I

To my parents /&,.a/eSe *t ,int:zf,ea zt""u e/Z/ .'

,l

7o ,eocw rtof p rv@ar & 6{rera{ Zou-.e 1*cr?74./ie (,t:*/ftt,'ea/ {eafarts/ fr./, &./' o{ a'/,2,/ lor.+lt ria. /n,fV "/ 7

Philosophy of Sci e nce the in Twenfiieth Century

i ^. , r . 'r r t-ouri",entral Inemes t-I

i'
i 'io"'. ; tli I

'",

L;,onaldGillies

if

t.

BLACI(WELL
Oxlort UK & Canbridge USA

-1i i' : .1,

i!,t
2

Copyright @ Donald Gillies 1993 The right of Donald Gillies to be identilied as author of this work has been asserted eccordancewith the Copyright, Designs and PatentsAct 1988. in Firsrpublished 1993 Blackwell Publishers 108Cowley Road Oxford OX4 fF

In the research worf,q of all branches of empirical science this spirit of a seientifc coipieptionof the wortd is alive. However only a very few leadintri thinkers give it systematic thottght or advocate its princip:4iNeurath b5al., The Scientifc Conception of the World: .t., The Vienna Circle. 1929. t
il

UK

238 Main Street, Suite 501 02 Cambridge, Massachusetts I 42 USA All rights reserved.Except for the quotation ofshort passages the for purposesofcriticism and review, no part ofthis publication may be reproduced, stored in e retrieval system, or transmitted, in any forrir or by any means, electronic, mechanical,photocopying, recording or otherwise, without the prior permission of the publisher. Except in the United Statesof America, this book is sold subject to the condition that it shall not, by way oftrade or otherwise, be lent, resold, hired out, or otherwise circulated without the publisher's prior consentin any form otbinding or cover other than that in which it is publishe{ and without a similar condition including this condition being imposed ofl the subsequent prlrchaser. BritishLibrary Catalogning Puhliation Data in A CIP cataloguerecord for this book is availeblefrom thg British Library. Library of Congress Cataloging-in-hftlicationData Gillies, Donald. Philosophy ofScience in the twentieth century: an introduction / Donald Gillies. P- cm. "" Includes bibliographical references and index. . (pbk.) ISBN 0-63t-15864-2. ISBN 0-631-183s8-2 1. Science- Philosophy - History - 20th century. I. Title,

A certain type of p.;inicillium produces in culture e powerful antibacterialsubstanlb.. . . The active agent is readily filterable and the .r.^. 'p.rriiililirr' has been sivin to filtrates of broth culturesof the moulh. . . . Penicillinls non-toxic to animalsin enormousdosesand.isnon-irritant....It is suggested that it for may be an efficient:iantiseptic applicationto, or injection infected#ith penicillin-sensitive into, areas microbes. Fleming, BrilkhJiurnol oJExperimental Pathology,7929.

4.4

. (..

Q174,8.Gss 1993 501- dc20

92-36318 CIP

t.

{ir
.':. t.

Typeset in lOVzon 12pt Bembo by Bcst-set Typesetter Ltd., Hong Kong . Printed in Great Britain by TJ. PressLtd, Padstow, Cornwall This book is printed on acid-free paper'

t
I

,/^vr Z.ul"aa zt c^*fuie. ./ o4e.z,z/c)64'oto.rs/rat,,, .] '7hL"7hra a-atJasS e- 'tantozbt ,' 4r."Fy'. J>c, ec aanf a'l|' <.62fut/*zd ] .. at *z Jz at>*a #4 & .t'-'1/? ll, ho *c sa aa'fifr,h 'fu 'A' .**kaa ea?Jo.aJ'2651" Z--4.. 7ra Jel 4*; hs &2'lez yd t I l/ zae. -.v.t (tt *r>aAta //Ef-/fl/. 4 a,tu{ 2 lLa/cte/ /,/b+e.r;4/a/J | *.zeazb !: 5 72/ *,-- a-y,.^f.- =d )z aZ. ia* aoeLbJa,4 | '*+4" iv i e.r erorc-t>2e4,2a1',:t/<r -tl za +F *t A -+#a/v /d ri:Cgntents 7en2{r'

i"t6.

a;';-1i7"---q/

,va,-izp

q-fte4tz -d

lu"

.np*,.,s;

/*9 -":"

zl a1'//a12

&"8 z'Ja7 t'< az '-

{Hq

^I oo*zt/z+.. ne ,4"2 /g6-./e,*'. ,

q
,/

?2./ H.2'9o'4

/*e

.;"at7/,4.'< t'7 ?t*'

,4 e'ottt'"/

c'' -

., ::--.''-"'-

'''

7z'W
_ -#l '*/

/*W'
prcfrcc Preface

7x*aee<" '

.1./, C.,rit o,-7 .

.i

-"/ Lz ,tb// ."fi fl*zy eeq|qaa ' /.a/ -a 4La l)4/.4/ /-(lZ,"U*4"/eh.U.N4 "ta'A,z /+;;..r.,*:/;7;';;;';-4in-n V*+., e".Vy.,z4:i,A'".f2*,t

Acknowledgements Acknowleds.ments. i:

t:

-..-J

caatobu*Fr! * . *l*.i

Prd I Inductlvl.h.nd '& crttrc!

. /a.:4t ' ;aa"r,e./c ;.a//... V a,. ,v,1',2

."4*z'.'+;iz,r" /"r-c./.h' tee't an '*W;;&:dr{TWh:{yHW' 6c't*,/.*,-'-J"*,ruz -1:ate?::?: . sao;t tt ni'v^*-Cl',i.'i7l.s"iw r*^i,*naou a/*ficz *-*-;4d),6cl"t -- 4
?:.'4.;:4?2:'1

a 4f-..iz,t':'.

',;1,/rcat

tD' u'hbiidst school thcvi'no' circr. .nd PoFF '..-*.n

I *e**-,.,i0:rrductiviim.Ru!,cn.nd
3

fia,,'t 6 a,xa,?,/ t / p1t-V #*-t,/./ C/-,'rosnl h 77.j'2 aa{' .r W J'fr.-f . aoo,r.:oov.r4 a/.aac 21a7 ,3-/ z+ zZz 2,r/"12///, ' 'Y 2 --4.a. -*t )^ t 2a. iJ 47frf ^:/z>'6a,'/tr "+ '''4f,*' -nd"4

':ievA.i;K#:-4,3#",;*;f
. zTsq t*J,baf {..h?e Ad4., plt*e lpcct>gE t-Aa''uz nz.a./' .-/4^ -/ eq*

, i;. ;;:: ; ;' ; ;: ;;;;

;*/;'- f "zia1i'a"t';.T.'A! 2/,' &:3ffi::tr::#:H*iffil"'

oftits.olth? Plar.t' 2.6 Fd,s;;'Di;:o;dr d aE P"ti'i in:cndtiwITiluction 2.1 ibDkae..roIt. 3e 3 Durcn',critiquc;illucr,vilm
'7
"tttPhotu6'& Dn$iMctJ'antdl ot Baeoni.n ttuh..tton 1E

" .t-\ductivisrn. rhcoryorcoie.rurci Hi! ' l_"1f":-9-lon': rnoKerutauons lorrr-rk'Iicdtionilm) oJ Ind"tioBn 2'2Pow"\ in . . ';:.::f:.91:!*i

Poppt2l nH"*:::4 tN ut'P'6at nn:"20 ci'c'. 2t oJ vi.n'a ^1.1t th'

n<*- ,o to&,4.'tt )'-i'{-a'-# ,t 7*'* 4 )*r"alt?.ii . ,atlo/az'*t ./ e,r".ea.t ."))i ii,la" -;;; -;,6.*2 a/& : /.k, '"/ i-.,t r.i, + /7 a@ry. aava,"/t6d 'h .4...f<
-tf:fa

. rytY. .r .--Z 4At/ /"a i6r' *z#* Z

r:,affi:;,Wf U!'W;ff tJ#i:#i"il"a;J t lwn,sm 6 tn. N.vtonidn M..t'ot 51 3.2 N.vLo 's Kcplt\ Laws dnd 'Y:::.!il'y'"{c::'r!'6

54

.*/ 5"2'9"4" 1)'oJrii.','ar zy".ou./&'-+z;at3t;E"f /" al"e-.n

' ht/vca/. L *1a.at*-<c
Zd, ue ,a..<r'..o/

";:ft#i{';ff::-"ir"#frlygmy*
,: ; ,tj,*_i t*iii.';i,i,t'i,ail

6c <alqae4

-.t" 4

ada-/F'd'd'<2

"t Jaaica.>rc,/*cc.. *''etuet

'of

'

ffilt

: (h rc#sot$/ozat-tw*7fi;n2;tiS..-.*'are;
, *' u,tn 474el'ca 7&- ! citetll4 t'"-Afl

:
.,

*i
j:

F.

;f trt= s .'

Contents and the Duhem-Quine Thesis Partll Conventionalism 4 Poincard's Conventionalism of 1902 4.1 Kant's Philosophy of Geometry 75 4.2 The Discouery of Non-EudideanGeontetry77 4.3 Poincar1's Conventionalist Philosophy of Geometry 85 4.4 Poincari's ConventionalismanclNewtonian Mechanics90 94 4.5 Poincarton the Limits of Conventionalism 5 The Duhem Thesis and the Quine Thesis 5.1 Preliminary Exposition of the Thesis. The Impossibility of a Crueial Experiment98 5.2 Duhem's Criticisms oJ Conventionalism. Hk Theory oJ Cooil Sense(le bon sens) 102 5.3 The Quine Thesis108 5.4 The Duhem-Quine Thesis712 The Nature of Observation .,
111

73
/J

i
It I

Contents

9.3 Duhem oni*oppr on theInfuenceof Metaphysics on Science 195 914;.Duhem's Defence Religion of 201

li

10

Falsificationisqii1n Light of the Duhem-Quine the l Thesis 10.1 Faki,frcatiittism theFakifability Criterion 205 and 10.2 Existential'Statements 10,j Probability 206 Statements '''10.4 FakifabilityaniltheDuhem-euine 207 Thesis 210 lc:}l A Suggisted Dinarcation Criterian lnuoluingthePr'fttciple Explanatory o! Surplus 214 10.6 How Muc'i:ofFakijcationlsm be Retainet!? can 227 10.7 Sotne Confiuding Philosophical Remarks 228

Part lll 6

Notes References Index

231 238 245

Protocol Sentences 6.7 Carnap's Views on OhseruationStatementsin the Early 1930s 120 6.2 Neurath's Views on ObseruationStatements in the Early 1930s 122 5.3 Popper's Views on Basic in Statements 19i4 124 Is Observation Theory-Laden? in 7.1 Duhem's View that AII Obseruation Physicsis TheoryoJ Laden 132 7.2 A Reinforceffient the Holistic Thesis and Neurath's Principle 137 7.3 SomePsychological 146 Findings 140 7.4 SomeCeneral Conclusiotts

lt9

132

Part lV.. The Demarcation between Science and Metaphysics

ii

1s1

ii,

Wittgenstein, Vienna the Meaningless? Is Metaphysics 153 Critique Circle.and Popper's Problem itsImportance and The 8.1 Introiluction: Demarcation Life 153 8.2 Wittgenstein's 157 8.j Wittgenstein's 172 TractatasT65 8.4 The.Wnna Circleon Metaphysiu 177 Critiqae'ofthe ViennaCircleon Metaphysics 8.5 Poppgr's of Later 8.6 Wittgenstein's Theory Meaning181 8.7 The 185 Life Infuenceof Wittgenstein's on his Philosophy in The Views of M6taphysics Relationto Science: Popper,Duhem, and Quine in to 189 on 9.1 Popper Me.taphysicsRelation Science 192 9.2 DuhemandQuineon theStatusof Metaphysics 189

- ltt/t,;/
| ( ,,s //p
ttt';.'.t.",(4 (

sieiuleq (

. 1,ff.,'/.1=
t/

(/ \,

'2.

/kraatt a2r'.a/,y. / VrNPle ' /

norlrL to"z rTc aJzt w'4ts z1'' ,/,rt A - 4rE A-s q //?u/ct cd z+tcc.^tAc/t./ apel*- /no'a{<.1../v .4,/y'td. /.* wy'Z*q ao/.zpqu/.

,/lt'r/,*yt-.2ts.,y.7>,z,pl !/,./')t^7V2"2./"ara.oothz'*"/'*',{*''a'16'ac//fi'(/lff'u ' " 'J

/ / /'/,V L"*.n o"o,*/7

H T
:

: ' lr a*tee vie/o,/, tt,o4,i ola .a,/,r#4//.a 7.1."." *l,l-i,*nll;,tf

At*t;,r ' i'rne" alt"'r'd ,rr,l Jh-lU { coalclrtula,41 ,la/"t r/*/ m,.aa/ a,z./' .zzatrr a/-e a,uur72a//y

Vez,,r,

/Q'.nL'e'ac

'*ay' bdrrt'/a"z

4L'

**$*:ir;1T;[tin:*

everything, and I have-;ccordingly selected what seem to me to be the most central themi$. Inevitably there is a personal element in this choice. Other wiiters on the subject might perhaps have selected a slightly different list of topics; but I think all would agree

0u, n t{t

aLli+n,1,/, /,y',t/ .',"

I
;

roe^ /7,o1 oaz .r24 tr/nr, |r"trnt,(-t"/r/d Za/L'a$t2A*4 ',1 ut r//. fuc ,6ohl a,zz nrlf ,+F F 4)'oaI eta,/' aar .hQrna/ ftrc /(tr."yaua eatc / z4 -+t lt t'tf, . .411ror/.7a2fu-. /u-v A ./' 4g -r

,,z"u2r14,/ ,,itr,t .*n ,7,+o/11,, -*./-i

? ,tl*la/l'4e t-.ezh /H/zta il 77c,/ Aey wee.z.Z?r.y A",{/ lty'ect.ra *'./n,n4 Yt ,1oze, ot4,7 ao.d/a/.Lv -e/,/r7e."e&f

4L J;';

4- *',

a.bz?oczlia iillfY,r",z2

t-+?:'il"Tu, r hope, .jrr,,,,. some things will beof interest that to

that the four central the;.nes discussed here are indeed of the greatest thosealreadyfamili", *ith the subject,6ut it is also intendedas an introduction to philcsophy of science. Indeed, the historical

--/",rlir4y""J

F F
I

,}kJ;;1/"r ,'{,n,;.yi.;:a,7)"-t}r",/".g
,S'r,.1acrt.'4t ,,nt/rrt,/ilul{ ( ZerE;*zz/;6/ .efulo *t -p.tc4 *' 'tn /"of( "ft1t/ OXlerztt"Vfl.-,

) #r lle

cuifr cz41 6.X

r;?, ;ff;:: 4ili'',1';.T1',':ff:T:l fr:Tfi,:;fl,.:i'',.;;:'X,:':i; alwayseasyro follow, iyut at leastt have tried io arroidireating
any artificialbarrierstoun;ierstandingwhichmightberaisedbymore /v< oithe piilosophy of science. Ideasin the philosoph,ri science of thus becomeempty unlessthey

throrgh th" rt,riy of tl.c developmenrs which led up to them. Since the book is intendedasianintroduction, I havenot presupposed any previous knowledge of philosophy or of science,and have avoided,

ilf,*i;'.n'*i'"f,1,,[.J;i5:#JiJ",':il','r'*':]f ;]fi;|il]

'/ic,t )t ., ,/(t,a,/ + 7Afu2ft tu,*?f gl,,z*.+aa ,.>yeztat -a-t",*'h)[!:fr::Ti,n:::ftril.';f,,il:f,?:liil:t::fJ[Ttdn,, 4vzee' rt aa /d4?vft z&f'frc.t i,1 gives rise,cons_tirute tht,;ubject-marter
7('s41ttt,t/ nr?/n/'4J' /rr*/p u.1*" t,iy.l,tr/ " t!/ ,/ nll ." / *at/ *a,/ e/c.?c4?&Jr'zfiZerqta :7f 244/L 6.vrhr?race'ft<l/ i,t il{ ' ia.,a.'fua/ *"lf:'o,i*rii -acy' ,rfta tl i, Gc va-re76.+6"7/&'t ft, 7l /n//a.e,trtt4/ 4/ hvre'r/ more historical scierlie, in order to iilustrate the philosophical discussions. Theseepisi;iiesinclude Kepler's discovery orthe eiliptic

F F h
Ir

4,t''6e,z:af:a ,t/:nt,/,ecq,tle n( -44,,--,'riyq*-I,/fttt,t1,/v(G'vl,,./2.f.t(,/,/.n.7'.{4a4,i"2,4a4*1r32-afi'r.-.--'-.-o

,,Qij .'a,t,ytt/ lt/rie d,?2e: --->J&-xx7 p./Q

;i:##,'';:,:j*T'ff.*:Hli":Til';:TLjr:'*1TiF;:lXTn

i,
il

xtl

Preface

.t,
1l

theory by Planck and Einstein. Although these examples are introduced to make specifip philosophical points, they have, I believe, an interest of their own. Most intellectuals are more familiar with the works and lives of great artists than with those of great scientists. It is understandable that this should be so; yet the achievements of a Kepler, a Fleming, or an Einstein are as remarkable as those of a Mozart, a Proust, or a Michelangelo. I have included in the book notjust an account qfthe ideas ofthe philosophers, but some details about their lives as.:well. These were often surprisingly dramatic and interesting. Philosophy of scienceis not such a remote and'ivory tower' pursuit as it might at first seem. Questions about science impinge on both religion and politics, and are therefore liable to arouse all the passions which these activities inspire. This is clearly seen in the fate of the Vienna Circle, perhaps the most influential group of philosophers of sciencein the twentieth century. The Circle in fact met in Vienna only.for about twelve years, from 1922 to 1934. ln 1934 its leader, Professor Moritz by Schlick, was assassinated a Nazi student, and the other members of the Circle were forced to flee the city. The book has taken several years to write, and I have benefited fron the help of numerous friends and colleagues. So as not to overburden the preface, I try to express my thanks in e special Acknowledgements section which follows.

',:.,

Acfrnowledgements
t

; i' +
i

i:,8
. r!

Donald Gillies. Departmentof Philosophy, King's CollegeLondon, Strand, London IVC2R 2LS .lune 1992

Anyone who writ$ a general book like this one needs a great deal of help from speciai$stsin different areas, and in this respect I have been extremely ford'r:nate. rg19tu11sn i4-plrysics, and My discussion of'the gwgngieth qgr1-tur,y_ more specifically {{,the science and philosophy of science of Herrri Poincar has ben6,tited from many long discussions with Jerzy Giedymin and Arlfriur Miller. It should be added that these two distinguished schollts have quite opposite views on many points. I have tried to give.f"'orne itrdications of these differences, as well as references to the itelevant Papers and books, so that an interested reader can learn m:bre about this fascinating controversy. ln my treatm"tt: of Duhem, I have been assisted by two distinguished Duhenqqscholars, Niall Martin and Anastasios Brenner. Niall Martin I harip known for very many years, and large indeed must be the numtfpr of conversations we have had about Duhem. But I have also tJeen able to use his recently published .book. I became acquainted:with Anastasios Brenner only through studying his recent paper ar*d book on Duhem, but he was kind enough to read some of the i$ctions of this book dealing with Duhem and to offer very helpful iomments. My original traihing was in the mathematical sciences.However, a former colleagu$; Hermke Kamminga, not only read through some early drafts',bf the book and offered useful comments, but other contexts, that an adequate philosophy of insisted, in this anrj5 science can be achieved only by taking accourlt of the biological as well as the physicil ,sciences.It was her exhortati6n which made me decide to includei some examples from biochemistry. Another colleague, Melvin parles, led me to the examples of the discovery of penicillin and the slrrlphonamidedrugs. He gave me every assistance in my task by lenciing me books, reading through earlier drafts and
i ::

!i ,
.j

--

'it.Ii:.

.t
XIV

Acknowledgements

canre to t,lk ro our seminar.peter Mitchelfs interest io tlr.-fiiior"pfi ;;,#; ;* great kindness of led him ," ;"il;f.;rouble to explain some detailsof his chemiosmoricf,yp"tlr.ri, 'earr iol"*rrl.. like read and comment .1 *f arig air."""f^*amples, myself and to He was also kind enough to invite.me to th'e meeting,lo silver jubilee of the laborar"ryl, -ciy"""*i... O.tob.. 1990for the h. carried out the researches which led to the Nobel ir:iir. On this occasion I mer JosephRobinsonand Bruce w"i"., d"in LiJ.rt.-isrs wirh a strons interesr history and philosophy;f in ,;..,with whom I havehai mr'rchusefuldiscussiona'a ."ir.ip""t"".J--i, *", with greatsorrow that I learnedof tl.resuddena.rrii"ip"r.. tvlitcfrel As regardsthe resultsfro^_"_pi.l.ri-pri.r,"rogy in April 1992. _i"ii.r.J r, my discussion the narureof ob;;;;i";,I of *", grertly aided by Richard Gregory who made _r"y ur.frir,igg.rtioir-;; -..", ;;;;.j rne with rhephotograph usedAr'pt"r.-+. I he general structur.e.o.f book emerged the the as resultof a series of long and rnost cniovabte ."""";;;;;;;?^)it c;urio Giorello, who mrny heipiut ,ugg"rri";, ,p"jn. points. Nearly all :,lT marerial the -id.. "" has bcen pr"..it.d ,, f..rul..,-boili-;; phffiJJ; srudentsirr Lord.rr U'iversity,s i;;.;l;;;. recrureprogramme arrd as part of p'ilosophyscience studentsar Kirrg's College London. I "r,.i.n..-ioir.;:r?. have b.r.fi;;J;; many asturecom_ rrrt'rrts f-ronr these lively and critical,rl;;;;' rvly owrr vrcws on history and philosophy of science have been rrruch irrfltrcrrced over years by the'*l"kfy seminars in the the. srr'jecr first at .hersea coreie ."Jid,"rf;. irs nleiger, ar Kins,s orllt:gr l-orrdorr' we hav^e"haJ-;;ffifu array of visitine sltclker.s, irclurli'g solne th. g..rt.rt-rr"_., i" both science anJ -of :rrtl l,hilosophyof sciencE. y*;fi;r.ussions, ll:lt both in the srrrr:lr itsclf ;r.cl over coffee ,ft..*r.,ls,-"iave often been r'e. ,t.-r.f".r,-rrii*."ra therefore as llfl:"l"lr,"i*,rs papers liketo s:ry :r s'<.cirl w'rd of t'anks to the s.minlrs-.r.g,rlr.r;*ir;';;; rct:.grrizetlrc irrffucnce s.nre of th.i, o_n'ii."s commentsin thc 1r;rgr:s "f wlrir.hfollow. "rrd f:irrally.I slr.uld rike to express my thanks to the followins <.'y'itrq'r-lr.ldcrs[<rr t'cir ki* p.r-i'rr;;;; l'l;rte I is original phJ"g;.ph^;i irs ";. il";;;#r: p.nr.ittin culrure pl.tc, ;rrrtlis r.qr*xiuceawit. the p"i-ilri"" 'lt:rrrirg's St Mary,s Hospital Mrrlir.;rfsr:lr<rol. l)llrc 2 cornes frorir M"; B;;k"i ,q"ii,inirlii,r^ri

to important papers. I was also very forrunateto nieet it. trr. p.". fv[ii.i.if, "Nlili prize-winner for chemistry ;n f SZA, whe; he

correcting mistakes,.and drawing my attention

current is reproducedwithithblpermission of the publishersof the Inc' Plate3 cbmesfrom Richard Do.,e, tirbli.ations, &;i;iri;", with the permission E;;;;t;r -f;",fi it e Intelligii,tEvt, and.is.rcproduied Weidenfeld the author",irid ihe publishers, ,f _George ,and fn:m a photographof RichardGregory's' whtch Nicolson. Plate4 is me permissionto use' he kindlv sent me "tt6tliave

r* ir Acknowledgements t

il

*f'^f:rtr:ffi
'a"t2aYd'6;ee? i-;,Tr"A7'Yffi eaaeazrercd,,Too 7fu/'( dd [w
o-5/ aAc6l .'o,o6ui

- (.o Pelt, lJLb,:l;ff*::#; clr ^"

V* tde.zrl'ff ,"aiLt);' o'i^"oa/

A,.,r ,/r+",Lrtrz''/,7;,r/7ye ! 7fu ;Jear1'fr freuy { c-a doel /;;;;; ';i;;"fr:/-:!^:^9,;Y Y: h' ,fr, ,oi
;*H\ry
/'t *{"'e/ /o1e"nst' "/ "n )a-*'i')'*"d ':" #L,?* rL/a /'?4, @/e odJ-)"*t vtctd7y su<7'z+e1+a( L* )"*.#rlu/ :ft";*ffi;i:
/hq/

aa-Ayryc- z.)#r/"try*'rtSlo;

')o)t

t#*

7;/,"/ trll;4 fl,ffZt,Iff i,'L,+7'7'*" .4*< H:ffi"r.

- e"'dw!' i"*>f /":"'f JL Ae* *n fod)i;fJyato*/,/"y'7".'fK:',:t: ala*cl n fr .ea' e*ste&zot,&/ .
T4Ea2t//eo.a"rd*+/ *f,": biaera '*4zttad

'+-

eftez4rel+frd ''7 rtu star'a !&'f* ;,47;6;/ aol ^ao"e J-)rT2il'"i'r' &*f zze'E.a#f (""e "'r edvc4o/apr.^, *rES
// - -j^r*-^t."/"i-rc,aAb/;reFse-reav4|rr? ""*

::;:7;"J ;t' 'l 1r"x"?, #tn3;u T;i^*i

ro"t#

i,

,'i ,l i; rl'

.l '
t; ,,i

Ino u0t ivisrn andits

Ill,.
Ir

t'

PartI
Critic s

I i'

t:

.!,

!,n

I ;' +'
i:

i
ii
4 !:. 'a .,1 :;f,i .71, : .1!

i&

i i.'
.l

ii'

rj

di&".p/&q{t

j '.

'. *t A"e,b c'4,2'l 4^w rhi *4o i '- .t&prr$rrc "/ t ^/"f:9 ";464/ / 4 ce"tJw)uvDr"t"t 7V,l ;',t 't;frrz'l-/r 74'-e^n7/g '', ALrS ht't ,4ord -od rc'/;d un{a'fro * Aomci 6V d""**'t ,roTuot mar4oa. Zu dfl^coelte*aj"fr*td&?Aft 7'*,o 4:'n/"" ftu ofiecr?, un,y'.e .T'"a{ <;a Tayc+za/ or"tr{uu*dt "f i-nu )c{en/iru *n'/*"7:;; "*-/r4,* -6 *t! *'f Ata'*t *tv tiwrluk J A**;o/. ,# *, s6"XH 5o aq 4t a<qffir' fi:fi a*rd ftt *%rw7{vz 'J .) -*"oi6.r.'ar a// aot'r"-a- k z6 -;;r;" eonldot'tJ Ar "*ne {5* 'ru,t'"*4ae{ e-e/&64t lo'nu n' evf&acE *"e// 6ff;;
a/V d-AtQfV .

.;1
i,:r .,4 li. l
' :I

rl

Background : Some H,istorical

Indi:tctivism, Russel I

Circle.afid Pbpper the Vien;ha

:i; f, i tA1 .tr

and thqiCambridge School,

*rJ

6rafq.

iVr

'1.1 lnductivism

ii'

its The first rheme to bp dealt with is Inductivism'and Critics. Ildttf t_i11iyn a theory 9d'"sci911ti!c*metho.dis w.hich was developed_lgng before the iwintietH .cen-tury. main featurebwere iptro-diited in llts it *as f.rhapi the the- JiivEniient[ i".i:,"eighteinih centuri"r, the dominant account cif' scientific method in ".rd nineteenth centurfl

inductivism andlv,ith some of its supporters in the twentieth century: namely,1.8i$r[S and the Qam!-rj{g-.--s._bqpl and the V-igrga 'i" Circle. inductivism ri'oory of scientific method goes back to Francis ", " Bacon (1561-1626);',Bacon was a contemporary of Shakespeare (1564-1616), and the'ie is even a palpably abs'rid theory that Bacon wrote Shakesp.are's..iiterary works as well as his own philosophical ones. Bacon came.from a family with connections among the great and por,r.erful of the lland, and he entered Parliament at the age of twenty-three intendi$g to follow a political career. He did not meet with much succes$ a! first, for he never seems to have gained the favour of Queen Elizabeth, despite her taste for intelligent men. fiis fortunes improved . under King James, despite that monarch's

--/-:;i;lF;-: . 'ri',J I i

.iiLrtliri'

r: ' J

lnductivismand its Critics

Backqround S' Sorn"Historical 'i' he outs this idea ti'follo*s' 'Human knowledgeand human power is the m."t in one; for ',t{},rere cause not known the effect c'nnot be obeyed; and that be commanded must produced. Natu oroduced. Naturej:to be commanded must be obeyed; and tha
*hi.h in .ont..rrplction is as the cause is in operation as the rule' ^$akes the same distinction .in another b. 259). Bacon -1{. ly of with experiments light' of fe c'.lls experiments *hrt

it
i;

predilection for handsome, bur unintelligent, nren. Bacon becanre Keeper of the Great Sealin 1617 and Lord Chancellor in 1618. Only two years later, however, he was prosecuted for taking trribes. Bacon did not deny the charge, but claimed that the bribes never influenced his decisiorrs- a somewhat curious defence, indicative of Bacon's philosophical turn of mind. Bacon was condemned, and forced to leave the public arena. He devoted the rest of his life to philosophy and scientific experimentation. The latter activity proved fatal, since he died of a chill caught when stu{fin$ a chicken with srlow to see if it could be thus preserved. As so often, Bacon seems here prophetic of methods which would later come to be widely used. He wrote a number of philosophical works, but we shall rely nrainly on w-trQt,js perhaps his most famous: the N,o,ua1n'O1ganun, published inll620.bnd dedicated to his patron, KingJames. Aristotle's collected writings on logic had been given the name Organ.um (literally 'tool'); thus Bacon's Nouum Organum, or new Organunr, was intended to supersedeAristotle as an instrument for reasoning. Bacon.was a man who loved money and luxury; but he was not selfish, and wished the benefits of comfort and ease ro be more widely extended emong humanity. He saw thar dris could be achieved only by the improvement of technology, and thus became one of the first to argue that scientists and technologists should be praised and supported. As he says: The introduction of hrnous discoviries appears to hold by far the first place among human actions; . . For the benefitsof discoveries . may extend to the whole race of man, . . . Iet a man only consider what a dillerence there is between the life of men in the most civilized province of Europe, and in the wildest and most barbarousdistricts of New India; ... And this difference comesnot tiom joil, not from climate,not from race,but from the arts. (Bacon, 1620,p. 300) Although Bacon's aim is largely ro improve technology ('the arts', as he calls it), he does not make the mistake of sonie modern politicians of supposing that the best way to achieve this is to fund only projects which are capahle of yielding practical applications in two or three years. On the contrary, Bacon says explicitly: .For though it be true that I am principally in pursuit of works and the active department of the sciences, yet I wait for harvest-time, and do go1 attgmpt to mow the moss or to reap the green corn' (p. 251). He believed that in the long run technology could only be improved by improving our knowledge of rhe narural world, by ..r.yirrg out basic researchin science, as we would now say. In a famous passage

.o.r,.rriirrg A s he says:

fruit

All industry in eiperimenting has begun with proposilrg to itself and certaindefinitew$'Ls to be acc<implished, haspursuedthem with it eage-rness; has sought, I say, experiDrematureand ur*easottable m.nts of Fruit, tibt experimentsof Light; not imitating the divine light only and assigned whichrinits first day's work created orocedure, io it o.re entire ddy; on which day it produced no material work, but proceededto drat'i:n the days following' (p' 245) Bacon was convinled that basic research in the natural sciences, or 'experiments of 4ght' as he put it, would reveal all kinds of phenoni'iira which could be usecl as the basis for new o,rli.,o*" technologies. As $e says: 'There is therefore much ground for hopins that there'fare still laid up in the womb of nature many secrets*of excellei;t use, having no aflinity or parallelism with anything that is ntiw known, but lying entirely out of the beat of the imalination. #hi.h have not yet been found out' \n.ZSZ1' The l"ast 370 yq$rs have shown how well justified was Bacon's advocacv of scieniiiic research with the aim of improving technoloev, *hil. his h6pe that 'there are still laid up in the womb of of excellent use' has been fulfilled to an extent ,,"i.r." t ,rny ...t.ii ereater than perha$ Bacon himself could have imagined' " Br.on *ti .rot;-{pntent with urging that more scientific research should be carriedl,out. He proposid a method which, if followed, would in his view [-esult in an expansion of our knowledge of the natural rvorld. Sorne of the precise details of Bacon's method - for example, his Tahlii anil Artangements of Instancer- are no longer of great interest. Hofi7ever, his general approach still has supporters Ioday. It is this geh'eralapproach which I will call indurtiuism.1wlll next turn from thd;particular case of Bacon to a general exposition of the 'main featutds of inductivism, but it will be convenient to illustrate this fronriiime to time with some further quotations from the Nouum Organil'a. The basic idea rif inductivism is that science starts with observations, and mone. oiifrom them to generalizations(laws and theories), and predictions. G-ood scientists, following the 'inductive method', f,

i;l
ii:

lnductivism its Critics and

i i

tl I

j A",i rolo g ;*'i !:f"jffi ifi:i: "r I "-"" liol'r' g'J' aerthe win
Il: All ravensare black. S: All swansare white. It was .btained, accordingto the inductivists,by observing a larse rrrurrber ravensand noting of that they were all brack.It wrs thJn irrfe*ed that arl ravensare uri.r.' ;rJ'r#i*tt;" the case swans. of l{owever, thereis ,n inte.esdng I#;;f*ween rhe rwo cases. Il, so far as we know, is S r.u" by 11e trp till t'e eighteenthcenrury. 1,i, +;;;h;",Ji"".*pror.rs Etrropeans Bur the .'rrii of Aus-tralia observedbtack swans, *f1;:; ilj .h:l* !i. leneratizationrhat all rerutation

nrerlrod, which he calledInterpre;;;;[^:f iature. As he says: .The conclusionsof human.reason as ordin"arilyapplied in matter of nature, I call for the sake of airU".tio"-I thin5;rashor premature).rrr"t *rr"";*;i, nltiripotiowof Nature (as a eliciredfrom factsbv a jusr arrd methodical proce-ss, ."ll t tntirprAiotionof Nature, (Bacon. 1620, p. 262). As w. ,hn'-... \;-;;'';:::r. chap.ter, ',;;.r.rparrons.popper, in consciousopposition,ro Ba-con, ;;; . . rash or prelneture'an irrtegralpart of scientin. *.tf.oa. Ler us lrext rllustrateinductivism by rwo examples.The first is concenredwith birds,-and is a .""i dv;; among philosophers

Nature,r"J.on,.Irr.di;-;;*;ffi ,;"#,Toi::*::,,i"!;::L:!

will begin by making.a .urnber of careful observations. "g"i.r"ti""tion They wit trren cau-tioJsry.ra1se i"d. ."*. from the data collected, and oerhaps make a prqai?d* on the basis of this generalization. rn"_. i"ar.,i". ;J;;'T_"'rtrongly advocatedbv SherlockHolmes. Thus, in .ffr" S.""arf i" to v/atson: 'It is a capiral-misakJio-;h;;lr. Bohemia,,Holmes savs before one has data., Holmes'smethod is to collect-tfr. a"1",'riirhen infer from it what happened. This is typical .f ;h.-;;;;riu. characteristic induitivism r;;;il;;'rp..,rr"rio,i ,pp.o".ti. It is atso of in the absence of data. Indeed, Bacon named;;;h^;';;;

..:, 7 Background ;iomeHistorical discovery Kepler's one will beginhereby deilcribirrg suchexample:
that planets move in $llipses with the Sun at one focus. The story is briefl'li'this. Between 1576 and.l597, Tycho Brahe, * Danish astronomer, &ade a long series of cireful observations of the heavens, particularlyr'f the movements of the planets. The telescope had not then been irrEented, but Tycho Brahels obseivations were the most ,accurate e.|Pr made with the naked eye. ln 1597 !e left Denmark to becomef.llmperial Mathematician lat the couit of the Emperor Rudolf II i.ft Prague. Here in 1600 he took on Johannes Kepler - -ft.pu as his assistant. task of woiking. o11!43q orbit of Mars from r.irti-r.rr$e take him a week. Tycho Brahe's data. i'\t first he thougfiiif-*""ia In fact, he worked.c.*r the problem fsr over six years, before concluding that the orbii was an ellipse. He published this result in his AstronomiaNoua ('Ndw Astronomy') in 1609. At first sight thisr,appears to be a classic example of inductive method. Tycho Bra$e painstakingly makes a long series of careful observations of the irlanet Mars. Kepler in an equally careful and painstaking fashion lrrfers from these observations'that the orbit of Vf shall see later whet an anti-inductivist, such as Mars is "n1llip.". Popper, might say aubut this example. Bacon (1620, p 2eP) explains the inductive method by a striking analogy with wine-rltiaking. To make wine, we first industriously gather 'countless gr$pes . . . ripe and fully seasoned'. From these grapes, the juice fo;d.the wine is 'then squeezed in the press'. The grapes correspond t+i'5he observations from which scientific generalizations (laws or thf.ories) are somehow squeezed out. But how exactly are scienti{id,iireories to be inferred froin observational data? which we must consider. This is the next queptl_,ion The most hopefuliview is that we might obtain scientific theories by logical deductiorirfrom the data. Let us suppose that George is the name of a pet ra,fin. Then we can give the following example of a logical deduction:''' 'b-[ack. All ravens are George is a ra'uil.n. Therefore, GeQrge is black.

i
I

";.. if we *.r"i t". pfi;;";;"iilir. arrrlsr:ierrrific rnethod, ir.i, d;;r;;;i; ffiir.

r While ;;;i;;;;;.;,.;: iJTifJnlT}|; .if; ii., ; .g,.J" "*,,,swarrs ravens !i{!::i: .irfpf.".lr_pf.u of ffi and pr.i"ia. empirical gerr*atizatio's,
-h";i;;'

scierr fi c infere,ces,r.rjit: i:9::tivists in terpreti"t ti

1'.;11:J;::,,*jl'f.llji,

or rarsincatiSn til;;;,;; wlr be

;;;

(1) (2) (3)

examptes im* of p()rtfllrt scientific discoveries drawn from ,frltirrory of science. I

,o.ory of science

(1) From the premisse.r; and (2), we deduce logically the conclu:'. sion (3). The Scottish phllosopher David Hume (1711-76) pointed out that scientific inferiiiices from observations to laws or predictions

:.,
iil

ri:

lnductivism and its Critics

Background l] lsome Historical Regarding the urillformity of nature, he writes: i: I for The problem *. i;ru. to disiuss is whether there is any reason irr w;at ij;:."ll.d'the uniformity ofnature'' The beliefin the ;"ii.;G or unifor-itv of r,"tii. is the belief that everythingthat hashappened no will haooenis an itstarrceof some generallaw to which there are is of . . . Tt! business science to find uniformities' suchas .*i.oiilitt. far as our ifr. fl*, of motio.i:rnd the law of gravitation,to which' so scientehas .*t.ndj,.th.r" areno exieptions' In this search ."p.ri."* and ii mav be conceded that such ;;:;--;;tk"blyiiuccessful, uniformities tt"".'i,r'aldhitherto. This brings us back to the question:

cannot be obtained by logical deduction. His argument was roughly this. If we accept the premissesof a logical deduction as true, then wc mlrst accept the conclusion as true as well. If we believe that all ravells are black and that George is a raven, then we sirnply must believe that George is black. The conclusiorr of a logical deductiorr fbllows with certainty from the premisses. Let us contrast this with the caseof inferring a law or a prediction from observations. Suppose we have observed several thotrsand ravens and found them all to be black. We then infer either the law that all ravens are black or the prediction that the next observed raven will be black. We cannot, however, infer this prediction with certainty from the data. We might have an experience like that of the early explorers of Australia, and find that the Dext raven we observe is not black at all. Simi.larly, we cannot infer wirh certainty that all ravens are black. T'hus scientific inference, Hume argued, cannot be the same as logical deduction. This conclusion is supported by modern logical theoryThis situation explains the iritroduction of the term 'inductiorr'. Since predictions and laws cannot be obtained from data by deduetion, inferente,it is claimed that they are obtained by induction, or deductiue or inductiueinferente.

; held in the past' Have we any relson, assuming that they have-always 98-100) io ,uppnr. ihat th;f will hold in the fiuture? (1912' pp-

Russell refers to It is interesting t,{ note that in this passage 'Ihree years later, in 1915' Einstein tt* if {ravitationiv.*,o";t his eene;al theory of relativity, w6ich predicted -signitn.r"i ""lfirfr.a frprrt Ne*to"'s law of gravitation one of these l."ir,ilri, *", .orriirn-r.d i. the eclipse experiment of 1919' tlussell J.ui"rion, himself took a greai interest in all these developrnents' A l t h o u g h R usse fi w a sa n xi o u sto j u sti fyb e l i e fi n th e u n i fo r m i ty .t"ir.tia from doubts .'ihich arerttremorably illustrated ,rr ;;;;;..L uniformity ir"-fri, .iri.f..n exd'mple: 'Rat6er crude expectations of The matr who has fed the chicken every rt. tirUl" to be misl.dading. showing that aru-ifrr""*f-,"ur its.llif at"last wrings its neck instead, of nature would have been *,iewt:"s to the uniformity -jr.-rJf"a the chick$'r' (1912' p. 98)' tusefulto -" we reflecti6ns on the chicken lead him to conclude.that n*r.ff;, than certainty:l should seek probatillity rather

1. 2

The Unif or m it y of Nat u r e a n d t h e P r i n c i p l e of Induction

Inductivism then leads to the view drat there are such things as inductive inferences; but this in ttrrn raises quite a number of difficult questions: what is the nature of these inductive inferences? how do they differ frorn deductive idferences?and how can they be justified? Many of these questions were taken up by the Cambridge school, which flourished in the first few decades of the twentieth century, and then by the Vienna Circle, which was influenced by the Cambridge school arrd continued irs work. In chapter 6, 'On lnduction', of his book The Problems Philosoof p/r1,, published in 1912, Berrrand Russell gives a very clear account of some of the questions concerned '"vith inductive inference which were preoccupying him and other members of the Cambridge school at that time- Russell's thinking on this subject irrvolved two irnportant concepts: the unifurmity of nature and the principle of indlct.ion.L.et us briefly examine his treatment of each of these in turn.

;; together' The most we c"ni4ope is that the oftener things are found ,1," ,r,or" probab'le it becomcs that they will be found together if they have beenfound iogether often enough' .".rft., ,it*, ".tdliitrt, (iltt'ost certainry' It can never quite to ,f* o.oU"Uifir" ,pli: "-outti bl."us. we know that in spite offrequent repetitions ,.r.fr c.rt"i,rty, iiaa failure at the last, as in the caseof the chicken ift.t. t"-",i-es (p' whose neck is wftng. Thus probability is all we ought to seek'
102) ., Russell now goqs on to present his positive view that inductive pfinciple oJ inf"."n.., are i,rltiiied by an appeal to *hat is called the principle as follows: induction.Russell f,jlmulates this

r
I I I
10 lnductivisrn and its Critics The principle we are examining may be ca.lled rhe principle of induction, and its two parts may be stated as follows: (a) When a thing of a certairr sorr A has been found tg be associated with a thing of a certain other sort B, and has never been found dissociated from a thing of rhe sort B, the greater the number of cases in which A and B have been associated, the greater is the probability that they will be associated in a fresh case in which one of them is known to be present; (b) Under the same circumstances, a suflicient number of casesof association will make the probability of a fresh association nearly a certainty, and will rnake it approach certzinty without limit. (p. 103)

'].::
11 :irSome HistoricalBackground i,'r' I)espite these enidgnia, the principle ofinduttion has been severely criticized by Popp'tj.t, as we shall see in the next chaPter. Before pursuing these philiiiophical arguments, however, let uS pause to fill in some more hisr.prical background. Why were Russell and his circle led to take arj interest in problems connected with induction? And who were the ]ltienna Circle who continired and developed the work of the Cambir.lge school?
ri . I

r|

and 1.3 Frlussell the Cambridge school

r
I

r|

r
I I|

This formulation of the principle of induction deals with predictions e.g. the next observed raven will be black. Russell (1912, pp.104-5) gives a sjmilar formulation dealing with laws and generalizations e.g. all ravens are black. It was poinred out to me that Russell's formulation of the principle of induction contains an error. Let A be raven, and B be black..Applying Russell's principle, we reach the conclusion that it is highly probable that R and S bill be associated in a fresh casein which one of thern is known to be present'. Thus if I krrow that an object is black, I should apparenrly conclude that it is likely to be a raven, which is clearly wrong. 'Ihis nnftrrtunate consequencecan be eliminated by rephrasing the firral clause of (a) above as follows: 'the greater is the probability that they will be associated in a case in which A is known to be 'l'his slip of Russell's is indicative of the great difficulty of Irrescrrt'. ftrrnrlrlatirrg the prirrciple of induction in a satisfactory manner. 't'lre prirrr:iplc of induction is none the less regarded as fundartrcrrtal by rrrost thinkers of the inductivist camp. One such thinker was I l;rrrs lleicherrbach, an associate of the Vienna Circle. To ilhrstrateRcir:lrt:rrblch's adherenceto rhe principle of induction, Karl Itolrlrsr' (1934, pp. 28-9) gives the following two quotations from hirrr, lroth written in 1930 and published in the Vienna Circle's rr irrrrr ; r fI ir k ennt nis : 'l'lris prirrr'iplc dctcrrrrincs truth of scientifictheories- eliminate the To rl liorrr scierrrr. woulcl rneannothing lessthan to deprive science of tlrc powcr ro tlcr:itle rhe truth or falsity ofits theories. tr/ithoutit, ,lrurly, rr.ierrce worrld no longer have the right.to distinguish its tlreoricrlirrrrrthc fancitular.rd arbitrarycreations the poet'smind. of Anrl rp,;riu:"l'he prirr<:iple irrduction is unreservedly acceptedby of tlrc wlrolc ol'sr:icn<'c . . . rro man can seriouslydoubt the piinciple ancl it r c v et yrlrry lilt. eit.h ' cr'.

(]'872-1970) *rr'. BertrandRussell

.,'

,Lion pf the EnglishWhig

aristocracy. Both hi|i parentsdied before he #as four yearsold, and His grandfather he was brought upilby his paternalgrandpaients. (Lord John Russell)'had been three times Prime Minister. Russell went to Tdnity Coilege, Cambridge, in 1890, where he becamea and then turned to philosophy- FIe was a wrangler in mather.ii'itics Fellow of Trinity fipm 1895 to 1901and a lecturer in philosophy therefrorn t910t o 1916. on Russellbegan as'a Kantian, and his first book, An Essay the of Foundations Ceonltry, published in 1897, was an attempt to give geometry from a Kantian standpoint. an account of non-,,i:uclidean After the completiottrof this work, Russellbecamea Hegelian for a this position under the influence briefperiod, but he ioon abandoned of his friend G. qi Moore. After these flirtations with German thinkers, Russellrelprned to the British tradition, and for the rest of his life held to some.,formof empiricism, though the particular form which he espoused ilariedfrom time to time. Russell'sfirst triiling had been in mathematics,and mathematics
poses severe probilin for empiricism. Accorii posesa se'ere prob;lmior empiricism, According to empiricisrn, all exPenence. rer marnemaucal trutns knowledge is based-'on experience. Yet mathematical truths - for Knowteoge ls Dasec_.on Ior example, 5 + 7 :,',12 - seem to be known'independently of ex-

perience. In the n:lieteenth century, Johr Stuart Mill, who was that, contrary to first actuallyRussell's la;ilgodfather,had suggested truihs such as * | appearances, matne.,!$atrcal mathJn;tical trutns sucn as 5 + 7 : 12 are actualy lz are actuallv
inferred inductivel'ii from observations, iust like other scientific

t|

generalizations.Rulirell did not find this solution satisfactory, and looked for another lfproach in the logical works ofPeano and Frege (for detailsseeGillie;i; 1982). f'he basic idea was to reduce mathernaticsto logic. Let us conare sidera logical truth sjrrch 'All bachelors unmarried',We do not as have to conduct an;e';npiricalsurvey of bachelorsto check that this is correct. Orrce w+,understandthe definition of bacheloras un-

i:'r
lq

F T

12

lnductivismand its Critics

married man, it becomes obvious that all bachelors are unmarried. On the other hand this truth is empty - a mere linguistic convention, or a truth-by-definition. : The claim which Russell hoped to establish was rhat mathematics can be reduced to logic - a view known es logicism. If logicism is correct, then mathematical truths are basically of the same kind as the truth that all bachelors are unmarried. So we can know rnathematical truths independently of experience, but only because such truths are essentially empty - mere linguistic conventions rather than assertionsabout the world. This approach leads to a new v_ersion of empiricism, which could be called logical empiricism. Knowledge is divided into two kinds. The first kind is knowledee of logical truths. This knowledge is indeed independent of e*perience, but it consists of mere truisms, empry trurhs-by-definition. Mathematical knowledge is of this kind, The second kind comprises all really significant knowledge about the world, and is basid on experience. The first step in developing logical empiricism is to establish that mathematics can be reduced to logic. Frege had already tried to show that arithmetic could be reduced to logic. However, Russell discovered that Frege's logical system contained a contradiction now known as Russell's paradox. Russell's paradox can be stated informally as follows. Consider sets, such as the set of men or the set of fingers on rhe left hand of the President of the United States. These sets are not members of themselves. Thus the set of men is a set and not a man, and so is not a member of the set of men. Similarly, the set of fingers on the President's left hand is nor a finger, and so not a member of itself. On the other hand. some sets are members of themselves: for example, the set of all sets. Let us therefore cbnsider the set of all sets which are not members of themselves. Is lhis set a member of itself? If it is, it isn't. If it isn't. it is. This is a contradiction. The discovery of this paradox in 1901 inaugurated Russell,s most creative period in theoretical philosophy, which lasted rill the outbreak of the First World War. Russell's programme was to reconstruct Frege's logicism in such a way as io avoid the contradiction. In the course ofcarrying our this task, he devised his theory of. descriptions (published in 1905) and his theory of types (published in 1908), The deduction of marhematics from purely logical premisses was carried out in the three huge volumes of principia Mathematica wrirten with Alfred North Whitehead and published between 1910 and 1913.

13 i Some HistoricalBackground ';^'i1f seemecl Principia Mathematicamight we[ The appearancq$f Principia Mathematica might well have seemed tlie truth of logicism beyond reasonable at the time to es't4.\ulish doubt. In retrospecirwe can see that there were difliculties from the start. It was not-clear that Russell's complicated theory of types was reallv part of logic, and Russell had to make use of some axioms for example, the-ad:.,gmsof infinity and choice - which again did not seem entirely logi{al in character. A German mathematician, Ernst Zermelo, had proflosed an alternative method for dealing with Russell's contradicfion. His approach, known as axiomatized set theory, was also publith.a in 1908- Initially, Russell's theory of types was th. motrl powerful syStem, but Zermelo's approach was sirengthened with.la new axiom by Thoralf Skolem and Abraham Fraenlkel in lgzz.i.nhereafter axiomatic set theory proved to be simpler and more;i powerful, and has been widely accePted as a foundation for ma;liematics, though the theory 9{types still retains a certain importan[e. The worst blow to Russell's logicism cam-e, however, with thei:publication 1n 1931 of the paper 'On Formally Undecidable Propb'sitions of Prindpia Mathematica and Related Systems I' by the Gzech mathematician Kurt G6del, a member of the Vienna Circle.':'We can state G6del's result in the somewhat simpler form produced by the American logician John Rosser in 193&. According tq the Gddel-Rosser theorem, it is possible to construct a propolition of arithmetic which is not provable in Principia Mathematila (PM) provided PM is consistent, but which can nonetheless bd' shown to be true by an informal argument outside PM. This.proposition is thus a truth of mathematics indeed, of arithmelic - which cannot be derived from Russell's logical axioms (ass{ming these are consistent)- The situatiott cannot bJrepaired simply iiy adding a few more axiorns b PM, for G6del's prooi.rn be repeatbd for this enlarged system- G6del's incompleteiress theorem s6owr'that no logicist system of the kind proposed by Russell can evet be,'l,adequ"t"for the whole oFmathematics. philosophers have abandoned logicism, Since G6del, $ott though there are siill a few who hope to revive it in a modified form. Bet*.en 19iC-and 1930, however, logicism seemed to be an account of matfie:iaatics, in which there were certainly some difficulties, but whiih, on the whold, was highly plausible. Thus, under the influenceiif Russell, the Vienna Circle espoused logicism in the 1920s. Their,iviewpoint is well expressed in Rudolf Carnap's defence of logicisml'iCarnap, 1931) in a symPbsium on the foundations of mathemai.cs published in the Vienna Circle's journal
I rrJrvr rvqt sqvrivr

tl
!l

ll

il
lt

!l

l,[

it
l{

ll
il 1[

il tl
ii il

ill :iil

i,l

Er k ennl n$

ll1 lYJl.

r aae

;i]

..' l \

r; 1

't
t I
t
I

il{

.:

1,r:
, 1t.,
tJ i

il

,J
rT,Irir

a
I I
l'4 Inductivismand its Critics

;
'$ome HistoricalBackground
't:

15

a
I I I

t
l|

t t
I I
T

l.er rrs rerurn now to 7912, when Russell published his discussio' ,l irrtfrrcti<rrr The Problems of philosophy. At this srage Russell in rlr..1;lrr that hc and whitehead had establishedlogicism"to be an ;rtlctlrr;rtc accourrt of rnathematics. The next step in ihe development .1 tlrc rr.w.logical empiricism was evidently tJshow that sciartific, :r:; opposcd to mathematical, knowledge could be based o,, ."_ lrr.ricrrc.. If, however, scientific knowledge is obtai'ed from .lrscrvati'' by inductive inferences (as inductivism suggests), this r:riscstlrc problern of how such inferences can be justifiJ As wc have seen, Russell's response to this problem was to l).srrlxrc a 'principle of induction' and to emphisize the role of 1'rrh;rbiliry. Russell himself did not pursue these.investigations nrrrtlr firrther, since, with the outbreak of the First World Viar. his irrt<'rcst shifted l-rom philosophy ro radical politics. l)rrrirrg the First World-theoretical War he campaigned for pacifisrn, ,nd, , ", resrrlt, his lectureship inphilosophy it Trinity College, Cambridge, wrs rot rcrrcwed in 1916, a'd he was imprisoned foi six rnonths"in l9lll..lr the renrai'ing fifty-two y.r.s oihis long life, Russell did *r::rsirrr;rlly rctrrrn to theoretical philosophy, buion the whole he rlcv.tt'd hir'self to social and political queitions. He was imprisoned :r1i;rirr, tirne for a week, in1961 at the age of eighty_nine,for his tlris p;rrt irr tlrc carrrlraign for nuclear disarmament. llrrt i'vcstigatiorrs i'to probability and induction were carried on ;rt )rrnbrirlgc by a seriesof younger philosophersof science:w. E. .( .f.lrrrs.rr, .folr. Maynard Keynes (before he iurned to economics).1 I l:rrrkl Jcfli'eys, arrd Frank Ramsey. These thinkers all adopted an rllrl)r().(lrt<rthc problem known as Bayesianism, theory whlch a still lr,s :r l.r'11crru.rbcr of supporters today. Now many (perhaps even rrr'st) irrrlrrt'tivisrs are Bayesia's, while many (perhapi even most) ll.ryr'sirrrs:rrc irrrluctivists.Thus the two theories a.e oft.r, plrc.i tr1',r'r.lrr.r rrrr<lcr si,gle heading, lt seemst. me, however, thai thev I slr,rrrltllre <';rrcfirlly distinguished, since ir is in fact pos3ibieto be al irrrlrr.ti'ist lr.r. a Bayesian, and vice versa. To clarify the ''t rlif lir err.c fretwcer indudivisnr and Bayesianisrz, us briefly examine let tlrr' lrist.ry rrf ll.ryesiarrism.This theory, though not as old as irrrlrrr tivisrrr, w;rs intr.duced long before'the twentieth century. It lir\t irlrp(.; u( . (in f hc t , ir r t he eighieent h c e n t u r y . 1, ll:ryesi:rrrisrr rrarrrcdafter the English mithernatician Thornas is lf;ryr.ri (l /O'.).6l), whose inrportant contribution to the theorv of p< lrro lrrrlrilit ywr r slr r r blis hed x t hum ous ly in 1 7 6 3 . B a y e i 's p a p e r w a s r.rrilrrrrDir;lrctl r. the Iloyal Society by his friend'Richard price (l7,l.l (,1). l)rict. wrotc arr introductory letter and an appendix. As

these are significanti'additions, the paper could with justice be considered a jointl pt;7er, and Price should perhaps be recognized, along with Bayes, as;p founder of the BayesianiapproachPrice was strongl.,uinfluenced by the discussions of induction in Hume, 1748. As I ,i:iavealready explained, Hume argued that a generalizationsuch il'All ravens are black'ot'a prediction such as next observed,liaven will be black' cannot',bq obtained by -T'he logical deduction froni reports on the observation of any number of blick ravens, howevsr large. Even if thousands and thousands of black ravens have bqeh observed, it-may happen that the very next raven we encounter i'i'iill have some other colour. Price thought that,.llayes's probability calcrrlations could be used to resolve these prolllems posed by Hume.' The idea is a simple one. Otiservational i'vidence can never render a prediction or a generalizatiort rcrtain!.;but it may be able to make either or both prcbable.lndeed, *e 4tiight be able to use the mathematical theory of probability to calcui,ue the probability which evidence gives to_ 1 prediction or a gencralization - to calculate, for example, the probability that a patjri:nthas a particular diseasegiven that he or she displays a particularlset of symptoms. The Bayesian school has devised ways to carrT.,out snch calculations, and these make use of what is known as Ba,rles'stheorem. The opponents of Bayesianism do not deny the validity of Bayes's theorem, which is a result in the standard mathematicql theory of probability. What they question is the legitimacy of the lrse made of this theorem by Bayesians. Having explained,lthe basic ideas of Bayesianisffi,we can now compare it to inductiiisrz. Inductivism is a theory of how scientific research should be colhducted. It claims that a scientist should. make a large number of ca{eful observations, and then obtain predictions and generalizations bilthe process ofinductive inference from these observations. Bayesiairism, on the other hand, is e suggestion as to how scientific geneializations or predictions should be assessed reladve to the evideril:e used to support them. The Bayesian urges that the mathematied theory of probability should be used to calculate the probability ofthe generalization or'prediction given the evidence- It is therefor'e clearly possible to be an inductivist without being a Bayesian. So,ineonemight believe that the inductive method is the correct rnzay j!: carry out scientific research, but have no concern for trying to calculate the probabilities ofany generalizations or predictions obtaini:d. This, indeed, was Bacon's position. In the gives his classic account of inductivism; yet Nouutn Organum, Baci,pn nowhere does he sug";est methods for calculating the probabilities of
l:,

.-:

t.
i:

to

lnductivism and its Critics

the probabilities, given evidence_ p blutiiti.', gi,,.n ; oene,.rio.ri^-;:l:.-,:i3l1,i- *: d:,"y1'g. s"uch;td.; ;;,-ofi ;;;;1,;,; ;: l. " that g.".,,rir"iion,"*;,"d,.;ffi; ::,fl.*:']::T!tl: could be obtainedby the inlucri.,r. rn.J.oJ.

g.eneralizations predictions,nor does he everi or suggesrthat this should be done. AJ a matter of fact, it would-hav. UJ& frirr"ri."fju impossible for Bacon to have been a B"y.ri; B"y.;i";i;;;;il aboura possible useof the mathematical .A."i", .f so. could only be formulated as a theory o7r, i-U"Uifiry, ,"J tn mathematical calculus of probabilitv had been inrreoi"dl Now historians of mathematrrafprobability_normally date its firsr appearallce a to correspondence betweenpierre de Fermat and Blaise'pascal;il.; in the year 1654.3 Bacon'sJtlovuy Orgoou*was published 1oo.k_gface in 1620, a full thirtv-four years before rhe ,,.?y,fi.st ;rds-in";; mathematicalrheory p.oUrUitiiy. ifr"r-#on - - - - - l' could nor have - '4gs .- a\ "f b e e ri a Bayesian. Conversely,it is perfectly possibleto be a Bayesian . bur not an io'ha.,,e tii, ;Jt;o; rr-r,rt* r,"ri i1,1.jiy1'j^Carnap-se,ems period (cf Carnap, 1950,-pp.M_204.-i;. ir.ld.n would be to try to calculate

ijome HistoricalBackground
t.

17

vance to the Bayesiiln programme are fascinating questions, but to pursue them furt}"g:.r would involve us in mathematical considerations which lie outiiide the scope of this non-technical book.a In what follows, there.$r., lr. ,h"il be concerned almost exclusively with inductivism, :rpd Bayesianism will be mentioned only in passing. I will next r$ive a brief historical account of the Vienna and early 1930s, continued the work of Circle which, in the11.920s Rtrssell and the Camiliidgs school.

" fl1t
,l

',i.4 The Vienna Circle

the two

are often

But once a

therefor" r^ #i'-J thato..^-^rr ,., therefore, find ,l^* Russell , ,lr.'cl*;;;; ro ""a inductivists and Bayesians

g::.:1, lyidence, is th. procedure r'c Pruce(rurt t*ir'ly the Bayesians attempt to il;-;;:,.p $ which carry o,rr.-ii".rp
not surprising,

ffiiTJT:fi

Given this-general outlook, it was natural for those working in m^athematically minded Cambridge ," i"".riig"r. ;il; il.;;;;ft# of probability needed camyouithe Bayesiafi ro. programme.Keynes (1921)adop-ted. logical interpretation.i p."iJUifiiy, a r"d r;;,;h;; probability logic or inductiveiogic as an .*terxion of the deductive logic used .b-y Russell and Vrhitehe i;"iM: Whereasdeductive logic provided a foundation for "d *rit._"ri.r, inductive looir was intend-ed provide a justification {b. .;;;;..-R;;;;; to (i;i;; criticized Keynes's logical-interpretation, a.".fop.a,i;;;;l what is known as the isubjectiu"'rt."if ""J orfrJo"uilry;l mrr'prr"i of view was introducedanidevelop.a ina.fi"Lntly i,,I;i; b; ,h. mathematician and philosopherBruno d. Fi;;;. I he nature of theseinterpretations probability of and their rele_

i*' The origins of the Vidhna Circle go back to e group of enthusiastic research studentswho, in 1907, used to meet in an old Viennese coffee-houseeve_ry Tfrursday night to discussproblems of science and philosophy.r Tht';e of theseyoung men were working in different fields (Philipp'Frankin physics,Hans Hahn in marhematics, and Otto Neurath iii.{economics), but they shared an interest in philosophy of scienc{.I Some yearslater, in 1921, Hahn obtaineda chair of mathematicgiinVienna, and, when the Mach-Boftzmann professorship ofthe ifiductive sciences becamevacantthe next year, Hahn exerted his inflrr,ence get Moritz Schlick appointed. to Schlick'sarrival.irrr,Vienna 1922marked the real beginning of in the Vienna Circle. Schlick in fact arranged a seminar for a small invited group, and ;fiiis group came to be known as the Vienna Circle. At first the ptincipal participanrs were Hahn and Neurath; and Philipp Frank, 14ho was professor of theoretical physics in Prague, paid frequegt visits. Later the group was extended to include, among oth$rs, Viktor Kraft, Herbert Feigl, Friedrich Waismann,and Kurt Gddel. Schlick was able to obtain the post of instructor at the Uni#ersity of Vienna for Carnap, who joined the circle in 1926 and rapidly becameone of its leading {igures. In 1929 the group published)fts manifesro: Wksenschaftliche Weltaffissung: Der Wiener Krers ('The Scientific Conception of the World: The Vienna Circle'), which was written mainly by Neurath; and in 1930 it startedits own jouipal, Erkenntnis ('Cognition') editedby Carnap and Reichenbach. .: The manifesto co:itains an appendix listing the members of the Vierrna Circle - fourteen in number. Therr is then a list of thosesympathetic tfueVienna Circle, which includesRamseyand to Reicherrbach. Reichen$ach in Berlin, and so wai not formally a was

.T

18

Inductivism its Critics and

,i'
.L

Background Historical :,Some

19

member of the Vienna Circle, but he shared their interestsand ourlooft, and can be consideredan associate rriember of the Circle. The maniGsto'sappendix qives, finally, an honorific list of 'leading representarives the scientific world-conception'. This list contairri of only three nanles:Albert Einstein, Bertrand Russell,and Ludwig Wittgenstein. Karl Menger, in his contribution to Gadol (ed.), 1982, gives a vivid descriptionboth of his invitation to join the Vienna Ciicle, or the Schlick Circle (Sehliek-Kreis) ir was then called. and of the as meetingsof the Circle. Here is his account of the invitationj When I returnedto the University of Viennain the fall of 1927, n order to teachin the chair for geometry,the mathematician Hans
Hahn asketl me whether I would like to ioin the Schlick-Krek.the group that abroad becamefamous under tlie name of Vienna Circle. 'l attend regularly', he said, 'and so do Carnap, Neurath and a Gw younger people; and Philipp Frank visits us wheniver he comes to Vienna fiiorn Prague]. We meet every orher Thursday evening on the ground floor of the wing of this building in the Boltzmanngaise.' We werc talking in the L-shaped university building that housed the rnathematics and physicsinstitutes.(Menger, 1932,p. 85) Of the meetings he writes: The room on the ground floor in which the Circle met - rarilv more than 20 persons- was rather drab. \Ve would stand in little groups talking until Schlick would clap his hands and we would be ieat.d. SometimesSchlick would begin by reading ro us a letter that he had receiveddcalingwith problemsthat we had discussed.or planned to consider. His correspondentsincluded Einstein and Rusiell. He would open other sessions readirrgannouncements new publiby of cations (especially British ones) and would promise to.report abour sonlc or ask volunteers to rcview them. Occasionally, Schiick would introcluce a guest passing through Vienna. Then there would begin r discussiorr thc topic proposedin the precedingmeetingor so;eon one's report about work in progress. But in none of the many nreetirrgs that I attcndedin the courseofthe yearsdid the dibatesevei touch political or economic problems. Even men with strong political convictionsnever expressed them in the discussions ofthe Circle. It was in thesc discussionsthat Schlick particularly excelled both as a stimtrlatirrg participanrend as a moderatorof ideal poise. (p. 86) 'l'hc philosophical views of the Viemra Circle came ro be known as 'logical positivism', though rcally the term 'logical empiricism',

introduced eerlier, iJrinore appropriate.As we rrright expect, Russell rvas a major influengd.In his memoir of Hihn, one of ihe'founding members of the Vi'lhna Circle, Menger writes: 'During the early 1920she developedl:lgreat admiration for the works of Bertrand Mathe.matik Russell.He reviewe{;some df them inthe Monatsheftefiir und Physik. In one.$f these reviews Hahn suggestedthat one day Russellmight wdll ,,fieregardedas the most important philosopher conduited a seminar of his time' (Mengqi 1980,p. xi). Hahn also. on Russell and {Vhiileheed'sPrincipia Itlathedaticain the academic year 7924-5 during iwhich the partieipantswent through that work chapter chapter. by i ,i. Another impoitalfi influence on the Vienna Circle was Ludwig Wittgenstein, whcde first important philcisophical work, the was published in 1921. The Vienna T'ractatus L,igko-Phtit;ophicus, Circle devoted itselflio reading this book'paragraph by paragraph' during the academil;'iear 1926-7 (Menger, 1930, p. xii). Curiously enough, Wittgeniteifii,though in Austria, did not take part in these discussions.After flnishing the Tractatus,he decided to give up philosophy and 6edir'mea village schoolmastbrin Austria. It was not ur.rtil 1929 thaf-'Vittgenstein decided to iesume philosophical and he th$ji retumed to Cambridge'in England to do so. research, Wittgenstein never ,itttendeda meeting bf the Vienna Circle, but with selectedmembers of the Circle he had occasional<iiscussions (notably Schlick ani.Waismann) in the perio{ 1929-32. Notes of some of these conivlrsations were made by' Waismann, and ere unil printed n Ludwig ir,tittgenstein iler Wiener Kreis (McGuinness,

re67).

ir

Roughly speakinf,;we can say that Russellinfluencedthe Vienna Circle as regards tili-'ir logicism in mathematics and their intetest in induction, wherhas 'fi/ittgenstein influenced them as regards the question of thefdemarcation between scienceand metaphysics. Wittgenstein's Tracilau will therefore be discussed some detail in in with the demarcationptoblem. Some further Part IV, which deair', life information about I0.ittgenstein's will also be given at that stage. As this book devLlops, it will become clear that there are many objections to the logical empiricism advocated {irst by Bertrand Russelland then byilhe Vienna Circle. lndeed,logical empiricismis held by few, if any;fphilosophers today, Yet although many of the detailed views of tbi Vienna Circle have been shown to be false, a casecan still be mide for some of the Circle'3 general principles. Thus, in their mariiiesto we find the followingi 'In thl rese-arch work of all branctresof empirical science this spirit of a scientific

:ii

'l 'l
_ *

i,:' .f ltjr.

1_S ' -_,

....-._,,_:i at' i l i :1, ....;,:.r.'

:
i

d 1

,i'l

20

Inductivism its Critics and

tsome

HistoricalBackground

zl

rcnception the world is alive. However only a very few leading of thinkers.give it sysremaric thought or advocareit, pri.r.lpi.r3 (Neurathet al., 1-929, 3). Even if the ViennaCircle were w.o.rg in p. their precise analysisof the.scientific conceprionof the world, tfief may-havebeenright that thereis such a conception,and right also to advocates principles. its

1.5 The Twentieth-CenturyRevolutionin physics so far we have discussed logical and philosophical influences rhe on Vienna circle, bur scientific influenies weie ario of the greatest importance-.The years 1900-30 were rhose of a great r".,ror,itionin physics, which called inro question the Newtoniln mechanicsthat had been acceptedfor nearly two centuries, and gave bi.th tt th. new theoriesof relativity and quantum mecharricsl rhe revolution got under way in the early yearsof the twentieth ce'tury, *;tt itr" development of the.specialtheory of relativity by Hendrik firr"r"-, Henri Poi'car6, endAlbert Finstein,6 and the-firit,steps q.,".rt,ri' theory by Max Planck and Albert Einsrein. A ,, alieady remarked, was the confirmation of Einstein,sn"- "i,.,rr, 'otable th.o.y oi gravitation by the eclipse experiment of 1919. During the 1fr0s, when the Vienna Circle-was being formed and was aJ"a"phg ir, scientific conception of the world', Werner H.ir.nb..g,'E;-i; Schr6dinger, and Paul Dirac were introducing ifr. n.!"""rumechanics. srrong interactions betweenthe philosophizing the of --.Th.t.^yrle and the great revolurion vienna.circle in physiis. onihe on".h"nd, the crrcldevotedmuch time to discussing conceptual the problems of the new physics- the nature of spacJand time i" thJ figh, ;i relativity and the paradoxes quantum mechanics. of On the other hand, many of the creatorsof the new theoriesin physicsadopted some form- of empiricism or positivism as theii philosophy of s_crgnce. This was true, for_example, of Einstein ,rri H.ir.rrb.rg. Admittedly, Einstein turned against empiricism in his later year!, but this was afrer his most creative period as a physicisi haj finished. . There were also personal contacts between members of the vienna circle and the leading physicistsof the time. schlick was a good friend of Einstein's, and the two of them eng;aged a in co'siderable correspondenceconcerning the philosophi.l irrr"r-

pretationofrelativity:EinsteinevenassistedSchlickinhisacademic Kiel in l92l' the i"r."t. rcf"its Schdik to obtain a professorshipin to Vienna. (For an excellent account of ;.;;'d.;';:lZfrfi.[ii""""a ichlick' correspondence'seeHoward' 1984') ilEtt"A--not pursue A, ,t i, book is nbn-teJhnical in character, we will technical iFsuesin the philosophy of physics' interesting th. revolutron -o.. thoueh theseare' ltlis worth noting, howcver, how the both the Vienia Circle's choiceof fundamental i" ,Krf.ll"n;;;...i problq,ns and the way in which they pursued those "ilif"i""fti.rf important facior was the changein attitude ;;"1#;:'11ri tiiboty of mechanicsand gravitation' Prior to the N;to'" -"q, ;;;;;;; as ,.rrot,ttio", this thg.gry had been regarded pY *ott sglsntisls. certain' Newton's theory had-indeed had if "lr.t"fit, remarkable!'run of empirical success' explaining a rnass "o, ""rl*fy, , passing the experimental tests.to -otr' obr.rrrrtiot, of "n.i,to...ssfully Yet t henewt heor iesot r elat r vlt y w hi ch' i thadbeen. subject ed. showed that Newton's theory was at best and quantum mechqnicJ ,r, ,ppto*i-ation, i'rnd actually gave incorrect results in many - fo.li ..x,tttple, for bodies moving with very- high .i..tlitr""."t ne", ,r..y large gravitating masses'and in the for boditr.l, ".b.iri.r, micro-world. The;iftpilureof such an important and preuously scidniific theory gave a new significanceto Hume's *"ii-.""n.*ed to philosophical doubii about induction' The problem was posedast,'rr.it.i. extent, scientific theories could be justified l"a to w.hat inductively by obsewational and experimentalevidencet' ,. I'i

1.6 Popper
.I

an important thinker who, though. not a Let us ,ro* "o.tridqr Circle, had close links with it' Karl Popper of the Viehna -.*b., Uo* in 1902,ss was some yearsyounger than Wittgenstein and -", found"r membets of the Viinna Circle. He was never invited to th. tt. Vienna Gircle's serninar, and was wittily, but accurately, "it.rrd by illeurarh as the Circle's 'official opposition'. .t ".r.t..i""a Discouery093!) was first book'Tfii Logk of Scientifc P;;;.;; in the Vicnna"Ciicle's seriesedited by Schlick' It deals ru[iirfr"a '*iif, tU" topics modi;characteristic the Vienna Circle - induction, of de-arcrtion, probagiility and confirmation, the conceptualproblems oi qor.rr,r-'mecha*ics, and so on. Thus-Popper shared.with the Vieina Circle an interest in the sameset of problems. The difference

22

Inductivism its Critics and

i., Background j;9me Historical

23

lay in the answers he gave to these questions. Popper criticized the views of the Vienna Circle on most fundamental philosophical issues,and he developedphilosophical theories which differed profoundly from theirs. It cannot be said, however, that Popper differed from the Vienna Circle on every topic. For example, in The Logk of ScientificDiscouery, Popper arguesfor a version of von Mises' frequency theory ofprobability. Richard von Mises, like Reichenbach, was an associate of the Vienna Circle, and the frequency approach to probability, which he and Reichenbach adopted,was accepted many memby bers of the Circle. Popper also sharedthe Vienna Circle's admiration for Einsteirrand Russell.Thus, in his QuantumTheoryanil the Schism in Physics, Popper speaksof 'my boundlessadmiration for the work of Eirrstein' (1982b, p. $n, and, in his Realismanil the Aim oJ. Science, writes: 'In the autumn of 1935. . . I was taken to a meeting of the Aristotelian Society. Bertrand Russell, whom I had long adntired as the greatestphilosopher since Kant, read a prper ort "Tlre Limits of Empiricism"' (1983,p. 12). But Popperdid not, as we shall see,sharethe Vienna Circle's admirationfor Wittgenstein. All his liG Popperhas remainedan essenrially oppositional figure. Not long aftcr finishing The Logic of Scientifu Discovery, left he Vlennr to trke up a lectureship in philosophy in New- Zealand. Since he did not find there any major school of philosophy to attack, he devotedhis time to writing a long criticism of plito'and Marx, whonr he characterized 'enemiesof the open society'. as A.fter.thewrr,,Popper went to England, and once-again found hirnselfin largely isolaredoppositionto the dominant philosophical trend. Wittgensteinhad rerurned to Cambridge in ttre 1930i and he, together with important. figures in Oxford (Austin and Ryle), creeteda"system of thought known as 'Iinguistic philosopiy;. Aftcr thc SecondWorld War, this becamethe standardlpproach'to philosophyin Britain, and Popperwas one of the few major figur.s irr tlre country to rejectlinguisticphilosophyvirtually in tito. popper had a srnallbut lively schoolin London, and exercised considerable influerrce the developmenr philosophyof science on of world-wide. Although he had few followers in the British philosophical comtrrtrrrity, his views had, and still have, many supportersamong scientists. So for a tirne afterthe Second World War, Popperwas in London, scparated only a few miles from Wittgensrein Cambridge.yet by in the two philosophers seemonly to havehad one encounrer. Already lcgerrdshave sprung up about this event. The details differ; but

most versions invol.ril,Wittgenstein brandishing a poker ferociously and then storming frQn the room in a rage. Popper hasprovided his own account of the-ineeting in his autobiography, Unended Quest (7976; pp. 122-4). .ln'1946 Popper received an invitation to speak to the Moral SciencgpClub at Cambridge, giving his talk qn 26 October of that yei{; In reference to this, Popper explains his 'custom whenever I iin invited to speakin sonle place, of trying to of develop some conse.duences my views which I expect to be unacceptableto thati,irarticular audience' (p. 124). This 'custom' certainly reveals a gieat deal abo-ut Popper's attitudes. On this it occasion, led him tii. speakon the subject'Are therePhilosophical Problenrs?'and to.aTgue that there are genuine, important such problems. This was ';idirect challengeto Wittgenstein'sview, which rvill be discussedinifart lV, that philosophical problems ere all pseudo-problems ariiing from the misunderstanding of language. Popper recounts Witfuenstein's very negative reaction to the paper. follows: His descriptionends,i"s
'1'

At that point Wittgcistein, who was sitting neartj'r. fi.. ,nd had been nervously playing vlith the poker, which he sometimes.used like a conductor's baton ti.'emphasize his assertions,challengedme: 'Give an example of e *i6ral-rule!' I repted: 'Not to thr-catenvisiting lecturers n'ith pokerlil Vhereupo4 Wittgenstein, in a rage, threw the poker down and stcliSed out of the room, benging the door behind : him. (p. 123) :1i:.

"'.
Richard Braithwaite; f,.r whose rooms in King's College, Cambridge, the meeting took place, used to explain in after years that 'Wittgenstein was meibly poking the fire - violently! In Part IV I will Alglue, somewhat paradoxically in view of the strained relations beF.'reen the two philosophers, that Wittgenstein's later ideas on langfulrge and meaning actually lend support to Popper's dccount of netaphysics.
''l ''
l{

1.7 Ttie Dispersal the Vienna of Circle

We must now briefir" describethe melancholy events which led to the expulsion of the'Vienna Circle and its 'offrcial opposition' from their native city. In '"934, the year of the publication of The Logic of Discovery, Austrian Fascist trlie, Party, under Dollfuss,seized Seientffic power. Parliament i,vas suspended, and socialists disbanded and

-'rTf "-

.-ilTr:;f,Ti

.:,
Inductivism and its Critics jailed. But Dollfuss himself was murdered by the Nazis in July 1934. None the less independent Austrian Fascism contintred for a while with Schuschniggas the new Fiihrer of the Fatherland Front. But long it could not be, and on 12 March 1938 Hitler invaded and occupied Austria. Many members of the Vienna Circle wereJewish, and even rhose who were 'Aian' held liberal or socialist opinions unlikely to appeal to Fascistsand Nazis. Moreover, the Vienna Circle's ,scientific-con_ ception of the world' was a threat to the hiehlv unscientific racial theorieswhich were a key elemerrtin Nazi propag"nda. The most left-wing member of the Vienna Circle was Ofto Neurath. Although not Marxist in any orrhodox sense, he had considerable sympathy for Marxism, and was in charge of central planning in the temporary Spartacisr government set up in Bavaria after the firit world war. !(/hen the Spartacists were overthrown, Neurath was sentenced to imprisonment, but was released on the intervention of the Austrian government. He would almost certainly have been imprisoned by the Fascists when they seized power in 1934, I:ut, luckily, r,vasin Moscow ar rhe rime explaining his system of international communication using pictograms.^Insteal of returning to Vienna, he wenr to Holland, but hal to flee again when the Nazis invaded in 1940. He then werir to Oxford. wheri he remained till his death in 1944. The next dramaric eveltt was the assassinationof Schlick, which took place in 1936. Schlick had examined (and failed) a thesis on ethics.written by a Nazi student (Nelbock). Later, when ascending the philosophers' stairway in the university building, he was sho-t dead by this student. The student was sent-to prisorifor a mere ten years, though he could have been sentenced to death, ancl was then releasedby the Nazis when they occupied Vienna in 1938. Nelbock applied for a full pardon in 1941 on rhe grounds that he had done a useful service in disposing of a Jewish professor. (In fact, Schlick was not Jewish, but was descended from the Prussian nobility.) Nelbock was given a post in the geological division of the mineral oil administration of the war economy, where he worked till the end of the war. He died in 1954. After Schlick's assassination,most remainins members or associates of the Vienna Circle left in a hurry. This brilliant group of philosophers had, like Dante, to eat the bitter bread of exile.T But the loss to the culture and civilization of Austria and the Germanspeaking world was, at the same time, a gain to the English-speaking world. 'fhe displaced philosophers, scientists, and marhemaricians ;j5o*e HistoricalBackground

they settled in the USAi'rBritain, and the Commonwealth, v'here a qreat iniil'ence on the development of philosophy of .".rt.a ..i..r... ThIs pro.e*i was reinforced by some'British and American ohilo.ooh"r, who h'ad studied with the vienna circle in Austria and ih.., ,"irrrr,.d horr,.l. This course was followed by Willard van Orman Ouine fronl Harvard and Alfred Jules Ayer from Oxford' Aver's veision of ljgical empiricism, published in 1936 in his book Iunouonr. Truth, aii Logic, spread Vienna Circle ideas in Britain' Veri, iiteresting todlare-Ayer-'s criticisms and developments of his "whibh tpp.rt"d as an introduction to the second o.igi.ral .,ie*s edition of his book. I rvill {iscuss some of these later on' 11liO) ' I have now introduced most of the philcisophers whose views we shall consider id, this book. There remain two French thinkers (pierre l)uhem andiHenri Poincare) who dourished before the first iorld war and whtjse work in the philosophy of science is of the hiehest importance. We have already encountered Poincard as one of th! men involved i:r the development of special relativity; but, in addition to his wcl-k in mathematics and physics, Poincar6 wrote books of the greate$t significance on the philosophy of science' It will be Jonveirieni to give biographical'nodes on Duhem and Poincar in chapteri3. For thl moment we have enough members of the cast to c.rttiittd&our philosophical drama, and what better wayto do so than to cdriside; Popper's criticisms of the inductivibm of Russell and the Vidnna Circle.

't

i;

.i;

Critiqueof Inductivism iL6pper's

"l

27

2
Popper's Critique fnductivism. of and 1|ir Theoryof Conjectures Refutations Faliitication (or ism)
2.1 popper,sCritique of fnductivism [.Jjh" ina""tivist thinks^rhat. science proceeds by first collecting observationsor dara (Bacon's ,.ounrtir, . . . ripe and fuily ;;;* '#l seasoned') and then inferring l"*, ;;J";;iiiiior,, #; il by inducti.on.Popper argues.against this that ;;; ;r;ru*;i; observewithout a theoreticalbackgroundlff... i, t* t;il'il: argutnenti
The belief that scienceoroceedsfrom observadon to theory is still so widcty and so firmly ireld. that ;;; is ofren met wirh incredulity. I have even -t;;; insincere_ of .t.." ,uJp..i.J-JU.i,rg derryingwhat nobody in his senses a""Ui -" .:r" Iilrt in fact the belief that we cen start with pure observations elorre,without anything in the nature , ifr.".y, is absurd;as mav beilhrstrated thestoiy .'"ir"iJi.rJ.a tfi;ft;;;ffi by "f "rtrr" ^"n science,wrote down everything he could bequeathed his -do.i"ty pricelesscollection of obslrvatilns t" trt" "f..rr.,l"a n"f"i ro be used as inductive evidence. This story shoul.l show ,r-,fra, though beetles nray profitably be collected,oLservations _"" ,roi 1'werrty-five years ago I rried to b.rng hol; ;e same point ro a group of physicsstudentsin Vienna by ;.;;;;;'" tecturewith the lollowirrg instructions:,Take penciland paper; caiefully observe,and write down whar you have observedt, ';;; tfr'.y'.rf..i, ..;;r;, l- warrted thern to observe. cr.rrif-rr'."itir,i"u'.iion, ",f .observe!, is a'surd. . . . C)'servatio'is always ,.i..,i".. ir rr"eJsa chosenobject, r_ defirrite task, an interest, a. point of view, a problem. And its p,i"

:l"I li rt presupposessimilarity and classificattn,- which i6;;;y' ;;;; i" ir, i; ;;;l sultposesinrercsts,points of view, and problems. (p"pp..,--f6if, p. 4(t

presupposes a descriitive^ta"grr!.,' *i,t

The samething a$plieseven if we go right back to the beginnings or of science of an :lidividual human life' Popperarguesthat somethine like modern:i$ciencedeveloped in ancient Greece through critilism and modiiriLation of an older pythological picture of the world. New-born labies do not have blank minds, but inborn expectations as th$i result of genetic inheritance. However, as expectarionsmay be disappointed- The Po-pperpoints out,r,iithese ,r"*-bo* child explcts to be fed, but may be abandonedand starve. Let us see how rlhis argumnt of Popper's applies to the.Kepler that Tycho Blahe made his obserexainple. It will beif^emembered particularly of the planets, between 1576 vations of the heay,qns, and 1597,.Now Cciiiernicus'snew theory of the universe had been publishedin 1543.Fy the 1570sthere was a major theoteticaldispute betvpen the upholders of the older Aristotelianin "ttrotro-y Ptolemaic view thqi the Earth was stationeryat the centre of the universeand the Stin went round the Earth,and the Copernicans, who thought that tlie Sun was stationary at the centre and the Earth went rou;d the Sutr, Tycho Brahe's observationswere relevant to this theoreticalcontipversy- At a more basiclevel, even his division of heavenly bodies-into stars and planets involved a theoretical classification. Inde$d, this classification was different in the two contending theoriesi.In the Ptolemaic theory, a planet was a heavenly bodv whiJh *tt td'i a fixed star and which niove'dround the Earth; so t-he Sun was a lplanet, but the Earth wa3 not a planet- In the Copernican theoryl-i planet was a body whiih was not a fixed star the and which moved nlbund Sun; so, on this account,the Sun was not a planet, but th;eEarth was. Let us turn to a pecond argument of Popper's. This concernsthe principle of induction. Some inductivists thiirk that their inductive inferencesare just:.ifiedby the principle of induction. But how is the principle of ioJuction itself justified? If we try to justify it inductively by expirience, we get iirto a circle, since any indlrction from experiencedipends on the principle of induction. We might try to avoid maki;;r4this circle vicious by justifying the basic principle of induction iqy a higher-order principle of induction. But then this would, in tr.fri;,,have to be justified by a still higher-order principle of induct;]on,and so on. Thus, if we try to justify the principle of induct),on inductively by experience, we get either a ii.io* circle or ap;infinite regress.The other possibility is to try to justify it indeliandently of experience, or . a priori. However, this looks like a blind act of faith. This is how Popper puts the
argument:

':''i
l. ,i.!, it
rl" :ri

.rit._.-.._._.,__

: i.i

.,-.". : -.:

_. .-:.,*_--.._,

28

Inductivismand its Critics

' l'

Inconsistencies miy easily arise in connection with the principle of induction. . . . For the principle of induction musr be a universal-state_ ment in its turn. Thus if we try to regard its truth 'as krrown from experience, then the very seme problems which occasioned its intro_ duction will arise all over again. To justify it, we shoultl have to employ inductive inGrences; and to justify these we should have to assume an inductive principle of a higher order; and so on. Thus the attempt to base the principle ofinduction on experience breaks down, sirice it must lead to an infinite regressKant tried ro force his way out of this difficulty by taking the principle of induction (which he formulated as the ;principle of universal causation') to be 'a piori valid'. But I do not think that his ingenious ettempt to provide tn a priori justification for synthetic statements was successfulMy own view is that the various difhculties of inductive logic here sketched are insurmount$le. (1934, p. 29)

of Critique Inductivism Prqpper's n"lPtt there is no need ,o tiit" blind faith in such a principle' ,'lT. prollim by suggestinga non-'nductlve tneory or sets round the and refutations, Li*,in. t".,nod. fhi. is his method of conjectures which l will now exftrund' .,
)t .

and of 2.2 Popper's'1-heoryConjectures Refutations ' (or Falsificatioriism)

'i twt t srrrrrvqtrv" rv!r1

{U,r-" tt;;.J..

logic, we oointed out i;at, from observations anil deductive many infer the4iuth of a generalization' Thus however infer the truth that all swans are white swans we seei we ."., tt*"' white

! I

P9nl9.- Q963, p.-289) argues that the same objection applies ro rhe principle of uniformity of nature, which he regards as a kind of principle of induction.. It is interesting ro see how Russell responded to objections of this sort. Russell agreed that we cannot justify the pri'cipie of induction by experience, and concluded that we musr ,...pi the principle a priori, or, as he put ir, 'on rhe ground of its intrinsic evidince'. This is what Russell said: 'We can never use experience to prove the inductive principle without begging the question. Thus we mus: either accept the inductive principle on the ground of its intrinsic evidence, or forgo all justification of our expectations about the future' (Russell, 7912, p. 106). He regards forgoing all justification of our-expertations abou.t the future - that is, c-omp'iete Jcepticism _ as norhing less than intellectually frivolous. Thus he advocates an a priori acceptanceof the principle of induction ,on the ground of its intrinsic evidence'- He thinks that the acceptanceof the principle of inductiorr is necessary in order to be a sciintist: .The general'prin_ ciples of science, such as the belief in thereign of law, a-ndthe belief that every event must have a cause, are ai completely dependent upan the-inductive principle as are the beliefs of daily Uf.'1i. fOZ;. Russell thus holds that we must, however reluctanrly, b.li"rre i. the principle of induction as a kind of blind act of faith in order to do science. And here we come to popper,s most fundame'tal criticism; for Popper thinks that we can be icientists and do science without making any inductive inferences. Hence we do not need a principle of induction to justify inductive inferences; consequently,

. .?!:tll:Tx1^:11 j";ic canr{rt establishthe truth of a scientific gene-ralization d.;;i; (or refute or falsify it)' iti, theylan establishits falsity i* ".rif" 'This it black's-an"'we can infer bv il;'iil.'"itll."",l"n ii;, " swansare whire' is false. generalization.,All logic that, il;;,i;; rrhe or falsify a.scientific generalization' I' other *oldr, we:-can-refute and i;.fp., *f.r, to tnii as the asymmetry between falsification
verification. or and it i, leadspoppel to his conjectures refutations, falsifuationist, does not start with observa..-rr.r, of sci..rtifittmethod. Science ,iorr", * the inductiv-ist claims, but with conjectures'The scientist by to refure tireseconjectures criticism and testing (experi;l;..i". h ,o.rrt, ,nd observations). conjecture which has withstood.a test or ization with certairity. It may break down on the-very next lni ihe case of the discovery of black swans in .Ur.r""ri." 1"s

althouc.h that,., i"OO.r, ho*euJl' observed

r"rr.tn*to' maybe-tt":11":y -i::tn^':9'^-b11::l "r"t"U.."f 'w. car,in.rrerkrrow a scientific theory, law, or elnelalLrrcvry' rdwt v! a scltllLurL We car,i'never Know tentatlvely. ,*i"i*ay.

i.-Aegr-aijd.,the.;tentative,coqiectural nature of scienti{ic knowlvadon and ex ;;;i;;;;;;p"iif*.",

fo--iiil1.rr".. Popperis edge, ,an., popp.t is forijl of citing theexample of Newtonian mechanics'

in. t'.li*,o"".',t'.ory.lLoduced extraordin4t-9::1,.1i*|,:T::; u.til published-(1687)

tS00. N."..rhelei$; relativisticmechanics.

from the time it was between 1900 and t920 it was found to be

inaccurate in some.respects, and corrections were inffoduced using

|i, .1:

,.,iir-l ' . .1 .'::'1 ., ;:; . . i.:t ' ::::}'|t\.

30

Inductivism its Critics and 2.3 The Distinction between Discovery and Justification

ij,;:pps1'5 Critique Inductivism of

..l;

31

theory of coqiectures and refutationsleadsnaturallv to a fop.per's distirrcrion between the discouery scientiric rryp#i*, of ,ri ,ir.i, justi,fication ualidation. or This is how popper himseifputs tt. i"tt..,
The work of the scientist consists in putting forward and testing theories. The initial stage, the act of conceiving or inventinq a theorvseenrs ro me neither to.cdr for logicar anarysis nor to be.,rt..p,iul.-li it.. Tlre question how it happens thrt ,r.* idea oceurs to a rnan _ " wlrether it is a rnusical theme, a dramatic .lrrni.t, or a scientific thcory - nray be ot great inreresr to. empirical psychology; b";l; i; irrclevant.to the logical analysis of scientihc knowledge. This latter is cr.rrrcerrrcd rrot with questions,of quidfaxi?i, b";;;il;t; Jact __(Kant,s t; rrcsr io ns o f 7r st if cal i ort or uaI iil i ty (Ka nt,s quia _ i u rirll -.tr. qu.r ri J.r, o f' .th c lir llo win g kin d . Ca n .a ,t"t"- .r rt U *3url i n.al ".. A ndi f so, ho;? l s it tt:sta b le ? it lo g ica lly d e p e n d e nt o.,..i ,"i n Is other statemeuts? Or ( tu e r itt p cl.tr a p s lr tr a d ict th e m ? ( 1 934, co p. 31)

intuition', in Beigsl,ir's sense.In a similar way, Einstein speaksof the 'seirch for thosehi'ihly universallaws. . . frbm which a pictureof the world can b. obta$ed by pure deduction. Theire is no logical path, he says, 'leading tir these. . . laws. They can only be reachedby love (Einfihlung) intuition, basediuJiln something like an intellectr.ral ofthe objectsofexlrcrience.(p. 32) So Popper's view islihat there is no such thing as a logic of scientific discovery- only a lc:$c of scientific restizg. The distinction lliitween iliscovery and justifcation is related to the distinction betweer,i 'iniluctiuism and Bayesianinr which we made iri is indeed a thcoiy ofjustification, not the last chapter. Bill'.tittrttof discovery. Baytiians seek to justify scientific generalizations or predictions by sho*l.'ing that, although they are not certain, they can none the less be siiown to be probable, given the evidence used to support them. ii is thus possible to be a Bayesian regarding the justifi cation of. Jcientific generalizations ind predictions, while denying the induct.ivist theory of how they are discovered. Indeed, as we remarked in:the llst chapter, Carnap held a position of this sort in his later p.fi"a. Thus, in his 1950 book, hi accepts, with acknowledgement,j!:he Einstein-Popper position on discovery. As

).:l:,.*l^:Tt.. screllceand the arts. rrr('r'f r)er*ce, Great scie'tists can, accoiding to l)oplrer, possess the kind of .creativitv *hich rs recognized as a rr.ssessi.rr oIgreat artists- A classic example of creative"i"ruition ii s.ierrccis provided by Kckur6's discovery itt tgos that the ,i* ."rio., nt.rrrs of the henzcrremolecule_are,rrang.d in a ring. K.krl6, ;hil. rrretlit:rti.g orr the structure of benzene,"fblt and dreamt of a srrake lritirrg its tail. On awakening, he hit "ri""p on th. ,otrrtion oifril lrrob ler r r . l\rlrpcr goes olt to consider the, r,ier,vthat philosophers . of science s''rrld attcDrpl to give a rational ,.conrt.uJtio' of th. p;";;;;";; tlis<'.very, As prrts it: .Some nrighr object that it would be more t. 1lls 1r1l1P1rse regard it as tf,e buJin.ss 'c to of epistemoid";; ,,rdtional I'r.<lrrr'cwlrat has bcer] called a ironrtrurtion,, of the steps tlurt ltavc lerl the scientist to a discovery _ to the firdi;;;;;; rlc.w, tntllr' (P 3l) Popper, however, rejectssuch a view, because he illillKs ilt:tt dlscovcry always contains an irrational, creative element. /rs ne l)ilt.stt:
M.y vicw ol- tlre rnattcr, for what it is worth, is that there is ^o. such thirrg.:rsa logical mcthod of having n.'o ij."r, a logical recon_ strtr.rir' of this process.-lvly_ viei may be ;*pr.rr.d.;; ;;;;; tllrt cvcry discovery contains'an irratjonal .1.-.rrr, or -a creative

bringsabouta cerrain jlT."" of Popper's rapproche_

it: heputs

ji,

But in one point'the present opinions of most philosophersand scientistsseem to-cgree, namely, that the inductive procedure is not, so to speak, a mebiianicalprocedure prescribedby fixed rules. lf, for instance,a reporrpfobservational resultsis given, and we $tant to find a hypothesisi which is well confirmed and furnishes a good explanation for thdcvents observed,thei there,isno set offixed rules which would leadius rutoinaticaily to the best hypothesisor even a good one. It is a, matter of ingenuity and luck for the scientist to hit upon a suitab| hypothesis;.. . . This point, the impossibility of an automatic indirqtive procedure, has been especially emphasized, among others by il{arl Popper. . . , who also quotes a statementby Einstein . . . . Tha, same point has sometimes been formulated by saying that it is nbt possible to construct an inductive machine. The latter is presumably rneant as a mechanicalcontrivance which, when report, would furnish a suitable hypothesis,just fed an observatio,..al as a cornputing :irachine when supplied with two factors furnishes their product. I iam completely in agreement that an inductive machirre tftiskird is not possible.(Carnap,1950,p. 192) of The development:iince 1950 of ever more powerful computers has it reawakened hopes,:',hat might after all be possible to construct an

I ! "1:-;fili'-' ':': 1 1 ': I

,il

3Z

lnductivismand its Critics

Critique InductiVism of f:Popper's

33

automatic inductive machine. lndeed, a branch of the subject of artificial intelligerrce, known as ffiachine learning, has as its aim the progranlming of computers to produce generalizations when fed with data. In the next chapter we shall examine one approach to machine learning which has been developed by Herbert Siinon ancl his team at Carnegie-Mellon University. Returning to Carnap in the 1950s, we must next emphasize that, although he rejected an inductivist, or Baconian, account of scientific discovery, he accepted the Bayesian theory of scientific justification. He believed that scientific predictiorrs could be justified inductively by showing that they have a high piobability on the known evidenie. Popper, on th other hand, has always rejected both irrductivism as e theory of discovery and Bavesianism as a theory of justification.

in (First Book, XXV{D, which was quoted in 1.1. The passage questionruns as &jlows: I
Like Bacon, rqe iiright describe our own contemporary science - 'the method of reasoriirrg which men now ordinarily apply to nature'- as consisting of 'anticipations, rash and prematute' and of 'prejudices'. But these marlellously imaginative.and bold conjectures or 'anticipations' of ours.are carefully and soberly controlled by systematic tests. Once put fgrward, none of our 'anticipations' are dogmatically upheld. Our metlod of research is not to defend them, in order to prove how rightiwe were. On the contrtry,' we try to overthrow them. Using all i'lti weepons of our logical, mathematical, end techto ptove that our anticipations wcre false - in rrical armouiy, witry in their steed, new irnjustified, and unjustifiable order to put for*ird, anticipations. ., .,(popper, 1934, 278-9)

2.4

Some General Observations on Popper's Theory of Scientific Method summary of his theory of scientific

It should next bi observed that at the end of the passage quoted Popper explicitly iejects from the preface,l to his 1963 book, Bayesianism. He .aigues that 'these conjectures . . . can neither be (in the sense of established as cirtainly true nor even as "probable" the probability cal{ulus).' Of course, the Bayesian thesis is precisely

lFbpp.r gives the following lmethod:

t,

can progress just because f we Knowledge can grow, and. . . science can learn from our mistakes. i' The way in which knowledge progresses,and especiallyouri scientific knowledge, is by unjustified (and unjustifiable)anticipa-\ tions, by guesses, tentative solutions to our problems, by con-i by jectures. These coqjecturesare controlled by criticism; that is, byi refutations, which include severelycritical tests.They mrf attempt'ed survive thesetests;but they can never be positivelyjustified:they cari neither be established certainly true nor even as 'probabie' (in thiS as (1963,Preface, vii) p. i sense the probability calculus). of This is a very interesting passage,and I will make a number of comments on it. To begin with, Popper speaks of our knorvledge progressing 'by unjustified. . - anticipations'. A learned reader might suspect that there is here a hidden reference to Bacon, and a desire to rneke whet Bacon regarded as undesirable into an integral part of scientific procedurc. These suspicions would be quite correct, for, to a passage in his earlier Logic of Scientijc Discovery (1934), Popper adds a footnote referring to a section of the Nouum Organum

that scientific coniecturescan be made probable in the senseof of the mathematicalriblculus probability. However, Popper is not content just to cfiticize the Bayesiauattempt to justify scientific the he conjectures; staj,?s much strongerthesisthat such conjectures . cannot be justified at all, or, as he puts: 'theseconjectures . . can in never be positive$'justified'. He also speaks both the 1934and (my anticipations' the 1963passagbs 'unjustified(andunjustifable) r5f be of It emphasis). shor-ild clearfrom all this that Popper'scr.itique quite a number of differenttheses, diflerent of inductivism contailds a kinds and of diffeifht strengths.I will next try to disentangle few on of them, and to coirirment their plausibility. Let us take{6rstiPopper'sthesisthat ther6 are no such things as \ analogousto deductiveinfetences, which I by inductive infe*dhces, scientific generalifitions and predictions can be obtained from i datal.Popper suggestsinsteadthat all such general- | observational izations and predictions are-celjsgllles, and that the important i point is not how srich conjectriffiffiB6'iaitred - any method will do i - but that they should be tested severely once they have been J -:'\,
proposecl..i

This thesisof Fopper'sseemsto me plausible,and it also brings about a great sim$iificationin the theory of scientificmethod. We any curiousprocessof inductive inference do not have to pc,ftulate

'{i

:
1
' - - ,i a ,:- !.r ;;.-

.1;;l;1,i'i;il;l

36

and its Critics lnductivism

of l,poP"a, Critique Inductivism

JI

drawn from a single branch of natural science or from a sir.rgle historical period of scientific development. I will therefore supple[rent the Kepler example with two further examples drawn from twentieth-century science, and frorn the biological and medical field. The first of these is Alexander Flemirrg's discovery of penicillin. I will recount the story in more detail later, but its outlines are well known. Flerning observed a culture-plate which had becorne accidentally contaminated by a mould. I{is own photograph of the culture-plate is reproduced as Plate 1. Flen-ring concluded that the mould was producing a substance which had destroyed the bacteria growing on the culture-plate arrd which might, therefore, be a suitable entibiotic for the treatment of bacterial infections. Surely, here, observation preceded theory, end Fleming rhade an inductive irrference from his observation. My final exarnple is closely related to Fleming's discovery, but has some rather different features. It is the discovery of the sulphonamide drugs by the German chemical company I. G. Farben Industrie. The chemists working for the company examined literally hundreds of chemical compounds, until they hit, to some extent by accident, on one which cured mice infected by haemolytic streptococci. This really seems to fit the Baconian model of a mass of careful observations which eventually reveal a 'secret of excellent use'' These three examples, therefore, appear at first sight to fit the inductivist model rnuch better than the falsificationist model. But our further examination of them will show that this initial impression is to some extent misleading, and that an analysis in terrfls of coqiectures and refutations is, in each caset more plausible thah might at first be thought. Yet the apparently inductivist features of the three examples,ere not illusory, and our analysis will lead us eventually to a kind of synthesis between inductivism and falsificationism'

he assurned that the;Sun was a centre of force which governed the motions of the pldllets. If Kepler had triedlto relate Mars to the Earth rather than tof the Sun, he would never have found the elliptic orbit. He also begah wirh the theoretical assumption that the rnotion of heavenly bodiesj was either circular or composed of a small nunrber of circulaujmorions. This assumption (which Kepler larer rejected) had been rfiade by astronomers since the time of plato and

T I

Aristotle. Wrjh thesebackdiound assumptions, I{epler formulared his frst . hypotlrcsis: i' ..' The orbit oflMars is a circle round a centre C somewhat displaced frorli, the sun S, and its motion is uniforrn with respecr e pdint U, to O-n this hyporhesis i(lllustratedin figure 2. 1) the planet moves faster when nearer the Sr.*r,in conformity with the idia that the Sun is a centre of force infrirncingits motion. made g0il folio pages in small handwriting of draft cal-Kepler c'lations relating torithishypothesis.He basedit on-four observed positionsof Mars i{r.opposition, and it agreedwithin two minutes of arc with anothel ten oppositions.But he then wenr on to test the hypothesisagalqilst somi further observations Tycho's, and of this produced a deviation of 8'. This led Kepler to reject his first hypothesis.
Perlhellon,or narest polnt to Sun

2.5

Kepler's Discovery of the Elliptic Orbits of the Planets

Let us consider, then, Keplcr's discovery that the planet Mars moves in an ellipse with the Sun at one focus,r At first sight it seems as if Kepler inferred his law inductively from Tycho Brahe's observations. But let us look a little nrore closely at what happened. To begin with, Kepler started his investigation with a number of theoretical assumptions. He was already a convinced Copernican, and so related Mars's orbit to the Sun rather than the Earth. Indeed,

pointfromSun Aphelion, turthest or Figure2.1 Kepler;s fjli'st hypothesis

I T r!

Jtt

Inductivism its Critics and

iPopper'sCritiqueof Inductivism

39

r r
I

t, rl
rll I

Kepler next tried (second hypodresis) to compose the motion out of two circles. Mars moved round the second circle (or epicycle), whose centre moved round the first circle (or tleferent). The combination of deferent and epicycle produced an egg-shaped curve with the pointed end at thb perihelion, the nearest point to the Sun, and the broad end at the aphelion, the furthest point from the Sun. This again could not be made to fit Tycho's data. * Kepler next tried a third hypothesis, which is mathematically equivalent tp an ellipse. However, through a mathematical mistake, he got a cuive which did not fit. He then tried.an ellipse (fourth hypothesis) which worked, and subsequently realized that the third and fourth hypotheses were eqrrivalent. A closer examination of Kepler's work shows a pattern bf conjectures and refutations in accordancewith Popper's model. But two difficulties should be noted: 1 'I'he assumption about the circular morion of heavenly bodies could not by itself be directly refuted. Kepler was able to refute only specific hypotheses based on ir, such as his first and second hypotheses. 't'he data which Kepler used to test his hypotheses were collecred by 'llycho Brahe before these hypotheses had been formulated. lrr tlris sense, observation did precede theory.

t t t

l.ct us say a little more about these two points in turn. In connertiorr ra,iththc first, we should make some mention of the view.s ol 'l'lronrrs l(rrhn, a distinguished American philosopher of science wcll krrowrr ftrr lris analysisof scientific revolutions. In his farnous It)(r,l lrook, Krrhrr divides scientific developnrent into revolutionary 1,r'riotlsrrrrl rrorr-rcvolrrtionaryperiods, calling the latter 'normal ,;ricrrrc'. llr. tlrirrks thrt Popper's theory applies to revolutionary '.rirrrrr., lrrrt rrot to rrorrrralscience.As he says: 'l suggest then that Sir l(;rrl lr;rst lr;rrrclcriscdthe entjre scientificenterprisein terms that ;rlrply nl1ly to its or'r:irsiorral revolutionary parts' (19)0, p. 6). Ari ;r rrr;rtter llct, however, it is questionablewhether Popper's ol lltr.ory ol' r'orrjer'trrrcs arrd re[utations gives a satisfactory account nl rrrinrtilrr' revolrrtions. Irr general, scientific revolutions involve tlrr ovrltlrrow ol sorrrt: lrigh-level theoretical assumption and its rrrrrrrt lry :r nc'"v ;rssurrrptiorr. rr'1rl,rr The difficulty is that such lrrlllr h'vt'l tlrcorctir';rl rssunrptiolls carlnot be directly refuted by nlprr,r r,,rliorr.Wt' t:rrr orrly directly refute specific hypothesesbased ' ol llr { r t ( ' ; r + iur r l) li( ) lr s . l' lr us a c r ea t i v e s c i e n t i s ti s n e v e r f o r c e d b y

refutation to give,,up a high-level theoretical assumption. Kepler need never have abandoned the age-old assumption ofthe circularity of the heavenly mo'tions in order to introduce his ellipse. In the face of the refutation oflhis first two specific hypotheses, he could simply have added more aird more epicycles until a better fit was obtained. A sequence of r:trnjecturesand refutations of specific hypotheses neeil never be revoiutionary in charactdr. Moreover, such sequences are often found to tie the fine structure of what Kuhn calls'.ro.m"l science'. : Turning now to.'dur second point, it seemsundeniable that obserprecede theory (at least in the sensein which vations do sometil:j:.es Tycho Brahe's obs$rvations preceded Kepler's theorizing). But how do such cases diffdr from the obviously valueless random observations which Popper parodies? We have already given part of the answer to this qtreili:ion.Tycho.Brahe's observations were made at a time when a coniioversy was raging between the Ptolemaic and the Copernican tt i:bries of the world, and his observations were clearly ielevant io this controversy. Another'point worth noting is that Tycho tsrahe't obseivations were also relevant to a practical problem. Trade'ar,ll shipping were developing rapidly at the time, and brought with l:hern the demand for better astronomical tables which could be usrld in navigation. It is significant that the King of Denmark, who firi !.ced Tycho Brahe's observatory, derived a large part of his royal ir,:come from tolls on ships passing through the Danish sound, whlrlr connects the Baltic to the North Sea. discoveries to practical use, employing Kepler put his astii:onomical them to compute a new set of astronomical tables. These were called the Rudolfrie'Tables a{ter his patron King Rudolph. Il, and . were issued jn 762'1 They achieved new le*els of accuracv, com'*.." pletely superseding'all previous astronomical tables, and ,t once used for navigirtional purposes. Thus we may spy that it is qnly worth collecting observations if these Baconian "grapes. . . ripe and fully seasoned' are clearly relevant to a theoritical controversy or a practical application.
t'

2.6 Fleming's.)iscoveryof Penicillin:CredtiveInduction

I

The research worll ,which led to the discovery of penicillin began because Fleming h4d been invited to contribute a section on the staphylococcus grciirp of bacteria for the nine-volume A System oJ

40

Inductivismand its Critics

')-

:.I.?opper's Critique of Inductivism

41

Baderiology which was being produced by the Medical Reserrch Council.' Flerning's contribution did indeed appear in the second volume in 1929. Staphylococci are spherical bacteria which are responsible for a variety of infections. In particular, the goldenaureus is the common cause of boils, coloured Staphylococcr.ts carbuncles, and other skirr infections. While readirrg the literature on staphylococci, Fleming came across an article by Bigger, Boland, and O'Meara of Trinity College, Dublin, in which it was suggested that colour changes took place if cultures of staphylococci were kept at room temperature for several days. This interested Fleming, because the colour of a staphylococcus cen be an indicator of its virulence in causing disease. He therefore decided to carry out an experimental investigation of the matter with the help of D. M. Pryce, a research scholar. The staphylococci were cultured in glass dishes, usually 85 mm in diameter, known zs Petri dishes.These disheS were filled with a thin layer ofa gelatinous substance celTed agar to which enough nutrients could be added to allow the microbes to multiply. Using a platinum wire, some staphylococci were spread across the srrrfaceof the agar, and the plate was then incubated at a suitable temperature (usually 37"C), to allow the microbes to multiply. After this period of incubation, the dish was set aside on the bench, and was examined every few days to see if changes in the colour of some of the staphylococci could be observed. While this fairly routine investigation was continuing, Pryce left the laboratory in February 1928 to start another job, but Fleming continued the work on his own throughout the summer. At the end of July Fleming went off for his usual summer holiday, leaving a number of culture-plates piled at the end of the bench where they would be out of the sunlight. Early in September (probably on 3 September) when Fleming had returned from lris holiday, Pryce dropped in to see him. Pryce found Fleming sorting our rhe pile of plates on his bench. Discarded plites were put in a shallow tray conteining the antiseptic lysol. This would kill the bacteria, and make the Petri dishes safe for the technicians to wash and prepare for use again. Fleming's tray was piled so high with dishes that some of them were protruding above the level of rhe lysol. Fleming started complaining about the amounr of work he had had to do since Pryce had left him. He then selected a few of the dishes to show to Pryce. More or less by chance he picked up one in the tray of discards but above the level of the lysol. Accolding to Pryce's later recollection, Flerning looked at the plate for a while, and then

said: 'That's fr.rn1,' The plate was in facr the famous penicillin plateil., This is how Flbming himself described what happened in the paper he published,inJune 1929: varients a number of cultureWhile working v{1th staphylococcus plateswere set asi'de the laboratory bench and examined from time on to time. In the ex'rininations theseplates were necessarilyexposed to tie air and they bi;ame contaminatedwith various micro-organisms. It was noticed that around a large colony of a contaminatingmould the staphylococcuilcolonies becametransparent and were obviously undergoing lysis. fp. 226)
l.

Fleming's photcigraph of the original penicillin plate is reproduced as Plate 1,r.rand it is easy to follow his description when examining the photogtaph. The colonies of staphylococci are the small circular blobs, and the contaminating mould is very obvious. Near the mould tllg staphylococci become transparent or disappear altogether. They a5e obviously, as Fleming says, undergoing lysis, which rneansthe digSolution of cells or bacteria. From his observarion of the plate, Flemirig inferred thar the mould was producing a substance capable of dissolving bacteria. The mould was ideniified as being e Penicillium:t{t first it was incorrecrly classified as Penidllium rubrum, but later it .r,vas found to be the much rarer speciesPenicillium notatum. Fleming dccordingly gave the name peniiiilin to the bacteriolytic substanoir which he thought was being produced by the rnould. I The events desiribed so far may make it look as if Fleming's discovery was simply a matter of luck. Indeed, there is no dou-"bt that a lot_of luck _tjas involved. Hare subsequently tried to reproduce a plate similir to Fleming's original one, and ficund to his surprise that it wa{'quite difficult (see Hare, 1970, pp. 54-87). The general effect of ltleming's plate could be prodtrcid only if the mould and the staphylococci were allowed to develop at rarher a low temperature. .E,ven room temperature in the summer would usually be too hig!':, but here the vagaries of the English weather played their part. S'y examining the weather: records it Kew, Hare discovered that fbrinine days after 28 Jrrly 1928 ( just when Flerrrlng had gone on holi'{lay!), tirere was r siell of'"exceptionally colE weather. A final pqint is that the smain of penicillium which contanrinated Flemingrs,plate is a very rare variety, and most penicillia do not produce pei'ricillin in suflicient guantity ro give rise to rhe

J,I

I| il

42

Inductivismand its Critics

t
I I'

Popper's Critique Inductivism of 43 '\ insight on Flemingls.part seizethe opporrunity which chance to had presentedto him. ,l;{othing shows this more clearly than a comparisonof Flemingl.srreaction the contaminated to plate with that of his coileagues thellaboratory in (includingthe headofthe laboratory, Sir Almroth Wrigh6;, wheu he showed it to them. With characteristic candour,Hare ilescribes completelack of interestshown by the himselfarrdthe oth,:rs: tThe rest of us, b:ling engaged researches seemed more in that far
important than a iontam.inatedculture plate, merely glanced at it, thought that it waii no Inore than another wonder of nature that seemedt9,,be forever uncarthing, and promptly forgot all ilnt The piate was 41soshown to Wright when he arrived in the afternoon-What hersaid, I do not recollect,but. . . one can assume that he was no morb enthusiastic he could not have beenless* than the rest ofus had bcen that morning. (1970,p. 55) Fleming was by r-ro means discouraged by his colleagues' cool reaction. He took a ::rinute sample of the contaminating mould, and started cultivating itin a tube of liquid medium. At some later stage he photographed the. plate, and made it permarlent by exposing it to formalirr vapour, tirich killed and fixed both the bacteria and the mould. Fleming keprithe plate carefully, and it is now preserved in the British Museumi, So we have here a case of fortune favouring the prepared mind. i3ut what exacrly had preparcd Fleming's mind to realize somethin! ,which his colleagues missed? We can answer this question by gir:ng a brief account of the .researches which had occupied Fleming ir'i,'the fourteen years preceding his discovery of penicillin. i, .When the First ?ITorld War broke out, Fleming was already working in the in, r;ulation department headed by Sir Almroth Wriglrt. Wright, Fle;ieing, and the others were sent to Boulogne to deal with the rn'arw:i)nded. and, in particular, to try to discover the best way of treating :infected wounds. At that time wounds were routinely filled witl-'' powerful antiseprics which were known to kill bacteria outside the body. Fleming, however, soon rnade the remarkable discovery that bacteria seemed to flourish in wounds treated with antisepiics even more than they did in untreated wounds. The explanation of this apparent paradox was quite simple. In an untreated wtr'und the bacteria causing the infection were attacked by the body"'r: natural defences, the white cells, or phagocytes,

t
I
I| I
pr.nir:illin ltltt I lrlrrrrirrg's crrlture-plate. Reproduced with the permission o l \t M , ly ' r : I los 1r it : M c dic alSc hool,L o n d o n . rl lllrr I r v lr irlr lilc r r r ilr llobs c r v c d. I low d i d s u c h a r a r e m o u l d f i n d i t s rr,.ry ir r t o |h, r r r ir r li' s l: r bor et t r r y ?' I he m o s t l i k e l y e x p l a n a t i o n i s a r rlinn r , r r r t ' ' llr c r c w; ls ; lt t llat t inr e a t h e o r y t h a t a s t h m a w a s c a u s e d l'1, rr r , t r r lr lrplor , r , ir r 1qr t lr c bas em en t so f t h e h o u s e s i n w h i c h t h e ir \ ,r'.tlur ; r t i( lr v r , t l.' l' lr is t lr c or y was be i n g i r r v e s t i g a t e d , b yh e s c i e n t i s t t ((' | | ;r 'l'orrrlrc) irr tlrc laboratory immediately below Fleming's, ,rrrrll , r ' lir r r r lr r ' lr ; r t l ls ; r r c s ult a lar g e c o l l e c t i o n o f m o u l d s t a k e n ol lurr:rsul-ferers. seemsprobable thatpenicillium Ir r rnr f lrr' lrrrrrsr.ri :rs( lt uttltl unt w: t \ ( ) n( 'ol t lr < ' s c uloulds . 'llr r , r r , ir r o r lor r lr t t lr c r r t hat a gr e a t d e a l o f l u c k w a s i n v o l v e d rrr tlr t ' r lr r r ov r . r of lr c r r ic illiu. Yet i t s t i l l n e e d e d c r e a t i v i t y a n d y

t
Il

I I
I]

t.:l,ii,i .;: l1

:il'.

44

Inductivismand its Critics

,FoPPer'sCritiqueof lnductivism

45

which ingested the invading bacteria. If the wound was treated with an antiseptic, some bacteria were indeed killed, but the protective phagocytes were also killed, so that the net effect was to make the situation worse than before. Wright and his group therefore maintained (quite correctly) that wounds should not be treated with antiseptics. They advocated the earliest possible surgical rentoval of all dead tissue, dirt, foreign bodies, and so forth, and then irrigating the wound with strong, sterile salt solution. The medical establishment of the day rejected this recommendation, and so the superior treatnlent was accorded only to those directly in the care of Wright and his team, After the war, Fleming returned to the inoculation department in London, and here in l92l he discovered an interesting substance which was given the name lysozyme. Lysozyme was capable of destroying a considerable range ofbtcteria, and wes found to occur in a variety of tissues and natural secretions. Fleming first came across lysozyme while studying a plate-culture .of some mucus which he took from his rrose when he had a cold. He later discovered that lysozyme is to be found in tears, saliva, and sputum, as well as in mucus secretions. He extended his search quite widely in the animal and vegetable kingdoms, and found lysozyme in fish eggs, birds' eggs, flowers, plants, vegetables, and the tears of more tlran fifty species of animals. Lysozyme destroyed about 75 per cent of the 104 strains of airborne bacteria and some other bacteria as well. Moreover, Fleming was able to show that, unlike chemical antiseptics, even the strongest preparations of lysozyme had no adverse effects on living phagocytes, which continued their work of ingesting bacteria just as before. From all this, it seemed that lysozyme was part of many organisms' natural defence mechanisms against bacterial infection. Lysozyrne had only one drawback. It did not destroy any of the bacteria responsible for the most serious infections and diseases. The hypothesis naturally suggested itself that the pathogenic bacteria were pathogenic partly becauseof their to resistance lysozyme. .lf we put together Fleming's research on war wounds and his research on lysozyme, a problem situation emerges which I will call the 'antiseptic problem situation'. On the one hand, the chemical antiseptics killed pathogenic bacteria outside the body, but were less effective for infected wounds, partly because they destroyed the phagocytes as well. On the other hand, the naturally occurring antiseptic lysozyme did not kill the phagocytes, but also failed to destroy the most inrportant pathogenic bacteria. The problem, then,

was to discover thF,-'perfectantiseptic' which wotrld kill the pathogenic bacteria witiiout affecting the phagocytes. The work on lysozyme suggestepi that such an antiseptic might be produced by some naturally occiirring organisms. It is commonly lrernarked that'creativity consists in establishing a hitherto urlsuspecif;d connection between two apparently diflerent areas, or problem*ituations. This was precisely what Fleming did when he realized'd;ie significance of the penicillin plate. Instead of dismissing the co$fiaminated plate as a failure in his current investigation of the cd,lours of staphylococcin he saw it as pbrhaps providing the sol*ion to the antiieptic problem situation which had arisen from his edr-lier researches. In elfect, he must have con.iectured that the mould might be producing the 'perfect antiseptic' capable of destroiing pathogenic bacteria without disturbing the phagocytes. ),, The assunrption'that Fleming made sueh a conjecture is borne out by his subsecitlent actions. Fleming grew the mould on the and then filtered off the mould to ptosurface of , -.at,;broth, duce what he cat'ried'mould juice'. He then tested the effect of this rnould juice 6ri a number of pathogenic bacteria. The results were encouraging. The virulent streptococcus, staphylococcus, pneumococcus, gciiococcus, meningococcus, and diphtheria bacillus inhibited. In fact, mould juice was a more were all powerfufll powerful germiciclir than carbolic acid. At the same time, mould juice had no ill elfqcts on phagocytes. Here at last seemed to be the 'perfect antiseptic'.i At this point inlt-he story, however, a series of difliculties began to emerge. Sorne.' lurther experimental work suggested, misleadingly, that pedi'cillin might not be effective in the body. At the same time, it provbd difficult to isolate and purify the compound, and there were difficulties about storing penicillin in such a way that it would not rapid[1, lose its power to destroy bacteria. In his 1929 paper, Fleming wiiite: 'lt is suggested that it [penicillin] may be an efficient antiseptic for application to, or injection into, areas infected with penicillin-seriiitive microbes' (p. 236). Yet the problems just mentioned caused him to become despondent about the possible uses of penicillin ai an antiseptic; and not long after his paper was completed, he aba;:.Coned research in that direction. At this jur.rcttire,ifortune on ce again favoured penicillin. .Although Flerning abandone,l his earlier hopes that penicillin might be the 'perfect antiseptic',rhe found another practical use for it. This resulted irr a continued cultivatiorr of the penicillin mould and production of

'iiill

r
J T T
46 Inductivismand its Critics

F:,.pper's Critique !nductivism of

47

r| r|
t
T I
Il

I I
:

nrould juice. Thus both mould and mould juice were readily availablc when Florey and his team at Oxford decided, a decade later, to make another attempt to develop penicillin as an antiseptic. The rnain source of income of the inoculation departrnent where Flcmiug worked was the production and sale of vaccines. There was indeed an efficient unit for producing vaccine (a vaccine laboratory, as it was then called) within the walls of the department, and Flerning had been in charge of the production of vaccines since 1920. ln particular, a vaccine was made against Pfeiffer's bacillus (Ilttcilhts influenzae) which was believed to cause influenza and othcr rcspiratory infections. It was difficult to isolate this bacillus bccausecultures were apt to be swamped by other micro-or5;anisms. Fl<:rning, however, had discovered that penicillin, despite its effect on so nlany vintlent bacteria, left Pfeiffer's bacillus unaflected. By irr<'orporaling penicillin into the medium on which he was growing l'l't'ifl'cr's bacillus, he could eliminate the other germs, and produce goo<l sarrrplcsoltthe bacillus itself. Fleming in fact used this nrethod lirr prcpariug the influerrza vaccine in his vaccine laboratory, and was rnadc in the vaccine laboratory for this purpose every ;rcrri<'illirr wcck rficr its cliscovery. Significantly, the title of Fleming's 1929 lr;rlrcr()lr lrcrricillirrwas: 'On the antibacterialaction of cultures of a 1'critillirrrr witlr spccial refcrenceto their use in the isolation of B. itrllttrntlr', Florrirrg also serrt samplesof the mould to other centres r()n(crrrc(f witlr thc isolation of B. influenzae,and in this way culrrucs ol tlrc rrrorrltlwere established the Lister Institute, Shifneta at llrrivt ' r s it y M c t lir ' al Sdr ool, and a t G e o r g e D r e y e r 's S c h o o l o f l',rtlrohrgy irr ()xfirrrl.'l'hus, when Florey and his team decided to of tlkr trlr ir;,',uitr tlrrcstiorr whether penicillin might be an efficient tlrc ,rrrtirrlrt tlrt,y wcrc t[rlc to find samples of Fleming's strain of ir', ltrrtlrillirlm t,otiltt,,n.jrrst clown the corridor in the Dreyer School nl Itllrology whclc thcy were working. Flerning's work, like l(r'1rln's, illrrrtr;rtcstlrc c:loscconnection which so often exists bet\v('r. r s r ir r r lilic t lis r ' ov c r yr nd pr ac t ica la p p l i c a t i o n s . | (,t il3 ilcxl tty to cxarrrirrc relevance this important scientific of the r'l,rrorlc trr tlrc rlttcstiorrof irrductivism versus falsi{icationism.As a f f f ,ff f ('f f rf l;rlt, lilerrrirrg's rescarches appearto involve both induction rlrrl r otrilr'lrrrrs lnrl rr'firtatiorrs.On examining the contaminated rnivc<l tt two hypotheses.The weaker was that the lrl,rtr,,liL.rrrirry, rrr',trl,l lrrorltt,'ctlu lr;rr'(criolyticsubstance,while the stronger was tlr,rt t lr ir rr r r lr s l; r r r cr r r iglr tbc t lr e' per l t c t a n t i s e p t i c 'w h i c h h i s e a r l i e r c rr'rr',rrr ltcli lr;rrl lt'rl ltirrr to tlcsire. Now here, observation defirritely L,rl tlrr' lor u r;rt iorr ol- thc h ypotheses,and the hypothesescould lrrr.rr,r

certainly not have b;icn formulated without the observation. It thus seems correct to say phat these hypotheses were obtained by induc,sbservation. The hypotheses were, however, tion from a chance_ conjectures, and Flen-,ring proceeded at once to test them by preparing his mould juice an$'investigating its eflect on various kinds of pathogenic bacteria arld on phagocytes. As we have seen, the early tests corroborated ti:;e 'perfect antiseptic' hypothesis, though some later results gave r;cher misleading counter-indications. In short, we seem to have a case of hypotheses generated by induction from observations and then deveioped by the method ofconjectures and refutations. Dcis this mean that some kind of syntiesis of inductivism and falsificationism is possible? Such a synthesis lias been suggested by Mitchell (1989).3Mirchell emphasizes what cci.iid be called conjectural hiluction, and describes some features of thelprocess in the following two pessages: It is a srrrprisingch;rracteristic the coqiectural of process induction of that a useful theor'1" appears to contain more information than the limited set of singuh,r data from which it was originally induced. The useful theory must gontain, not only the pattern of the singular data as presented, but alsc the pattem of some natural generalprinciples of which the singr;lar data were symptomatic. The task of the imaginative researchworker is to guess what the gencral principles might be. (1989,p. il)
l'.

And again: Although, . . the inductive process may proceed by small and apperently undramatic.steps, thesestepstake the form ofguesses,'and the process theoiy-generationis non-deductiveand basicallyconof jectural in the sense.4escribed Popper. (p. 12) by In fact, reflection sho:ws that reconciling Popper's theory with some form of induction is not difhcult. Popper gives no account of how scienti.fic conjectures ioriginate, and he considers this to be a matter for empirical psychulogy rather than philosophy of science. It is therefore possible,,to supplement his account by claiming that scientific conjectures fometimes originate by induction from observation. This would be Mitchell's conjectural induction- As we have seen, this view fits v;ry well with the details of Fleming's discovery of penicillin. It is int*r-esting that Mitchell speaks of 'the task of the imaginative research ..4orker' (my emphasis); and indeed, I argued earlier that Fleming':; conjectural induction involved considerable

T
I
I

,.1
-. . . . _. *. , i. . , _

il,

48

lnductivismand its Critics

Pgpper's Critique Inductivism of

49

creativity. In cases like this, we could also use the term creative induction. We can, moreover, contrast it with .another way of generating conjectures which could be called creatiue theorizing. In the latter case the conjecture is formed by meditating on earlier theoretical developments rather than by considering observations. A nice example of creative theorizing is provided by Copernicus. Copernicus was not much of arr observational astronomer. His De of Orbium Caelestium 1543 contains only twenty-seven Reuolutionibus observations made by himsele and neither these nor any other new observations seem to have had any influence in the genesis of his ne* hypothesis. Copernicus arrived at his heliocentric theory in something like the following way. He becetne dissatisfied with the Ptolemaic models used irr the astronomy of his tirne. He therefore read many ancient Greek texts to see if an alternative approach could be found. In this way he lighted on the Pythagorean view that the Earth moved, and, starting with this hint, developed his own hypothesis. Here, then, we certairrly have ireative theorizing rather than any form of induction. Perhaps, horvever, Copernicus and Fleming occupy the two extreme points of a scale, and rnost hypothesis formation in science falls somewhere in between, being generated by reflection both on earlier theory and on new observations. Kepler in fact affords an example of such a mixture of creative theorizing and creative induction. So far, then, all our examples have involved some human creativity; but the question naturally arises as to whether some or form of mechanical Baconian induction rnight not be possible as well. 'fo pursue this question, we will examine in the next section another example - that of the discovery of the rulphonamide drugs,

2.7

The Discovery of the Sulphonamide Drugs: Mechanical or Baconian lnduction

chemists on thousands of infected animals to see if the compounds had any therapeutic 1:alue. The I. G. Farben chemists Hoerlien, Dressel, and Kothe produced a rich recl dye which,pas very effective with piotein materiais such as wool and silk. Thig was known as Prontosilrubrum. Domagk and his team then discov'bred that this same compound possessedthe definite ability to cure mice infected with haemolytic streptococci. Dornagk published this finding in 1935, but referred back to experiments carried out in {932. , This sequerrceof iiverrts can be described using the schema of coqjectures and refutitions. As each new compound was produced by I. G. Farben's inilustrial chemists, it was coqiectured that ir might have the abilipy to cure one or more bacterial infections. This conjecture was dhen tested by administering the compound to infected animals and'seeing whether any improvement resulted. In the case of nearly atli;ihe impounds piodui.d, the conjecture was refuted, but at last !;' compound appeared for which the corresponding coniecture #as confirmed. In this example, then, the conjectures were not pdpduced using any scientific imaginarion or creativity, but in a rc{itine fashion. The process can therefore fairly be described as wecl'.tinical But was it mechanical intluction?The answer seems to be '|.,1o'. At no point was there any inference of a conjecture from obseJjvations. The conjectures were gelrerated in a routine fashion, and';!ren tested experiinentally. The observations came after the conjebtUre, as in Popper's model. I.,will, therefore, call this procedure m$chanieal falsificationism raiher than mechanical induction.Ir should, hqwever, be added that mechanical falsificationism bears some resenlblence to the rype of induction advocated by Bacon, ,, | The discovery of tfiA sulphonamide drugs does i{deed have many Bacorrian features. To begin wirh, Bacon srressesthe desirability of tearn work in the scieices, As he says: It is not a wry ovet'1 which only one man can passat a time (as is the casewith that o{ reasoning),but one in which the labours and industries o{ men (es;:;cially as regards the collecting of experience), may with the best effect be first distributed and then combined. For then only wili men "begin to know their strength when instead of greatnumbersdoing;Ill the samethings, one shali take chargeof one thing and anotherof;rnother. (1620,p. 293) Far {rom etnphasizingl imagination, creativity, or genius, Bacon seems to want to m'ake science into a routine abtivity which could
, . :r.

The sulphonarnide drugs were discovered in Germany as a byproduct of the activities of the giant chemical cornpany I. G. Farben.a The discovery was made by a team headed by Gerhard Domagk, who was born in 1895 and appointed at the early age of thirty-two as director of research in experimental pathology and bacteriology in the institute attached to the L G. Farben works at Elberfeld. Domagk and his team had huge laboratories in which they routinely tested compounds produced by the firm's industrial

;i::i::r,

t;0

Inductivismand its Critics

Po;rper'sCritiqueof lnductivism

51

lrc <:;rrricdout by anyone of average intelligence. As he says: .But tlrc <''rrrsc I propose for the discovery of sJiencesis such * l."rr., Irrrt little t. thc acuteness and strengrh of wits, but places all wits rrrrrl rrrrderstandings nearly_ on a level' (p. 270). Thus the army .l sr:icrtists in Domagk's laboratories performing rouiine tests oi-i gcnerated hypotheses seems to conform well to Bacon,s '.rrtincly itlcrrs. M'rcovcr, all this activity did i'deed result in what Bacon crrllccfa 'secret of excellent use' (p. 292). ll;r<orr strcsscd that he wanted to introduce a new form of induc_ ti.rr wlrir:lr was not induction by simple enumeration, but involved r,xr'lusiorr ;rrrrlrejcction. 'llhis is how he puts it: 'llrrt rhe 8r'ent('st changeI irrtroduceis in the forrn itself of induction rurrrl thc.jutlgrrrcrrt nrade theteby. For the induction of which the lop,ili;rrrs spc;rk,whi<Lrproceedsby simple enumerdtion,is a puerile tlrilrp; r'orrr'lrrrles hazard;is a.lways rt liable to be upset by a con_ Irurli|tory irrstarrr:e; tlkes into account on.ly what ls known and rrt(liil;ry;;riltllc;rtls rro restrlt. io Now wlmt thc scicnccsstand in need of is a form of induction whir'lr shrrll;rnrlysccxperience and take it to pieces,and by a due rncl rejectionlead to an inevitableconclusion.(o. lrr.r err rl ext'lrrsi.rr tr) ,r,l lly 'irrrlrrr ti.r by sirrrlrlccrrumeration'Bacon means something like tlr. irrrlrrr'ri.rr li'.rrr tlrc observation of severalthousand white ivans l, tlr r . ' r r r lr r s i' r r r lr r t t he nex t obs e r v e ds w a n w i l l b e w h i t e . B a c o n r.p , ir r r l' i . t lr ir . , s ' r r lr t r c r ilc t hir r g' . No w w h a t h a p p e n e di n t h e c a s eo f l\uttril lrlrr,rr w:rs rlr* 'exclusion arrd rejeciion' of a very large rrrrrrlr'r ,l r.rrrp.rrr<ls urrtil eventually one of therapeutic valie \ '.rh rlrir rivcrcrl. llccausc of all these parallels, I will use tbe term llrtrrtrtiilrtittlrttriott ;rs r syrorrym for nechanical falsificationkm. Arlr r r lt t . r lly t lr is t c r r r r ir r . logy c ould b e a l i t t l e m i s r e a i i n g s i n c e , a s ,rlrr,rrll' |',irrtrrl rrr, Ilar;r:lrianirrductiorr is not reaily iiduction at ,rll ll;rrolr. ril.rr()vcr, fhils to take account oI'a rather crucial diflrr rr lt y wlr ir ' lr we lllt s t llex t c olt s ide r . ll,rr,r r.r'rrrs r'lrevc llrorrght that in any particular i'stance there u ,ill lr t ' r r r r ly ; r li. w lr r s s iblc hy pot h e s e sa v a i l a b l e .T h u s , b y q u i t e a llrrrr llrrr'f is .l 'cxt'hrsiorr anclrcjectiorr',we will be led toih. t.rrth ,r,,', r nr r r r . v it . r lr lr . r or r < ' lus ior rr' . ac t , h o w e v e r , i t i s o f t e n p o s s i b l et o lr f llr.r r , r , r t r wit lr r . ; r s t;.ur ( . n( ) r nlousnu m b e r o f h y p o t h e s e s ,a n d t h e r e . rr'ry r r ) r lr t ' s ullit ir : nt .t ir r c c r r r c s o u r c e st o t e i i a l l o f t h e m i n t h e lr'rlx' . { . r llr ir t ir r ll . r r . r r c whic h wor k s . T h u s i t w o u l d s c a r c e l y b e lrlr,llrlr' t{r t('\t l )l tlrer:rpeuticproperties every single compound

which the chemists of today are capable of synthesizing. In this situation there has to Lrerecourse to what are called heuristics. A heuristic(from tl;r: Greek heuriskein,'to discover') is a guide to discovery. In the conii:xt of mechanical falsifcationisrr, hypotheses are generated by some rc,iitine or mechanical procedure; but in practice this procedure is unlikely to be totally random. It will usually with some heuristic. Even the search probe devised in accordai:"ce cedure which lead tc, :he discovery of Profttosil rubrum, the first of the sulphonamide dr:-gs, was guided by various heuristics. One of these was the idea th:it dyes capable of sfaining textiles might also have useful therapeuiri properties. The 'dye heuristic', as it might be called,s had been introduced before Domagk by Paul Ehrlich. Ehrlich discovered tha': if certain dyes are injected into living orgrn: isms, they are taken rp and stain only some particular tissues and not others. Ehrlich gives the following example, which played an important role in his discovery of the method of 'vital staining': a Thus, for example, methylene blue causes really wonderful staining of the peripheral ner','oussystem. If a small quantity of methylene blue is i4lected into a frog, and a the small pieceof the tciigue is excisedand examined,one sees finest twigs of the nerves beautifully stained, a magnificent dark blue, againsta colourless lackground- (1906, p. 235) Ehrlich goes on to observe that this specific staining property is lost if the chemical com:-.osition of the dye is changed even to a small extent. Thus he says, I was able to prove that the nerve-steining property of methylene blue is conditioned by the presenceofsulphur in the methylene-blue molccule. Synthetic chemistry has, in fact, given us a dye which, ofsrrlphur, correspondsexactly in its chem.ical apart from the absence constitution to methyleneblue. This is BINDSCHEDLER'S green. of the With the absence the sulphur, thcre is associated inability to stain living nerves.ilbid.) In the light of lhese interesting discoveries, Ehrlich reasoned somewhat as follows. Suppose we know that a particular disease is caused by the in.rasion of some micro-organisms. To cure the disease, we need to.find a chemical which is highly toxic to these micro-organisms, br,t which does not harm the patient. This can be achieved if we can i,nd a chemical which kills the micro-organisms and which is taken -:p only by the rnicro-organisms and not by the

ii i 'r '

52

Inductivismand its Critics

$$pp"r'. Critique tnductivism of

j-

53

other tissues. Now dyes like methylene blue are highly specific, in that they are taken up by some tissuesand not by others. Marry dyes are also toxic. So it is not unreasonable to think that some dyes might have good therapeutic properties. Indeed, Ehrlich was able to show that his favourite, methylene blue, was helpful in curing malaria. As he says: 'In my further experiments . . . I started {tom the supposition that dyes with a maximal tinctorial activity might also have a specialaflinity for parasiteswithin the host-orgenism, . . . . I chose the malaria parasites and was able, in association with to show that methylene blue can cure Professor GUTTMANN, malaria' (p.2a\. So the'dye heuristic'proved successfulfirst for Ehrlich, then for Domagk. L:onically, however, it turned out that the therapeutic properties of Prontosil rubrum hle nothing to do with its ability to dye fabrics. The molecular structure of Prontosil rubrum is shown in figtre 2.2, where the hexagons are the benzene rings whose discovery by Kekuld we described earlier, and, as usual, N denotes one aton of nitrogen, S ofsulphur, O ofoxygen, and H ofhydrogen. It is clear that the molecule consists of two halves joined by the double bond denoted by :. In the trody, four hydrogen atorns are added to the molecule through the action of enzymes, and the molecule splits into two different molecules, sulphanilarnide and tri-amino-benzene (see figure 2.3).
NHa

Now it turned tji,rt that only one of these molecules (sulphanilfor amide)is responbib! killing the pathogenicbacteria.Sulphanileither textiles amide, however, hi; no power of dyeing or :staining or bacteria! lr' This concludes iily accountof Popper'scritique of inductivism. In the next chapter'{will considerDuhem's critique of inductivism, of a critique which sh.iilslight on somefurther aspects the problem.

ll

1,
I

.ir :
tl ';

H2Nsoz+,r,:Nt>^*
1,

Figure 2.2

The rnolecular structure of Prontosil rubrum !:.1 l.tH2

t.

rl i
HzNSO2

H"N

/\ \:/

NHz

Sulohanilamide

Tri-amino-benzene

Figure2.3

Reduction of Prontosilrubrumin the body

Duhem's Critiqueof Inductivism

55

3
Duhem ' sC ri ti q u e f o I ndu c ti v i s m

3.1

Inductivism as the Newtonian Method

|) t r l1r ' nr ' s r it ic luc < ; f indt r c t iv i s r n i s c o n t a i n e d i n h i s n r a s t e r l y b o o k t rrrr lrlrifosttlrlryof s<:icncc: The Aim and Slructwe oJ Physiral 'fhcory, ns wlrit lr lirst :r1rpc:rrcr{ a seriesof articlesin the years 1904 and 1905. ' l' lr c , r t t ; r r k on ir r t lr r c t iv is nloc c u r s i t r c h a p t e r 6 , s e c t i o n s4 a n d 5 . ' llr or r lilr or r ly a f t ' w pagc s lone , t h i s i s o r r e o f t h e m o s t i m p o r t : r r r t p I r , r t is lt l( 'ir r t wc r r t it : t h- c ( : nt ur y h i l o s o p h y o f s c i e r r c e . s l) r r lr c r r rr c f i. ' r s o ir r < lr r c t iv is ran 't h e N c w t o n i a n n r e t h o d ', a n d h c t -s r r lr ( ) ( lu( ( ' sit ir r t lr c lir lk r wing w a y : It rv.rstlris. . . th:rt Ncrvtorr had in nrirrd whcn, in the 'Gcncral \r lrolrrrrrr' r.r,lritlr r'rou,nshis Prirrripir, rejccted viqorouslyas hc so r r ul, , t r h' r r : r t r r r lrhilos ophy; r n y h y p o t h e s i sh a t i n d u c t i o nd i d r r o t ol t Il r , x lr , rl r lr ( ) nr r ' r lr c r ir r r c r rwho r l l c a s s e r t e dh a t i r r a s o u n d p h y s i c s t; t rvcry proprrsiliorr slrorrltl drarvnfionr phcnonrena generalizcd bc and ( l( lr y t t r r lu, t r or r . ) O , l- 5, pp- 19 0 - l ) ('MatheNcwlrrrr's Itltilt,soplri,rt' Natroa/i-sPriuripia Mallrcnratica r r r , r t ir , r ll) rir r tilr lt ' s ot Nat ur al P h i l o s o p h y ') , o r P r i n c i p i af o r s h o r t , r v . r r lr r s t lr r ir r t r , rir r 16u7.ln it . . N e r v t o r r p u t s f o r w a r d h i s t h r e e l a w s l r ' l lr r ot ior r , r r r t lr is l: r ' "v gr av i t y , a n c lu s e st h i s s y s t e n 'r f t l r c o r e t i c a l l of o lnr ' rlr , r lr rr t o t ' x pl: r ir rt hc r no v c n r c r r t so f t h e s o l a r s y s t e m a n d t o , r r( nr nt lir r t lr c t it lc s nnd nLl n l c r o u s o t h e r t e r r e s t r i a l p h e n o r n e r r a . l' lr r ' ; ' , r r ' , rtt' r r r ; t ir ir ' : r lu( c c s sof N c r v t o r r 's t h e o r y l e d t o i t s b e c o m i n g s , r r{ r ' lr l1' 11 y v ilt r r : r lly t hc c ut i r c s c i c r r t i f i cc o m n r n n i t y i n t h c e a r l y lr , L' r , r , lr r r r l t lr c t ' ililt t c c ut h c e n t u r y ; a n d N e w t o n i a r r m e c h a n i c s t I r r r r , lr r , ' , 1lr c r ' or r r t ' r s t onc ph v s i c su n t i l t h e o n s e t o f t h e r e v o l u t i o n of r r I lr y r 11 ir r llr r ' t ' ; r r ly 1' c ar s t h c t w c r r t i c t h c e n t u r y . r of l) t t lr r ' r r r i, ; r pr r it c ( ' ( ) r r c c t t o l i r r k N e u , t o n w i t h i n d u c t i v i s r n .

Ncwton inclr-rdes account of scientific method in his Principia, an and this account is indccd inductivist in character.It is more doubtful rvhcthcr Dtrhenr is hir to call inductivism 'the Newtonian nrethod'. As we have secn, the indtrctive method was formulated by Francis lJacon before Newtor.r was born, and Bacon is indeed thc probable sourcc of Ncwton's views on rnethod. Bacon had, in [act, studicd at 1'rinity College, which was where Newton spent his Canrbridge days. Although irrductivistn shor,rld not be entirely identified with Newtorr's theory of scientific method, it will, none the less, be useful briefly to cxanrine Newton's own version of irrductivisrn, both bccause of its historical inrportance and because it is this version of inductivisrn which Duhem attacks. Newton's accourlt of scientific method is set out in a section of the Principia which hc entitles 'Rules of Reasorringin Philosophy'. Rule 3 is formulated as follows: I'he qualitieso-fhodies,uhieh admit neitherintensijution nor renissionof deXrccs, uhith arefound to bcktn.q all bodies and to within the rcath o-[our the experincnls,are to he esleemeil universalqualiticsoJ all hodiesu'hatsoevu.(1687,p. 391)) Tlris rulc is clcarly dcsigncd to infer universal laws or gerrcralizations from finitc sets of observations. Newton goes on to clairn tlrat lre had obtained his law of gravity in this way. Newton's law of universal gravitation statcs that every body in the universe attracts cvcry other body. The magnitude of this force of attractiorrbetweetr arly two bodics is giverr by tlre procluct of their massesdivided by tlrc sqtrareo[the distancebetweerr thcm. Thus if massesr\ afid n2, are a distance r apart, the force of gravitational attraction betweert thern is given by F : ()ft11n2112, where G is the universal constant qf gravitation. Newton clairrrs,in the following passage, to have inferred some aspects of this law irrdtrctivelyusing his Rule 3. obscrvall it universally appcars, expcrinrcnts by and astronornical tions. that all bodiesabout tlrc earth gravitatetowards the earth, and that in proportion to the quantity of mrtter which they scverally contairr; that the rnoor.rlikcwise, according to thc quantity of its lnatter,gravitatcs towrrds the earth;that, on the other harrd,our sca

' Ja

t - a: a' . . 3-

?^C is a' i: s

Duhem's Critiqueof Inductivism

37

lll one torrartls.rnotllr'r: ur.rvltelcstt.ir-;rrlsthc r.rrrrorr:;rrrtl tlrc pl;rrrcts :rrrtl thc c()lncls irr likc rrr,lnncr ros-arc"ls tlrr- strrr:rve lliust. irr rr.rrsc(prcnccof- this rtrlc Ii.c. llulc 31, universallv allorv that al] bodies wllr(socvcr arc crr<lorvcdwith a principle of ntutual gravitation.

(1,..)ee) Irr thc ncxt rulc (llulc 4), Ner,vton states his inductivisrn more c l ca rl y a r r dex plic it ly : In txptrinrcntrrl philosophy dtc to ltttrkulort propositions u,c infcrred hy inductionJi'ou phenomena acaraltly (tr uerynearlytrue,,lottuithns ,qureral that startdin.g (ontrdryltypotlrcses nray Itc inapined,till suchlinrc as any other o(afi, hy whkh theynay eillrcrIrcnrade phcnonc,ta nore arcurdla, ot Iiahlcto cxccptions. This rulc wc rnustfollow, that the argument inductiorr of may not (p. bc cvadctlby hypothescs. a00) Thc airn of tratural scicr)ce (or 'experinrental philosophy', as Ncwtorr calls it) is to obtairr propositiorls 'inferred by general rnductiorr frorn phcrr<lnlcna1. N cw t ot r dc v elops th c s c i d c a s a l i ttl e fu rth c r i n thc (l orcral Scfr<rlitrrrr which l)rrhcrn rcfcrs. This was added to the Princiltiain to tl rc se co r r c {dit ion of 1 7 1 3 .F l e re , N e w to n s a y sth a t al though he has c cl i sco v c r c d hc laws go v c rrri rreg ra v i ta ti ()n a a ttra cti on, hc sti l l does t l rro t kn o w t hc c aus c o f g ra v i ty i ts e l f. A s h e p u ts i t: ' Ll i therto w e h a ve cx plainc d t hc phe rro rn c n a f th e h e a v e n sa n d of our seaby the o p()wcr o F gr av it y , but h a v e n o t y e t a s s i g n e d e c a useof thi s pow er' th (Pritrtipia, (]crrcrrl Scholiurn, 1713, 1>.546). l -l c tl r c r rc or r t ir r r r c s : llut hithcrto I hevc rrot bccn ablc to <liscovcr thc causeof thosc propcrtics gravity frorrr phcuor-nurr, of and I franrcno hypotheses; is ftrr ',vh;rtcvcr rrot dt'drrcccl frorn thc phcnornena to bc callcd is :rrr hypothcsis; arrd hy1-rothcscs. rvhctlrcrrrrctaphysical phvsical, or rvhr:thcr occult c1u:rlitics rncclranical. of or hrvc no placein cxpcrinrcrrtelphilosophy.In this philosophyparticularpropositionsarc infcrrcd frorr thc plrcnonrcrra, aftcrwardsrendcrcdger-rcr:rl arrd by i rrd u c t ior r . ' l- hus sth a t. . . tl rcl a u ' so f n ro ti o na n dof gravi tati on, it wa wcrc cliscovcrctl. 547) (p. I-'l ypot lr c s c t hor , fb r N c w to rr, a rc s p c c u l a ti o nsw hi ch arc l tot s irrlcrrccl by gctrcral indtrctiorr frorn pherrorrrena. He clain'rsthat he h i rrrscl I c loc sr ot put fb rw a rd s u c h h y p o th e s e s a n d thi nks that other t , l l a tu ra l s c ic r r t is t s hou l d i rtri ta tch i s e x a rn o l e .

It is wor:th nlaking anothcr couplc of comments on this larnous and irrteresting passaeefron.r Ncwton. First of all, Newton rnakcs the typical llaconian inductivist's corlflation of discovery and justifi* cation. The laws of motion and of gravitation were discovcred by inductive inference frorrr phenomena, and are also justified irr thc sanlc way. Secondly, at ollc point irr tlre first sentcnceof the passagcfrorrr page 547, Newton no longer speaksof inference by gcncral induction from phctronrena,but directly of deduction frorn phcrronrcna. l -hi s i s sonrcthi l rg of i rnpor t ar r cc[ or , as Lakat os point cd ( ) ut , induction and dcduction are oftclr conflated ilr thc seventccnth arrd ei ghtccnthcorturi cs. A s L akat os hir nself says: and In thc scvcrrtecnth cightccnthccnturicsthcrc was no clcar disfor arrd'dcduction'.(lndcccl, Dcsclrtcstinctionbctwecn'inductiorr' ittter alios- 'inductiorr'and 'dcduction' were synonymous tcrrns; of he did not think much of the rclcvance Aristotelirn syllogistic, which incrcaselogical contcnt. lt'rforrnal arrclprcfcrred inFcrcrrccs 'Cartcsian'valid infcrcnccs hoth in nrathcmatics atr<lscicrrcc contcl rt arrdcarrbc char act cr izcd by an inf ir r it yof vali<l i rrcrcasc only (l96tt, p. 13{)) pattcrns.) Lakatos perhaps exaggeratesslightly here. After all, Hunrc, as wc have secn, did have a concept of what could be obtained try rcasorr, and denicd that enrpirical laws or predictions could be dcrivccl irr thi s w ay. Thus Fh.rme,alt hough he lived in t he eight ecnt h ccr r t ur y, does, in effect, distinguish betwecrr dcduction and induction. On the other hand, Lakatos is broadly correct. The trraj<lr thirrkers(with the exception of Leibniz) rcgardc(l severrteenth-cerrtrlry Aristotelian logic as just arrother piece of stcrile scholasticisnr, and they no longer bothered to study it. A renewed intercst irr and developnrent of fornral logic did llot occur until the tinrc of I3oole and Frege in the ninetcenth century. In the seventccrrtharrd eighteenth cerlturies, therefore, logic was treated infornrally, arrd such an informal approach does not bring out clcarly thc distinctrorr betr,vccn deductiorr and inductiorr. ' l hc si tuati on w as ve r y dif [ c'r or t f <r r I lusscll, his ( , - ar r r br ir lgc fol l ow ers, the V i cnna C i r cle, and l) opper . Thesc t wcnt ict h- ccr r t r r r y thi rrkershad a vcry prcci seconceptof dcduct ivelogic, as f t : r nr alizcr l, for exanrple, in llussell and Whitchead's PM. They could tlrus rn:rkt' a very clear distinction betweerr deductive and inductivc Iogic, Perhaps dre distinctioll was even a little too clear. This corrr:lrrrlcs

-.-

:1.,t ". i

' i {F

;f

iI I t

lls

58

Inductivism and its Critics

of Duhem's Critique Inductivism

59

l,::

or r r br i c f a c c o u n to f N c rv to rr' sv c rsi otrof i nducti vi srn.W c w i l l next c or r s i rl c r )u l rc l rr' sc ri ti c i s rrrs . l

3 .2 N e w to n ' s In fe re n c e of the Law of Gravi ty fro m K e p l e r' s L a w s a n d D uhem' s Obj ecti ons from which Newtorr clainred to infer his law of T'he pherrorrrerra Iudeed, Newtotl clevotes gravity indrrctively irrcludcd Kepler's lar.vs. a corrsiderablesectiorr of the Principia to deriving the law of gravity fronr Keplcr's laws. I-lis argument is too tcchnical for us to expoutrd in dctail, but we carrexplain the general idca in inforrnal ternrs. Lct us bcgin with Newton's frrst lar.vof motion, which he himself statesas follows: in Every body cortlitures its state of rest, or of unifonn motiott in a to that stateby right [i,e. straightlline, unlcss is eompelled chatrye it inrprcssed upott it. (1687, p. 13) _forces No w c o n s i c l c r p l a rrc t,l ), tn o v i ng round tl tc S un, S . A ccordi ng a to Kcpler's laws, its orbit will bc arr cllipsc with thc Surr at orrc f<rctrs ( s ecf i g u rc 3 . l ). ' fa k e d re p l a n e t at a parti cul arpoi rrt i n i ts orbi t. If no forces werc actiril on it, it wor-rld, by Newton's first law of nrotion, corrtinue moving with unifornr velocity along the straight lilre showlr as a scries <lf dashes.However, it actually moves in an ellipse. So, Newton reasoned,a force nrust continually be acting on the plarret,pr-rllingit away from its natural straight line path, round into tlre curve. A nrathematical calculation now shows that this force rnust be dircctcd towards the Surr, and must vary ilrversely with the squarc of the distance of the planet from the Surr. Have we lrere valid infcrcnce of Newton's law of gravitation from the phenornena? Not according to Duhelrr, who writes:

gravitation merelya generalization thc of Is this principleof universal provided by Kepler'sLaws and their cxtensionto thc two statcments motion of satellites? Can induction derive it frorn these two statements? Not at all. In fact, not only is it more gerreral than thesctwo statements unlike them, but it contradicts and them. . . grauity, veryJar Jrom being derivable The principle of universal by generalization iniluctionJrom observational anil the laws oJKepler,formally rcntrailictstheselaws. If Newton's theory is correct,Kepler's Laws are neeessarily (1904-5, p. 193) falsc. Duhem is quite .orr".t in what he says here. If we consider the Sun and a plarret in isolation from all the other bodies in tlrc universe, then from Newtorr's theory it follows that the path <lf the planet will be an exact ellipse. Flowever, it will be remembered that, according to Newton, every body in the universe attracts every other body. So our particular planet P will be attracted trot just by the Sun, but by all the other planets of the solar system. These other gravitational attractions will perturb the orbit of the planet, causing its path to deviate slightly from a true ellipse. So, according to Kepler's laws, a planet moves round the Sun in arr exact ellipsc; whereas, accordirrg to Newtorr's tlreory, it moves on an cllipsc with srnall pcrturbatiorrs due to thc gravitational attractions of the othcr bodies irr the solar system. Thus, Newton's thcory formally contradicts Keplcr's laws, and if Newton's theory is correct, Kepler's larvs are necessarily false. It appearsto follow from this that Newton's theory cannot be derived either inductively or deductively from Kepler's laws. If we suppose otherwise, we have an instance of deriving from given premises a conclusion that corrtradicts those premises; and this seems absurd. Duhem goes on to observe that Newton's theory is actually validated by using it to calculate the perturbations of the planetary orbits, arrd then showing that these predicted perturbations agrec with those actually observed. This is how he puts it: Thercfore, if the certainty of Newton's theory does not emarratc from the certainty of Kepler's Laws, how will this theory prove its validity?It will calculatc, with all the high degree approximation of that the constantlyperfectcdmcthods of algcbrainvolve, thc pcrturbationswhich at eachinstantremove every heavenlybody from thc orbit assigncdto it by Keplcr's laws; thcn it will compare thc perturbatiorrs calculated with the perturbations observedby mcansof the most preciseinstrumentsand the most scrupulousmethods. (pp. 193-4\

Figurej.l

A plarrct rnovirrg round thc Sun

60

Inductivism and its Critics

r. , '

t:i

'

: -, , .

61

Wh c-' r rlat c r or r , in 1 0 .5 , w e c o n te to c x a mi n e the w ays i n w hi ch , scierrtific dreorics carr be confirrned by evidence, we will reach a corrclusionwhich Inrgcly agrccswith what l)uhem says here. Such, thcn,. is l)uhern's nrair"rarguntent against Newton's indrrctivisrrr.Ile lras, however, in the sarle place, a subsidiary argunrcrrt which is itself of corrsiclerable interest. l)ulrcnr argues that, irr the coursc of his supposcd c{erivatiorrof thc law of gravity, Ncwtotr has to rc-dcscribe Kepler's laws using th e n r c c hanic al i o n s o f ' fo rc e ' a rrd ' n ra s s ' .Yet thi s i s a dubi ous not nlove, since Kepler's laws can be stated in terms of positior.rs, vclocities, distances, areas, times, and so or1, without ever lnentionirrg foqcesand nlasses.As Duhern hirnself says: 'only dynarnics pernrits us . . . to substitute statemerlts r:elative to "forces" and "n ra s s es "f or laws r cl a ti v eto o rb i ts ' (p . 1 9 a ).T h e probl ern i s how i t is p<lssibleto dcrive, either incluctivcly or deductively, a theory irrvolving the new concepts of force and rnass from a set of laws which do not irrvolve these concepts. Any such thing seernsindeed to b e higlr ly pr ot r le n ra ti c .r This concludcs rrry account of Duherrr's critique of inductivisrn. In the rrext sectiorr I will try to put tlris critique irr its historical setting and to corrsider its validity, particularly in relation to the discussions thc prcvious chapter. of

3.3 Criticisms lnductivism of and the Revolution Physics in

The errornrous succcssof Newton's scientific work lent authority to thc irrductive nlethod by which he clain-redto have reached his rcsrrlts. It rnay therefore not be a coincidence that Duhem's arrd l)opper's criticisnrs of inductivisnl are contemporary rvith the revolution in physics wtrich showed that Newtonian mechanicswas inadequate in sonrc respects. I will next attempt to explore some possibleconnectiorrs. Tfrc years 1904-6 saw dre enlergence of the special theory of rcl a tiv it y . E ius t ein p u b l i s h e dh i s a c c o u n to f th e new theory i n 1905. As wc shall sec itr the next chapter, Poincardshifted from defending Newtorrian nrecharrics arguing for its replacementbetween 1902 to and 1904. Those who maintain that Poincar6 was arr independent discoverer of special relalivity base tlreir claim on his 1906 paper. Now, as we have seen, l)uhenr's Tlre Aim and Structure Physical of
I i I
I

'I'lrcory first appe:rrcd irs t Sontbs",*fan&C&Sryi2i,',r yrr",,, $04 -{ ard w as then publ i sl rccli n book f or r r i ir r ivii6. I s t lr is'ii: i; : : i', . . f io4?r / coinciderrce? was tltcrc sotrtc cotrtrcctiolrbctwccn I)uherrr's work Or and the begirrnirrgof the rcvolution in ph/$ics? ttf ln fact, most of the ideas contained in The Aim and Stnrcture PhysicalTheory wcre introducecl in a sericsof papers which l)uhclrr publishcd bctwccn lB92 arrd 189(r,and which are cotrvctricntly c()llecteclin Jaki (cd.), l9tl7. Howcvcr, as l]rcturcr poirrts otrt (199()b, l) pp. 33()and 334), thes cear ly ar t iclcsdo not cor t t ait r uhcnr 's cr it icisr n of i rrducti vi sm.Thi s was f ir st publishedin M ar ch and Apr il of l9( ) 5. It is very unlikely that Duhem was influetrccd by Einstcin, sitrcc, as we shall see later, l)uhern continued as late as 1915 to rcjcct Einstein's theory of relativity as an aberration of tlre Germalr nrirrd, and it is improbable that he read Einstein as early as 1905. On the other hand, it is very irnprobable tlrat Duhem did rrot rcacl some of Poincar6's rcflections on the subject before 1905. l)uhcnr undoubtedly followed Poincar6's work on physics and thc philosophy of physics; and, indeedi nrany of Duhenr's ideas emcrged as developmcnts or criticisnrs of Poincar6'sviews. In his 1904 lecturc, Poincar6 annourrcedhis corrviction that Newtonian mechanics was inadequateto explain tlre nerv experimental findings in physics, ancl called for the developnrent of a new nrechanics. Perhaps it was readirrgthis article which inspircc{Duhem to make his owrr critiquc not of Newtoniar"rmechanics,but of the Newtonian method. The first phase of the revolution in physics began with the enlergenceof the spccial theory of relativity around the year 1905. and development of quanturn The second phasesaw the appearance mechanicsby Heisenberg, Schrodinger, Bohr, Dirac, and othcrs ilr the years 1925-8. Once again, Newtonian mechanicswas shown to be inadequate- this time in thc micro-world of the atorn. It secms not urrreasonable suppose that this second blow to Newtonian to mechanicswas one of the factors which stimulated Popper's critiqtre of the Newtor.rian rnethod (indtrctivism), which appeared n 1934. This corrjectureagreeswith Popper's own reminiscences, for Poppcr writes: At the timc (1930)whcn, orcouragedby Hcrbcrt Feigl, I bcgan writing my book, rnodcrnphysicswas in turmoil. Quantum mechanics had been createdby Werlrer Heisenbergin 1925, but it was several years bcfore outsiders- including professionalphysicistsrealizedthat a major brcakthroughhad been achievcd.And from the very beginningtherewas dissension confusion. (1976,pp. 90-1) and

)
i'--'- " "'--'-'--'-:-'.. .t;:i.-:',. .. :t.i. :l

t
I I

62

Inductivism and its Critics

of Duhep'sCritique Inductivism

63

rl rf
I

r|
I I I
I

:l
I

T I I I
I I

M<rrcovcr, Itoppcr's book 'l'fie Logic of Scientific Discouery, published in 19. |4 , c o n ta i n s rro t o rrl y h i s c ri tici srn o[ i nducti ' ri sm, but al so l wholc ch;rpter (the ninth) devoted to a discnssionof tl.rephilosoplr ic al pro b l o rrs o f q rra n tu n rth c o ry . I-ct us now try to evaluate l)rrherrr's critique of inductivism, partictrlarly with rcgard to the discussions cbapter 2. Inductivists of srrch rs Ncwton hacl high hopes o[ developing a method of inductiorr wlrir:h woultl bc very sinrilar to logical deduction. tn logical tlc<lrrt'tiorr, corrclusiorr derived frorrr the premisscs,using a few thc is s ir r r lr lc rrrd g c rrc ra lru l c s . If th c p re rn i sses accepted certai n,thc are as r'orrr:lrrsiorr t()() llrust bc regardcd as ccrtain- The inductivist plarr, t lr or , w; rs to tl c v c k rp a rn c th o d o f il rducti orrw hi ch w otrl d enabl e l;rws ;rrrrl tlrcorics to be derived fronr observations using a few sirrrplc :rntl llcrrerll rules. Even if these laws and theories could rrot bc cstablishcclwith certainty, the hope was that they could lrc slrowrr to lrold, givcrr the observations, with great probability. Nt ' wlor r l rl q rrc tl th a t h e h a d o b ta i n e d hi s l aw of gravi ty from' the Ke ; r lr c r r or rre rrai 'rrr:l rrc l i l rg p l e r' s l a w s, by j ust such a method of , ir r r lr r c t i o rr, rc rr a tta c k c dth i s c l a i m of N ew ton' s. The heart of hi s l )trl c t it ic is rrr i s r:rrrrta i rrt'itl th e fo l l o w i ng passage:' The pri nci pl e of rr rtttivt'ts'il and .gtttlit1,,vrty .liu'.from heing tleriuableby generalization ittlrtrliott lrou lhr ohscrvati<nml lau,s oJ Kepler,formally contradicts these / , t r ls '( l( )0 4 5 , 1 r. 1 9 3 ). l) r r lr u rr' s l ro i rrts c c rrrs l l l e to te l l heavi l y agai nstthe possi bi l i ty to ol ; r r r r t ' tl ro rl l ' i rrrl rrc ti o na n a l o g o u st o l ogi cal deducti on. It scarcel y o s n' r r r sp o s s i b l ci o rl c ri v c , b y a n y th i n g l i ke l ogi cal deducti on, a con, lr r s ior r w l ri r l r l o rrrrl l l y c o n tra d i c ts the premi sses.Orr the other lr ; r nr l,l) trl rc rrr' sc l i ti c i s rrrd o e s n o t u nderrrri nethe possi bi l i ty of the k r r r t f of to tti t' tl u rtl i ttd u tti ttrt r w h i ch w e argued i n the previ ous fo r lr , r pt c r , ' l ' l rc re i s rro rc rs o n w h y .a creati ve sci enti st shoul d not lr onr t lrc s trrtl yo l ' :r s c t o f o trs e rv a ti ons l ed to a conj ecturew hi ch be s lr;ut ly ( orrlr;rtlir't l ltc obscrvations. Strch a conjectrtre could be r r r r r r r c r l i ;rtc tc s tr.rlb y s c e i rrgw l ' re therthe devi ati onsfrom accepted ly olr s r , t v ;rtro rrsl ri c h i t l rrc d i c te da c tual l y.occurred. fhct, N ew ton w In s ( ' ( ' lt s t o l r:rv co l l t:ri rrc dl ri s l a w o f g ra.vi tyby a mi xture of creati ve llr r oliz l rp ' ,l ttttlc o rrj r:c l trr;rlrd u c ti o n. l l oughl y speaki ng,hi s creati ve il t lr . . r r z r rrg r.rrr tl rc c rrrl i c rw o rk s o f I)escartes, i l eo,and H uygerrs l Gal lr ' r l lr r r trt(!;t s y s t(' n lo l ' rrrc c h a n i c s hi ch i nvol ved the new concepts w , l l. r r r ' ;rrl rl rrr:rs sO o rrj c t:tu ra il n d ucti on from K epl er' s l aw s thcn . , h' , 1lr ir u l o tl rt.ro rrr:l rrs i o rr a t th e force of gravi ty must vary as the th r r r v r ' r , i r' s (l rl rrc 1 ' tl rc tl i s ta rrc eT h e si rrrpl eaccount w hi ch I gave i n o . . 1, . 1 r lr o v c l rro b rl rl y c o rrc s p o n d sto a part, though onl y a part, of ;

his reasoning.A much fuller accourrtof the historical origin of Newton's law of gravity is given by Duhem in 1904-5, pp.220-52. Regarding this, Ilrennershrewdlyobserves:
schcma that after rejectingthe inducti.zist It is by no meansacciderrtal of the transitionfrom Kepler'slews to Newton's principlein part two clraptersix of 71e Aim and Strudure,Duhem givcs a long account of the historicalgencsisof the principle in the next chaptcr.This to accountis clcarlyintendcdas an altcrnative the inductivistrccon. struction. . . History of scicnccthen providcsthc missinglirrk for Dtrlrcm'srcjcctionof induttivist methodology.(1990b,pp.331-2) A s w c might cxpect , I ) uhem 's account on t he wh<llc em phasizcs what we have called 'creative tlreorizing'; yet he sometimes mcntions developn-lents which could be described, in our tcrms, as 'conjecturalinduction'. Thus, for example, he says:'Newton through his own efforts discovered the laws of uniform circular motion; he compared these laws. . . with Kepler's third law and recognizedas a result of this comparison that the sun attracted equal masses of different planets with a force inversely proportional to the square of ( the di stan ces' 1904- 5, p. 251) . This concltrdesnry exposition and evaluation of Duhem's critiqr.rc of inductivism. It is just one of several important contributions by Duhem; we shall meet some of the others later in the book. Irr the next chapter we shall consider some of the contributions to philosophy of science of l)uhem's brilliant contemporary Poincard. lt is thcrefore worth pausitrg now to say something about the lives of these men, who must surely be reckoned to be two of the most outstanding philosophers of scienceof the twentieth centrlry.

3.4 The Livesof Duhemand Poincar6

P i erre D uhem was bor n in Par is on 10June 1861, and died at his country home of Cabrespine (Aude) on 14 September 1916.2At thc age of twenty, Duhem entered the Ecole normale supdrieurewhere he studied theoretical physics, He was a brilliant student, and received first place in 1885 in the competitive examitration for teaching physics. Yet already Duhem had given offbnce to the French scientific establishment. The trouble began with a dissertation on thermodynamics composed when Duhem was twentythree. This work contained very outspoken criticisms of Berthelot,

:'iri'i
,.'.i

.r,i,., i

64

Inductivism and its Critics

Duhem's Critioueof Inductivism

65

tlrclr r. powcrfirl lrrcl ilrflucrrtial figtrre, and it was, as a result, rcjcctcd. Mrrcovcr, ro position in Paris became available for the youlrg f)ulrcln. After brief srays at the science facuhies of Lille rrrd llcnrrcs, l)uhcrrr trecanreprofessor of theoretical physics at ljordcaux.at the asc of thirty-two. He hcld this post till his deatn. 'fhouglr he later becar'e reco*ciled with lJerthelot, Duhem made furthcr encrrries, and was rrever on fricndly tenns withJhc parisian scierrtificestablishnrent.He was a n1ar1 strong personality, very of houcst, outstandingly brilliant intellectually, a.d with firr'lv rreld. but urrusual,convictions. [n short, he was just the sort of i'dividr-ral likcly to corne ilrto conflict with an intcllectual establishment. In additio' to all this, Duhenr was a dcvout catholic, arrtl held vcry conservative political views. l*le was therefore sonrewhar out of place in the liberal, anticlerical atmosphere of the Third llepublic. Duhem is thus the mirror image of the members of the vi"r,rro Circle, who were, for the most part, liberal and anticlerical, and consequentlyat odds with conservativeand clericalcirclesin Austria. Yet drere is sontethirg to be said for Duhem's opponents, for Duhern was one of those ur.rfortunate scientistswho,^ crespitegreat intcllectrr.ef brilliance, seem to have an unfailing instirrct [o. aclop'ti,rg approaches which prove to be unsrrccessful.Duhenr l-racla love of abstract nrathernatical theories, and tried to develop gerreral thernrodyrrarrrics and an.errergeticist programme sirnilar to those of ostwald and Mach. l{e rejected the attempt of Boltzr'an' and Gibbs to reduce thernrodynamics to statiitical mechanics, and attackcd the i'troductiorr of atorns i'to physics. yet, of course, it was atornic approach which proved successful.Simirarly, iu ^thc thc field oI electricity, Duhenr attacked Maxwell's electromaslretrc theory, and supported the ideas o[ Helmholtz, which ,r."r,,o* largely forgotter. He also failed to appreciate the importance of Lore'tz's theory of electrons, a.d, as late as l915, wrotc a r.rolemic agairrstEinsteirr'stlreory of relativity. 'I'his is not to say thatbuhenr did rrot make sonre co,tribtrtions to physics but, o' ihe w.hole, he was urlsuccessful that field. irr I{is studies o[ histo,ry and philos'phy of science hacl a very differerrt outconle. Duhenr's first book on the history of science (L'Euolutiotr de la nticanique) was published in 1903. ifr;, *r, followed by Les origines de la statiErc(1905-6) and a massive,threevolur : nes t udy of L e o n a rd o d a V i n c i (1 9 1 3 ). A l so i n 1913. there appcaredthe first volurrre of his nronumental systimedu montle.This was planned as twelve huge volurnes covering the development of astronolrly and physical theory lronr the pre-Socraticsto Galileo.

lly the time he died in 1916, l)uhem had completed ten of thc volumes single-handed, and had published five of these. Duhem's nrain achievement in the history of science was a revaluation of the medieval period. Prior to him, sciencehad becn thought to have ended with the Greeks and to have begun again irr the sixteenth century. Duhem showed that the medieval scholastics had considerable scientific achievements to their credit, ancl that these achievements had a notable influence on the scientific rcvolution of Copernicus and Galileo. It is clear that Duhem's religious position influenced his work orr the history of science.Errlightenment thinkers had seen Catholicisnr as an enemy of science,and lrad thought that sciencecould flourish and develop only if it managed to shake itself free of the obstructivc influence of the Church and its superstitious dogmas. I)uhenr wanted to show that, on the contrary, sciencehad flowercd in tlrc medieval period under the aegis of the Church. He also wanted to show that the Enlightenment's hero Galileo had taken many of his ideas from the medieval scholastics, and that the Church had not been altogether wrong in its criticisms of Galileo's theories. I)uring this period of researchon the history of science,Duhcnr was also developing his ideas on the pbilosoplry of science. As wc have seen, 'fhe Aim and Stntctureof Physical Theory appeared first as a in seriesof articles in the Repuede philosophie 1904 and 1905, then in book form in 1906, with a second edition in 1914. Despite this intensive activity in history and philosophy of sciencc, Duhem never abandoned physics, and indeed he always, rather perversely, regarded his work in physics as of rnuch greater ilnportarlcedran his contributions to history and philosophy of science. Thus when, in the 1890s, an enquiry was made as to whether hc would be interested in a professorship of the history of scienceat the Colldge de France in Paris, he replied with the remark: 'l am a physicist. Paris will obtain me only as such, if I ever should retrrrn there.' Then again, when in 1913 he pr epar edhis Not icein suppor t of his candidacy for membership of the Academy, his account of his work in philosophy was only one tenth as long as his account of his work in physics, and was everr shorter dran his acconnt of his work as a historian. On this point, the judgement of history is definitely against Duhem's assessnlent himself. of It is interesting to compare Duhem with his great French conternporary Jules Henri Poincar6 (1854-1912), for there exists a curious series of similarities and differences between the two thinkers. Poincar6. like Duhem. was trained in mathematics and

trl)

Inductivism and its Critics

of Crittque Inductivism Duhem's

67

i |' lr y s it s ,[ ' rrr, rrrrl i k cl )u h u n , h e w a s h i ghl y successfulrr thesefi el ds, . r lr ( l( ' nr irl rc tl tl rc rrr w i th a s c ri e s o f b r i l l i ant contri buti orrs.Indeed, lr . r ' ; r r rl; r y r' l l i rrr, ;rl o rrgw i th E i n s te i' ancl Lorentz, to be one of llr r ' lir r r r rrl .rs.l ' th c s p c c i a l th e o ry o f rel ati vi ty. D uhem conforms t qt r it cr ' los t' l y to th c ro n ra rl ti c s te re o typeof the i nnovati ve thi nker s I or t r by i n n c r (' ()l rtn ttl i c ti o n a n d a t odds w i tl r hi s soci al envi ronr nt ' r r t . l' o i rrr' ;rrtr, y c o l rtra s t, a p p e a r sto have had a cal m, w el l b lr , r l; r r r c t 'p c rs o rra l i tya rrtl to h a v e fi tted comfortabl y i nto hi s soci al rl r r r ilit ' r r . f t r r ir r r ' :rrd:rs b o rrr a t N a n c y j rrto w h at hi s obi tuari sti n Tl rc Ti mes w rr's rfcscrif us 'nrr okl hourgcttis family'. I3y contrast, Duhem's fanrily r,v:rs rlir;tirr<tly pclit hourge<tis, since his father was a textilc salesman wlr o nt ' v t ' r trrl tl t' rn rr< ' ln ro n e y . Po i n c ardbegan hi s educati onat the r ll, t f u ol' N rrn c y , rrn tl tl rc rr p a s s e dw i th great di sti ncti on through the l' iloh. lr oly tr.c h rri r;rrcl.c o b ta i rre dh i s doctoratei n 1879,and shortl y l ; r llc lw; r r rl so h l ;ri rrc rl l i s fi rs t a c a d e n i i cappoi ntrnentat C aen. H ow l c v c r , lr r s s tl y i rr tl rt. p ro v i rrc e sw a s a short one. Tw o years l ater, at llr r . : r 1r ic l ' tw e rrty -s c v o r, h c rc tu rn ed to P ari s. Frorrr then on a o s r r r t ' s s io rr l ' w t.l l -rl c s c rv e cllro n o u rs w ere aw arded to hi nr. In 1B B 7 c o I t , ' w; t s c lc < ' te rl l rrc rrrl tc r f th e Ac a d 6nri edes sci ences. 1889 he l o In wolr lr r ir rtc rrra ti o rrrl ri z c o f 2 ,5 ()0c row ns and a gol d medal offered p lr y t lr c l( irrg o l ' S w c tl c rr fo r w o rk o rr t hc three-body probl ern. The l: r c r r r ' lrllo v c n rn rc n t tl rc n rn a d c h i m a rncrrrberof the Legi on of llor r nr r r ' . l n l ()Otll rc w :ts c l e c te d to th e A cad6rni efranqai se. Thus, wlr r lc l ) u l rc rrr l :rn g rri s l rc tln th e p rovi nces, P oi ncar6 establ i shed i lr r r r r s r . ls tl re rrrrrl i s p rrrclrl a d e ro f F re nch mathemati csand physi cs ll c r t r l' . r r r s , l' oir r lr r rti ,l i k r. l )rrh c rn , b c g a rrto ta ke an i nteresri n phi l osophi cal r gur ' s t ior rs rrrrtll ()(X ),a n d l rc re , I t hi nk, w e can see agai n the :rro t t r llt t t ' t r c c l tl re b c l l i rrn i rrgo I th e tw enti eth-century revol uti on i n o r ; r f y s irl. l l c tw c t' rr l 9 O 2 a rrc l h i s e a rl y death i n 1912, P oi ncar6 tlrler. plrikrsoplricalworks: Science and Hypothesis(1902), 1'rrlrltslrcrf ' l' hr l' t lnt o l ' S ri u rtt' (1 9 0 5 ), n d S c i e n ce Method(1908).A fter a and hi s r lr , r t lr ,lr is l rrs ttl ro rrl i l rts r) rh c s u b j e c tw ere pr,rbl i shed book {brm. o in l' oir r r ' . r r rtk .l l s w i tl r p h i k rs o p h y o f rn a themati cs w el l as phi l osoi as plr y ol s r i c rrc r' ; rrt, u rrl i k c l )u h e m, h e never di d researchon the b I ils t ( ' ty ol s t' i c rr< ' c . l' r r r r r ' : rrri ' ls ri l tl s o p h yo f s c i c n c ei s k r.rowas conventi onal i sm, I f and rl will t lt ' s cri l rt' i t i rr t' h a p tc r4 . It a p p e a r ed the same ti me and i n the at s ; r r r r r ' s o r' i rrl l i c rr s l )rrl ro rr' s , a n d there are, as w e mi ght expect, rrri n r r ; r r y s ir rri l :rri ti r.s rt.tw c c rr th c i r p h i l osophi cal posi ti ons. S onre l phi wr ilc r s lr :rv t'c v t' l r r' l :rs s i fi c <)r-rh e m ' s l osophv as a ki nd of conll

ventionalism. This, however, seemsto me a mistake.3Duhem often criticizes Poincar6 in sharp terms: for example, 1904-5, p. 1a9f' where l)uhem attacks an article of Poincar6's published in the Reuae de mttaphysiqueet de morale in 1902. Duhem's position could better be described as modifcd.falsifcationismthan as conventionalism' Let us move now from differences to similarities. Duhem and Poincar both write with an adrnirable clarity and precision. Their elegance of style shows them to be the heirs, in this respect, of the classicFrench philosophical tradition of Descartesand Voltaire. Another characteristiccommon to Poincar6 and Duhem is also one which distinguishesthem from Rttsselland the Vienna Circle. This is the absencein their writings of any use of formal logic. Both men write very logically of course, bgt their logic is informal rathe,r than formal. Duhem seems to have taken no interest in the new formal logic being developed by Frege, Peano, and Russell;4 whereas Poincard studied it, but rejected it as sterile and useless.Poincar6 also rejected Russell's logicist view of mathematics, and conducted a controversy with Russell on this point. Poincar6'sside of the debate and is reprinted in his 1908 book Science Method (part II, ch. 3-5)- It ..r,rr1itut.r a brilliant and witty polemic against Russell, Peano, and the other logisticians. For example, Poincari writes: 'l find nothing in logistic for the discoverer, but shackles. ' . . if it requires 27 equations to establishthat I is a number, how many will it require insight, to demonstrate a real theorem?' (p. 178). With characteristic Poincard comes close to anticipating Godel's incompleteness theorem. Yet history has shown that his attacks on formal logic were misguided. The new logic is now constantly used tlrroughout mathematics, and has become an indispensabletool for computer science. Poincar6 drew his inspiration for his philosophy of science and mathentatics from the then current practice of science and mathematics in which he was actively engaged. Duhem, as we have secn in this chapter, used episodes from the history of science to criticizc certain philosophical views and to support others. The contrast between Duhem arrd the Vienna Circle is an illustration of the difference between historical and logical approaches to the philosophy of science. Those who adopt the logical approach attempt to give a logical analysis of scientific inference and of the structure of scientific theories. The title of Quine's book Fron a Logical Point of View (1953) clearly indicates the approach of its author. Carnap's Logical of Foundations Probability (1950) is an extreme example of the logical

:l;:.:'1'

68

Inductivism and its Critics

of Duhem's Critique Inductivism and 3.5 Artificiallntelligence the Revival of lnductivism

69

approach. This nrassivebook is filled with forrrral logic and abstract synrbolisnr, but drcre is no dctailed discussion of even a single c p i so d e fr onr t he his t or y o f s c i c n c e . The historical approach to plrilosophy of scienceis quite differcnt in character.The adhcrentsof this approach like to test their rrrodels in of scienceby seeirrghow well they can account for case-studies is o.f tlre lristory of science.lf Carnap's Logicall;oundatiotts Probability t he 'i d e a l t y pe' of t he l o g i c a l a p p ro a c h , Ku h n ' s ' I' he S tmctureof Scientijc Rcvolutionshas a corresponding position in the historical approach. Kuhrr actually begins his book: 'History, if viewed as a repository for more than anecdoteor chrouology, could produce a in decisive transforn"ration the irnage of scienceby which we are (1962, p. 1). He illustrates his philosophical views now possessed' with a great rnass of material from the history of science, but nowhere does he enrploy any of the syrnbolisnl or techniques of forrnal logic. Much can be said for arrd against both these approaches. Those who favour the logical approach will n<l doubt accusethe historians of being vague and woolly, in contrast to their own precisiorrand T'he historians might perhapsreply that the lnodels of the exactncss. logicians rnay be very precise, but they bear very little relation to any actual science,past or present. The ideal philosopher of science would no doubt cotrrbir.reelements from both approaches. The historical approach unquestionablyproduces greater realism, but the logical approach has received a boost from the ongoing computer revolution. Formal theories of science are tailor-made for impleIncntation in"the field of artificial intelligence, whereas the tnore intuitive insights of the historical school would need a great deal of processirrgbefore drcy c<luld be used in this way. I'erhaps the logical approach gives tlre science of computers, while the historical approach presentstlre scienceof hutnan beings. This is one exarnple of how the development of artificial intelligence is affecting philosophy of science, but certainly not the only 'We have seenhow inductivism was undermined by the revoluone. tion in physics. In the final section of this chapter, I will show how the contenrporary revolution in conrputing is restoring an interest in inductivisrrr.

t;i

.::

a'': :'i

'.:j

In recent ycars a branch of artificial intelligencehas developed which learning.Those who study machine learning try to is called machine write programs which will enable a computer, when fed with data, to output a law or laws which thesc data satisfy. lrr othcr words, machilre learning attempts to implement lJaconian itrduction olr a computer. Some workers in -the ficld have even claimed remarkable Thus Langley, Simon, Bradshaw, and Zytkow write: 'We successes. will describe the computer progranr BACON.1, which as its name implies is a system capable of making scientific discoveries by induction orr bodies of data' (1987, p.25). Moreover, they go on to give the following list of physical laws discovered by BACON.l: B oyl e' s l aw , K epl er' s thir d law, G alileo's law, and O hm 's law ( p. 86). This makes it look as if Baconian induction has been definitely established as a reality. However, the claims made here (like so many in the field of artificial irrtelligence)must be treated with a certain degreeof scepticism.After all, BACON.l has only succeeded in discovering laws already known to those who wrote the computer program! Let us look at the caseof Kepler's third law a little more closely, to see what has really been achieved. Let D be the distanceof a planet from the Sun, and P its period that is to say, the tin're the planet takes to complete its orbit. Kepler's third law statesthat D3 lPz: c (a constant) for all planets. is described in Langley et al., 1987, The program llACON.l pp. 66-86. In the caseof Kepler's third law, the computcr is given data corrsistingof valucs of D and P, and has the task of finding a law relating D and P. It is also given heuristics (pp. 66-7) which amount to asking it to look for a law of the form D-Pn : a constant, where m and n are integers. The program BACON.l now enables the parameters rz and n to be estimated from the data, and it comes up with m : 3, n : -2. Can we correctly describe this as remaking Kepler's discovery of his third law? It seems to tne that we cannot. The problem, of course, is that the computer is in effect told (i) which two variables to relate, and (ii) the general form of the law it should look for. The really diflicult part of Kepler's discovery was finding the information in (i) and (ii), and this BACON.l does not do at all. Once the information in (i) and (ii) is given, the problcnr reduces to estimating two parameters from the data. This step

,1

::it;:i: ii

I I .....".-".-1

/0

Inductivism and its Critics

Duhem's Critiqueof lnductivism

71

lln ( : ( ) N . I c l o c ss trc c c e d n c a rry i r rg out, but i t i s the easi est i step; nr ( ) r c o v c r, tl rc rc i s rro tl ri n g s u rp r i si ng i l r a computer program c s t it r t ; tti ttgt [c ' w p a ra n re tc rsrr a g i ven nrodel . i i 'l'o sr:t'rrrorc clcarly the great differerrccbetween Kepler's original t lis c ov c ry rrrc l c o rrrp u te rp ro s ra m s o f the B A C ON type, i t w i l l be lrcllrlirl lo rctrlrn bricfly to our earlier example of Kepler's first law. again the caseof the third law. Kepler's first Wt' r'rrntlrcrr corrsiclcr l:lw st;rtcstlrat all planetsrrrovein ellipseswith the Sun at one focus. l(t'1rlcr olrtairrcclthis law by studying Tycho Brahe's data on the ollr it o l Ml rs . [.c t u s n c x t s e e w h y a nornral sci enti stof K epl er' s h t ir r r t ' , c v c rr i f, p c r i m p o s s i b i l e , e h a d had the hel p of a computer rn I r r ( ll' ; r ;rnrrrrr< ;l r o re p o w e rfu l th a n B A C ON .1, w oul d have fai l ed I o < lis r' o v t'lr(c p l c r' s { i rs t l a w . ( ) r r r l ry p o tl rc ti c ;lrro rrrr:rl c i e n ti s t oul d, of course,have adopted s w l o t lr t ' r k .trri n ;rrrt r;tra tl i g trtf h i s ti m e - that i s, P tol emai castronorny. 1 ( ) or r s ctl rrt' rrtlly w o u l c l h a v e re l a tedthe orbi t of Mars to the E arth rc t uk t ' r r ;rs fi x c tl , :rrr< w o trl c l h a v e l o okcd for a l aw consi sti ng of a l crrriccl by a deferent. His lrypothetical stronger rcr it's of' c1ri<'y<'lcs (;o I t A ( : ( ) N Irro [!ra r)r l l l d tl l c ]nh a v e checkedthesedefererrt/epi cycl e th r r r or k ' l s ;r1 .1 ;ri rrs tc c l :rta b y a d .j u sti ngal l the paramcters - fbr arrd epicycles, their respectivevelocities, cx;rrrrplc,r:rtlii of' tlcficrcrrrt r ur r r lo o tr .a rrtl h a v c l o r.rn d e v c rs i on w hi ch fi tted best. N o doubt, s th and pow erful cornputi ng techwit lr : r s rrl l i c i c n ttrru rrb c r f e p i c y c l es o trirlttt's, ;t gootl lit worrlcl have been found, but Kepler's first law wor r l, l n c v t' l l r:l v c l rc c rrd i s c o v e re d. he t ( r ' 1 rl t' o l rt;ti trt' tlri s l a w s o rrl y b ecause had previ ousl y come to r l ; r ( ( ( ' l) t tl r(' trc w O o p t:rrri tr:tu re o ry. The consi derati ons hi ch l ed tl w I r ir r r t o rk r s o w t' rc s rr[rtl ca rrd p h i losopl ri cal : del i cate,qual i tati ve a I r ; r l; r n rr1 1 l ' rtrl l ttn tc rrts r a n d a g ai nstthe r.rewtheory, an admi rair o fb p l ri I r or r lo r l ' y tl ru 1 1 o rr' :rrr l o s o p h y ,a n d perhapseven N eopl atoni c sun , wor r ilri p . Al l tl rc s c l rc rrri s ti c s w h i ch w ere of vi tal i mportance hi sl t r r t t ur l l y , ;rtc ;rl rs c trtro rrr th c IIAC O N programs. M or c o v n , r' v r' rrOo p c rn i c l n i s rn w as not enough. To obtai n hi s I r r s t l, rw , l (r' p l c r l rl tl to ta k c a fu rth er revol uti onary step. H e had to ; r lr , r r r rl olnrt' s r' :rrc l rl i rr a l l w i n te rnrs of a combi nati on of ci rcul ar l r r r ot iu rrr;rrrrl i rrs tc :rrl l a w o f a di fferent form (the el l i pse).Thi s try r r lr ' gr11 v 1 1 1 r;rrl i cttc o tl rr> trs a rrd arsof astronorni cal ye w tl tradi ti on, and w, r \ lr .l t t;rk t' tr v t' u l ry Oo p c rn i c u sor Gal i l eo. t' ll w t' r(' l u r' n to tl rc c u s co f l (e p l e r' s thi rd l aw , w e can seethat the ! i, unl l )o i n l s rrl r| l y ' 1 ' 1 tr' 'a ri a b l e sc onsi deredare D and P . B ut D , v t lr c r lr ,;t:rrrr' t' o l ' l1 rl :rn c fro rrr th e Sun, w oul d have had no si gni fi r t p lr r r r c i n tl rt' rl o n ri rr;rrrt a ra d i g m of P tol emai c astronomy. N o

normal scientist of the time would have considered it. Still rnore remarkable was Kepler's search for a law of the form x'nyn : a constant. Eliipses had been studied in antiqrrity, but a law of the form x3y-z: a constant was a complete novelty. Of course, today, more than three and a half centuriesafter Kepler, laws of a generalized polynornial form have become a commonplace of scientific culture. This is apt to blind us to the greatnessand originality of Kepler in introducing such a law for the first timeThe work of Langley, Simon, Bradshaw and Zytkow in their 1987 book is certainly of value and interest; but their claim to have produced computer programs capable of making major scientific discoveries by induction on bodies of data must be treated with some caution.s This is not to say that nothing has been achieved in the field of machine learning. On dre contrary, Stephen Muggleton and his colleaguesat the Turing Institute in Glasgow are developing what seemsto be a much more hopeful approach to the subject.6 I will concltrde the present chapter by considering whether our discussionof lJaconiarrindr.rctionin 2.7 gives any indications about the possibility of inrplenrerrting it on a computer. I argued carlier that l3aconian inductionis really eqr.rivalentto mechanical.falsi.ficationism and can be consideredas a procedure of conjecturesand refutations. The conjectures are generated not by the creative insight of a scientist, but by some routine or mechanical procedure carried out in accordance with some heuristic. The conjectures formed in this way are then tested against the data until, hopefully, one of thcm turlls out to fit the facts. How might such a procedure be implcmented on a computer? The best approach would seen'tto ir.rvolvc co-opcra t ion bet wccn com put cr scient ist sand cxper t s wor king ir l the field in question. The experts could provide the heuristics, and the computer scientists could then write programs for generating with these heuristics and for testing these hypothesesin accordatrce data. hypottresesagainst the We can illustrate this approach by imagining a team of computer scientiststiansported back in tinre with their computer and an electricity generator to the year 1600 and given the task of helping the astronomers o[ that epoch to disqover simple laws governing the motion of the planets. llow woulH this team find the most aPproWell, they would clearly priate heuristics for generating hypotheses? begin by interviewing the leading mainstream astronomers, and, as already pointed out, this would lead them to generating hypotheses in which the planets Inoved routrd a stationary Earth on paths which were a combinatior-r of circular motions' Some progress

72

Inductivism and its Critics

would urrdoubtedly be macle along these lines, but the research programrle would not lcad to a rnajor breakthrough. To obtain the breakthrough, the team wotrld have to be a little more broadrnirrded and consult not just the establishrnent,but some of the outsiders dismissed by the rnajority as technically competent but crankislrin their approach. Looking around among such individuals, tlre teanr wftrld quickly corne across Kepler, who with great enthusiasm would propose a quite different set of heuristics. He would suggestthat, insteadof relating all the variablesto a stationary Earth, they should relate them to a stationary Sun. Instead of considering jtrst circular rnotions, they should try other known crrrves such as the ellipses which had been studied in antiquity by Apollonius of Perga, and they rnight also try generating laws in the form of peculiar conrbinations of algebraic expressions.If the tearn were sensibleenough to adopt the heuristics of this eccentric philosopher-mathernatician, they would be quickly rewarded with lr l success.lt can be seen from this hypothetical example that there is i, ' no necessary antagonism between human creativity and computer i, l :ir | approaches. Orr the contrary, human creativity is a wonderful resource and one which car.r consciously used to produce better be systcrnsof artificial intelligence.

Pa rtll
Co n v e n ti o n a l i s m a n dt h e D u h e m Thes is Oui ne

,:trilti

''ril:'i":: 4

Poincare's Conventionalism of 1902

oo

In this chapter I will give an exposition of Poincar6'sphilosophy o[ Poincar6 was one of the science,which is known as conuetttiotralisn. leading mathematiciansand physicistsof his time, and in his mathenratical work he made considerableuse of non-Euclidean geomctry. This led him to an interest in the nature and foundations of geometry. It is probable that he first dpvised conventionalism to give an account of geometry, and only later extended it to some other parts of science.As an introdr"rctionto Poincar6's ideas. therefore. it will be necessary to say a little about the philosophical influence of Euclidean geometry prior to the discovery of non-Euclidean geometry, and then about the impact of non-Euclidean geometry on the theory of knowledge. This is an important topic in itself, since Putnam has claimed that 'the overthrow of Euclidean geometry is the most important event in the history of science for the epi stemol og ist '( 1975, p. x) . Er,rclid's axiomatic-deductive development of geometry was written around 300nc, and was accepted, almost without challenge, as a true account of the geometry of space for more than 2,000 years. Moreover, Euclidean geometry was used by Newton in developing his mechanics, and so received additional support from the successof Newtonian physics. It is not surprising then that Euclidean geometry came to be seen as a certain and indubitable piece of knowledge. This attitude receives its classic expression irr Kant's philosophy of geometry expounded in his Critique of (1783). We will now briefly Pure Reason (1781/7) and Prolegomena descri beK ant 's posit ion.

of 4.1 Kant'sPhilosophy Geometry

Kant's theory of geometry depends on a pair of distinctions, nanrcl y:
---.;.ialf

76 (i) (ii)

and Conventionalism the Duhem-Ouine Thesis betwccn a priori xnd (t posteriori knowlcdgc, and betwecn analyticand syntheticjudgenrcrrts.

l i 'i ;i i i i i J i r ,;--.

.i

77

The first distinction is realiy a traditional one. Kant explairrs it as fiolfows: 'ln wbat follows, therefore, we shall understand by a priori krrowlcdge, not knowledge independent of tl'ris or that experience, trut knowlcdge absolutely irrdependentof all experience' (1781/7, AZll33, p. 43). A Trostcriori knowledge, by contrast, does depend on expcncnce. 'Ih e s ec ond dis t ir r cti o rri s re a l l y d n e to K a n t h imsel f - though thcre are tracesof it in earlier authors. He says: 'Either the predicate 1) bclongs to the subject A, as sornething which is (covertly) contained in this concept A; or B lies outside the concept A, although it docs indeed stand in connection with it. In the one caseI entitle the ju d g e n re nt analy t ic ,ir r th e o th e r s y n th e ti c ' (1 7 8 1 1 7 , 61810, p. 48). A Kant's notion of an analytic judgement can be illustrated by the favourite modern example (mentioned above): 'All bachelors are un ma rri e d. ' Her e, t lr e s u b j e c tA i s ' b a c h e l o rs ' ,a n d r he predi cateB i s 'unmarried'. IJut bachelorsare by definition unnrarried men, so that the predicate is here (covertly) contained in the subject, and the judgement is therefcrre analytic. Another approach, due to Frege, is to define an analytic judgenlent as one which can be reduced to a truth of logic using only explicit definitions. We have already rrlentioned the view held by Russellafter 1900, and then by the Vienna Circle, that mathematics is rcclucibleto logic. We now see drat this view, known as lo,qicistn, calr be fortnulated as the theory that nratlrematical propositions are analytic. This was ,rot Kant's opinion rcgarding mathematics, which, for him, included Euclidean.geornetry as one of its most inrportant parts. Kant believed that mathematical judgemetlts were syntlretic a priori. I{ant first argues that 'Properly mat}renraticalpropositions are always judgements a priori, and not empirical, becausethey carry with them necessity,which cannot be taken from experience'(1783, pp. 18-19). Kant thought that any generalisationbased on experience such as 'All swatls are white' cannot be necessarily true, since an exception, such as a black swan, might be found in the future. Tlre thcorems of Euclidearr seornetry, however, he regarded as rrecessarily true, arrd therefore as a priori rather than a p<tsterittri. Consider, for exanrple, the theorem that angles in a triangle are cqtral to 180'. This is not, on Kant's view, an empirical generalisation which rnight be contradicted by finding a triangle with

arrgl csacl di ngttp to 179". Tlr e r cst i, t ''*r uf |, gt d, . 'ov: u f - 0. t , '"' , E ucl i d by show i rrg tl rat i t f cr llows logically f r or n axiom s r vhiclt : ir <: t|emselvcs secn te trc oltviously corrcct. So the theoretn is shown to be necessarily true, and since the necessity 'cantrot be taken from experience', 'the judgc'ihent is 'a priori, and not-empirical'. Karrt has next te show that the truths of Euclidean geometry are syrrthetic rather than apalytic, which he tries to do with the following argument: Nor is any principleof purc gconlctryinalytic. That the straightline proposition.For my is!a synthetic bctwecntwo pointsis the shortcst cotrtainsnothing bf quantity but only, a quality' concept of straight Thc ioncept of the shortest is thercforc wholly an addition' and from the conceptof the straight line' cennotbe drawn by any analysis (1783,p. 20) These, then, are the considerations which convinced Kant that Eucfidean geonretry is synthetic a priori. But this doctrine is a puzzling orie, and leads Kant immediately to raise the question of i'to* t,rth synthetic a priori knowledge is possible- This is indeed a problem. tf a judgement is synthetic - that is, not based on a mere analysisof concepts- it would appear to be 'about the world', and horce krrowable only on the basisoFexperience' How, then, can we know about the world a priori? ln his transcendentalphilosophy, Kant proposed sorne ingenious answers to this question, but my aim here is not to expound the Kantian system. I will therefore give next a brief account of how non-Euclidean geometry was discovered in the nineteenth century. We will then see why non-Euclidean geometry led Poincar6 to reject Kant's view of geometry as synthetic a priori and give instead his own conventionalist account of geometry.

Geometry of 4.2 The Discovery Non'Euclidean

Ironically enough, the discovery of norr-Euclidean geometry arosc out of attempts to make Euclidean geometry even more certain and Euclidean geornetry was derived from five axioms, or well for.rnded. postulates.Four of these did indeed seem to be obviously correct; but the fifth postulate, or parallel postulate, was a little less obviotts than the others. Already some ancient Greek gometers had raiscd

- ;t,:l:,_,

I I I I i I I
;

ill

(l rrrv o rrti o n a l i s rnd th e D u h em-Oui neThesi s an P

Conventionalism1902 of Poincar6's

79

I ' t . r 1t4 t r r /

gc i l t.rr;rl kl'ilrrc s rr Ii trc l i c l c a n o metry

geometry Fignre lincsin Bolyai-Lobachevsky 4.2 Parallcl

r lr r , it r ot r s ; rl ro ttt tl l i s p o s tu l a tc , a n d the attel npt w as begun to r nr llov ( , orr l i rrr' l i rlb y tl c l i v i n g th e fi fth postul atefrol n some even t t t olc r ' v ir lctrla x i o l tt. ' l' o s c c w l rrr[h :rs i n v o l v c d h c re , l c t L l sstatethe fi fth postul ate, rot l fo ir r llr l or r l' ,i rr;rl rtrt. b rrt i rr r s i rn p l e r, t hough equi val entform due rr ir r t lr r ' llr ili s l r ttt:rtl tc ttt;rti c i aI)l a y fa i r ( 1795). W e w i l l assumethat g wr ' , ur ' r k ' , rl ttrgw i tl r 1 tl ;rrrc c ()tn c try ,s o that al l the poi nts and l i nes , r or r r t r lr t r ' r ll i c i rr l 1 rl ;rrtcl.)l a y l a i r' s fo r nr of thc paral l el postul ate llr lt t lr c lot n r' sl l rt' l o l l o w i rrg : ( lr v r r r .t l i trc I :rrrrl p o i ttt l ) Iro t l y i ng on l , there exi stsone and :t or r ly u rre l i tre rrr tl rro rtg h P p a ra l l e lto l . ' l' lr r r r s tl l tts tti tc rl i rr l i g trrc 4 . l . (T w o l i tresare paral l eli f they do r nt nlt ' r ' 1, to w c v t' l l :rr tl rc y l rc c x tc ttd edi n ei ther di recti on.) l ll wc t lrtrrk i tl ro trt tl tt' n l l ttc r, tl rc re do appear to be tw o al terr , r lr v c st o t l ri s u x i rtttr,l ri rs t < l f a l l , tl te re nri ght be no l i nes through P ' l ' l ri s ;tl tc rn :tti v ci n fa c t g i ves ri se to a form of nonlr , r r , r lllllo l . geometry. S econdl y, k lr ur lt r lr ' , r r r!,r' (l n r(' tty tt.rw rt a s l l i e n ta r.rni an l o ttc l i n e through P paral l el to l . For t lr r ' r c r r r r li l rt l rc n ro rc l l trrtr lr , r r r r lt k ' , r rr l i l i rrrt' 4 ,.1 ,rrr:u rc lttt' rn i g h t both be paral l elto l . Tl ri s geometry know n , r lt lr l, r t t v c l i rv c l i ti s c to :t l i rn rr o f rro n-E ucl i dcan y , r r llr r ly , r t l .o l r;t,' l rt' v s kt' ,(' ()rr)c try . ' llr r ' lr r r t tn ;rtl tc tttrtti c i :l n s c o trs i d er how ever' thesepossi bi l i ti es, t() wnr ' v r ' r y l ;rr l to rrr tl ri n k i rrg l l ra t th e y cottl d gi ve ri se to geometri es , r lt lrr r ; r t iv cto l l rrr' l i rl ' s .( )rr th c c o n tra ry, they bel i eved that these a r r l' . , r t r ou\ o l tl re 1 r:rr:rl l t' l x i o rn w c re absurd, and coul d qui ckl y H lr l r lr owt r to l r' ;trl l () :r (:()n tra d i c ti c l n. etrce arose the pl arr of p usi r , lr ' r r lltrfln rn l i l :rrrl i rl ' s 1 r:rr:rl l c l()s tu l a te ng the rnethod of recl ucti o r
, n l , t h t t t t l t t t tt

F F F F

F I

The two alternativesto Euclid's parallel postulatc would be considered in turn, and would be shown to be absurd by deriving a contradiction from thenr. This would leave Euclid's parallel postul ateas t he only viable possibilit y. This prograrnme did have an initial success. was shown that the It assumption that there were no lines through P parallel to I did lead to a contradiction. ln retrospect, this result was misleading, becausc it restedon the hidden premise that a straight line can be extended as far as we like in either direction. No one considered the possibility that there might be a finite upper botrnd to the length of any straight line; but this possibility, as we shall see, gives rise to Riemannian geometry. At the time, however, the elimination of the 6rst alternativc to Euclid's parallel postulate seemed conclusive, and only the sccond alternative remained for consideration - namely, that through a point P not on I there rnight be more than one line parallel to l. It proved much harder to derive a contradiction from this assumption. One of the most gallant atternpts to do so was made by the Italian ah Jesuit Gerolamo Saccheri(1667-1733) in his book Euclides Omni Naeuo Vindicatus('Euclid Freed from Every Flaw') (Milan, 1733). Assuming the second alternative to Euclid's parallel postulate, Saccheri derived a succession of strange theorems, concluding eventually that these results were absurd and that Euclid must therefore be correct. But Saccheri's results, though strange, were not contradictory, and were indeed tlreorems of a non-Euclidean geometry. Yet such was the hold of Euclid on people's minds that almost a hundred years elapsed before any mathematician realised that this was the case.

B0

and Conventionalism the Duhem-Ouine Thesis

Poincar6's Conventionalism 1902 of

81

Thc lronour of publishirrg the first systcms of non-Euclidean vsky, and a gconletry is sharccl bctwectr a Russian, Lobache Hungarian, John llolyai. Thcy worked completely independently, and published withirr a few years of each other. Lobachevsky's 'Principlcs of Geometry' was printed in Russian in the Kazan Bulletitt for 1829-30, while John Bolyai's 'The Science Absolute of Space' appearedas the appetrdix to a book by his father, Wolfgang llolyai, on geonletry, published in 1832. The great Gernran tnathelrraticianGauss had obtained similar results at an earlier date, but he he did rrot publish bccause,as l-resaid iu a letter ol 1.829to lSessel, was afraid of the clamorrr of the Boeotians. Boeotia was a region of ancient Greece whose inhabitants were considered by the Athenians to be stupid arrd uncultured. Like Saccheri, both John Bolyai arrd Lobachevsky took the second alternative to Euclid's parallel postulate and derived consequencesfrom it, but their poirrt of view was different. Saccheri had hoped to derive a corrtradiction from this alternative assumptiou, and hcnce to vindicate Euclid. Bolyai and Lobachevsky, however, regarded this assurnption as one of the axioms of a geometry alternative to Euclid's. This is the form of norr-Euclidcan geometry now knowtr as llolyai-Lobachevsky geometry. lIr this geometry, the angles in a triangle are always less tlran t80o (or n ra d ians ) .I ndeed, t h e s u n r o f th e a n g l e si n a tri angl e i s not corl stant (as in Euclidearrgeor.netry),but becomes smaller as the area of the triangle beldnres bigger. To obtain the exact fortnula, we have to at associatewidr each poir-rta constant, K, known as the curuature
,l

that point, and consider

AI} C

JJ

K dS which is the integralof K taken

over the area of the triangle ABC. We then have

JJ A1]C

l l Kd s : A +L t*c-r

fr

(1)

lrr llolyai-Lobachevsky geometry, K is always negative, so tlrat A + B + C < n. In Euclidean geornetry, K = 0, so that A * B + C : r radians : 180o.The third case, in which K is ahvays positive, gives llienrannian geotnetry, which we shall consider in a
I non t e l l t .

'I'he ptrblications of tlolyai and Lobachevsky did not in fact raisc a clantour anrons the lJoeotians.They were simply ignored. llolyai did not putrlislr anything further on the subject, whereas Lobachevsky, tltroughout the rest of lris life, wrote a series of expositions arrd dcveloprtrcntsof his new geometry, includine one

in French, 'Giorn6trie imaginaire', published in Crelle's prestigious journal (1837), and one in German 'Geometrische untersuchungen zur Theorie der Parallellinien', published as a booklet in Berlin (1840). Despite these efforts to overcome the language barrier, Lobachevsky did not succced in arousing any interest among Western European mathematicians. This disappointed him considerably, but he continued undaunted and in the year before his death (1855), wherr already blind, he published in both French and Russian yet another exposition of his systen: - the Pangdom6trie. Non-Euclidean geometry did not become widely known until after the work of Riemann (1826-66), a student of Gauss and one of the most brilliant mathematicians oF his time. Whereas llolyai and Lobachevsky developed their system from a study of Euclid's parallel postulate, Riemann had a completely different approach. His ideas on non-Euclidean geometry originated by abstracting fronr some earlier work of Gauss on curved surfaces. This need not surprise us. As we have seen, work on the parallel postulate suggested, misleadingly, that a geometry in which there were no lines through P, a point not on l, parallel to I was impossible. Only a ncw approaclr showed that tlris was a possibility after all. lliemann expounded his ncw system of non-Euclidean gcometry in a famous lecture entitlcd 'Ubcr die Hypothesen, *cl.he der Gcometrie zu Grunde liegen' (On the Hypotheses wlrich lic at tlrc Foundations of Geometry), dclivcred as qualifying lecture for the title of Priuatdozent the faculty at G<ittingen in 1854. Gauss himto self was preserltat the lecture. In Riemannian geometry, there are no parallel lines, so all straight lines intersect each other. Straight lines are not infinitely long, and, indeed, there is a finite upper lound on the length of straight lines. In Euclidean geometry, there is one and only one straight line between any two distinct points A and B. In Riemannian geometry, there rnay be more than one straight line between two points, so that two straight lines can enclose a space. This last property is impossible to represent visually on a plane sheet of paper. If we draw two 'straight' lines l, m, between the points A and B (see figure 4.3), the lines look curved, not straight. The kind of difficulty seen here should perhaps give the reader some sympathy for those who rejected non-Euclidean geometry. Finally, in Riemann geometry, the angles in a triangle are always greater than 180" (or n radians). The formula is again (1) except that, in Riemannian geometry, K is always positive, so A * B + C ) n. The bigger the triangle, the greater is the difference between the sum of its angles and 180o.

.ixl.

ttz

C o n v e n ti o n a l i sa n d th e D u hem-Oui neThesi s m

of Conventionalisrn 1902 Poincar6's

83

a 4, c lii, r ; ur r | ' l ' w o l i rrc s n c l o s i n g s p a c e

llt c r r ru rrn w u s i rr a rn u c h b e tt er posi ti on than Il ol yai or (() l. olr ; r r ' lrt' v s k y l 1 c t h i s i c l e a sk n o wn to the rnathemati calcomnluuit y , (i c l rrr:rrryw ;rs tl rc l c a d i n g c o untry i n the w orl d at that ti me lr r r t t t ; t t l rc l rr;tti r'rc s c rrc h . l l i e m a n tr w as a pupi l of the most famous :rl ( it r r r r ; r r rn t;ttl r(' n r:l ti c i :rrr, trs s , a n d w as w orki rrg at the l eadi r-rg Oa ( lc r llr n rrrri v i rs i ty fi l r rrra th e rn a ti c a l research, Gotti ngen. N one the t: h' r r , nr o rc tl r;u r:r < l c c ;rrlw a s to p a s safter hi s l ectureof 1854 before rrl tr I lt r ' t r lr ' ; tr ' n o tt-l r,trr' l i tl c lg c o rn e try becarne i del y know n among w rrr . r ltr c r tr.tt i ;ttrs . it ( ) r r t ' r r' ;rs o rr rr tl ri s d t:l ;ryw a s th a t l l i emann' s key l ectureof 1854 li o w, t r lr r t l tl ts l tc rl rrl y i rr l tl (r7 , ()n e y e a r after i ts author' s earl y death. ( inr c r ; r l l y , l ro w c v c r, i n tl rc l fl (rOs ,t he ti dc at l ast began to turn r r r l, r v or rr o l ' rro rr-l l rrr' l i rl c :rrrc ()rn c t ry. In the years 1860 to 1863, g t lr r r r r r t r' ,i l ru trtl c l rtr' l rc tw t:c(]a u s s and a certai n S chumacherw as rr ' l ' l rc l i c rl rrt' rrt l t' rc n c c s the unknow n names of B ol yai rc to 1' r r lr lr r l rc rl i , t t t r l I olr,r, rt' v s k y :rro u s t' rl rrtc rc s t,a nd causedH oi i el to undertake l t lr r ' l' r lr rrl t tr,rrrs l :rti o rrl tl re w o rk s of these men. In 1866 there o lr l , r 1' pr ' , rr tl rr' l i rc rrr' l r t mrrs l l ti o n o f Lobachevsky' s 1840 German lir r r r l' h' t,to g c tl rc r w i tl t c x tr:rc tsfro rtr the Gauss-S chumachercort r ' t lr nn( l (' n (r'w l ri h ,rrr l tl (r7 th c F rc n ch transl ati onof B ol yai ' s paper ; r v . r - .lr r rl ' l n l rc tl , l ;i rr;rl l y ,i rr tl rc y c a r s 18(rB -72, the Ital i an rnathe-

matician Beltrami and the German mathematician Klein published proofs of the consistency of the various forms of non-Euclidean geometry relative to Euclidean geometry. The old hope of Saccheri and odrers that a contradiction would be found in non=Euclidean geometry was shown to be an illusion, and, from a logical point of view, non-Euclidean geometries were proved to be on a par with Euclidean geometry. Many of the relative consistency proofs of Beltrami and Klein are mathematically quite complicated, but there is one particularly simple such proof which I will give in outline. It shows the general method of these relative consistency proofs, and illustrates other points connected with non-Euclidean geometry as well. This proof applies to a particular kind of Riemannian geometry known as doubleelliptic geometry. Let us suppose that we have an axiomatic system for double elliptic Riemannian geometry in the plane. The idea of the relative consistency proof (in this case, as in others) is to produce a model of this system within Euclidean geometry. This model consists of a kind of dictionary according to which the terms of double elliptic Riemannian geometry, such as point, line, and so on, are translated into Euclidean geometry. This translation has to bc of such a kind that all the axioms of Riemannian geometry are true in Euclidean leometry when translatedaccording to the dictionary. In this simple case, the model for the whole plane of the Riemannian geometry is the surfacc of a sphere in three-dinrensional Etrclidean space. A point in the Riemannian plane corresponds to a point on the sphere. A straight line in the Riemannian plane correspondsto a great circle on the sphere. (A great circle is a circle on the sphere whose centre is the centre of the sphere.) The angle between two straight lines intersecting at a point is simply the angle between two intersecting great circles. This is illustrated (in figure a.\ by the diagram of a sphere which can be thought of as the Earth. Two great circles have been drawn through the North and South poles (N and S) which intersect at right angles. These intersect another great circle (the Equator) at A and B, and the angles NAts and NBA are also right angles. The axioms of double elliptic Riemannian geometry are all satisfied in this model. Without giving a full proof, we can make this plausible by considering the apparently paradoxical features of Riemannian geometry listed earlier. First of all, it is true in the model that there are no parallel lines, since any two great circles intersect. Moreover, there is clearly a finite upper botrnd to the

B4

Thesis and Conventionalism the Duhem-Ouine

N

of Poincar6's Conventionalism 1902

85

\ \ \

-tI

4.4 A model for doubleellipticgeometty Figure

length of any great circle. More than one - indeed, an infinite number - of great circles pass through the points N and S, while the great circles NAS and NllS clearly enclose a space. Finally, in the triangle NAB, all the angles are right angles, so that the sum of the arrglesin the triangle is 90' + 90" + 90" : 270", which is clearly greater than 180". geometry becornesthe The curvature K ) O of the Riemanniat.r curvature of the sphere. This curvature is the same at every point orr the sphere, so we are dealing here with a space of coustant curvature. Tliere are also Rienrannian spacesof variable curvature, in which the curvature K varies from poirrt to point. We can use this model to show that if Euclidean geometry is consistent, then so is double clliptic Rienlannian geonrctry. For suppose, as Saccheri would have hoped, that we can derive a contradiction from the axioms of this Riemannian geometry. Take this proof of the contradiction and translate it into the model, using our dictionary. We then obtain a proof of a contradiction using only assumptions which are true in Euclidean geometry, and thus Euclidean geonretry is also shown to be contradictory' This modEl has a certairl natural feel. Consider beings who are confined to the surface of a sphere. We ourselves are for the most part in this condition. For such beings, the shortest practical path two points A and.B on the sphere would be a great circle t"t*..n connecting A and ll. 'Ihe Etrclidean straight line would be a 'tunnel' betweerrA and B, whose constrttctionwe may supposeto be hardly

practical. If these beings defined a straight line as the shortest path on the sphere between two points, they would have Riemannian rather than Euclidean geometry. The model therefore gives Riemannian geometry a homely feel. It is harder, however, to imagine three-dimensional space being Riemannian rather than Euclidean. To do so, we have to suppose that three-dimensional space is curved in the fourth dimension. This is impossible to visualize, but can be described'mathematically by means of equations. This leads us to the conception of the geometry of space which underlies Einstein's general theory of relativity. ln 1915 Einstein introduced his general theory of relativity, in which the geometry of physical space is assumed to be Riemannian rather than Euclidean. The new theory explained the motion of the perihelion of Mercury, which was an anomaly for Newtonian theory. (l will give more details about this in the next chapter.) Four years later, the new theory received further confirmation from the eclipse experiment in which the predictions of general relativity were shown to be much more accurate than those of Newtonian theory. Thus, from the 1920s, general relativity came to be accepted as more accurate than Newtonian theory, thereby showing that the true geometry of physical space was non-Euclidean rather than Euclidean. It is important to note for what follows that the Riemannian geometry used by Einstein was one of variable, rather than constant, curvature. General relativity postulates an interaction between matter and space, such that space becomes curved near very large gravitating masses like the Sun. For short distances in relatively weak gravitational fields, however, Euclidean geometry remains approximately true. This concludes our account of the discovery and eventual triumph of non-Euclidean geometry.

Philosophy Geometry of Conventionalist 4.3 Poincar6's

I I
I I
I
I

We how have enough background to understand the conventionalist account of geometry which Poincari expounded in his 1902 book and Hypothesls. As with any philosophical text, it is imScience ' portant to bear in mind the exact historical conjuncture at which it was written and the corresponding background knowledge assumed by its author. As we have seen, non-Euclidean geometries had been

-t_ -

;l;11:;;1i;

lllt

Oo rrv o rrti o n a l ia n d th e D uhem-Oui neThesi s sm

Poincard's Conventionalism 1902 of

87

F F

lrr

F F F

F I

h
F
F
tt

rc lr r uv r ' ( lr o rrs i s l c rrt l a ti v eto Eu c l i d e a ngeonrctry by the earl y 1870s. A k ' , r r lr rr1trt;rtl rt' n r;rti c i as u c h a s P oi nca16, w ho had w orked 1 n r x lllt r v r ' ly w i th rro rr-l i rrc l i d e a rre o m etri es, w oul d i n 1902 have g t , r k lr r t lr c i r k rg i r:rl p o s s i b i l i ty c o n rp l c tel y for granted. Ora the other lr , r r r r l,I ' l( l ,l i s tl ri rtc c rr y c a rs b e fo re th e appearancen 1915 of the i 11c r r r , r .trl lrc o ry o l ' rc l ;rti v i ty , w i th i ts postul ate that l l i emanni an 11( ' r f r f ly i s tl rc l ru c l l c o rn c try o f s p ace.Indeed, i n 1902 nearl y al l r cf t r r , t llt c r t r;rti ti lrl ry s i c i s ts , n c l u d i n g Poi nca16,w oul d have regarded i tl t t nt t lt t t t lr.l t' ru r1 ' ,t' o tttc tri ca s l o g i c a l ly possi bl e but not pl rysi cal l y s r r ' , r 1, w ;rs s ti l l ;rl rrro s rrrri v e rs a l l y c ceptedthat the true geometry lt t a ol s p; r c w u s l l ,rrc l i tl c a rr. c s h a l l s e ei n a moment l row thi s assumpt W ttorr ;rllrr lerl l)oirrclr6's reflectionson the nature of geometry. rcgartls rron-Euclicleall geometry as refuting the Kantian l'oitrr';ud' l l c w ri tc s : ' w h a t i s th e n a tu r e of geometri calaxi oms?A re ; r os it ior r , tlrev syntlrctit' n priori intuitions, as Kant affirmed? They would then lrc irrrposcrlup()u us with strch a force that we could not conceiveof tlrr r'orrtr;rry pr<llrositiorr, nor could we build upon it a theoretical crlilicc, 'l'hcrr: wotrkl bc rro non-Euclidean geometry' (7902, p. 48). llowc v e r, l )o i n c a r6i s j u s t a s rrru c hopposedto empi ri ci sm as he i s t o l( : r r r li :rrri s rrr.c g ,o c so n to s a y :' O ught w e, then, to concl ude ll lh: r t t lr c : rx i o rrrs f g c o rrrc trya re e x p e ri mentaltruths?B ut w e do not o ()n t t t ; r k cc x ;rc ri l u (' n ts i d c a l l i rre so r i deal ci rcl es;w e can onl y make llr c r r ror r n r;rl c ri a lo l rj c c ts ' (p .4 9 ). ' l' lr c r r ' ;trc v l ri o rrs rc p l i c s w l ri c l r e r npi ri ci stscoul d make to thi s ullunlent ,' l ' l rc y c o rrl c ls a y th a t g e o m etry i s arr empi ri cal theory of lc t t r ; r l s 1 r;rrcb rrt th ;rt, l i k c rn a rry th e o ri esof physi cs- for exampl e, , t lr c ir lc ll ;1 ;rs:i w s - i t rrtu s t b e u n d e rstoodas hol di ng onl y approxi l r r nt c ly . A l tc rn l rti v c l y , th c y c o trl d a c ceptthe physi cal real i ty of the r lr ; r t ill c o trl i rrtrrn r rts tl c s c ri b e d b y g eotnetry, but argue that our r . v ir lr t r c cl o r tl rc c x ;r(' ttru tl r o f g e o m etry consi stsof i nexact obsero v . r t ior r s l i rr c x :rrrrp l c , f n ra te ri a lo bj ects w hi ch are approxi mate o t r , r lir r t t i, rtrs l ' s tru i g l rt l i rrc s ,c i rc l e s ,a nd so forth. S uch observati ons ur r ly lr lr t i u l l y c o n l i rrrr tl rc trtrth o f g e ometry (i nterpretedas hol di ng cx;r(tly) lrrrt tlrr:n rro s<;icrrtific theory is ever confirmed more than p, r r t r . r lly , ll rs irrtt'rcstinl.l, (:()lllp,lrc to Pr>irrcar6's total rejection of Kantianism , r r r r lc r r r p i ri c i s rrr g :rrrl i l rgg e o n re try w i th the posi ti on adopted by rc f lr r s r c f l r rrrl y l i v c y c l rs c a rl i e r i n h i s E ssayon the Foundati ons of ( )t otrtrlryol lt'l()'i. ' lt t r s s r ' l l , l i k c l ' o i n c a r(, th o u g h t th at K ant' s vi ew of E _ucl i dean l1r ( r nr t r y ;rs ., p ri o ri i s u n tc n a b l e i n the l i ght of non-E ucl i dean he 1' , r ' or r r ty tr I h rw c v c r, rrrrl i k cP< l i n c a r d, bel i evedthat somethi ng of '.

Kant's position on geometry could be rescued. More specifically, he used arr argument along Kantian lines to try to prove the a priori validity of a weaker sub-system of Euclidean geometry: namely, the sub-system which deals with the relations of points, lines, and planes, but without introducing the metrical notions of distance and angle. This sub-system is known as projectiue geometry.So Russell's claim is that projective geometry rather than Euclidean geometry is a priori. In chapter 3, section A, of his book, he attempts to give a proof of this, and concludes as follows: With this, our deduction of projective Geometry from the a priori conceptual propertiesof a form of externalityis completed.. . .I wish to point out that projectivc Geometry is wholly a priori; that it dealswith an object whose propertiesare logically deducedfrom its definition, not empirically discoveredfrom data; that its definition, again, is founded on the possibility of experiencing diversity in relation, or multiplicity in unity; and that our whole science,thcreforc, is logically implied in, and dcduciblefrom, the possibilityof (1897,p. 146) srrchexperience. Withirr projective geometry, we can go on to give definitions of the rnetrical notions of distance and angle; but these definitions can be formulated in various ways. One particular set of definitions yiclds ordinary Euclidean geometry, while others give various systems of non-Euclidean geometry. According to Russell, we cannot decide a priori, but only on empirical grounds, which of these various geometries is true of actual space. He writes: 'The Euclidean and non-Euclidean spaces give the various results which are a priori possible; the axioms pecrrliar to Euclid - which are properly not axionrs, but empirical results of measurement - determine, wirhin tlre errors of observation, which of these a priori possibilities is his realisedin our actual space' (p. |afi. Russellhere expresses belicf that the empirical results of measurement support the claim that actual sp aceis Euclidean. So, while Poincar6 rejected both Kantianism and enrpiricisnr, Ilussell devised a view which combined elements of both positions. Russellused an argument alorrg Kantian lines to try to establishthat projective (nor Euclidean) geometry is a priori correct. This left open a range of possibilities, both Euclidean and non-Euclidean, concerning the geometry of actual, physical space;and Russell held that the choice between thesc various possibilities could be made only on empirical grounds by observation and experiment. when he propoundcd Russell'sview was certainly very reasonable

---iri.;1--,

88

Thesis and Conventionalism the Duhem--Ouine

89 ri meanin6.We rnight xs wc,r ask if thc metricsystemis true, ancr iirc old wcights and rncasurcs f alsc;if Cartesianco-ordinatcsarr: tme arc frlsc. Onc gconrctrycantrot morc truc than and polar co-ordinatCs be another;it can only bc morc convenient.(1902,p. 50) Poirrcard'sconventionalist account of geometry has two parts. First there is the claim, which we have just quoted, that Euclidean geonretry is a set of cotrventions rather like those which define the rrretric system. Second, however, Poincar6 holds that Euclidean geometry is the simplest possible convention and that, since it agrees well enough with observation, it will never be given up. Poincar6 believes that the simplicity of Euclidean geometry is arl objective mathematical fact. What he probably has in mind is tlrat in Euclidean geometry the curvature of space is zero (K : 0), which might arguably be considered as objectively simpler than K having any non-zero valuc. At any rate, this is what he says: Now, Euclideengcometry is, and will remain, the most convcnicnt: lst, because is the simplest,and it is not so only becausc our it of mental habits or bccause the kind of direct intuition that wc have of of Euclidcan spacc; is thc sirnplcst itscll just as a poil'-omial of it in tlrc frrst degrce is simpler than a polynonrial of thc seconddegrec; 2nd, becauseit sufficiently agrces with the properties of natural solids,thosebodics which we can compareand measure meansof by (p. our scnses. 50) Poincard buttressesthis later on by appealing to a type of argument introduced by his contemporary Duhem. We shall consider the Duhem thesis in the next chapter, but Poincar6's use of it forms a convenient introduction to what will be said there. Poincard writes: geometry is true, the parallax of a very distant If Lobatschewsky's star will bc finite. If Riemann's is true, it will be negative. These are the results which seem within the reach of experiment, and it is hoped that estronomicalobservationsmay enable us to decide betwcen the two geometries.But what we call a straight line in astronomy is simply the path of a ray of light. If, therefore,we were to discover negative parallaxes,or to prove that all parallaxesare higher than a certain limit, we should have a choice between two conclusions:we could give up Euclidean geometry, or modify the laws of optics, and supposethat light is not rigorously propagated in a straight line. lt is needless add that every one would look to upon this solution as the more advantageous. Euclidean geometry, thereforc,has nothirrg to fear from fresh experiments.(pp. 72-3)

it, but it did not farc wcll in the light of Einstcin's general theory of rclativity. Thc problem was that, as we have already mentioned, the gerreral theory of relativity itrvolves Rienrannialr geonletry of variable curvature; whereas Rttssell's doctrine that projective gcometry was d priori conect rtllcd out as a priori inrpossible itiernannian geoniet.y of variable curvattlre and allowed only curvature. So although Russell's lLienranrtiangeometry of constant geometry, it allows several diffbrent types of norr-Errclidean theory actually rttles out the particular one which was used by Einstein. of lrr lris Essayon the Fonntlations Ceowefry, Russell quite explicitly of criticised Rienrann for allowing spaces non-cottstantcurvature. In clrapter 3, section Il, Russell gives an alleged proof of the a priori truth of the axiom of free mobility - that is, the axiotn which postulates that spatial lrraglritudes can be moved frorn place to place without distortiott. Frotn this axiom it follows that spacesof notrconstant curvature are irnpossible. Russellaccordingly says: to Ilicrnannlus failcdto obscrvc,what I havc endcavoured provc in of that, unlessspacchad a strictly constantmeasure thc ncxt clr:rpter, alsothat thc abscncc impossiblc; Gcometrywoulclbecornc curvaturc, of of constantmcesure curvatureinvolves absolutcposition' which is that all geometrical ,Flence is lcd to the conclusion hc an absurdity. axioms are etllpirical, and may not hold in the infinitesimal whcre (1897,p. 69) is obscrvation impossible. Russell was never dogmatic !n his attitudes, and he welcomed the general thcory of'relativity with enthusiasm, repudiating his earlier vie*t ott gcometry. ln any case,by that time Russellhad eliminated all traces of Kantianism from his philosophy of mathematics, and view of mathematics. was the leading advocateof the logicist Let us, however, retllrtr to Poincar6' Having rejectedetnpiricism a1d Kaptialrisnr, he proceededto expound his own conventionalist philosophy of geometry: a neitlrcrsynthetic priori innitiotts arc axionts thercfore T'he Xeotnetrkal Thcy are convcntions.Our choicc among all nor experinrcntal -farts. is possiblcconvcntiorrs guidetlby experimentalfacts; but it remains and is only lirnitcd by thc neccssityof avoiding every contrafrtc, rigorouslytrue even rnayrenrain diction,and thusit is that postulatcs their adoption laws which have dctermined *hen thc cxperimcntal (l of Itr arc only approxirnatc. othcr words, the axiottrs geometry do not in What, then, are of spcak tlrose arithmctic) onlydafnitiotts disguise. of lt gcotnetrytrtre? hasno Is arewc to think of thc.qtrestion: Etrclidean

- i:::l:li;l;f

Ti,'j-

90

and Conventionalism the Duhem-Ouine Thesis

of Poincard'sConventionalism 1902

91

Poinc:rrd is of course corrcct in saying that to test Euclidean gconrctry cxperinrentally, rve have to assunle more than just tlre axiorns of Eucliclcangcornetry. We need auxiliary assunrptionssuch as that the padr of a light ray is a straight line. It is thus always logically possible,in the face of an apparentexperimental refutation, to give up an auxiliary assunrption rather than Euclidean geometry itself. Flou'ever, Poincard was wrong in thirrking in 1902 that the scicntific corrrnrunity would hold on to Euclidean geometry, come what nray. As a nratter of fact, the cornmunity decided, after the irrtroduction of the general theory of relativity in 1915 and its few years, to lirnit the cxpcrinrental confirmation in the subsequerrt applicability of Euclidean geometry and to regard Riemarrrrian geometry with nor-l-constantcurvature as giving a more accurate accourrtof the geonretry of physical space. T'his, however, is to nrove our historical point of view to a date ffter 1920 and to judge the nlatter with historical hindsight. Let us now retunl to the early years of the twentieth century and corrtinue orrr account of Poincar6's views h 1902 by explaining how he cxterrdcd his conventionalist philosophy from Euclidean geometry to Newtorrian mechanics.

o-f The Principle Inertia- A body under the action of no force catronly move uniformly in a straightlinc. Is this a truth imposedon the mind i priori?If this be so, how is it that the Greeksignored it? How could with the causeof motion? or, they have belicvedthat motion ceases again,that every body, if there is nothing to prevent it, will move in a circle, the noblestof all forrns of motion? (1902,p.9l) Poincar6's statement of the beliefs of the Greeks is perhaps a little misleading. According to Aristotle, only bodies in the heavenly region, and hence composed of the fifth heavenly element (aither), moved naturally in a circle. Poincard's general argument is, however, convincing. Aristotelian mechanics is quite different from Newtonian mechanics, and yet Aristotelian mechanics was believed to be correct for many centuries. It is hard to see how this historical fact is compatible with Newton's laws of motion being a priori truths. Having denied that the principle of inertia is an a priori truth, Poincar6 goes on immediately to deny that it is an experimental fact: Is, then, the principlc of inertia, which is not an d priori truth, an experimcntelfact?Have there ever beenexperimentson bodiesactcd on by no forces?and, if so, how did we know that no forccs were is acting?The usualinstance that of a ball rolling for a very long time on a marble table; but why do we say it is under the action of no it force?Is it because is too remote from all other bodiesto experience any sensibleaction?It is not further from the earth than if it were thrown freely into the air; and we all know that in that caseit would be subjectto the attractionof the earth. (pp. 9l-2) Moreover, Poincar6 argues that if the principle of inertia were an experimental law, it might, irr future, be modified in the light of observation and experiment and replaced by a more accurate law. But Poincar thinks that the revision of the laws of Newtonian rnechanics, like the revision of the axioms of Euclidean geometry, is not a serious possibility. As he says: 'An experimental law is always subject to revision; we may always expect to see it replaced by some other and more exact la;w. But no one seriously thinks that the law of which we speak will ever be abandoned or amended. Why? Precisely because it will never be submitted to a decisive test' (pp. 95-6). Suppose, for example, we observe what seems to be a deviation from the principle of inertia. Such an apparent deviation, Poincard argues, need never force us to abandon the principle of

4.4 P oincar6'sCon v e n t io n a lis ma n d Newtonian Mechanics

l)oincari argtres that the laws of Newtonian mechanics, like the axiorns of F:iclidean geometry, are conventions. His general line of ilrliurrrcrlt is exactly the sarne for Newtonian mechanics as it was firr Euclidean georrretry. He begins by clairning that the laws of nrcclrarrics are ncither a priori nor experimental truths. Although tlrcsc laws were suggested by experierrce,future experiments and <lbscrvatiorrs call llever invalidate them. The laws of Newtonian nrcclrarrics disguiseddefinitions, or conventions, and, moreover, are thc situplcst such conventions. l)oincardconsidersdre laws of Newtonian mechlnics in turn. Let rrs fbllow his argunrent in more detail in the caseof Newton's first llw o[ rrrotion, or the principle of inertia, which we discussedin c lr apt c r3 . ' l)oincari bcgins by arguing that the principle of inertia is not an a priori rruth. l"lc says:

92

Conventionalism the Duhem-Ouine Thesis and

93
physicist, was one <lf thc prirrcipa, i,".;i:icors the t\n/eni,+iii-csi:i of ',ri'.., ;ris ievolution in physics. \lfliat happencd rvas that Poincard chenpr;:": mind regarding rrrcclranicsbctween 1902 and 1904. In tlii introductiorr to his 1905 book The Value oJ Science', Poincard explicitly draws attenti8n 'tdr'fiis change of opinion. He writes: It was altogether natural. . . that celestialmechanicsshould be the . first model of mathematicalphysics; but . . . this science . . is still to developing, even rapidly developing. And it is already necessary modify in certain points the scheme I outliried in 1900 and from and which I drew two chapters 'Science Hypothesis.'Inan address of atithc St Louis exposition in lq)4, I sought to survey the road the readershall see farthcr travcllcd;thc result of this investigation on. (1905,pp. 13-14) ln fact, Poincar6's St Louis addressof 1904 is reprinted as chapters 7-9 of his 1905 book. lt is in this address that we find the change in his opinion regarding Newtonian mechanics. Poincard states quite clearly that this change of opinion was brought about by reflection on soffrerrew experirnental results. As he says: 'I have long thought that these consequences theory, contrary to Newton's principle, of would end sorne day by bcing abarrdoned, and yet thc recent experiments on the movement of the electrons issuing from radium seem rather to confirm them' (1904, p. 102)- A few paragraphs later, he speci fi es that these ar e't he exper im ent sof Kauf m ann'( p. 103) . t Walter Kaufmann conducted ap experimental investigation of the mass of high-speed electrons (or cathode rays, as they were still widely called) emitted by radium salts. His results were published in the years 1902-3. An attempt could have been made to explain the variation of mass with velocity, which he found, as an electrodyrramic phenomenon and so as not applying to Newtonian or mechanical mass. However, Poincari concluded that the same laws of variation of mass with velocity must apply to mechanical as well as electrodynamic mass. This implied the falsity of the law of conservation of mass (or Lavoisier's principle), which in turrl irnplied the falsity of Newton's laws. As Poincar6 himself puts it: So the mechanicalmassesmust vary in accordance with the same laws as the electrodynamic masses;they can not, therefore, bc constant. Need I point out that the fall of Lavoisier'sprinciple involves that of Newton's? The latter signifies that the centre of gravity of an isolatedsystem moves in a streight lirre; but if there is no longer a

inertia, becausewe can always get round the difficulty by postulating that the deviation is due to invisible molecules: Il, thcn, thc acccleration bodics we cannot seedepends somcof on thing clsc than the positionsor velocitics other visiblebodiesor o[ of invisiblc molcculcs, the existenceof which we have been lcd previouslyto adrnit, thcre is nothing to preventus from supposing that this somcthing elscis thc positionor velocityof other molecules of which wc havc not so far suspected cxistence. the Thc law will bc (p. safcguardcd. 96) Poincard then considersthe other laws of Newtonian mechanics, and reachcs the conclusion that: 'The principles of dyriamics appeared to us first as experimental truths, but we have been conrpelled to use them as definitions' (p. 10a), and again that: 'Experiment may serueas a basis_lbr principlesof mechanics, the and yet utill neuerinualidate theni (p. 105). Poincar6 sunrmarizeshis overall view regarding the principles of Newtonian mechanics as follows: lf thcsc postulatcs posscss generalityand a certaintywhich was a absentin thc experimcntal truths from which they wcrc derived, it is because thcy rcduccin final analysis a sin.rple to convention that wc have a right to make, bccausewe arc ccrtain bcforehandthat no cxperirnentwill come to contradictit. This convention,however, is not absolutcly it arbitrary; is not the child of otrr caprice. admit it !7e certainexperiments because have shown us that it will be convenient, and thus is explained how experiment hasbcen able to build up the principles rnechanics, why, nevertheless, cannotoverthrow of and it them. (p. 136; I have herc altered the standardEnglish translation slightlyto give a more accurate rendering the French.) of Poincar6'sconventionalist account of Newtonian mechanics,like his conventionalist account of Euclidean geometry, is no longer tenable in the light of the twentieth-century revolution in physics. The view that Newtonian mechanics holds exactly in all circumstances was, contrary to Poincard's predictions, abandoned because of the development of relativistic mechanics. Indeed, Newtonian mechanics is now thought to hold only approximately for bodies moving at speedsmuch less than the speedof light and in relatively weak gravitational fields. Newtonian mechanics was not in fact saved by the postulation of irrvisible molecules or any similar device. But here we come to a paradox, for Poincar6, in his work as a
I ; : '; ; '; i i 'i l i '

94

C o n v e n ti o n a l i sa n d th e D u h em-Oui neThesi s m

Poincard's Conventionalism 1902 of

95

c ons t ai l trra s sth c rei s n o l o rrg c ra c c ntrcof gravi ty,* . no l o,,g.. r . know cven what this is. This is why I saiclabovethar thc ."p".inrentso' thc cathoclc rays appearcd justify the doubtsof Loientz to concerning Ncwton's principle. Fronr all theseresults,if they wcre confirmed,would arise an cJrtircly ncw nrechanics, which would be, aboveall, characterized by this f:rct,that no velocitycould surpass that of light. (p. 10a) l)oincarr<'s trext piece of researchwas to carry out a mathematical clcveloprnentof this new mechanics. His paper on the subject was s ubr nit t edi n 1 9 0 5 ,a n d p u b l i s h c di n i 9 0 6. ln l9O2 Poincard had argued that the principles of Newto,rian trrcchalrics were definitions, or collventiolls, which would never be irrvalidatedby-Lxperinre't. Yet by 1904 he had decided i' the light oI I(aufrrralrn's experirnentsthat Newtonian nrechanics neededtobe rrrocli{iccl, arrd by 1905 he had developed the nlathemaricsof a new 'lherc could scarcely be a more striking nrecharrics. exarnple of a sciclrtist-urathernaticiarr carrying out a brilliant piece of research wlrich contradicted his own philosophical principles. In effbct, l)'irrcar6 was collservative i' his philosophy of scie'ce, but revoI r r t ior r ar y rr h i s s c i c l rti fi cp ra c ti c e . i 'lhe cxar:t collverse holds in the caseof lloi.car6's conternporarv. l) t r her r r .l)u h e rrr, a s w e s l ra l l s e ei n th e' ext chapter, propoi nded' a assumpti ,ons i mmune i lr lr ilos oph yo f s c i c n c ci n w h i c h rl o th e o reti cal l} , r r t lr c p o s s i b i l i ty .f re v i s i o ' r a ' d rn odi fi cati o' . l ndeed, D uhem c x plic it ly s tl tc s tl l a t N e w to n ' s l a w s o f mechani csmav be chansed i. t lr c lir t r rrc . Y c t l )rrh c rrr, fa r l ro m w el comi .g E i ni tei n' s thei ry r l' r ' c l: r t iv i ty , rl t' rr.rrrrt' c dt a s a n a b e r rati on of l he German .ri nd. i S . l) t r lr t ' rrr w ;rs l )r()g rc s s i v c rr h i s p h i l osophy of sci ence,but rei r r t t ir r r r ly i n l ri s s c i t' rrti fi < :ra c ti c c . L a te r on, I w i l l further examrne p t lr c s t ' t ' r r r.u s (' ()n tr:l (l i c ti ()rsb e tw e e n phi l osophi cal theory and i s lit ' r r t ilic l)r;r(' ti (ctl r:rt a re to b e fo u n d i n the w ri ti ngs of poi ncard r ut ( l I ) lt ll( ' i l 1 .

sort, but Poincar did not lbllow in his footsteps. For Poincard, in fact, most of the remaining laws of scienceare not conventions or disguised definitions, but genuinely, empirical laws founded on induction from observation and experiment. In the following passage, Poincard statesthis limit to his conventionalism: Principlesarc conventionsand dcfinitions in disguise.Tlrey are, laws, and theselrws havc, so howcvcr, deduccdfrom expcrimental into principles which our mind attributes to an to speak, beenerectcd far absolutevaluc. Some philosophershave generalised too much.2 They have thought that thc principles were the whole of scicnce, and therefore tbat thc whole of sciencewas conventional. This paradoxicaldoctrine, which is called Nominalism, cannot stand a examination.How can a law become a principle?It expressed relation betweentwo real terms, A and B; but it was not rigorously true, it was only approximate. We introduce arbitrarily an intermediateterm, C, more or lessimaginary, and C is hy defnition that by wlrich has with A exactlythe relation expressed the law. So our law is decomposed into an absoluteand rigorous principle which the experimental expresses relationof A to C, and an approximate and revisable law which expresses relationof C to B. But it is the clear that however frr this decompositionmay be carried, laws will alwaysremain.(1902,pp. 138-9) Regarding experimental laws, as opposed to principles, Poincard gives a Bayesian inductivist account. Observation and experiment can never make a prediction or a law certain, but experience can, nonetheless, render predictions or laws probable, so that 'Every time that he [a physicist] reasons by induction, he tnore or less consciously requircs the calculus of probabilities' (pp. 183-a). and Hypothesis(the Poincar6 accordingly devotes a chapter of Science ninth) to a considerationof the calculus of probabilities. He also places a limit to conventionalism on what could be considered 'the other side'. Although he regards geometry as conventional, he does not extend this account to the whole of mathematics. On the contrary, for arithmetic and hence for analysis, he advocates a modification of Kant's theory of synthetic a priori i ntui ti on. and Hypothesisappeared in 1905, The English translation of Science and Russellreviewed itin Mind in the same year. Poincard replied to Russell's review the next year, and this was indeed the openinp;of the controversy between Russell and Poincard to which we have already referred in chapter 3.3

4. 5

P o i rrc a 1 6o n th e L i mi ts of C onventi onal i sm

5r r r r . I t r r r rrt.l ti c x rc rrrl t' tll ri s c .rrv e n t i onal i sl n from geornetry tc) t r r t ' , ' lr , ur i tr, c rrri l l l rtl r;rv ec x l rs s l e dh i l n to extend i t further to the w r ( ' nr ; nnur l ' , rr;rrrr' l rr,sl s t:i c l rr.c . n d c ed, a contenporarv French l o l .t' 1' lt r lr r , , , , 1 tl 1lr' 1l,(o y , rl i rl 1 ' ro tl u c ca g lobal conventi 6nal i snr thi s of

" r't;;rTl::

:]il--

96

Cdnv ent ion a l i sa n d th e D u h e m-Ou i n eT hesi s m

. "_11

97

tly 1905 llusscll was a convinced logicist in the philosophy of rrrathernatics,arrd, therefore, as we nright expect, he criticises Poincar6'sKantian account.of arithmetic and analysis.Moreover, in accordancewith thc views on geolrletry which he had expressedin lris 1897 book, Ilttssell rejects Poincar's claim that geotnetry is is wholly convetrtional,and argttesinsteadthat experierlce lreededto dccicle bctween the various geonretries which are a priori possible. As llusscll puts it: ordcr in by that mattcris arranged pcrception a spati.rl All tlrisshorvs and it is orders; diffcrentfronr sotncof thc possible which is ccrtainly that wc selectat all whosc origin is irr pcrceptiolr only tbr rcasolrs to fronr anrong tlrc ordcrs that arc a prittri possiblc.And this srtffices as provc that gcomctry is not ulrolly cotrvetttioual, M. Poincar6 p. (19U5, 415) contcnds. In addition, llusscll conlrnents: 'There is also aIl interesting but rinsatisfhctory discussion of probability, whose importance, in irrductive proofs, M. Poinca16very justly emphasises'(p. al6). To tlris l)oincard rcplied, not rulrcasonably:'Mr ltussell has not the air o[ bcirrg very satisfiedwith what I say about probability. I anr not very satisficd with it cithcr and I would be happy if Mr Russell had sonrethirrgrnore satisfying to propose' (1906a,p. 143). The inrportance of this exchange is that it may have helped to stinnrlate the interest of Russell and his followers in probability and inductiorr which we describedin chapter 1. If this is so, it is not the orrly instance of the controversy betweerr Russell and Poincar6 provirrg intcllectually fruitful as well as entertaining. It will be rcnrclrrbcrcd that ll,ussell discovercd a paradox in logic which vitiatcd Frcge's original vcrsion of logicism. Indeed, several other paradoxescanre to light at about the same time, and, in order to produce a new and satisfactory version of logicism, Russell had to develop some nrethod for solving these paradoxes. In his 1906 polcuric against llussell and logicisrn (reprinted in a shortened vcrsiorr irr his 1908 Science and Method), Poincar6 argues that the paradoxesonly affect Cantorian set theory and mathenraticallogic two tlreoreticalconstructionswhich he is preparedto abandon in toto (cf. his 1908, p. 189). ' Givcn such an attituclc; Poincar6 should perhaps have merely ignorcdthe paradoxes. flut he could not resist the temptation of hrio, 'Logisticis trying to solve therrr. Rcrnarking lvith characteristic anlinomies,'I)oincar6 proceeded to sketch no longerbarrcn, it engetulcrs

his own solutiorr ro {.iisc l-tf i.ao"ale.s0tZfu), 7te. Svlo{,-m iavi vol vcd thc so-cal l ecl cior r scir clc pr it r ciplc,. , . . r ir - ssc, i incoi, i. . . . - ; , this principle into his own solution, presented irr his 1908 papcr 'Matherr.ratical L<-rgic based gn the Theory of Types'. as hr coriel.g$$11;tr1" I would like to stress again that the account jusr given of Poinca16'sconventionalism is based [argely on chapters 3, and Hypothesis,which was published in 1902. 5, and 6 of his Science A ftrller accourlt of Poincar6's philosophical views would havc to take account of both carlier and later texts. It is clear that the advcnt of special relativity in the'years 1904-6 caused Poincar6 to change his opinions in some respects, but different views are possible as to the rnagnitude of the change. Giedymin (1982 and 1991) has argued in favour of a considerable degree of continrrity in Poincard's thinking. Giedymin believes that we can construct from Poincard's writings as a whole a sophisticated conventionalist position which is not nndermined by special and general relativity. He expounds and develops this position in the works just cited, which are strongly to be recommended to the reader who wishes to explore Poincari's philosophy further. I, however, will turrr in the next chapter to an exanrination of some furrher irnportant philosophical ideas of Poincar'scontenrporary, Duhenr.

\'
"i

-t

':i- -'

r
t
I

r|

The DuhemThesisand t he O uineThesis

99

5 TheDu h e m h e s i s n dth e T a OuineThesis

with his predictions, what he learns experiment in disagreenrent is constitutitlgthis group is is that at least one of thc hypothescs doesnot but the experitnent turracccptable ought to be modified; and (p. which one shouldbe changed. 187) designatc In order to discussthe Duhem thesis, it q'ill be useful to introsl.dteffient. Observation statements duce thc notion of an obseruat;ot1 in 6and7 below. For t he w i l l be consider ed m or e det ailin chapt er s nloment. however. let us take atr observation statenlent to be a statement which can provisiorrally be agreed to be either true or falseon the basisof observatiotraud experiment. According to the lJuhern thesis,an isolatedhypothesisin physics (h, say) can never be falsified by an observation statement, O. As a generalisation covering all the hypotheses of physics, this is sornewhat doubtful. Physics does appear to contaitrsome falsifiable hypotheses.Consider, for example, Kepler's first law that planets rnove in ellipseswith the Sun at one focus. Supposethat we observe a large nunrber of positions of a given platretand that these do not lie on an ellipse of the requisite kind. We have then surely falsified I(epler's first law. The schemaof falsificationcan be written, where 'not-h' is short for 'lt is not the casethat h': Iflr, then C), but not-O, thereforetrot-h.

t
T
;

,rtt I t t , t t t t ( r l \\,tl l i tr1 ' ,s tl rc p l ri l o .o p fty o f sci el rce, references 9ftel i t I r t , r r lr 'n r r , l r,rt r;rl l t' rl' tl tc :l )trl rc rn -Q u i ne thesi s' .R eal l y, how ever, r,; t lr t , ,r , , , , , r rrrt' tl ri rr1 1l n ri s tro rrrc r; r, a s w e shal l see, the D uherrr ol fo rr t lr r ' , , r . , , lr l l ,' r,i n r;u ry i rn l ro rta rrt s p e cts re fronr the Qui ne thesi s.l n rx l | t lr r t , lr , r 1' t,' r ' ,r,i l l l ro rrrr<tl rc tw o th e s es n turn and expl ai n how i t lr , ' y , lr ll, ' r I rv i l l r o l rr' l rrrl c c t;h a p te r y suggesti ng th that the phrase b ' t lr , =l) r r lr , ' r rr (J l i rrc tl rc s i s ' < :o rrl l< l u s e d to refer to a thesi sw hi ch re , lr r r l' nr r ' . ,r' l l rrr,' rrl sl l o n r b < l t.l r ttre D u hem tl resi s and the Qui ne t lt . , r , , \ nr r rr rr,,r' i l l Itc n r:rd co Ith i s s u ggesti onn the fi nal chapter rv i nl t lr r ' l' , ' nl . (, l r.r;' 1 t' l1 )), i rr w h i c h Po p per' sfal si fi cati oni sm i l l be { w , ' , , r lun, , l rrr tl rr' l i l i l rt o l tl rc l )rrl rc l n -()ui ne thesi s.

(1)

I :tr
I
EI

lr I l'rolirrrirrary Exposition the Thesis. of llrn lrrrlrossibility a CrucialExperiment of

I ll l)11 11 6 1 11",nr , ur \ / , ; i1' , r r ilit r r nt ont r ibut i o n s t o t h e p h i l o s o p h y o f c .u i, tl | , l" tlt . t l' , , llr l r r r os l nlll) ( ) r t . 1nt was h i s f o r m u l a t i . ; n o f w h a t I rltf f ,,rll tltr l r t t lt t t t t t / r r ' r ir . W it h his us r - r a lc l a r i t y a n d i n c i s i v e n e s s , Ilttlt, rrt ',1s 1r ' ,trlr t ' ; t lr , ' . ; t s : t s: t : ; r ' < ' t ion hea d i n g t h u s : ,lrr 1 t1tr1111,,',,1 l'lt1'sits Oan Neuer Condemn an Isolated ,, I lyy,t1l1111, t 1hr ly t l1[ 4r ole T h e o r e t i c a l G r o u l t ( 1 9 0 4 - 5 , lut l' ll{ t) | .rt' I rr tlrr . , . , , ' ,t r or r lr t ' r ' x por r r r < lst . hc t he s i s a s f o l l o w s : Itt',ttttt, llr , ' 1' l1y ' , 11r !( : lr r r r ( ' v c r s r r t ljc c ta p i s o l a t e d h y p o t h e s i s t o l r rl,ltnrrrr t , r l t r ' , , t , lr r r t or r ly ; r wlr ole gr ou p o f h y p o t h e s e s ; h e n t h e w

I
t

I I

|r
1'
fl I I

tollens. Tlris usesa logical law called modus Horvever, the Duhem thesisddes apply to some hypotheses,and this createsa difficulty for Popper's falsificationism,which we will examine in chapter 10. Consider, for exatnple,Newton's first law of rnotion (T1, say). The arguments of Poincar6 given in the last chapter and some further arguments to be given in the next chapter indicate that T1 is not falsifiable. We canllot find an O such that schema(1) above holds when we substituteT1 for h. Newton's full theory (T, say) consistedof three laws o[ motion (Tr, Tz and T3) and the law of gravity, Ta. So T was a conjunction o[ thesefour laws (T : Tr & T2 & Tj & T4). Even from T' by itself, however, we cannot derive any observableconsequences regarding the solar system. To do so, we need to add to T a number of auxiliary hypotheses:for exarnple, that no other forces but gravitational ones act on the planets, that the interplanetaryattractions are small compared with those between the Sun and the planets,that the mass of the Sun is very much greater than that of the planets, and so on. Let us call the conjunction of srrch auxiliary hypotheses which are appropriatein a given caseA. We now have the schema:

100

Conventionalism the Duhem-Ouine Thesis and If Tr & T2 & T3 & T4 & A, then O, but nor-O, therefore not-(T1 & T2 & T3 & T4 & A). (2)

The Duhem Thesisand the OuineThesis

10'l

Moreover, from not*(Tt &Tz dc T3 & T4 & A) it follows that at least one of the set (Tt,Tz, Tl, T+, or A) is false, but we canrrotsay . which one. As the history of scienceshows, it is often e very real problern in . scientific research to decide which one of a group of hypotheses should be changed. Consider, for example, Adanrs and Leverrier's discovery of Neptune in 1846. From Newcon's theory T together with auxiliary hypotheses, astronomers were able to calculate the theoretical orbit of uranus (the most distant planet then known). i This theoretical orbit did not agree with the observed orbit. This meant that either T or one of the auxiliary hypotheses was false. Adams and Leverrier conjectured that .the auxiliary hypothesis concerning the number of planets was in error. They postulated a new planet Neptune beyorrd (Jranus, and calculated the mass and position it would have to have to cause the observed perturbations in (Jranus's orbit. Neptune was duly observed on 23 September 1846 only 52' away frorn the predicted position.t This part of the story is quite well known, but there were some subsequent events which are also relevant to the Duhem thesis. Another difficulty which occupied astronomers at the time concerned the anomalous motion of the perihelion of Mercury, which was found to advance slightly faster than it should do according to siandard theory. Leverrier tried the same approach that had proved successfulin the caseof the Uranus anomaly. He postulated a planet Vulcan nearer to the Sun than Mercury, with a mass, orbit, and so forth which would explain the advance in Mercury's perihelion. However, no such planet could be found. The discrepancy here is very small. Newcomb in 1898 gave its value as 41.24' + 2.09'per century; that is, less than an eightieth part of a degree per century. However, this tiny anomaly was , explained with great success the general theory of relativity (T'), by which Einstein proposed in 1915 as a replacement for Newton's theory, T. The value of the anomalous advance of the perihelion of Mercury which followed from the general theory of relativity was 42.89" pdr century :. a figure well within the bounds set by Newcomb. 'We see that, although the LJranus anomaly and the 'Mercury anomaly were primafacie very similar, successwas obtained in one case by altering an auxiliary hypothesis, in the other by

In the next section, Duhem goes on to draw an irnportant consequencefrom his thesis. This section is in fact headed'A "Crucial Experiment" Is Impossiblein Physies' (1904-5, p. 188). Duhem uses the term cntcial experimentin somethirig like the sense given by Bacon in the Nouam Organum to his'fact of the cross'.He forrnulates this notion of crucial experiment as follows: 'Enumerate all the hypotheses that can be made to account for this group of phenomena; then, by experimental contradictiorr eliminate all except one; the latter will no lotrger be a hypothesis, but will become a certainty' (ibid.). However, there is -an obviotts objection to crucial experiments in this strong sense: namely, that we can never be sure that we have listed all the hypotheses capable of explaining a group of phenomena. Duhem makes this point as follows: contradictiondoes not have the power to transform a .Experimental physicalhypothesisinto an indisputabletruth; in order to confer this completelythe to power on it, it would be necessary enumerate various hypotheses which may cover a determinate group of all phenomena; but the physicistis never sure he has exhausted the (p. imaginableassumptiorls. 190) In view of this difficulty, it seems desirable to adopt a rather weaker sense of crucial experiment, which may be defined as follows. Suppose we have two competing theories T1 and Tz. An experiment (E, say) is crucial between T1 and T2, if Tl predicts that E will give the result O arrd T2 predicts that E will give the result not-O. If we perform E, and O occurs, then T2 is eliminated. If we perforrn E, and not-O occurs, then T1 is eliminated. In any evellt, one of the two theories will be eliminated by E, which is thus crucial for deciding between them. lt does not of course follow that there may be some, true, because theory is necessarily the successful of, theory, T3, which differs from T1 and T2 but as yet unthought explains the whole matter much more satisfactorily. Duhem's point is that if T1 and T2 are such that his thesisapplies to thenr, then we cannot derive O from T1 but only from T1 and A, where A is a conjunction of auxiliary assumptions. So, if not-O is the result of the experiment, this does not demonstrate beyond doubt that T1 should be eliminated in favour of T2. It could be that one of the auxiliary hypothesesin A is at fault. Duhem illustrates this by what is perhaps the most famous example of an alleged crucial experiment in the history of science: Foucault's experiment, which was designed to decide b.etween the wave theory and the particle,theory of light. The wave theory of

the maintheory. 1plterinS

r rlr i i;,;t,ir,.;

,:t:i:',:::.:

rf

lO)

Corrvonlionalism the Duhem_OuineThesis and

The DuhernThesisand the OuineThesis

103

F
I :

r F
:

lip. lr llr r r r l i rt.rl tl r;rrth c v c l o c i ty o f l i g ht i n w ater shoul d be l essthan ilr v . l, r it y i rr a i r' , w l rt:rc a sth e p a rti cl e theory predi cted that the vehrrtty of lilllrt irr w;rtcr should be greater than its velocity in air. f r ' r r r ; r r r lrr l .v i s c rl ;r rrrc th o d fo r m e a s uri ng the vel oci ty of l i ght i n w, r lr , r , ar rrl l i rrrrrrl th :rt i t w a s a c tu al l y l ess than the vel oci ty of lig, lr t ir r ;ri r l l e l e , th c n , w e s e e m to have a cruci al experi ment wlr t , lr , lt ' ri rl r.srl c l i rri tc l y i n fa v o u r of the w ave rheory of l i ght I t r r lr r ' r l, s o tn r. o l Ii o trc a u l t' sc o n te rn p orari es, notabl y A rago, di d t r t ahr t , r f utl r;rt l ro rr< ;rtrl t'e x p e ri me n t was a cruci al experi ment i n r s jrr rl I lrit r r t.r . rre l) r r lr c r u | ,i rrtc tl h o w e v c r, th a t to deri ve from the parti cl e . ' rrt, llr e' r y t lr ,rt tl rr v c k r.i ty o f l i g h t i n w a ter i s greaterthan i ts vel oci ty lll wr rrr.rrl , rro t .j rrs t tl re a s s u r npti on that l i ght consi sts of ' t ll, l lr dr t lrlr r ( tl rr' l i rrrtl l rrrc rrtah y p o th e s i s of the parti cl e theory), but t lur y it r x i l i ;rry a s s u rrrl rti o l rs s w e l l . The parti cl e theory coul d a c lw, r y , tlr t .,i ;rv ,.rlry ;rl tc ri rrg o m e o f th eseauxi l i ary assumpti ons- s l s A llulr er r r lr rrtr i t: ' | o r i t i s rro t l re tw e e n[w o hypotheses, thi emi ssi on arrrIrv,r_ve Iry;r'tIrcscs,that Fotrcault's experimint judges trenchantly; ll r lr r hler rrtl rrr b e rw (' c rrtw o s e ts o f theori eseach of w hi ch has to lr e lt ler r rrr i r w l ro l e , i .c . b c tw e e rl tw o enti re systems, N ew ton' s nlr liqx s 11 1l1rry p ,e rrs 'o p ti c s ' (p . l 8 9 ). S o, accordi ng to D uhem, l1 l' nr t rir r r ll' srx l ' e ri rrrc rrt s rro t a c ru c i a l e xperi nrenti n a stri ctl y l ogi cal i reilifl Yer, ux wc slrnll sec in the next section, there is arrother, welkr.r sr.rrrr. wlriclr thc cxperiment is crucial, even for Duhern. irr
I

experiment. According to Duhem, the study of the history of sciencernakes any such claim highly dubious:
The history of scienceshould show thit it would be very imprudent for us to say concerning a hypothesis commonly acceptedtoday: 'We are certain that we sl-rallnever be led to abatrdon it becauseof a new experiment, no rlatter how precise it is.'Yet M. Poincar6 does not hesitate to make this assertion concerning the principles of mechanics. (p.212; I have here sliglrtly altered the standard English translationin the interests of clarity.) Poincar6's mistake. according to Duhetn, was to take each principle of mechanics singly and in isolation. It is indeed true that when a principle of mechanics - for example, Newton's first law of rnotion - is taken in this fashion, it cannot be either coufirnted or by adding other hypotheses to refuted by experience. However, such principle, we get a group of hypotheses which can be any conrpared with experience. Moreover, if the group in question is contradicted by the results of experiment atld observation, it is possible to change any of the hypotheses of the group. We cannot say with Poincard that certain fundamental hypotheses, because they are appropriately simple convelrtions, are above question and can never be altered. This is how Duhem puts the matter: It '"vould be absurd to wish to subject certain principles of rnechanics to direcl expcrirnental test; . . . Does it follow that these hypotheses placed beyond the reach of direct experimental refutatiorr have rrothing trtore to fcar from experiment? T'hat they are guaranteed to renrain immutable no.matter what discoveriesobservation has in store for us? To pretend so would be a serious error. Taken in isolation these different hypotheses have no experimental meanirrg; there can be no question of either confirming or contradicting thern by experiment. But these hypothesesenter as essential foundations into the construction of certain theories of rational mecharrics.. . these theories. . . are scltetnatismsintended essentiatly to be compared with facts. Now this comparison might sorne day very well show us that one of our representations is ill-adjusted to the realities it should picture, that the corrections which come and complicate our schetnatism do not produce sufficient cotrcordance between this schematism and the facts, that the theory accepted for a long time without dispute should ' be rejected, and that an entirely different theory should be constructed on entirely different or new hypotheses. On that day some

I F,
F
Il
h

tl

li,2 l)uhorn's Criticisms Conventionalism. of (le llln lhoory of GoodSense bon sensl

llr r lr ' r r r iq rrrrrrri rrrc sr' h s s i fi e d a s a c o nventi onal i stas regards hi s |, lr ihr r o|lt y o l n l i e l c t' , b rrt h e i s c e rta i n l ynot a conventi onal i stn the i neluc ,l I r lloy lrrrl l)oirr<:rri.lndeed, he devotes two sectionsof his ,littt rnrl ,\lntttutr ol l'hysiral 'l'heoryto criticising these thinkers very r lr ; r r ly .ir r r rlc rp rl i c i tl y . l l c l i l rn ru l a te sth ei r conventi onal i stposi ti on l' ll. r v : ' (l e rt:ri tr fi rrrd ;u n e n ta lh y p otheses of physi cal -theory 'rt r ir ilr r ol lr r ' ri l rrl r;rrl i ttc rlb y a rry e x p e ri ment, because they consti tute irr rr',rlif .irliltiliotts,;urrl y lrr:r:ause certain expressions the physicist's in r r n; r p, r . k r .tl rc i r rrrt.l rri rrlrl rrl y th ro u g h them' (p. 209). f;r l ) r r lr r r r ,l rj c r rs s rro n g l y ro Po i n c a rd' s ai m that the pri nci pl esof .l cl Nr . wt , r r r : r rr rrrc < h ;rrri <ws l l 'i b e g i ven up, becausethey are e er r lr . s ir r r I l. s t r.rrv t' rrri ' rrs a v 'a iva b l ea n d cannot be contradi ctedby l

:..":;i

104

Conventionalism the Duhem-Ouine Thesis and

The Duhem Thesisand the OuineThesis

105

one of our hypotheses, which taken in isolationdefied direct experimental refutation, will crumble with the system it supportedunrier the weight of the contradictionsi'flicted by reality on the corrsequences this systemtakenas a whole. (pp.215-16) of ' Thus Duhem's position seems to me more accurately described as nodifed falsifcation, rather than conuentionalism.Duhein claims that some hypotheses of physics, when taken in isolation, can defy direct experimental refutation. He is thus not a strict falsificationist. oll the other hand, he denies that such a hypothesis is immune from revision the light of experimenral evidence. A hypothesis of _in this kind may be tested indirectly if it fo.rrs part of a systern of hypotheses which can be compared with experiment and observation. Further, such a hypothesis may o' so're occasior 'crtrmble with the system it supported under the weight of cortradictions inflicted reality'. Duhem does not deny that ,among the theo-by retical elements. . . there is always a certain number which the physicists of a certain epoch. agree in accepting without test dnd wlrieh they regard as beyond dispute' (p. Zll). Howe.,rer, he is very concerned to warn scientists against adopting too dog;matic an attitude towards any of their assumptions. His point is that, in the face of recalcitrant experience, the best way forward may be to alter one of the most entrenchedassumptions.As he says: Indeed, we must really guard ourselvesagainst believing forever warrantedthose hypotheses which have becomeuniversallyadopted conventions,and whose certainty seemsto break through experi. mental contradictionby throwing the latter back on more doubtful assumptions. The history of physics shows us that very often the human mind has beenled to overthrow such principlescompletely, though they have been regardedby common consentfor cinturies as inviolable axioms, and to rebuild its physical theories on new hypotheses(p.212)

Of all the laws of physics, the one best verified by its innumerable consequencesis surely the law of urriversal gravity; the most precise observations on the movements of the stars have not been able up to now to show it to be faulty. Is it, for all that, a defirritive law? lt is not, but a provisional law which has to be niodified and completed unceasingly to make it accord with experience. (p. 267)

The

episode

of

the

anomalous

motion

of

the

perihelion

of

. , : .

Duhem gives as a' exarnple the principle that light travels i' a straight line. This was accepted as correct for hu'dreds - i'deed. tho_usands of years, but was eventually modified to explain certain di(fraction effects. Duhem even cites Newton's law of gravity as a law which is only provisional and may be. changed in future. unfortunately this passagehas been accidentally omitted from the English edition of the Aim and struclure of Physical Theory.It is here translared from the rFrench edition:

Mercury fits Duhem's analysis perfectly. It would surely have seemed reasonable to explain such a small discrepancy between Newton's theory and observation by altering some auxiliary assumption. In fact, however, the anomaly was only explained satisfactorily when Newton's whole theory of gravity was replaced by Einstein's general theory of relativity. Indeed, from a logical point of view, Duhem's philosophy of science can be seen as offering support to the Einsteinian revolution in physics. It therefore comes as a surprise to discover that Dtrhem rejected Einstein's theory of relativity in the nrost violent terms. In his 1915 booklet La Science ('German Science'),Duhem argues that Einstein's theory allbmande of relativity must be consideredas an aberration due to the lack of sourrd judgement of the German mind and its disrespect for reality. Admittedly, this booklet was'written at a time when bitter nationalistic feelings were being generated by the First World War. Indeed, it belongs to a genre known as 'war literature', and is actually a relatively mild example of this unfortunate species of writing. All the same, it is clear that Duhem did reject Einstein's theory of relativity in no uncertain terms. So, as already observed, we find in both Duhem and Poincar6 a contradiction between their philosophical views and their scierrtific practice. Duhem was led by philosophical considerations to the conclusion that Newtonian mechanics is provisional and may be altered in future; yet he repudiated the new Einsteinian mechanics.2 Conversely, Poincar6 suggested in his philosophical writings of 1902 that the principles of Newtonian mechanics were conventions so simple that they would never be given up; yet, only two years later, in 1904, he decided that Newtonian mechanics needed to be changed, and started work on the development of a new mechanics. Some light is thrown on these strange contradictions by one further elemerrtin the Duhem thesis which we have still to discuss.This is Duhem's theory of good sense(le bon sens). Let us take the typical situation envisioned by the Duhem thesis. From a group of hypotheses, {hr . . . h,,}, say, a scientist has

106

Conventionalism the Duhem-Ouine Thesis and

The Duhem Thesisand the Ouine Thesis

107

t t t

dcrltrccd O. Expcriment or observation then shows that O is false. It firllows that at least one of {h1 . . . h,} is false. But which one or ()n('satc f:rlsc?Which hypothesis or hypothesesshould the scientist try to chlrrge irr order to re-establishthe agreementbetween theory ;rrrtlcxp<:ri<:rrce? Duhern statesquite categoricallythat logic by itself (:lrr()t lrclp thc scientist. As far as pure logic is concerned, the r'lroict'llctwccrrthe various hypothesesis entirely open. The scientist his itr te:rr'lrirr11 clccisionmust be guided by what Duhern calls'g;ood serrsr"(/r bortscns): certairr opirriotrs Itrrrr, lol',icis rrot thc only rule for our judgernents; of the wlrir'lrrkr rrot lirll urrder hammerof the principle contradiction prcrfcctly unreasonableThesemotiveswhich do not lre irr;rrryclrsc w y these' reasonshi cl r I t t r r r c r r fi o rrr l o g i c ;rrrd c t d i re c to u r choi ces, l trot to rr,uorrrlotl.r krrow'and which speak the ample'mindof finesse' mind,' constitutewhat is appropriately lrul rrut to llrc 'g<:ornetric r nff lr f go o tls c rrs e 9 0 4 -5 ,p . 2 1 7 ) . (1 , s l) r r lr c r r t i rrr:rg i rrc ' tw o s c i e n ti s ts who, w hen faced r,vi th the exlrrritrrcrrlalcorrtrrdiction of a group of hypotheses,adopt differerrt rtrrrlegitlr, Sricrrtist A alters a fundamental theory in the group, wlr nr ns s r i t' rrti s tl l ;rl tc rss o n re o f th e auxi l i ary assumpti ons.B oth ilr ir t egir r rtc k rg i c n l l y p o s s i b l e ,a n d o n l y good sensecan etrabl eus Thus, i n the di spute betw eetr lu r lr r it lc l re l w c c trtl tc tw o s c i e n ti s ts . t lr r lr r r r t irl e th c o ry o f' l i g h t a n d th e w ave theory of l i ght, B i ot, lr y , r r r f n ti n u u l ;rl tt' ra ti o na n d a d d i ti on of auxi l i ary assumpti ons, d lr t r lr iot t r l y l rrrl i rrl l c rri o trs l y e fe n d e d the parti cl e theory, w hereas lr r *r r r r . lr o rrs l rrrrl l ytl c v i s c t{ n e w e x p eri ments favouri rrg the w ave llr t ' or y , lr r l l rc e rrtl , l ro w c v c r, th e d i s p u te w as resol ved. fasterin Allrr |orrr';rrrlt's cxlrcrirrcnthad shown that light travelled hypothesis; rlrr tlr,urirr wutcr',lliot g:rvcup supportingthe emission hi r t r ir t ll' , p rrrr' l o 1 ' ,i co trl tln o t h a v ec o m pel l ed rn to gi ve i t up, for w that A rago rx w lior r r lr r l l ' s g rt' ri rrrc rrla s ,rd t th e c ruci alexperi ment wave opticsfor a longer tirne lrr tlrrrrry,lrt srrwirr it, but by resisting (p. lrtvc lrctlnlackingin good sense. 218) llrot rvrrrtltl 'l'ltis cllbct qtralifiessome of Duhem's earlier retnarks 1r:rss;r1',t'in , t lr or r tt r r r c i :rl c x p c ri l n c n ts . L e t u s ta ke tw o theori es,' I' 1 al td T2, that i s, w hi ch cannot be to wlr ir lr ; r r cl ro tl t s rrb j c t:t th e l )u h e n t th esi s; lr . r t r r l ir r i s o l ;rti o rrl ru t o n l y b y a d j o i ni ng further assunrpti ons.l n sc ; r r t r ir t ly k r1 ' ,i r' ;rl trs t' ,tl rc rc c a n n o t b e a cruci al experi l nent w hi ch of r h. t i, k . , ; r ' t' ,v c c rr' l ' s:rrrd ' [' 2 .T h e g o o d setrse the sci enti fi ccomlr

I|

|t
;

I I I F tt F F
LI

tl

rnunity can, however, lead it to judge that a particular'experirnent, such as Foucault's experiment, is in practice crucial in deciding the scientific controversy in favour of one of the two contending theories. In his 1991 book (particularly chapters 4-6), Martin argues that 'lifelong meditation on certain texts of Pascalshaped many of the most important and difficult featuresof Duhem's thought' (p. 101)was derived In particular, f)uhetn's theory of good sense(le bon sens) irr part from Pascal. Indeed, in the passageintroducing le bon sens, Duhem quotes part of Pasc4l'sfamous s-ayingthat the heart has its reasollswhich .Lrsott knows nothing of.3 Although Duhem was undoubtedly influenced by Pascal, it is possible to suggest factors of a more personal and psychological nature which may have led hirn to his theory of scientific good serrse. his writings on philosophy of scienceshow, Duhem was a As man of outstanding logical ability; yet, as a physicist, he was a failure. In almost every scientific controversy in which he was involved, he chose the wrong side, rejecting those theories such as atornism, Maxwell's electrodynamics, and Einstein's theory of relativity which were to prove successful and lead to scientific progress. Although Duhem stubbornly defended his erroneous icientific opinions, he must have known in his heart of hearts that he was not proving to be a successfulscientist. Yet he must also of nis own exceptiorral logical powers. This have been t*tt. situation could only be explained by supposing that something in addition to pure logic was needed in order to become a successful scientist. Here, then, we have a possible psychological origin of Duhem's theory of scientific good sense:namely, that Duhem saw that good sertse is necessary for a scientist precisely because he hirnself was lacking in good sense. Duhem's rejection of a new theory which agreed so well with his own philosophy of science (drat is, Einstein's theory of relativity) is just another instanceof that lack of good sense which unfortunately characterized Duhem's scientific career. Poincar6, by contrast, was one of the great physicists of his generation, and was arnply endowed with the scientific good sense which Duhem lacked. The contrast between the two men is particularly evident in their respectivediscussionsof electrodynamics. As we have already remarked, Duhem attacked Maxwell's theory harshly, and advocated the ideas of Helmholtz. Poincar6 devotes a chapter (the thirteenth) of his 1902 book to electrodynamics. He begins (pp. 225-38) by discussing the theories of Ampdre and

-i

rirri;ti;,i:rii.r

;:,..r .:::::;-,-

108

Conventionalism the Duhem-Ouine anci Thesis

Thesis Thesis the Ouine and TheDuhem

109

Helmholtz and by mentioning the difliculties which he finds irr rhese theories. Then, on p. 239, he introduces Maxwell's theory with the words: 'Such were the difficulties raised by the current theories, i when Maxwell with a stroke of the pen caused thent to vanish.' , Subsequent developments completely endorsed Poincard's support for Maxwell, while Helmholtz's ideas on electrodynamics, so strenuously advocated by Duhem, are now remembered only by a few erudite historians of science. It was Poincar6's scientific good sense which led him, contrary to the principles of his own converrtionalist philosophy of science of 1902, to a modification of Newtonian mechanics. Duhem's theory of good sense seems to nle correct, but, at the same time, more in the nature of a.problem, or a starting-point for further analysis, than of a final solution to the difficulty with which it deals. What factors contribute to forming scientifiC good sense? 'Why are some highly intelligent individuals like Duhem lackirrg in good sense?These are importarit questions, some of which will be raised again later in the book. In the next section, however, I will turn to a consideration of the Quine thesis.

defining an analytic staternent. Frege defines an analytic statement as one which is reducible to, a truth of logic by means of explicit definitions. These two ways of defining an analytic statement are both illustrated by the standard example of an analytic statement, namely 'All bachelors are unmarried'. But Quine defines analytic statement in yet a third way. He writes critically of 'a belief in sorne furrdanrental cleavage between truths which are analytic, or grourrded in meanings independently of matters of fact, and truths w hi cl r are syntheti c, gr ounded in f act '( 1951, p. 20) . I n ef f ect , or is here taking a sentgnceto be analytic if it is true in virtue of Quine the meanings of the words it contains. This is the definition of 'analytic' which is adopted by most modern philosophers interested in the question. Once again it is admirably illustrated by the standard example: S : 'All bachelors are unmarried'. Someone who knows the rrreanings of 'all', 'bachelors', 'are', and 'unmarried' will at once recognise that S is true, without having to make any empirical investigationsinto matters of fact. Thus S is analytic. All this seems very convincing; yet Quine denies that the distinction between analvtic and svnthetic is a valid one. He writes: It is obvious that truth in general dependson both languageand 'Brutus killed Caesar'would be fact. The statement extralinguistic falseif the world had beendifferentin certairr ways, but it would also be false if the word 'killed' happenedrather to have the senseof 'begat'. Thus one is tempted to suppose generalthat the truth of a in is into a linguisticcomponentand a statement somehowanalyzable factualcomponent. Given this supposition,it next seemsreasonable that in some statemerlts the factual component should be null; and But, for all its a priori reasonabletheseare the analytic statements. ness,a boundarybetweenanalyticand syntheticstatements simply hasnot beendrawn. That thereis such a distinctionto be drawn at all article faith. is arrunempirical dogmaof empiricists, metaphysical a of (1951, pp. 36-7) The empiricists to whom Quine refers are, of course, the empiricists of the Vienna Circle, especially Carnap. As we have seen, their particular brand of empiricism (logical empiricism) did indeed involve drawing a distinction between analytic and synthetic statements. However, support for the distinction is not confined to some membets of the empiricist camp. Kantians too support the distinction, which was indeed introduced by Kant himself. But what has all this to do with the issuesinvolving Duhem and We can begin to conventionalism, which we have been'discussing?

5.3 The Ouine Thesis In his famous 1951 article, 'Two Dogmas of Empiricism', Quine puts forward,"with a reference Duhem, a thesiswhich is related to to Duhem's. Nonetheless;it seems to me that Quine's thesis is sufficiently di{ferent from Duhem's to make the conflation of the two intellectually unsatisfactory.a will next briefly describethe I thesis,5 and explain how it differs from the Duhem thesis. Quine The first obvious differencebetween Quine and Duhem is thit Quine develops his views in the context of a discussionabout whether a distinction can be drawn betweenanalytic and synthetic statements, whereas Duhem does not even mention (let alone discuss) analytic/synthetic the problem. We have alreadymet two ways of defining an analyticstatement. The first was due to Kant, who actually introduced the analytic/ syntheticdistinction.According to Kant, a statement analyticif its is predicateis containedin its subject.This formulation presupposes an Aristotelian analysisof statementsinto subject and predicate. It is not surprising that Frege, who rejectedAristotelian logic and introduced modern logic, should have proposed a new way of

r:i:ll::i::

t
t
i I

I l( l

(l rrrv o n ti o n a l i s a n d th e D u h em-Oui neThesi s m

The DuhemThesisand t he O uineThesis

111

F F

F
F
h

ts

h F H

lr r r ihl ; r lr r i rl g e b y o b s c rv i n g th a t th e meani ngsgi ven to sounds and irrrt riptiorrs ure tlctcnnirred purely by social convention. Indeed, '.he rrorill rorrvt.trliorrsdiffer from one language to another. So if a r r ir lr ' r r ( eis trrr< : rr v i rtu e o f th e me a n i ngs of the w ords i t contai ns i (tfr.rt is, is rrnlytic), itis afortiori true by convention. T'hus if a law ir urrulytir', it is truc by convention. The converse may not hold, rirrrc it is <'ont:civable that a law might be rendered true by a set ol r ' onv c r r ti o rrs h i c h i n c l u d e n o t j u s t l i ngui sti c conventi ons conw r c r r r inll t lr e rrrc a rri rrgo f w o rd s b u t a l so perhapsconventi onscons r r c tt er l wit h n rc :a s u ri rrgro c e d u re s . p I )rrhcrlrrrscdhis dresisagainst the claim that a particular scientific l;rw wlrs i.rrrcby converrtion.It is now obvious that exacdy the same irrllilnlcilt <'ould bc used against the claim that the law is analytic. lrrtlt'crl, (]rrinc rlocs argue against the analytic/synthetic distinction lulor t g. it t s t rc s el i n e s .6 tl lhrt to (:urry his argument through, Quine makes a claim (the ' ()rrirre lhcsis) which is much stronger than the Duhern thesis. The lkcy tlilli'rcrrce bctween the .two theses is clearly expressed by 'Duhent's thesis("D-thesis") has a limited and lVrrilk'rnin as [<>llows: spct:irl s(:opc rrot covering the field of physiology, for Claude llcrrrard's cxpcrirnents are explicitly acknowledged as crucial. (]rrirrc's tlrcsis ("Q-dresis") ernbraces the whole body of our k t r owlc dge ' (1 9 7 9 , p . 5 9 9 ). l)rrhcrrr docs intlecd place explicit lirnitations on the scope of his tlrcsis. f lc writcs: "l-he Experimental Testing of a Theory Does Not Iltvt tlrc Somc Logical Simplicity in Physicsas in Physiology' (1904-5, p. ltl{)). Ilc tlrink.sthat his thesisdoes not apply in physiology or in t:crtairr branches <lf chernistry, and defends it orrly for the hypotlrcscsof'physics. My own view is that Duhem is correct to lirnit tlrc scopc of-lris thcsis, but wrong to identify its scope with that of a pnrtit:rrlnrbrarrclrof scicnce- namely, physics. There are in physics lhlsilirblc laws - f<rr exanrple, Snell's law of refraction applied to glass - whcrcas physiology and chemistry no doubt contain Irypothcscs sulr.icctto the Duhem thesis. When we return to this r;rrc:siion clraptcr 10, I will argue that the important thing is to irr tlistirrg,rrish, within any branch of science, two different types of l;rws, lrypotlrcscs, or theories, which l call level 1 and level 2. l.cvcl-l lrypothcscs are falsifiable by observation statements, lcvcl-2 hypotlrcsescantlot be compared with experiencein wlrt'r't'trs isolut,iorr, l)ut ()nly whcn they are taken in conjunction with other lrypothcscs. l)rrlrcrn is thus only partially justified, inasmuch as plr y s ic sis ' r rro rc tl rc o re ti c a l ' th a n m o s t other sci ences, and so con-

tains a higher proportion o[ level-2 hypotheses. For the moment, however, it is not of great importance where exact boundaries are drawn. The crucial point is that Duhem wanted to apply his thesis to sonle statementsand not to others, whereas the Quine thesis is supposedto apply to arly staternentwhatever. 'fhis is closely connected with a second difference between the l)uhern thesis and the Quine thesis. Duhen-r maintains that hypothesesin physics cannot be tested in isolation, but only as part of a group. However, his discussionrnakesclear that he placeslimits on the size of this 'group'. Quine, however, thinks that the group extends and ramifies until it includes the whole of human knowledge. Quine writes: 'The unit of empirical significanceis the whole of science' (1951, p. 42); and again: The totality of our so-called knowledgeor beliefs,from the most casual mattersof geography and history to the profoundest laws of atomicphysics evenof pure mathematics logic, is a man-made and or fabric which impinges on experience only along the edges.Or, to change figure,total science like a field of forcewhoseboundary the is conditions experience. conflictwith experience the periphery are A at reacljusirnents the interior of thi field: . . . gui the total in occasions field is so underdetermined its boundary conditions, experience, by that there is much latitudeo[ choiceas to what statements reto evaluate the light of any singlecontraryexperience. particular in No experiences linkedwith any particular are statements the interiorof in the field, except indirectly through considerations equilibrium of affectingthe field as a whole. (pp. a24) The Quine thesis is stronger giranthe Duhem thesis, and, in my view, less plausible, Let us take, as a concrete example, one of the casesanalysedearlier. Newton's first law cannot, taken in isolation, be conrpared with experietrce.Adams and Leverrier, however, used this law as one of a group of hypothesesfrom which they deduced conclusionsabout the orbit of LJranus.These conclusionsdisagreed with observation. Now the group of hypotheses used by Adams and Leverrier was, no doubt, fairly extensive, but it did not include the whole of science. Adanrs and Leverrier did not, for example, nrention the assumption that bees collect nectar from flowers in order to make honey, although . such an assumption might well have appearedin a contemporary scientific treatise dealing with a question in biology. !7e agree, then, with Quine that a single statement may not always be (to use his terminology) a 'unit of empirical significance'. But this does not mean that 'The unit of

rli11li,.

. it . i :

112

Cenventionalism tl.e Duhem-Ouine Thesis and

The Duhem Thesisand the OuineThesis

113

empirical significance is the whole of science' (1951, p. 42). A group 'of statementswhich fllls considerablyshort of the whole of science may sometimes be a perGctly valid unit of ernpirical significance. Another difference between Duhem and Quine is that Quine does not have a theory of scientific good sense.Let us-take, for example, Quine's statemnt:'Any statementcan be held true conle what may, if we make drastic enough adjustments elsewhere in the system' (p. 43). It is easy to imagine how Duhem would have reacted to such an assertion when applied to a statement falling under his thesis. Duhem would have agreed that, frorn the point of view of 'pure logic, one can indeed hold a particular statement - for example, Newtotr's particle theory of light - to be true, come what may. However, someone who did so in certain evidential situations would be lacking in good sense,and indeed perfectlyunreasonable. BecauseQuine does not have a theory of good sense,he callnot give the Duhemian analysiswhich we have just sketched. Indeed, it is significant that his 1951 article, 'Two Dogmas of Empiricism', is :reprinted in a collection entitled From a Logical Point of View. Where Quine does go beyond logic, it is towards pragmatism, though Quine's pragmatism is usually mentioned only in passing, rather than elaborated, as in the following passage: 'Each nran is given a scientific heritage plus a continuing barrage of sensory stimulation; and the considerations which guide him in vrarping his scientific heritage to fit his continuitrg sensory promptings are, where rational, pragmatic' (p. 46). ,- Although the Duhem thesis is quite clearly distinct fronr the iQuine thesis, it might still be possible- indeed, useful - to forrn a icomposite thesis containing some, but not all, elements from each f of the two theses. The phrase Duhem-Quine thesiscould then be lvalidly used to denote this composite thesis. In the last section of Ithis chapter, I will elaboratea suggestion along theselines.

which the holistic thesis does apply. In discussing these differences, I have so far sided with Duhem against Quine. There is one point, however, on which l would like to defend Quine against Duhem. Quine, as we have seen, extends the holistic thesis to mathematics and logic. Duhem, however, thought that mathematics and logic had a character quite diflferent from that of physics.'Crowe (1990) gives an excellent general account and critique of Duhem's views orr the history and philosophy of mathematics. I will here confine myself to a brief account of some views corrcerning geometry and logic which Duhem expounded in his late work La Science allemande ('German Science'). Duhem begins his treatment of geometry with the following remarks: Among the sciences reasonirrg, of arithmetic and geornetry are the the most simple and, consequently, most completelyfinished;. . . . What is the sourceof their axioms?They are taken, it is usually knowledge(cowtaissance commune): is to that said,frorn cornmon sense say that any man saneof mirrd is sure of their truth beforehaving studied the scienceof which they will be the foundations. (1915, pp, 4-5) Duhem agreeswith this point of view. In fact, he holds what in 1915 was a very old-fashioned opinion, that the axioms of Euclid are establishedas true by common-sense knowledge (connaissance commune) comrnon sense (le senscommr.ut) intuitive knowledge or or (connaissance intuitiue). A proposition from which Euclid's fifth postulate can be deduced is that, given a geometrical figure (say a triangle), there exists another geometrical figure similar to it but of a different size. Duhem argues that the intuitions of palaeolithic hunters of reitrdeer were sufficient to establish the truth of this proposition. As he says: One can represent plane figure by drawing, or a solid figure by a sculpture, and the image can resemble tl're rnodel perfectly, even though they have different sizes. This is a truth which was in no way doubted, in palaeolithictimes, by the hunters of reindeer on the banks of the V6zire. Now that figures can be similar without being equal,implies, as the geometricspirit demonstrates, exacttruth of dre E ucl i d' s postul ate. (pp. 115- 16) Naturally enough, this attituie to the foundations of geometry leads l)uhem to criticize non-Euclidean geometry, and, in particular, Riernanniangeometry. This is what he says:

5.4 The Duhem-Ouine Thesis

Let us say that the holistic thesisapplies to a particular hypothesis if that hypothesis cannot be refuted by observation irrd eiperiment when taken in isolation, but only when it forms part of a theoretical group. The diflbrences between the Duhem and Quine theses concern the range of hypotheses to which 'the holistic thesis is applied and the extelrt of the 'theoretical group' for a hypothesis to

:::;llli;;;:;.; ,,
.". tJ:,:' I

":.:
,':.::r.

It trf
I i

(lorrverrtionalism Duhem-Ouine andthe Thesis

The Duhem Thesisand the OuineThesis

115

is fli,',rr.rir,'s rloctrirrc a rigorous algebra, all the theorems for which it lirrrrrulllcs:lr(! very prccisely deduced from its basicpostulates; it so liirtirlies gcorrretric tlre spirit. It is not a true geoffietry, in putting fbr, lorwlrrl its lxrstrrlates,is not concerned their corollaries it that should JHrceilt cvcry poirrt with the judgernents,drawn from experience, wlrirlr corrstitlrtc our intuitive knowledgeof space;it is therefore r r lr r Ht l ;l trt c o l l n ro l rs e n s e(p . l l 8 ) to .

It is lrerlrnpsno acciclentthat the non-Euclidean geometer cited hy l)trlrerrr (rr;rrrrcly,Riemann) was a German; for, as already t r t r t r r k c r l , L t Sti a n c c l l e m a n d e , ri tte n i n 1915, w as an exampl e of a w llrr wrr lilcrntrrrc of the time, designed to denigrate the enemy lat ior r ; r lit y . l )u h e n r a tta c k s G e rma n sci enti stsby cl ai mi ng that, wfrifr tlrey l)osscsstlre geometric spirit (I'espritg1omitrique),their ics r'orrlra<lictcorn.lnon sense (/e sens cornmurr) l'esprit de llrerrr or lilrctr, wlliclr is l)uhem's new term for something like his old t t ot ior rol' 1 1 o o tl e trs c . s (livcrr tlris gcrreralpoint of view, it is not surprising that we find | )rrlrcrrr t:orrrlnrrning the theory of relativity. He speaks of 'the lr r ir r r iple o l ' rc l :rti v i ty s u c h a s h a s b e e n concei ved by an E i nstei n, r t M lx A h n rl rrn r, a Mi n k o w s k i , a L aue' (p. 135). Forgetti ng the lor r t r ilr r r l i o rrs l ' l ri s ()w n c o n l p a tri o t P oi ncar6, he denouncesrel ao t iv it y : r s : r ty p i t' a l l b c rra ti o n o f th e G e rrnan rni nd. A s he says: 'l'lrefhr'tth;rtthe prirrciplc rclativityconfounds the intuitionsof of all ((rrtnr()m ncrse, docsnot arouse against the mistrustof tlre German it it To lrhyritists- t;rritt:tlrc t:ontrary! accept is, by that very fact, to wherespace, ovrrturn ;rll the <krr:trilrcs time, movementwere treated, and physics;such a devastation has ell tlrc tlrcorit'sof rrrcchanics rrotlrirrg ;rlrorrt wlri<:h displease it. can Gerrnan thought;on the ground wlrirlr it will lr:rvcclcarcdof the ancientdoctrines,the geometric w npir it ol ' tl re (i c rrrra n s i l l d c v o te i ts el f w i th a happy heart to reof I'rrilrlirrp,wlr.tlencw plrysics which the principleof relativitywill l lf'tlris rrcw physics, disdainful common sense, of lre tlrr lirrrrrtlltiorr. ((rlnlcl trr:rlltlr;rt<.rbservation experience and haveallowedto be t!irrF and terrestrial mechanics, the { orrrtrur lltl irr llrc tlornlin of celestial n lr ur r ' lyr l rrh rr' ti v c rc tl ro dw i l l o n l y b e more proud of the i nfl exi bl e rip,nrrrwitlr rvlrir'h it will have followed to the end the ruinous ( nr qr r lr re r( so l i ts p o s trrl a te . . 1 3 6 ) (p r' ' f ' lr , L' v c l ,' ,l rtrrc rrl rc c e p ta n c e f non-E ucl i deangeometry and :rrrc l o r r . l; r t t v it yl rl rv c rc rrrl c re rll )u l re m ' s a tte mpt to found geonl etry on r ililil11{ r r r(' rrrc l i l rl (' n i [)l e . It i s s u rel y now more reasonabl eto

extend the holistic, thesis from physics to geometry and to say that, in the face of recalcitrant observationsr we have the option of altering postulatesof geometry as well as postulatesof physics. This is, after all, precisely what Einstein did when he devised his general theory of relativity. The picture is the same when we turn from geometry to logic. We have already quoted (chapter 3, note 4) Duhem as saying that 'There is a general method of deduction; Aristotle has formulated its laws for all time (pour toujours)'(p. 58). Yet by 1915 the new logic of Frege, Peano, and Russell had clearly superseded Aristotelian logic. Moreover, Brouwer had criticized some of the standard his alternative intuitionistic approach. logical laws, and sup;gested Quine writes: 'I{evision even of the logical law of the excluded rniddle has been proposed as a nleans of simplifying quantum mechanics' (1951, p. 43). Admittedly the new 'quantum logic' has not proved very successful in resolving the paradoxes of microphysics; but there is no reason in principle why a change of this kind should not prove efficacious in some scientific context. In artificial intelligence, non-standard logics (foi exarnple, non-monotonic logics) are being devised in order to model particular forms of intelligent reasoning, and this programme has met with some success.'Ihus it serns reasonableto extend the holistic thesis to include logic as well and to allow the possibility of altering logicai laws as well as scientific laws to explain recalcitrant observations. I am now in a position to formulate what I will call the DuhemQuine thesis,which combines what seem to me the best aspectsof the Duhem thesis and the Quine thesis. It will be convenient to divide the statement in two parts. A The holistic ttresis applies to any high-level (level 2) theoretical or hypotheses,whether of physics or of other sciences, even of mathematics and logic. (A incorporates ideas from the Quine thesis.) The group of hypotheses under test it.l any given situation is in practice limited, and does not extend to the whole of human knowledge. Quine's claim that 'Any statement can be held to be true come what may, if we make drastic enough adjustments el se wher ein t he syst em '( 1951, p. 43) is t r ue f r om a pur ely logical point of view; but scietrtific good sense concludes in to many situations that it would be perfectly unreasonable hold on to particular statements. (B obviously follows the Duhem thesis radrer than the Quine thesis.)

116

Conventionalism the Duhem-Ouine and Thesis

In what follows I will use the phrass'the Duhem-Quine thesis' to to denotethe conjunctionof A and B. The thesisseems me to be both true and important, and in chapter 10 I will examine what it consequences has for Popper's falsificationism.Now, however, we will turn to the third of our four centralthemes,and examine the often problematicnature of observation.
ii:

P a rtlll
The Nature of Observation

'il

+
i

--

-t

llffi;::;

120

TheNature Observation of Statements Viewson Observation 6.1 Carnap's in the Early1930s

lentlst

Protocol Sentences t should B. lf A's

ble of bein c checked

121

Carnap was persuadedby Neurath to changefrom psychologism to physicalism, and, like many converts, he becamean extreme of advbcate the new view, arguing that physicalismappliedeven to that the psychologyitself, Now it does not seemtoo unreasonable like ' observation of statements sciences physicsand chemisry should be about physicalobjecti. Should not psychology,however, rely on A its subjects'reports of their immediate subjectiveexperietrces? study dreamsby asking subjects psychologistmight, for example, io'rtpott their dreams,as well as by noting his own. Thus for psychology,at least,psychologismmight seemto apply. However, in his article of 1932/3, Carnap arguesthat physicalismis correct even for psychology. Neurajh had itttr g the gl@in espon-din t.ffi from the syitemlanguage:'Of first importance for fiffilTnguage ipistemological analyses are the protocol language,in which the (in plimitiu. protocol sentences the material mode of speech:the about the immediately given) of a particular person are sentences of in language, which the sentences the formulated, and the system (193213,pp. 165-6)' In terms system of scienceare formulated' of this distinction, Carnap formulates the thesisof physicalismas follows:
some of To every sentence the system languagethere corresponds are of sentence the physical languagesuch that the two sentences . inter-translatable.. . The various protocol languagesthus become sub-languagesof the physical language- The physieal languageis This anil uniuersal inter-subjective. is the thesisof physicalism. lf the physicallanguage,on the grounds of its universality,were would become all adoptedes the systemlanguageof science, science (p. as physics.Metaphysicswould be discarded meaningless. 166) One point to notice here is the claim that statements about physical objects are inter-subjective. This is really the crux of the matter. The line of thought is this. Let us first consider the protocol sentenqes of a science like physics or chemistry. Such sciences are carried on by a community of scientific workers. Now a particular observation or experiment may be carried out by scientist A. the community, it is
j

ocol

A's protocol, in thesecircumstances, becomes meaningless B. for This is his reason for applying physicalismeven to psychology. Let us next see how he statesthe irhysicalistthesis in tl"recaseof psychology. Carnapconsiders example,'the sentence "Mr A is now the P1: excited."' (p. 170),and comments:
P1 has the same content as a sentence P2 which . . . asserts the existence of that physical structure (microstructure) of Mr A's body (especially of his central nervous system) that is characterized by a high pulse and rate of breathing, which, on the application of certain stimuli, may even be made higher, by vehement and factually unsatisfectory answers to questions, by the occurrence of agitated nlovements on the application of certain stimuli, etc. (p. 172) But why do we have to give this interpretation of the sentence P1,

'Mr A is now excited'? Would it not be more reasonable say that to P1 refers to an inner psychological state of Mr A? Carnap argues 'No', because in that case P1 would be testable only by Mr A himself, and not by another individual, B. Carnap assumes that a sentence P1 is only rneaningful for an individual B, if B can rest P1. So on the psychologistic interpretation, P1 is meaningless for everyone except Mr A. This is absurd, however, since P1 is clearly meaningful for people other than Mr A, and so Carnap concludes that P1 must be interpreted in a physicalist rather than a psychologistic way. The same applies even to 'I am now excited' when uttered by Mr A hirnself. As Carnap puts it: Let us say that psychologist writes sentence '(I am) now excited' A p2: into his protocol. . . . the view which holds that protocol sentences cannot be physicallyinterpreted,that, on the contrary, they refer to somethingnon-physical(something'psychical', some'experiencecontent', some 'datum of consciousness', etc.) leads directly to the consequence that every protocol sentence meaningful only to its is autlror. If A's protocol sentence werc not subject to a physical p2 interpretation, it could not be tested by B, and would, thus, be mearringless B. (pp. 192-4) to

-l

12?.

The Nature cf Observation

ProtocolSentences

123

'l'lris prpcr of Carnap's may well have influenced Wittgenstein's htcr {brrrmlation of 'the private language argulnerlt' - that is, the ilrgrrrncr)t that a private language is irnpossible (cf. Wittgenstein, l()53, scc. 243f.). Indeed, as we shall see in a moment, Neurath the possibility of a private language. explicitly clenies Onrrrap buttressed his argument about testability and meaning hy l consideration, again similar to some passages in the later Wittgcnstein, of how a child leams language. He takes the example (193213,p. 196) of a mother who, observing her child to be in the physical state of tiredness,puts him to bed, and says:'Now you are lnppy to be in bed.'Later, when tl-rechild is tired and is put to bed, f re slys: 'Now I am happy to be in bed', an expressionwhich he has lclrut as a way of describing his physical state. 'l'llcsc arguments of Carnap's do not seem to me to be decisive.It like 'l anr happy' or 'Mr A is corrltl still be clairned that sentences excitcd' refcr to subjectiveinner states,while agreeilrgthat we could rrot rcfcr, in a public language, to such states unless they were correlatcd with publicfy observable physical conditions and beImvioural patterns. We will return to this question later on. Let us rf('xl corrsider Neurath's important 193213 paper on 'Protocol S er r ler r cc s ' .

die is lying on the table' are not completeprotocolsentences. They are not even candidates a position within the innerrnostset of for brackets. For this they would, orr our arralysis, leasthaveto read at ' Ot t o now joy, 'or 'O t t o now sees r ed cir cle, 'or 'O t t o, now sees a a red die lying on the table'. (p. 202) Of course 'Otto now joy' would not be a suitable candidate on other grounds, since Neurath rejects psychologism. We corne r)ext to a most important point. Protocol sentences, according to Neurath, are corrigible, and may be discarded. Here Neurath's position is very different from the earlier psychologistic view. Accordirrg to psychologism. a orotocol sentencedescribesthe irrrrrrediate experienceof an observer. It is thus completely verified by the observer's irrrrnediate experieng:. So protocol sentences constitgle a.firm arrd incorrigible hasison which scienceca" he huilt to gg This ceases be true once we switch to physicalism. Otto can protocol that there is a table in the room and yet, bccause is lying he or suffering from a hallucination, there is no table in the room. The scientific community may initially accept his protocol, and then conle to reject it. Ncurath puts the point as foilows: ln unifiedscience try to construct non-contradictory we a systemof protocol sentencesand non-protocol sentences(including laws). Whcn a new sentenceis presentedto us we compare it with the systemat our disposal,and determinewhether or not it conflictswith that system. If the sentence does conflict with the system, we may (or would be done with 'ln discardit as useless false),as, for instance, Africa lions sing only in major scales.'One may, on the other hand, and acrcpt sentence so changethe systemthat it remainsconsistent the The sentence even after the adjunction of the new sentence. would then be called'true.' The fate of being discardedmay befall even a protocol sentence. which Carnap ordains for No serrtence enjoys the noli mi tangere (p. protocolsenterrces. 203) Neurath, of' course, here refers to the earlier, psychologistic Carnap. To reinforce the point about protocol sentencesbeing corrigible, Neurath considers an example which curiously anticipates the finciings regarding the split brain. He imagines a scholar called 'Kalon' who, at the same time, writes with his left hand that there is nothing in the room except a table, and with his right hand that there is nothing in the roon except a bird. under these circumstances, the scientific comrnunity would have to discard either or both of the protocols.

Statements Viewson Observation 6.2 Neurath's in the Early1930s

are about physical objects, Although Neurath's protocol sentences they clo not have a simple form such as: 'There is a table in the rooln."Ihe best way to explain their curious structure is to give Ncrrrath's own example of one: protocol sentencemight, for instance, read: 'Otto's A <:orrrplete at3:17o'clock:[at 3:16o'clockOtto saidto himself (at 3:15 lrrotoc()l by <r'clock therc was a tableirr the room perceived Otto)].' (193213, 1t. 202) Ncurath goes on to stress that protocol sentencesmust contailr tlre rralrreof a persotr (the observer): that the name of to [:or a lrrotocol scntence be completeit is essential s()rncpcrsolloccur irr it. 'Now joy,' or 'Now red circle,'or 'A red

124

The Nature of Observation

Protocol Sentences

125

This article of Neurath's seems to have had an enormous influence on the subsequent course of philosophy. As we have already remarked, . Neurath explicitly denies that there can be a private ,language, and so very probably influenced the later Wittgenstein. The passage in question is the following: 'In other words, every language as such is inter-subjective. The protocol of one moment must be subject to incorporation in the protocols of the next, just as Ithe protocols of A must be subject to incorporation in the protocols ,of B. It k therefore meaninglessto talk, as Carnap does, of a priuate olayguage'(p. 205). ;' Neurath's rejection of psychologism leads him to formulate the lfamous comparison betwe-en scientists and sailors who have to rebuild their ship at sea: 'J

N
t >. t9

,N
t\

iv r5

rs

lished
' open sea, never able to dismantle it in d and to reconstructit to have influenced him
l{baref6 a.b6 a aterer/lst

f8 ,\

iq

we are committed to the view rhat. statementscan be justifed only by , stdtements. The demand that all statements are tq be logically justified (described by Fries as a'predilection for proofs') is therefore bound to lead to an injnite regress.Now, if we wish to avoid the danger of ' . dogmatism as well as an infinite regress, therr it seems ai if we could onl y have recourse to psyclolo2i sra,j.g.jhglbld_nelh4t statements gj,r. b. j,rstifi.d ,rot o; perience; Faced with this trilemma - dogmatism zs infinite regress zs psychologism - Fries, and with him almost all epistemologists who wished to account for our empirical knowledge, opted for psychologisrn. In sense-experience, he taught, we have 'immediate knowledge': by this immediate knowledge, we may justify our 'mediate knowledge' - knowledge expressed in the symbolism of some language. And this mediate knowledge includes, of course, the statementsof science.(%a, pp.93-4)

there out of,the best materials'(p. 201). This passage cited by Quine, and is in developinghis holistic philosophy.

N
i\

i$ s{

l) N
i)

6.3 Popper's Viewson BasicStatements 1934 in

Let us turn next to Popper'streatmentof the questionin his Logicof (1934). Popper dealswith the question in chapter ScientifuDiscovery 5, which is entitled 'The Probleri of the Empirical Basis'. But he talks neither of 'observation statements'nor of 'protocol statements', but of 'basic statements'.Popper's book was published shortly after the papersof Carnap and Neurath which we havejust and, as we shall see, Popper refers frequently to these discussed; papers. Popper beginshis discussion, however, by mentioningJ. F. Fries (1828-31) and his book Nere oderanthropologkche Kritik der Vernunft ('New or Anthropological Critique of Reason').Popper extracts from this work what he calls Fries'strilemma, and expoundsit as follows: The problemof the basis ofexperience troubledfew thinkersso has
are of ' deeplyas Fries.He taught that, if the stetements science not to be accepted dogmatically, we must be able to justdy them. If we demandjustification by reasoned argument,in the logical sense, therr

:p s

r$
i\) i\ 'r q

epistemologists who wished to account for our empirical knowledge' by rejecting psychologism. Thus Popper is here in agreement with Neurath and the later Carnap. However, it is worth noting that the version of psychologism criticized by Popper is rather \trongef than that attacked by Neurath and the later Carnap. These . two oppose the view that observation statements describe the immediate experience of a particular individual and are completely ,justified by those experiences. Popper, however, attacks the view that 'statements can be justified not only by statements but also tatement

l.s

ment cannot be iustifiedbv them - no more than bv thumoins the

tare' | 05) b (p. .,b,rffi#..$ f" a:fr ",lx,g;:+$ { q t,",b {_t+ We have earlier enilorsefl the anti-psychologism of Neurath'end trua.
^rnrc,$[ used in science must be inter-subjectivein character. Popper's(fr '1;-^ stronger version of anti-psychologismseems more questionable, however. For example, is it true that statements can be compared only with statements?Why can't I compare a statement with some of my perceptual experiences?Let us take the statement 'There is a the later Carnap, and have agreed that'the observation tt^t

126

The Nature Observation of

Protocol Sentences

127

copy of "The Logic of Scientific Discovery" in this roorn.'At this very moment I am engaged in comparing this statement with a group of my visual and tactile experiences (and, indeed, am finding that it accords with these experiences). Then again, consider the view that a basic statement cannot be justified, even partially, by perceptual experiences. Suppose I make the statement S: 'There is a tortoise in my bedroom', and that S is disbelieved by Mr A. It is not in fact true that I have to produce .another statement, S', in brder to justify S. Without uttering a 'word, I could lead Mr A to my bedroom and point to something crawling about on the bed. S would then be justified by Mr A's perceptual experience. Of course, this justification would only be partial. It might be that the object in question is really an electronic toy purchased by me with the aim of deceiving Mr A. However, to say that a justification is only partial is not to say that it does not exist at all. For these reasons, then, I cannot accept Popper's stronger version of anti-psychologism. Let us turn next to some criticisms which Popper makes of Neurath. Popper writes: Neurath's view that protocol sentences not inviolable represents, are in my opinion, a irotableadvance. But apart from the replacement of perceptions perception-statements merely a translationinto the by formal mode of speech the doctrine that protocol sentences may be revised is his only advanceupon the theory (due to Fries) of the immediacy of perceptual knowledge.r It is a step in the right direction; but it lelds nowhere if it is not followed up by anotherstep:we need a set of rules to limit the arbitrariness 'deleting' (or else of 'rccepting') a protocol sentence. Neurath fails to give any such rules and thus unwittingly throws empiricismoverboard.For without such rules, empirical statementsare no longer distinguishedfrom any Every system becomesdefensible one is if other sort of statements. allowed (as everybody is, in Neurath's view) simply to 'delete' a protocol sentence it is inconvenient.(p.971 if Popper seelns to me to be right here. Some rules are indeed needed to lirnit the arbitrariness of 'deleting' (or else 'accepting') a protocol sentence. But how can Popper himself provide any such rules? In terms of Fries's trilemma, Popper has rejected both psychologism and dogmatism, and so seems to be left with an infinite regress. Any stetement is corrigible, and we can justify S only by producing another statement, S'. S' in turn is corrigible, and can be justified only by another statement, S", and so on.

Popper has the problem of evading this infinite regress, which he does in an interesting and ingenious fashion. Popper thinks that it is indeed possible to go on testing any particular basic statement (S; say) indefinitely. However, such a procedure would render scienceimpossible, and so scientistshave to make a decision and reach an agreement to accept S when it has passed a certain number of tests. I'his decision has a cqnvg4lignal .l..rr..rt. and so basic statements are partly .o.t.'frii6ffir--.. says, 'Palic statements are accepted as the result of a decision or asreement: and to that extent thev are conventions'(p. 106). The decision about when to stop is not entirely arbitrary, however, and r mentions, in parti

As he says:
Any basic staternent can again in its turn be subjected to tests, using es e touchstone any of the basic statements which can be deduced frorn it with the help of some theory, either the one under test, or another. This procedure has no natural end. Thus if the test is to lead us anyu'here, nothing remains but to stop at some poirrt or other and say that we are satisfied, for the time being. It is fairly easy to see that we arrive in this wey at a procedure according to which we stop only at a kind of statement that is especially easy to test. (p. 104)

which are especially easyto This criterionof stoppingat statements suggested test leadsPopper to criticizethe form of basicstatements by Neurath. Popper's point is that it is usually easier to test a statement,S, of the form 'There is a tortoisein my bedroom', than a statement,S', of the forrn 'Mr A is observing that there is a tortoisein my bedroom'. Indeed,to test S', we may first have to test S, and then, if S turns out to be true, to test that Mr A is indeed Aq the observing reptilein question. Poppersays:
There is a widespread belief that the statement 'I see that this table some profound advantage over the statement here is white', possesses 'This table here is white', from the point of view of epistemology. But frorn the point of view of evaluating its possible objective tests, the first statement, in speaking about me, does not appear more secure than the second statement, which speaks about the table here.

(p.ee)

In order to evaluate this interesting disagreement between Neurath and Popper, let me first introduce some terminological

:r_rli?;.":. ,_
:,r l :!:,:..; ...t,.'

: r ir - -

128

r;f TheNature Observation

ProtocolSentences

129

or conventions. I will use Neurath's term protocolstatement, protocol, for a statement in which the observer is explicitly specified. Thus a

typical protocol would be: 'Mr A in such and such a laboratory at such and such a time observed that S'. I will confine the term
obserldliott slaletaet t to statements which ttt'tt't"t thg. rqsult of att

in.any way alh'ding- tq the observation or expgriment wjth.-o,.gt 'This table here is white', although eximple, Popper's o6ffi6lf"r

T6ffi-es,

observatiotr for emphasis, I will speak of impersonal

to statefttenls emphasize that the statement does not refer in any way to the person making the observation. The question before us is whethei protocols or impersonal observation statements should be preferred as the basic statements of science' Neurath argues for as we shall lrotocols, Popper for observation statements - though, see later, Popper does qualify his view at one point' I will argue that both types of statement are necessaryfor science, but that their role is rather differelrt. It should first be noted that protocols do iIr fact occur in papers by leading scientists. To derhonstrate this, we can quote once again the opening of Fleming's 1929 paper on penicillin. Fleming begins the paper as follows: variants a number of cultureWhile working with staphylococcus plateswere setasideon the laboratorybenchand examinedfrom time exposed to io time. In the examinations theseplates were necessarily the air and they becamecontaminatedwith various micro-organisms. It was noticed that around a large colony of a contaminating mould and were obviously coloniesbecametransparent the staphylocbccus undergoinglysis. (1929,p.226) The last sentence here is a protocol. Admittedly the impersonal form; 'It was noticed that...' is used; but, in the context of the paper, this is clearly equivalent to: 'Alexander Fleming noticed that . . .'. The fact that protocols do occur in leading scientific papers is not by itself decisive, for it could be argued that they are in principle dispensable. To show that this is not so and that protocols are really needed, it will be helpful to return to @ statements are to some extent conventions. Popper illustrates this by trial by jury. As he says, 'By its ?rawing an intereiirng ant[6[F detision, the jury accepts, by agreement, a statement about a factual occurrence - a basic statement, as it were' (1934, p' 109). The verdict of the jury is then used as the basis for further legal pro-

cedures: for releasing the prisoner, for example, or for sending him to gaol. [n the same way the jury of scientists accepts, by agreement, an observation statement which then becomes Part of the data used to evaluate scientific theories or hypotheses. This arralogy seems to me a good one, and it illustrates very nicely the element of social agreement which goes into the acceptance of an observation statement. The analogy does not, however, in my opinion, lend support to Popper's extreme anti-psychologisrrr. Let us suppose that a jury of scientists agrees to accept a Particular observation statement S; then S will, in general, have been partly justified or verified, by the perceptual experiencesof at least some of the jury. In the same way, the verdict of a jury in a legal casewill, in general, be at least partly justified by the evidence they have heard. Of course, in neither case is the justification complete. Both observation statenlents and the verdicts ofjuries are corrigible, and can be rejected as a result of later investigation. Theoretically, each member of the scientific jury could perform the experimetlt or observation, and partially verify on the basis of his or her own perceptual experiences the observation statement in question. In practice, however, lo-r reasonsof expense,lack of time, and so on, the observation or experiment will be performedby only a few members of the scientific jury, and the others will rely on the reports of these active observers or experimenters. This is where the need for protocols arises. Consider Mr X, a member of the scientific jury who has not made a particular observation himself. Mr X has to decide whether the corresponding observation statement (S, say) should be accepted, and, to do so, he has to rely on a number of protocols of the form: 'Mr A observed that S', 'Ms B observed that S'. and so on. If Mr X knows that Mr A, Ms B and so on are honest and reliable observers or, at least, that they work in institutions noted for the maintenance of high scientific standards, he will be inclined to accept S provisionally as correct. To form this judgement, however, he will need to consider not just the observation statement S, but the protocol statements 'Mr A observed that S', of 'Ms B.observed that S', and so on, for his acceptance S depends partly on his judgement as to the reliability of Mr A, Ms B, and the rest. Moreover, if at some later stage, reasons arise for doubting the truth of S after all, Mr X may want to consider the protocols again to see whether Mr A, Ms B, and so on might have been mistaken. In fact, Popper qualifies his criticism of Neurath at one point, and hints at a 'combined theory' of the type we have just sketched. Thus, he writes:

130

of The Nature Observation

Protocol Sentences

131

We stop at basic statementswhich are easily testable.Staternents about personalexperience i.e. protocol statements are clearly rzof of this kind; thus they will not be very suitableto serveas statements at which we stop. SZe of coursemake useof recordsor protocols, do of such as certificates testsissuedby a departmentof scientificand industrial research.These, if the need arises,can be re-examined. for example, to test the reactionThus it may become necessary, times of the expertswho carry out the tests(i.e. to determinetheir (pp. 104-5) personalequations). It might be objected that the use of photographs, tape recordings, and so forth could eliminate the need for protocols by individual scieutists. Earlier we quoted Alexander Fleming's protocol: 'It was noticed (by Alexander Fieming) that...'. Yet Fleming published a photograph of the penicillin culture-plate in the same paper (seeour l'}latc l). It could be argued that a member of the scientific jury, Mr X, when considering whether to accept the corresponding observatibu staternent, could rely on the photograph, and forget about Alexander Fleming'$ personal protocol. It is true that photogratrrhsarid similar devices do help here; but it would be wrong to think that they elirninate the need for protocols altogether. In order to interpret the photograph in the paper, Mr X has to acceptthat the photograph is of a culture-plate which had been prepared in such arrclsuch a way. Moreover, to accept these statements, he has to rely orr tlre protocols of Alexander Flerning. Science accepts ancl uses inrpcrsorral, inter-subjective observation statements; but to accept suclr stntcrrrents,it has to rely ultimately on reports which are based on thc persoiial sensory experiences ofindividual scientists. 'l'o surrr up, thcn, I have diverged from Popper in allowing sorne clcgrec of psychologisrn and in according an important role to prrtt<lcols wcll as to impersonal observation statements.It should as Ire stressed, however, that this is notin order to provide sciencewith arr irrcrlrrigible eurpirical basis, as was the original intention of psych<llogislrr. lrrdividual scientistscan very easily misinterpret their Bens()ry exl)crience,and so produce mistaketr protocols. Indeed, an observatiottstatementacceptedby the scientificjury will irnpersorral olierr be nlorc ccrtain than the protocols on which it is based,and an irrtlivirltral protocol rrray sometimes be corrected in the light oi an observati<rnstatement. Suppose, for example, that five a<:cepterl scielrtists, Mr A, Ms B, . . . , Dr E, independently perform the srrrre experirrrcrrtarrd reach exactly the same conclusion, which is expressedirr the staternentS. Their five protocols: 'Mr A observed tlul S', 'Ms l] obscrved that S', . . . are pooled by the scientific

rl
I

cornmunity, which comes to accept the impersonal observation statement S. Because of the convergence of independent experiments, S is much more certain fhan any of the protocols upon which it is based. Now supposefurther that a sixth scientist, Mr F, has also carried out the experiment and obtained a result different from S. In the light of the acceptanceof S, he may re-examine his experimental procedure to find out whether he rnade some mistake. Very likely he will discover an error in what he did, and, on repeating the experiment in corrected form, obtain S like the others. But then again, things might go the other way. Mr F might discover that the result S was obtained only because the others had neglected to take sorne important precaution. They might be persuaded by him to repeat their experiments with this modification, and the end result could be that everyone agreed on Mr F's statement, S', rather than the original S. I am thus in agreement with Popper that all observation statements are corrigible, and would like to close this chapter by quoting the memorable analogy with which he illustrates this corrigibility:

';l
tl l

li,

(
I

tive science has thus nothine 'absolute' of its theories rises, as it were, a a swamp. It is like a building erected on piles. The piles are driven down from above into the swamp, but not down to any natural or 'given' base; and if we stop driving the piles deeper, it is not because we have reached firm ground. We simply stop when we are satisfied that the piles are firm enough to carry the structure, at least for the time being. (p. 111)

Theory-Laden? ls Observation

IJJ

7
Theory-L,aden? ls Observation

V n 7,1 Duhem's iewthatA ll O b s e rv a t ioin P h y s ic s is Theory-Laden

observation and l.(:l ilri lul'n llow to tlrc views of Duhem concerning in cxlr('tinlcut srricrr<:c, as in the case inductivisrn, irr the of so Just of we ptr'scnl('irsc obsc'rvatiorr, lrave foulrd it convcnientto brcak orclcrarrclto clealwith the Vienna Cir:clearrd lltr r'lrrorrolrtllir:al bclirre' l)ulrcnr.It shouldnot be forgotten.however,that Pr,1r1rcl w:tswritirrgscvcral years these l)ttltettt befiore Austrian thinkers. l)ttlrcnr'str('ltnrcntof .observation science, in like his critique nl irrtlrrltivisrtr, <lf the very highestimportance, is and must be to of rrurior contribution the philosophy science. is It rotrrirlcted;r lrrnt,rirrr.rf lris .rtin and Structure Physical of Theory(1904-5),part irr 'Experinrent Physics'. in With his usualclarity ll, r lr,rptcr crrtitlcd 4, arrrl irrrisivcricss, l)rrhcnr beginsby stating his centralpoint as ftrllrrs,s: of ,4n litpuiutrtrt in PhysirsIs No, Simply the Observation a the lt Inlerpt'etation 'this of Itltutottttrtort; Is, Iiesides, Theoretical ( | 9()4--5, la4) p. l'ltrtronrrttotr.
'l'lttr vtcr v , ,rl tr ir'l r rr ' ,

obscrver plungcs thc metallic stenl of a rod, mounted with rubbcr, into small holes; thc iron oscillatesand, by mcans of the mirror ticd to it, scncls a bcaru of light to a celluloid rulcr, ancl the obscrver Rrllorvs tlrc rrrovcrncntof thc light bcanr on it. Thcre, no doubt, yor.r havc rn experinrent; by rncarrsof the vibration of this spot of light. this physicist nrirrr"rtcly obscrves the oscillatioms thc piecc of iron. of Ask hirrr now rvhat he is doing. Is he going to arlswer: 'l am studying thc oscillationsof the piece of iron carrying this mirror?'No, hc rvill tcll you that he is rrreasuringtlre electricalresistance a coil. If you of arc astotrished,arrcl ask-him what nreaning thcse words havc, and what relatiorr they havc to the phenolrrena hc lras pcrceivecl ruci which you havc at the sarne time pcrccivcd, he will rcply that vour qtrcstion rvould requirc soulc vcry long explanations, and he will recornrnerrd that you. take a course in electricity. (p. 145)

In thesecircumstances, physicistmight make an observation the statementS, such as: 'The resistance the coil is 2.5 ohms.' of Howcver, as Duhem'sanalysis shows,S is the resultof interpreting the llature and movementsof many piecesof apparatus, using a group of complicrted physical theories. The physicist to devise has a theoreticalmodel of thc experimental apparatus perforrna series and of calculations about this rnodel. Only then can he concludefrom the movemellt of the light beam on the celluloid ruler that the resistance the coil is 2.5 ohms. The simpleobservation of statement S is highly theory-laden. Duhem goes on to observe,again quite correctly, that 'Ilre Theoretiul Interpretation Phenomena of Alone MakesPitssible Use the o.f Instruments' 153). He argues that this is true even of an (p. instrutnent simpleas a magnifyingglass, as and is still more true of a nricroscope. he says: As The objcctssccn through thc magnifying glassappcar circledby coloursof the rainbow;is it not the theory of dispersion which
tcaches to rcgardthesecoloursas creatcdby the instrurrrcnt! us end to disregard them when we describethe object obscrved?And how nruch more inrportant this remark is when it is no longer a rnatterof a simplc magnifying glassbut"of a powerful microscope! 15a) (p. So far in Part III of this book we have engaged in a lengthy discussion of the observation statenrents of science; yet this is the first time that we have had occasion to menriolr the use of instrurnents. This is somewhat curious, since virtually all observatiorls in modern scienceare made with the help of instruments.

to Irrs ct.rurc be generally accepted..'y philosophers rs u a l l v fo rrn u l a ted as t

by giving the exr-rnpleof nreai \ lr inll,tl rr' r' l t' r' ttt' :rlrc s i s tl n c co l - a coil:
(irr rr r ln t lr ir Lr lr om t or t i dr aw near t h i s t a b l e c r o w d e d r ; i t h s o m u c l r nlrlr,l t , r t lr \ '; r r r r ' lc t ' t r ic [ r ; r t t c r y , c oppc r w i r e w r a p p e d i r : s i l k , v e s s e l s lrllr'rl r v it lr t nc r ( r r r y , c oils , : r s nr all i r o n b a r c a r r y i n g : . . m i r r o r . A n
I I

,l

:I

tJ4

The Natureof Observation

ls ObservationTheory-Laden?

Horvever, our plescnt-clayrelianceon instrunrelttsUas bv no rneens alr,vays feature of science.Sciencein tl-reancient v ,rrld arrd during a the Middle Ages rvas basedentirely on naked-eye(r ,r, rather, naked human sense-organ)observations.The first really ;.gnificlnt nsc of an instrunrent for scientific observation occurred w,ren Galileo used a t c lc s c o p eto s u rv e y th c h c a v c rrs G al i l co bcgal r hi r obscrvati onsi rr . 1609, and published his first findings in'fhe Starryi.'[essenger (1610). I lis discoveries were truly rernarkable, and sbr wecl the grcat advantage of inrprovirrgl naked-eye observation b' the use of an instrumeut. Galileo r.vasable to observe rnountaii:l: on tlle ltroon with the help of his telescope.Ile could see at l(1rst ten tirnes as nrally stars as had previously bcen known, aud wa,; aLrlcto otrsctvc t hat t he M i l k y Wa y ' i s , i rr fa c t, nothi rrg but a : ongcri cs of i rrnur uc r a b l c s ta rs g ro u p c d to g e th c r i n cl ustcrs' (p. ,1)). Gal i l co al so discovered that the planet Jupiter has moons circlin;, round it. Galileo's resrrlts were acceptedby ntclst people t,r the time, but sorne of his Aristotelian optponerrts questioned whether it was right to rely on observations rnade with the tele$cope.,f:eyerabencl has argued that such objections lvere by no nleans as;bsurd and irrat ior r al a s th e y mi g h t a t fi rs t a p p e a r (1975, pp.99-1,' 4). Gal i l eo had tro well-estaLrlishcd lrody of opticnl theory with wlrich to irrtcrprct hc slr,r' through thc telcscopre, ancl he lrnrl to irakc good this '"vhat dcficicncy by r lrrgc nnlount of spcculatiorr.Thus .lllileo's ust: clf the telescopeiliustrates both the advantagesof instr.inlents and the problems of interpreting theoretically the results ths ' yield. A further consideration of the use of iustrurncnts leads to a criticisnr <rI the pltysicalinn Ncurath arrclthc latcr Carnap. l)hyoF sicalisnr, it rvill bc renrelrrbcred, is .the doctrine tlr;rt o[rservatiorr statenlerltsare about physical objects. If by 'physica object' lrere is s lr lc ant a rrra c ro -o trj c c t u c h a s a ta bl c or a chai r, tl rc cl octri nc i s certainly wrong. Observation statelllelltscan also b, about rnicropar t ic le sto o s n ra l l to b e o b s e rv e dwi th the naked ey:. C onsi der the photograph of a cloucl charnbcr shown in Platc 2. Th s is interpretet{ as follows:'A 63-rnillion-volt positron passes throu;!r a 6-rnm lead plat e an d e rrl e rg e s s a 2 3 -n ri l l i o n -vol t posi trorr.' H ei :, then, w e are a obs,:rving a positron: a nricro-particleof anti-ntatter ' rf the kind first postulated by Dirac in a high-levcl physical thcory. Iu ihe light of all this, let us corrsideragain Carn;i :'s claiur that a s s er lt erl c e u c h a s ' M r A i s n o rv e x ci ted' has tl te satt.ecol rtent as a sentencer,vhich assertsthe existcnce of a physical ri :ucture of Mr b A ' s bod y , c h a ra c tc ri z e d y c e rta i u behavi ouralsyrr":tonrs. C arnap is driven to this analysis becausehe doubts the legitirrracyof attri-

through a Gmm leadplatc and 2 Plate A 63-nrillion-volt positron passilrg emerging as a 2}-million-volt positron. From Max Botn, Atonic Physics of (1935),facing p. 47; rcproducedwith the pcrmissionof the publishers Inc. the current, Sth edition, Dover Publications,

buting subjective psychological states to Mr A. Yet, if we can interpret the photograph of a cloud chamber as showing a positron passingthrough a 6-mm lead plate, why should we not intcrprct Mr A's behaviour as showing that he is in a certain mental statc- that of

. _,-;,._rr;.;

136

T h e N a tu reo f O bservati on

Theory-Laden? ls Observation

137

F ft I

In exciterrrent? both cascswe are interpreting obscruation by ltlealls of theory. In thc positron case, the theories are t::,tse of rnodcrtr physics. In the caseo[Mr A, the theories are the tra'litional theoties or of conrnron-sense folk psychology.'fhe attributicr of exciterrreltt N t o M r A i s c o rn p l c tc l y i n te r-s u b j c c ti ve. earl y al l ouservcrs,arntcd with thc theorics of tolk psychology, will interprct Mr A's bchaviour in thc same rvay, just as nearly all obscrvers who are sufficiently well informed about modern physics will interpret tlrc cloud clrarrrbcrphotograph in thc sallle way. There is actually nc need to ask Mr A for itrtrospective reports on his psychological state- though this could be done. Let us llow return to Duhent, who clevelopshis theme by intro* dr.rcing anotltcr interesting poitrt. We often speak of elirninating causesof error itr atr experirnetrt or observation by making approprirtc corrections. These corrections, iIr Duhenr's view, consist csscntielly irr inrproviue tlte tlteoretical interpretatioti of the experirrrcnt. I'lc illustratcs this by thc cxantple of the correctionswhich the l;rcrrclrplrysicist Ilcgrrault ttracleto his Itratronretcrreadirtgs: by this r:ottlcl rcprcscrrt realInanometer an idezlone, formed llclirrrLrlt evcrywhere fluid ol' lrr irrr:orrrprcssible having the sametemperature rt ;rrrtlsrrlrjcctctl cvcry point oI its free surfaceto an atmospheric of rc Itr('ssu irrrlcpcntlcrrt thc height; between this uversimplified and lc s t lr c r r rt' ;rrrtl;rl i tyth c rc w o u l d b e to o greata di screpancy conThen l tc y tl \ r ' ( lu( ' rrtl y , rt' c x p t:t" i rrt<w otu l d b c i nsuffi ci entl preci se. :n ('rvrs ;r rrcrvirlc;tlnlattonlctcr, ( ()lr( tharrthc first, rrrorccotnplicated th otre;he forrl s thi s trew lt r t r ( ' l )r(' \(' l l i n l ')(' ttc r c rc l l a l l d c ol )crcte , fltrid and allows thc tclrrperaturc rr,rr()rr('t('rrvitlr ;r cotttprt:ssiblc hc ln \ , , n y l ro n r o u r' l )o i u t tro :trto tl tc r; al so al l ow s tl * : barometri c ro rv I r r { ' 11 u rt' r l r,rrrl ' ,c l rc no ttc g o c s hi ghcr up i n th: atmosphere. cotrst,,l ,l te nl any so scl o A ll t lrl rc rt' t' trrl ri trl l s { l l rc p ri rrri ti vc tcrtrc
rri tttrl l (r tl\ (| l!r /)

F F F F F I
I ||

retical interpretation by means of a striking analogy: 'The physicist of who cornplicatesthe theoreticalrepresentation dre obscrved facts by corrections, in order to permit this representatiollto come to closer grips with reality, is similar to the artist rvho, after finishing the line sketch of a drawing, adds slradingitr order to expressbctter on a plane surfacethe profile of tlte modcl' (p. 158)' As usual, Duhem lirrrits his discussionto physics, but it is clear just as much that his analysisof observatiolras theory-laderrapplies to the othcr scienccsas to physics. Let us cotlsideragain Alexalrder Fleming's protocol: 'It was noticed that around a large colony of a contaminating mould the staphylococctls colonies became transparerlt and were obviously undergoing lysis' (1929, p. 226). I-lere various blobs which would convey rrothing to thc uninstructed lay person (see Plate 1) are interpreted in terms of bacteriological theories involvirrg concepts such as 'staphylococctts colonies' and 'lysis'. Moreover, instruments are now useclin all thc sciences. We need only think, for example, of the use of the eiectron microscope in biology. T'hus Duhctn's points could bc illustratcd just as well by examples frorn chemistry or biology as by his orvn exalnples from physics.

Thesis of 7.2 A Reinforcement the Holistic Principle and Neurath's

The view that all observatiolr in scienceis theory-laden reinforccs the holistic thesiswhich was formulated in 5.4' We there considcred the caseof a hypothesis, H1, which could not be refuted by observatiorr when taken in isolation, but only whetr taken as part of a wher e G = t l{1, I 12, . . . conj unct ion of a gr oup, G , of hypot heses, I-I,,) say. Now supposeG is refuted by an obscrvation statcmetrt,O. by This statement, O, is established the interpretation of scnsations in terms of a further group of hypotheses, G', where (l' : {Kr, Kz, . . . K,), say. Thus, to test H1, we need not only the hypotheses H2 . . . H,,, but also the hypotheses Kr . . . K,. Itl this serrsethe holistic thesis is reinforced. To put dre same point attothcr way, in the cvent of the refutation of thc group G by O, a scientisthas, itr addition to the option of changing onc or more of the hypotheses in G, the optiorl of querying one of the assutnptionsin G'irl sucir a way that O is rejected and perhaps replaced by an observatiotr statcn-rcnt,O', which is conrpatible with G. There are nrany irl-

( lni r ' , r l1 ,rrrr, rrttr i i (' (' rl l s() [rc ri g l r t on tl ti s poi nt. ' i -hc correcti otts l | )ttl ttttl o tl ttr' t' tl o b ta i t t bcttcr resul tsfi om hi s ttratrotrl r r ' lr r r l{ ,.1 i ,rr.tttl t lr ;rl r t r \ v rr r' l r,rrtt l y i tttPto v (' l l l c tttsi rr thc thcorcti cali nterpretati on ' ilr r = r , l lr r ' , r ' r1 ,r'ttttt' ttl ;tl :rl )l );tr:tttts'.l ' l r rrs, i rrstcad of regardi ng the 1 t ir r ' r nr \ , .r,, .u r rn (()n rl rrc s s i b l cl l tri t l , hc i rl trocl ucctl a theorcti cal Il l 1r o, lr ' l rr' l rtr r to o l , ;trr' rtttl l t o [' i ts t ' orrrl rrcssi l l i l i ty. l steaclof rcof aS g, , , r r , lirr1 ,. ,tl ttto s l rl rt' ri tIrt' s s tttt' rttrtl tc l l l C rC ttry rrl cl C pcndctrt tl i r' tttotl cl , cl ttcto l -aphcc, w l ri ch rt l tr.rrrtro .l rtrr.rl l l rc o t' c ti r'rtl r t l lr r , r ll ,l rt, w t ' nh, , t rt rrrrrrlo l l l tt' v ;tti ;tti o l t o l l r:r rotttt' tri c )rcssl l r(: i tl r l l ci gl rt; l of l l { , r ! r l ' , r r l n | )rrl r,,rrrl l rr,,tr;rl t'tl ri r l r rttr' t' ss i rrrP rrl vi rrl (l rc tl rt' o-

138

The Natureo{ Observation

ls Observation Theory-Laden?

139

stalrces i' tlrc history of science wherc this seco.rd strategy was followed rvith grcat succcss. In 6.2 I quotcd Nerrrath's farrrousconrparison b:tween scientists and s a i l o rs w h o l ra v e to re b u i l d thci r sl ri p at sea::rahrel y,' v/e arc like sailors who nrr.rst rebLrildthcir ship o' the opir se3,-neverable to disnrantleit in dry-dock and to rcconstruct it drerc out of dre bcst rrurtc'rials'(1932/-1, p. 201). This passageconres lionr Ncurarh's article on protocol staterrreuts,and this is no acrident, since the :uralogy is essentially -justified by the point that ail observatiou is theury-laden. To see this, let us first srate the colrt(irt of Neuratb's conrparison explicitly as a principle which l w.,l call Neurath's prirttiple. principle is a conjunction of rwo pa,r;, (A) auc{ (B), .Neurath's rvhich nray bc statcd as follows: (A) In o rc l e r to tc s t a n y s c i c n ti fi c statcnl cnt, w e ' rave to assunrc for the time beirrg some other scientific st.tements. (This correspondsin the analogy to the fact that we ilrl 0uly rerirove o'e plarrk of thc ship if we leavc others in place, si'cc otherr.viscthe ship would sink.) T h c re i s , h o w e v e r, n o s c i c,ti fi c statel ' e' t v;l ri ch ca' ' ot bc subjcctcd to tcstirg, and pcrhaps aba'd<l'ed a: a result of the tests. (-fhis corresponds in the arralogy to th:. fact that any plarrk of thc slritrrcr' lre renroved and chcckr: to sce if it is ,! rottut.)

well. IJut we now coule to a point where Duhem seernsto r,'reto be mistaken. Duhem argues that his view of observation as rheoryladen applies to physics, but not to the observation statements of ordinary life. In the rrext section, however, I will present some findings of empirical psychology which suggest that everyday obscrvation statements are just as theory-laden as those of science. Thc only difiterence is that in everyday life the theories applied, usually unconsciously, are co[lfiron*sense theories shared by all, rather than the high-level theories of a particular branch of science. llut let us first examine l)uhehr's statenent of the opposite point of view, He writes: An experimentin physicsb"ing quit. anothermatter than thc mere ob ser vat ion a f act . - . of When a sincerewitness,sound enoughin mind not to confusethe play of his imagination with pcrceptions, knowing the language and he uses well enough to expresshis thought clearly, says he has observeda fact, the fact is certain:ifl declare you that on suchand to such a day at such and such an hour I saw a white horsein a certain street, unlessyou have reasonsto considerme a liar or subject to hallucinations, you ought to bclievethat on that day, at thar hour, and in that street,thcrc was a whitc horse.(1904-5, pp. 158-9) In lcgal trials, it is of course important to ascertain whether the observation statenlents of wiuresses are true. The main technique is to cross-examine the witness, and, correspondirrgly, it is often fcarccl that thc witlress mey be lying. Such fears, accordirrg to l)uheur, are lessjustificd in the caseof physicists, whom hc rcgards as gcnerally trustworthy. Morcover: Aftcr submitting the physicist'stestilnony to the rulcs detcrmining the credibility a witness's of storyt we shallhavedoneonly a pa;1,thc easiest part at that, of the criticism which should determinethe valuc of his cxperiment. . . . we nrust inquire very carefully into the theorieswhich the physicistregardsas established which he usedin interprctingthe and hcts he has observed.(p. 159)

(B)

it is casy tb scc thar both (A) ard (13)follow fror;r rhe view trrar all ob s c rv a ti o ' i s th c o ry -l a d c n . As for (A ), supposc.r' c arc tcsti ug a particular scieutific staternentS, we have to compar(r S with at least one obscrvation statclncltt, O, and to accept O we treed to accepr the group of theories (G', say) with rvhich O is'lader.'. Thus, to tesr s, we need for thc tirrre being to accept the scientifi,;statenrerlts of G'. As fbr (B), it follovrs fronr thc vicw rhat all observarion is theory-laden that no observation statemerlt o is inc.rrrigible, since the theoretical i'tcrprctation involved in o ca' alwl ys be queried. Moreover, such a .query can al..vayslead to a furt,,r.er tesf of O. Thr:s, any observatiorr statenlcnt carr be tcstcd and perhaps abarrdonccl. 'I'hc sarrrc cven urore obviously true of the nrln-observation is staterrlents science,and thus (I)) follows of So far, tlrerr, I have argued in favour of Duhem'; view that all qbserv?tion in qhvsics is rheory-laden. and irrdeed, | ,ue suggested that this should be cxte.ded to other lrrarrclrt',of scieiice as 'iew

There is undoubtedlysorrrctruth in what Duhem says here. In legal cases,the main worry is usually whether witnesscsare lying. ln experimentalohysics.thg main worry is usually whether app; p;at.l[eo ries h a;e-be; Howevcr, the two cases not quire so sharplydistinct as Duhern are

1::iri:i:.,i:i;1' '';. ' ,'l::l:i:.

14O

The Natureof Observation

141

claims, sirrcethe observationsof everyday life do sr:enlto be theoryin laden. For examplc, a theoreticalintcrpretation is rrr,eded ordc'r to rccclgnizca movitrg patch of colour as a white horsc. This clairrr is supported by Popper, whcl takes as atr cxanrple the statetrrcnt'l-lere is a gl a s so f w a te r' : (or synrbols,rir idcas); evcry naurcs uscs rutiversal Evcry descripti<ln The stateof has the character a theory, of a hypothesis. statcrnent ment, 'Herc is a glassof watcr' cannot bc vcrificd by any obscrvawhich appcar tiorral expericnce.The reason is that thc uniuersals (An in it cannot bc correlatcdwith any spccificsetrse-expcriencc. 'inrnrcdiatc givclr';it is unicluc.) is experictrcc' only onee'immediately By the word 'glass',for example,we denotephysicalbodicswhich holds for thc worcl and behaviour, the sarne cxhibit a ccrtainlaw-like thcy 'watcr'. Urrivcrsals to of canlrotbc rcdtrccd classcs cxlreriorccs; , . c an n o tb c c o n s ti tu tc d(1 9 3 4 p p . 9 4-5) I)opper is correct here. Suppose, as so often happens in detective stories, that the visitor drairrs what seetnsto be an itrnocttous glass of water ancl falls dcad irnrncdiately. W'hat has been itlterpreted as wat c r w a s , i n re a l i ty , a d i l u te s o l uti on of cyani de.

7 .3

So me Ps y c h ol ogi cal Fi ndi ngs

' [ ' lr c t h c s i s th a t a l l o b s c rv a ti o rr s theory-l adctri s strongl y supportcd i by s o n rc o l - tl tc i ttv c s ti g a ti o n so f enrpi ri cal psychol ogi sts.l ttdccd, t lr is p o i rrt o f v i c w [te s a l o trg h i s t ory i rr erupi ri calpsychol ogy, anc{ llichlrd (ircgory traccs ideas of this sort lrack to Hernranlt votr on Physiological Optics (1856-67), in which Ilt'f rrrlroltz's T'rcntise tl ra t p c rc e p ti ons are deri vcd by unconsci ous lk ' lnr l ro l tz a rg u e s am{iot"n scttsory signals.r Lct t-tsstart by considerir.rg irrfcrcrrr:cs billrorrs ligurcs; irr particular, tlte trvo lllost fanrous such figures: the Nc c k c r r:ttb c .a ttd tl rc tl u c k -ra b b i t. ( lr r rrs i tl c rth c d ra w i n g o f a c u be i n fi gure 7.1. l t i s possi bl e to i 1r t ' r cc i v ct i n tw () rl i l fb rc rttw a y s . We catt ci ther sce the face trrarketl r v it lr : r' s ;rs i tr l i o rrt l n c l th a t n ra rked w i th b' s as behi rl d, or vi cc v c r s ;r. Wi tl r a l i ttl c l )ra c ti c e i t i s possi bl e to sw i tch at w i l l frorl r s c t ' in 1 tl rc c rrl rci n o n c w e y to s e e i ngi t i n tl te othcr. 1 ' ' l lri s p h c rro rrtc rrow a s fi rs t d e s cri bed a S w i ss crystal l ogl apher, by tt I . A . N t' tk c r, i rr l tt3 2 . F tc n o tc d the effect w hi l e tryi ng to draw a l t r y r 1,1rv l ri t' l rl rc w ;ts v i c w i n g th ro ugh a nti croscope.

Figurc7.2 The duck-rabbit

The drawing in figure 7.2 can be seen either as a duck or as a rabbit. In the duck interpretation, the two large protuberances are the duck's bill, and the duck is facing left. In the rabbit interpretation, the two large protuberancesare the rabbit's ears, and the rabbit is facing right. Once again, it is possible,with a little practice, to switch at will from seeing the drawing as a duck to seeirrgit as a rabbit. and vice versa. These ambiguous figures give strong support to the view that all observation is theory-laderr.When we look at the duck-rabbit, we do not simply register curved lines and a dot. We see these curved lines either as a duck or as a rabbit. '!7e are providing an active

._

,_.-,,,i,;-ill;

-.t;aa:,:,4'

142

The Natureof Observation

ls Observation Theory-LadenT

143

theoretical intcrpretation of the phenot:retron. TL,is interpretation occurs unconsciously, before we become consciouily aware of the interpretations drawing, and, in this case, two different theoreticr:.I a rc p o s s i b l e .T h e l i n e s c a n b c brought under the c:rtrccpt' duck' or under the concept 'rabbit'. Exactly the satrreconsidi:rationsapply t<> the Necker cube. Ambiguous figures like thc Necker cube and the luck-rabbit have been much discussed (and rightly so) by philosoi:,hersas well as psychologists. This is largely due to the influence of Wittgenstein wlro nrentions the Necker cube in his 1921 Tractatus(5.5423) and the (1 Inuestigatiotrs , xi). Following duck-rabbit in his 1953 Philosophical on fronr Wittgerrstein's discussion, Russell Hansrtn discussesthe ambiguous figures in relation to the problem of o:servation stateof nlents irr chapter 1 of his 1958 Patt.erns Discovery. We have used the ambiguous figures as an argunr:'rrtfor thc claitn that all observation is theory-laden. Horvever, ther: is an interesting to ob_iection usitrg trot just thc viewirtg of ambiguous figures, but the vicrving c"'fany figure or picturc, as a nrodel ftr: our perccption of ordinary objects. The point is put forcibly by he psychologist Richard Gregory in the following passagesfrotr his book T/re Intelligent Ele: 'Pictures are in sorne ways highly ar :ficial inputs for the eye. Although we can learn a lot about percepticn from pictures, and they are certainly convenient for providing st.mulus patterns, they are a very special kind of object which can gi ;e quite atypical r e s u l ts ' (1 9 7 0 ,p . 1 8 ). And again: ar Pictures havea doublercality. Drawings.paintings d photographs are objectsin their orvn right - pattemson a flat sh::t - and at the to salnetime entirclydiffcrentobjects the eye.We set both a pattcrn o, of marks on prper, with shading,brush-strokes photographic 'grain', ancl at thc sarnctinrc we sec that thesecolt"posc facc, a a arc Irouscor a ship on a stornrysca.Pictures uniquc .lnrongobjccts; and as some o her thing, enfor they are secrrboth as themselves tircly .differentfrom the paper or canvasof thc pictu;e. Picturesare paradoxes. at No olycrlcarrbe in two places thc sametitne;no ribjectcanlie in spacc.Yet picturesare b.tth visibly flat two- arrd thrce-dirnensional Tltey are a ccrtainsize,yet alsothe sizeof a arrdthrce-dirncnsional, (p. are or a houseor a ship. Pictttres impossible. .'2) To avoid this difficulty, it would be as welJ to give some exarnplesinvolving the perceptiolrjust of three-dirensional objects
I i

viewingpoinl

Fi,gure Planof the Amesroom 7.3

which lcud supportto the thesii that ordinaryeveryday observarion is theory-laden.ln fact, Gregory (1970, 1981, and elsewhere) describes quite a number of such examples. !7e will nrentiorltwo which are particularly striking and interesting.They are rhc Ames room and the insideof a mask. The Ames room was constructed Adelbert Ames, a man who by startedas a painter and then turned to creatingvisual illusions. Figure 7.3 shows the true shapeof the room as seenfrom above. One sideof the room is much longerthanthe other,but the room is designed, accordingto the principlesof perspecrive, that from the so viewing point, it preselltsto the eye the same irnpression a as nornral rcctangular room. No'*' two peopleof approximatelythe sarnesizeare placedirr the room in the cornersindicatedin the diagram. What thc spectaror sces from the viewing point is shown in Platc3. The Anrcsroolrris pcrceived being of normal shape, as while one of thc irvo sinrilarly sizedpeopleis seenas very much biggerthan the other. It is easyto understand what is happening here.The brain hasthe choice of irrterpretingthe visual input .accordingto cither .rf the following two theories: Tr: The two peopleareapproximately samesize,but one the is rnuchfurtheraway because room is an odd shape. the Tz: The room is the usual rectangular shape,and the people aredifferentsizes.

'l.iili;i't:t

The Natureof Observation

t'

ls Observation Theory-Laden 7

145

T
T F
F
T
;

i i.

{ I \

F
F F
T

I>late 3 Thc Ames roorn. From Richard Gregory, The Intelligent Eye of (1970), p. 2(r; rcproduceil with thc pcrmissiorr both the author and the prrblishcrs,George Weidenfeld and Nicolson.

Plate 4 On the left the inside of a mask of the face on the right. Photograph by Richard Gregory, reproduced with his permissiorr:

F tr J-r

Flcre, 'f1 is correct, and T2 itrcorrect. Yet anyoire, on seeitrg the Anrcs roottt lor the first titrte, r,vill urrcollsciouslyopt for T2. T'his is pcrhaps not surprising. Experience has tnade us all very fanliliar r,r'ith thc fact tlurt roonls arc ttearly always rectattgttlarin slrape, ofltertvary ilt size. Thus, or1the basisof experience,T2 while peop.rle is bcttcr corrfirnred thatr T1, although T'1 is in fact correct. Illterestingly enough, if the roorn is explored witlr a long stick, it will gridually conre to look its true queer shape. Thus further infornraits iiot, .rr.r alter the way irr which the brain unconsciouslyprocesses v is ual i n p u t. f'hc Anres roolrr shows very clearly that ordinary everyday 'We are contillually interpretilrg our observatiop is tlreory-laden. sensory experienr:c in ternls of cotntrron-sensetheories. Solre of these theorics ln:lv irrdeed be itlrate, but others have certainly been that roonls are very likely learnt by expericttcc.'Ihe gerreralisation in shape is sonrethirrg that could only havc bcclr tq bc rcctatrgular le;rgrt by expericrtce. It is very irnplausible (to say tfie least) that s uc h a g c n e ra l i s a ti o nw a s g e n e ti cal l y programnrl )d i nto pri rni ti ve l nlol, w h o n ra y l tl v c l i v e c l i n tl rc j tr ngl c rathcr tl -ran touscs.

In Plate4 the object on the left is the insideof a rnaskof the face on the right.2 Although the hollow mask is uiewetl it from the inside, is seenas a face sticking outwards towards the spectator. This is bccausc experience stronglyconfirmedfor us all that noscs has stick out of facesrather than into drem. We automatically interpretthe mask in terms of this well-confirmed hypothesis,unless there is such very strong depth information that the farrriliarhypothesisis over-ridderr,and a less familiar one put in its place. As Crcgory says: inwards. we Iook If about noses sticking There nothing is impossiblc
at the back of a nrask, close-to in a good light, wc sce it as a hollow face. There is sufficierrt depth-information to overcome the improbability of a face being hollow. But when the mask is vicwcd in a poor light, or an even light casting no shadow or tcxture, it gcnerally appears as a normal fece, though in fact it is inside out. lt takes the full power of stercoscopic' and texture depth-information from the eyes to convince the visual brain to reject its usual perceptual hypothcsis. .. that it is a face with the nosc sticking out. (1970,

p. 128)

l
140

The Natureof Observation

vs
$ i $ I -) $ '" S S $ \ S $ 5 u{ $ t f,\$ ."" I $ { h t\ I N d q "t \ \T N $ N{ > N \ t ) t I { "$ -1

t1

ls Observation Theory-Laden?

i47

r:rrat clearlydetnonstrates ordinary everyOnce again,this exatrrple is in fact theory-laden. day observation

7 .4

S o m e Ge n eral C oncl usi t:tts

:: l i

The discussion of an observation statements ltas been lorlg and corrrplicated,arrclit is thcrefore worthwlrilc tryi :rg to sttllrnrarizca fcw of the conclusionsrcached. Let rne begin, th:n, by attetnptillg a definition of an observation statemcnt. The p: eceding discussion suggests the following: an observation statem:nt is a statetnent which is the result of some sensory ilrput in:erpreted, whether consciously or unconsciously, using a set of th:ories. lnstrtttneuts may be - indeed, usually are itr present-day science - used to produce sensory inputs which .arould otherwise lrot have occurred. In that case,theories of the insqq4rent are part cl:'the sct of thcories r iic d tri i u tc rp rc t th c s c n s o ry i rrp u t. The use of ever ttrore complicated instrulnents call seenl to reduce the importance of hunran serrsory input; but tl'is, in fact, always rerrrains inelirninable. Suppose, for' example, ltat a conlplicated ilrstnrtnctrt is cottstrttctedto olrscrve Sotllc proc.:tS. Tltc itrstrttlttcttt is connected to a conlputcr, and tire ltunran o{ server has orrly to read the conrputer print-out r.vhichstatesthe rer rtlt of the obscrvation. Even in this cxtrerne case,the human obst ver has to rcacltlte conrputer print-out, and this involves hinr o: her havitrg sonre sensory input. Moreover, this input has to be ir :erpreteditr accordance with a set of theories concerning the proct ss, the instrunlcut, and the computer. So there is a necessarysui:iective or psychological element in observation; but this is not il:conlpatible, as we shall see in a tnoment, with observation statemelltshaving an illtersubjective character. Th,: Tlris brings us to the topic ofp.syrholo.qinn. origirral versiorrof theory was that obscrvation statenlelltsrec(unt the irtrlttcc{iatc this sensory experiencesof a particrtlar irtdividual, I nd are complctely This kind of psyclologism was quitc verified by these experietrccs. corrcctly criticizcd by Ncuradr and thc later iartrap. Ibefg-il

nopper, in favour of weak psychologism:Let S be an observatiorr statement,and E the correspondingsensoryexperience. There is nothing absurd in claiming that E partially justifies or partially verifies S. Indeed, this seemsto me correct. Of course,the justification or verificationis only partial,sinceS involves the theoretical interpretation of E, and this interpretationmay well be (indeed, often is) mistaken. As far as Fries'strilemma is concerned, havethereforeopted for I psychologisnr its weak form. But the question irr now arises to as horv, in the light of this, it is possibleto defend the inter-subjective characterof observationstatements. Neurath and the later Carnap defendedinter-subjectivism. the form of physicalism the view in -. th"t observationstatements must be about physicalobjects.However, I argued that this was too narrow a fiamework. Observation statemelrts could be about micro-particles(for example,positrons) (for or psychological states example, anger) well asaboutordinary, as macro-, physicalobjects.For positrorrs,the sensoryinput hasto bc intcrpreted termsof the advanced in theories physics. of For anger, the sensoryinput has too be interpretedin terms of the commonsensctheoricsof folk psychology.However, there is no ciifference in prirrciple between the two cases.Indeed, I argued at length, usirrgexamplcsfrom empirical psychology,that ordinary everyday observation is just as theory-laden as observationin one of the advanced natural sciences.

ience. l'opper nraintairicd, statenlent,-not bv a psychological experience. H:re, I sided against

sensory.input, on which the observation I, statement, is bascd. S, So thesesensoryinputs rlust have a straightforwardcharacter, and not be arcane any way. One could not usemysticalexperiences in which are accessible only to privileged individuals. One consequencc is that a given observation statement not associated is with a single sensoryinput, I, but rather with a potentially infinite set of such sensoryinputs, which could be thoseof different individualsor of the sameindividual at different times. A secondrequirement that is the set of theoriesusedin the interpretationof I should be accepted by all members the community.and that the applicatiorr these of of
a

si!

l4t J

T h e N a tu reo f Observati on

ls Observation Theory-Laden?

149

t lr t ' o ri c sto tl rc c a s ci rr h a n d s h o u l d have a standardrouti ne characrer wlr it l r r:rl rrl tl c c a rri e c lo L l t, o r a t l east checked.by any mernber of b I lt t ' r ' o rrr ttrrri ty . n ' l' lrt' l o rc 1 1 o i rr1 g l y s i s o f' o [ rso' vl ti ol t stal .cnrcntshas n]:ul y a rrn li':rtrtrt'sirr colrtrrrorrwith l(ant's analysis of pcrccpti<.ur (scc I(ant, l' / 81 1 7 ). I(:rn t b c l i c v e c lth a t c o n s ci ouspercepti orlconsi stsofj ucl gcr r r c r r tsrv h i c h a re fb rn re d b y b ri ngi rrg i ntui ti ons under concepts. S ir r r i l l rl y , I a n a l y s eo b s e rv a ti o nstatementsas formed by gi vi ng to s('nsory inptrt (which corresponds to Kant's intuition) a theoretical irrtcrprctation (wltich corresponclsto Kant's 'bringing under conrcpts'). ()f-course, there are differences. Kant's intuitions had two prrrc firrrrrs:spacear-rd tirne, whereas I do not postulate that sellsory irrprrts :rre characterizcdby arly pure forms. Kant's concepts had t o ir rc l trd ca t l c a s t o n e o f h i s tl v cl ve pure conccpts of the underst;rrrtlirrg, categories,whereas I do not claim any such lirnitation or ( ) n t l rc : c h a ra c tc r o f th c th c o re ti cal i ntcrpretati ons. Indecc{ , tl rc tlrt:orics involvcd carr be of tlrc rnost varicd character, fioln the t lt c ori c s o f c o rrrn ro n s e n s e to th ose of an advanced ruathentati cal s s t ' ic rrc c rrc h:rsp h y s i c s . 'l'lrc rclationship between an observation statement, S, and the c r lr r c s p o n d i n gs e n s o ryi n p u t, I, i s a very probl emati cone, and rai scs r r t ; r r ryd c c p p ro b l e rn s c o n c c rn e d w i th the nature of corrsci ousness : r r r r l tl rc rrri rrc l -b o d y p ro b l e m. W e cannot pursue these i ssuesi n t lc t r r i ll rc rc , b u t i t i s o n l y fa i r to i n di catesome of the di ffi cul ti es. ' l- ., t," e i n * i tl r. i t * c onv i o rrs c x p e ri c u c e l ta s a l rc a d L i nvol ved sorne theo.eti cpl i nter'l'hus. ptrrc scusory inpnt has to be considered as sornc11t'tlrtion. ()l -i l c o n s tru c t. o r D c rh a D s s occurri rtg ul rcorrsci ousl v,f r,vc a t lr inu l lt lopt a p h y s i c a l i s ts ta n d p o i n t, w e can thi nk of the sensoryi nput as w:rvc's or particlcs * for exarnple, photons - irnpinging on the I r r r r r rl n s c l l s o ry a p p :rra tu s . T ' h i s i rrput must:r!rcady have becrr corrsitlcrably proccsscd be[ore it reaches consciousncss,arrd the t r ; r r r s(i rrrn ;rti ofro n r c l e c tri c a le x c i tati onsi n the l l rai n to consci ous rl i 'c c x pl' rl 1 ' 1 1 1ts c l I ra i s c sth e rv h o l e rni nd-body proi :l enr. Thi s i s as far ; r sI r v i l l l rrrrs rrc i s a rra l y s i sa n d I w i l l now consi deranother aspect , th ol t lr t ' Itro b l c ru o f < l b s c rv a ti o rr tatements. s Arry givcrr otrscrvation statclrrelltcan always be further checked : r r r t ltc s tt.' tl . o F ri c s i s q u i te ri g h t to sce here the possi bi l i ty of an S ir r lir ri tc rc [{ rc s s .I.l o w c v e r, s c i e n ti stsare practi cai peopl e, and ttre r'('llrcsslr;rs to lrc crrt short by the 'scientific j',rry' deciding at a r ' t ' r ' t ; ri rr< l i rrtto a c c c p ta p a rti c u l arobscrvati onsti l tement.P opper i s p r ililr t to s :ry tl l a t th i s d e c i s i o n i s partl y conventi onal ;but I w oul d

i i

I !
1'

li:

also stressthat it is partly justified (in most cases) rhe sensory by experiences at leastsoqreof the jury. A decisionto acceptan of observation statement never final and irrevocable. is observation st:rtc.nc'ts;rrc always corrigible,and Popper is right to sce thc edificeof scie'ce as erected-on piles dri"en into a i*anrp, r"ther tlrarron bedrock. Nore the less the accepta'cc pro tem of sonrc obscrvationstatenlents data is necessary the scientificenteras for priseto go forward. I will concludethis chapter'*'ith an a'alogy which is co'srruced by,- so to speak, reversingthe direction of gravity in popper's analogy. Popper imagines driving down piles to preveni- the weighty edificeof science from sinking into a swamp. Let us instead conceive scientific of theories hydrogenballoons as with a tendency to escapefrom reality (the earth) into the airy regiorrsof nletaphysics.rThese hydrogenballoonsare attached tti. earthnor by io large c-ablcs, by a multitucleof fi'e threadsarrd thin strings, but rathcrlike thosewhich held Gullivercaprive whe' he first awokeirr Lillipur. Eachfi,e threadis a prorocolof the fornr 'Mr A observcd that o'. so each thread represellts the sensoryexperiences a of particular individualinrerpreted the light of a setof theories. in Thc thi' strirrgsare formed by twisting togcthera few of the threads. They rcpresent irnpersonal observation statements, whicl-,are O, forrnedon the basis individualprorocols of suchas 'Mr A observed that O'. 'Ms B observed that O', and so on, but whictr are nlore cer.tai'than any of the protocols which they arebased, asthe on just strings are stronger than the threadswhich composethem. Any thrcador s-tring may snapor be cut, but althougltthir ,rrry alterthe positiouof the balloon,it will still remainattached earthbecause to of thc nrultitude of other threadsand strings. If we cut all the threadsand stri'gs, however, our theoreticll balloon will floar away from the rcality of the earth towards the airy rep;iorrs of mctaphysics. our scientific theory will havebecome metiphysical a theory. But what exactlyis metaphysics? How doesit differ'from sciencc, how doesit relateto science? and Thesearequestions which we shallexarnine Part IV of the book. in

i] L i {a

g
F

&,

fr
aL .Ih:

Fart lV TheDem'arcation between Science H andMetaphysics F

E i

-t'

,,1

B
l s Met aphy s ic s Meaningles s ? Wittgenstein, Vienna the Circlo, Popper's ond Critique
8.1 Introduction: Demarcation The Problem and its lmportance
Wc now turrl to the fourth theme, which concerns fundamental a problem in the philosophy of science; that of demarcatingscientifc tlreories from other sorts of theories, particularly tnetaphysical theories.All four themesare of the greatest imporrance loi the analysis science, the demarcation of but problemis, as we shallsee, the therne which irnpinges.most ol't generalintellectualconcerns outside the realm of science itself. Although the dentarcation problenrhasbeenmuch discussed the twent[th ctlntury,it is not in new to this century,but was considered Hunre and Kant in the by eighteenthcentury. As far as the eighteenthcentury is concerned, the problem was posed by the great successes the scierrtificrevolutiotr and of of Newtonianphysics.It seemed most eighteendr-century to thinkers tlrat Newton's theory was a new type of theory, a scientifc theory, supcriorin kind to previous theories. the samctirrre,rcligionwas At for thc first time in WesternEurope coming unclerheavy attack, partly, no doubt, because the disillusionment of brought about by the wars of religion of the sixteenthand sevcnteenth centuries. The contrast,then, was betweenscience, considered a sound form of as krrowledge, and religious heliefs, whose clairn to be knowlcdge was nrore dubious. Hume, an enemy of religiorr, statedin the famous Iastparagraphof his Enquiryconceming Human L)nderstanding the that volrrnres 'divinity or school metaphysics' of contairr'nothing but sophistryand illusion' (1748,p. 165). I(ant had a rlore syrnpathetic attitude towards religion. He

ii

$ [i
$

E $

fi H
e

154

'' Scienceand Metaphysics

ls Metaphysics Meaningless?

1b5

demarcated rcligion front science, but tlrought tl'.at, although religion rvas diffi'rcnt frorrr science,it could still be jurxifi"4 up to a point . 'I'hesc rcligious problcrTrsare by no trtealls dcad, ls the rccettt creatiollist corrtroversy in the USA indicates. Educit.ed courltrolt scnsehas, for a lotrc titrre, rcgarclcdDarwiu's tlrcory cf evolutiort as a scientiflc theory, and the accourlt of the origin of :ire speciesiu Gerresis as a religious nlyth. The creatiotristshave attacked this vicwpoint, clainring that thc Genesisaccount is a thecry quitc ort a par with the Darwiniatt theory, arrd that both viet,vs sh<luld bc taught at schools. Can r,ve really say that the Dan'inian theory is scientific, and hence superior to the nott-scicntif,c account itr A Gerresis? prelirninary study of the dentarcation ptnbienr is ttcces s ar yit r o rd e r to a n s w e r th i s q u e s ti o n. So far, we have cotrsideredthe issue of science vi:rsus religion, but anothcr important question regarding the denrarc;.tionprobleltr concerlls science versus pseudo-science.There exist certain irttellectual activities which are claitned by their proponetrts to be su,-;lt physics, as scienceson a par with thc standard tratural sciettces chemistry, and so on, but which are denottnced by tlicir detractors The earliestexample of such :rn activity is as mere pseudo-scietrces. astrology. Critics of astrology maintaitted that where,ts astrononry was a genuinc science,astrology was atr illusory pseu,jo-scietrce. in the twen

a pieccof mere wish-fulfihnenton the part of certainntalcontents who are disgrurrtled with crpitalist society.Sinrilarly, oppotrcnts of psych<lanalysis Freudiantheory as a tissueof myths, retherthan see a geuuincscierrce. Curiously, one of the principal recentopponents (Eysenck)has tried to rehabilitate of psychoanalysis astrology as a scicrrcc. Although psychoarralysis Marxism are rhe two principal and activities whose status, whether scienceor pseudo-science, disis puted, thereare others- for example,intelligence testing.Evansand Waites,in their 1981book-lQ andMental Testittg, claim that thereis ruogenuinescientificfoundatiorrfor intelligencetests,and that the thcories about IQ and heredity collstitute an turutaturalscience'. Naturally, the advocates IQ testswould not agree.The issueis of clearlyone of considerable practicalimportance,because concerns it thc question of whether IQ tests should be used for selectionin schools. It is clear that no progressin these disputesis possiblc without a carefulinvestigationof the demarcation problem, whosc importanceis thus clearlydemonstrated.
1 tinquishing between science and metaohvsics. In this however, metaphysics must be taken in of

influentialbodiesof t.heor whose . The status is dis continued to rise irr tlre ninetecnth sciences of the ..natural [Eitig.
t

l"nt,tiy with advances irr physics, chelttistry, and bi':logy' Freud and and Marx shared the generalenthusiasntfor the traturalsciences, both of thenr aspired to extend sciellceinto new area;. Freud saw hinrself as the fourtder of a new scienceof psycholog'r, Marx of a new science of society. Freud compared his discoveries to those of Coper nic u s a n d D a rw i n (1 9 1 7 , p .3 5 1 ). E ngel s i n l ri s speech at the graveside of Karl Marx itr 1883 also made a comparison with Darwin. He said: Just as Darwin discovered the lan of development of organic nature, so Marx discoveredthe law of development of lrunran history' $. a2\. Engels, then, considered Iv rrxisnr to be a scientific analysisof society on which a prediction of .ae inevitable downfall of capitalisrn could be securely based. The >pponentsof Malxism see the matter rath6r differently. For them' 'Marxism is a pseudo-science,rather than a geuuitte science; whilt the famous prediction of the downfall of capitalisln has no scielttificbasis,and is

is to say, the generaltheoriesproduced by philosophers,such as Plato'stheory of ldeas,Leibniz's Monadology,and Hegel'saccount of the Absolute. Philosophers the pasr have certainly prodrrced in such theories,but should they have dorreso?Do such theorieshave any value at all? Or would it be better if intellectuals concentrated - ttii pfbdritin["theorieswhich were scientific? We will examinethese questions well in the course-of discussiqn the demarcation as our of problern. Wittgensteinin his Ss far as the twentieth century is concerned, 7921 Tractalus and the Vienna Circle concentrated more on distinguishingscience from metaphysics the narrow sense in and on the science versusreligion aspectof the problem. Popper, on the other hand, devoted nrorc attention to the question of scienceversus pseudo-science. This is not to say that Popper ignored the other aspects the problem. Indeed,De Oliveira hasarguedplausiblyin of his 1978articleon 'Popper'sTwo Problemsof Demarcation'that in his 1934 Logic of ScientifuDiscouery,Popper, who was closely involvedwith the ViennaCircleet that time, concentrates more on

lut)

Scienceand Metaphysics

ls Metaphysics Meaningless?

157

J
I

t I T

tlrc rlistinction [r.ctween science and metaphysics in the narrow s c r r s c,w l rc rc rs i ri s o rn e o f h i s o th e r w ri ti ngs, parti cl l l arl y hi s 1963 Oonjtrtrtrrsard. Rejrtatiorrs,he tackles the issue of science versus distinctions should be borne in rnind; but, as 1'sctttkr*s<'icn,;g,,..These llr c v ; rri o rrsa s p c c tso f th e p ro b l e m are al l i ntercorrncctcd, e w i l l w t t ot ilt s i s l o rr tl rt:rrr o m u c h i n w h a t fol l ow s. to As wc rcrrrarkcd carlier, Russell influenced the Vienna Circle as rcgrtrtlstlrcir intcrcst irr logic and in the problem of induction. Their ir r t c r c s ti rr th c rl c rrra rc a ti o n ro b l e rn, how gver, u' as sti nl ul ated by p tlrc wor k of llrrsscll'sstudeni Wittgenstein.[h1deed, the dernarcariou lrroblcrrr is orrc of the central issues in Wittgenstein's first philosrrlrltic';rl wrrrk, tlrc T'rttctalus Logico-Philosophicus published in 1921. 'l'ltis ngrccs rvitlr Wittgcnsteirr'sown opiniorr, for, vu'ritingto l{ussell ; r lr r r r rti l rc ' I' ru tl ttts o n 1 9 Au g u s t 1 919, he says; ' The rrrai npoi rrt i s (3csagr) t lr r t lt t ' o ry o f w h a t c n n b c c x p re ssecl by props - i .e. by f irrp,rrirgr: - (arrd, lvhich conrcs to the sanre, what can be thou,ght) sir'l f ar r r fr v l ti rt c l n l r()t b c c x p rc s s e db y p rops, but onl y show n (gezei gt); wlt it ' lt , | l tc l i c v c , i s th c c a rd i n a l p robl enr of phi l osophy' (Il ussel l , lr,l(rll,1r, llll). As wc shall see, the distinction between what can be exptcssctl ntrl rvlt;tt can only be shown is, within the framework of tltr: 'lhtrfrtlrs, tlrr: clistinction between scienceand metaphysics. 'f 'lrc irrllrrcrrt:c o('tlrc 'I-rtrctatus the Vienna Circle is very well on t k r c r urtt' trtc tll.rrrl c c < l M c rrg e r, w h o w as i nvi ted to j oi n the V i enrra , ( jir r fc i rr tl re l rrtrrn m o f 1 9 2 7h a s thi s to say: ()rrfy in 19?.6127 thc prograrnof the Circle becndiffcrent,Hahn lmd about Wittgenstein's Itrltf rrrc, Various rlislgrccnrents Trattatus bad srtrllcctl irr tlrc grorrp. T'htrs,at Carnap'ssuggestion, Schlick had nrcctings necessary read es to corlsccutivc to rlccirlcd rklvotcils rrriury 'l'lrcjoint rcacling, tlrc lrook ;rkrrrrl. sentcnce sentcnce, by fillcd that clrtircl(ir(icnricycrr lrcforcI joined. (Menger,1982,p. 86) I t is t l rrrs s c rrs i l rl ct() s ta rt o u r i nvesti gati on of the dcrrrarcati orr the views of Wittgenstein in the Tractatus, lrlrrlrlcrrrby r:orrsirlcrirrg ro lur r t 1' rrrti c rrl a rlly w tl rc s cv i e w s i n fl uencedthe V i enna C i rcl e. l ' l' lr is i s o rtr (i rs t c n c o u rrtc rrv i th th e i deasof Ludw i g W i ttgenstei n, r v lr o is c c rtrri rrl y o rrc o f th e n ro st fanrous phi l osophers of the t r v c nt ic tl r (' e n trrry . l rr u c c o rd l n c c w i th the usual practi ce of thi s lrook, I will givc n fbw details of his life before discussing his lr lr iloso l ' l ty . l t s l ro trl d b c a d c l c d , how cver, that i n thc casc of W it t llt :rrs tt' i rr,tl rt' r' c a rc s o n l c d angers i n thi s procedure. For W it t gc rrs tc i n w rs a l r u rru s n a l m a n , and l ed a strange l i fe, w hi ch

calls out for psychological and sociological analysis.The dangeris that of becorningtoo fascinated the man, and henceneglecting by his philosophy, or perhaps, becauseof the pequliaritiesof the mar1, over-estimatingthe importance of his philosophy. Yet, so interestingis Wittgensteilt'slife that I cannot refrain from giving sonle account of it. I will try, however, to concentrate those on aspects it which I believeto be relevantto his philosophy.When, of later, I come to discuss philosophy,I will then refer back to the his biographical episodes which may haveinfluenced particularideas rhe of Wittgensteinunder discussion.

8.2 Wittgenstein's Life

* Ludlvig Wittgcnsteirlwas born irr Vienna in 1889, and died in Canrbridgein 1951.rHis socialbackgroundwas an unusualone, since his fanrily as one of the richestin Austria, and indeedin Europe. While most of Wittgenstein's relativeswere fairly well-todo, the colossal fortune of Wittgenstein's immediatefamily was due to the efforts of Wittgenstein'sfather, Karl. Through a seriesof rcmarkabledeals,he made himself the leading figure in the steel industry of the Austro-HungarianEmpire, organizingthe Austrian Iron Cartel in 1886. Then in 1899, having accumulateda vasr fortune, he retired from business the age of orrly fifty-two. at Karl Wittgensteinwas a patron of the arts as well as a successful entrepreneur. The family residence Vienna was visited by some in of the leadingmusicians the day. GustavMahler, Bruno Waltcr, of Johannesllrahms, and Clara Schurnannfrequently came to the PalaisWittgenstein;while JosephJoachim and the youthful Pablo played in its great salon. Karl Wittgensteinwas interested Casals in painting and architectureas well as music. ln 1897 a group of progressiveartists defectedfrom the conservative Vienneseariists' association form what they called the 'Vienna Secession'. to Karl Wittgensteinwas their most important benefactor, arrdgave money for the Secession building which was erectedin 1898.This srriking building ernbodiedthe architectural principlesof the group, as wcll as providing a place where they could hold tlreir exlribitiorrs. Thc group advocatedwhat was known as the Jugendstil, Viennese the vcrsiorr art nouvcau. most famouspainter,GustavKlimt, did a Its of portrait of Wittgenstein'ssister Margareteon the occasionof her rrrarriage 1905. in
irr'1

i B
F 5

a
$

i
F

,e. g f,

-..i:iiir:

158

S c i e n c e n d M e taphysi cs a

ls Metaphysics Meaningless?

159

sto:y, but the lirmily So flir, the Wittgenstein history is a success (in this respect sonrewtrat resernbling the Kennedy farnily) was also to be urarked by tragedy. Ludwig was the 'ioungest ot- eight children, five boys and three girls. Of the five bo1 s, the three eldest comnritted suicide. 'fhe fourth, Paul, also had an unhappy fate. He rvas training to bc-'a conccrt pianist, but lost his right rrnr in thc First World War. Ravel wrote tbe Concerto the t.eft Hand for hiru. Jbr Ludwig himself suffered from periods of profoun,:l depression,and secms to have bcen ncar to comnritting suicide on leveral occasiorrs. None of tlre fivc Wittqensteirr brothers marriecl, a;rciat lcast four <.rf them, including Ludwig, seem to have had homo:;exual tendencies. Significantly, the three girls were much better ad.i;sted, ancl two of ttrern nrarried and had children. Ludwig was at first educatcd at home by priva,,: tutors, but r,vas thcn sent at the age of fourteen to a grammar school in Linz. Ironically, this school was also attended by Adolf Hitler, who was only a few days older than \Vittgenstein. The rwo were llever actually in the same class, sirrceWittgenstein was a year ahead for his age, whereas Hitler was a year behind. Wit:4enstein was not outstanding at school, and in lrjs Matura certifica:,: he obtained the highest grade in only one subject: religious knowlcdge. After leaving school, Wittgenstein began by stucyilrg engiueering at the Technische Hochschule in lJerlin-Charlottet burg. In 1908 he went to England, and became a researchstudent i: the departrnent J of engineering at the University of Mancheste:,",where he did .rescarchin ; i! 4cronautics (the clcsigtr of a ptopellcl). Horvcvcr, his ,-4i t)? study of the requisite nratltetrraticsled hinr to alt illterest in the :ij j founriationsof mathematics,and in 1911he transfer.edto Cambridge .i to study this subject under Bertrand Russell. W-ttgenstein's relationship with Russell was decisive for his later work in philosophy. Ilussell has the following to say about an earh' encounter with W it t gen s te i n :
J

Thc relationship between llussell and Wittgerrsteitr sootr became a close une, as is shown by arrotherof Russell'sanecdotes: Flc used to come to seeme cvery eveningat rnidnight, and paceup rnd down my room like a wild beast for three hours in agitatcd silcrrcc. I C)nce saidto hinr: 'Are you thinking about logic or about your sins?''Iloth', he replicd, and continucdhis pacing.I did not like probablcboth to him that it was time for bed, as it seerncd to suggest arrdnre that on leaving me he would commit suicide.(p. 99) llusscll probably knew that two of Wittgenstein's brothers had by then comrnitted suicide, and this may account lor his tolerance of Wittgenstein's strange behaviour. It seems clear that Russell had 'Wittgenstein. McGuinness, beconre an intellectual father figure for however, makes the very interesting point that Russell's own enlotional state drew him towards 'adopting' Wittgenstein as an intellectual son. McGuinness quotes the following passagefrom a letter which Russell wrote to Lady Ottolirre Morrell on 22 August 1912: 'l had a letter from V/ittgenstein, a dear letter which l will show you. I love hirn as if he were rny son.' He then goes on to conlnlent: 'There is no doubt that Russell meant exactlv what he said. He was rlow forty and fassionately warrted childrcn but saw r)o prospccts of having any. . . . Wittgenstein to some exterrt made up for the children he had missed' (McGuinness, 1988, p. 103). One remarkable feature ofRussell's relationship with Wittgenstein is that Wittgensteirl carne very rapidly to assunle the donrinating rolc. Irr a lctter to Lady Ottolirre Morrell datcd2T May 1913,I{usscll has this to say about a recent encounter with Wittgclrstcin: 'l showed him a crucial part of what I have been writing. He said it rvas all wrongt not realizing the difficulties - that he had tried ury view and knew it wouldn't work. I couldn't understand his objection - in fact he was very inarticulate - but I feel in my bones that he must be right, and that he has seen something I have missed' (qnoted from McGuinness, 1988, p. 174).ln 1916, in another letter to Lady.Ottoline Morrell, Russell recalled this episode as follows: Do you remcmber that. . . I wrote a lot of stuff about Thcory of ' Knowledge, which Wittgensteincriticizedwith the greatest scverity? His criticisln, tho' I don't think you rcalizcdit at the timc, was an event of first-rate importancein my life, and affectcdcverything I have done since.I saw he was right, and I saw that I could not hopc work in philosophy.My impulservas evcr again to do fundamental shattered,like a wave dashedagainsta breakwater.I becamefillcd (1968, 57) with utter despair. p.

i
lii

t:

!:,

$ lr
li ii
t:

irl

At the end of his first term at Trinity, he came to n e and said: 'Do you think I am an absoluteidiot?' I said: 'Why do you want to but if know?' He replicd:'Because I am I shallbccomean aeronaut, if I am not I shall lrccornea philosophcr.'l said to hirl: 'My dcar fellolv, I don't know whcther you are an absoluteiriiot or not, but .' if you will write mc an essayduring the vacationt pon any philoyou, I will rcad it and t,-'llyou.'He did topic that interests sophical so, and brought it to me at the beginrringof the neli term. As sootr persuadcd that Lc was a rnan of I as I read thc first senteltce, becanre hirn that he should on l1o acciitnt becornean gcnius, ancl assurccl (1968,p. 99) acrotlaut,

160

Sc i e n c e n d Metaphysi cs a

t t t t t t t t t
I I

llr r s s c lla < l c l sn a fo o tn o tc :' I s o o n [4 ot over thi s rnood,' but the fact i r c r r r r r ir rs :rt h c tl i d v c ry l i ttl e fu n c l a n rental ork i rr phi l oscl phyafter th w l( ) 1. ] , t lcv o ti rrg h i rn s c l f i n s te a dto p o l i ti cs and other i ntercsts. l) t r r ir rl i tl rc F i rs t Wo rl d Wa r, W i ttgcnstei n served w i th thc A r r s t lir rrr;rrrrry , rrra i rrl yo l r th c c a s tcrtr front. It w as duri rrg thi s llt ' r ior l t l r:rt l rc t' o rrrl rl c tc dh i s fi rs t nraj or phi l osophi cal w ork, rhc ' l' t t r lr t lr t .r, l ri r' l r rv . of ' ' v :rsl rtrb l i s l te di n 1 921 The ci rcrtmstal tccs i t.s r ' r ' t nl' os i ti o rrl l rrtl y c x p l a i n w h y th i s book, w hi ch.i s l argel y about I o1ii, , l, trry ,rr;r1l1 r,, rrrc l rrri l rl c o l rta irrs ui tea nunrber of refl ecti ons ;, u rrl cl ur r r lr ' ; r tl r l i rl c x rrrrrp l ctl rc fo l l o w i n g (as i s custorrrary e w i l l gi vc , w t ; r t t r l, r lirrrrs rrr l l rc ' l ' ru rttttrtsy c i ti n g the rel evantparagraphrather l ro b I lt . r t r1r , r 1 1 r' n rl )(' ): nu r f r ' l ll, ! N n t o rrl y i s tl rc rt: o g u a r:u ltec tl rc tcruporalrnnrortrl i ty n of i ul llr r ,l rttrtt,ttro rtl ,tl r:rti s to s a y o f its ctcrnalsurvi vrl aftcr cl crath: s , co fi thcr lr r r t ,ir r ,rn y (;rl r('tl ri s i rs s rrn rp ti o rr n rpl ctcl y ri l sto accotnpl i sh i b lr lt lr n\r, l rrr rv l ri r' l r t l r:rsn l rv a y s e eni ntcndcd.Or i s sonreri cl dl c l' ly lr l l r),n ry \u rv i v i rrl i(o r c v c r?Is u ot tl ri sctcl rrali fc i tscl f' l srnucl r l ul , t t t rl rl l r' .r\o rr l )rc s (' rrtl b ? T h e sol uti onof thc ri ddl e of l i fc i rr li r lr r r{r' , i l l rlti i l rr'l rt' srrrt/ri r/r' s p a r:c l me. a n c ti ( ll t ' r rl tt;ri rrl yrro t tl rc s o l u ti o no f a ny probl cms natural encc of sci llt . r llh rf(l u l rr' (1 ,) p Nnt t ' lt ntv tl rc th ' trtl rr' ;rti rl n ro b l e l rl conres i n herc. The ri ddl e of nut lr n' i(' n l l i l r i s rro t l o b c s o l v e c lb y the sol uti on of any probl eurs ol n, r lnr , rlrl r' rrrr' , l rc r:rrrs th c ri d d l e b el ongs to metaphysi cs. c A llc r tl rr' rv .rr, Wi l tg c trs tc i rr' s i fe entered w hat rvas perhaps i ts l r r ont c iit t.ro trl i n rry l rl rrs c . l l i s fa th e r had di ed i n 1913, but, ski l ful lr r r t r r r c rr n r,u r tn tl rr' l rrs t, h c h a c l , beforc hi s cl cath, transfbrrcd v t r t r ully , rl l l rrr l rr;rri tlc ;rp i ta lto A n re ri can stocks and bonds, l rri nr r lr , rffy f f f tl r(' t l rri tc tl Strrtc s c l C o rporati ou (Il artl ey, 1973,p.21). Stc ' llr r r r llr r' \\,,u \v l l r(l r rtri rrc tls o n ra n y in A ustri a greatl y i ncreased the W t t t plt ' rrs l t' tn ttu u c , N o s rl o trc rh a c {Ludw i g W i ttgorstei n rcturncci lo I u r t v t lt , u r l rl r' , l ro w c v t' r' ,tl ra n l rc rc n ounccd l ri s shareof thc farui l y or nl{ r r ( ' y , ,rrr.l,rrr,rrr1 -' l,r,1 i t to l l c tra n sferredto hi s si bl i ngs. A s the J t t t lr (r , rl l l rt' ' l ' t,rrl ,ttrtr rrtl l l c rtra l rd l l ussel l ' sfavouri te studer-rt. f u hc r ' or t lr l,{ r l ( o u trr' , l r,rv c o l rt:ri n c tla u rri versi typost, and probabl y a o v t ' r 1,11, ,,' ,1rrr' . Y r,l l rc rt' rro rrrrc ctl ri s possi bi l i tyas w cl l , and startcd d , r t r , r ur in l l( rl n r\r' l u t tc trc l rt' rsrr c l c rrrentary i school s.Then, bctw cen ) l( l. l( lr r tl l tl .!(r, l t,.w o rk c tl a s a school nrastcri tr thrcc rcurore (' l r A r r s t i, r n v rl l ,r1 1 ,' s r ;rttt,rrb u c l r, c h b erg,and Otterthal ). Pu llr r r v i ,, tl rtr v t' ry s tr' ;rn l l ,cc h a v i o r .rr bc cxpl ai ncd?I)uri ng thc ll to wr r , W r ttl ' ,r' rr,,tc rrr c rrtc rc c la s nral l bookshop i n Tamow i n l r;rrl
i
ti

t6j Galicia'Hqre he discovered jr.rstone book;'Tolsroy'sThe cospel irt l3rieJ'. bought it, and reid and rereadit durini the rest Ji tr" He war. Thus wittgerstein became convcrted what could be calleda to Tolstoyalrchristia' populism. I believethat this rerrrained of o'e his nrostdceplyheld beliefs until his dying day. Tolstoy'schristian populisrn involvesi-condenrnatio, the rich of ard praiselor the virtu.eand nobility.of the poor, particularly of poor peasants. Thus Tolstoy himselfsays: I turncdfronrthelifc of our circle, acknr:wredgi'g oursis not lifc that but a simulation lite - that the conditions-of of iuperfluityin which wc-Iive clcprive of thc possibility urrdcrstandirig a'd thatin us of litc, orclcr undcrstand I musturrdcrstand rn to life not lifc such "*.-.ptional as-ours-who arc parasiteson life, but the life of the sirnplc labourine
folk - those who rnake life - a.d the meani'g which thcy att.iu,,t. ii it. (ltt79, p. 67)

ls Metaphysics Meaningless7

ts J-

wc carrsec that v/ittgenstei''sactionswere very nruch in accorcl with this point of view. Indeed, Tolstoy himsclf had openeda schoolfor the peasants his estate. on If wittge'stein had hopedto fi'd rrolrle a'd virtuouspcasa'ts in Austrianvillages, wassoondisappoi'ted.on)3 ocro'er he Mrotg 1921,he wrote to Russell:'l am still at trattenbach,surrou'ded, as cver' by odious'essand baseness. k'ow that hu'ran beingson I the averageare not worth rrruch anywhere, but here thei are much good-for-'othing and iriesponsible than elsewhere' -r'ore (Russell, 1968,p. 120).lt seems rharthe villagers had an opinionof wittgenstein which wasjust about as unflatteii'g as the oiinion he held of them. 'What had gone wrong? For the answerto this question, are we . indebtedto the i'vestigatio's of w. w. Bartley tit. tn the 1960s Ilardey decidedto visit Trattenbachand the other villageswhere wittgcnstei'had taughr,qosee-if anyoneremcmbercd hiri. IJartley thought that after more,thanforty years,wittgcrrstein might wcil have been forgorten, and he rook some photogiaphsto jo[ failing memories.It turned out, however, that n.rily-.u".yonJ of the appropriate. age-groupremembered Wittgenstein, though no one k'ew that he had subsequently becomei frmous phitJsophcr. It sec.nrs that wittgerrstein made as big an impact ii the Austrian villages he did at cambridgeot ,-ong thc vierrnacircle. IJut the as i'rpressionseems the whole to havebeen unfavourable. on Typical rvere'the,yeird fellow' (|)artley,197J, g6), ,a'crazy descriptions p. fellow' (ein verriickter Kert)(p. S4).

l:r :i;lt.jl:l l

to t

Scienceand Metaphysics

ls Metaphysics Meaningless?

toJ

!:

tr
.t' ?

At that time a village schoolmasterwould nor-nally have livecl in a house with at least one setvant, and at a statrdatilrather above that of the peasants.The villagers ktrcw that Wittgeristeilt canle fronr a very rich fanrily, and were therefore most surpri: t:d that he chose to live irr what llartley aptly describesas 'ostentaticus poverty'(p.72). lived in a tiny, prinritive rootn, an : rcgtrlarly hacl lris Wittp;errstein lunch rvith orte of the poorest fanrilies itt the arca (the Trahts). His evening nreal was, if anything, even worse, sini:e it consisted of a mixture of oatmeal and cocoa prepared by hinrself. It seems that Wittgenstein's eccetrtricities,whiclt in Cantbrid!;e wcre takeu as a as sign of gctrius, rverc rcgarded by the Austriatr Fr!;.lsallts a sign of The average villager had to work hard to nrental derangetnent. achieve a low starrdard of living, atrd tto doul i: clrcanled of inrprovirrg his conclitiotr at lcast a littlc. Wltat, tlrcr , woulcl hc or shc think of a rich nran who had the nlealls to livc r,vell, but lived in poverty for no aPparentreasotr? homoThe villagcrs had also foutrd out about V/i':r-genstcin's sexual tendencies,and this may have beeu a sour,;eof anxiety to the parentsof the boys r.vhomhe taught. For a long period, Wittgetrsteitr spent every day between 4 p.m. and 7 p.m. giviirg extra lessoltsto his favourite pupil, Karl Grubcr, whom he evcr, walrtcd to adopt. Alnrost ccrtuinly this rvas all cluite iltltoccttt, tr:.lt otlc catr uttclcrstand how such conduct could have alartrred tl',: villagers. When Wittgenstein watrted to take another of his favc:rrite pupils, Oskar Fuchs, to Vienna to see a play at the llurgthealtr, Fuchs's ltrother refnscd to entrust ltcr sott to 'the lveird fellow'. Wittgcn*5tciuclcvoteclhinrself witlt great intcrtsity to ltis cluticsas a teachcr. He seelnsto have lreen detcrttrinecltlrrt his ptrpils shoulcl reach a high standard atrd, in particular, leartt sonlethiltg about subjects (such as algebra) which were not ntrrmally taught in elementary schools. All this proved very succlssful with a few talented childrcn such as KarlGruber, but led to <onflictswith those of urore averagc abilities. Wittgenstein devoted the first two hours aud some oihis former pupils of each day to teaching nrathelrratics, recalled remetnberiug these two hours with lrorror for years afteron (Monk, 1990, p. 195). There was sorne resistance the part r.vards of the chilclren to this ilrtensive teaching, and Wittgenstein resorted to corporal punishnreltt. He seemsto have been particularly hard orr the girls, pulling hair till it cante out and boxing ears till the blood , r an. Matters firtally canre to a head, in Otterthal i'r April 1926. The irrciclent(cf. llartley, 1973, pp. 89-90, and Monk 1990, pp. 232-3)

concerned eleven-year-<ild calledJosef an boy Haidbauer.Haidbauer was an unfortunate case. His iather ivas dcad. and his mother worked as a live-in maid for a local peasanrcalled piribauer. H-aidbauer paleand sickly, and later died of leukaemia the age was ar of fourteen. For some reason,Wittgensteinlost his temper with Haidbauer,and struck him two or threetimes on the headwith such force that the child collapsed.Wittgensteirr had previously struck Piribauer'sdaughter so hard that she bled behind the ears., When Piribauerheardof the Haidbauerincidenr,he determined action, on and, together with some rreighbours,instituted legal proceedirrgs againstWittgenstein.These villagers, who seem ro lrave regardCd Wittgenstein as a mentally derangedman with homosexualtendencies, were probably very worried that, if not sropped,he would lose co'trol and hurt one of their children seriously.wittgenstein was acq-uitted the trial, but gave up school teachingthireafter. at Ironically, in the autumn of the year in which Wittgenstein's career as a schoolteacher ended so grimly, the Vienna Circle begandevoting their seminarto a line-by-line readingof his Tractatus. On his return to Vienna, Wittgensteinworked for a while as a gardenerfor a monastery,and seriouslycontemplated becominga nrorrk. However, his sister found anotheroccr.tpatiou hirn. She for had decidedto build a large mansion in Vienna, a'd Wittgenstei' and his friend Paul Engelmannbecame involved in the projcct. It is not clear how much part Witrgensteintook in the design of the house, but he spent a great deal of time organizing the actual building, According to Bartley, 'he supervised workmen in the the most meticulouslycarefuland exactingway during rhe construction of the building' (1973,p. 95). By now, however,Wirtgeustein was beginning to becomeinvolved with philosophyagain. Although he never attendeda meeting of the Vienna Circle, he startedto have philosophicaldiscussions with Schlick and l7aismann in 1927, In March 1928he attended lectureon the foundations mathenratics a of by Brouwer, who impressed him deeply.By the summerof 1929, Wittgenstein had decided to rerurn to philosophy on a full-time basis. As with nearly all aspectsof Wittgenstein'slife, his return to academia had some distinctly bizarrefeatures. decidedto apply He for a CambridgePh.D., with the Tractatus, now a *orld-famous by philosophicalbook, as his thesis. His oral examinationwas conducted in June 1929 by Russelland G. E. Moore, and Moore is supposcdto have presented following report: 'lt is my pcrsonal the opinion that Mr Vittgenstein'srhesisis a work of genius;but, be

rf
:l
:l

1{}4

Scienceand Metaphysics

ls Metaphysics Meaningless?

165

r|

F F
T

F
F
F
h

ftt

F
F

requi red for tl i e t lr ir t u s i t rrn l ' , i t i s c c rta i n l y w e l l rrp to the standarci ( i. r r r t l rri tl g e t:g rc co f D o c to r o f Phi l osophy' (B arti ey, 1973, p. 98). rl lr r lr l,l 0 W i ttl ;c rrs tc i nw a s e l e c tc dto a fel l ow shi p ai Tri ni ty C ol l ege, ( l, r r r t Irri tl 1 1 rl,rr th c s trc rrg tho f a r cport w ri ttcl t by l {usscl l . f' hcrcc W lf ilr l rt' r' c n tl ri rrc d (w i th s o rrrc cl taracteri sti cal l y i ttgcrrstci tri l rr ir r lt ' r rtrp l i o rrs s rrc h a s a y e a r s p e nt i n a hut i tr l .l omray) a dotr at , ( iar r r h ri tl p l cl,rc c o rrri rrg p ro fe s s o ri n 1939. H e w ol ked as a hospi tal a lr or t cl ' ;rtrrli rr n tu c d i c a l l a b o ra to r y duri ng the S econd' W orl d W ar, ;tr,l tftrtr retircrl frorrr his chair in 7947. He was found to be sufTert r r g f io ti t (:i l n (:c irrr 1 9 4 9 ,a n d c l i e di n C arnbri dge i n 1951. Alrcady l:y Jr,rrrc1929, V/ittgerlstein had rnoved away fronr the ancl during his later period at grrrsititrtr ittlvocrtcclin thc 'fraetatus, ltc ( lnrrrlrrirlBe, hc' rvorkcd out a conlpletely diffbrent philosophy, only after his death. The two main works of rvlrir'lr wrs prrltlishccl published in Inpcstigatiotts, tlrr l.rtet Wittgcrrstcirt are Philosolthical published in of Mathematics, on l()li.l, rrttf ltt'rrr,trfts tlrc Foundations ' l ' l re s cl ru o k s d c a l p ri n c i p a l l y w i th the phi l osophy of l anguage, I t ) 5( r , t lr r ; r l ri l o s o p h yo l ' p s y c h o l o g y , a n d the phi l osophy of mathernati cs. ' l' lt ey a rc l l ttts l c s s rc l c v a n t to th e phi l osophy of sci encethan the 'l'rnlilts, wltich ltarl, lrccauseof its influence on the Vienna Circle, ru lrig cllt'ct otr tltc dcvclopntettt of philosophy of scietrcein thc ' l ' h e rc a re , h o wever, sol ne thi ngs i n the P l ri l ot went i c tl t (rc n ttl ry . which car-r be related to the demarcation sryhil.tl Inursli,gatiotts s rrrtl I w i l l c l i s c trs tl te s el ater i n the chapter. lr r qr lrl c n t, l lit t t ' v c y i rrgW i ttg c n s te i rl ' s i [e a s a w hol e, the thi ng that stal l dsout llrr rrte is lltc c:otttrlst bctween the adulation which he receivedfrorn with which he r.lite glorrlrs of wcalthy intellectualsand the harshness was rejccte(lby thc Attstrian peasalltry. Wittgenstein had no sooner student than l l ussel l and a lr r iv c ti i rr Oa rrrb ri c l g c s a y o u tl g research r lr e ot l rc r l e .l tl i rrgp h i l o s o p h e rsw e re hai l i ng hi m as a phi l osophi cal gc r r ir rs ,Wl rc rr W i ttg c tts tc i tt' ss i s tcr vi si ted hi rn i n C ambri dge i n I t ) 12, s l rc h u d tc r w i th l l trs s c l l ,w h o sai d to her:' W e expect drc trcxt big s t e p i n p l ri l o s o p h y to tre tl k e n by your brother' (McGui nness, as f ()fltl,1r. f .10),'l'hc Vicnna Circle regarded tlrc Tractatus a turningi rr p l ri l rts o p h y .W h e n l 7 i ttg e nstei nreturned to C ambri dge, he lr t r ir r t rvus trertctl witlt tr<l lcss awe atrd reverence,as Malcoltrr's nretloir ( l95t t ) t l c ;rrl y s l to w s . (lrliottsly, lil/ittllcrrsteirlseems to have treated these elite groups h , , " v lr o ;rrl ttl rttc rl i rrr w i th a fa i r measure of i ndi fference, i f not t lor v r rr 1 l l rt t:o trtt' rrrp t.Il trs s c l l s ccured W i ttgenstei n' s cl ecti otr to i ( ir r r r r h ri rl g e ' su ()s t c x c l u s i v e s e cret soci ety, the A postl es - Iater l Io bet.ornc lt lrrcctlirrg ground for Russian spies. But Wittgenstein

resigned after attending only one meeting.'This was at leasr one more tharl the number of meetings of the Vienna Circle which he attended. My irnprcssion is that Wittgenstein'spopulisnt was not arr affectation, but rvas quitc genuimely felt; that Wittgenstein really belicvcd the u'calthy elite to be degenerate, and that virtue rvas to be found amoirg the poor rather than the rich. Yet his attemprs to mix with the Austrian peasantryended with their taking legal action to get rid of hin. S/e know nothing of his relations with the workers who built his sister's mansion; but Bartley's description of his sr.rpervising therrr 'in thc-most meticulously careFulind exacting way' (1973, p. 95) suggests that his relations with them may nor havc been very good either. What moral can be drawn from all this? Perhaps the sinrple one that it is hard for anyone to cscapchis or her classbackground. So rnuch,then for Wittgerrstein the man; let us now turn to his phi l oso phy.

8.3 Wittgenstein's Tractatus Let us begin with a number of quotationsfrorn the Tractatus which show the views on rnetaphysics held by Wittgensteinat thar time

(1e2t).

4.003 Most of the propositions and questions to be found in philosophicalworks aie not falsebut norxensical. Consequently wc cannot give any answer to questionsof this kirrd, but cen only establishthat they are rronsensical. Most of the propositionsand qucstionsof philosophersarise from our failure to understahdthe logic of our language. (Thcy belong to thc sarneclass tlrc questionwhcthcr the good is as nlore or lcssidcnticalthan the beautiful.) And it is not surprising that the deepcstproblcms arc in fact not problcrns all. at This point of view leads him to characterize what he regardsas the correct method in philosophy as follows:
6.53 The correct mcthod in philosophy would really be the following: to say nothirrg except what can be said, i.e. propositions of:;; .. natural science- i.e. something that has nothing to do with philosophy'l { r then, - and then. whencver solneone else wanted to sav sonrethins'rictrsay sonrethirlgr'nictai :,{

i,

,.: I ! rt::j l:.: ii

\r n

166

Scienceand Metaphysics

ls Metaphysics MeaninglessT

167

physical, to dcmonstrate ro hirn th:rt he had failed to givc a nrcaniug to ccrtain sigrrs in his proposirions. Although it rvould lor trc satisfying to tl"re other person - he would noi have the feclirrgthat wc rvere tcaching him philosophy - thrs nrethod v :rulcl be thc only strictly correct olrc. For Wittgcrrsteirr, thcn, the corrcct lnethod rn philosophy is to

clernonstratc that any given rnetaphysical proposjtior) is nreaniuglcss. For this purpose, it would clearly be dcsirable -o havc a thcory of meaning, and Wittgenstein does indeed provide :uch a theory in the Tractatus.It nrtrst be added, however, that later il:. his life he criticized and gave up his 'l-ractatustheory of meauing, and in lris sccond nrajor plrilosophical work, the Philosophical,lnuestigations(1953) advocated a different theory of meaning. My ,,'rn opinion is that Wittgcnstcirr's self-criticism is correct, and that, although rhcrc arc some inrportant ideasin his earlier (Tractatus) the,rry of rneaning, his Iater (lruestigotions)theory of meaniug is prt':ferable. Nolv the interesting thing (as I will argue later) is tha the doctrine rhar rretaphysics is nreaninglessdocs indeed follov'n, from the Tractatus tlreory of rneaning, but not from the Investigatlons theory. Let us begirr, however, with a brief exposition of the Tractatus tlreory. This tlrcory is based on thc notion of alr c crncntary pr<lposition, which is supposcd to assert tltat a sirrrple lact is thc casc. As W it t g e u s te i rr ri rn s c l fs a y s ; l 4.21 The sinrplest kind of propositiotl, r:ll elernentary propositjorr, asscrtsthe existenceof a statc of lffairs. A ll ot h e r n te a u i rrg fu lp ro p o s i ti o us.are bui l t up cut of cl cnrcntary propositions. We can explain this by an analog.ywith chernistry. The elenrentary propositions corresponcl to atclns, and the nonelenlelltary (o. complex) propositions to nolecules. Indeed, elementary propositions can also be called atomi, propositions, and complex propositions moleular propositions. Jhe process by which molecular propositions are built up out of ator,ric propositions is known as truth-frmctional composition. Wittgenr tein says: As 5 A proposition is a truth-function ol elementary prop o s i ti o n s . B y ' pro p o s i ti o n ' h e re , W i ttg e n s te i nmeans' meani ngfulproposi ti on' . The 'propositions' of metaphysics and even, rrther surprisingly, thosc of nrathcinatics(cf. 6.2) rre pscudo-propostions.

ln a nroment I will explain what is niearrt by truth-f'rrcio'al conrpOsition, but first let us consider a nlorc basic question. wittgenstein's Trdctatus theory of meaning is basedon the ric.tio'of elenreutary(or atorrric)propositions, and wc nright well expcct that hc would clarify this notion by-giving sor'c exanrprcs eleme'tary of propositions. IJut he never in fact does this; irrstead,he nrakcsa few gerrcral rcrnarks about clemcntary propositions. vfc l*arn, [or cxanrple, that elementary propositions corrsist of narnes, and that nanles mearr objects. 4.22 An elenrentary proposition consists of namcs. lt is a nexus, a collcatetration,. nanlcs. of 3.2W A narne nlcans an object, The object is its nrcarring.

Malcohn has an interesting anccdoteregarding this: I askcdWittgcnstein whcther, whe. hc wrorc thc Tractatus, had hc cvcr dccided uporr anything as an example a .sinrplc objcct'. His of rcply was thar at ther time his thou6;ht had becnthat hc *., l,gi.ior; " and that it was not his business, a logicia', ro try to dccidewircthcr as this thi'g or that was a simplething or a conrplerthing, that bei'g r purely e'rpirical matter! It was clcar that hc rcgarded lris fl.rricr opinionas absurd.(1958, 86) p. Granted this otrscurity in the Tractatusitsclf,, it was up ro thc vie'na circle to clarify the nratter, and this they did by identifyi'g clcure'tary propositions with simple observaiion stitcmerrts, <lr protocolstatenlents. This idgrtification led in turn to the irrvolvecl debates about the nature of observation statemcnts which rve consideredin chapter 6. ve shall not go ovcr thesesubtletiesagain herc, but confine ourselves to such colnmon-setrsc examples- of simple observation statements as 'The apple on my desk is green'. This is because our aim now is not to analyse sirnple obseivation statenlents, but to discuss how complex observatiolistatenrerrts can be built up out of simple ones. Later scholars have_ argued that the interpretation of 'elementary propositio'' adopted by tlre vienna circle is not really compatible with tlre Tractatus, and so probably docs not rcprescnt what v/ittgenstein had in mind. This view seems to have'originated in Anscombe's An Inffoduction to Wittgenstein,sTractatus lieSl; lsce particularly, chapter 1,'Ele-mentary propositions', pp. ZS-40). Arrscornbe quotes the followirig passagefrom the l'racioius:

. : r; i.i,;i:..

t t
I
I
-

Ititl (r.. \ ' llil

Sc ienc e and M e t a p h y s i c s lt is c lc ar t hat t he logi c a l p r o d u c t o f 'f w <t e l e m e n t a r y I'r' oPos it ir lr r s c au neit her be a t a u t o l o g y n o r , l c o n t r a d i c t i o r r . 'l l r c s t at c r r r c nt t hat a point in t h e v i s u a l f i e l d h a s t w o d i f f e r e u t rol our s ; lt t hc s ar lr c t iur e is a c o n t r a c l i c t i o n .

ls Metaphysics Meaningless?

169

Slrt' Ilrctr ( ( ) nlnlc nt s : ll t'lt' r r r t ' r r t lly I r r o[ r , r s it ior r s c s ir r r p l c o b s c r v a t i <l r r t ; r t ( 'r n c n t s , t i s ar s i vt'ry tlillit:trlt t() scc llow what Wittgcnstcin says hcrs can possibly oflr,tlrl 11ootl tlrcnr; for, for any proposition which could reasonably Irt't:rl lc t l ; r ' s ir r r plcobs c r v at ion s t at en l e n t ',o n e c o u l d f i n d a n o t h e r t h a t rvorrltl lrc irrcorrrpatiblc rvith it and be precisely analogous to it hrliit'llly. 'l'lrcrchrrc, whatcvcr eletrrentarypropositiorrsmay be, thcy (1959, p. 27) ;ll(. rr()t sirnlrlc obscrv.lti()nstatcn'lcnts. 'l'lris :rr'1' , t r r r r c nt c c r t ainly v c r y c ollv i l ) c i n g . is 'l'lrc vic r v t hat t he c lc r lr c nt ar y pr o p o s i t i o r r s o f t h e 'I - r d c t a l i l sw e r e
l l ()l

.rtrtl 'f 'trrrlnrirrirr thcir intcresting book LVittge,Btein's Lliemm(1973) ( s('(', lirr cxarnlrlc, pp, 145and 212-21).The est, as an alterna-

ir r t c rrtl c rl [rc o b s c rv a ti o n s ta tenrentss al so adopted by Jani k to i

tl ra t c l c n rc n tl r

world atementsof the kirrcl thcoretical of phvsics. Thus the 'objects' the of dgpcst liQssiblc
I ' t ilt lilltts < l rrl tlb c : rrro re i n th e n a ture of el ernentaryparti cl esthan w ol r ' olot tt P l tt< :l tc s . ' l' lr t ' r c i s i rr { i rc t c o rrs i d e ra b l e v i d ence i n favour of thi s second c pr ir r t c r t ' trrti o rrl t fi ts w c l l w i th Ma l c ol rn' s al l ecdote recoul l tedearl i er, ; r r r r l1livt' s:r l rc ttc r c x p l l n a ti o n o f th e gnonri c utteratrce: . 1.0 ;,1 .]2n :r n r:rrrn c r f s p c a k i ng,obj ectsare col ourl ess. l o ' I ' lr i: is ct' rrrs c fi rri tc l yto ru l e o u t th e suggesti onthat' obj ects' coul d tl < Otr the other I ' t ' t ' illr cr l rh y s i r:;rl l r.j c c ts l r i n rn tedi ate seuse-data. o lr . r r r . l,it i s r;rri tc (()n l p ;l ti b l c w i th th c vi ew that' obj ects' are sonl c p k ir t l r r l l tl ry r;i r:rl l c rrrc trt;rry a rti c l e sw hose cornbi nati ongi ves ri se c t o t olorn , As l :rr u s th c < l c v c l o p rn e rlof phi l osophy of sci ence n tl rc t i r actual l y lr r , r ' r r t i t' tl rc rrtrrryi s c o rrt:c rn c ch,o w ever, w hat W i ttgetrstei n l rv rr is t lr or r r ' ,l rt l re rv ri ti rrg th c l i o c trl rrr.r of nruch l essi urportanccthan I r or r , lr t ' rv l s i rrtc rl rrc tc d b y th e Vi enna C i rcl e. I w i l l therefore t ' r llr or r rrrl l r;rt i s i rr r:l [c ' c ta h y b ri d theory of meani ng, i Ir w hi ch rv r r r r ' , r r r rr1 ,,l rrl l t,rs i ti o na re rc g a rd e dastruth-functi onsofel ementary Irro s p I t r , r l' os rti rn s .:rrrtlc l t' rrrc n ta ry ro p o si ti onsarc i denti fi edw i th si l rrpl e

observatio$ siatements.This hybrid theory, as we shall see, led ro the ViennaiCircle's famous vcrifiability criterion. It is now tinre to explain what is lneallt by truth-functional courposition, and, in accordancewith our programnle, we shall do so in terms of observation statements.Lei us begin with a simple observation statement Or : ''Ihe apple on my desk is green'. Following comnlon senseand ignoring for the monlerlt some of the subtleties corrsideredin Part III, we call say that the truth of this statelrlcnt can be checked by a straightforward observation. If the statenrent is true, this can tie verifed by observation. Next consider the negation of O1 - that is, not-Ol : 'lt is ltot thc case that the apple on my desk is green'. Opce again, the truth of this statement can be checkedby a straightforward observation,so that, in general, if O is a simple observation statenlent, then so is its negation, trot-O. Let us now go on to consider complex observatiorrstatemcllts. To do so, let us take another simple observatiorrstatcmcrlt, 02 = 'The apple on dre kitchen table is red', This can also be chcckedby a straightforward observatiou. However, we cannot chcck O1 and b2 simultaneously. I can check 01 by making observations in my study, but then, in order to check 02, I have to desccnd the stairs and enter the kitchen, a process which clearly takes time. This suggests that we might characterize a simple observation statement as one whose truth*value can be checked by observationsmade at a particular tinre alrd place. A conrplex observation statenlent is then one whose truth-value (whether true or false) can indeed be ascertainedby observations,but by observationswhich have to be made at different places and/or times. An example of a complex observation statement would be: 01 & not-O2 : 'Thc apple on my desk is green, and it is not the case that the apple on the kitchen table is red'. If this is true, its truth can certainly be verified by observation, but two observationsat different placesand times are needed. In our earlier discussions,the phrase 'observatiorrstatcnlent' was sometimes used to mean simple observation statement, and at other tirrrcs to covcr both simple and complex observations statcnlcnts. This arnbiguity was deliberate, since insisting on the distinction would have con-rplicated argument ururecessarily. the Now that wc have introduced the distinction, however, the anrbiguity can be cleared up retrospectively. In Part III, 'observation statenrcnt' is always used to mean sirnple observation statcment. Thus, at thc beginning of chapter 6 I say: 'L,et us call a stat

170

Scienceand Metaphysics

ls Metaphysics Meaningless?

171

sra After result of an observation or experinretrtan ,o&servarioa ,lrrreal-' prelimirrary definitiort is dr veloped itrto a good dcal of aualysis,this t lr c f ollor v ing (7 .4 ):4 l q b s e r" a ti o tts ta tc is the rcsult of somc sensorv irtput irrterpretcd.whethe' conscior"rslv or unconsciorrslv, using a sct of theories.' IJoth dr ilnitions are statel rl c:l ts. Itr 5.1, horvcvcr, I say: 'lct tls take atr observatiort statelncttt O, to be a statenrent wlrich can provisionally bc- agrccd to be e::hcr trtle or false on thc' basis of observatiotr and experinrcnt.' Tlris definition covers both sinrple and complex observatioll statenlent;. I ar n nou ' i n a p o s i ti o n to c x p l a i n w hat i s mea.tt by truthproposicontposition - the processby rvhich lnole<-'rlar fr.rnctiotral tions ere built up out of atotnic propositions. It shoulci be clear tltat tl s if O il at r r > b s c rv a ti o tt ta tc trrc rrt, tc tl s o i s rrot-(). Tl ' : vcry satrrc observatiotrslvhich dcterrnittc whcthcr O is trr.rcor fal ;e dcterlnitre rn,hcthcrnot-O is false or truc; sincc if O is true, not-(. is false, atrd if O is f als e. n o t-O i s tru c . S i rn i l a rl y , i f Or, ()2... C ). are a fi ni tc set of observatiotr statcnleltts, then ()1 & Oz &. . . 'k O,' is irl gelreral an observation statelrtort. (Tlrere is a srrtall exceptional casc which wc shall nlelltion later.) The moleculat proposition Or & 02 ct. . . & O, is true only if all its colnprlnellt atonls ()1, ()2. . . Ou arc truc, and is falsc if arry of these ato'trs are falsc. we cr.ndetcrmitte Brrt since Or . . . O,, arc all observation statcrnellts, by obscrvations wlrethcr they are true or false, aud thcse salne observationsdetermitre wheiher Or &. . . & O. is true or false. tllcy are, so . TIre particles 'uot-' atrd '&' are ktrowtr as tonnettiues; to spcak, thc gluc r.vhichsticks atorttic propositions (sirrrplcobscrvatogcther to fbrltr lttolecrtlarpropositio'rs (cornplcx tion strrtcrrrcrrts) The procedure cali be imagine,{ as follows. obscrvation statemetrts). Suppose wc start with a finite set of simple observatior statements. 'Wc first forrn the ncgatiotrs of each of these statenel-ts, and add them to the set. We then takc any finite subset of sta:enrentsand fornr thcir coujuttction; tltat is, we join them togethcr with thc to c ounc c t iv e' & ' , a ttd a d d th e s e s ta te n rc n t s the sct. W : thetr forur the negations of all the statemctrtsin tlre rtew set, and rdd these to forrn a set which is still fr-rrthcrexterrded. We then fo-m the coujr.rnctionsof all subsetsof this new sct, and so ou, ittdcfinitely. We can bricfly sttutnrarizc thc proccss by s:rying that, frttnr a sct of atomic propositiotrs, rve fortrr .atr extettded set of mc'lecttlar Protlf positions lry trteatrs tlrc conl)cctivcs'not-' and '&'. set cotrsistcdof silnple obscrvatiorrsti ,llllellts, tlle If thc original cxtcntlcclsct lvill, as lvc ltavc lrgttcd, ilt gcrrcralcollsist clf obscrva-

tion statenrents. There are, however, sonle exceptional caseswhich we rnust now mention. The connectives'not-' and '&' can sometirnes generate a logical contradiction or a logically valid statement instead of an observation staternent. For example, if O is any observation statenlent, then O & not-O is a logical contradiction (always falsc), while not-(O & not-O) is a logically valid statement (always truc). If, however, we take a set of simple obscrvaticrr statenlents, gerlerate the extended set of statements using the connectives 'not-' and '&', and then elinrinate all the logical contradictions and logically valid statemerrts,we are left with a set consisting cntirely of observation statements. As a rnatter of fact, we can get,the same extended set by using

connectives other than 'not-' and '&', and lTittgensteinin the Tractatus trseda singleconnective known as the Shelfer stroke synftol. Tlre gcrreralrrotion involved here is that of a trath-functiotr. sav We that a proposition is a truth-function propositions P2.. P; P of P1, if the truth-value of P (whether P is true or false)is detcrnrined by the truth-values Pr, Pz, . . Pn. of If we start with a set of atomic propositions, can be shown it 'no1-' and '&' will generate set quite easily that the connectives a consistingpreciselyof the truth-functionsof theseatomic propositions. The centralthesisof the hybrid theory which we are considering($at is. Wittgenstein's Tr4crat#s theory asinterprctedby the Vienna Cir-cle) that all meaningfulpropositions truth-functions is are of simple observationstatements. follows'that any meaningful It propositiorr is either a logical contradiction or a logically valid statcrncntor an observationstatement.It further follows that if a nrcaningfulpr:oposition true, this can be verified by observation. is The meaningfulproposition(P, say)is, by the centralprernise the of theory, a truth-function of simple observationstatements. We can determine observation truth-values these by the of simpleobservation staterrerlts, in this way we verify that P is true. This leadsto the and VienrraCircle's famous verifiabilitycriterion. It is easv to seg that, on this accountof meaning, mcllph-ysics becomesnreaningless. But the theory has othEi consequences as well. For instance, all ethical propositions become meaningless. Wittgensteindraws this conclusionquite explicitly: 6.42 Arrdso it is impossible thereto be propositions ethics. for of Propositions express can nothingthatis highcr.

6.421 It is clearthat cthicscannotbe put into words. Ethics is trarrscendental. (Ethicsand acsthctics one and thc samc.) arc

T I
II II
I

||)

Science Metaphysics and

the givenby Popper, will reject l .rtcrurr, lirr csscrrtially reasons I r"(trttillprcruutrrfuhc hybrid thcorv: nanrclv,th3! allgcaningftll* tltc ptulru$itiotts tt.U!.h lr_c: ntS. Norrr tlre lt'ss,I llclicvcthat thereis a valuable ideaherewhich can frc tlt,ritrcrl,I lhirrk tltnt it is corrcct to say that all obseruatiou \(illt'ntt'ttlt;tt'c ttttth-frrrrcti<>ns of sirnple observation statements, I and rviff r,rlf tfris thc 'l'rattiluslYicnna Circle characterizatiolt the class of
rr|'tr|rst,tv;t |it r t t s t llt c llr c t lt S. I r c t hc

of observation
r .- .+ .!_- .-

Thev had far too

n t ur r owa ( ' (frr(;c l )ti oo f w h a t w a s

F
:l :l
;

8 .4

T h o Vi e n n a C i rc l e o n Metaphysi cs

F
h h
I

' l lt c V it ' t r n ;r( l i rr' l c i rc (:c l )tc c l ttg c rrs tci rr' s cw that rnctaphysi css Wi vi i l ur t ir c ly r ne l rri rrg l c s s' . ' l rc c l a s s i c x p re ssi on thei r poi nt of vi ew i s e of lrr lrt' lotttrtl itr ( )nnt:tp's1932 article'The Elimination of Metaphysics t lr r ot t lllt l. rrg i c l l A n l l y s i s o f L a n g u a g e' . H ere C arnap w ri tes emtl i ncl lr lr ir t it ' ; r lly t;rt ' l tt tl tc rl o rn a i no f me ta p hysi cs, udi ng al l phi l osophy ol' v r r lr t c nn tl tro rrtu rti v ctl rc o ry , l o g i c a l anal ysi syi el ds the negati ve slalc,rrcntt this domainare entirelymeaningless. in rt'stfft l/rrtt thr ,illa,qel ' l' lr r r c wit lr ;r r;rrl i c :rl l i rrri n a ti o no f l n etaphysi csi s attai ned' (1932, c ( r 0. l) , ' l ' l ri s i s rro t to s a y th a t C a r nap' s vi ew s are i denti cal to 1r 1r . rr W it t p, r ' r r s l c i rr' ls . l i rc t, a s w c s h a l l s e el ater on, C arnap changedand r f t . v r ' f r r lr c rl rc i rk :;rso f' tl rc T ra c ta tu s.In parti cul ar, C arnap used tl v ; t t iot t s k in tl s o f' k l g i c n l a p p a ra tu sw h i ch w ere more conrpl i cated t lr ; r r rW it t l' ,c tts tt' i tt'trrrth -l i tn c ti o n a lc o nrposi ti on.I w i l l not go i ntcr s rl t lr c s t . t ' r ' lrrri r' ;rlc ta i l sl rc rc , h o w e v e r, p a rtl y because t theseaddi ti onal c ot t r lr lir ' ; t ti o trs < l rro t a l fc c t d re b a s i c phi l osophi calargunl ent atrd tl th lr lr t ly lr c c :rrrslc i rrk tl rrt trtrth -fu n c ti o nalcornposi ti ondoes i n fact tl r r t llir ' c io c l ra t' l c tt:ri z .crc c l n s so f o b s e rvati onstatenl ents.

lrr lris 19.)2p;rpcr, Oarrrapstatesthe verifiability criterion as lirllows: "l-lrc rrrcrrring a staternent in the method of its of lies
ls vcn

F
F
tr

Heidcgger's Was Ist llfietaplr y t illz ( 1 1 )' )t)), :rrrtl tl rc rr p ro c c c d s t<-rdcnronstratc to hi s ow rr s ut is lr r c t io trtl t;tt th c y a rc rn e a n i n g l ess Here are sonre of the

173 Heideggerian.pronouncements which carnap deals:,where with do we scekthe Nothing?How do we find the Nothing?. . . W. t nu* the Nothing ' ' . . Anxiety reuears Nothing... . Th; for the whiJ and because which we were anxious,was 'ieaily' ;";ild.'-iii..a, of tlre ,.. .Nothilrg itsclf - as such - was prescnt. wnt artouttrris 'Og, Nothing? The Noth,ing itselfn.orhings' Carn ap, l9t3A,-p. 1cf. ,, Cnllnp poinrsout, the ernphasis in ttreoriginalj. is The thesisthat thesestatements is by no nleans "r. -.r,ri.gress implausible.with a ce*ain ponderousness whrch is noi talcing in -irr,o hum.our, C.arlap tries to trarr-slate Heidegger'sp."p"ri,iou, -iiti",rtru " and co'clud"r-ihrt r."riu.", fl.:i:" logical larrguage, nultints' carnor even be^expressed such , to,rgurg". in !::!l:f^it,self 5tt' Iess can,wespggifythe finite setof simpleobservation ,t.tJrrr",rt, would verify the claim that the Nothing itself nothings, if, $tg rndecd,rt were true. Ir follows, accordingto-Carnap, thai;The Noyhiup itself _, is meaningless pr""udo_p.oporiiion.-In rle r'ar,nothings' rrrs sectron of paper,carnap considers another defence of rnetaphysics: namety,thai ii serves for'the exprcssio, thiiiirrot of a.ttihtde a person of toruarrls (p.7g). This view ii nor so r'.,rpiliriur". rtft' -a1xi.ety After ail,. feeli'gs of intense such as Heidegger'.*fi"rr., were, as it turned our, eminentlyjustified in the Geriiarry oli nz9. Yet Carnap is not preparcdro idhit this defcnce. i, quit. it regltlrnate someone express "u.n tor to his or her basicattitudeto life, but this should be done through art (lyrical poerry, ;rr;;, ," forth), through metaphysici.The problernwith the nretaphvsical ";; 'ot mode of expression i basicartitude is that it aelua., of ir,i;;rr_ physicianinto thinking that 'he rravelsin territory in which truth .In a'd falsehood at stake',whereas, reality. . . h; n", are not -zA. ou: o,"tr. "rr"it"a iik. :*p':":9. somerhing, r,rai,iri;lf. i:l1ll't':E: I hus, the nretaplrysicia' foolisrrlytrics to argucfor rrisposiiiorr, to refute the views of his opponeritr,*hil. tiie lyrical po.t _-quit" correctly, because is simplyexpressing .ttitud" ,u iif" _ ,"drin, he an fronr. doing any of theset'hings. ca."nap co'cludes trrat ,Metaphysicia.sare musicians withouimusicalatility' (p. g0),;;l ;;;ir., for writing Thus SpakeZarathustra tll. f"r,rr;?;;ry i, Y:f:l: than traditional rather metaphysics. It is rrot difficuk to detecta politicarmorivarionin trre Vicnna circle's artackon metaphysics. The poriticalviews of the ci..r" were, broadly speaking,liberal and left liberal, and their principal opponenrs were the Catholicreactionary parties. They surelythcn had a rcas.'for warti'g to dc'ro'strate'tliatcatholic ih"orosv *", mea'irrgless verbiage.In the secularsphere,they wourd to?irrity

ls Metaphysics MeaninglessZ

-ri

.ri.;.;,1,i: 1 "

ri l -i .

-_

tll,r

tt4

Scienceand MetaPhYsics

ls Metaphysics Meaningless?

175

il i'
li:

conrbarting thc leading philosophers o.. the authonfirrtl thcrDsclves Heidegger al:'l Nie-usche' ,r.i"tr .igtt,, rvho then' ai ,to*, *"" and Nietzsche was Hitler's Heidegglr gave his supporr ro Nazism,. lir,roufii. piilos.rpll"..'it is ttnderstandablcthat a lil'eral such as , C"r,rrp would wa.t to disrnissthe writi'gs of thesepl 'ilosophersas nlere enrotional effusions. Yet thr' very success 'either nreaningless.or of irrflirerrce Catholic thcology and of the pl,ilosophiesof and social rnay l{eidegger and Nietzsche cast doubt on Canlap's view'roitrt' It s writings in ,ro, b.""".ry or, indced, possible to exprcss Heidegger the precise language of fornral logic, but we should 'rot therefore nre rningless. .on.lud. that lieid*egger'sphilosophy is _altogether different 1lom those of Wittgenstein's att-i;des'*er. s-ont.what Carnap"and most other melnbers of the Vienna Circle Wittgenstei' was cert"i.rly, as we have secn, held the view that me;rphysics a man of stro'g religlous leanilrgs' mcaniniless. He was, however, the even rh;ugh not a nreutber of any organized religiorr; and in

a h"e Tractatus developecl theory of the mystical.'Ibe t'T +gfnerre

,' wrt are tl un less be shown or the of'

fi

1S

1 Ii 9 :--'t. t wlilcn

emselves

Thus thc ainr of the book is to seta limit to thought,olxther - tlot for of -to thought, but to thc cxPrcssion tl-roughts: itr orclc:to bc ablc "to thought, rve should have to find both r ides of thc to sct a"lirnit lilrit t|inkablc (i.c. rvc'shouldllavc to bc ablc to think l"hat cannot o" il;tfitti;refore only bc in la'guage that the limit car:be set, a'd of what lics on thc orhcr sicle thc lirrritwill sinrplybc nontrnse.(p. 3) in the wirtgenstein gives an inrercsting examplc of this tlreory follorving passage: of of 6.521 Thc solutiorr thc problenrof life is sccnitr the r arrishing thc problcrn. have found efter a long 1is 'ot tltis tle rcasor why those who of of doubt that the sens'e life becameclearto therir have then p..iod that tccn unableto say what cotrstituted scnsc?) 6.522 Thcre arc, indecd,thingsthat cannotbe ptrt into $ords. They Thcy rre what is mysticrl' nnnifeslrhemselvcs nt,rkr

cxpcfierrcc, but 'rrrecould not cornrnunicate this understanding in *o.dr. Sirnilarly, it rnight beconre manifesr to us that certain ethical clainrs wcre correct. Ilut we would llot then be able to put thcse cl ai rrrs rrto wor ds. i Wittgenstein's argurnerrt in thc Tractatus drat metaphysical and, horce, ruost religious statementsere meaninglcssmight well sccrn to bc an attack orr rcligion; and indeed, it was interpreted as such by Carnap arrd other nrenrbers of the Vienna Circle. In reality, lrowever, Wittgenstein,was offering an intellectually sophisticated defence of religion. We can appreciate the nature of this defence better by cornparing it with the, defence of religion offered by the Catholic Church. This is quite relevant to Wittgenstein, fotr, although his farnily were ofJewish origin, he was brought up as r Catl-rolic. Pope Leo XIII's encyclical Aeterni Patris of 1878 insisted that Catholicisrn should be based on scholasticphilosophy, parricularly orr the writings of St Thomas Aquinas. Aquinas's philosophy rs a metaphysical system which is essentiallya mqdification of Arisrotle's. So here religion is fitted into a rnetaphysical theory which is, ' naturally, accepted as meaningful. The trouble with groundirrg religior.rsbelief in this manner is that Aristotle's metaphysics has been $ornewhat discredited by the advance of science. As Martin aptly puts it: 'Aquinas's work was. . . embedded in an obsolete natural philosophy, of generally Aristotelian character,so that if his thcology was rccoverable,it had in sonre way to be rcconciled with later scientific ideas' (1991, p. 43). "Io put the problem in more generalterms, if a particular religion is based on sorne metaphysical system, then the advance of science nray uudernrine the credibility of the metaphysicaltheory adoptec{, and so undcrnrine the religion itself.@bg Wittgenstein. that we regard the doctrines of the religion as strictlv nreanitrglcss hy aud onlv to be apprehcnderJ r nrystical exnerio'.re, and hold fi'rther that these religious doctrines are clearlv derrrarcated from the nteaningful statements of science. No advances in sciencc will now have the effcct of underrnining thc religion. Its doctrincs are, so to spcak, protected by being removed from the sphcre of scientific criticism. We have already remarked that Wittgenstein had discussions with Schlick arrd Waismann from 1927. Carnap attended some of these nreetingSs, arrd has given an interesting accoullt of thern in his Intellutual Autobiography. writes: He

| lti

Scienceand Metaphysics

ls Metaphysics MeaninglessZ
waisma'ri, no lo'ger with me or Feigl, who had also becor.c acquai'tcd with him in thc mcantime,let alone with thc circlc. Although thc differencein our attitudesand persoralities exprcsscd itselfonly on certain occasions,undcrstood I very wcll thatwittgpnstci' felt it all the time and, unlike rne, was distuibcd by it. He'said to Schlickthat hc could talk o'ly with so'rcbodywho'holds his hand'. (p.27) wittgcrrstei''s theory of philosophy poses a problenr for his owrr book, the Tractatas.In the Triuatui wittgenitei' clairns that philosophical writings are meaningless, but, siirce the Tractatusis itself a piece of philosophical writirig, it would seernto follow that thc Tractatusitself be r'car_ringless. with a surprisiug (though 'rus.t n.t uncharacteristic) rigour, wittgenstein acccpts tlis cJrrclusion. Tlre doctri'es of the Tractatus'make themselves marifest' as true; but thcy carlrlor be rnea'iugfully stated i' words. This position is put forward in the famous penultirnateproposition of rhi'I'ractatus. 6.54 My.propositionsserveas elucidaiions the follorvingway: in anyo'e lvho understa'ds me ev'rtually recogniscstl .i ,.,o,r_ ",n sc'sical,when he has used them - as steps- to clirnb up bcyoncl therlr. (He Inust, so to speak,throw awry thc ladde,ft., h" hrs clirnbedup it.) Hc rnust transcc'd thesepropositions, arrd thcn he wiil scc thc world aright. This proposition of wittge'stein's has undoubtedly a cerrain dramatie appeal. Yet' we- should really fecr considerabledisquict when an author stateson the last page that his book is meaningiess. Surely somedring has. gone wrong! sornething had i'deed "gone wrong, a.d this was demo'strated by popper in his critique oi the Vienrra Circle, which we shall consider in the next section.

( )rrrt. rr,lrr.rr talked about religiort, tfie c:,ltrast bctrvccn Witt11<.rrstcirr lrrr ,r r r t lSt lr lir k ' s l. t , s it iot tbc c at t r cs t r i k i n g l y a p p a r e l l t ' l l o t h a g r c c d o f r orrrrt. in tlrc vicrv tlr;rt thc doctrincs of rcligion in their variotts fornrs Ir,rrlr r , , t lr c or c t ir ' ll c ont c r r t . I lut W itt g e n s t e i n r e j c c t e d S c h l i c k 's v i e r v tlr,rt r r ' li1, , i. rl, c lor r gc dt o t he c hilc lho o dp h a s eo f h u r n a n i t y a n d w o u l d r r sl,rr v ly t lis r r |1, t ' : tit t t lt c c our s c of c u l t u r a l d c v c l o p l n c l r t ' ' ' ' 'l lr c s c ; t nr l s ir r r il; r roc c ur r c nc esit t o t t r c o t t v c r s a t i o l l ss h o w c d t h a t tlrr'r c r v ; r s ; t s t r ol) g ir r t t c r c olf lic t in Wi t t g e l s t c i n b e t w c : c nh i s c n r o tro rr : r llilr ' ; r r r r llr is ir r t c llc c t ualt hir t k i r r g . H i s i n t c l l c c t , w o r k i r r g w i t h lirr',r t inlc t t s it y ; t t t r l pc r r c t r at ing po w c r , h a d r c c o g r r i z c d t h a t t t r n t r y ir :,l,rl ( ' nr ( ' nls r t lr c f ic lt l oI r c ligion a n d r n e t a p h y s i c sd i d n o t , s t r i c t l y s ily s1,r ' ; r k ir r 1g, : llly t lt ilr g. l1 his c ha r a c t c r i s t i ca b s o l u t e h o n e s t y w i t b liirrr s t ' ll,lr c . li, l r r or t r y r o s lr ut his cy c s t o t h i s i r r s i g h t , D t r t t h i s r c s r r l t rv;rslxtrt'tttt'ly p:rirrfirlli.;r hinr enrotiorrally, as if [c wcre corllpclIcd to :rr lr r t it l r v c lk t t c s silr a bc lov c c l pc r s o n . S c h l i c k , l r r d I , b 1 'c o t r t r a s t , Ir;r,lrrrr lovc lirr nlctirl)llysicsor nlctaphysicalthcoiogy, and thcrefore r trrr lr l , r lr : r nr lortrlr c r r r wit lt ot lt ilt nc r c o t r f l i C to r r c g r c t . E a r l i c r , r v l r c t r W s \v('w ( ' r ( ' r 11t lir r 1q it . t gc r r s t c ilr 'boo k i l r t h e C i r c l e , I h a d e r r o l c o u s l y l,r.lit . v t , ttl lr : r t lr is it t t it l( lc t owar d llc t a p h y s i c s w a s s i r l i l a r t o o u r s . I lr,rrlr r ol p: r i, l s r r f lit ' ic lr tat t c llt ic r nt o t h e s t a t e m c n t si n h i s b o o k a b o u t : tlrt'r r r y s t ir ' : r l,lr t ' c at t s this f c c lir r gsal l c l t h o u g h t s i n t h i s a r e a w e r c t o o ,lrvt' r 1t , t ' r1r ' ot t tt nit t c . ( ) r r ly pc r s or ta lc o n t a c t w i t h h i m h e l p e d m e t o rt t \(,(' n t ( ) r ( . r ' lt . lr lylr is : lt t it u( lc at t his p o i r r t . I h a t l t h c i t t l p r c s s i o t r h : r t h i s w v ,rrrr l, iv ulc r r t 'rc illr r c s pc c t t o nr c t aph y s i c s a s o l l l y a s p c c i a la s p c c to f ,r nl()r'(' lrrsir' itttt'r'ttal conflict irr his persorlality fronr wlrich hc peir f 11111 r Lr lly . ( 1963 , p p . 2 6 - 7 ) t' srrlf r t ' r l t lc r : 1, 1y (i:rrrr;r p br ir r g, s or . n' c lc ilr ly t hc di f f e r c n c e b e t w c e n Wi t t g e n s t e i n 's vicrvs ur r t l lr is owr r ; bt t t hc t c t t ds t o d i s t n i s s Wi t t g c n s t e i n 's o p i l l i o l ) a s nr('r'('ly t lr c , : c s t t lt ( ) [ - ' a s t r or ] g ir t ne r c o n f l i c t ', p e r h a p s f a i l i n g t o s e e tlr:lt rt ( ( ) ns lit t t t t ' s : t s t lllt lc and s oph i s t i c a t e d d e f e n c e o f r e l i g i o n . but ( i;r rr 111r c 1i( ) y ( : ( l t f ic s e dis c us s i o n s w i t | Wi t t g e n s t e i n ; p( ) t r c c ipr oc ar e C a r p a p 's f r i e p d l y f e e l i n g s , a l c l , Wirtll,t . r r s it , i, , ili( i ,rltt.r ;r , uv lr ilc ,: r s k c < lt hat C) ar lt ap s ho u l d n o l o n g e r a t t e n d h i s n r e e t i n g s rvitlr St lr lit ' k ut r t l W : r is t r t ann. Thi s i s h o w C a n 'r a p r e c o t l l l t s t h e
l i l ,l tt('t:

Iri

, :11 !i!

8.5 Popper's Critique the Vienna of Circle Metaphysics on Poppcrput forward two basiccriticisnrs thc Vicnnacirclc's of

'

Wlr t ' r r W it t l, . c r r s t c ir rt r lk c t l about p h i l o s o p h i c a l p r o b l e n r s ' a b o u t lnt l; krr, r r v lc , l11c ,r n11r r : t gc l t lr c r v or ld , I t l s t l i l l y w a s i t l l g r c c t l t c t r t w i t h lrir v ic r v s : r r r , l c c r t : r ir r ly his r c t lr a r k s w c r c : l l w a y s i l l u m i n a r i n g a r l d stir r r t t l, r t inl, ,f,iv c t r ; lt t lr c t ir lr c s wh g r t h c c o l l t r a s t i n We l l a n s c h a u u n q l .rn , l lr ; r s i, pt . r s , r r 1rlt t it u( lc bc c at t tca p p a r c l l t , I f o u n d t l t c a s s o c i a t i o n rvit lr lr ir r r nr os t ir r t c r c s t it t g, c x c it i n g a n d r c r v a r d i n g . T h e r e f o r e I t(.li r ( . 1( . ( lit r v lr c r r lr c br 6k c of f t f i e c o n t a c t . F r o m t h e b c g i r r l i n g ,,1 ' l' l. lt ) , , r r , W ilt llc r r s t c ir r wis hc t l t o m e e t o n l y w i t h S c h l i c k a r r d

views on scie.ce and rnetaphysics.First of all, he proposed that u-erifability should bc replaced by Jatsi,frabilityas thl critcrion of demarcationbetween scienceand metaphysics.Secondly, he claimed that metaphysics, though different from scie'c", *r, i' gencral meaningful, and might even be positively herpful to scienceirl some

178

Scienceand Metaphysics corrtradicted by singular statements' (1934, p. 4l). This point can be amplified by introducing the consideration of existential staternents. An existential statement is one which asserts that sonlethi'g exists. The statenrent 'There is (or there exists) a white raven' is a simple example of an existentialstatentent. Now the. interestin-g rhin-g is that the situation regardirrg the verifiability and falsifiability of existential staremerlts such as tThere is a whitl raven' is exactly the opposite of what obtains for universal statemellts such as 'All ravens are black'. As we have seen, the universal statement 'All ravens are black' is not verifiable by an observation statement, but is falsifiable by, such a statement - namely, one reporting the observation of a non-black raven. The existential statement 'There is a white raven', on the other hand, is verifiable by ar-r observation statement - namely, one which reports the observatiorrof a r.vhiteraverl - but ir is nor falsifiableby ariobservation statement. However many ravens we observe and whatever their colours, our observations canrrot possibly contradict trre statement that there is a white raven. This remark about existential statements brings us to the other half of-Popper's criticism of the verifiability criterion: rramely, his clainr that 'il did not excludeobviousmetaphysical statements'(1963, p, 281). What Popper has in mind here are, intcr dl,d, existential statenlelltsdrawn from religion, magic, and the occult. An example would be 'The Devil exists', or, as Popper puts it with more prcci si orrarrcl ail: dct My cxanrplcconsists the following purely cxistcntialthcory: of 'There existsa finite sequence Latin elcgiac of couplctssuchthat, if it is pro'ouncedin a'_appropriate mannerat a ccrtaiutime and place, this is i'rmediately followed by the appearancr the Devil - that is of to say, of a man-like creaturewith two small horns and oire clovcn hoof.' Clearly,this untestable theoryis, in principle,verifiablc. (1963, p. 249) This incantatio' sratement of Popper's (as it inight be cailcd) is indeed verifiable; bur presumably.it is rhe kind of occult claim that we would wish to exclude frorn science.Another example, along the same lines, is 'There exists conscious experienceafter death'.'lt is logically possible for this statement to be verified, but ir ca.not be falsified. Again, this is a sratementwhich belongs more ro religio' than to science.

cases.For Popper, the derrrarcation between scienceand lnetaphysics is not a dernarcationbetween sensearrd nonsense. It seems to me very irnportarlt to distinguish c.lrefully betrveen these two criticisrns, for the following reason. 'l'he falsifiability criterion has been subjectedto a number of quite tecllrricalcriticisrns, follows that we but everrif sonre of theseare correct, it by no llrearlri should reject Popper's viels on scienceand metaph'rsicb folo. As a il nratter of fact, ig_qgfg!q., is largely indepaxlent of the exact details of the denrarca'tion criterion. It is rnerely necessary that some, at least rough, demarcation betwcetr science and ntetaphysics shculd be possible. Moreover, this thesis of Popper's is of great in:portatrcc, siuce, somewhat surprisingly, even today, nrany philo"ophical schools di::nriss rrretaphysicsas nreaninglessor, at least, ,rndesirable. Yet arc, :ls wc Poppcr's general argunents in favour of mctaph.,'sics shall 5sg, alnrost wholly convincing. Poppcr first publisheclan extended account of his criticisrnsof the Vierrna Circle in his Logic of Sciantifc Discovery c" 1934. Auother and ReJ ilationsof 1963. useful text is chapter 11 of his Cotijectures This chapter is errtitled 'T'he Demarcatioll betwt en Science and Metaphysics', arrd rvas written in 1955 as Popper's contrilrution to t he v olurrre o r1 th e p h i l o s o p h y o f l l udol f C arrrapr r the Li brary of Living Phikrsophersscriescclitedby 1'. A. Schilpp. .io in sotne parts of this chapter, Popper criticizes thc vielvs of C;,rnap rvhiclt we irr cxpe1111d.',1thc prcvious scction. In particular, Pt ppcr fonnulatcs his criticisrn"of vcrifiability as a derrrarcationcriterion as follows; 'My criticisrn of the verifiability criterion has always becn this: against the interrtiorr of its defenders, it did not gxclude-lhvious stateffients: hut-il did exelvdetheuasJ iffiporrtnt andjllLerglLll-g metaphysical t_hat _ofrrl/ sricrrtilic -srrrtcrl_crrlr, _is_tp-say.!p_scie,ltlfig__lihggr:iel, t _ univ c r s dlla w s o f rra tu rc _ (1 9 6 3p . 2 8 l ). Let us takc ti e secondpart oI , f f is eas ie s ti o s ta rtw i t| rth e s i n rp i e p h i l osophers' exarnple of a universal gcneralization:namely, 'All r,rvensare black'. This is not verifiable by any finite conjunction of observation strtemcnts about ravens; but it is lalsifiable by observing a whitc ravell. hrdeed, the analogous generalization 'All s'rans are white' r,vas falsified in just tlris way. There is thus a Ltgical dsyt,tffietql betrveen verifiability and falsifiability, as regards such urriversal generalizations.As Popper puts it: 'My proposal is based upon an dsynlnetry betwcen verifiability arrd falsifiability; an asymnretry which rcsults frorrr tbc logical forrn of universal staterrrcnts.For

rrrese "."., ,.1,'"f':T:l,i',:-r:::;n,enrs,0,, *,';: are

.:

,E

's.

1,,

lllo

t
t I

S c i e n c e n d Metaphvsi cs a

ls Metaphysics Meaningless?

181

A l tt' r tl rc s cp rc l i ru i rra ri e sl , t u s l l ow turn to P opper' sfornrul ati on e ol l ,rl rrl i rrl rrl i ty ;ts ;r l c n ra rc a ti o n teri ol r. Thi s j s how hc puts the c cri
i l t,tl lt' l

r r ' or r r , r, nt , . , n( t t r r < : lnpir ic al l yv c r i f i a b l e . I f w c w i s h t o a v o i d t h c l'\ t t lv i\ t ' s lr r is t uk t :of c lir ninat in g , b y o u r c r i t c r i o n o f d e r n a r c a t i o n , tlr r , t lr lor r , t ir ' ; r ls y s t c r r r sof nat ur a l s c i e n c e , t h e n w e n r u s t c h o o s e a { nt r ' r . r wlr ir ' lr r r llr ws us t o adm i t t o t h e d o m a i n o f c m p i r i c a l s c i c n c c {'v r , nr l, t l( ' lt ( . nls wlt ic ll c anuot bc v c r i f i c d . llr r t I r lr , r llr ' t . r t ; r ir r ly lnr ita s ys t c n r a s e r n p i r i c a l, r r s c i c r r t i f i c r r l y i f a< o fl f'r Lllr;ll)l(' rrl' bcirrg lestcd by experictrce. Thesc colsidcrations brllllr,('rt r;rnrot tltc ucri.linbility tf but thefalsifability of a systent is to be t, r k lr r ; t s , t r r it c r ior r ol' t lc r t r ar c at i o n .( 1 9 3 4 , p . 4 0 ) Nn tr t lr , r t l' o1r 1r 1.t1 : s lr ot hc r e s p e a k o f a t h e o r y , b u t o f 'a s y s t e r n '. lot 'llrrr ' r , , , r n ir npolt ; r lr t lr oir r t , t o w h i c h w e s h a l l r e t u r r l l a t e r . N o l v , Ir.rv r ' r ' r ' t , I lv ill t ullr t o l) oppc r ' s s e c o n d , l n o r e g e n e r a l , c r i t i c i s m o f tlrl Vic r r t r , r ( lir r ' lt ' ,

'lf

H
I

I|

H
t

t+ l{
tt

'f f rl Virrrrr;r ( lirr:lc rcgarclcd verifiability as a nteaning criterion. |. lr l' (,t, l ry r o n tr;rs t,l l u t fo rw a rc {fal si fi abi l i tyas a demarcati on,but r , r r t .r n l (' ,u l i rr1 1 ,.ri tc ri o n .In d c ed, Fopper al w ays mai ntai ned that < rrr,nI y | | l(.t,t rysir':r s(ilt(:lncuts are perfectly meaningful. pl I lt o l l l r(,r l r;rs l r s i rrrl rl c , b u t pow erful , argumel l t to show that t r t ' t t l rc tv c r i l i ;rl ri l i ty n o r { h l s i fi a bi l i tyare adequate meani ng cri teri a as ( , 1 l fl l l l , p 1 r. l ' l l tl ).It d c p c n d son the pri nci pl e that i f a statement, lr , rq rrrr' ,rrri l rl i l rrl ,rr,rri ts n c g a ti orr,l rot-S , must be meani ngful . If tl r v r ' l ,rk l S t. l rc l r rrrri v c .rs a le n eral i sati on, e get the i nteresti ng g w r lr r r l t tl r,rt i tr rrc 1 1 ;rti o l n o t-S, i s an exi stenti al stateurcnt. More r, r 1' r ' rl tr,rl l y l v r' l t;rv cth t: fb l l o w i n g: r S rro t S A l l r;rv (' l rs rc b l a c k a l t i s rro t tl rc :< :a s c at al l ravensare bl ack th ' l ' l rc rc i s (o r c x i s ts ) a non-bl ack raven

for science.This consideration provides further argume'ts for the clainr that.m.etaphysical speculaiions ofren are ,in." it seerrrs unlikely tharsuch speculations -.r?ingfui, could be helpfulForscie'ceif -rrrir, they were,nreanirgless mere nonsense. or Before .ottring io lrorvever,l wrll presentin the sectiorl a' ergumerit of nry 'ext ow, for the thesis much nretaphysics in fa"ctnr.rni,.,gfrf, is ,that Curiouslyenough,this argument based sonleid"rs of popirr,, is on principalopponent,rVittgenstein. At fir"t sight, nothing could seemnrore paradoxical than to use somc of Wittgenstein's. ideasin supporrof bopper. eopp.., ,, *. have seen,criticizedwittgenstein's tractatus in'a decisive^*.y. rn. meetingof Popper-a4d Wittgenstein.at Cambridge(deJcribed Irl:t ir 1.5)hasa'almosr mythic quality;*but, wharever ilroo.".a, it is clcarthat the two,famois philosopheis not pr.,'o', did "*..if ii.naty ternls. Now, admittedly,wittgensteinhimserfcriticized his earrier Tractatus views, and developed quite differentapproach hl, i.r.. i ;,, work, thc Philosophical poppcr, -liowever, was, Inuestigaiions. if anything, even lessappreciative-of later-work than he this had been of tlre Tractatus. lndeed, he went so far as to say in a radio pro_ grammebroadcast rhe wi'ter of r970-r: 'Russeil in readthe iniorop.llical Investigations withour dcriving a'y enlighrenmenr rro* ii. so did.l, I rnustadmir' (Magee1ed.), lVll,' p. 13b).yct, when the dust settcso' a heated phirosophical controversy, is somctimes it possible to seethat the embattled.opponents more in common had thr,, they tlrcrrrselves realizedat rh; time. In his philosopt;rot iir-rrrigoii.io^ V/ittgenstein d.evelops r.r* theory of meaning. I will Bii.ny ". this theory, and then argue that it can bI used ro J.rr"a lxpou'd Popper's thesis that ma'y meraphysical theories m.rningrri.' rr.

8.6 Wittgenstein,s LaterTheory of Meaning \JUittg.cnsrein's theory in his phitosophical Investigatiorts that the is meaning of a word is given by its use in a lan[uag._g",rr"By , 'language-game' meanssome kind of rule_guid"ir.?iri..A*ry he i' which rhe use of la'guage plays an essenialpart. He |rl_rrlf i'tr'duces the co'cept as followi: 'r shall also'call ttr" *i,ot., consistingof languageand the actionsinto which it is woven, the "language-garne"' (1953,sec. 7). A'd again:'Here trre term "language-game" meantto bring inio promirie'ce is the fact tr,,rtih. speaking oflanguage partofan activity,or ofa form oflifei is tr... Z:1.

t* H
fr tr

Now , ,ti rv c l r;rv c ;rl rc ;rd y o i rrt cclout, S i s fal si fi abl e,trut not-S i s p lnt l ,rl ri l r,rl rl t' ,' l ' l rrrs , i l ' w c w ere to adopted fal si fi abi l i ty as a t lr ' , u u n l l r ti l t.ri trrr,w c w o u l c l h a ve to say that S i s meani ngful and llt , t l tro l S i ,i rttr' ;ttti trl ' .k ' s s ; th i s , to say the l east,i s hi ghl y courl terb rrt ir r t n rtrv r' , rx ,rrtl y tl rt' s :rrrrc ;l rg unl cnt be used agai nstveri fi abi l i ty l can . r , ,, r rrrr' ,rri rr1 1 tc ri o rr, s i n c c rro t-S i s veri fi abl e,w hi l e i ts negati on, r ri nol n r| l S S, i s l ;rl s i l i ;rb l t: u t not veri fi abl e. b lrr tl rc rrrx t rl r,rl rtt' rw c : s l ta l l corrsi derP opper' s vi ew that rnetatrl l' lr \ ' ,.rr,rl r' ;tr,rrrrll l rc o ri c s c a rr b e - and i ndeed have been - useful

l,;iri'l:

r ,',tr:il::

182

Scienceand MetaPhYsics

ls Metaphysics Meaningless?

183

by wirtgenstein illustrateshis concept of languar.-ganle-. his famousexampleinvolving a bossand a worker oil a building site. The bossshouts'slab" for example.and the work rr has to fetcira of slab. wittgenstein'spoint is that the rneaning tl.r:rvord-'slab'is given by itt,tt. in the activity carriedout by bossarrdworker' the Inv with Wittgenstcin's stigatiotrs ory I anrLroadlyin syrrrprthy to seems r r: inappropriate' 'latrgtta ge-game' but of language; his tertn at 'ganre' nr,t a garrre all, is of Irr {hci, his ti.st exarnplc a languagc I prefer iSereforeto speak i. ger.reral 'lang'agebut rvork. _of for activities',reservingthe ternr 'language-game' those activities enough,the 'rneaninB.=use' which really are grtt-,.t. Irrterestingly by tlreoryis partly anticip'tated Wittgensteili1 the T; utatuswltcrc hc rvritcs: 6.211 (ln philosophythe question,'What do we actuallyuse I:ads to valuable this word or this propositionfor?' repeatedly insights.) : how Wittgenstein's'meaningu:e'theorycanbe Let us rrow see are I statements that nretaphysic: thesis usedt0 defendthe Popperian ple who. ncet in general nreaningful.hnagine a group of..pct tnetaphysical after'oons,say) to discuss .r"gr',i"rly(o' Wec{ricsday Catholic theologiars; [:r exatnple,or q.lrtio,ir. They c.uld be They will certainlyus9 a l rrgeflumber of philosophers. I-legelian 'grr)undof being" suchas 'essence', ,.rtrd, .,ti .,,rtu, .tlialectic" and so "*p."ttions, on. Y"t, this discourseis not arbitrary,,but r A rule-guicle<i. begirrnerwho uses an expressiot ittcorrectly is if irray evcn be ostracized he ':r she docs trot repriiiranded,"'and co,rforrl. within thc group it is well known ll'hc .rrc thc cxpcrts )ect,atrdso on' artlistenedto with nrostre:; wfiosepronoultcelnclts (the rve havea language-gamc tcrnl 't utue'is perhaps Hcre, zurcly, vity, i' nlore appropriate t6is ilrsia'ce),a rule-guicled.ac: or a fornr have words and e-rpressions of lite.^Wiihi' this la'guage-game, by rules and ccrventions. on a use which is circurnscribed wittgepstein'slater theory of meaning,therefore,;ve lnust surcly and have tneanin-.4, that t6e ,.y ihr, theseworcls and expressions is discourse nrearringful. rnetaphysical himself have agreedwith this use of 6is Wtuia Wittgerrstein Thcre t6iory of mea'i'g to supporthis rival Popper? Itwestigdtiotts 23 In havedorre. section that which suggests he might ir.on.'prsrage phiiosophicat he reviews the multiplicity of Iniistigations of the T by language-ganres giving r lottg list of examples. his list includes

ii
ii

'praying'. This suggeststhat Wittgenstein would have included religious cerernonies language-games, so would have been as and comnritted to the view that religious discoursein general, and thcologv irr particular,was rneaningful. Thcre are a few indications(partictrlarly passage just cited) the tl-ratV/ittgensteinmoved, in his later period, towards the view that rcligior,rs and nretaphysical discourse wcre meaningful.But these indications few, and my overallimpression that Wittgensteirr are is irr tlre P/rilosophical Investigations arrd large stuck to the T-ractatus by view that rnetaphysical stat-ements meaningless. are Thus he saysin sectiorl 464: 'My aim is: to teach you to pass from a piece of clisguiscd nonscnsc somethirrg to that is patentnonsense.' This is very similar to the passage from 6.53 of the Tractatus citcd ear:lier, namely: 'The correct merhod in philosophy would really be the following:. . . whenever someoneelse wanted to say something nretaphysical, demorlstrate him that he had failed to give a ro to meaningto ccrtainsignsin his propositions,' This situation lcads to what I regard as a central flaw in Wittgenstein's later philosophy. The view that rnetaphysicsis nrearrirrgless does indeed follow frorn the Tractatus theory of nrcaning,lrut that theory of mcaningis inadeqr"rate. thc PhilosoIn phicalInuestigations, \Vittgenstcinputs forward a rnuch bcrtcr thcory of rncaning, but makesthe mistake holdingon to his old view of of the senselessness metaphysics a view which is actually inof compatible with his new account meaning. of Such, at all events,is my own opinion; but some defenders of Wittgenstein might srill think thar the nonsensical characer of rrrctaplrysics could be defended tentrs of the Inuestigations in thcory of meaning, despite the considerations given earlier. Let us next cxarnine how this could be done. Now it will be remembered that Wittgensteilt's first cxampleof a languageactivity was work on a building site. The boss shours 'Slat-r', and a worker goesoff and brings a slab. This rule-guided socialactivityof bossand worker givesmeaning thc word'slab'. to Let us contrastthis with the metaphysical discussion group which nleets on Wednesdayafternoonsand considerssuch questionsas whether the essence the ground of being impliesexisrence, so of and on. Perhaps.Wittgensteirr might say that words acguire meaningirr the practicaleverydaysocialactivitiesof the building sire, but not in the purely theoretical discussions groupsof philosophers. of There are indeed hints of such a view in his writing - in the following passagc exanrplc: for

t,,
lt T
t l*
I

184

Scienceand Metaphysics

ls Metaphysics Meaningless?

185

csscrttill, itr Wc arc untlcrthc illusionthat lvhat is pcculirr,profoutrtl, our invcstigation, rcsidesin its trying to grasp t;;e itrconrparatrle of That is, the order existingbetween concepts dre cssence languagc. word, proof, truth, experience of proposition, , and so on. This ordcr is a .rrpcr-order bctwccn - so to spcak- Jt,pef-collccpts. Whercas,of 'expcricncc', 'world', havc a use, it corlrsc,if tlrc words 'languagc', 'tablc', 'larnp', 'door'. rnustbe as lrrlrnblca ollc as that of the rvorcls (1953,scc.97) I t loo k s h c rc a s i f a g c rru i rrctts e of a rvorcl can otrl y bc a ' hurrrbl c' The view we are considering has a certain populist flavour. Words acquire genuine nreanings in practical, everyday social activities in rvhich workers heave slabs across building sires. The purely thcoretical language-gamesof (effete?)ilrtellectuals do not suffice to give their abstract terrrls meanittg. As we have already rernarked, Wittgenstein seens to have been attracted emotionally tor.vardsa populisrrr of this kind. He admired Tolstoy's eulogies of virtuous peasarlts, and disliked the conversation o[ his fellow from Karl Britton's 'Portrait of a The following passage acaderrrics. Philosopher' shorvs a personal attitude not unrelated to the philosophic;rl vicrv we arc rrow cotrsiclcrittg: Hc h;rd,hc said.only oncebecnto high tablcat Trinity and the clevcr convcrsation the dons had so horrificd hirn that hc had colnc out of with both handsover lrisears.The donstalkedlike that only to scorc: they did not even enjoy doing it. He s::id his own bedmakcr's and about the private lives of her prcviousgentlemen conversa,pion, at hc aboutlrer own farnily,was far preferable: least could undcrstand why shc talked that way and could belicve that she enjoyed it. (Quotcd frorn Pitcher, 1964,p. 12) sympathies we nray have with the populist contrast between workers, wlto are 'real people', as agaitrstdotrs, who are 'atrophied intellectuals', the theory we are considering is untenable, as the following couuter-exarnpleshows. Let us supposethat a piece of pure mathematics is developed by a group of pure mathernaticiarrs.It is thcrr taken up by a group of theoretical physicists, and uscd in thc crcation of a new physical thcory. Finally, this thcory is used in a practical application - perhaps even in house building. Supposc, furthcr, that wc adopt thc view that a tcnrt is in rncanirrgful only if it is r.rsed a practical, everyday social activity and not rvhen it is used in purely theoretical discourse. It then follorvs that thc tcrnts of the nrathcnraticaltl"reoryare Irreatrirrgless 'lVhatever

li

T
t
;

ll

il il J
T
I"
li

demarcation, great interest.The two of distinctionsare not, however, the same,as I will no* da_orrrt."r.. Let us begin with an exampleof a scientifictheory *rri.r, i, ,ri'"r.a i' pracrice,arrdth'r 89 on to an exampleof a trr."r/ *-lri.r.,i, i' practicebut is not,scienrific.. "r.a Einsteirisgeneraltheory .ir.rrri"iry was introducedi' 1915,but has been urJd i,t practicai rpfrl."ri",,, only re-ce'tly.For more than fifty years,the.efore,,lri; rh;;;;;, , scie'tific-theory,but one-notusedin practice.Converr"ly, liir.u.ry easy, find examplesof theorieswliich are not scientihc, to bur are usedin practice.consider a tribe which hasa mythotogi.rt th.o.y of the elements,and usesthis theory in its rain-makingianc.s and

yet the. practicar/pur.ly is!rye@t. ure.scrence/metaphysics

then usedby Einsteini'his generar theoryof reraiivity-Jigi5. irr. general theory of relativiry ha.srecently been appri.a to p...ti..t problemsconcerned witrr satellites. what stageiid th. kJy t.r,r* At of the tensor calculusbe;ome1mea'ingful? It ieems clerr tliarih"y were meaningful throughoutthe p.o."rr. therefore. use rse to disti

d.y.Jop:{.rr.a piece ly-iruo lj:i].r;lt I,y1:,a'd Levi- civit i, ilr a. .of pure,rrath.,,rrtii, rtarl a' s' ( r ccr
ir r r por t antpapcr of r 90r . it *r ,

while.the theory is bei'g developed the pure by nrarhernaricians, that they rernain mea'ingress*i.n ,ir. nraihenrarics ir," physical.theory, then suddenlybecom; *;;,,s_ "lrJ bnt :::rj:_,11"^r:w rime Iur lor the frrsr wherrthat trreoryis appried rrouse to buildi's. Sucha corlsequence seems rne quitt unacceptable to Our-hypothetical example-is, .ou.r",,ro nr... farrtasy, is ol ' but quite close to solne actual historical cases.Fo. .*"rrrf l.i *,rro,

il

:ll.^ir,*^r:ll,llil,ltr.,tces. whrch rs unscientific,but which is used in practici. As this may seem to exhibit a rarher patronizing attituie rowards prinriiiu. peoples,it should be poinGd out thlt there may *.il u'.-,;nrirr. examples in our own society. politicians rcguiarly ,rr" auriou, eco'ornic theoriesin decidinggove'rment poliJy, ,ni ir i, ,i*r'uy no rneans clearthat theseeconomictheoriesare scientific.
8.7 The lnfruenceof wittgenstein's Life on his phifosophy Having now co'sidered someof wittgenstein's phirosophicar views, let us rerurn to the intcrestingquestiJnof how these ,rc i.i.,.i' ,"

Herewe have exanrpte a theory an of

186

Scienceand Metaphysics

ls Metaphysics Meaningless?

187

his strange life. The nrost striking thing about \[l t:tgensteill a _as philosopher is that he produced two quite distin:; philosophies' And what What lid him to abandon his views in the Tract,:'.us? provided the stirnulus to produce rhe nerv philosophl to be found in Von Wright has this trr say in attswcr Investigations? iis philosophical to these questions: r's, of great itnportanccin the originationof wittgerrstci new iclcas by views were subjcctcd two of was1le critiiis'r t' rvhichhis earlier w3s a whoscprqn;rturcdcatlrirr 193() was Rarrrscy, his fricuds.onc licavy loss to coptcllrporrrythought. The 6thcr was P r:ro Sraffa,arr Italiair econor'ist rviro had co'e to Carnbridge s:ortly beforc \li ittgenstcinrcturttcclthcrc. It was above all Sraffr's acutc atrcl to Wittgenstcin abantcn his carlicr compelled fo.c"ful criticisnrtl.rat with views and sct out upon nclv roads.Hc said that his di: ,:ussions had all branchc,; bccncut. nradchirrrfcci like a trce fronr which Sraffa grecnagainwas duc to it: ow1 vitality. T|at this trcc could becornc likc fro .n outside did larcrWitrgensteiu noi rcccivcan inspiration Thc rhatwlriclr rhc carlicrwittgcnstcingot frorDFrcgcand ilusscll.(l95tl,

p.s)
Nolv, rvithout wishing to deny that thc criticistus of llanrscy arrd Srafla lverc of sourc intprortatrce,I would likc to s lggest that the prirrcipal stinrulrrs for changing ltis philosophy canrt:front.Wittgetriteiu's life expericnces berween 1920 and 1929, rnd that these experiellces wire the sources of the vitality rvhi, h epablcd thc Wittgenstciniall tree to become greett agaitr. Thcoric Gratr,teurcrFrcund,i$,alle goldnerBautn' Und griin dcs Lebens Part I Goetllc' Faust, (All theory, dear friend, is grey, and green the gold'::r tree of lifb.) wlrerr wittgepstcin began r.vriting the Tractatur, his prevtous trai'ing h.,d blerr irr the physical sciences,engineerlng, and logic; and hii' v ierrs a b o u t th e n a tu re o f l a n g u agerefl ecttl ri s backgrorurd. L,anguageconsistsof elenrelrtarypropositions wfiich picttrre reality and'-wliich are glued together by logical conne(r;ives to fort' co'rplex proposiiions. Mianingful propositions ar( either logical cor*radictior-ri or logically valid propositions or ::mpirical p..rpositions of thc naturil scienccs.A model for all this is provided by ' ^Iilewtonirn nrechanics, for Wittgenstein hiursel: says: '6'341

Newtonian nrechanics, for exarnple, imposes a unified form on thc description of the world.' It is not surprising that the Vienna Circle, absorbed as its members were in the study of logic, mathematics, and science, should find this attracriveand plausible.-lJutin the 1920srhe author of the T'ractatus was involved in trying to teach language to young children. htdeed, Wittgeristein's next publication after the Tra*atils was his wiirterbuch fiir volksschulen(1926). Despite the unforturrate end to dre author's career a-sa teacher, this was officially approved as e sclrool textbook. (For more details about the book and how it was c<rrnpiled,see llartley, 1973, p. 77.) Now Wittgensteili could hardly have failed to notice that the T'ractatus theory of language was not vcry plausible as an account of how his young pupils used language. Moreover, there are sigrrs every*'here in the Philosophicat Investigalionsof the influence of Wittgenstein's years as a schoollnaster. Thus, already in section 9, when Wittgenstein begins a discussion of the natural nurnbers, hc remarks: 'Children do lcarn thc use of the first fivc or six cardinal nurnerals in this rvav.' Sections 156-78 are devoted to a discussionof reading; while in sectiolr 185, where Wittgenstein introduces what have come to be knowu as thc rule-following considerations,we find the following; Now - judged by the usualcriteria- the pupil hasmastcrcdthc scrics of rratrrral nunrbcrs.Next we teachhirn to write down other series o[ cardinalnun'rbers arrdget hirn to the point of writing down series of the forrn 0, n,2n, 3n, ct c. at an orderof t he f or r n'*n'; so at t he or der '*l'he wr it esdown t he seriesof natural nurnbers.- Let us supposewe have done exerciscs and givcn him tcstsup to 1000. Now we gct the pupil to continuea series (say +2) beyond 1000andhe '*'rites 1000,1004,1008,1012. We sayto him: 'Look what you'vedone!'- He doesn't understand. Wc say: 'You werc nlcant to addtwo: look how you bcgauthe scrics!' 'Yes,isn't it right?I thoughtthat was how I was rrreant - Hc arrswcrs: to do i t.' The exalnple is, of course, fanciful, but it does not sound too far rernoved frorn Wittgenstein's own classroom experience. The influence of Wittgenstein's work on his sister's mansion is evell nlore obvious, since, as we have seen, the very example with

:.i.r.:.:,: ,i: iri

I
t t t
I I

llt lt

a S c i e n c e n d M e taphysi cs

u lr r rlr W i ttl l t' rrs tc i tti tttto c l u c e sh i s concept of ' l arrguage-ganl c'i s llr , , l ol ;r Ito s s u rrd a ' u v o rk c r o l l a bui l di ng si te.' fhus, al though V / r t t lilt ts tc i n l l l l l c rrrs to h a v c a b a n donedphi l osopl r, for a w hi l e i n r r r , lr ' rt r r rl rr o l l rc r tl ri rrg s ,th e s eo th e r thi ngs w ere ?,:tual l ya ki nd of i o I ' r r ' lr , r r ,rtrulrr t l ri s l ;rtc r i tttto v a ti o ttsn phi l osophy.

I
Metaphysics Relation in to Science: Viewsof The Popper, Duhem,ord Ouine

9.1 Popper Metaphysics Relation Science on in to

One of thc nrost important aspects of Poppcr's critique of thc Vienna Circle is his insistence that metaphysics can be not only nrea'ingful, but even of positive value to science.Popper holds that theories may start life as metaphysical, but then corne gradually to be tra'sforncd into scienrific hypotheses. In the following weilknown passage, he compares this process to that of particles in a liquid being gradually deposited ar rhe bottom of the contairrer: To obtain a picture or modcl of this quasi-inductive evoluticl^ of scicnce, the various ideas and hypothesesmight be visualizcdas particlcssuspended a fluid. Testablescience thc precipitationof in is thcseparticlesat thc bottom of the vessel: they settledown in laycrs (of universality).The thickness the depositgrows with rhe numbcr of of thcsc layers, every new layer corrcspondingto a thcory more universal than those beneathit. As the result of this processidcrs previouslyfloating in higher metaphysical regionsmay sometimcs l,e reached the growth of sciencc, by and thus mlke contactwith it, rrrd settlc. Examples of such ideas are atomism; rhc idca of a single physical 'principle' or ultimate element (from which the othirs derive); the theory of terrestrial motion (opposed by Bacon as fi-ctitiotrs); agc-o)dcorpuscular'iheory the oflight; thc fluid-thcory of clcctricity (revivcd as the electron-gas hypothesisof rnctallic ionduction).All thcscrnetaphysical concepts idcasn.ray and havchclpcd, cven in their early forms, to bring order into man's picture of thc world, and in some casesthey may everr have led to successfirl prcdictions. Yet an idea of this kind acquires scicntific statusonly rvhcn it is prcsented falsifiableform; that is ro say, only when it in

I|
|l

F H
I

190

Scienceand Metaphysics

Metaphysicsin Relationto Science

191

bccorrrcs possiblct<>dccirlccrnpiric:rlly hctwccn it ;rrrtlsonrt' r.iv;rl ( t lr c t r r y . l( ) . 1. 1, 27 7 - U ) pp. Of the various examplcs which Poppcr cites ill this passagc,thc nrost striking, perhaps, is that of iiomism. Atornism *.r firrt introduccd in the West by the pre-Socratic thinkers Leucippus arrd l)crrrocrirrrs.It contirrucd as a powcrful trcnd in tlre aucic:rrt workl with l!picurtrs in Glcccc rrnd Lrrcrctius in llonrc. 'l'his alrcicrrt :rtorrrisrn nrust, I think, bc classed as nrctaphysical ratlrcr tlrln s c ic r r tfiic . Aucient atonrism was revived in Wcstern Europe in thc scventccnth century, and discussed thc lcading scierrtists thc day. lt by of should still, at thrt tinrc, bc consiclcrcd a rnctrphysicalr;rthcr: as than :r scicntific hypothesis. At the beginning of the nirreteenth ccntury, I)rltou rcintroduccd atourism to solve sonrc problcn.rs chcnristry; in wltilc lralf-w:ry through thc rrirrctccrrth ccrrtury, Maxwcll brouglrt atorrrisnrinto rnathematicalphysics in connection with the kinctic thcory of gascs.lly thc cnd of the ninctccnth cclrtury, atonrisrrrcan dcfinitcly be considered as a scientific hypothesis; but this scientific dcvelopmcnt would scarcelyhave bcen possible without the earlicr history of atonrisnr as metaphysics. This exarnple constitutesyet another argument for thc thesis that rnctrphysical theories can be perfcctly nrcanirrgful. Arryonc who tnaintains that mctaphysics is always nreanirrgless faced with somc is difficult clroiccs. One option would be to say that atornism was always scicrrtific, and so meaningful; but then it has to be rnaintaincd that thc atornism of thc ancient Greeks was scientific, and this is hardly plausible. The other option would be to say that atomism was meaningicss from ancient Greek tirnes until soll1e point in thc ninctecnth century, when it became scientific and so meaningful; but ther it must be held that a perfectly meaningless theory was somehow used by Dalton, Maxwell, and others in ordei to formulate

vle w o n m e ta p h vsl

a- non:t6table

e savs: 'Atomism is an metaphysical theory whosc

influenceupon science has exceeded that of many testable scierrtific theorics' (1983, p. 192). After giving some further example of nlctaphysical theorieswhich have influencedscience, continues: he

'F,ach of thcsc trtctaphysical thcorics scrvcd, beforc it becalne t c st;tl tl e ,:ts:t t'csc;tr t'l t ) r o [{r ;l l l tl l l c l i l r sci cttr :c,l t' i r r tl i c;r tctltIc l dircction of our scarch, and thc kind of cxplanation that might satisfy us; and it made possible something like an appraisalof thc dcpth of a thcory' (pp. 192-3). This is an inrportant passagc' bc."ur" it ilrdicates rhe heuristicrole of metaphysics in guiding the constructiorl of scicntific hypothcses. Such hypotheses almost .lw:rys crrlcrgc irr tltc coursc of a scit:ntistor a group of scicl)ti$ts progralnmc, _andsuch researchprogralnmcs are pursuing a re.scarch "guidcd by iornc gcncral (or mctaphysical) principlcs.and usu;rlly wliich indicatc thc kind of spccific hyp'thcsis which should bc iclcas clcviscd to explain existing facts and to be testcd against flurther obscrvations attd cxpcrinrcnts. TIrus thc gcncral idcas of atomism guidcd l)alton in constructing a hypothesis-to explain some facts of iircr'ical cor'binarion, and guidcd Maxwell in trying to account for tlrc otrscrvcdrclxtionsof prcssurc,vo]umc, and tclllpcraturc irr gascs' Without thc mctaphysical idcas of atomism to guidc thcir rcscarch progrxrnnlcs, ir is vcry doubtful whcthcr Dalton or Maxwcll could iravc dcviscd thcir spccific scicntific hypothescs' Popper was undoubtedly right to rehabilitate metaphysics against thc stiictures of the Vicnna Circle, but it still seems to me that some rcstraints should be put on enthusiasm for metaphysics' Many thcories - for example, Lcibniz's Monadology - have nretapl.rysical rrot iffcctcd scictrcc vcry tnuch onc way or thc othcr, whilc somc lnetaphysical theories have perhaps irnpeded, rather than helped' science.-For example, religious theories of divine creation and the humau soul werc (and pcrhaps still arc, to somc cxtcnt) obstacles to the Darwinian theory of biological evolution. So far we havc discussed the influence of metaphysics on science; but, conversely, science can influence metaphysics. I will conclude this scctiorr with a brief discussion of this opposite influence. An obvious example here is the debate bet'ween determinism and indetcrminism. Laplace adopted a strong version of determinism, and was clearly influenced by the successof Newtonian nlechanics. Converscly, the discovery of limitations on Nervtonian mechanics arrd the adoption, as a futtdamental theory in physics,-of quantum rnechanics, with its probability and uncertainty, has led to a revival of irrdetermiuism. Indeed, Popper has written a book, The Open (Jniuerse (1982), arguing the case for indeterminism. Some possible misconceptious on this matter must, however, be avoided. I rm not maintaining - and it is not the case- that a new scientific theory logically implies a new metaphysics- Popper, for

;
192 and Science Metaphysics Metaphysics Relation Science in to i93

i, : jl

ii
lir

example, argues convincingly in the book just cited that Newtonian mechanics, even if accepted as the fundamental theory of physics, does not necessitatethe adoption of metaphysical determinism. Yet there is no doubt that Laplace was influenced by the success of Newtonian mechanics in formulating his views on determinism. Conversely, Popper does not argue from quantum mechanics to metaphysical indeterminism. Indeed he writes: 'My refutation of l'scientific" determinism. . . nowhere makes use of probability theory; nor do I appeal to quantum theory. "Free will", too, is mentioned only incideirtally. . . . My argument holds for every physical theory, however strongly deterministic it may appear' (1982, p. 106). Yet Popper, as well as other recent advocates of ofquantum indeterminism, are undoubtedly infiuenced by the success mechanics.

ir'
il

!,

9.2

Duhem and Ouine on the Status of Metaphysics

i:

il;; iiir
iri

il

ii:' ,ill
.ti lil

iri'

From the very beginning of his classic The Aim anil Structure ol Physital Theory, Duhem presupposesthat metaphysical systems cxist rnd that these systcms, though distinct fronr those of sciencc,rrc norretheless rnerningful. I say advisedly that Duhem presupposes tlrese doctrines. since he nowherc states them. but rather assllnlcs them rvhen arguing for some other thesis.Thus part 1, chapter 1, of T'hc Aim antl Stntctureis entitled 'Physical Theory and Metaphysical Explarratiori". Here Duhem introduces metaphysics in explicit contrast to physics, in the following passage: Now thesctwo questions Does thcrecxist a matcrialrcality distinct and appcarances? What is the naturcof this reality?- do from sensible not have rhcir source in experimentalmethod, which is acquaintcd and can discover nothiltg beyond only with sensibleappearanccs thc thcm. The rcsolutionof thesequcstionstranscends methodsuscd (1904-5, p. 10) by physics;it is the object of metaphysics. Duhern goes on to argue that theorctical physics should not be subordinated to metaphysics. His reason is the followirrg: 'lf theoretical physics is strbordinated to nretaphysics, the divisions separating the diverse mctaphysical systelns will extend into the domain of prhysics.A physicrl theery leputed to be satis{:rctoryby the sectarians of one metaphysical school will be rejected by the partisans of

'ii
:rrl ii1

rnother school' (pp. 10-11). He thentgoes on to consider an cxample: nanely, the theory of the actioii of a magnet on iron. With a characteristic wcalth of historical detail, he shows how attemprs rvere made to subordinate this physical theory to differenr systems of n.retaphysics, such as Aristotelianism, Boscovich's point-atomism, material atomism, and Cartesianism. These attempts were all, so Duhem argues, unsatisfactory. I do not warrt to consider here whether Duhem is right or not in making this particular claim. My poin-t is only that he assumes, throughout the discussion, the existence of meaningful metaphysical systcms such as Aristotelianism, atomism, and Cartesianism, which are distinct from the scientific systems of physics. Today it may sccnr surprising to make assumptions of this kind without any discussion or comment, but it rnust be remembered that Duhem was writing bcfore Wittgenstein and the Vienna Circle launched thcir attack on mctaphysics, Duhenr wxs a man well versed in the history of both philosophy and science, and was familiar, rherefore, with a considerable number of diflerent metaphysical systems which rvould have appearcd to him as obviously both meaningful and distinct from scicnce. Doubts about mctaphysics arose later from thc logical analysisand theorics of n'reaning Russell,Wittgenstein, of autl thc Vienna Circle. For the srrl1crcason, Duhem does not artcmpt irr his (1904-5) to fornrulate a criteriorr for demarcating science from metaphysicsthough hc assumcsthat the two are distinct. The problem of finding :rn adcquate demarcation criterion in terms of some concept such as vcrifiability or falsifiability arose out of the attempt to analyze the structurc of scicntific theories with the tools of formal logic. Duhem rc:rsonsvery logically, but he uses logic in an informal fashion. He norvhcrc employs the formal logic of Frege, Peano, and Russell (see clraptcr3, note 4). Lcr us next compare Duhem's views on science and metaphysics rvith those of Quine. Once again it will become clear that the views of the two thirrkers are quite distinct. Vuillemin puts the point as lbllows: 'The D-thesis . . . arose fronr historical consideratitrrs .,rd only physics fell under its scope, metapl'rysics and sciencebeing kept complctely distinct, Quine . . . refrained from drawing any sharp distinction bctwcen scienceand rnetaphysics'(1979, p. 609). This is curtainly correct. Dr-rhem, as wc have seen, clearly assumes that scicncc and rnctrphysics arc distinct, while Quine has this ro say:

:-

'&,!h clogrl.".-t shailrrsue. arcill-founded. mffE,l;AE;;ei-""-

irrg thcrn is, :rs wc shall sec. a blurring of thc supposed boundarv

Metaphysics Relation Science in to

195

i T

t
||

trct_yscjJ,.spgcr,rlative rtrctaphysics narural alrC scicncc' (1951 p. 2t)). , ytt ,,rI lssuc,lny own synrpathies witlr Duhenrratlrcrtharl are -Typical nretaphysical ihcorics sucir as plato's th"o.y oi Slll:1" or thc t'().r.rs doftrine of trrc Tri.ity do sccnrto havc a criffcrcnt typical scicntific thcoricssuch as Lavoisicr,sthcory :.1:,,11..1:l ll:1 corrrbustrorr Bohr's theory of thc atonr. r'hc bourrcrarv or 't [)ctwcctr rrrctaphysics scicnce thr-rs arrd is rcal,,",tr.-,tt"r,,-,rJil;i Ol'coursc, thc boundaryis not entirclysharp,ancli,rt".m.ji.t" o. doubtfirl casesdo undoubtedlyo..u,; but 't'his is true of nranv .listinctiorrswhich are noncrhcress perfcctry ,rJ ,rr"ri.-'i,1 "i"ur" cvcryday lifc wc can easilyenough iisti,rguirh berweenthor" *ho are bald a'd thosewho are ;ven thoigh it is hard to s"y ho* 'or, trr;rrry hairsfl nlan lllr,lstlose beforehe bccoirres bald! Arothcr corrsideratiorr this subjectis tbc followi,g. .5onrc Qrrinc's of nrodes -rerevant.to .rgurbly q"la.rriri*ai,rE. of expr,ession^rr"

and Popper the Influence on of 9.3 Duhem Metaphysics Science on

I)uhcnr's vicws on rlctaphysics and scicncc arc in fact nruch morc like Popper's than Quine's. Duhem and Popper agree, in contrast to Quinc, that science can be distinguished from mctaphysics, though I)uhem docs not in his 1904-5 attempt to formulate a dcmarcation critcrion. Duhem is perhaps a little less enthusiastic than Popper about metaphysics in relation to science. As we have seen, Duhem strcssesthat physics shotrld not be subordinated to metaphysics. On the other hand, he- does admit that metaphysics can influence scicnce, and gives many historical examplcs, some of which overlap those considered by Popper. We will next consider a few of these l)uhemian examples of metaphysical influence on science, as they are both intcresting and instructive. Let us start with Descartes' metaphysical views about matter, which, Duhenr argues, later influenced the hydrodynamics of W. Thomson and Maxwell's theory of magnetism. This is how Duhem describes this example: According to Descartes,matter is essentiallyidentical with the cxtendedin length, breadth,and depth, as the language ofgeometry gocs; wc havc to considcr only its various shapcsand motions. Mattcr for the Cartesiansis, if you pleasc, a kind of vast fluid, incompressibleand absolutely homogeneous. Hmd, unbreakable atonrs arrd thc crnpty spaccsscparatingthem arc mcrcly so many appcirallccs,so many illusions. Ccrtain parts of thc univcrsalfluid may bc animatcd by constantwhirling or vortical nrotions; to thc coarsccycs of thc atomist thcsc whirlpools or vorticcs will look likc individual corpuscles. Thc intcrmediary fluid transmits from one vortcx to thc othcr forces which Ncwtonians, through insufficicnt analysis, will take for actionsat a distance.Such are the principlesof the physics first skctchcd by Dcscartcs, which Malebranchc investigated further, and to which W. Thomson, aided by the hydrodynamicresearches ofCauchy arrdHelmholtz,hasgiven theelaboration and precisioncharactcristic present-day of mathematical doctrines. This Cartcsian physics cannot dispense with a theory of maghad alreadytried to constructsuch a thcory. Thc nctism; Descartcs corkscrcws of 'subtlc matter' with which Descertes,not without some naivct6, in his theory replaced the magnetic corpusclesof among.the Cartesians thc ninetecnth Gassendiwere succeeded, of ccntury, by the vorticcs conccivedmorc scientificallyby Maxwell. (1e04-5,. .13) p.

F F F F
F

th (1 p. p...42). ere eui H .;u o rl g c -e g q e s , 951, 42). Herequ;n. pr"ruppoffilf : (1951, U, rc:l(tcrs, pcrhaps wider circleof ilrt arct?erhaps and a.widercirclc intcllectuals a sharca toralityof lL.l,t,.-ri; ljur or/f bclrcrs' is it irr_ fact correct to spcakof o,r beriefs when bclicli,vary so widely lronr one prsonto arrotlrer? ConriJ.., fo, cxntttplc, group of theorcticaiphysicists.ff,i, g.",rf ,rritfhi 1 courri' a devour Roma' catholic, an'ortrrodox Jew,*a riiut"-6.t, l)rotcsrant,.areligiouslyagnosricadnrirerof the n"e ,"".t.i-l libcurlMuslirrr,l Marxist-Li'i'ist, l'cr so or. Auyorrc n,,rritinrl*itl-, tlrc thc.rctical. physics cor'rrrurrity wirr rccog'iscrrlattrris.*rnrol" rs (lultc rcilrlstlc.l hc point, howcvcr, is that thc various bcliclsof ltlc group do rrot fbrnr.a totality sharedby the group as a whole. 'l'lrcrc wil.l, of coursc, bc.a cor" of physicalt'co"ries^which ,"rrly cvcryonc in thc group will agrccon acccpting {g1J:ro,v:!ign1ll1i brrt thcrc will also bc a host of rcrigious ,,r"d',,r"rrpuyslcal tlcllefs ltrorrt rvhich diffcrent members of Ih. group *lll^dirrg... nrori Tl'e1c,religious and rnetapJrysi".rl iiff"r",r.., ;r;t l;;;; :Tl]l1glI: ctlccts cvcn *.,,1i1 physicsitseli and will vcry likely, as rv. I.,.u" llrcady argued,influe'ce_the way in which tlie variousphysicists cnrry o,t thcir rcscarch. Howevcr, the gc'cral situatio'l, ,lnror, inrpossiblc to describc witlrout making" the scicnce/'r;p6;i;; distirctio', a'd Quine's picrure of mowieage or beliefs ,; i;;;;; a scanrlcs.s garnlcnt does not accordar alt ,vell with the rcalitics o?
tllc sltu iltto n _

lll:

iilr

il ;
1r

ilf

X.

* ]:]
i

ii
ii

tvo

Science Metaphysics and

in Metaphysics Relation Science to

rationalscientificmethods,Aristotelians,Copernicans, atomists,and Cartesians, havc in cmulation fought its advent. (p. 240) It is somewhat rnisleading of Duhen.r to include Copernicans in this list. He is clearly thinking of Galileo, who denied that tides were caused by the Moon's attraction; but another famous Copernican, Kepler, held that the Moon attracts the waters of the sea by a magnetic action. Duhem's mention of the Cartesians in this context is, however, quite appropriate, and is rvorth pursuing. We have seen that Descartes' metaphysical views of the nature of lnatter had a beneficial influence in the nineteenth century on rhe dcvelopment of hydrodynamics and Maxwell's theory of magnetisrrl. At arr earlier stage, however, Cartesian metaphysics acted as r definite block to scientific progress, and was an obstacle to the :j acceptance Newton's theory of gravity. It was part of Descartes' of systcm that one piece of matter could affect another piece of matter only through diiect contact. Action at a distance *"r ,,ot allowed. Yct Newton's hw <>fgr:rvity states th:lt I piece of rnatter excrts a force ou every other piece of matter, however distant, according to the inverse square relation. Such a law was clearly quite inadmissible according to Cartesian nretaphysics. This situation led Newron hinrselfto doubt his own law, and to hope that a deeperexplanation which avoided the use of action at a distance would be found. Nrturally, thc rcaction of rrrany Cartesiansto Newton's law was still nrorc hostile. 'fhcrc rre undoubtecl exanrplcs of metrphysics excrting a benefici:rl influence on the progress of science,but we should not forget thlt rnctaphysic:rl idcas can obstruct, as wcll :rs hclp, scicnce.Whar is interestingabout the present example of Cartesianmetaphysicsis that it shows thar the very same metaphysicalsystem can be helpful in onc scientific contoxt and obstrr-rctive another. in Irr gcrrerll tcnrrs, thcrr, l)oppcr :rnd l)uhcrrr lgrcc llrout thc irrllucnccof'rrrct:rphysics scicrrcc,btrt thcy dilll'r in ln intcrcsting on llrsl:ionubout orrc prrrticulxrcxnrrrplc- thnt of iltourisnr. For Poppcr, :rtourisur onc ofthc nrost striking exanrplesofhow a nretaphysical is thcory can have a beneficial influence on science. Duhem, however, hclclthlt thc influence of atomisnr on scierlcervasa balefirl one. Indeed. hc lrcqucntly attacks atomisrn, as, for example, in the following Pirssxgr': (lonsiclcr sonrctrnc, instrucc,who lvoultl t:rkcphysical tbr thcoryjust as rvc lravcit, in thc yc;rrof gracc1905,prcscntecl thc. by majoriry of thosc

Duhem also discusses at some length the surprising fact that unscientific astrology had, at one stage, a beneficial influence on mainstream science. Duhem introduces this theme thus; 'Discoverv is not subject to any fixed rule. There is no doctrine so foolish that it may not some day be able to give,birth to a new and happy idea. Judicial astrology has played its part in the development of the principles of celestial mechanics' (p. 98). Judicial asrrology', in facr, proved helpful in the development of the Newtonian theory of universal gravitation, and, in parricular, of the theory that the ddes are caused by the attraction of the Moon. Now astrology is based on the idea that the heavenly bodies influence the fate of mankind here on earth; so astrologers would naturally be sympathetic to the vierv that rhe tides are caused by the Moon's attraction, since this would appear to them asjust one instance of their general idea that the heavens influence events here on earth. Conversely, many of the more standard schools of science rejecred the doctrine of the Moon's attraction as occult and superstitious. So, in this instarlce, the ideas of the less rational astrotogers proved more successfulthan those of their more rarional conremporaries. This is how Duhem describes the development: Ptolemy and Albumasardid not hesitate invoke r particularvirtue, to a specirl influence of the moon on the waters of the sea. Such an explanation was not intended to pleasc thc truc disciples of Aristotle; . . . The virtue that the t.ides manifestwas, on the other hand, made to order for the astrologers who found in it thc undeniablc proof of thc influencesthar the hcevenly bodies cxert on sublunar things. This hypothr'sisivrs in no less fevour lrnong the physicianswho compared the role played by heavenlybodics in the tidal phenomenon with the role attributed to thcm in crisesof disease; not Galcn did attach the 'critical days of pituitary diseascs' thc phasesof thc to nrrxu? (pp. 233-4) Intcrestingly cnough, thc astrologcrsclid notjust fornrulltc thc nrlin thcory (that thc Moon's attractiorlis thc causcof the tides), but also recognized that variarions in the tides are due to the influence of the Sun. As Duhem says: Morin fell back on principlesofjudiciary astrologyin order ro affirm thc rolc pl:rycd thc sun irr thc vrri;rtions tidc, :rn<l is irrrlccd by of it to tlrc irrdisputablc crcdit of thc astrologcrsthat thcy prcparcd all thc materials thc Ncwtonian theory of tides,whereas defenders for the of

ll

t,

'

rf
J
198 S ci ence and Metaphysi cs

in to Metaphysics Belation Science

199

ts
T

J T
I

F F F F F I

who tc:rch it. Anyorrc who would listcn closcly to thcrtllk irr cllsscs and to thc gossip of thc laboratories without lookirrg back or carirrg for what r u scd b c ta u g h t, w oul d hcar physi ci stscotrstantl ycmpl oyi ng i n thci r to tltco r ics r r to lccu lc s,atottts, :urtl cl cctrori s, counti ng thcsc srrral ll l odi cs and dcterrnining thcir size, their mass, thcir chargc. I3y thc almost tunivcrsalasscnt favouring thcsc thcories, by thc cnthusiasm thcy raisc, rnd by thc discovcrics thcy incitc or attributc to thcnl, tbcy would undoubtcdly bc regardcd as prophctic forcnrrrrcrs of thc thcory clcstinedto triunrph ir-r thc futurc. FIc rvouldjudgc that thcy rcvcrl r first draft of tlre idcrl form which physics will rcscnrblc nrorc cach d:ry; and as thc analogy betwecn thcse theorics and the cosnrology ofthc atornists str ikcs h im a s o b vi ous, he w oul d obtai n arr crni rrcntl y favourabl c prcsumption for this cosmology. I-{o'v diffcrcnt his judgrncnt will bc ifhc is not corltent with krrowing physics through thc gossip of the momcnt, if hc studics dccply all its brarrchcs,not only thosc in voguc but also thosc that an unjust otrlivion has lct bc rrcglected, and cspecially ifthc study ofhistory by rccalling thc crrors of past centuries puts him orr his guard against thc unrcasorr..:d cxaggcrations of thc prescnt tirnc! Wcll, hc will see that thc attcmpts at explanation bascd on atolnism Irrve accompanied physical thcory for thc longcst timc; . . . he will scc thctn constantly bcing rcborn, but constantly rlrortcd; cach tinrc thc fortunatc daiing ofan experimcntcr will havc discovcred a ncw sct of cxp cr im cn ta l la ws, he w i l l sec thc atomi sts, w i th fcvcri sh hastc, takc posscssionofthis scarcely cxplored dornairr and construct a meclranism approximatcly rcprcscnting thcsc ncw frndirrgs. Thcn, as thc cxpcrittrcnter's discovcrics bccomc more nunlcrous arrd dctailcd, hc will scc tlrc atonrist's combinations gct complicatcd, disturbed, ovcrburdcncd with arbitrary cornplications without succccding, howcvcr, in rcrrdcrir r g r p r ccisea cco u ntof thc ncw l aw s or i n corrnccti ngthcm sol i dl y to thc old laws; . . . It will appcar clearly to hirrr that thc physics of atorrrisrn, condcmned to perpetual fresh starts, docs not tcnd by contirrued p r o g r css to th c id eal form ofphysi cal thcory. (1905, pp. 303-4)

as which ainr to represent simply, as snrallnunrbcrof principlcs, l.lws' a conlplctely, and as cxactly as p<.rssiblc sct of cxpcr.inrcntal

(le0a-5, le). p.
It is not difficult to scc tlrc ideological factors behind this dispute. Materialism, mechanism, arrd atdtnisnr wcrc favourcd by Duhcnr's liberal, republican, and arrticlericalopponents, whereas Duherrr. as a dcvout Catholic. irrclined towards idealism and mathematics. It was Duhem's mis(brtulle that, during the best years of his scientific lifc, the rnetaphysical position which he abhorred - namely, mechanism and ato;ism - lcd to thc most striking advancesin physics. This, indced, affords a partial explanetion of Duhem's lack of scientific I{is particular metaPhysicaloutlook irrclined eood sense (le bon sens): hirn to theories and approaches which were inaPpropriate to the nlost ilnportant scientific problems of his day. No doubt there are mally othcr cascsin which an inappropriate metaphysics can explain a lack of scientific good sense. It would be wrorrg, however, to conclude that nretaphysical oricntation givcs a full cxplanatiort of scientific good sensc (or the lack of it). In fact, such an explarration is not adcquate even itr the case of Duhem. Duhenr's adnriration for abstract lnathellratical thcories should have led him to support Maxwell's theory of clcctromagnctism and Einstein's theory of relativity, whereas he rcjected both in harsh tcrms. Adrnittedly, in the casc of Maxrvell, he nright have been misled by the mechanical models which Maxwell cmploycd irr ltis or,'*'rraccoutrt of tlrc thcory. Howevcr, Dtlllcnr should have realized that these models were inessetttial; indeed, lie docs quotc Hcrtz's falnous statcnlent that 'Maxwcll's thcory is the system of Maxwell's eqtlations'(1904-5, p. 80)' As we saw il1 5.4, Duhern's erroneous view tirat arithmetic and geometry are based oll comnron-sense knowledge was one of the reasons why he rejectcd llon-Euclidean geometry and Einstein's theory of relativity. Although Duhem might not have appreciated this point himself, the development of the key theories of'modern physics (relativity arrd quantum mechanics) did to some extent bear out his polemic against atomism. Consider, for example, the change from Bohr's theory of the atom to the quantum mechanics of Heisenberg, Schrddinger, and Dirac - a change which took place from 1926 an. In the earlier phase of the Bohr atom, models were constructed which involved minute, but still visualizable, particles moving under the action of electrical and mechanical forces. The Bohr atom was conceived of as a miniature solar system. With the quantummechanical revolution, eveiything" dis5cilved iittcl systems of

Alnrost everyone would now agree with Popper's view that :rtonlism had a beneficial influence on science and disagree with I)uhem's claim that the influence of atomism was negative. None the less, this polemic against atomism is of great importance irr understanding Duhem and his intellectual work. Atomism as a metaphysical view of the world led naturally, at that time, to a preference for mechanical explanations in terms of concrete visualizable particles. Duhem strongly opposed this approach to physics, advocating instead the use ofabstract mathenratical theories. Ile puts his position with customary clarity as follows:'A physical thcory. . . is a system of mathematical propositions, deduced from a

200

Sciehce Metaphysics and

in to Metaphysics Relation Science

201

equations, whose solution gave the correct answer but whose interpretationwas far from clear. At all events,it was no longer possibleto think of concreteparticlesof the classical type. (For-an interestingrecent accountof thesedevelopments, chapter4 of see Arthur I. Miller's book Imageryin Scienti,frc Thought(1984)-, entitled 'Redefi ning Visualizability'.) Now one section of Duhem's polemic againstatomism seems curiouslypropheticof thesedevelopments. runs as follows: lr
Thcn, as the experimenter's discoverics becomemore numerousand detailcd, he will see the atomist's combinations get complicated, disturbed, overburdened with arbitrary compli-ations without succecding, however, in renderinga preciie accountof the new laws or in connecringthem solidly ro the old laws; and during this period he will see abstract theory, mrtured through patient laboui take possessionof the new lands thc experimentershave explored, organizctheseconquests, a.nnex them to its old domairis,and nrakee perfcctlyioordinated empire oftheir union. (1905,p. 304) f)trlrcrn dictl in 1916, but, cvcn if hc hrd livcd to 1926. it is to bc doubted whether he would have welcomed the advent of ouanrun.r mechanics. After all, he witnessed a sinrilar change from forenrz's theory of electrons to Einstcin's thcory of relativity, and, as we have seen, he did not appreciate Einstein's work. We should not dwell on Duhem's failings as a scientist,however, but rather concentrateon his brilliant philosophical insighrs, which make him one of the greatest philosophers of science of the twentieth century. Let us now try to see what general conclusions we can draw from this example. We are dealing here with rwo approaches to the construction of scientific theorr'es, based orr different metaphvsical ideas. These approaches could be called: (i) concrete models'and 1i;; abstract mathematics. Concrete models worked wonderfullv well in the-development of atomic physics up to about 1925, bui proved ineffective thereafter, giving way to abstract mathematics. Atsrracr mathematics did not work well in this field before 1925, as, among other things, Duhem's own scientific work demonstrated. Conversely, concrete models were ineffective afrer 1925. The example is in fact sirnilar to that of Cartesianmetaphysicswhich we consijered earlier. Cartesianism, it will be remembered, proved to be an obstacle to scientific progress during the period when the Newtonian theory of gravity was emerging, but, at a larer stage, was helpful for tlrc dcvclopnrcnt of hydrodynanrics and Maxwell's theory of rnaqnctisnr. Sintilarly, corrcrcte rrrodels l.raveworked well irr sonre

in scientificcontexts,abstractmathematics others. Indeed,the two approaches continue today in different branchesof science.Biochemistry, for exarnple,works almost exclusively with concrete rnodels,using little or no abstractmathematics. The models are chcmical rather rhan mechanical.Thus, for example, Mitchell's chemiosmotichypothesis(seechapter 2, note 3) explains energy flows acrosscell boundaries meansof an ingeniouscombination by of chemical reactionsand ion transport whose net effect is the cnergy flow. In particle physics,on the other hand, almost everyof thing (too much, onc is almost temptedto say!)consists abstract inathematics. The conclusionseemsto be inescapable that metaphysical ideas rre not only meaningful,but necessary science. for They provide an framework within which specific scientific theories indispensable and comparedwith experience. Metaphysics canbe constructed acts Br-rt guide rs a guide,or heuristic, science. while a metaphysical for to is necessary move in any directionat all, sucha guide canjust as lcad in drc wrong. rs in thc right. dircction.Tht' vcry stule c:rsily nrctrplrysic;rl systcnl (wlrcthcr Pythrgorclnisru,nrcchlrricrlnrrtcprogrcss or crn promotescientific ri:rlism, Cartesianisrn, whatevcr) in one context or problem situation while acting as an obstacleto scicncein another, What this shows is that there is no magic in formula for doing good science.It is often necessary scientific rcsearch cxplore f;rlsetrails so thit the correctone carlbc fourrd, to

i. ,a

9.4 D u hem 's Def enceof Religion Duhem was an ardcnt Catholic, and wc cannot give a satislactory without examining accountof his views on philosophy of science their implicationsfor religion. In 1904, Abel Rey publishedan in xrticleon Duhem's philosophyof science which he arguedthat this was the philosophyof a believcr.Dtrhem replied to Rey in his 1905articleentitled'Physicsof a lJcliever'.Here Drlhcnt deniesthat with an apologetic intent, of was developed his philosophy physics that it was 'forcedon the authoroutsidcof arty and insistsinstcad conceru,and ahnost despitehimself, or nrctaphysical theological through the daily practice and teaching of the science' (1905, sometruth in this. Duhem devoted nearly p.275).Thereis certainly and its historyaud of :rll lris time to thc study arrdteaching physics in to rcsearch physics.Many of his ideasirr thc philosophyof

:
J

r|
J

202

Science Metaphysics and

to in Metaphysics llclation Scicnce

203

t
I|

T T
T
T

IJ

li;

scicrrcc arrlscfronr thcscactivitics, arrtlarc oI grcat intcrcstaplrt frorrrarryrcli.qiorrs qucstious. Yct, if l)uhcrl is tcllirrutlrc trtrtlr,hc is not tclling thc wholc truth, for lris philosophical idcaswcrc also i.firrc'ccd by rcliuio,s corrsidcrrtions. Indccd"his 1905:rrticlcis ;t p ro o fo l ' tl ri s ,l o r i rr i t l rc rrscsri sphi l osophy ' sci crcc pr' vi dc r l ol to defcncc rclision. of I)u l rc rn ,a s w c hrvc sccrr, l crrrarcxtcs cr)cc < sci frorrrrrrctupl rysi cs. Scic'cchrs its orvrrsphcrclud its owu nrcthods. scicrrtific tlrcorics 'r'lrc lrlrvctlrcir orvrrsulrjcct-rrmttcr gcrrcrllchur;rctcr. saruccrrr :rrrtl l rc s :ri ti f u rc t;rp hysi cs. thc sphcrc rrd rrrctl rods rrrctl physi r:s o brrt l oI ;rrrd subjcct-tuxttcr gcncrrlch:rractcr.of tlrc :rrrd urctaphysical thcorics arc quitc diflcrcnt fror:r thosc of scicrrce and of scicrrtifictlrcorics. Now irr Duhenr's day it wasoftcn rnainteirrcd religionhad bccn tlrat disproved,or at least underrnined, by the advanccof sciencc. Duhern argucstliat rcligiorrbclongsto thc sphercof rneraplrysics, rv h i c l ri s q u i tc d i f fc' cnt fl ' onl ;rrrrld< l cs rroI i ntcri l ctw i th rl rl t of scicrrcc. tlrc idcr .f scicrrcc So urrdcrnrirrirg rcligiorr('r lrry rrictaphysic:rlvicw lirr thrt rrrlttcr) is sirnply :r rrristakc. As l)rrhcrrr h i rrrs c lp- tsi t: lu
It has bccn fashionablc some tirne to opposcthc grcat thcoricsof for physicsto thc furrdanrcntal doctrincson which spiritualistic philosophy arrd tlrc Catholic faith rcst; thcsc docrriucsarc rcally cxpcctcd to bc sccncrunrlrlirrg urrdcrtlrc rirrnrnirrg blows of scicrrtilic systcrtrs.. . . Now thc. sysrem wc havc expoundcd gcrs rid of thc allcgcd . objcctions that physical thcory would raisc ro spiritualisricnitaphysics and C:rtholic dogrrra; mlkcs thclrrdis.rppi,lr casilyas tlrc it rs rvirrd swccpsaway bits of strrw, for accordirrgto this systcnr thcsc objcctiorrs lrc, and can ncvcr bc anythingbut, misundcrstandirrgs. (1e05,p. 283) This sound-s like a plausiblc defenccof rcligion, and ir nright have . becrr expected tlrat Duhcnr's argumcltts *ouJd b" lr".y *.1-.o,rr. to the Catholic Church; but, as Martin has showrr in his 1991 book on Duhem (seeparticulariy chapter 3), this was rlor the case.Far from the Catholic Church acccpring Duhem's dcfence of religion, lris argunlents were thrown under the suspiciorr of heresy. As Martirr s ay s : ln thc crrcyclicalPascendi DonriniciCre.gis 1907, two ycars afrcr of l)uhcrrr's 'Physicsof a Bclicver', thc official position was rnadeclcar in_thcnameof Popc Pius X. Of thc dangcrouirrp".t, of thc hcrcsyit callcd 'nrodernisnr'idcntificd by thc cncyclicaltwo corrccrnmc hcrc,

wh a ti tca l l ctl tl r c.a g n <r sti ci sl tr 'o ftl r ctr r o d cr tl i sts'a tr d tl r cscl r r r l tr o tr . bccausc.of .rf .rli.n." and faith.'Thc frrst for cxamplc was dangcrous to uatttraltlrcology ' ' Thc sccondwas undcr ' ,i," J.,rtng. it clid pp: 38-9) of suspicion Iid"it-- (1991, doctrines Natural thcology is the attempt to argue for theological n:rttri:rl worlJ' But if' as Duhcrrr clainrcd' ir.rt"-iit" lrt.t.li". .rf tltc

scparate',:lt::.lt:'.:* ;i; ;pkt.t of scic'cc arrdrcligio' are.quite that faith cntcrprisc' is E tlrcology hardlya vi:rblc

i,#firu;;;-+,'

thelristometaPhysical-system' in rclisiousteachings a particular Aquinas' The aim of orthodox Latnollc of tclianisnr St thornas Thomism reconcilc in*ti..,uttt should then have been to try to to :lrgtlcthat the two bclong to rathcr thin rvit| rtro{crrrscicpcc tliffcrcrrtsphcrcs' -"-S,r-,ttull, to thcn, for thc attitudc of thc Catholic Church this dcfcnccin thc l)ulrcrn'sd.i.n.. of rcligion. Lct us now cxaminc Duhem'sline of thoughton iiJ, of ,o*. of our caiier discussions' a good dealin common with that of,witts;n1te1-t1 has ;ii;:;;J." s
flncs arc nlc shares w lttgensteln

thccathorto !1-t':;1'1*".'j: thl'*

I t fr F F
tr

.rrrrroi undermine religious belief' in advances scie-nce

is that a la the to arate. As we have

earlier i thanothers'Giventhissituation' views more plausible rnetaphysical .h;,;;; in scic'ce ca' have arr cffect on the credibiiity of religious fact' ;;ii;Fr, .onritty to what Duhem claims' As a matter of on writings arc not really consistent this point'.In his Duhem's own are and metaphysics J.f*." of religion, f,c maintainsthat science

F b

204

Scienceand Metaphysics

separate, non-interacting spheres. Yet, in his detailed analyses of historical developments in science, he gives many striking instances of the influence of metaphysical ideas on science, as we saw in tlre third section ofthis chapter. Indeed, in his 1905 article'Physics ofa Believer', he actually ruggers a way in which some recent developments in science might support a particular metaphysical view of the world. Duhem says that we can argue in favour of what he calls a 'cosmology' - that is to say, a metaphysical view of the world - by analogy with theories in physics. However, we must be careful to use, not the present state of physical theory, but rather the ideal form towards which it tends. The present state of physics in 1905 was characterized by the triumph of atomism, against which Duhem polemicized. The ideal form towards which physics r.vas tending, according to Duhem, was general thermodynamics. Duhem goes on to develop an interesting and quite plausiblc analogy between general thermodynamics and Aristotelianism, and

10
Fa lsification i n th e L i g h to f ism e' t he Duh e m-Qu i n Th e si s

10.1

F a l s i f i c a ti o n i sm a n d th e Fa l si fi a b i l i tv C r i te r i o n

If wc rid tlie physicsof Aristotle and <>f Schohsticisnr thc outworn of and demodedscientificclothing coveringit, and if we brirrg out in its vigorous and harmoniousnakcdness living flcsh of this cosmothe logy, lve would be struck by its rescntblance our ntodcnr physicel to theory; we rccognizein thesetwo doctrinestwo picturesof the samc ontologicalorder, distinct bccause thcy arecachtakenfrom a diffcrent (1905, 310) point of vicw, but in no way discordant. p.
ll

'li

ri

It seems herc as if Duhem has almost forgotten his 'two separatc spheres' ju'itification, given earlier in the same arricle, .ni l,n, instead adopted the more orthodox Catholic positiorr of trying to reconcile modern science with the scholasticisrn on which Catholic doctrine is based. Yet the very way in which he devclops this orthodox app.roach indicatesits dangers.If generalthermodynamics had indeed triumphed in physics, then this might have been used rs an argument for scholasticism. However, by thc samc token, tlrc failure of general thermodynamics and the success of atomisrn should be considered as undernrining scholasticisnr, and hencc Catholic doctrine.

Falsificationism is a theory of scientific method - more specifically, the theory that scienceproceeds through conjectures and refutations. Now fior this theory to work, it is clear that the conjectures put forward by scientists tnust be c:rpable of being refuted by obscrvation and expcrimcnt. This'suggests that we should allow as scientific only those conjectures which are refutable or falsifiable. Indced, Popper himself says: 'But I shall certainly admit a system :rs empirical or scientific only if it is capable of being tested by experience. . . . I shall require that its logical forrn shall be such that it can be singlcd out, by means of empirical tests, irr a negative scnse: il ttrust be possible Jor an empirical seientijc syste,nto be reftted by (1934, pp. a0-1). experience' frlsificationisnr as a and falsifiabilitv as a

-l

ed to distinsuishs
tic one, since arry criticism of Yet this conncction is a firlsifiability as a denrarcation criterion (and there have been many) would seern to cast doubt, at the-same time, on falsificationisrrr as a mcthodology. My aim in this cl'rapteris to investigate the range of problems which arise out of this situation. I will begin by consiclcring threc strndard objcctiorrs to falsifilbility as a dernarcation criterion. I will argue that the first two of these difficulties (which concern pure existcntial statementsand probability statements)can be surmounted and that the net result of the discr-rssion in fact, to is, providc morc argumcnts irr favour of falsifiability. The situation is differeut regarding the third obiection basedon the Duhem-Quine thcsis.I will arguc that this difficLrltyshows falsifiability to be, if rrot cr:tircly wrorrg, at lcast inadequate.

Science Metaphysics and

was to dcny that an scieuce arrd mcta

and Falsificationism the Duhem-OuineThesis

207

both forrnularcdhis thcsis anrl co,rtiffid-li6 ;ut!iunlc throughout his book that a distirrctioncould be rrraclc lr.twccrr scic'ccand metaphysics. the fifth sccio' of this chaptcr, I' I will nttcn:pt to defendDuhem's rarher than euine's porliion, Iry.srrggcstirrg a demarcatiorl criterionbased rcnfimability, on 'ew To illustrate ritf rer rfr;rnf Isifability. this criterion,I will sliow that'ii k';ttlsto the intuitivelyplausible result trrat Newtonian nrechanics ls rcicrlific, whercasAdlcr's .rheory of tlrc infcriority courplcxis trrtlpltysic;rl. Thcn in the sixth sectiorl will considcr r irow niuch of lhlriticlti.rrisrn a rnerhodology be rerairrecl the demarcarion as carr if erltr.riorr is.charrgcd thc-rnanner in suggested. Ratlrerpleasingly, it Iunri ort that rnuch of falsificatiorrism retains validity, r"'itt"r, its cvcn sssurrring new dcmarcation our criterion, canstill 6. it "rr"rt"i lltnt ncie'tists shoulduse their crearive powersin formulatingbold r'otliecttrres lbout the natural world, and that th.r. co,rj.".tu..s

in which c;ttions,which arc usually inrplicit irr the scicntific colltext -fhus 'Thcrc cxist cgg-layingthcy occur, thcy bccomc faliifiablc. circrttnscribed) region-of i,i.irlirrt, i,., tl.,i, p*rti.ular (qrritc closely falsifiablc, sittcc we cotlld search thc spccificd Ar',r,rrli.' lrc.or.,lci i, his owrl ..gion rvith carc a'd find no such rnamnrals. ltr tcrnrs o[ the point: cxirnple, this is how Popper Puts 'Tlrcrc cxists an clcnrentwith thc atomic nttmber 72" " is scientific indicdtions. thcory, arrda.thcory u'hichgi.ves as part of a highly tcstable on to.liultik ilenrcnt.lf, t6c othcrhand,wc took tlis cxistcntial ,oi,[inru us ii.t.rn"rrt in isolation,or rs part of a thcory which docs trot givc any hint as to how -and whcrc this clement can bc found' tlren we it simply bccause would *Juld ltnu. to dcscribcit as metaphysical (1983'pp. 178*9) rrot bc tcstablc' Th i ssce n - r sto m cto b e r ca so r r a b l yco tl vi tl ci n g ,a n d l co n cl u d e th a t difficulty for cxistential statements do rrot .ottriitnt" an insuperable thc falsifiability criterion'

10'3 Pr obabilit YSt at em ent s

+ 6av,te//r 'r6,a o77ruc/ ai:El ar'g4^(1** d4/a*wfrea? /I q.o,bs, v/e.,/ .

10.2 ExistentialStatements Ar tve ruw irr 8.5, cxistcntial statcments suchas'Thcre is (or there e.xhm) wlritc r;rvcn'are vcrifiable,but not falsifiabre. carrverify n lue tltc t'lnitrr tlmt thtrrcis a whitc ravcrr simply by obscrving ..u..,,o'f .' iltrtt eolotrr;brt no Iirritcsct of obscruatioiu .ou",r, .rn p"rsit iy of folrlfy tlrh r:l;rir', Now this logical poirrt might b" ur"i ,, .i, rlller tirrrrt.. lirlsifi'[rility a dcniarcation as criteritn. After all, rhcre rfrr reerf to be honn f which are existc'tial in .fidr scientificstatements lirt cxarrrplc, that thereexistegg-laying mammals, an or, 'ltrttgrter; errrrrtpkwlrich l,'ppcr hirnsclfgives(1934, Oq, tt rt thereexists p. dfr elelfref wirlr the atonricnunrber72. It codd be argued,thereft tfut firlsili:rbility too srrong, and rhat it irfadvcrterrtly [ye,, .is t'lltllliliterroluc gcrrrrincly scicntific statclnetrts. llrr;rper''r .nnw* to this difficulty is trrat existential statenrenrs ttr lllcir'wll ;rrr irrlccrlrnctaphysical. However,with somequalifi-

with the falsifiabilityof probability Tlrcrc is a difficulty contrectcd which Pbppcr himself statcsvery clearlyas follows: statcmcnts .Ilrcrclationsbctwcctrprobrbilityrttdcxpcrictrccarcalsosti|lirrnccd
discover of cl"rifrcatio,r. ln invcstigating this problem wi shall almost insttpcrablcobjcction. to tny at first ,"",,, what will ",'r play.such probability statemcnts mcthodologicalvicws. For althoughbc in turn .-"i,.iiy iitip.rtant rolc in crnpirical-sc-icncc,.thcy out to ycr. to striuJalsiiication. rhis vcry stunrblingblock prirrcipic intpcruious '*iiLi'".."r. a rouchsronc upon which ro tcst rny theory, in ordcr co ottt what it is worth. (1934,p. 146) firrd To see why probability statements cannot be falsified, let us take Suppot. wc-are tossing a- bent.t:il: tlr. ,irrrpl"rt i*"*pl". 'tt{ the tosses are-independent and that the probability of oostulrte that rr headsin n ircadsis p. Let prob.(mln)be the probability of gctting tosscs. Then we havc Prob(rala) :"Co,P"'(l P):-'"'

.1i:

t{ 1i 208 and Science Metaphysics

and Falsificationism the Duhem-Ouine Thesis 209

The exactmeaningof this formula is not important. The important point is that however long we toss the coin (that is, howeve"r n big is) and whatever number of headswe observe(that is, rvhatever the value of nr), our result will always have a finite, non-zero probability. It will not be strictly ruled out by our assumptions. In other words, theseassumptions 'in principle imperuious strict are to fakifrution'. Popper's answer to this di{ficulty consistsin an appealto the notion of methodological falsifiability.Although, strictly speaking, probability statementsare not falsifiable,rhey can none the less be used as falsifiablestatements, and in fact they are so used by scientists. puts the matter thus: 'A physicistis usuallyquite well He able to decide whether he may for the time being acceptsome particular probability hypothesisas "empirically confirmed", or whether he ought to reject it as "practically falsified"' (1934, p. 191) . I have worked out a particularversion of this approachin some detail in my 1977 article 'A FalsifyingRule for Probability Statements' and subsequent book An Objactiue Theory Probability oJ (1973) (cf. part III, pp. 161-226).The full solntion involvesa grext deal of

technical mathematical probability and statistics- particularly a consideration of the theory of statistical testing. However, the basic idea is not difficult, and can be exphined as follows in terms of the cointossing example.

Alrhough, as alreadypointed out, eachvalueof m/n has a finite, non-zeroprobability, some of theseprobabilitiesare much higher than others. For example, if mln is very near to p, prob(mln) is rnr.rch higher than it is if mln is very far frorn p. Collecting these higlr probabilitiestogether,we obrain an interval round p, (p - d, p + d) say, snch that chere a probability of more than, say, 95 per is cent that mlniswithin the interval (p - d, p + d), and a probability of less than 5 per cent that mln is outside this interval. pe
d the unde outside this as confi al, This procedure can be

t\'rl

P ---=H is regarded as falsifiodil the result lies here F ;r lsifyir r g;r p r o b lliility h yp o th csi s

, l : i r ; r r r , 'l ( ) . i

given in figurc 10.1. An actual example is given in my 1973 book (pp. 124-7).1 tossed an ordinary coin (an old penny) 2,000 times. Calculations show tlrat, if we assume prob(heads) = 7/2, there is a probability o1 97.3 per cent rhat m/n will lie in the interval (0.475, 0.525). The observed valrre of m/n was rctually 0.487, giving a confirmation, rather than frlsification. '!ilhen developed in full mathematical detail, the procedure agrees re:rsonably well with the standard thcory of statistical testing developed by the Pearsons(father and son), 'Student' (W. S. Gosset), Fishcr, and Ncynran. There arc, however, some rarher techrrical difficulties connected with the so-called one-tailed tests and (particularly) the Ncynran paradox. Ilut solutions crn be for-rnd for these difficultics, and so frlsificltionisnr crrr proviclc a sin.rplcand reason:rbly srtisfactory account of probability which accords rvell with r r n i r r s t r c a n s t ; l t i s t i c r lp r i l cti cc. r It shoulclbe addecl,howevcr, th:rt thcre is no consensus regarcling the foundations of statistics,and some schools oI thought - notably thc Blycsian - :rclopt rn approlch which is vcry diflcrcnt frour fhlsiflcrtionisrn.lu rn irnportant reccnt book, Hor,vsonlnd'Urbach (1989) criticizc falsificationisn"r statisticsand defend the Bayesian in r p p r o l c h , w h i l c i r r r rr y r n ,i cw o I th ci r tr o o k ( ( i i l l i cs, 1 9 9 0 ) , I nlturnlly dclcnd firlsificrtionisnr against their rttacks. !!!flB!_tbgnlost iudicious corlclusion is that everv school of rs. arrd th:rt thosc of thc worsc thln what is tlrc rrornr in this ficld. At all cvcnts.

',';:':;

ttv

Scienceand Metaphvsics block for

state me'ts cannot be regarded as a fatal stumbli'g falsiflrbilitv,

211 Thesis and Falsificationism the Duhem-Ouine This might be done in the following of considcrations simplicity. sinplieityif no thcory rvay. Lct us say that a thcory T is of adcquate be floundsuch that U can (a) U is as sinrplcor sirnplir thanT, and ,tr. obicrvablc results derivablefrorn T are likewisc it ) "tt derivablefronr U. and of if tile thcn say that a theory, T, is scierrtific it is falsifiable like suflicesto eliminatecases rdcquatcsimpliciry. This definition simplicity' However' the are'clearlynot of adequate f A fra which oroblern of climinating metaphysicalcomponentsfrom scientific fa'tasy' but ilr.lori", (asit nright be callcd)is no nlere philosopher's easyto situationswhere it is not altogether irr sc'ientific can crop up absolute the Newtou introduced colrcePts-of ,"rolue. For example, alrd time. fere theseconceptsgenuinelyscientific'or were srrace additionsto t6e theory? ,, Maclr chirircd, nlercly rnetrphysical ,ir.u. Thc answeris by no mcausclear' ofverifiability: criticism part ofPopper's Let us runl to rhesecond of an6 interesting all, ,rr*.iy, rlrar 'it did exclutlithe nost-inportant the scic11f1 theorics,thc uniuersal that ,rirrtii, stutaffiettts, is to say, does allow into IorrLf naturc' (1963,p. 281). Now falsifiability laws, including mally that are concerned ,ai",ta" ,on',a at,rir,".sal cntitiesand propcrtics.Unfortunately, however,--it witlr obscrvablc laws, which areuormally .pp..., to cxcludehigher-lcvcltircoretical To see this, let us take an example liltnt"a as part of Jcierrce. that all namcly,Newton'sfirst law, which states earlicr: considcrcd bodiescontinuein a stateof rcst or uniform motion in a straightline urrlcssclisturbcdthcrcfrom by a force. Thc difficulty herc' ls rv-c with our iiscussion of Poi'cari's conventional,r- i, connectiol.l a isrn in chaptcr4, is the following. If wc .observe body rvhich is rest nor in uniform motion in a straight line and is rreither at not actedon by'any for9c, $.e.nwe can alwayspostulate apparently an'invisibleforce acting on the body. This is exactly what Newton which rnove in ellipses,rather than did in the caseof the plarrets, actedo' by the rhat thcseplanets_were straightlines. He postuiated gravitation. (sceFigure 3.1) A similar irrviJble force of universal exceptionto Newton's first devicecould be appliedto any apparent law (the law of inertia),whiih thus doesnot appearto be scie'tific accordingto Popper'sfalsifiabilitycriterion' The p"roblernhere is, as we saw in chapter 5, posed.by the Duhem Q uinet hesis. Newt on'sf ir st lawcannot bet est edonit s

10. 4

Fals i f i a b i l i t y a n d t h e D u h e m - O u i n e

Thesis

Falsifiability has been able to overcorne rhe difficurtics so far co'sidered. Much more serious, however, are a number of problems co'nected with the Duhem-Quine rhesis, which was dcscribed in chapter 5. These difficuities do show, as I will now argue, rhar falsifiability is not adequate as a demarcation criterion. Mv intention is not, however, entirely negativc. I will try in thc next section to fornrulate another demarcatio' criteriorr, based o' confrmabirity, rather tlran falsifability, and will thcn argue, iu the-followiig scction, that, wirhin the new framework proposed, rnany of thl basic ideas of falsificarionism car be rctained. In particulai, it will turn out that thc falsificationist approach to statistics rnentio'cd irr the prcceding section is still viablc. Ler me begin, however, with my critique of falsifiability. This is perhaps best approachcd by considerirrg again Poppcr's basic criticisnr of vcrifiibiliry, whicir he sct out clearly in the following passage: .My criricism of the vcrifiability critcrion lias always been this: against the irrrerrtiorr of its defcrrders, it did not excludcobuiousmctaphysical state,ne,rts; it did but uelndc rhe nost inrportantand interesrhg ill icicntifc stdtct,tcttts, is oJ rhat say, tlrc scientific theorics, the universallaws of nature'(1963, p, !o

Firsr of all, does falsifiability fail to excludeobvious-mctaphysical statemcnts? Unfortunatelythe arlsweris 'yes', because i iesult of known as the 'tacking paradox'. Let T bc a falsifiabletheorv - for example, Kc-pler'sfirst law. Let M be an arbitrary metaphysical statement for example,'The Absoluteis sleepy'.SerT, = f & ifA. Then T' is, like T, falsifiable.Since T is faisifiable,rhere is an observation statement such that o follows logically frora T, So if o we find by observation thar O is false,it follows thai T is false.But since O follows logically from T, it also follows logically from T & M. So T' : T & M is falsifiabie.In other *oids, giverr a falsifiable theory, we can always 'tack on' an arbitrary metapihysical statcment, and still havea falsifiable theory. What canbe dorieabout this situation? My own view is that falsifiabilitymust be supplernented hcre by

28t). L.i. 15 5.. if we cantur' this criticism poppcrhi'rsclf. against

rli
llr

il

{{

212

and Metaphysics Science

Falsificationism and the Duhem-Ouine Thesis

213

own, as an isolated hypothesis, but only as part of a theoretical group. Lct us call Newton's first law T. To obtain observable couscqucnccs fronr Ncwton's thcory, wc havc to add to T: (1) furthcr thcorctical assurxptions:namcly, the second and third Iaws of motion and the law of gravity - call thc conjunction of theseT', and auxiliary assumptions: for example, that non-gravirarional forces havc no appreciable effec on the motion of the Sun and the planets, that the mass of rhe Sun is very much greatcr than the mass of any planet, and so on - call the conjunction of theseA.

prcmisc of thc sct of all premisesused irr derivirrg thc rcfutcd conclusion. The :rttributiorrof the frlsity to sonrc prrticulrr hypothcsis is risky, cspccirlly wc if tlut bclongsto tlrissct of prcrnises thcrcforc of which cutcr into cvcry consitlcrthc grclt rrurrrbcr rssrrrrrptions cxpcrinrcnt.. . . 'l'lrc rnsrvcr is tlmt rvc crn indccd hlsify only s),stcrrr.t of thcories lncl that lny :rttribution of f;rlsity to rny p:rrticul:rr statcnrcnt within such a systcrnis always highly unccrtanr.(1983,p. 187) Lct us sec lrow this works out in terms of our exanrple of Ncwton's first law, T. As we hrve scerr,T is not lalsifiable,but tlrc conjunction of T & T' & A is falsifiable. One option for Poppcr woulcl thereforc be to say thxt T is not scientific but metaphysicll, whcreas T & T' & A is scientific, This suggcstion is not vcry srtisfactory from Popper's own point of view, sincc hc has always strongly criticized the vcrifiability criterion on the grourrds that it laws of nature. Now given the vicw we are cxcludcd the rrrrivcrsal considerirrg,Newton's first iaw, a classicparadigm of an important lrrcl intercsting law of rlilturc, is excluded as a piccc o[ n)ctlphvsics. 'l'lrc prcscnt suggcstion cor.rltlnot, thcreforc, bc considcrec{as sxtisfactory by Poppcr or, indeed, by rrryself. There is, however, rrrothcr possibility. Could we not s:ry th:rt :r thcory Ts is scicntific if it is part of one or nore systcms of thcorics of the form Tl & T2 &. . . & T, which are falsifiable? Tlis nrodified demarcation critcrion certrinly rllorys Newtor"r'sfirst lrw to bc scicntific; but, unfortunately, it has the consequellcethat any lrbitrary metaphysicalstatement is scicntiflc. This carr bc shown by trsirrgwhat I will call the Ayer cxanryle.This exanrple was originrlly put forward by Ayer irr the introduction to his 1946 book (pp. ll-12) as a criticisnr of his own previous versiortof thc verifirbility criteriorr; and he then goes on to fornrulatc a new versiorr of thc principle of verification. Ayer notes that he was irrfiuenced by Berlin's criticisms (cf. Berlin, 1939), though Ber:lirr's countercxlrnplc is in frct cliffcrent fronr Aycr's. But llthor-rgh the Aycr exanrple was first proposed in conncction with the verifiability critcrion, it can uscfully be considered in corlnection with thc falsifiability criterion, and :riso in coruection rvith the confirmability critcrion to be' introduced in the next section. Let M bc rn arbitrary nretrphysical st:ltement - for exarnplc, 'Thc Absolute is slcepy'. Let 01 . . . O, lrc auy arbitrlry observation statcrncnts fcrr cxlmple, 'This is white', 'This is :r pen', and so forth. Then M is a cornponent of the following falsifiable systerni: M & ( i f M , t h e n O 1 ) , M & ( i fM , th e n O2 ) ,... M &( i fM , th cn O") .

(2)

Now, from the conjunction T & T' & A, we can irrdced deducc observable consequencesregarding the motions of the planets. Call orre such consequenceO. Suppose now that we observe not-O. We cannot conclude 'not-T', but only 'either not-T or not-T' or not-A'. To put the point another way, we know that one of the assumptions uscd in the dcduction is falsc, but wc do not know wlricli one. So none of the assumptions has been falsified. Moreovcr, as thc Duhern-Quine thesis holds for any application of Newton's first law to cxplain observable phcnonrcna, it follows that this law is unfalsi6able. How does Popper himself deal with this difficulty? It will be rcmcrnbcrcd that in forurulating his view of falsifiability as a dcnrarcation critcrion, Popper speaksnot of scienrific thcories, but of 'theorericaisystems'. Interestingly, thc term 'system' (in French syslinc) is',also used by Duhem in this corrtext, as the following passxqe,already quored in chapter 5, slrows: 'For it is not bcrwcen two hypotheses,the emission and wave.hyporheses, tlut Foucault's experinrentjudges trenchantly; it decides rather between two sets of theorics each of which has to be takeri as a whole, i.e. between two cntire systcms, Newton's optics and Huygcns' optics' (1904-5, p. 189). An appeal to this rrotion of 'tl.redretical systems' does irrdecd constitute Popper's answer to the problern posed by the Duhemsho'ws: Quinc thesis,as the following passage More scriousis an objcction closcly corrncctcd with thc problcrn of coiltext, end thc fact thar my criterion of dcrnarcationepplics to s),ste'ntslhcttries oJ rathcr than to staterncnts of context.Thii obiccout tion nray be ptrr rs follows. No singlehypothcsis, may bc saij, is it lalsitlable, bccausc evcry rcfutarion ofr conclusion may hit any singlc

--1

214

Scienceand Metaphysics

and the Duhem-Quine Thesis Falsificationism

215

llcrrcc,by our nroclificd critcriorr, ouglrtto bc scicntific, M wl.;-h it is ttot, l conclutlc thrt l}lsifi:rbility not rdcquatc :r clcrrurcutiorr is ls c ritc ri o rr.

'falilc 10.1 Classificationof statcmentsitrto lcvcls 0, 1, 2, md 3

Level Status
?

Criterion

Example atomisrn Greek first law Newton's first Kepler's law recording Statement position Mars of at a particular time

T I

1 0 .5 A Su g gestedD emarcati on ri teri onInvol vi ngthe C Principleof ExplanatorySurplus l l ' fl h ri l i l l ri l i ty i s irr;rdcqrr:rrc;rs i l dcrrrarcl ti on rcri orr, cri thc ncxr rrlrvirrrrs ()t to is lrossibility consitlcr rou.linuahility rc,.rohitttlility.'l'hc srrggesti<rrr is tlrat:r theoryis scicntific arrdonly i(itis confinnnltlt:: if tltut is, c;rp:r[rlc :rcquirine of sorncpositivcdcgrccoltsrrpport fronr ;l possiblcobscrvltior.r starcnlcnr. a theory (T, say) is falsifiablc, If tlrerrit is ccrtairrly confirmable. For ler O bc an observatioll statctrrcrrt which is a potcntialfalsifier T, thcn, if rrot-O is obscrvcd, of tlris will liirl)port On thc othcr harrd,thercarc thcoricsor laws T. rtteh as Ncwton's first law which arc corrfirnrablc witlrout ircinu frrlsiti,r[llc, rrrr thus sugecstirrg irrtrodrrction . rrc* l"u.ii I thc of (lcvcl 2) bctwccn fhlsifi.rblc srlrclllcnts(lcvcl l) rnd rnct;rphysic:rl !tlitlf nrrnts (lcvcl 3). If obscrv:rtion stlttcntcnts classificd lcvcl rrc as (f, tltc tr:lrcrrrc bc presented in table10.1,rvhere, concisecan as for rrels, I lnvc ornittcd considerations relating to simplicity. Thc ntr(rwscorrncctilrg lcvclsI rrrd 2 to lcvcl 0 arc dcsigncd irrdicatc to tltnr scicrrtifictlrcoricsof any lcvcl, whcn oncc confirurcdand itfc'elrt$l lrnckgrourrd ;rs knowlcdgc,can bc uscd in thc intcrprctatiott ol"ttcnsory cxpcricncc which is nccdccl produccobscrvation to F l l fID c rr , n t!i l-,et rrow cxarnine rrs whethcrit is reasonablc regardNewton's to firrt law, 'l', :rslrcingconfirmablc, thoughnot falsifia6le. is part of T vrrriorrr corrjrrrrctiorrs asT & T' & A which areconfirmable; such so, if'wtr supposc that rhc supporracquircdby the conjunction disis tt'illtttcd iulrollgthc conrponents the conjunction, of then T will bc r:uttlirrrnhlc. llut now ir could be objected that this approach will rtlr irrto thc difficultics created the Ayer exarnple. by (lttcc lgnirr, lct M bc an arbitrarymetaphysical sratcmcrrt for ex;ttnlrlt', "l'lrc Absolute lazy' * aud let O be an arbitraryotrseris veliorr fltltcnrctrt- for examplc 'This is whitc'. Thcn we havc thdt () lirlkrws logically frorn M & (if M, thcn O). It .vould rrnlr, tltcrclirrc, that O supports & (if M, then O); hence, M again rttpporirrlq thc-strpport rhe conjunction distributcd tlut of is among tlte contporrcrrts thc conjunction, is supportcd someextcli of M to

('
v

l11
\Y

\0

MetaphysicalNotconfirmable but Confirmable. Scientific not falsifiable Falsifiable and Scientific confirmable Observation Truth-value determinable statement by observation

ts

ts

F
lr

lr

would appearto statement by O. Thus arry arbitrary metaphysical scicnccfrom bc confirmable,and our: wlrolc atteinpt to dcmarcate to on nrctaphysics thc groundsof confirmability appcars collapsc. a I To rvoid this clifficulty, will suggcst principlcwlrichlimits thc as a can bc considcrcd supporting in circuurstlnccs which cvidcncc theory, and, which, in particular, blocks the confirmation of M in the Ayer example.To illustrate what is involved here, let me first give an exampleof a gcneralprinciple of confirmation which This is of by would, I thirrk, be acceptcd rnost philosophers science. the that the more severe testing'whichstates thc'principleof scvcre the h tcstswhich a hypothcsis haspassed, grcateris the confirmation of h. Popper seemsto have been the first to formulate this principle, and he puts it as follows: 'It is not so much the number of which determines the degreeof corroboracorroboratinginstances tion as the severityof the variousteststo which the hypothesis in (1934,p.267). questioncan be, and hasbeen,subjected' .My aim, then, is to introducea principleof confirmationtheory having the samelogical statusas, though a di{ferentcontent from, Popper'sprinciple of severetesting.This principlewill be calledthe It surplus. is intended to supplement'rather principlaof axplanatory than replace,Popper'sprinciple of severetesting, with which it is cntircly compatible.The idea behind it is this. The principle denies that if e follows logically from h, this automaticallymeansthat e supports h. Not all facts which follow from a given hypothesis of so support that hypothesis, the principleclaims,but only a subset thesededuciblefacts - a subsetwhich constitutesan explanatory

Iti

F T

rl

I
216 Scienceand Metaphysics

Falsificationism the Duhem-Ouine and Thesis

217

surplus. In particular, it will transpire that O does nor support M & (if M, then O). I will now artempr to formulate the version of the principle of explanatory surplus which applies to the kind of case we have been considering.I Let us suppose that a scientist is trying to give a theoretical explanation of a set of facts which I will denote by fi . . . fn. To do so, he or she makes a number of theoretical assumptions which l will denote by Tr . . . Tr. The facts in qucstion ttrty b" concerned rvith single events or may be universal laws connecting observables, such as 'All ravens are black' or Kepler's laws. In either case we assume that fi . . . f,, are well confirmed by observation and experiment, and so can be assumed to be true (rt least when interpreted as approximations) while the attempt at theoretical explanation is being nrade. Let us now further suppose that each of fi. . . (, follows logically fronr sonre subser of T1 . . . T" together with initial conditioni which are established by observarion and experimerrr. So our assumption is thar for each i (1 < i < tr), d follows logically frorn a theoretical systemof the form Oi & T;1 &. . . & T;o, where O; is an observation statement, and each Tu (1 < j < p) ir'a nrember of thc setTl ...T. . The question before us is: Given this general situation, to what extent, if any, are Tr . . . T. supported bV f, . . . fn? The principle of explanatory surplus sratesthat Tr . . . T, are supported not by all the facts they explain, but only by that fraction of the facts which can be considered an explanatory surplus. There is implicit hcre an economic analogy. The successful theoretician is like a successful entrepreneur. To be successful, an entrepreneur has to choose invesrments It. . . I, for his or her capital in such a way that he or she obtains a surplus, and the bigger this surplus, rhc morc successful the entrepreneur. Likewise, to be successful, the dreoretician has to choose the theoreticalassumptionsTr . . .T, in such a way as to generate a surplus, and rhe bigger this surplus, thc more successlul the theoretician. In the first case, the surplus is an economic one, and takes the form of an excess of receipts over olltlays. In the second case, we are concerned with an explanatory surplus, which consists, roughly speaking, of an excess of facts explained over theoretical assumptions employed. Next we have to consider how the explanatorv surplus should be estimated. The sirnplest and rnost stiaightforward methocl is to subtract thc number of theoretical assumptions usecl fronr thc nurnbcr of facts cxplained. So, in our notatiorr, thc size of the

explanatory surplus is estimated as n - s. Thus, if a theoretician has to explain n facts, and needs to postulate n or more than n theoretical assumptions to do so, then the facts do not support the theoretical assumptiolls at all, even if rhe theoretical assumptions do explain the facts (in the sense of the dedr.rctive model of explanation). A theorc.tician in such a situation is like an unsuccessful cntrepreneur, who either breaks even or makes a loss, but in anv casefails to make a profit. Let us see how the principle of explanatory surplus deals wirh Ayel example. Here we have drat O follows logically from {e -O M & (!f M, then O), and the guestion is whether supports M & (if M, then O). Well, we have two theorerical assumpiions M and (if M, then O), but only one fact is explained - nameiy, O. So there is-an cxplanatory deficit rather than an explanatory suiplus, and therefore, according to the principle of explanatory surplrrr, O does trot supporr M & (if M, then O). I will now show that we can deal with the tacking paradox using tlre prirrciple of explanatory surplus. Ler T be a bonajtle scientific theory, and let us tack on to T nr arbitrary meraphysicalassumprion M to obtain T & M. Intuitively we prefer T ro T & M, but wliy? If we accept thc principle of explanatory surplus, then T is always bctter supported than T & M by given evidence, for, since T & M hrs one more rheoretical assumprion (namely, M) than T, the c'xplanatory surplus generated by T & M will always bc smaller than that generated by T. This certainly gives a rcason for preferring T to 'l' & M, and, in gencral, the principlc of explanatory surplus nlotivates the search for theories which are as simple as possible, in thc senseof containing as few theoreticalassumptionsas possible. Lct us next consider an objection which might be made to our nlethod of cstin'rating the size of the explanatory strrplus. It cotrld bc said that the division into s seprrare rhcoretical assumptions Tr. . . T, or n separarefacts f1, . . fn is rather arbitrary. We might, lor example, considcr rhe conjunction 4,-r & f,, as a single fact f,_1, thereby reducing rhc nunrber of facrs, and hence the explanarory surplus, by one, [t might in some casesbe possible to represent f.-'1 & f" quite naturally as a single facr. For example, let f,,-r:Xisasibling, :Xismale; [ therr f ''- r : X i s a b r o t h e r .

t1 tr

r
||

Thesis and Falsificationism the Duhem-Ouine Thcrc is undotrbtcdlya rcardifficurtyrrcrc, but it does in nry opi'ion, makc rhe suggcsted 'or, of esti'ratirrgtrrc cxplanatori 'rctrrod surplusvaluelcss. a concrcte Irr scicntific situation]whcre^stanclarl linguistic fornrulationsare in usc, thcre wiil gerrcrally,so I wourcl cl:rirrr,bc a rraturalwry of cffccti,g thc clivisilr' into'scparatc f,rcts or separatctheorcticalassumptions.of coutsc, this division will nevcr bc.completcly determinatc,but thc', i' the prescntcorrtext, we are airrringonly ar a ro^ugh _not at anything qualitativecstimate precise a'd qua'ritarive. If we were writing a' artificiarr"t.llig"ii"? program' for exanrple, might wa't to makeo'r estimatc we Jf thc cxp.larratory surplusp-rgcjse quantitativc,but in that contcxt, thc and logrcallang,uagc which.everythingwas formaliscdwould provide ilr a method for distirrguishing separaie facts and scparare thJor"ti.rt assumPtrons. There is another consideration which goes sonre way towards resolving thc problenr in hand. I' practicc,wc oftc' *airt to cstinrate.degrecs confirmation in order to evaluatetwo competing of theories,such as, for example, the Copernicanarrd the ptoicrrr"# theoriesat sonle time ir thc period 1543'-16g7. Now i. sucha casc, the-precisedetailsof how we make trre division i"ro r.p.r"r. i".i, and separatctheoreticalassumptionsdoes not matter too mucrr, provided is done in rhe sameway for eachof the two it ;ffi;; theories, This is all I want to say-rbout thc problcnr in gcrrcralterrrrs. Herc, as so often, the proof-of the pudding is in th"e it principle of explanarorysurprushas becn d"esigned "rti,rg. " to con"nabi-e thcory ro bc applied actual to scientifiiexarnplcs, whcthcr lTrrlio,] I'sto'cal or present*day. in practice provespossibie estimate If it to -riJ the.size the. of exp-lan.to.y ru.piu, in a serisible n"tu,"t-*ry, .,ia the.,principle explanatorysurplus leadsto satisfactory of ."rultr, il ttren ttrcrels a strong casefor adopting it. If not, not. The evalua_ tion of the princip.le thus r mrtti. fo.-r is furthcr investigation, anJ I wlll conrent myself here with showing very briefly hJw the prin_ ciple canbe applicdi' rwo differe't .ri"r, o,i. involving , ,u..Jrriul thcory a.d thc othcr arl urlsuccessful tircory. Thelc casesarc Ncr.vton's laws and Adler's rheory of the infcriority."-pl";. J;, ;;; ncxt scctionwe will give as a third examplc:tli" quanrurn thcory of Planckand E"instein. "".i 10.5.1 Newton's Laws Newton presented theory as ionsisti'g of trrreelaws bf his mbtion and the law of gravity. This was a periectly niltural formulation
;rt the tiurc, atrd we may say' thercfori, that his thcory divided naturally into four thcoretical assumptiolls: Tr, Tz, T3' and Ta, a'd Ta is the larv ,uh..e 'i1, T2, a'd Ti:rrc the ihree laws of 'rotion Perhaps, as suggested earlicr, we should add an A, of gravity. but, for siurplicity, I will <lm.it coniaining thc auxiliary assur.llprions, this. As Jur estimatioir of the size of the explanatory surplus is, in rny casc, rough and qualitative, this onrission will not affect the ' gcncral linc of argument. Let us turn to the facts which Newtoll sought to explain' Here again, there is a natural division into Kepler's three laws of planetarynrotion and Galileo's law of falling bodies' Applying the principle of cxplanatory surplus, we conclude that if Newton had -expla-irred fcplcr's laws and Galilco's law and nothing else,he would not have g"r,"rrted an explauatory rurplus, and his theory would not have bccn supported. At first siglrt, tlris rcsult rnay sccrn surprising, but further reflectiorr shows it to be reasonable. Newton produced a complicated thcoretical systcnl irrvolving new concePts (force and mass) and a bold and curious assunrptioll concerning gravitatiorral attraction. Wl.rat would have been ihe point of adopting such a complicated system if it cxplained no Inore than thc observational laws of Kep.le.r aird Galileo. In such a situation it would surely be better to stick to the observatiorral laws and reject the theory as a piece of metaphysics. This is cxactly what thc principle of expla-natory surplus ingg"rtt by assigning zero suPPort to the theory in this case' "ficrc, howev.t, we are talking merely hypothetically and not in rccordancc with historical reality. Ncwton's theory did not cxplain just Keplcr's laws and Galileo's law, but a great deal more besides' \n the Principia Newton explailred, with reasonablesuccess,the laws of impact, ihc tides, the inequalities of thc Moon's motions, and some planetary perturbations. He was also able to derive results .o,rc..niltg the figure of the Earth and comets. We clearly have here alarge explanatory surplus, and Newton's theory is correspondingly strongly supportcd. ThIs'appiication of thc principle of explanatory surplus justifies to a considerable extent ,onie retti"tks of Duhem's which we quoted carlicr irr 3.3. This is what Duhem says: Thcrcforc, if thc ccrtainty of Ncwton's thcory docs not cmanate from thc ccrtainty of Kcplcr's Laws, how will this thcory provc its validity? It will calculate, wi1l1.a!!- lrigh dgg-ree tPP.I-oliT1!i"1 $e 9f that thc conitantly perfectcd meitt"dt of aigibri involvE, tlii pei: turbationswhich at iach instrnt rcmovc cvcry hcavcnly body from

FI

F F

I F I

fr
tl

220

Scienceand Metaphysics

the orbit assigned ir by Kepler's Laws; then it will compare the to calculatedperturbationswith the perturbationsobservedby mcans of the most preciseinstrumentsand the most scrupulousmethods.

(1904-s, 193-a) pp.

There are two things in this passage wirh which I disagree. First of all, it seems me too strong ro speakof a theoryprouingitsvalidity; to it would be better to refer to a rheory acquiring high degree a oJ confnnation. Evidencecan neverprove that a scientifictheory is true, but it can show that the theory is well confirrrred. Secondly, Duhenr nrentions only the perturbations orbits;but ofher things,suchas of the theory ofthe tides,the calcularion ofthe parhsofconlets, and so on, should be included as well. With.these reservations, would I support what Duhem sayshere, which is indeedan instance the of principleof explanatorysurplus.

22i such a task with success. B is such a person,and this explains Mr why he aced as he did. Here, then, are rwo characteristic Adierian explanations observedhuman behaviour.The questionbeforeus of is whcther such explanarions give any support to Adler's theory. hr order to give a reply, we have to analysethe numbir of theoretical assunrptions usedin theseAdlerian explanatiorrs. clearlv we havethe_ ass'mption, thatall humanbeingihave inferioriry T1, an complex.However, to obtain the explanations, general this "rru-p_ tion must be supplementedby rwo particular asiumption, .on_ ccrning Mr A anclMr 13respectively. Thescare: Tz: Mr- A struggles to overcome his inferiority complex by perforrningdifficulr and dangerous Gatswheneverpoisible. Tr: Mr B is so masteredby his inferiority complex that he will avoid evcn artemptirlg anything, that appears difficult or dangcrous. ln this case, then, rwo facts(f1, Q) are explai'ed by three theoretical :rssuntptions Tz, T3). Thus no explanarory (T,, surplusis generated, but rathcr, arr__cxplanatory deficit, and the'explinationi give no support to Adler's theory of thc' inferiority cornplex.Asiuming, which l thi'k to bc tlrc case, th:rt rll Adler's exphnitions i' rermsif the inferiority complexareof the aboveform, it follows that Adler's theory of thc inferiority con.rplex not confirmable,and is hcnce. is by our demarcation criterion,nretaphysical,

Falsificationism the Duhem_Ouine and Thesis

10.5.2 Adler's Theory the Inferiority of Complex Popper,who worked with Adler for a while, discusses theory of the the irrferiority complex briefly iu Conjectures Refutations antl (1963, p. 35). I will herepresent variantof one of Popper's a cxamplcs. In this example we have rwo facrs: f1 and f2. They can bc described follows: as Mr A is walking beside river. He sees child fall in. Without a a hesitating, jumps in, and gallantlyrescues child. he tlre [z: At the sameplace,but on anotheroccasion, Mr lJ is walking besidethe river. He sees child fall in. Although he can swini a as well as Mr A, hc fcarstlrat hc.might drown if he trics to rescue thc child, and so walks quictly away. (ln onlcr nor to make the story too tragic, let us supposethrt drc child is washed ashorc withorrtdrownirrg.) How are the facts f1 and f2 explained on Adler's theory? According to that theory, cvcryollc hls lrr infcriority conrplcx. Sonrcpcoplc,lrowcvcr,strugglcto ovcrcorlrc this conrplcxby pcrforming difficult and dangerous fearswhenevcrpossible. A falls Mr into this class, and his behaviour thns explained. is Other peoplc, however, are totally mastered their infcriority complex,and will by never undertakcanything which appcarsdifficult or dlrrgcrous bccausc thcy lccl too incrpablc'and infbrior to bc ablc to carry out

fi:

10.6 H ow m uch of Falsif icat ionism be Ret ained? can Sirrccwc.:rrccfrrrrgi'g rhc dc':rrcltio' critcriorr tromJarsifahility to rcnfirnnhility,it nright bc tlrought rlmr wc wot,ld liave ,o ."i".t rrrrrny tlrc nlcrll()(lologicrl of prcccpts ot.liilsiJirttiottisrrr; but tlristrirrrs ort'ot to bc thc casc. I'fact, ncarlyall the prescriptions offalsificxtionisrlr be rctairrcd; cln thcy ncedto bc strpplcnrerrted, r:rtlrer th;rn 'l'lris :rb:rrrtlorrctl. is :r vcry dcsir:rblc situltioir. sinccljrlsific:rtiorrisrrr h:rsprovcclits r,vorthirr nrany branchcs scicntificenquiry. Let us of norvv how this c<lntcs sce rbout. "fbc n'raindiffcrcnccbetrvcenPopper's approachand the one rdvocatcdherc is tlre following. l)opper op".nt". r.vitha three-lcvcl rrrotlcl.Ilc h:rs ohst'rvution (or b:rsi$ st:rtciltcnts, scicrrtilic l:r,rvs or thcoricsrvhich lrc assunrcd be falsifiablc, to a'd rrretaphysical

:
i t. ! :i .t ,ti ?i

Scienceand Metaphysics st:ltclncnts. Hcrc, howcvcr, I lrave suggcstcd a for-rr-lcvclrnodcl. i)oppcr's thrce lcvcls are idcntificc{with nry lcvcls 0, I and 3, but I include a level 2 (scicntific bur r.ror falsifiable), which is nor i11 Popper. The key diffcrcnce, thcn, is thc adclitionof orrc cxrrr lcvcl. I t lollor v s f lo' r t h i s t l m t , : r s f i r r . *s l c v c l - l h y p . t h c s c s o r r l l c o r i c s are concenrcd (and thcre are mally such hypothcscs arrd thcories in science), Popper's methodology of coujectures and rcfutations can be nraintainedurrchanged. this level, hypothesesare put forward At as conjectures. They must then be tested as severely as possiblc. If thc testsrcsult in ref'tarion, the hypothescsnlust bc lfthc 'rodificd. tests result in verification, the hypotheses can be acceptcd provisioually as corroborated, but the need for furthcr tcsring rerirains. All these features of falsificationism as a mcthodology can bc rctair.rctj unchanged for hypotheses ar rhis levcl. Now if wc introducc a Jalsifl,ittg rule for probability statcments(cf. Gillies, lg7l and 1973), then the majority of statistical hypothcses lrecor.ncfalsifiable, lcvel-1 hypothescs. So a falsificationist nrethodology for staristicscan bil retained within thc prcseut frarnework- Thcn again, the prcsent franrework contains a level 4 of nrcaningful mctaphysics, io that the claims of Popper and Duhcm abour gcneral nrctaphysicalidcas actirlg as a heuristic for the construction ofscicntific hypothcscscan srill be supported. Let us now consider levcl-Z hypothcscs, such as Newtotr's first law, which are confirmablc, so scientific, but not falsifiable. Evcn as rcgards-thcoricsand lrypotbcscsofthis lcvcl, nrarry ofthc lrrcccpts of falsificatiorrism ca'be retaincd. The airrris to co'flrrn thcsc lcrrcl2 hy.pothcses; but, in ordcr to do so, it is ncccssrryto tunl thcrn into ievel-i hypotheseswhich can rhcr.r tesredas scvcrely as possible. bc Thcrc are a number of ways in which a lcvel-2 hypothcsis can bc converred into a level-l hypotllesis. It nray bc possible, by adding some extra assumptions, to derive a level-l law front the levcl_2 hypothesis. This level-1 law can then be rested experimentallv. We shall give some.examplcs of this in a momort. Then again, the levcl-2 hypothesis may be converted inro a level-l theorerilalsystenr by the addition of further theoretical assumptio's. The conseque'ccs of this.thcoretical ,system can then be compared with expcriencc. Note, however, that the confirmation of the levcl-l law or thc lcvcl-l thcoretical system coufirms the ievel-2 hypothesis only if the conditiorrs specified by rhe principle of cxplanaiory surplus are satisfied. This produces ,sorne 'ew aird i'rpoitarrt rnethodorogical precepts not to be found in falsificationism. Irr the transition fro'r lcvel 2 to level 1, scientists are required to usc as few theoreticai

and Falsificationism the Duhem-QuineThesis assunlptions as possibie and to dcrive as ntany facts as possiblc. We shall scc in a momcnt how thcsc preccpts wcrc followcd by such great scientists as Planck and Einstein. Since the position hcre advocated supplements, rathcr than it rbarrclons,f;rlsiflcationisrrr, could appr<lpriatclybc callcd modiirtl falsifcationisnr.I have already suggested (in chapter 5) thc use ofthis term to characterizeDuhem's position, but since the present account of science is really nothing other than a systematic development of Duhem's, there is really no contradiction in terminology here. 'fhe rc is orre, frequendy voiced criticism of falsificationism which is avoidcd by modified falsificationism. It is often said, particularly by working scientists, that-the aim of science is not to prove theories wrong, as falsificationism seems to suggest, but to find theories which actually work. This apparent paradox is neatly resolved by modified falsifrcationisnr. The aim of science on the present account is indeed to find theories which are well confirmed, for only such theories can form ihe basis of satisfactory practical applications. However, in order to get theories which are well confirmed, it is necessary to submit any theory we propose to harsh criticism arrd severe experirnerrtal testing. Only if a theory survives this ordeal can it become well confirmcd. Thus criticism and testing are not ends in themselves, but means to an end - that of theories which are well confinned and can form the basis ofpractical applications. Modified falsificationism and its four-level model can also be corrrpared with falsificationisnr and its three-level modcl by seeine l.row well they fare in the analysis of key episodes in the history of science. I think it will be found that modificd falsificationism can in rnost cases furnish a more realistic and convincing account. I will illustrate this by considerirrg just one episode: the introduction of quantum theory by Planck and Einstcin in the years 1900-5.2 Planck's introduction of quantum theory was occasioned by his study of the problems of lreat radiation and, more specifically, of black body radiation. A black body is one which absorbs all the radiation falling on it. Such a body is to some extent a theoretical construct, but it can be approxinrately reaiized experinrentally by a small hole in the wall of an oven at uniform tetnperature. Arry radiation entering such a hole is unlikcly to find its u.ay out again, so tilat the hole aats as a black body. In 1879 Stefan proposed the law that the amount of black body radiation is proportional to the fourth power of the absolute temperature, T. This result was improved b.y Wien, who in 1893 obtained from, ggnelal thermolaw which gave Stefan's dynamical considerations his displacement

224

Scienceand Metaphysics

Falsificationism the Duhem-Ouine and Thesis

225

+ l: , , 'i

i:r i .i

law as a special case. Wien's displacement law relates the quantity of radiation to its wavelength. For a given temperature, there is a definite wavelength (1,..", say) at which the amount of radiation is at its maximurn. As the temperature increases, so l,-"* decreases. This is why we speak of a ilisplacementlaw. Now Wien's displacement law contains an unknown function which cannot be fixed by general thermodynamical considerations. We can calculateit, however, if we assurrre model of the radiating a body. The simplest model of a body radiating at a given frequency is a linear harmonic oscillator. If we adopt this model, we can obtain a fornrula relating the amount of energy radiated at a given frequency to the value of the frequency and the value, T, of the absolute temperature. This is the Rayleigh-Jeansradiationlaw, which apees very well with experiment for low frequencies(or equivalently long wavelengths), but diverges dramatically for high frequencies (or short wavelengths). In fact, it follows from the Rayleigh-Jeans radiation law that the amount of energy radiated should go to infinity as the wavelength goes to zero. This was called the 'ultraviolet catastrophe'. In fact, however, the anrount of energy falls to zero as the wavclerrgtlr goes to zero, Note that thc Rayleigh-Jeans radiation law is a falsifiable law, and, indecd, was falsifiecl in an unnristakable fashion. As so often in science, the falsification of a law led to an advance, for it was at this point in the story that Planck made his cruci:rl intervention. Planck suggested that the energy of the radiarion might not be emitted continuously, but in multiples of some minimal unit of energy, or quafllum. If this quantuin of energy has value ee, sly, the energy emitted by the oscillators can only take values which are multiples of e6, such as ee, 2e6, 3e6, and so on. It is clear that, in formulating this theory, Planck must havc been influenced by atomism in a rather abstract sense,so that, contrary to Duhem, we have here another instance of the beneficial influence on scienceof metaphysicalatomism. On developing his idea, Planck discovered that he had to set ee : hv' where v is the frequency of the radiation, and h is a universal constilr"rt, rrow known :rs Plarrrlrl constanl. Frorn this assunrption,it is possible to derive Planck'sradiation lara, which agrces very wcll with experiment. Let us review the situation at this iuncture. Planck's radiation law

is a level-l, or falsifiable, law, and hasbeentestedand confirmedby experiment.This law, however, wls derived from a levei-2 law postulatingthat the energy of the oscillatorsis quantizedand that the quantum of energy is hv. This level-2 law has so far only succecded explaining a single fact, and so, according to the in principle of explanatory surplus, it has not at this stage been confirmed- although undoubtedlyit has proved heuristicallyvery fruitful. It was lt this poirrt that Einsteintook a hand, and very much altcred picrure. thc The year 1905 was Einstein's annui mirabilis.In that year he publislredthree prpers (all in the samejournal, rhe Annalen der Physik), cachof which made a proforrndcontribution to a different the introduced special theory branch ofphysics.One ofthesepapers of rclativity; another dealt with Brownian motion; but it is the third, publishedin March 1905,which we shall considerhere. This paper has the somewhat strange title 'On 'a Heuristic Viewpoint concerningthe Production and TransCesichtspunktf [heuristischen formrtiorr of Light'. Wc shrll seein n rnonrentrvhy Einstein r,rsed viewpoint'. thc phrrse'hcnristic Irr this paper Einstein was concernedwith a quite different phcuonrenon This was the photoelectric frorn blackbody radiatiorr. by cffect, which had beendiscovered Hertz in 1887and investigated bv Lenard in 1902. Fieure 10.2 shows Lenard's experimental

Light t---t--'

=/t-itt1

I:i,gnrt 10,2

Lcnard's cxpcrime rrtal arrangetncnt for

invcstigating

thc

226

Scienceand Metaphysics

and Falsificationism the Duhem-Ouine Thesis

227

t
I

arrangement.The photoelectric effcct occurs whctr short_wavc (rrltraviolct) light f:rlls ccrrain o. nrctrls,rrrcrcby causirrg crrrissi.rr rlr. of clcctro's. I' figurc 10.2 iight entersa wi'dow, F, and strikes the metal at C. This liberatcsclectrons,which are accclcrated (or rctardcd) thc field bcrwccnC and A. in
Now one interesting result which emerged was that light bclow a ccrtain wavclcng_th, or, cqui,valcrrtly,abovc e ccrtain fr-cqucrrcy, is nccded for arry clectrons at all to becrrrittccr.Tlrc cxact vaiuc.,ithc nllrilnluln lrcqrrcrrcyrrccdcdlbr cilrissiorrdcpcrrcls tllc pxrticrrhr olr rurctll. Eirrstcirrhld rlrc,ingcninus idc:r of uiirg pll'ck's qrrnrrnl hypothcsis-to-cxplai' this rcsult. FIc thcrcforc posrulatcd that tlrc of light - and indecd of any elccromag;retic radiation :lelgy - is dividcd into discrctc quarta, cacrrcontainirg ar crcrgy rrv, wrrcrc v is thc freqrc'cy of thc light, and h is rtl:rrrcl'sco'stiirt. This is, i' f;rct, :r gcncr:rliz;rtiorr I)l;r'ck's hypothcsis,si,cc l)l:r'ck :rppricd'lri.s of q';rrt.') orrly tt> thc crcrgy cxt--lrlrruc l)ctwcc' thc rlicill:rt.rs .idc;r involvcd in his nrodel of black bodv radiaiion. .Having rnade this assumption, Einstein further postulated that when a quentum of light struck an electron in thc rnetal,s surface, it gave up all its energy to the electron. If work of a quantity A n"e,l, to be done for the electron to overcomc the forces Lolcring it r,-, irr. nrctal's surface, rhe' che clecrro' will bc libcrated ;fhv > R, "nty and it will rhen move off with kinetic erlersy E, whcre

t t T

r
:

E= h v -A
I

(l )

T T
I

r
tI

we ca' seeat oncc frorrrthis trratcrcctrorrs bc cnrittcdo'ry wiil if the frequencyof the light, v, is greater than A/h, where A is a constant whose value depends the mctal, and that, if clectro's on are enritted, ..the frequency a'd kinetic erergy will be related by^equatiorr (1), which is Einstein's equariorr the photoclcctric foi effect. If an apparatus that of Lenardillustratedin figure like , ^ y".."ryider 10.2' it is clearthat.thc frcquency,v, of the i'cident ligl-,tcarreisily be controlled. The velocity, and hence the kinetic f, o'f "ir..gy, the electronscan be measurcdby deflcctionexperimc,rts"'o. , Uy countcr-field.Thus Einstcin'scquarionfor thc photoelectric effccl ls a level-l law whose validity can be checked cxperimentally. is It worth noting, however, that this level-l law was derived iro, level-2 hypothe"ispostulatingrhe quantum natureof light. .Wecan now return to the title of Einstein'spaper, arrd explairi rhc reason why he spoke of a 'heuristic vicwpoini'. The quantum theory of

him to dcrive it sitrce enabled for heuristic Einstcin, light was irrdeed cffcct.Evcrrsuclra darirg lcvcl-1cqrario' ftrr thc photoclcctric hi.s thirrkcr as Einstein was reluctant,howcver, to regard the levcl-2 quantum thcor:y of light as anything more than heuristic, for an Thi quantumthcory of light contradictcd rcason, obuiout cnougl"r thc wave theory of light, which was supportedby a wide variety arrd which s'as a corlarrd diffractiott cxperinrcrrts of irrtcrfcrctrcc then that thc seenrcd of scqucncc Maxwcll's equatiolls'It must lr.ave considcretions, Iigltt w:rsrulcd out by tficorctical quitrtu,ntficoryof dcvicc. bc could thcrcforc only a hcuristic :iucl Einsteinhimsclf had doubts about the quantum thcory of light, had evctr stronger doubts. Among those who and odrer scientists wcrc convinccdthat thc qualltulll thcory of light must bc wrong was thc firnous cxpcrimcntalphysicistR' A. Millikarr. Millikan that if thc quantum thcory of liglrt was wrong, thcn its rcasoucd Ilirrstcin'scqtlation for tlrc pltotoclcctriccffcct, was conscqucncc, alsolikcly to be wrong. Hc thcrcforeset out to refutethis equation cxperimentally.After ten yearsof testirrgthe equationas severcly as possible, Millikan was forced to conclude, contrary to his expictations, that the equation predicted exactly the- observed verification,irr experimental rciults, and to assertits unarnbiguous spiteof its uureasonablcness. of Therc could hardly bc a nlorc strikingexample falsificationist principlcs. Millikarr sct out to falsify a law which hc bclicvcd orr as groundsto be wrorrg. He testedthe law as severely he lheoretical could, and only wherrhis attemptsat refutationfailcd was he forced to concludcthat thc law had bcor strongly corroboratcd.Millikan's dccision to attempt to falsify Einstein's equation should not, of course, bc taken as showing that Millikan was a bad physicist, but rather that he was a good one. As Cohen astutely renrarks: 'Einstein's theoreticalsuggestiontended to be ignored more than actively combatted.Millikan, a truly great physicist,was aIl excePtion to the rule' (1985,p. azQ. has the a Not every scicntistwho proposes bold new hypothesis good fortune to find a Millikan who is willing to test it severely'It of the shouldintcrnalize prescriptions that scientists is for this reason and make a point of criticizing and trying to rcfute falsificationism, their owrr theories. An example of a scientist who did act in this way is Mitchell. When he first proposed his chemiosmotic hypothesis,this was too far rcmoved from current ideasfor any othcr scientistto think it worth testing,and Mitchell had himself to testing.' submit his hypothesisto experimental

il

" it , l . t t{ .i ; :i l,
i!i

228

Scienceand Metaphysics

and Thesis Falsificationism the Duhem-Ouine

229

i:

Let us now turn our attention from Einstein's level-l law of the photoelectric effect to the level-2 hypothesis from which it was derived: that is to say, the quantum theory of light, or rather, more generally, of electromagnetic radiation. After Millikan's work, the posirion was the following. From a single level-2 hypothesis (the quantum theory of electromagnetic radiation), two level-1 laws in diflerent areas of physics had been derived and confirmed experimentally. These were Planck's radiation lalv rnd Einstein's law of the photoelectric effect. The identity of the level-2 hypothesis in these two separate caseswas demonstrated in a striking fashion by the fact that the sa[le constant, h (Planck's constant), could be determined experimentally in two quite different situations and was found to have the same value. We have, then, a single theory explaining two distinct facts, and so, by the principle of explanatory surplus, the theory becomes confirmed. Here then we have the beginrring of the confirmation of the wave-particle duality which was to receive further confirmadon in due course. This example shows the great importance of falsificationism and, at the same time, the need to supplemerrtfalsificationismby introdrrcing a new level of hypotheses and r prirrciple (the principle of expianatory surplus) showing how such hypothesescan be confirmcd.

was used in any practicalapplication.It seems,however, that the holds good; that a theory which is appliedsuccessfirlly in converse practiceis therebyshown to be scientific. Broadl ition for

care s often theory mayErpear to be successfulwithout really being so..This carr be illustrated by an incident from the life of the explorer David Livingstone.a Livingstone was a missionary with the Bakwain tribe in Africa in the 1840s. He succeededin converting the tribe's chief, Sechele, to Christianity. Sechele was, irr fact, Livingstone's only convert itr his entire career. Now, as Sechele was the chief, he had to perform the rain-nraking ceremony, an important office in a country with vcry little rain. After Sechelc'sconversion, Livingstone persuaded Scchelcto give up rain-making as a heathen ritual. Unfortunatelv, a prolonged drought set in immediately aftcr Sechele's decision, arrd the tribe eventually forced Sechele to start the rain-making ccrenony again. Now, from our modern point of view, lve can say that the orrset of drought was a coincidence, and had nothing to do with thc cancellation of the rain-mrking ceremonies. However, Scchcle's tribe can hardly be blamed for thinking that the events denronstratcdthe practicalefficacy of their rain-making ceremonies, and confirmed the truth of thc rnythological theorieson which these ccrcmonieswcre no doubt based. We should bcware, moreover, of lssunring that similar difficulrics do not exist in our rnodern society. I have already mentioned (8.6) the difficulty of deciding whether a government's economic policv is really proving successfulor not - a situation which politicians rcadily turn to their advantage. Similar difficulties can occur in medicine. An example is provided by the recent surgical treatments of breast carlccr. If l! cancer lvas detected in a breast, it bec:rrne st:rndardpracticc to remove the wirole breast, on the grounds that any less drastic treatment would leavc cancer cells in the patient, which would c:luse a regcncretion of tlre turnour. Subseqr.renrlv, howcvcr, it was c;rlledinto qrrcstiorr whethcr snch a dcvastatingand disfigurirrg opcration wxs ncccssrry. It was argr.red that rcmoving

10.7 SomeConcluding Philosophical Remarks

I will conclude this chapter, and indeed the book, with two points of a rather more philosophical nature. The first colrccrnsthe question, which arose in connection with Wittgenstein's later theory of language, of the relation between the scientific nature of a tlreory and its application irr practice. Now the :lttel'npt to apply a theory in practice can bc regarded as analogous to sr"rbmittingthe theory to an experimental test. Bearing in rnind the qualifications discussed earlier (the principle of explanatory surplus and so forth), we can say that if a theory passes an experimental test, it is confirmed, rvhile if it fails the test, it is disconfirmed. Analogously, it seems reasonable say that if a theory is applied successfully practice, to in it is confirmed, and so, by our demarcation criterion, mr"rst be rcgerded as scientific. I rernarked earlicr (8.4) that a theory can be scientific even though it has not bccn applied in oracticc. For exrnrplc, Eirrstcirr's .gerrerll theory of rcl:rtivity was corrfirrrrcdby obsr'rv:rtiorr,and so sholvn to bc scicntific, nlillty yctrs bcforc it

l,i 1l

,;l

I
.i,

a a a
I I I

?.410

Scienceand Metaphysics

t ia

.irrst thc tlr'rour itscll, though it wonld indced lcavc somc crrrccr cclls ir.thc iraticnt, would at the sarnctiurc leavethc paticnt stro'gcr, lrrrth plrysically ancl psychologically, arrclso bcricr atrlc to ,".,J,r.. lirlly fronr rhc discasc.Tlrc cxpo'rerrrsof rhis poi't of vicw thus ,r.gtrcc{ tlrat a nruch less drastic operatio'r wo,ril producc a better rc(:()vcry ratc. of coursc, in theory it vrouid bc possible to tcst tlrcsc diffcrc't vicws by dividing patierrts at randonr intc, trvo llrl)ups, giving those irr one group onc operation and thosc in thc otlrcr group the otlrcr, and then seeing which group had the bcttcr rcc()vcry ratc. I{owever, such random trials on human guinea_pigs rlisc cthic:rl qr.rcstions.5 it really nroral to give a group-of paticn'is Is trcilr.'lclrt which doctors may belicvc olt g"trc..l thcorcticll ' Thc inlport:ult poiirt is that, without llr0rrnrls to bc unsatisfactory? systcr)ratic cxperimentation, which is not always possible,it nray bc di{ficult to tell in sonle caseswhethcr tlie application of a llcrrtrirrcly tlrct>ryis proving to bc successful not. Despite thesc difficulties, or tlrr:srrr:ccss failure of practicalapplicatiorrs a theory is ccrtainly or of . sigrrificarr factor, a'd should not bc neglectedi, tryi'g ,o .rr"r, tlrc:<;<trrfirnratiorr thc theory. of I . c onr c . r o nly l a s t p o i r r t , w h i c h i s a n a t t c l x p t t o r e l a t e t h c . S0 sr'icrrcc/rnctaphysics denrarcationto the disti'ctiorr between knowlt't.l1ic bcli"f. .rrrrl.. 't1,.".r.r,

Notes

Background 1 Chapter SomeHiitorical

I (and indccd, Thcrc is rro doubt of thc influcnccof Russellon Kcyncs and.induction'Thus Russell ui." ,r".r"; as rcgardsissucsof probability'l of to writcs in'the P"rcfacc ltis broblems Philosophy: havc d1ri1c-d valuable assistenccfrom unpublished writings of ' ' ' Mr J' M' f.y,r". . . . as regards probability and inductiorr' (1912' p' v); whilc '! on K.y,r". says in ihe Prcfacc to his Treatise Probability: 1ay bc pcrceivcdihat I have bccn iliflucncedby W' E' Johnson' G' E' Moore' lrd BcrtrarrdRussell,that is to say by Cambridge' which' with grcat dcbts to thc writers of Contincntal Europc, yet contlnuesln drrcct thc succcssiorr English tradition of Locke and Bcrkeley and Hume' of Mill and Siclgwick(1921,p. v). Onc point to notc lrcrcis that Kcyncs Incntions ttoi only Bertrand Rtlsscll but also G' E' Moorc' Rcccnt of ,"r""rrh on thc hisiory ofidcas hasshown that this ecknorvlcdgcment is Moorc's irrfluencc irighly appropriatc.ln 1903 Moore publishedlris Kcytrcs was most imprcssed flrrlrorrsrrcarisc otr cthils: Priic;pii ethica. in.some *;rl, tlti, work, but fclt thc rrcci to rlter a'd dcvclop its ideas rcspccts.This intcrcst in cthics lcd him to an intcrest in probability. Kcyncs'sconccrnwith rbc problemsof probability and induction Th^us influenccs'One was the ,.ore from the conjunctiott of two sePeratc oroblcm arca of inductivisrn and scientific mcthod, which was being at irrvestigatcd thc timc by Russclland which.is our conccrnin this -Thc other was the problcm arer of cthics which was being book.

lllt,r'r

*itl, "n-id.ntit"-.t *-IifiIi-ilil

;irItc.rrrcrn
,ilt Llnil( t( :cD[a

lr r r os t; r ll t hc scientific thcories introduced bcforc

Iravc

lr
T I

.wls rro scicrrtific knowled

r)rr (l('nr;uc;ltion critcrion, a metaphysicaltheory is or-re^which not is r rrtrlrr rrr;rlrl.by cxpcrience, Hence,.a fortiori, a rnctaphysicaithcory r , r t r r ( ' v ( ' t 111. 11r 111ls c o r r f i r r n c db y e x f l c r i wc i arrd so carl ncvcr t r t , r ylr r lr c r r r is t ic ally a l u a b l cf o r s i i e n c e i [ 'r o , . , r . c r s e s ,w c c a n v nevcr l, r y , l, r ir r rt ( ) nlc t t phy si c a l k n o r v l c d g e s

l rt' t" rr' .k rr.rv l c rl g c . hus.,.al though T r' ct,/physi cs urcari rgfull ' ci is

t t
I
rl

stu d i e d b yM o o r e .In th i sb o o kl w i l l n o td i scu ssM o o r e 'svi e w so n of with qucstions probability and induction, cthicsand their corrncction but rathcr rcfcr thc intercstcdrcadcrto Skidclsky,1983;Batcman,1988; Development. Philosophical arrda forthcoming book by Davis on Keynes's bctwccn Hurnc 2 For furthcr historicaldctailsconccrrrilg thc rclationslrip and Pricc. scc Gillics, 1987. probability,see3 For informatiorrabour thc early history of mathcmatical of David's admirablc1962book, which contrins an English translation of corrcspondencc 1654as appcndix4' the Fennat-Pascal