICK

out the transcendent; and such an opinion m ay indeed often enough have inspired the argum ents of the positivists, as well as those of their opponents. But this belief finds us well on o u r way to dangerous errors. T he term “the given” itself is a cause of grave m isunderstandings. “T o give” usually connotes a three term relation: it presupposes first, someone who gives, secondly, one to whom is given, and thirdly, something given. T he m etaphysician finds this quite in order, for what gives is the transcendent reality, w hat receives is the know ing mind, which m akes w hat is given to it into its “content.” B ut evi­ dently the positivist will from the very outset have nothing to do with such notions; the given is for him but a w ord for w hat is m ost simple and no longer questionable. N o m atter w hat w ord we choose, every one will be capable of m isinterpretations; if we speak of “experiences” we seem to presuppose the distinction betw een w hat experiences and w hat is experienced; with the use of the phrase “con­ tent of consciousness” we seem burdened with a sim ilar distinction, and in addition with the com plicated concept of “consciousness,” which in any case did not exist until invented by philosophy. B ut even ap art from such difficulties it is perhaps still not clear w hat is actually m eant by the given. D o only such “qualities” as “blue,” “w arm ,” “ pain,” com e under this heading, o r e.g. also relations betw een them , o r their order? Is the sim ilarity o f two qualities “given” in the sam e sense as the qualities themselves? A nd if the given is somehow w orked u p o r interpreted o r judged is this working-up o r judging no t also in some sense som ething given? But it is not obscurities of this sort that give rise to the current m atter of dispute: the bone of contention appears am ong the various parties only with the question of “ reality.” If the rejection o f m etaphysics by positivism signifies the denial of transcendent reality then it seems the m ost natural conclusion in the w orld that the positivist attributes reality only to non-transcendent being. T h e fundam ental principle of the positivist then seems to run: “Only the given is real.” If one enjoys w ord-play one can lend to this proposition the sem blance of tautological self-evidence by m aking use of a peculiarity of the G erm an language in thus form u­ lating it: “ Es gibt n ur das G egebene.” (T here is only the given.) W hat shall we m ake of this proposition? M any positivists m ay have expressed and advocated it (especially, perhaps, those w ho represented physical objects as “ m ere logical constructions,” o r “m ere auxiliary concepts” ), while this view has been attributed to others by their opponents. W e m ust insist, however, th at w hoever

Positivism and Realism

[ 85 ]

states this proposition seeks to establish an assertion which is m etaphysical in exactly the sam e sense and degree as its apparent contradictory: “T here is a transcendent reality.” T he problem about which the m atter revolves here is evidently the so-called problem of the reality of the external world, and there seem to be two parties: that of “ realism” which believes in the reality of the external world, and th at of “ positivism” which does not. In truth, I am convinced th at it is quite senseless to set two views in opposition in this m anner, for neither party really knows w hat it w ants to say (w hich is the case with every metaphysical proposition.) But before I explain this I should like to show how the m ore obvious interpretations of the proposition “only the given is real” actually lead at once to well-known m etaphysical views. This problem can take the form of the question about the existence of the “external” world only if somehow we can distinguish between inner and outer; and this distinction is m ade by considering the given as a “content” of consciousness, as belonging to one or several subjects to whom it is given. Thus the im m ediate datum would have attributed to it some sort of m ental character, the character of a representation o r an idea; and the proposition would then state that this character pertained to all reality: no being outside of conscious­ ness. But this is nothing but the fundam ental principle of metaphysical idealism. If the philosopher thinks himself able to speak only of w hat is given to him self we have before us a solipsistic metaphysics, but if he thinks he may assum e that the given is distributed among m any subjects we have a metaphysics of the Berkeleyan variety. O n this interpretation positivism would be simply identical with the older idealistic metaphysics. But since its founders certainly desired som ething quite different from a renewal of that idealism, this interpretation is to be rejected as contrary to the anti-metaphysical attitude of positivism. Idealism and positivism are incompatible. T he positivist E rnst Laas has written a work of several volumes to dem onstrate the irreconcilable opposition which exists on all points between them ; and if his student H ans V aihinger gave to his “Philos­ ophy of A s I f ’ the subtitle an “ idealistic positivism” it is but one of the contradictions from which this w ork suffers. E rnst M ach es­ pecially em phasized that his own positivism developed in an opposite direction to that of Berkeleyan m etaphysics; and he and Avenarius laid great stress upon not taking the given as a content of conscious­ ness. They tried to exclude this concept from their philosophy altogether. In view of the uncertainty in the cam p of the positivists them ­

J

86

]

MORITZ SCHLICK

selves it is no wonder th at the “ realist” fails to observe the distinctions we have discussed, and directs his argum ents against the thesis: “T here is nothing but the contents of consciousness,” o r “T here is only an internal w orld.” But this proposition belongs to idealistic m eta­ physics, and has no place in an anti-m etaphysical positivism, which is not affected by these realistic argum ents. O f course the realist can think that it is simply inevitable to conceive the given as contents of consciousness, as subjective, as m ental— or w hatever expression is used; and he will then consider as a failure the attem pt of M ach and A venarius to take the given as neutral and to resolve the distinction between inner and outer, and will believe that a view free of any m etaphysical basis is im ­ possible. But this line of thought is rarely m et with. A nd how ever it may fare, in any case, the whole business is m uch ado about nothing, fo r the “ problem of the reality of the external w orld” is a meaningless pseudo-problem . This must now be m ade evident. 2. O n t h e M e a n i n g o f P r o p o s i t i o n s

It is the peculiar business of philosophy to ascertain and m ake clear the meaning of statem ents and questions. T he chaotic state in which philosophy has found itself during the greater p a rt of its history is due to the unfortunate fact that, in the first place, it took certain form ulations to be real questions before carefully ascertaining w hether they really m ade any sense, and, in the second place, it believed that the answers to the questions could be found by the aid of special philosophical m ethods, different from those of the special sciences. But we cannot by philosophical analysis decide w hether anything is real, bu t only w hat it means to say th at it is real; and whether this is then the case o r not can be decided only by the usual m ethods of daily life and of science, th at is, through experience. H ence we have here the task of m aking clear to ourselves w hether any m ean­ ing can be attached to the problem of the reality of the “external world.” W hen, in general, are we sure th at the m eaning of a question is clear to us? Evidently when and only when we are able to state exactly the conditions under which it is to be answered in the affirmative, or, as the case may be, the conditions under which it is to be answered in the negative. By stating these conditions, and by this alone, is the m eaning of a question defined. It is the first step of any philosophizing, and the foundation of all reflection, to see that it is simply impossible to give the m eaning of any statem ent except by describing the fact which m ust exist if

Positivism and Realism

[ 87 ]

the statem ent is to be true. If it does not exist then the statem ent is false. T h e m eaning of a proposition consists, obviously, in this alone, th a t it expresses a definite state of affairs. A nd this state of affairs m ust be pointed o u t in order to give the meaning of the proposition. O ne can, of course, say that the proposition itself already gives this state of affairs. This is true, but the proposition indicates the state of affairs only to the person who understands it. B u t when do I understand a proposition? W hen I understand the m eanings of the w ords which occur in it? These can be explained by definitions. But in the definitions new words appear whose meanings can n o t again be described in propositions, they m ust be indicated directly: the m eaning o f a word m ust in the end be shown, it m ust be given. This is done by an act of indication, of pointing; and w hat is pointed at m ust be given, otherw ise 1 cannot be referred to it. A ccordingly, in order to find the meaning of a proposition, we m ust transform it by successive definitions until finally only such w ords occur in it as can no longer be defined, but whose meanings can only be directly pointed out. T he criterion of the truth o r falsity o f the proposition then lies in the fact th at under definite conditions (given in the definition) certain data are present, o r not present. If this is determ ined then everything asserted by the proposition is determ ined, and I know its meaning. If I am unable, in principle, to verify a proposition, th at is, if I am absolutely ignorant of how to proceed, of w hat I m ust do in order to ascertain its truth o r falsity, then obviously I do not know w hat the proposition actually states, and I should then be unable to interpret the proposition by passing from the words, with the aid of the definitions, to possible experiences. F o r in so far as I am able to do this I am also able in the same way to state at least in principle the m ethod of verification (even though, often, because of practical difficulties I am unable to carry it o u t). The statem ent of the conditions under which a proposition is true is the same as the statem ent of its m eaning, and not something different. A nd these “conditions,” we have already seen, must finally be discoverable in the given. Different conditions m ean differences in the given. T h e meaning of every proposition is finally to be determ ined by the given, and by nothing else. I do not know if this insight ought to be called positivistic; but o f course I should like to believe th at it underlay all those efforts which ap p ear by this nam e in the history of philosophy, whether or not it was ever clearly form ulated. We may indeed assume that it constitutes the real nucleus and motive force of m any quite perverted form ulations which we find among positivists.

[ gg ]

M O R IT Z SCHLICK

If we but once attain the insight that the m eaning of every p ro p ­ osition can be determ ined only by m eans of the given we can no longer conceive the possibility of another opinion, for we see that we have discovered simply the conditions under which opinions in general can be form ulated. H ence it would be quite m istaken to see, somehow, in w hat we have said a “theory of m eaning” (in A ngloSaxon countries this insight, that the m eaning of a proposition is determ ined wholly and alone by its verification in the given, is often called the “experim ental theory of m eaning” ). W hat precedes every form ulation of a theory cannot itself be a theory. T he content of o ur insight is indeed quite simple (an d this is the reason why it is so sensible). It says: a proposition has a statable meaning only if it m akes a verifiable difference w hether it is true o r false. A proposition which is such th at the w orld rem ains the sam e whether it be true o r false simply says nothing about the w orld; it is em pty and com m unicates nothing; I can give it no meaning. We have a verifiable difference, however, only w hen it is a difference in the given, for verifiable certainly m eans nothing but “capable of being exhibited in the given.” It is obvious th at verifiability is used here in the sense of “verifi­ able in principle,” fo r the m eaning of a proposition is, of course, independent of w hether the conditions under which we find ourselves at a specified time allow o r prevent the actual verification. T here is not the least doubt that the proposition “there is a m ountain of a height of 3000 meters on the other side of the m oon” m akes good sense, even though we lack the technical m eans of verifying it. A nd it would rem ain just as meaningful if one knew with certainty, on scientific grounds, that no m an would ever reach the other side o f the moon. T he verification rem ains conceivable; we are always able to state w hat data we should have to experience in order to decide the truth o r falsity of the proposition; the verification is logically possible, whatever be the case regarding its practical feasibility, and this alone concerns us. But if someone should say: within every electron there is a nucleus, which, though always present, never has in any way any external effects, so th at its existence never m anifests itself in nature— this would be a meaningless assertion. F o r we should have to ask the m aker of the hypothesis: w hat do you really mean by the presence of that “ nucleus”?; and he could answ er only: I m ean that something exists there in the electron. W e should inquire further: w hat does that mean? W hat would be the case if it didn’t exist? A nd he would have to answer: everything would rem ain exactly the sam e as before.

Positivism and Realism

[ 89 ]

F o r according to his assertion, the “som ew hat” in the electron has n o effects, and there would simply be no observable change: the realm of the given would not be affected in any way. We should judge th at he had not succeeded in com m unicating the meaning of his hypothesis, and th at therefore it m ade no sense. In this case the impossibility of verification is not factual, but logical, for by reason of the u tter ineffectiveness of that nucleus a decision regarding it based on differences in the given is in principle excluded. One cannot here suppose that the distinction between the im ­ possibility of verifying som ething in principle and the mere factual, em pirical impossibility is not clear, and is therefore sometimes difficult to draw ; fo r the impossibility in principle is logical im pos­ sibility which does not differ in degree from em pirical impossibility, but in very essence. W hat is em pirically impossible still remains conceivable, but w hat is logically impossible is contradictory, and cannot therefore be thought at all. As a m atter of fact we find that in scientific thinking this distinction is always clearly and instinctively felt. T he physicists were the first to reject the statem ent given in our exam ple regarding the forever hidden nucleus of the electron, with the criticism th at it was no hypothesis at all, but mere em pty word play. A nd in all tim es the most successful scientific investigators have adopted this standpoint with respect to the meaning of their statem ents, since they have acted in accordance with it, even if for the m ost part unconsciously. F o r science, then, o u r standpoint does not represent something foreign and out of the ordinary, but it has in a certain sense always been m ore o r less taken for granted. It could not be otherwise, because only from this standpoint is a proposition verifiable at all; and since all the activities of science consist in exam ining the truth of propo­ sitions, it continuously acknowledges the correctness of our insight by its practice. If express confirm ation were still necessary, it would be found m ost conspicuously at critical points in the developm ent of sci­ ence where investigation is forced to bring the self-evident pre­ suppositions to light. This is the case where difficulties of principle lead one to suppose that som ething may be wrong with these pre­ suppositions. The m ost fam ous exam ple of this sort, which will rem ain forever m em orable, is Einstein’s analysis of the concept of tim e, which consists in nothing but the analysis of the meaning of our statem ents about the sim ultaneity of spatially separate events. Einstein said to the physicists (an d to the philosophers): you must first state w hat you mean by sim ultaneity, and you can do this only

| 90 ]

M ORITZ SCHLICK

by showing how the proposition “two events are sim ultaneous” is verified. But with this you have completely determ ined its m eaning. W hat is true of the concept of sim ultaneity holds of every other concept: every proposition has m eaning only in so far as it can be verified, and it says only w hat is verified, and simply nothing m ore. If one should say th at it did contain something m ore he m ust be able to say w hat m ore this is, and to do this he would have to tell us how the world would differ if he were m istaken. But this cannot be done, since by assum ption all the observable differences are already in­ cluded in the verification. In the exam ple of sim ultaneity the analysis of the m eaning, as is appropriate for the physicist, is carried only to the point where the de­ cision regarding the truth o r falsity of a proposition about tim e is based on the occurrence o r non-occurrence of a definite physical event (e.g. the coincidence of a pointer with a point on a scale). But it is clear that one can ask further: w hat does it mean to say that the pointer indicates a definite point on the scale? A nd the answer can only be m ade by reference to the occurrence of certain data, o r as one gen­ erally says, certain “sense-im pressions.” This will be generally ad ­ m itted, especially by physicists. “ F o r positivism will always be right in this, th at there is no other source of knowledge than sense-im pres­ sions” says Planck1 and this evidently m eans that the truth or falsity of a physical statem ent depends entirely upon the occurrence of certain sense-im pressions (w hich constitute a special class of d a ta ). But there will always be m any who are inclined to say: granted that the truth of a physical statem ent can be tested only by the occur­ rence of certain sense-im pressions, this is not the same as asserting that the meaning of the statem ent is also thereby exhaustively given. This latter m ust be denied: a proposition can contain m ore th an can be verified; th at the pointer stands at a definite point on the scale m eans more than the existence of certain sensations (nam ely “ the existence of a definite fact in the external w orld” ). In answer to this denial of the identity of m eaning and verifica­ tion we m ust point out the following: 1) This denial is found am ong physicists only when they leave the actual sphere of physical state­ ments and begin to philosophize. (In physics, obviously, there occur only statem ents about the properties o r behavior of things o r events, an express statem ent concerning their “ reality” is not a scientific statem ent but a philosophical o n e ). In his own sphere the physicist admits entirely the correctness of o ur standpoint. We m entioned this earlier, and illustrated it in the exam ple of sim ultaneity. T here are 1. Positivismus und Reale Aussenwelt, p. 14.

Positivism and Realism

[ 91 ]

indeed m any philosophers who say: of course we can determ ine only relative sim ultaneity, but it does not follow from this that there is no such thing as absolute sim ultaneity, and we continue to believe in it! T h e falsity of this statem ent cannot in any sense be dem on­ strated, but the overwhelm ing m ajority of physicists is rightly of the opinion that it is meaningless. However it must be sharply em pha­ sized th at in both cases we have to do with the same situation: There is in principle no difference w hether I ask: does the proposition “two events are sim ultaneous” m ean more than can be verified? O r whether I ask: does the proposition “the pointer points toward the fifth line on the scale” m ean m ore than can be verified? T he physicist who handles these two cases differently is guilty of an inconsistency. He will o f course justify himself, believing that in the second case where the question concerns the “ reality of the external world” m uch more is at stake, philosophically. This argum ent is too vague for us to attach m uch weight to it, but we shall see presently whether anything lies behind it. 2) It is perfectly true that every statem ent about a physical object o r an event means m ore than is verified, say, by the occurrence of a single experience. It is rather presupposed that the experience occurred under very definite conditions, whose realization of course can only be verified by something given, and it is presupposed further that ever m ore verifications are possible (confirm ations e tc .), which in their turn, naturally, reduce to certain given events. In this m anner one can and m ust give an account of illusions of sense, and of error, and it is easy to see how those cases are to be included in which we should say the observer was merely dream ing, that the pointer in­ dicated a definite line, o r th at he did not carefully observe, etc. The assertions of Blondlot about N -Rays which he believed himself to have discovered were certainly more than statem ents that under certain conditions he had experienced certain visual sensations; and because of this, of course, they could be refuted.2 Strictly speaking, the m eaning of a proposition about physical objects would be ex­ hausted only by an indefinitely large num ber of possible verifications, and we gather from this that such a proposition can in the last analysis never be shown to be absolutely true. It is indeed generally recognized that even the most certain propositions of science are always to be taken as hypotheses, which rem ain open to further refinem ent and im provem ent. This has certain consequences for the logical nature of such propositions, but these do not interest us here. O nce again: the m eaning of a physical statem ent is never dc2. C f. P la n c k , op. cit., p. 11.

j 92 ]

MORITZ SCHLICK

term ined by a single isolated verification, but it m ust be thought of in the form : If conditions x are given, the data y occur, where we can substitute an indefinitely large num ber of conditions for x, the proposition rem aining true for each case. (T his holds even when the statem ent refers to a single happening— a historical event, for such an event has innum erable consequences whose occurrences are veri­ fiable). Thus the m eaning of every physical statem ent is lodged finally in an endless concatenation of data; the isolated datum therefore is here uninteresting. Hence if any positivist ever said th at the only objects of science are the given experiences themselves he was certainly quite m istaken; w hat alone the scientists seek are the rules which govern the connections am ong experiences, and by m eans of which they can be predicted. N o one will deny th at the sole verifica­ tion of natural laws lies in the fact that they yield such true predic­ tions. The com m on objection that the imm ediately given, which at most can be but the object of psychology, is thus falsely m ade into the object of physics is in this way refuted. 3) M ost im portant however: if anyone is of the opinion that the meaning of a proposition is nevertheless not exhausted by w hat can be verified in the given, but extends far beyond it, he m ust at least adm it th at this additional m eaning cannot in any way be described, stated, o r expressed in language. F o r let him try to com ­ m unicate this additional meaning! T o the extent to which he succeeds in communicating som ething about this additional m eaning he will find th at the com m unication consists in the fact that he has indicated certain conditions which can serve for verification in the given, and thus he finds o u r position confirmed. O r else he believes him self to have given a meaning, but closer exam ination shows th a t his w ords express only that something m ore is there, concerning whose nature simply nothing is said. A nd then in fact he has com m unicated nothing, and his assertion is meaningless. F o r one cannot assert the existence of something w ithout saying what one asserts to exist. T his is ob­ vious in the case of our exam ple of the “nucleus of the electron” which in principle lies beyond experience; yet for clarity’s sake we shall con­ sider another exam ple which brings out an im portant point of p rin ­ ciple. I observe two pieces of green paper and determ ine th at they have the same color. T he proposition which asserts the sameness of color is verified, am ong other ways, by the fact th at at the same time I have two experiences of the same color. T he proposition: “there are two spots of the sam e color before me now” cannot be reduced to any others; it is verified by the fact that it describes the

Positivism and Realism

[ 93 ]

given. It has a clear m eaning: by virtue of the meanings of the words involved in the proposition, it signifies just the existence of color sam eness; and by virtue of linguistic usage the proposition expresses just th a t experience. Now I show one of these two pieces of paper to a second observer, and ask the question: does he see the green as I do? Is his color experience like my color experience? This case differs in principle from that just considered. While there the state­ m ent was verifiable by the experience of color sameness, here, brief reflection shows, such a verification is simply impossible. Of course the second observer, if he is not color blind, calls the paper green, and if I describe this green to him more closely by saying: it is yellower than this carpet, but bluer than the billiard cloth, darker th an this plant, etc., he will find the same to hold in his experience, i.e. he will agree with my statem ents. But even if all his judgments about color agree entirely with mine I cannot infer from this that he experiences this same quality. It could be the case that on looking at the green paper he would have a color experience which I would call “ red,” that on the other hand, when I see red he would see green, calling it “red ” of course, and so on. Indeed it m ight even be that my color sensations correspond to his tone experiences, o r to any oth er data. It would nevertheless forever be impossible to discover these differences, betw een his and my experience. We should always understand one another perfectly, and could never be of different opinions regarding o ur environm ent if (and this is the only assumption th at need be m ade) the inner order of his experiences agreed with th at of mine. T here is no question here of their “ quality,” all that is required is that they can be arranged into systems in the same m anner. A ll this is indeed adm itted, and philosophers have often pointed it out. But, fo r the m ost part, while they have allowed that such sub­ jective differences are theoretically possible, and that this possibility raises a very interesting question of principle, they have held it to be “ highly probable” th at the other observer and I do in fact have the same experience. But, we must point out, the statem ent that different individuals have the sam e experience has its sole verifiable meaning in the fact th at all their assertions (and of course all the rest of their behavior) exhibit certain agreem ents. Hence it follows that the state­ m ent means nothing but this. It is only to express the same thing in a different m anner if we say th at we here are concerned with the simi­ larity of two system -orders. The proposition that two experiences of different subjects not only occupy the same place in the order of a sys­ tem but are, in addition, qualitatively sim ilar has no meaning for us. N ote well, it is not false, but meaningless: we have no idea what it m eans.

j

94

j

MORITZ SCHLICK

Experience shows, however, th at m ost people find it very difficult to agree to this. We must m ake it clear th at here we have to do with a logical impossibility of verification. It m akes good sense to speak of the similarity of data in the same consciousness, for it can be verified through an im m ediate experience. But if we w ant to speak of the similarity of data in different consciousnesses we are dealing with a new concept, which has to be newly defined. F o r the statem ents in which it occurs are no longer verifiable in the old m anner. T he new definition is simply the sim ilarity of all relevant reactions o f the two individuals; we can find no other. M ost people, of course, believe that no definition is required here; one knows the m eaning of “sim ilar” without it, and the m eaning in both cases is the same. But, to recognize this as a m istake we need only rem em ber the concept of sim ultaneity, in which the situation is exactly the sam e. T o the concept of “sim ultaneity at a place” there corresponds the concept of “similarity of the experiences of the sam e individual,” and to “ sim ultaneity at different places” there corresponds the notion of “similarity of the experiences of different persons.” T he second notion is, with respect to the first, a new concept in each case, and must be specially defined. We can no m ore indicate a directly experiencable quality which would verify the sim ilarity of tw o greens in different consciousnesses than we can for sim ultaneity at different points: both m ust be determ ined by a system of relations. M any philosophers have sought to overcom e the difficulty which seem ed to confront them here by all sorts of speculations and ideal experim ents, speaking, say, of a universal consciousness com prehend­ ing all individuals (G o d ) or thinking perhaps that by m eans o f some artificial connection o f the nervous systems of two individuals the sensations of one would be m ade accessible to the other, and thus be rendered com parable. But of course all this is in vain. F o r even in this fantastic way in the end only the contents of one and the same consciousness would be directly com pared. T he question, however, concerns the possibility of the com parison of qualities in so fa r as they belong to different, and not the same, consciousnesses. Hence it m ust be granted that a statem ent concerning the sim i­ larity of the experiences of two persons has no other communicable meaning than a certain agreem ent of their reactions. O f course everyone is free to believe that such a proposition also possesses another m ore direct m eaning; but so m uch is sure: no such m eaning is verifiable, and one cannot in any way state o r show w hat this meaning is. Hence it follows th at such a m eaning simply cannot in any way becom e the object of discussion. We can say absolutely

Positivism and Realism

[ 95 ]

nothing about it, and it can in no way enter into any language by m eans of which we com m unicate with one another. And what has, I hope, become clear here holds generally. We can understand in a proposition only what it com m unicates, and a meaning is communi­ cable only if it is verifiable. Since propositions are nothing but vehicles for com m unication we can include in their meanings only what they can com m unicate. F o r this reason I should maintain that “meaning” can m ean only “verifiable m eaning.” But even if someone should insist that there is a non-verifiable m eaning this would not help in the least. F or such a meaning can in no way enter into anything he says or asks, or into what we ask him o r answ er him. In other words: if there were any such thing, all o u r utterances, argum ents, and modes of behavior would remain quite unaffected by it, whether we were dealing with daily life, ethical o r aesthetic attitudes, with science or philosophy. Everything would rem ain as if there were no unverifiable meaning. F or if there were a difference this very difference would m ake it verifiable. This is a serious situation, and we must insist that it be taken seriously. Above all one m ust guard against confusing this logical impossibility with an em pirical incapacity, as if some technical diffi­ culty and hum an im perfection were responsible for the fact that only w hat is verifiable can be expressed, and as if there were still some rear entrance through which an unverifiable meaning might slip in and m ake itself evident in o ur discourse and behavior. No! The incomm unicability is absolute; he who believes (o r rather imagines that he believes) in a non-verifiable m eaning must nevertheless admit that with respect to it only one course is open to him : utter silence. N either he nor we gain anything, no m atter how often he asserts: “yet there is a non-verifiable m eaning!” F o r this statem ent itself is devoid o f m eaning, it tells us nothing. 3 . W h a t is t h e M e a n i n g o f “ E xternal W orld” ?

of

“R

e a l i t y ,”

We are now prepared to apply what has been said to the so-called problem of the reality of the external world. We ask: W hat is the meaning of the realist’s assertion, “there is an external world?” o r w hat is the meaning of the statement (attributed to the positivist by the realist) “there is no external w orld”? In order to answer the question it is of course necessary to m ake clear the meanings of the words, “ there is,” and “external w orld.” We begin with the first. “T here is an x” means the same as “x is

[ 96 ]

MORITZ SCHLICK

real” or “x is actual.” H ence w hat do we m ean when we attribute reality to an object? It is an old, very im portant logical o r philosophi­ cal insight, that the proposition “x is real” is of quite a different sort from a proposition which ascribes some property to x (e.g. “x is hard .” ) In other words: reality or existence is not a predicate. T he statem ent “the dollar in my pocket is round” has a com pletely different logical form from th at of the statem ent “ the dollar in my pocket is real.” In m odem logic this distinction is expressed by m eans of two very different symbolisms, but it was already clearly draw n by K ant, who, as we know, in his critique of the so-called ontological proof of G od’s existence, correctly found the source of error of this proof in the fact that existence is treated as a predicate. In daily life we constantly speak of reality or existence, and for this reason it cannot be very difficult to discover its meaning. In a law-suit it is often necessary to determ ine w hether a certain docu­ m ent actually exists, or w hether it is merely wrongly asserted to exist; and it is not altogether unim portant to me w hether the dollar in my pocket is only imagined o r is real. Now everyone knows how such an assertion of the reality of som ething is verified, and there cannot be the slightest doubt that the reality of the dollar is verified and verified only by the fact that, as a result of certain suitable manipulations, I obtain certain sensations of touch and sight upon whose presence I am accustom ed to say “this is a dollar.” T he sam e holds of the docum ent, except that in this case we would content ourselves with certain statem ents of others who claim to have seen the docum ent, i.e. to have had perceptions of a very definite sort. A nd the “statem ents of others” consist again of certain acoustic, or, if they were written statem ents, of certain visual perceptions. N o special analysis is required of the fact that the occurrence of certain sense-perceptions am ong the data always constitutes the sole criterion of statem ents concerning the reality of a “physical” object o r event in everyday life, as well as in the m ost subtle propositions of science. T hat there are okapis in A frica can be determ ined only by the fact that such anim als are observed there. H ow ever it is not necessary th at the object o r event “itse lf’ be perceived. W e can, for exam ple, imagine the existence of a transneptunian planet to be inferred with as much certainty from the observation of perturbations as from the direct perception of a spot of light in the telescope. T he reality of atom s furnishes us with another exam ple. A nd the sam e is true of the other side of the moon. It is of great im portance to realize th at the occurrence o f a definite single experience in the verification of a proposition about

Positivism and Realism

[ 97 J

nature is often not accepted as verifying the proposition, but that throughout we are concerned with uniform ities, with connections obeying natural laws: in this m anner genuine verifications are dis­ tinguished from illusions and hallucinations. W hen we say of any object o r event— which m ust be designated by a description— that it is real this m eans th at there exists a very definite connection betw een perceptions o r other experiences, that under certain condi­ tions certain data appear. Such a statem ent is verified in this m anner alone, and therefore it has only this com m unicable meaning. This was in principle already form ulated by K ant, whom no one would charge with “ positivism.” Reality for him is a category, and if we apply it in any way, and say of an object that it is real this m eans, according to K ant, th at it belongs to a collection of percep­ tions connected according to some natural law. W e see that for us (as for K ant; and the same applies to every philosopher who understands his business) it is simply a m atter of saying w hat it m eans in everyday life o r in science to ascribe real existence to a thing. O ur task is in no sense that of correcting the statem ents of everyday life o r of science. I must confess that I should repudiate and consider absurd any philosophical system that in­ volved the assertion th at clouds and stars, m ountains and sea were unreal, th at the chair by the wall ceased to exist whenever I turned my back. N o r do I credit any serious thinker with any such statem ent. It would fo r exam ple surely be quite a perverse interpretation of Berkeley’s philosophy to see in it such a system. He too didn’t deny the reality of the world of bodies, but merely tried to explain what we m ean when we ascribe reality to it. H e who says that unperceived ideas exist in G od’s mind does not thereby deny their existence but seeks to understand it. John Stuart Mill himself did not wish to deny the reality of physical bodies, but to clarify it, when he declared them to be “ perm anent possibilities of sensation,” though in my opinion his m anner of expression was very ill chosen. T herefore if one understands by “positivism” a view which denies the reality of bodies I m ust declare positivism to be simply absurd. But I d o not believe that such an interpretation of positivistic views would be historically just, at least so far as their ablest representatives are concerned. Be this as it may, we are not concerned with it, but with the view itself. A nd in this connection we have seen that our principle, th at the m eaning of a proposition is identical with its veri­ fication, leads to the insight th at the assertion of the reality of a thing is a statem ent regarding a regular connection of experiences. It does not lead to the conclusion th at the assertion is false. (T here­

j

9

g j

M ORITZ SCHLICK

fore reality is not denied to physical things in favor of sensations.) But opponents of the view just expounded are not at all satisfied with this. They would probably offer the following answ er: Y ou do, indeed, adm it the reality of the physical world, but, as it seems to us, only verbally. You simply call th at real which we would describe as a mere conceptual construction. W hen we use the word reality we m ean by it something quite different from w hat you m ean. Y o u r definition of reality refers back to experiences; but we m ean som e­ thing altogether independent of experience. W e m ean som ething which possesses that same independence evidently attributed by you to the data alone, in the sense that you reduce everything to them as to something not further reducible.” Even though it would suffice as a rebuttal to invite o u r opponents to reconsider how existential propositions are verified, and how verification and m eaning are connected, I think it necessary to consider the psychological attitude from which this argum ent arises, and to request attention for the following rem arks, which m ay result in a modification of that attitude. We ask first w hether on o u r view a reality is attributed to a “content of consciousness” which is denied to a physical object. T hus we inquire w hether the assertion of the reality of a feeling o r a sensation has a different m eaning from th at of the reality of a physical body? This can m ean for us only: do we have different m odes of verification in each case? T h e answ er is: No! In order to m ake this clear it is necessary to undertake some slight analysis o f the logical form of existential propositions. T h e general logical insight th a t an existential proposition about a datum is possible only if it is referred to by a description, and not if it is given by direct indication holds of course for “data of consciousness.” (In the language of symbolic logic this is expressed by the fact th at an existential proposition m ust contain an “operator.” ) In B. Russell’s symbolism, for exam ple, an existential proposition has the form (H x ) fx, o r in w ords: “ there is an x having the property f.” T he com bination of w ords “ there is an a ,” where “a ” is the proper nam e of an object directly present, and therefore m eans the sam e as “this,” is m eaningless, and cannot be written in Russell’s symbolism. W e m ust attain the insight th at D escartes’ statem ent “ I am ”— or, to use a less m isleading form ula­ tion, “my contents of consciousness exist”— is simply m eaningless; it expresses nothing and contains no knowledge. T his is because “contents of consciousness” occurs in this context sim ply as a name for the given: no characteristic is expressed whose presence could be tested. A proposition only has m eaning, is verifiable only, if I can

Positivism and Realism

[ 99 ]

state the conditions under which it would be true and under which it would be false. But how shall I describe the conditions under which the proposition “ my contents of consciousness exist” would be false? Every attem pt would lead to absurdity, for example to such state­ m ents as “ it is the case th at nothing is the case,” or something of the sort. T herefore it is self-evident th at I cannot describe the conditions which m ake the proposition true (try to do so !). There is indeed also no doubt th at D escartes failed to gain any knowledge through his statem ent, and was no wiser at the end than he was at the beginning of his inquiry. No, a question concerning the reality of an experience makes sense only if its reality can significantly be doubted. I can for ex­ am ple ask: Is it really true that I felt happy upon hearing that news? This can be verified o r falsified in exactly the same way as, say, the question: is it true that Sirius has a satellite (that this satellite is re a l)? T h at on a given occasion I experienced pleasure can for exam ple be verified by exam ining the statem ents of others concern­ ing my behavior at the time, by finding a letter written by me at the tim e, o r even simply by a veridical memory of the em otion ex­ perienced. Hence there is here absolutely no difference in principle: to be real always m eans to stand in a definite relationship to the given. A nd this also holds, say, for an experience at this very moment. F o r exam ple, I can significantly ask (say in the course of a physio­ logical ex p erim en t): do I, o r do I not, experience a pain at this m om ent? Observe th at here “ pain” does not function as a proper nam e for a this-here, but represents a concept which stands for a describable class of experiences. H ere, too, the question is answered by determ ining th at an experience having certain describable prop­ erties occurs in conjunction with certain conditions (experim ental conditions, concentration of attention e tc .). Such describable prop­ erties would be, for instance, similarity to an experience occurring under certain other conditions; the tendency to produce certain reac­ tions, etc. No m atter how we twist and turn: it is impossible to interpret an existential proposition except as a statem ent regarding a con­ nection of perceptions. It is reality of the same sort that one must attribute to data of consciousness and, say, to physical events. H ardly anything in the history of philosophy has produced greater con­ fusion than the attem pt to distinguish one of the two as true “being.” W herever the word “ real” is significantly used it means one and the same thing. T he opponent of this view will perhaps not feel that w hat has

[ 100 ]

MORITZ SCHLICK

been said upsets his own view in any way, but will be of the im pres­ sion that the preceding argum ents presuppose a point of departure he is from the outset unwilling to adopt. H e m ust indeed grant that a decision regarding the reality o r unreality of a fact in ex­ perience is always m ade in the way described, but he claim s that in this way one arrives only at w hat K ant called empirical reality. This m ethod defines the realm of the observations of everyday life and of science, but beyond this limit lies som ething m ore, the transcendent reality, which cannot be deduced by strict logic, and therefore is not a postulate of the understanding, but is perhaps a postulate of reason. This is the only real external world, and it alone is relevant to the philosophical problem of the existence of the ex­ ternal world. Thus o u r discussion leaves the question of the m eaning of the word “ reality,” and turns to that of the phrase “external w orld.” The phrase “external world” is evidently used in two different ways: first in the language of everyday life, and secondly as a tech­ nical term in philosophy. W herever it occurs in daily life it has, as do m ost of the expres­ sions used in practical affairs, a sensible m eaning which can be stated. In opposition to the “ inner w orld,” which includes m em ories, thoughts, dream s, desires, feelings, the external w orld is simply the world of m ountains and trees, of anim als and men. Every child knows w hat is m eant when we assert the existence o f definite o b ­ jects of this world; and we m ust insist that it really m eans absolutely nothing more than w hat the child knows. We all know how to verify the statem ent, say, th at “there is a castle in the p ark outside the city.” W e act in certain ways and then if certain clearly describable facts are experienced we say: “Yes, there really is a castle there,” otherwise we say the statem ent was wrong, o r a lie. A n d if someone asks us: “W as the castle also there at night, w hen no one saw it?” We answ er: “Undoubtedly! F o r it would have been im ­ possible to build it since this m orning; furtherm ore the condition of the building shows th at not only was it there yesterday, but for hundreds of years, hence before we were born.” T hus we pos­ sess quite definite em pirical criteria with which to determ ine w hether houses and trees existed when we did not see them , and w hether they already existed before o ur birth, and w hether they will exist after our death. This m eans that the statem ent th a t those things “exist independently of us” has a clear verifiable m eaning, and is obviously to be affirmed. W e can very well distinguish em ­ pirically things of this sort from those th at are only “subjective” and

Positivism and Realism

[ 101 ]

“dependent upon us.” If, for instance, because of some visual defect I see a d ark spot when I look at the adjacent wall I say that the spot is there only when I look at it, but I say that the wall is there even when I do not look a t it. The verification of this distinction is indeed quite easy, and both these statem ents say just w hat is con­ tained in the verifications, and nothing else. H ence if the phrase external world is taken with the signification it has in everyday life then the question regarding its existence is sim ply the question: are there in addition to memories, desires and ideas also stars, clouds, plants, anim als, and my own body? We have just seen th at it would be simply absurd to answer this question in the negative. T here are, quite evidently, houses, clouds, and ani­ m als existing independently of us, and I said above that any thinker who denied the existence of the external world in this sense would have no claim on o u r respect. Instead of telling us w hat we m ean when we speak o f m ountains and plants he would convince us that there aren’t any such things at all! B ut science! Does it, in opposition to com m on sense, mean som ething other than things like houses and trees when it speaks of the external world? It seems to me that nothing of the sort is the case. F o r atom s and electric fields, or whatever the physicist may speak of, are just w hat constitute houses and trees according to th eir theory; and therefore the one m ust be real in the same sense as the other. T he objectivity of m ountains and clouds is exactly the sam e as th at of protons and energies— these latter stand in no greater opposition to “subjectivity,” say to feelings and hallucina­ tions, than do the form er. In fact we are at last convinced that the existence of even the m ost subtle “invisible things,” assumed by the scientist, is, in principle, verified exactly as is the reality of a tree o r a star. In o rd er to settle the dispute concerning realism it is of very great im portance to draw the physicist’s attention to the fact that his external w orld is simply nature, which also surrounds us in daily life, and not the “transcendent w orld” of the m etaphysician. The distinction betw een the two is again especially clear in K ant’s philosophy. N ature, and everything of which the physicist can and m ust speak belongs, according to K ant, to em pirical reality, and w hat that m eans is (as we have already said) explained by him in just the way that it m ust be by us. A tom s in K ant’s system have no transcendent reality, they are not “ things in themselves.” Hence the physicist cannot appeal to the K antian philosophy; its arguments lead only to the em pirical external world which we all acknowledge,

[ 102 ]

MORITZ SCHLICK

not to a transcendent w orld; his electrons are not m etaphysical entities. Nevertheless m any scientists speak of the necessity o f assum ing the existence of an external world as a m etaphysical hypothesis. T o be sure, they do not do this within their own science (even though all the necessary hypotheses of a science ought to be found within it) , but only where they leave this realm and begin to philosophize. In fact the transcendent external world is som ething dealt with only in philosophy, never in a science, nor in daily life. It is sim ply a technical term into whose m eaning we m ust now inquire. How is the transcendent o r m etaphysical external w orld dis­ tinguished from the em pirical world? In philosophical system s it is thought of as somehow standing behind the em pirical world, where the w ord “ behind” indicates th at it cannot be known in the same sense as can the em pirical w orld, th at it lies beyond a boundary which separates the accessible from the inaccessible. This distinction has its original source in the view, form erly held by most philosophers, th at in order to know an object it is necessary to perceive it directly; knowledge is a sort o f intuition, and is perfect only when the object is directly present to the know er as a sensation o r feeling. H ence according to this view w hat cannot be immediately experienced o r perceived rem ains unknow able, in­ com prehensible, transcendent; it belongs to the realm o f things in themselves. H ere there is simply a confusion, which I have revealed elsewhere m any times, between knowledge and m ere acquaintance o r experience. But m odern scientists will certainly be guilty of no such confusion. I do not believe th at any physicist is of the opinion that knowledge of the electron consists in the fact that it enters bodily into the consciousness of the investigator through an act of intuition. H e will, rather, hold the view th at for com plete knowledge it is only necessary to state the laws governing the behavior o f the electron so exhaustively th at all form ulae into which its properties en ter in any way are completely confirm ed by experience. In oth er w ords: the electron, and equally all physical realities are not unknow able things in themselves, they do not belong to transcendent reality, if this is characterized by the fact th at it contains the unknow able. Therefore we again com e to the conclusion that all physical hypotheses can refer only to empirical reality, if by this we m ean the knowable. In fact it would be a self-contradiction to assum e hypo­ thetically something unknow able. F o r there m ust always be definite reasons for setting up an hypothesis, the hypothesis has a certain function to fulfill. Therefore w hat is assum ed in the hypothesis m ust

Positivism and Realism

[1 0 3 ]

have the property of fulfilling this function, and must be so con­ stituted th at it is justified by those reasons. But in just this way cer­ tain statem ents are m ade regarding the assumed entity and these express o ur knowledge of it. A nd of course they contain complete knowledge of it. F o r only that can be assumed hypothetically for which there are grounds in experience. O r does the “ realistic” scientist w ant to designate the theory of objects which are not directly experienced as a metaphysical hypoth­ esis for some other reason than that of their unknowableness, which is not under consideration at all? T o this he will perhaps answer affirma­ tively. In fact we learn from num erous statem ents in the literature that the physicist does not add any statem ent of its unknowable char­ acter to his affirmation of a transcendent world; quite the contrary, he is rightly of the opinion that the nature of the extra-m ental things is correctly represented by his equations. Thus the external world of the physical realist is not that of traditional metaphysics. H e uses the technical term of the philosopher, but what he means by it has ap­ peared to us to be nothing but the external world of everyday life, whose existence no one, not even the “positivist,” doubts. W hat, then, is that other reason which leads the “ realist” to conceive his external world as a metaphysical hypothesis? Why does he want to distinguish it from the empirical external world which we have described? T he answer to this question leads us back again to an earlier point in o u r discussion. The physical “realist” is quite satisfied with o u r description of the external world except in one point: he does not believe that we have granted it enough reality. It is not because it is unknowable, o r for any such reason that he thinks his “external w orld” differs from the empirical, but only be­ cause a different, higher reality pertains to it. This often shows itself in his language; the word “real” is frequently reserved for that ex­ ternal w orld in contrast with the merely “ideal,” “subjective” con­ tents of consciousness, and in opposition to mere “logical” construc­ tions, “positivism ” being reproached with the attem pt to reduce reality to such logical constructs. B ut the physical realist, too, feels obscurely that, as we know, reality is not a “predicate,” hence he cannot well pass from our em pirical to his transcendent external world by ascribing to it, in addition to the characteristics which we also attribute to physical ob­ jects, the characteristic of “reality.” Nevertheless he expresses him­ self in this way; and this illegitimate leap, which carries him beyond the realm of significance, would indeed be “metaphysical,” and will be felt by him to be such.

[ 104 ]

MORITZ SCHLICK

N ow we see the situation clearly, and can base o u r final judg­ m ent on the foregoing considerations. O ur principle th at the truth and falsity of all statem ents, includ­ ing those concerning the reality of a physical object, can be tested only in the “given,” and that therefore the m eaning of all proposi­ tions can be form ulated and understood only with the help o f the given— this principle is m istakenly conceived as if it asserted o r presupposed that only the given is real. T herefore the “realist” feels im pelled to contradict this principle and to establish the contrary: th at the meaning of an existential proposition is in no sense exhausted by mere propositions of the form “ under these definite conditions th at definite experience will occur” (those propositions constituting an infinite set according to our view ), but that their m eaning lies beyond all this in something else, which is to be designated, say, as “ independent existence,” as “transcendent being,” o r similarly, and to which our principle fails to do justice. A nd here we inquire: Well, how do you do justice to it? W hat do these phrases “independent existence” and “ transcendent being” mean? In other words: w hat verifiable difference does it m ake in the world w hether transcendent being pertains to an object o r not? Two answers are given to this question. T he first is th at it m akes a very great difference. F o r a scientist who believes in a “ real” ex­ ternal world will feel and work very differently from one w ho be­ lieves himself to be “describing sensations.” T he form er will o b ­ serve the starry heavens, whose view m akes him conscious of his own puny nature, and the incom prehensible sublimity and grandeur of the world with very different feelings of fervor and awe from the latter, fo r whom the m ost distant galactic systems are merely “com ­ plexes of his own sense-im pressions.” T he form er will devote him ­ self to his task with an inspiration and will feel a satisfaction in the knowledge of the external world which is denied to the latter, b e­ cause he believes himself to be dealing only with his own con­ structions. In answ er to this we offer the following com m ent. L et us assume th at somewhere in the behavior of two scientists there does exist a difference such as has been described here. Such a difference would o f course be an observable difference. Suppose now som ebody in­ sists on expressing this difference by saying that one of the scientists believes in a real external w orld and the other does not. In that event the meaning of this statem ent would consist solely in w hat we observe in the behavior of the two men. T h at is, the w ords “ absoWte reality” or “transcendent being,” o r w hatever expressions we

Positivism and Realism

[ 105 ]

m ight choose to employ, m ean here simply certain states of feeling, which occur in the m en when they observe the world, or make state­ m ents about it, o r philosophize. It is, indeed, the case that the use of the w ords “independent existence,” “ transcendent reality,” etc., is sim ply and only the expression of a feeling, of a psychological attitude of the speaker (this, m oreover, may, in the final analysis, be true of all m etaphysical propositions). If someone assures us that there is a real external world in the trans-em pirical sense of the word, he of course believes him self to have com m unicated some truth about the world. But in actual fact, his words express something very different; they merely express certain feelings which give rise to various lin­ guistic and other reactions on his part. If this self-evident point requires any further emphasis I should like to call attention to the fact— and with the greatest stress on the seriousness of what is said— th at the non-metaphysician is not dis­ tinguished from the m etaphysician by, say, the absence in him of those feelings which the other expresses in term s of the statements of a realistic philosophy, but only by the fact that he recognizes that these statem ents simply do not have the meaning they seem to have, and are therefore to be avoided. The non-metaphysician will express these sam e feelings in a different way. In other words: the contrast draw n in the first answer of the “realist” between the two types of thinkers was misleading and unjust. If one is unfortunate enough not to feel the sublimity of the starry heavens something other than a logical analysis of the concepts of reality and external world is to be blam ed. T o assum e th at the opponents of metaphysics are unable justly to com prehend, say, the greatness of Copernicus, be­ cause in a certain sense the Ptolem aic view represents the empirical facts as well as the Copernican, seems to me to be as strange as to believe th at the “positivist” cannot be a good parent because accord­ ing to his theory his children are merely complexes of his own senseim pressions, and it is therefore senseless to take measures for their w elfare after his death. N o: the world of the non-metaphysician is the sam e world as th at of all other men; it lacks nothing which is needed to bestow meaning on all the propositions of science and the whole conduct of life. He merely avoids adding meaningless state­ m ents to his description of the world. We com e now to the second answer which can be given to the question concerning the meaning of the assertion of a transcendent reality. It consists in granting that it m akes no difference at all for experience w hether or not one assumes something further to exist behind the em pirical world, that metaphysical realism therefore

[ 106 ]

M ORITZ SCHLICK

cannot be tested and is actually unverifiable. H ence one cannot in­ dicate any further w hat is m eant by this assertion; but nevertheless it does m ean something, and this m eaning can be understood even without verification. This is nothing but the view, criticized in the previous section, that the meaning of a proposition has nothing to do with its verifica­ tion, and we need only apply o u r earlier general criticism to this special case. Therefore we m ust say: you designate here by ex­ istence or reality something which simply cannot in any way be given or explained. Yet despite this you believe th at those words m ake sense. We shall not quarrel with you over this point. B ut this much is certain: according to the admission just m ade this sense can in no way become evident, it cannot be expressed in any w ritten or spoken com m unication, nor by any gesture o r conduct. F o r if this were possible we should have before us a verifiable em pirical fact, and the world would be different if the proposition “ there is an external w orld” were true, from w hat it would be if it were false. This difference would then constitute the m eaning of the phrase “ real external world,” hence it would be an em pirical m eaning; th a t is, this real external world would again be only the em pirical world, which, like all hum an beings, we also acknowledge. Even to speak of any other world is logically impossible. T here can be no discus­ sion concerning it, for a non-verifiable existence cannot enter m ean­ ingfully into any possible proposition. W hoever still believes— o r be­ lieves himself to believe— in it m ust do so only silently. A rgum ents can relate only to w hat can be said. The results of our discussion m ay be sum m arized as follows: 1) T he justified unassailable nucleus of the “ positivistic” tendency seems to me to be the principle th at the m eaning of every proposi­ tion is completely contained within its verification in the given. But this principle has seldom been clearly apparent within that general tendency, and has so frequently been mixed with so m any un­ tenable propositions that a logical purification is necessary. If one wishes to call the result of the purification positivism, which would per­ haps be historically justifiable, at least a differentiating adjective m ust be added. Sometimes the term “logical” o r else “ logistic positivism ” is used.3 Otherwise the designation “ consistent em piricism ” seems to me to be appropriate. 3. See th e a rtic le by B lum berg a n d F eigl in th e Journal o f Philosophy, V o l. X X V III (1 9 3 1 ), th e a rtic le by E . K a ila in th e Annales Universitatis Aboensis, V ol. X II I , Ser. B. (T u rk u , 1 9 3 0 ), a n d th e o n e by A . PetzäU in th e Schriften der

Universität Göteborg.

Positivism and Realism

[ 107 ]

2 ) T his principle does not m ean and does not imply that only the given is real. Such an assertion does not make sense. 3 ) H ence also, consistent em piricism does not deny the existence of an external w orld; it merely points o ut the em pirical meaning of this existential proposition. 4 ) It is not a “Theory of As If.” It does not assert that every­ thing behaves as if there were physical independent bodies; but for it, too, everything is real which the non-philosophizing scientist calls real. T he subject-m atter of physics is not sensations, but laws. The form ulation, used by some positivists, th at bodies are only “com ­ plexes of sensations” is therefore to be rejected. W hat is correct is only th at propositions concerning bodies are transform able into equivalent propositions concerning the occurrence of sensations in accordance with laws. 5 ) H ence logical positivism and realism are not in opposition; whoever acknowledges o ur fundam ental principle m ust be an em ­ pirical realist.4 6 ) A n opposition exists only betw een the consistent empiricist and the m etaphysician, and indeed no m ore against the realist than against the idealist m etaphysician (the form er has been referred to in our discussion as “ realist” in quotation m arks). 7 ) The denial of the existence of a transcendent external world w ould be just as m uch a m etaphysical statem ent as its affirmation. H ence the consistent em piricist does not deny the transcendent w orld, bu t shows that both its denial and affirmation are meaningless. This last distinction is of the greatest im portance. I am con­ vinced that the chief opposition to o u r view derives from the fact th at the distinction between the falsity and the meaninglessness of a proposition is not observed. The proposition “Discourse concern­ ing a m etaphysical external world is meaningless” does not say: “ T here is no external w orld,” but something altogether different. T he em piricist does not say to the m etaphysician “ what you say is false,” but, “ what you say asserts nothing at all!” H e does not con­ tradict him , but says “ I do n ’t understand you.” 4. r e fe rre d

O n th is p o in t a n d o n the e n tire su b je ct o f th e p rese n t essay th e re a d e r is also to H a n s C o rn e liu s’ “ Z u r K ritik d c r W issen sch aftlich en G rundbcgrifT e,’' Erkenntnis, V ol. I I. T h e fo rm u la tio n s th e re are , how ev er, o p en to o b jectio n s. See a lso th e sp le n d id re m a rk s in C h a p te r X o f P hillip F r a n k ’s fine w o rk , Das Kausalgesetz und seine Grenzen, a n d R u d o lf C a rn a p 's Scheinproblcme der Philosophie.

5

The Empiricist Criterion of Meaning BY C A R L G . H E M P E L

1. I n t r o d u c t i o n T h e f u n d a m e n t a l t e n e t of m odem em piricism is the view th at

all non-analytic knowledge is based on experience. L et us call this thesis the principle of em piricism .1 C ontem porary logical em piricism has added2 to it the maxim th at a sentence m akes a cognitively m ean­ ingful assertion, and thus can be said to be either true o r false, only if it is either ( 1 ) analytic o r self-contradictory o r ( 2 ) capable, at least in principle, of experiential test. A ccording to this so-called empiricist criterion of cognitive meaning, or of cognitive significance, m any of the form ulations of traditional m etaphysics and large parts of epistemology are devoid of cognitive significance— how ever rich some of them may be in non-cognitive im port by virtue o f their emotive appeal o r the m oral inspiration they offer. Similarly certain doctrines which have been, at one tim e o r another, form ulated w ithin em pirical science o r its border disciplines are so contrived as to be incapable of test by any conceivable evidence; they are therefore qualified as pseudo-hypotheses, which assert nothing, and which T h is a rtic le first a p p e a re d in V ol. 4 o f Revue Internationale de Philosophic ( 1 9 5 0 ). It is rep u b lish ed h ere w ith th e k in d p erm issio n o f P ro fe ss o r H e m p e l a n d th e e d ito r o f th a t jo u rn a l. 1. T h is te rm is u sed by B e n jam in ( 2 ) in a n e x a m in a tio n o f th e fo u n d a tio n s o f em p iricism . F o r a re c e n t d isc u ssio n o f th e basic ideas o f em p iricism see R u ssell ( 2 7 ) , P a rt Six. 2. In h is stim u la tin g a rticle , “ P o sitiv ism ,” W . T . S ta c e arg u e s, in effect, th a t th e te sta b ility c rite rio n o f m e an in g is n o t lo g ically e n ta ile d b y th e p rin c ip le o f em p iricism . (S ee ( 2 9 ) , especially se ctio n 11.) T h is is c o r r e c t: a c c o rd in g to th e la tte r, a se n te n c e ex p resses know led g e o n ly if it is e ith e r a n a ly tic o r c o rro b o ra te d b y e m p iric a l ev i­ d en c e; th e fo rm e r goes fu rth e r a n d identifies th e d o m a in o f co g n itiv ely sig n ific an t d isco u rse w ith th a t o f p o te n tia l kn o w led g e; i. e., it g ra n ts co g n itiv e im p o rt o n ly to sen ten c es fo r w h ich — u nless they a re e ith e r a n a ly tic o r c o n tra d ic to ry — a te s t b y em p irica l ev id en ce is conceivable.

[ 108 ]

The Empiricist Criterion of Meaning

[ 109 ]

therefore have no explanatory or predictive force whatever. This verdict applies, for exam ple, to the neo-vitalist speculations about entelechies o r vital forces, and to the “telefinalist hypothesis” pro­ pounded by Lecom te du Noiiy.3 T he preceding form ulations of the principle of empiricism and o f the em piricist meaning criterion provide no m ore, however, than a general and rather vague characterization of a basic point of view, and they need therefore to be elucidated and amplified. A nd while in the earlier phases of its developm ent, logical empiricism was to a large extent preoccupied with a critique of philosophic and scien­ tific form ulations by m eans of those fundam ental principles, there has been in recent years an increasing concern with the positive tasks of analyzing in detail the logic and m ethodology of em pirical science and of clarifying and restating the basic ideas of empiricism in the light of the insights thus obtained. In the present article, I propose to discuss some of the problem s this search has raised and some of the results it seems to have established. 2. C h a n g e s i n t h e T e s t a b i l i t y C r i t e r i o n E m p ir ic a l M e a n in g

of

As o u r form ulation shows, the em piricist meaning criterion lays dow n the requirem ent of experiential testability for those among the cognitively meaningful sentences which are neither analytic nor contradictory; let us call them sentences with em pirical meaning, o r em pirical significance. The concept of testability, which is to render precise the vague notion of being based— o r rather baseable — on experience, has undergone several modifications which reflect an increasingly refined analysis of the structure of em pirical knowl­ edge. In the present section, let us exam ine the m ajor stages of this developm ent. F o r convenience of exposition, we first introduce three auxiliary concepts, nam ely those of observable characteristic, of observation predicate, and of observation sentence. A property o r a relation of physical objects will be called an observable characteristic if, under suitable circum stances, its presence or absence in a given instance can be ascertained through direct observation. Thus, the terms “green,” “ soft,” “liquid,” “longer than,” designate observable char­ acteristics, while “ bivalent,” “radioactive,” “ better electric con­ d u cto r,” and “ introvert” do not. Term s which designate observable characteristics will be called observation predicates. Finally, by an 3. C f. ( 1 9 ) , C h . X V I.

[

110

]

CARL G. H EM PEL

observation sentence we shall understand any sentence which— cor­ rectly o r incorrectly— asserts of one or m ore specifically nam ed ob ­ jects th at they have, o r th at they lack, some specified observable characteristic. The following sentences, for exam ple, m eet this con­ dition: “T he Eiffel Tow er is taller than the buildings in its vicinity,” “T he pointer of this instrum ent does not cover the point m arked ‘3* on the scale,” and even, “The largest dinosaur on exhibit in New Y ork’s M useum of N atural H istory had a blue tongue” ; for this last sentence assigns to a specified object a characteristic— having a blue tongue— which is of such a kind th at under suitable circum stances (e.g., in the case of my Chow dog) its presence o r absence can be ascertained by direct observation. O ur concept of observation sentence is intended to provide a precise interpretation of the vague idea of a sentence asserting som ething that is “ in principle” ascer­ tainable by direct observation, even though it may happen to be actually incapable of being observed by myself, perhaps also by my contem poraries, and possibly even by any hum an being who ever lived o r will live. Any evidence th at m ight be adduced in the test of an empirical hypothesis may now be thought of as being ex­ pressed in observation sentences of this kind.4 We now turn to the changes in the conception of testability, and thus of em pirical meaning. In the early days of the V ienna Circle, a sentence was said to have em pirical m eaning if it was capable, at least in principle, of com plete verification by observa­ tional evidence; i.e., if observational evidence could be described which, if actually obtained, would conclusively establish the tru th of the sentence.5 W ith the help of the concept of observation sen4. O b serv atio n sen ten c es o f th is k in d b elo n g to w h a t C a rn a p h as c a lle d th e th in g -lan g u ag e (cf., c. g., ( 7 ) , pp. 52 -5 3 ). T h a t th ey a re a d e q u a te to f o rm u la te th e d a ta w hich serve as the basis fo r em p irical te sts is c le a r in p a rtic u la r fo r th e in te rsu b jective te stin g p ro c e d u re s used in science as w ell as in la rg e a re a s o f e m p iric a l in q u iry o n th e com m o n -sen se level. In ep istem o lo g ical d iscu ssio n s, it is fre q u e n tly assum ed th a t th e u ltim ate ev idence fo r beliefs a b o u t em p irica l m a tte rs co n sists in p e rc e p tio n s a n d sen satio n s w hose d e s c rip tio n ca lls fo r a p h e n o m e n a listic ty p e o f lan g u age. T h e specific p ro b lem s c o n n e c te d w ith th e p h e n o m e n a listic a p p ro a c h c a n ­ n o t be discussed h e re ; b u t it sh o u ld b e m e n tio n e d th a t a t a n y r a te all th e c ritic a l c o n s id e ra tio n s p rese n ted in th is a rtic le in re g a rd to th e te sta b ility c rite rio n a re ap p licab le, mutatis mutandis, to th e case o f a p h en o m en a listic b asis as w ell. 5. O riginally, th e perm issible ev id en ce w as m e a n t to b e re stric te d to w h a t is o b se rv a b le by th e sp e a k e r a n d p e rh a p s h is fello w -b ein g s d u rin g th e ir lifetim es. T h u s c o n stru e d , th e c rite rio n rules o u t, as cognitiv ely m ean in g less, all sta te m e n ts ab o u t th e d ista n t fu tu re o r th e re m o te p ast, as h as b een p o in te d o u t, a m o n g o th e rs , by A yer in ( 1 ) , C h a p te r I; by P a p in ( 2 1 ) , C h a p te r 13, esp . p p . 333 ff.; a n d by R ussell in ( 2 7 ) , pp. 445-47. T h is difficulty is a v o id ed , h o w ev er, if w e p e rm it th e -ividencc to con sist o f an y finite se t o f “ logically p o ssib le o b se rv a tio n d a ta ,” e a c h o f th em fo rm u la te d in an o b se rv a tio n sentence. T h u s, e. g., th e se n te n c e S „ “ T h e to n g u e o f th e la rg e st d in o sa u r in N ew Y o rk ’s M u seu m o f N a tu r a l H isto ry w as b lu e

The Empiricist Criterion o f Meaning

[

111

]

tence, we can restate this requirem ent as follows: A sentence S has em pirical m eaning if and only if it is possible to indicate a finite set of observation sentences, O u 0 2, . . . , O n, such that if these are true, then S is necessarily true, too. As stated, however, this condition is satisfied also if S is an analytic sentence o r if the given observation sentences are logically incompatible with each other. By the following form ulation, we rule these cases out and at the same tim e express the intended criterion m ore precisely: (2 .1 ) Requirement of complete verifiability in principle: A sen­ tence has em pirical m eaning if and only if it is not analytic and follows logically from some finite and logically consistent class of observation sentences.6 o r b la c k " is co m p letely v erifiable in o u r sense; fo r it is a logical co n seq u e n ce o f the S e n te n c e S 2, “T h e to n g u e o f th e larg est d in o sa u r in N ew Y o rk ’s M u seu m o f N a tu ra l H isto ry w as b lu e ” ; a n d th is is a n o b se rv a tio n se n ten c e, as has b een sh o w n above. A n d if th e c o n c e p t o f verifiability in principle a n d th e m o re g en eral co n c e p t o f confirmability in principle, w hich w ill b e co n sid ered la te r, a re c o n stru e d as referrin g to logically possible evidence a s expressed by o b se rv a tio n sen ten ces, th en it follow s sim ilarly th a t th e class o f sta te m e n ts w hich a re verifiable, o r a t least co n firm ab le, in p rin c ip le in c lu d es such assertio n s as th a t th e p la n et N e p tu n e a n d th e A n ta rc tic C o n tin e n t existed b e fo re th e y w ere d isco v ered , a n d th a t ato m ic w arfa re, if n o t c h e c k e d , m ay le ad to th e e x te rm in a tio n o f th is p la n et. T h e o b je ctio n s w hich R ussell (cf. ( 2 7 ) , pp. 445 a n d 4 4 7 ) raises ag a in st th e verifiability c rite rio n by referen c e to th o se ex a m p le s d o n o t ap p ly th e re fo re if th e c rite rio n is u n d ersto o d in th e m a n n er h e re su g g ested . In cid e n tally , sta te m e n ts o f th e k in d m e n tio n e d by R ussell, w hich are n o t a c tu a lly v erifiable by any h u m a n being, w ere ex plicitly reco g n ized as cognitively sig n ifican t a lre a d y by S ch lic k (in ( 2 8 ), P a rt V ) , w h o arg u ed th a t th e im possibility o f v erifying th em w as “ m erely e m p iric a l.” T h e c h a ra c te riz a tio n o f verifiability w ith th e h e lp o f th e c o n c e p t o f o b se rv a tio n se n ten c e as suggested h ere m ight serve as a m o re ex p licit a n d rig o ro u s sta te m e n t o f th a t co n c ep tio n . 6. A s h as fre q u en tly b een em p h asized in em p iricist lite ra tu re , th e term “ veri­ fiab ility ” is to in d ic ate, o f co u rse , th e co n ceiv ab ility , o r b e tte r, th e logical possibility o f ev id en ce o f an o b se rv a tio n a l k in d w hich, if ac tu a lly e n c o u n te re d , w o u ld c o n sti­ tu te co n c lu siv e ev idence fo r th e given se n ten c e; it is n o t in te n d ed to m ean th e te c h n ic a l possib ility o f p e rfo rm in g th e tests n eed ed to o b ta in su c h evid en ce, an d ev en less d o es it m e an th e possibility o f ac tu a lly finding d irectly o b se rv a b le p h e­ n o m e n a w hich c o n s titu te con clu siv e ev idence fo r th a t se n ten c e— w hich w ould be ta n ta m o u n t to th e a c tu a l existence o f su c h ev id en ce an d w o u ld th u s im ply the tru th o f th e given se n ten c e. A n alo g o u s re m a rk s ap p ly to th e term s “ falsifiability” an d “ co n firm a b ility .” T h is p o in t h as been d isreg a rd ed in so m e rec en t critica l discussions o f th e verifiability c rite rio n . T h u s, e.g., R ussell (c f. ( 2 7 ) , p. 4 4 8 ) c o n stru e s verifia­ b ility as th e a c tu a l ex isten ce o f a set o f conclusiv ely v erifying o c c u rre n ces. T h is c o n c e p tio n , w hich h as nev er been a d v o c a te d by any lo g ical em piricist, m u st n atu ra lly tu rn o u t to b e in a d e q u a te sin c e a c o rd in g to it th e em p irical m ean in g fu ln css o f a s e n te n c e co u ld n o t b e estab lish ed w ith o u t g a th e rin g em p irical ev idence, an 1 m o re­ o v e r e n o u g h o f it to p e rm it a conclu siv e p ro o f o f th e sen ten c es in qu estio n ! It is n o t su rp risin g , th e re fo re , th a t h is e x tra o rd in a ry in te rp re ta tio n o f verifiability leads R u ssell to th e co n c lu sio n : " I n fa c t, th a t a p ro p o sitio n is verifiable is itself n o t v erifia b le” (/. c.) A ctu a lly , u n d e r th e em p iricist in te rp re ta tio n o f co m p lete v eri­ fiab ility , an y sta te m e n t assertin g th e verifiability o f so m e sen ten ce S w hose tex t is q u o te d , is e ith e r an a ly tic o r c o n tra d ic to ry ; fo r th e d ecisio n w h eth e r th ere exists a c la ss o f o b se rv a tio n sen ten c es w hich en ta il S, i. e., w h eth e r su ch o b se rv a tio n se n ­ te n c e s c a n b e fo rm u la te d , n o m a tte r w h eth e r th e y a re tru e o r false— th a t d ecision

[ ||2

]

CARL G. H EM PEL

This criterion, however, has several serious defects. T he first of those here to be m entioned has been pointed out by various w riters: ( a ) T he verifiability requirem ent rules out all sentences of uni­ versal form and thus all statem ents purporting to express general laws; for these cannot be conclusively verified by any finite set of observational data. And since sentences of this type constitute an integral part of scientific theories, the verifiability requirem ent m ust be regarded as overly restrictive in this respect. Similarly, the cri­ terion disqualifies all sentences such as “F o r any substance there exists some solvent,” which contain both universal and existential quantifiers (i.e., occurrences of the term s “ all” and “som e” o r their equivalents); for no sentences of this kind can be logically deduced from any finite set of observation sentences. Two further defects of the verifiability requirem ent do not seem to have been widely noticed: ( b ) Suppose that S is a sentence which satisfies the proposed criterion, whereas N is a sentence such as “T he absolute is perfect,” to which the criterion attributes no em pirical meaning. T h en the alternation SvN (i.e., the expression obtained by connecting the two sentences by the word “o r” ) , likewise satisfies the criterion; for if S is a consequence of some finite class of observation sen­ tences, then trivially SvN is a consequence of the sam e class. B ut clearly, the empiricist criterion of m eaning is not intended to coun­ tenance sentences of this sort. In this respect, therefore, the require­ ment of com plete verifiability is too inclusive. (c ) L et “P ” be an observation predicate. T hen the purely ex­ istential sentence “ ( E ;t) P ( ;t) ” ( “T here exists at least one thing that has the property P ” ) is com pletely verifiable, for it follows from any observation sentence asserting of some particular object th at it has the property P. But its denial, being equivalent to the universal is a m a tte r o f p u re logic a n d re q u ire s n o fa c tu a l in fo rm a tio n w h atev er. A sim ilar m isu n d ersta n d in g is in ev idence in th e fo llo w in g p assag e in w h ich W . H . W e rk m cister claim s to c h a ra c te riz e a view h eld by lo g ical p o sitiv ists: “ A p ro p o sitio n is said to b e ‘tru e ’ w hen it is ‘v erifiab le in p rin c ip le ’; i. e., w h en w e k n o w th e c o n d itio n s w h ich , w hen realized , w ill m a k e ‘v erific atio n ’ p o ssib le (c f. A y e r ) .” (cf. ( 3 1 ), p. 145). T h e q u o te d thesis, w h ich , ag a in , w as n ev e r h eld by a n y logical positivist, in clu d in g A yer, is in fa c t logically a b s u rd . F o r w e c a n re a d ily d escrib e c o n d itio n s w hich, if realized , w o u ld verify th e se n ten c e “T h e o u tsid e o f th e C h ry sle r B uilding is p a in te d a b rig h t yellow ’’; b u t sim ilarly , w e c a n d e sc rib e v erifying c o n d itio n s fo r its d en ial; h en ce, ac c o rd in g to th e q u o te d p rin cip le, b o th th e se n ten ce a n d its d en ial w ould h av e to b e co n sid e re d tru e . In c id e n ta lly , th e p assag e u n d e r discussion d o es n o t a c c o rd w ith W e rk m e iste r’s p erfec tly c o rre c t o b se rv a tio n , /. c., p. 40, th a t verifiability is in te n d e d to c h a ra c te riz e th e m e an in g o f a se n ten c e— w h ich show s th a t verifiability is m e a n t to b e a c rite rio n o f co g n itiv e sig n ifican ce r a th e r th a n o f tru th .

The Empiricist Criterion of Meaning

[ 113 ]

sentence “ (* ) — P ( * ) ” ( “N othing has the property P ” ) is clearly not com pletely verifiable, as follows from com m ent ( a ) above. H ence, u n d er the criterion (2 .1 ) , the denials of certain empirically -—and thus cognitively— significant sentences are empirically m ean­ ingless; and as they are neither analytic nor contradictory, they are cognitively meaningless. But how ever we may delimit the domain of significant discourse, we shall have to insist that if a sentence falls within th at dom ain, then so must its denial. T o put the m atter more explicitly: T he sentences to be qualified as cognitively meaningful are precisely those which can be significantly said to be either true o r false. But then, adherence to (2 .1 ) would engender a serious dilem m a, as is shown by the consequence just mentioned. We would either have to give up the fundam ental logical principle that if a sentence is true o r false, then its denial is false o r true, respectively (an d thus cognitively significant); o r else, we must deny, in a m anner rem iniscent of the intuitionistic conception of logic and mathematics, that “ (* ) — P ( * ) ” is logically equivalent to the negation of “ (E * ) P ( x ) .” Clearly, the criterion (2 .1 ), which has disqualified itself on several other counts, does not w arrant such drastic measures for its preservation; hence, it has to be abandoned.7 Strictly analogous considerations apply to an alternative criterion, which m akes com plete falsifiability in principle the defining char­ acteristic of em pirical significance. L et us form ulate this criterion as follows: A sentence has empirical meaning if and only if it is capable, in principle, of com plete refutation by a finite num ber of observational data; or, m ore precisely: (2 .2 ) Requirement of complete falsifiability in principle: A sen­ tence has em pirical m eaning if and only if its denial is not analytic and follows logically from some finite logically consistent class of observation sentences.8 7. T h e a rg u m e n ts h e re ad d u c ed ag ain st the verifiability c rite rio n also p ro v e th e in a d e q u a c y o f a view closely re la te d to it, n am ely th a t tw o sen ten ces have th e sam e co g n itiv e significance if an y set o f o b se rv a tio n sen ten ces w hich w o u ld verify o n e o f th em w o u ld also verify th e o th e r, a n d conversely. T h u s, e. g., u n d e r th is criterio n , an y tw o g en eral law s w o u ld h av e to b e assigned the sam e co g n itiv e significance, fo r n o g e n e ra l law is verified by an y set o f o b se rv a tio n sentences. T h e view ju st referre d to m u st b e cle a rly distin g u ish ed fro m a p osition w hich R ussell exam ines in his c ritic a l discu ssio n o f th e positivistic m eaning criterio n . I t is “ th e th e o ry th a t tw o p ro p o sitio n s w hose verified co n seq u en ces a re id e n tic al h av e th e sam e significance” ( ( 2 7 ) , p. 4 4 8 ). T h is view is u n te n a b le in d eed , fo r w h at co n seq u en ces o f a sta te m en t h a v e a c tu a lly been verified a t a given tim e is obviously a m a tte r o f h isto rical acci­ d e n t w h ich c a n n o t possibly serve to estab lish identity o f co g n itiv e significance. B ut I am n o t a w a re th a t an y logical positivist ever su b sc rib ed to th a t “ th e o ry .” 8. T h e id e a o f u sing th e o re tic a l falsifiability by o b se rv a tio n a l ev idence as the “ c rite rio n o f d e m a rc a tio n ” se p a ra tin g em p irical science fro m m a th em atics a n d logic o n th e o n e h a n d a n d fro m m e tap h y sics o n th e o th e r is d u e to K . P o p p e r (cf. (2 2 ),

[

114

]

CARL G. H EM PEL

This criterion qualifies a sentence as em pirically m eaningful if its denial satisfies the requirem ent of com plete verifiability; as is to be expected, it is therefore inadequate on sim ilar grounds as the latter: ( a ) It rules out purely existential hypotheses, such as “T here exists at least one unicorn,” and all sentences whose form ulation calls for mixed— i.e., universal and existential— quantification; for none of these can possibly be conclusively falsified by a finite num ber of observation sentences. ( b ) If a sentence S is com pletely falsifiable whereas N is a sentence which is not, then their conjunction, S.N. (i.e., the ex­ pression obtained by connecting the two sentences by the w ord “and” ) is completely falsifiable; for if the denial of S is entailed by some class of observation sentences, then the denial of S.N. is, a fortiori, entailed by the same class. T hus, the criterion allows «empirical significance to m any sentences which an adequate em piri­ cist criterion should rule out, such as, say “A ll swans are white and the absolute is perfect.” ( c ) If “ P” is an observation predicate, then the assertion th at all things have the property P is qualified as significant, but its denial, being equivalent to a purely existential hypothesis, is dis­ qualified (cf. ( a ) ) . H ence, criterion (2 .2 ) gives rise to the same dilemm a as (2 .1 ). In sum, then, interpretations of the testability criterion in term s of com plete verifiability o r of com plete falsifiability are inadequate because they are overly restrictive in one direction and overly in­ clusive in another, and because both of them require incisive changes in the fundam ental principles of logic. Several attem pts have been m ade to avoid these difficulties by construing the testability criterion as dem anding merely a partial and possibly indirect confirmability of em pirical hypotheses by ob ­ servational evidence. (2 .3 ) A form ulation suggested by A yer9 is characteristic of these attem pts to set up a clear and sufficiently com prehensive criterion of confirmability. It states, in effect, th at a sentence S has em pirical im port if from S in conjunction with suitable subsidiary sectio n 1-7 a n d 19-24; also see ( 2 3 ) , vol. II, pp. 2 8 2 -2 8 5 ). W h e th e r P o p p e r w o u ld su b sc rib e to th e p ro p o sed re sta te m e n t o f th e falsifiability c rite rio n , I d o n o t kn o w . 9. ( 1 ) , C h . I.— T h e c a se ag ain st th e re q u ire m e n ts o f v erifiab ility a n d o f falsifia­ bility, a n d fa v o r o f a re q u ire m e n t o f p a rtia l co n firm a b ility a n d d isc o n firm ab ility is very cle a rly p rese n ted also by P a p in ( 2 1 ) , C h a p te r 13.

The Empiricist Criterion o f Meaning

[

115

]

hypotheses it is possible to derive observation sentences which are not derivable from the subsidiary hypotheses alone. This condition is suggested by a closer consideration of the logical structure of scientific testing; but it is much too liberal as it stands. Indeed, as A yer himself has pointed out in the second edi­ tion of his book, Language, Truth, and Logic,™ his criterion allows em pirical im port to any sentence w hatever. Thus, e.g., if S is the sentence “The absolute is perfect,” it suffices to choose as a sub­ sidiary hypothesis the sentence “ If the absolute is perfect then this apple is red” in order to m ake possible the deduction of the ob­ servation sentence “This apple is red,” which clearly does not fol­ low from the subsidiary hypothesis alone.11 (2 .4 ) T o m eet this objection, A yer has recently proposed a modified version of his testability criterion. T he modification re­ stricts, in effect, the subsidiary hypotheses mentioned in (2 .3 ) to sentences which are either analytic o r can independently be shown to be testable in the sense of the modified criterion.12 B ut it can readily be shown that this new criterion, like the re­ quirem ent of com plete falsifiability, allows em pirical significance to any conjunction S.N, where S satisfies A yer’s criterion while N is a sentence such as “The absolute is perfect,” which is to be dis­ qualified by that criterion. Indeed: w hatever consequences can be 10. ( 1 ) , 2d ed ., p p . 11-12. 11. A cc o rd in g to S tac e (c f. ( 2 9 ) , p. 2 1 8 ) , th e c rite rio n o f p a rtia l a n d in d irect te sta b ility , w hich h e ca lls th e positivist p rin cip le, p resu p p o se s (a n d th u s logically e n ta ils ) a n o th e r p rin cip le, w h ich h e term s th e Principle o f Observable K inds : “ A se n ten c e, in o r d e r to b e significant, m u st assert o r d en y facts w hich a re o f a kind o r class su c h th a t it is logically possible d irectly to o b serv e som e facts w hich are in sta n c e s o f th a t class o r k in d . A n d if a se n ten c e p u rp o rts to assert o r d en y facts w h ich a rc o f a class o r k in d such th a t it w ould be lo g ically im possible d irectly to o b se rv e an y in stan c e o f th a t class o r k in d , th e n th e se n ten c e is n o n -sig n ifican t.” I th in k th e a rg u m e n t S tac e offers to p ro v e th a t th is p rin cip le is en tailed by the r e ­ q u ire m e n t o f te sta b ility is inconclusive (m ain ly b ecau se o f th e in c o rrec t ta cit as su m p ­ tio n th a t “ o n th e tra n sfo rm a tio n view o f d e d u c tio n ," th e p rem ises o f a valid d ed u ctiv e a rg u m e n t m u st b e n ecessary c o n d itio n s fo r th e co n c lu sio n (/. c., p. 2 2 5 ). W ith o u t p ressin g th is p o in t an y fu rth e r, I sh o u ld like to a d d h ere a re m a rk o n th e p rinciple o f o b se rv a b le k in d s itself. P ro fe sso r S tac e d ocs n o t say how w e a rc to d eterm in e w h a t “ fa c ts” a given se n ten c e asserts o r denies, o r indeed w h eth e r it asserts o r d en ies an y “ fa c ts” a t all. H ence, th e ex act im p o rt o f th e p rin cip le rem a in s u n clear. N o m a tte r, h ow ever, h ow o n e m ig h t ch o o se the c rite ria fo r the fa c tu a l re fe re n c e o f se n ten c es, th is m u ch seem s c e rta in : If a se n ten ce expresses an y fact a t all, say f, th e n it satisfies th e re q u ire m e n t laid dow n in th e first se n ten ce o f th e p rin cip le; fo r w e c a n alw ay s fo rm a class co n ta in in g / to g e th e r w ith th e fa c t ex p ressed by som e o b se rv a tio n se n ten c e o f o u r choice, w hich m ak es f a m e m b e r o f a class o f facts a t le ast o n e o f w hich is c a p a b le , in p rin cip le, o f d ire c t o b se rv a tio n . T h e first p art o f th e p rin c ip le o f o b se rv a b le k in d s is th e re fo re all-inclusive, so m ew h at like A y e r’s o rig in a l fo rm u la tio n o f th e em p iricist m e an in g criterio n . 12. T h is re stric tio n is exp ressed in recursive fo rm a n d involves n o vicious circle. F o r th e fu ll s ta te m e n t o f A y e r’s c rite rio n , see ( 1 ) , se co n d ed itio n , p. 13.

116 ] CARL G- h e m p e l deduced from S with the help of perm issible subsidiary hypotheses can also be deduced from S.N. by m eans of the sam e subsidiary hypotheses, and as A yer’s new criterion is form ulated essentially in term s of the deducibility of a certain type of consequence from the given sentence, it countenances S.N together with S. A nother difficulty has been pointed out by Professor A. C hurch, w ho has shown13 th at if there are any three observation sentences none of which alone entails any of the others, then it follows for any sentence S whatsoever that either it o r its denial has em pirical im port ac­ cording to A yer’s revised criterion. j

3. T r a n s l a t a b il it y in t o a n E m p ir ic is t L a n g u a g e N e w C r it e r io n o f C o g n it iv e M e a n in g

as a

I think it is useless to continue the search for an adequate cri­ terion of testability in term s of deductive relationships to observa­ tion sentences. The past developm ent of this search— of which we have considered the m ajor stages— seems to w arrant the expectation that as long as we try to set up a criterion of testability for indi­ vidual sentences in a natural language, in term s of logical relation­ ship to observation sentences, the result will be either too restrictive o r too inclusive, o r both. In particular it appears likely th a t such criteria would allow em pirical im port, in the m anner of (2 .1 )(fc ) o r of ( 2 .2 ) ( 6 ) , either to any alternation o r to any conjunction of two sentences of which at least one is qualified as em pirically m ean­ ingful; and this peculiarity has undesirable consequences because the liberal gram m atical rules of English as of any other natural lan ­ guage countenance as sentences certain expressions ( “T he absolute is perfect” was our illustration) which even by the m ost liberal em ­ piricist standards m ake no assertion w hatever; and these w ould then have to be perm itted as com ponents o f em pirically significant statem ents. The predicam ent would not arise, of course, in an artificial language whose vocabulary and gram m ar were so chosen as to p re­ clude altogether the possibility of form ing sentences of any kind which the em piricist m eaning criterion is intended to rule out. L et us call any such language an empiricist language. T his reflection suggests an entirely different approach to our problem : Give a gen­ eral characterization of the kind of language th at would qualify as em piricist, and then lay dow n the following (3 .1 ) Translatability criterion of cognitive meaning: A sentence 13. C h u rc h ( 1 1 ) .

The Empiricist Criterion of Meaning

[1 1 7 ]

has cognitive m eaning if and only if it is translatable into an em ­ piricist language. This conception of cognitive im port, while perhaps not explicitly stated, seems to underlie m uch of the m ore recent w ork done by em piricist w riters; as far as I can see it has its origin in C arnap’s essay, Testability and Meaning (especially p art IV ). A s any language, so also any em piricist language can be char­ acterized by indicating its vocabulary and the rules determ ining its logic; the latter include the syntactical rules according to which sen­ tences m ay be form ed by m eans of the given vocabulary. In effect, therefore, the translatability criterion proposes to characterize the cognitively meaningful sentences by the vocabulary out of which they m ay be constructed, and by the syntactical principles governing their construction. W hat sentences are singled out as cognitively significant will depend, accordingly, on the choice of the vocabulary and of the construction rules. L et us consider a specific possibility: (3 .2 ) We m ight qualify a language L as empiricist if it satisfies the following conditions: ( a ) The vocabulary of L contains: ( 1 ) T he custom ary locutions of logic which are used in the form ulation of sentences; including in particular the expressions “not,” “ an d ,” “o r,” “if . . . then . . . ,” “all,” “som e,” “ the class of all things such that . . . ,” . . is an elem ent of class . . ( 2 ) C ertain observation predicates. These will be said to con­ stitute the basic em pirical vocabulary of L; ( 3 ) Any expression definable by means of those referred to u nder ( 1 ) and ( 2 ) . ( b ) The rules of sentence formation for L are those laid down in some contem porary logical system such as Principia Mathematica. Since all defined term s can be elim inated in favor of primitives, these rules stipulate in effect that a language L is empiricist if all its sentences are expressible, with the help of the usual logical locu­ tions, in term s of observable characteristics of physical objects. Let us call any language of this sort a thing-language in the narrow er sense. Alternatively, the basic em pirical vocabulary of an empiricist language might be construed as consisting of phenom enalistic terms, each of them referring to some aspect of the phenom ena of per­ ception o r sensation. T he construction of adequate phenom enalistic languages, however, presents considerable difficulties,14 and in recent em piricism , attention has been focussed prim arily on the potential14. I m p o rta n t c o n trib u tio n s to th e p ro b le m h av e b een m ad e by C a rn a p (5 ) an d by G o o d m a n (1 5 ).

118]

CARL G. H E M P E L

ities of languages whose basic em pirical vocabulary consists of ob ­ servation predicates; for the latter lend themselves m ore directly to the description of th at type of intersubjective evidence which is in­ voked in the test of scientific hypotheses. If we construe em piricist languages in the sense of ( 3 .2 ) , then the translatability criterion (3 .1 ) avoids all of the shortcom ings pointed out in our discussion of earlier form s of the testability cri­ terion: (a) O ur characterization of em piricist languages m akes explicit provision for universal and existential quantification, i.e., for the use of the terms “ all” and “some” ; hence, no type of quantified statem ent is generally excluded from the realm of cognitively significant dis­ course; ( b ) Sentences such as “T he absolute is perfect” cannot be fo r­ mulated in an em piricist language (cf. ( d ) b elo w ); hence there is no danger that a conjunction o r alternation containing a sentence of that kind as a com ponent might be qualified as cognitively sig­ nificant; (c ) In a language L with syntactical rules conform ing to Principia Mathematica, the denial of a sentence is always again a sentence of L. H ence, the translatability criterion does not lead to the con­ sequence, which is entailed by both (2 .1 ) and ( 2 .2 ) , th a t the denials of certain significant sentences are non-significant; ( d ) Despite its com prehensiveness, the new criterion does not attribute cognitive m eaning to all sentences; thus, e.g., the sentences “T he absolute is perfect” and “ Nothingness nothings” cannot be translated into an em piricist language because their key term s are not definable by m eans of purely logical expressions and observation terms. 4 . T h e P r o b l e m o f D is p o s it io n T e r m s a n d o f T h e o r e t ic a l C o n s t r u c t s

Yet, the new criterion is still too restrictive— as are, incidentally, also its predecessors— in an im portant respect which now calls for consideration. If em piricist languages are defined in accordance with ( 3 .2 ) , then, as was noted above, the translatability criterion (3 .1 ) allows cognitive im port to a sentence only if its constitutive em pirical term s are explicitly definable by m eans of observation predicates. B ut as we shall argue presently, many term s even of the physical sciences are not so definable; hence the criterion would oblige us to reject,

The Empiricist Criterion o f Meaning

[1 1 9 ]

as devoid of cognitive im port, all scientific hypotheses containing such term s— an altogether intolerable consequence. T he concept of tem perature is a case in point. At first glance, it seems as though the phrase “ O bject x has a tem perature of c degrees centigrade,” o r briefly “T ( * ) = c” could be defined by the follow­ ing sentence, ( D ) : T (x ) = c if and only if the following condition is satisfied: If a therm om eter is in contact with x, then it registers c degrees on its scale. D isregarding niceties, it may be granted that the definiens given here is form ulated entirely in reference to observables. However, it has one highly questionable aspect. In Principia Mathematica and sim ilar systems, the phrase “if p then