CONSTRUCTIVE FORMALISM

ESSAYS ON THE FOUNDATIONS OF MATHEMATICS

BY

R. L. GOODSTEIN, D.Llt.

,rofeutu of A19JhemoCla

In Un/ranley Collff' Leicester

UNIVERSITY COLLEGE LEICESTER 1951

: ,~

"

.1

i

!

il '(

. 1

CONTENTS

, :i::NTnoDUOTION: •• ;'EXISTENOE IN lIATIIElIIATICS. ZENO'S

:, PARADOX. '1'IIE bFlNITUI>E OF PRlllES ••

.' i

CHAPTJm I., '.',NUMBER AND FUNCTION

I

0ltAPTlm rr •. ,.'fIlm EQUATION CAI.OUWS

"

. ,,~ ,

ClIAPTBB nr .. 'RBLATIVE CONVERGENCE

. ~ ; .: . )~

ClIAPTBB " IV;:':', ~~ CoNSISTENOY PnOBI.EM

; -. \ ~

'CHAPTEB V'" 'l'MNSFINITl!l NUMBERS

, '4,'

CHAPTEB, Vl~., TJm~OB OF A CLOSED CURVE

, "

- .~.', .

CHAPTEB VU , ; ,'jPHB GBNESIS' OF THE NUMBEU SroNS

6w.4.m' vill ~ . '::,LOUAOE AND EXPERIENCE ••

. r', " " h,. '~::." ;~'l

BlBLIOOBAPllY ,oc"",$ ' ••

"!::'':~''

.,; r .~

" i

'.,." -.'

"bmn',·;.;,

... " ,

" ,

J

:1

7

PAGE, o

11

18

27

42

ii2

!iD

60

72

87

89

·.

PREFACE

The main forct'8 which have shaped tho foundatlons of mathematics over the past twenty-five years have been nntithetieul in purposo but complementary in effect. To Hilbert's formalism we owe the dotulled analysis of tho structure of mathematical systems and tho Imaginative conception of mathematics as its own object of diseourse ; to the constructivism of Brouwer, tho critique of classical logic and tho intuitive notion of a finitist proof.

The formalist-finitist controversy in the foundation.", of mathematics was resolved, in principle, by Wittgenstoin's analysis of tho chaeteristics of. a formal language. Wittgenstein showed that in a. formal language the meanifIIJ of the signs is a purely fundWnaI property of tho language; . it follows that Brouwer's denial of the validity of a formal axiom-the tertium non datur---was totally mistaken. The conclusion to bo drawn from the finitist critique is not that certain parts of mathomatics are incorrect but that tho currently accepted interpretation of tho signs, in ]?articular the interpretation of the quantifiers A and E in terms of uirlversaJity and existence is untenable_ One cannot dispute a formal

equivalence like , I

- Az(P(z» = Bz( - P(z»

but one may well b8 able to show that the use of the quantifiers A and E iJi the formula. is not consistent with the ordinary usage of tho terms "for all" and "there exists", so that in a system in which this formula holds, . ".A" and "B" are nqt synonymous with the universal and

~tial . operators.. '.)

. ·ConstructiviBin and: formaliam found a point of contact in recur8i,,~ number themy which· was developed by Skolem, and by Herbrand and GOdelin their construction of non-demonstrable propositions. Recursive number theory plays a fundamental part in the fusion of these two modes of thought in ~present work. .

The aim ofCOJf8trumve forlMlism is to replace the intuitive notion of a finitist p~ by the strictly formal property of demonstrability in a formal system. This is accomplished by the construction of a mathematical system-the' equation calculus-which operates independently of the axioms and constants of logic. This system affords a means of proving oortain types of logical formulm and consequently effects a ~)ictjon .0( logic to matheDJatlcs. ._ -

The necessity Jor_some equivalent of the theory of types involves any syQem feUnded upon the concept of class in intolerable complications, but even apart from questions of expediency there are good grounds fOl: denying the class concept a primary part in a mathematical system:

The equation caloulus gives to fU1u:tion the fundamental role - that classical analysis assigns to related classes. A function _ is defined by

10

PREFACE .

111t! introductory equnf ions of it!' sign, which h~' means of the trana· formation rules of the calculus. serve t,1I transform the functlon-sign int 0 a definite numeral. when definite numerule are assigned to the aruurnont JlI:H'I'~ in t.lre function-sign,

Though the reduction of mathennrtics to a formal system independent of h'l!il' solves mllTl~' of tho problems in tho foundations of mathematics -in the sens« that it eliminates them-e-there remains the 'ultimate' problem of tho relation of mathematics to I'u:dity, which finds expression in such questions as: "If mathematics is II purely formal system how is the upplivat ion of mathematics possible !" "How can an 'arbitrary svstem of convent ions' foretell the future I" "Whv does common a'rithmctiC' rerve thc shopkeeper. tho svientist lind the mathematician all equally well !" Such questions form only It part of the general problem of tho relat inn (If language to reality. It is a eommonplaoe that the scientist, like tilt' theologian, creates the world in his own image, but what in fll(,t is created is a language and neither the experiments of the one nor tho postulates of the other show in what way language is tied to reality.

These essays presuppose some know lodge of, and familiarity with, the problems of the foundations of mathematics and do not undertake a hist oricul survey of current theories. The applications of the formal system we construct are given in outline only and demand a not in('onsidcrnhle mathematical technique for their appreciation, but the system itself, Loth in form and content. is probably one of the simpJest of its kind, The equation calculus WIIS eon ceived some ten years ago and many of the thoughts herein expressed date from the Bli.me time. The fir~ draft of the essays war completed a year later but the final revision and rewriting were delayed by the war ; in the intervening years I have subtracted from, rather than added to, the original draft. so that the essays may be said to be more sculptured than constructed.

, Of the many friends who have helped. encouraged and inspired this1

work, first and foremost I must mention Ludwig Wittgenstein, to whose lectures at. Cambridge between 1931-34 and the many conversationa I . was privileged to have with him, I am immensely indebted; only in recent years have I grown to understand how much he taught me.

To Paul Bernays I offer gratitude and appreciation for his support, advice and help given so generously for many years.

)Iy last word is for my dear friend Francis Skinner, who died at Cambridge in 1941, and left no other record of his work and of his great gifts of heart and mind than lies in the recollections of those who had the good fortune to know him.

R. L. GoODSTEIN.

Unit'ersity Oo1kg~,.

Leicutu.

.A ugU8t, 1949.

,.

INTRODUCTION

I

I

:

EXISTENCE IN )1ATHEMATICS. ZENO'S PARODOX.

THE INFINITUDE OF PRIMES.

The great discoveries in mathematics are not in the nature of uncovered secrets, pre-existing timeless truths, hut are rather constructions: and that. which is constructed is It svmbolism not a proposition. Tho power of a living sym holism is the source' of that insight into mathematics which is termed mathematical intuition.

, In, the foundations ~f m~t.hem~tic-s a formal calculus plays the part which 18 taken by symbolism in the informal development. A symbolism leads on, a formal calculus leads hark, and JURI, as a formal calculus rightly is felt by creative mathematicians as II harrier to the fr~ expression of ideas, so in the critical study of the foundations, symbolism

is a source of error and misconception. •

The foremost question of the foundations of mathematics for the last twenty-five years concerns the lej!itimney of certain methods of proof in mathematics. What makes this question so difficult is the absence of any absolute standard, outside mathematics, with which mathematics can be compared. Philosophers have held that such a standard is to be found in a study of the :\1ind; that just as the laws of Nature are discovered by observation of. and experiment in, natural phenomena, so too the laws of mathematics are to he found as laws of thought, by n study of thinking processes, Yet. if we consider, we find ·t~a~ t~e CLaws of, Nat';1re' are ,bu,t empiricl~l hypotheses, subject to limitation and modification, admitting excepnons related to time and describing the world as it is, whereas the rules of mathematics are mathematics, timeless because they are outside time, independent of all. obser\'lI:tion and experiment and accordingly neither true nor false, expressing no property of the world, neither validating, nor validated by, any fact. The Claws of thought.'. if by the term \\'0 mean 18ws ~o~ulated by experimental psyC'hologists, 110 more form a standard by _ which tho rules of mathematics eun he tested. than the deductions of & Martian, from observations of the game, test the validitv of the rules of

chess. - •

Wl18t then is the meaning of the controversy between formalists and construetlvlats t The forma lists say thnt the criteria by whioh formal systems lire tested are the criteria (If freedom-from-contradiction and completenoss, IIl1d all their effort·!; ill the JlHRt twenty-five vears have been directed towards proving that a formal calculus, liko' Principia

12

INTRODUCTION

Mathe1ll1llica. a calculus of implication, disjunction and quantification, contains no insoluble problems, and in particular towards the construction of 1\ proof of the non- contradictoriness of this calculus. This preoccupation with contradiction springs from two widely different sources. From tho timo when language first ceased to be only a vehicle of communication and became itself an object of discourse, men have invented paradoxes. Already in the oldest paradoxes of which we have written record, the paradox of tho "Liar" and the infinity paradox ot Zeno we find the prototypes of tho paradoxes of the present day. The construction of formal sYStt-IDS, the very object of which was the resolution of those paradoxes, has accomplished only their multiplication. It seems as if the elimination of a paradox can, so to speak, be achieved on only one plane at It time and at the cost of fresh paradoxes on higher planes. Rather, this is the impression which the logistic technique of paradox resolution has produced, for in fact tho roots of the paradoxes lie in this very technique.

The second source from which the fear of hidden, yet to be dis. covered, contradictions springs ill tho uncertainty which every thinker has felt, particularly in recent years, regarding tho significance of postulational methods in mathematical philosophy, 1\ feeling that tbe postulation of the existence of even a mathematical l'ntity is entirely specious, metaphysical, and in no. way compar'l~le to the in.,·ontion of a physical entity to serve as a medium of expression or a. physical model.

Exi8tence in Mathematics. Problems regarding the existence of mathematical entities are of many different kinds. Contrast the qU08jOns. Are there numbers, do numbers exist? Does the real number "e" exiSt 88 something apart from the sequence I, 1 + I, 1 + 1 + 1/2 l,' 1 + 1 -I- 1/2 ! -I- 1/3 !, ... 1; Is there a prime number greater than 1010 1; Is there a prime pair less than 1010? greater than 1010? To the first question one may answer: Amongst the Big7!8 of our language we distinguish tho numerals, or number-signa, which are con. structed from the number-sign "0" by the operation of placing a vertical stroke after a number sign; the term 'number' is thus a classification index of signs. The sense in which we can say that numbers exist is that number signs are used in our language. Such questions as "have numbers an objective reality", "are numbers subjects or objects of thought"a.re disguised questions concerning tho grammar of the word "number" and ask whether or not we formulate such sentences as : That which you sec, hear, taste, touch, etc., are numbers.

The second question is concerned with tho meaning of limit.processes in mathematics and with tho concept of an infinite sot. To say that the real number "e" has an existence independent of the convergent sequence 1, l·j 1, 1 + 1 + l/~!, .. : is equivalent to saying that some infinite

l '.

ij

U

•

" •

J

.~

I

INTRODUCTION

13

proce88. is completed, Col' instance that the process of writing down all the aigita in the decimal expansion of e has been carried through. In what.sense Can an infinite process be thought of 88 eompletod t An infinite process is, by definition. a proooss in which each stage of the process is followed by another stage just as each numeral is followed by another, formed by adding a vertical stroke to the end of the numeral, An infinite process is therefore an unflni81rable proCOSR. a process which does not contain the possibility of being completed. A completed infinite process is· a contradiction in terms.

Pile relation of Zeno'B paradoz to the formali8l·flnitisl con1roversy. It is, however, commonly argued that we can conceive of a completed infinite process; that in fact, were it not so, Zeno's famous argument. would f?~ce us to deny the possibility of motion. For in passing from one p~8Jtlon A, to another B, a body must pass through the mid-polnt A1 of AB and then through the mid-point Az of AlB, and then through the mid-point A3 of AsB, and 80 on. Thus the motion from A to B ~y be considered to co~ in an unJimited (infinite) number of stages, VIZ., the stage of reaching A" the stage of reaching Az• the stago of reaching As, and so on .. AfOOr any Btage A. follows tho stage A .... l and no matter how ma.ny.of the stages we have passed through we have not reached B, and 80 we never reach B. But if motion from a point A to any point B ill not possible, then no motion is possible. Thus Zono argues; and .by retludio ad absurdum (for motion is certainly possible) it follows that the motion from A to B must be regarded as a' completed infinite process. The fallacy in this discussion is by no means easy to detect and seems to have eseaped the notice of mo.ny competent thinkers.

. l! we say that motion is possible we are appealing to our familiar expenenee of pbysieal bodies changing their positions. Let us imaeine a man rwming along a race track across which tapes are strung IL "few feet from the ground. We may 8UPP080 the track is 100 yards long and that we CODllll8JlCO to string the tapes at the 50 yard mark. If we call the ends of,the track A, B and the 50 yard mark A1 then Ao is the mid point of AlB and 80 on as above. At each of th~ points At. Aa, As, ••• a tape is strung acroaa the track. As· a man runs from A to B he will break each of the tapes we set up, and if we 8UppOSO that a tapo has been set up at each of tho points A1, Az, As, ••. then the runner will have b~ken. an infinite number Df taP!B' In putting the ~en~ in this form we have only pJac:oo the difficulty in a more ob,:o~ light, for we are now confronted WIth the task of setting up an unl!mi~ number .of. tapes, or, looking at it from another view point, of ISOlatmg an unlimited number of points. . On the one hand we have the po~bilit.r of passing from A to B and ~e unlimited possibility of specifymg pomts between A and B (an unlimited number of fractions between ~ and 100) and on the other !wtd.the impossibility of isolating these points on the track. How 18 this apparent incompatibility

resolved 1 ' .,

16

INTBODUCl'ION

The infinitude oj primeJl. We come now to the third question "Is there n prime number greater than 1010" l Consider first the question:

Is thoro a primo number botwoon 10111 and 1010 + 1O"? Tho nine numbers 1010 + 1, 1010 + 2, IOIll + 3, 1010 + 4, 1010 + 5, 1010 + 6, 1010 + 7, lUlU + 8, 1010 + 9, ('1111 be tested to find whether or not they ure prime, that is to say. each of the numbers ma.y bo divided in tum by tho numbers 2. 3, 4, 5, 1I}1 to 1O~ and if one of tho nino numbers loaves a remainder not I~ than unity for each of tho divisions then that number is primo; if however each of tho nine numbers Ieave» It zero remainder for somo division thon none of tho nino numbers is prime. In the same wav wo can test whether unv of tho numbers between 1010 + 10 and 1Oio + 20 is prime, lind, of {:ourse, tho test is applicable to a.ny finite series (i,e., a series in which the la .. t member is given), Thus tho question "is there a primc number between a and b" ma.y bc decided one way or the other in a spoeifiahle number of steps, depending only upon a and b, whatever numbers a and b mal' boo Whon. however, we ask whether there is It prime number grouter than lU10 tho test is no longer applicable since we have placed no bound on the number of experiments to be carried out. However many numbers greater than 1010 we tested, we might not find 1\ prime number and yet should remain always unable to say that there WIIS no prime greater than 1010• We might, in the course of the experiment, chance upon a prime number, but unless this happened tho test would be inconclusive. To show that the test can really be decisive it is necessary that we should be able in some way to limit the number of experiments required. and this was achieved by Euclid when he proved that, for each value of n, the chain of numbers from n to n ! + I inclusive, contains at least one prime number. The underlying ideas of this proof aro just thnt 11 ! + 1 leaves the remainder unity when divided by any of tho numbers from 2 to n, and that tho kast number, above unity, which divides any number is necessarily prime (every number has a divisor greater than unity, namely, the number itself, and the least divisor is prime since its factors will aJso divide the number and 80 must be unity or the least divisor itself) ; thus the ka8t divisor (greater than unity) of n ! + I is primo and greater than ". What Euclid's proof accomplished is not tho discovcry or specification of a prime number but tho construction of a function whose valuea are prime numbers. We shall have further occasion to observe how often mathematics answers the question "is there a number with such and such properties" by COfl8tructing a function; the manner and kind of suah constructions "ill form the subject of later considerations.

.

When we turn to the question concerning the existence of a prime pair greater than 1010 we are faced with the endlese tnak of testing, one after the other. tho primes greater than 1010, of which, as we have soon, we can determine 118 many as we please. to find whether there are two primes which differ by 2. In this instance no function has boon constructed Wh086 values 'form prime pairs, and there is no way of deciding the question negatively. We have asked a question-if quution it be

I

I

!!

I

Ili'I'BOD17O'l'IOll'

17

-to which there is no possibility of answering no and to which the answer ye8 could be given only if we c1umced to find. in the course of the endless task of seeking through a succession of primos, I~ pair of primes which dift'ered by 2. The formalists maintain that wo can conceive of this endless task as completed and accordingly can say that the sentence "there is a prime pair greater than lOW' must be either true ~r false; to this, constructivists reply that a. "completed "endless. taak" ~s a selfcontradictory concept and that tbe sentence there 18 a pnme prur greater than 1010" may be true but could never be shown to be false, so that if it be a defining characteristic of sentences that they be either true or false (the principle or"the excluded middle) then "thoro is lL prime pair greater than 1010", is no sentence. This dilemma h~ led s0!ll0 constructivists to deny the principle of the excluded middle, which means they have changed the definition of "sentence"; others to retain the principle, and, albeit unwillingly, reject the unlimited existential proposition, whilst the formalist retains both the principle of excluded middle and tho unlimited existential proposition together with an uneasy preoccupation with the problem of freedom.from.contradiction. Tho real dispute between formalists and constructivists is not a dispute concerning tho l~timacy of certain methods of proof in mathematics ; tbe constructivists deny and the forma.lists affirm the possibility of completing an endleaa process. "

~. :

.. , , .

.;

• Y:

·.X

",J:l .. . "~ :. t;: .'

; ,

~t __ •

'"

' . .,_.

"

18

CONSTBUOTIVB WBMAT,'UJM

ClIAl'TBB I.

NUMBER AND FUNCTION.

The conlenl oj a Jornud .~ystem. Mathematics hJI8 been defined as the class of propositions of the form "p implies q". where p and g are propositions. The underlying idea of this' definition is that. no mathematical statement is true in isolation. but is true only if some other statement is true; and this regressive process leads back to certain aMnS, statements which are in a sense outside mathematics. concerning whose truth mathematics hll.F nothing to say. but which are simply accepted in the way the pieces of a board game are accepted. Some writers on mathematical philosophy regard tho axioms as hypotheses, unprovable but subject to the empirical test of experience, and modifiable in the light of experience. whereas to others the axioms are just arbitrarily chosen "initial positions", the mathematic based upon thl:Dl forming just one of many possible mathematics. Common to aU existing systems of mathematical philosophy is tho view that mathematics is 0. system of conditional propositions, existential or universal, built upon the notions of conjunction, disjunction, negation and implication. Non-formal presentations of mathematic!! include, in addition to the mathematical signs proper, Iinkago terms "and", "or", "not", "if. then" • "there is" and "for all" and also directions such 11.8 "by substitution in such and such an oquation", "by means of such and such a rule of inference"; in formalised systems only tho mathematical signs, and signs for linkage terms. are elements of the system, tho directions being regarded as commentary on tho system. That the directions aro not part of the system is evident since tho directions are not operations of the system. but doscriptions of operations; what is doubtful is that the linkage terms should form part; of tho system, and not be treated also as commentary; for is it nut a. commenl on the system that euch-and-such propositions and 'such-and-such propositions imply such.and.such ,i" proposition? IT we look closely at a formal system of the conventiorial type we find in fact the anomaly of a dual, use of disjunction. implica~ etc .• what may be called an internal and an external use. For exampl8 if "-+" is the implication sign in the system. the systom' will contain a rule oj inference that in a prooJ uq" may be written down after "p" and "r -+ q" and this rule of inference is completely independent of.e axioms for the sign "-+ ... .such as "(p {it (p-+q» -+q"; the'rule of infeNJloa , being an external use of implication. the axiom for "-+" an internal lISC. The necessity for the' rule of inference' shows that the inclusion of an implicative sign in the system fails to maim deduction or derivation lID internal part of the system and so the inclusion of the implicati~, sign ill without object. This suggests the possibility of presenting mathematics as a. formal system in which only mathematical terms are clements of the system. both linkage terms and directions formiDga commentary on tho system, and ill which, accordingly. there are 110 axioms, only rules of inference. We shall describe such a system in outlino in the next chapter.

l!l

, Numbets., Many of the sips.. Of. ir. langu. age are classification indices~;£

of other ~ sip Thent are' tor instance such classifications as Mme,. injltt.tion-.si{pa, adverb i in partioular the word number is a classi- . lcation . index of signs, the numerals or number-signs of the langua~.

The claaaification of a sign is given, not alone by the shape of tho SIgn

but by the role played By the sign in.. tho language as a whole. Thus .. we Ulfght say,that the number-signs 8!e the signs which playa number -; role. or that ''number'' is the name of one of the roles taken by language . signs. As to the number.signs themselves •. these may vary fro~ be~ds

on a wira.to rows of dots or dashes on a pIece of paper. The dlvel'8lty ofnuinbarsigns has not however pzevented philosophers from ~ttempting to find something common to all number sigDa (apart from their common role) which might be regarded as the essence of the signs. number itself. Both Frep and RU8II8ll ae&od a number to be a class of related elesses, • which. since in their terminology a c1asa of objocts is ide!ltical with the common property of the objects, is equivalent to defining number 88 the common property of Jlumberaigris jwbat for example the numbertwo.aigns(such as two,beads or a pair of shoes) were considered to have in common was the property ,of .~. OD&oOne related. A partioular class is eho8en.8s the,~dard two-., perhapa a certa.in pair of nails. and another olasa; is, .or is ~t, a two-sign according as the elements of the clasa are. or a.r8.11Ot.one,One related.to the elements of the standardtwo 'class. that is p.oOOrdiug as thlmtis, pr is not. an element of the class corresporuling'to '8aohehml~t, ot.,tha standard class and an element of the standard class COl'.I;'OIIpOil to ,each element of the other ola&<J. The ,.confUSions underlying' thiS definition are concealed. beneath the Word "cottaspo~·~ HOw'(lo w&1mow'whather two elements eorrespopd or not" A cup standins on.a saucer ha"fa an obvious correspondence,

so too a'l;fiod ~ & tWig fil its beak. 01"80 pencil mark on a. sheet Of paper ; but it is·.~!tly f~ to say.th9re is a. correspondence bctW8OJ1 the~'of"an)(tWool88sas,'(ot"the same number of terms); and

~,oDesa.Y_B tWi. ". 8. WD..~.' the. re. is ~t¥stual correspondence. such a. cc)rrespoDdence almio·""'t..in be established ··the 'bilit to which we

r8rer liltlst' .1HJ;".!;I,. Posai~ty •• '~~e~f •. !at a condition fOl':\t.1tIiI~ ~.' ~tm 'equa);'QUDlbet'of terms.

, _.. ':. -.'-.~.' '~ i.:: .J:·;1.",: .» :~.::".:. ~':" ~ ro~ . c • '1 " .. -' •

,. ':,' ~,,(1'1WJ;',nt!eP,/~.fi~tmIIl»~.~ cl4Wd by the transJormat~ ',; . rides M;mu.mbet- ~ryJ"N1lmion:,_~.." ,ItUQ.ftOI a one-to-one ,elated pcnr < 'oJ ~"'t1&M~mta /I,futtt:tion; ~ 'r.( function whit:A dderminu a

pair'(lIOf&e.f4.oM·~ ~:l .;""" .

" __ ,:.. '_. .,#,,' (""""'.11"

·'1.;'Nosign is qUintA!8seriii&uY'8 numb8rsign. or the sign of a particular nUmber; ,A. pack ofpIaying cards, aWords a good example of thiR, the titUuu of the card.a. depending not on the cards themselves but upon the game. that is being 'played with them. In the game of Bridge the

Jac1c card has a greater ~afne t~ail the ten. whereas in Soixante-six the ten exceeds in value any picture card: 'In a less obvious sense the relative values of the cards may depend not only upon the game played but upon the state' of the game.

r·

.... ';' 'r.~~·"

t., .. , •

20

CONSTBUCTIVB FOBMAT ISM

When we pass from a 'finite' arithmetic with a specified list of ( number-signs to our familiar arithmetic of unbounded extent we find in addition to actual number-signs, a prOCf'88 for the construction of numbersigns, of snch 1\ kind that the applicntion of the process to 0. number-sign generates a further number-sign, The conventional number-signs are \

I the signs "0" and the signs formed by replacing the letter "z" in the complex sign "e"' by u number sign; thus "0", "0"', "0"", "0"''' are conventional number signs. An alternative system of number signs \ could he defined to be constructed. for example, from the number sign

l"O" and the signs obtained hy replacing "x" in "8(x)" by a number sign, giving "0", "8(0)", "8(8(0»", "8(8(8(0)))", etc. To my that the ,I system. of number signs is unlimited, or infinite, mys only that no limit ( 'I to the application of the number sign generating process is specified. I '. The po88ibilitg of constructing number signs is unlimited, but what is

constructed will always terminate.

The number signs do not form an isolated system but are contained in the parent body of language signs. That is to say, the number s~ are given roles alongside tho other signs of the language. The sign '0" is used synonymously with such terms as "empty", "free from", "absence of", "no" or "none", If the beads have been removed from a wire the wire is said to contain no beads, or 0 beads, or we say that the wire is empty of beads. Tho terms "one", "two", "three", etc., are synonymous with "Om, "0"", "0""', etc., respectively, where by synonymous signs we mean signs that may replace or be substituted for one another.

Counting objects, "one", "two", "three", "four", "five", and so on, is trsnslating from one number notation to another by means of the transformation rules "one and-one is two", "two and-one is three", "three and-one is four", and 80 on (perhaps 'saying' each and-one by pointing to an object) i but making a tally, like cutting a notch in a stick Cor every sheep that passea the gate, is just copying, a number sign. Recognising two marks as instances of the same sign (or two noises as instances of the sameaound, etc" or matching two colours) is the fundamental operation in a language. Such recognition iJlvo~ves overlooking some differences but not overlooking others. To ptU'eeite, that "0"''' and "0' .. " are the same number sign involves overlooking' the different places the signs occupy but llot overlooking their fonn (or rather overlooking some differences, but not others, in their form). Sometimes this 'recognition' is immediate, sometimes it involves several quite elaborate operations (replacing each objeot counted by a bead on a wire, then replacing ton beads on one wire by a bead on another wire and so on). Remember that a 1fIllI1Jine may accept pennies ,but reject half-pennies, or 'bad' pennies; in accepting as the same, or rejecting as different, two signs, we are acting as slot machines.

",

Fact and com:ention. The sentence forms of a langUage are of several types, two of which nrc the 'matter-of-fact' and the 'sign. convention' sentence forms. There are sentences, containing number signs. of both

e

,

• ~ •. , .... ,~ ~.'·~,to,::·l· ,.~-.'~ c

21

these types, and this is what is meant when an opponent of formalism says that mathematics is not just a game. For example, contrast tho aensence Corms "There are three pennies in my pocket", "there is one and two pennies in my pocket", "one and two is three". The first two are 'matter·of·fact' sentences, whereas the third is a 'sign-convention' sentence, and according to the convention expressed in the third sentence, the convention that "one and two" is synonymous' with "three", the first two sentences are synonymous. The equivalence of the first, two sentencea is a consequence of a rule of arithmetic but the truth or falsehood of the first (or seeond) sentence is a consequence of no arithmetio rule whatsoever. The first sentence expresses the fact that the physical object "the pennies in my pocket" is a three.sign-and this is an empirical hypothesis, like "there is a smudge on the third page of this _book' :-the third sentence is non-emplrleal, neither true nor false, but SImply 18, or is not one of the rules or a deduction from the rules of a oertairi arithmetic. This distinction is perhaps best brought out by mying that in the case of the first sentence what we test is the pennies whieh I have in my pocket, and in the case of the third sentence the test is to see .whether the sentence itself is one of a certain list. of sentences. The aign eonvention sentences (which contain the wholo of mathematics) stand to language in the relation of a dictionary i just as 0. dictionary places side by side the language signs which may replace o~e anoth~r without changing the sense of a sentence, 80 too mathomatu~al equntions place side by side • whioh may be substituted one for the other.

i .~. '. '

Variable.8. "l'henumeral variable is all instan~ of a general noun, and as 8Uoh its use is akin to that of a blank space whieh may be filled by certain signs. The college rule forbidding atwknl8 to leave their place ofleSidence after,a qortain hoar is e~ed 'by moans of the variable aIgn "student", If the ~ read ~'students in residence from January

1944 to JanuAi'y 1946". the variable term "student" could be replaced in the tc:mnaJation o,r the 'I'ille..by& list of names, the names of the students· in residence 19t44. ·and iii ~ this fact which has tempted Dl&thematical:p1iil~hel$ tQ\,.b,eti8Vl:t,.tMt"a,variable is a claBa of objects, but, if the l'l1ktilf mtend8!itOapply not o~ to present,. but also to f~ture studentJ,' th&~,!eP~ ·ofthe.variabJ& tdp, ~Y a list !,f names l!i no 1~ possibl9 an.d~.~ are obJfced· tl). 6nd'Q,mterpetatlon of ~nab"e 'signS which, does Df)£' id~ & ,~b~~:~'a ~ of ~es RJV:eD l!l

", extenalon.·~ ·'ne'trnt!L.,,...IiQ, of'~,se~·~ a vanable wgn 18 QDobaaged,if tb,Q;.,vAl'iahta:lQgRiS reillacied 'bl some C)e~ other sign, for iDBtimce~~th&!8eQtmi~."D1!ltal" ~. ill the earth". the variable "metal" may be replaqed by D&IIl8S of 1,iarticnllar Dletals, suob as "iron" '.' ,.' or,,~·gc»Ja·~.'80 tbat,~e ~ ·~eta1 .. may':~regard~ as a space to ~

ftl1ed in h¥ the name or a ~ •. What 18U1lS&tiaf'actory about this imalysi8 of the ~ariable is that, we require a .knowledge of all the particulara which oom.e undorthe variable sign before the variable itself is

. de6ned •. thatd to .y we are no better abie,to deJin.e a variable sign as a l»la$.~ ~ be filled by certaip sigDa than to identify the variable

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

..

, l

~-., ;.

>.-:-

" ~ - ..

"

aanteuee. will make the sign essentiaUy a numeral, will fix its role It .ia ~ useful to indi~ on a map which is the northerly direction, and this may be accomplished by aIlxing to the map some such sign as u t N". But.i£ one then doubted whether the arrow pointed to or from the north, it might..be thought neoesaary to add the direction "the arrow

I pointa·nort.h'~; yet. cannot,ODe now doubt which wayan arrow points ¥

t . . +'.. . ,~,. ,; ..

For example, if we define the nunieral variable "x" to bei<'a; bJaDk Funt:lion8, Just as ~ematiCl8 Permits the introduction of an

space which may be filled by a definite numeral, w~ are. invoifAid in a unlimited number of definite numerals, so toO it permits the introduction

vicious circle since we have defined 0. definite numeral to.·~ 'either "0'" of definite function signs without limit to number. The construction,

or ~e sign obtained b~ repl~g ':x" in the ,cOmplex,. ""~. by a or introduction of defihlte f'unotion signs however is effected by many,

definite numeral, and this definition Itself c;ontains avanable •. The and widely diverse, means which fall into two main categories. A

explanation is that the concept of an infinite class n8ceasari1y·cdntains function sign is. a complex sign containing one or more numeral-variablea

the variable concept. The correct formulation of the de6nitioJi of a ,and its method of introduction is such that when all the numeral variables

definite numeral is that the definite numerals are the signs obtaDied from in the sign are replaced by definite numerals, the resulting sign may'

the sign "x" by eliminating x using only the operations or nwlacinl be transformed, by certam simple and specific rules of procedure which

"xt• by "0" or by "x"'; thus the sign "0"" is obtained from z' by we B!I.o.U 8U~quently describe, into a de6nite numeral.

first replacing x by "x"', re~ting this operation gives "i"",replacmg. ., N' ., • .~. • . ; •

"x" by "0" then gives "0"". The variable sign "x" is a blank space either in .mathaD:laticalliteratu:re nor in works on the foundations

which may be replaced by "0" or by "x"'. Tho numeral obtained ttODi·, 9r,math~tica have the methods of introduction of definite function

th . .. '"~ b I' .... b d fi . ral is __ 11_.1 th sips ~ recogniieJl for. what they are, and analysed, We find for

e SIgn x y rep acmg x y a. e mte nume ~ ~. ,.~ . instance the. eo--.I1...:1 .. .I_a..t":_n of function .. couched in vano' us fio .......

successor of that numeral.· . . :' . . ~ \WUIUIIIU........ co_

- .,~"of varyiDg ~ .ot OohereDce; we 6nd the fonction concept itSa1r

In ROme works it if! maintained that 0. consistent use of variabt9 ., sqppqsedly dad.~~:the JIOIloept of a correlation and under the

signs is possible only if the variables are qualified by a propositional. ' .(: ' notio~ ~ ~1at.ion We· fiJ,ld every, conceivable juxtaposition and con-

function asserting that th~ variable signs are variable signs. It is held ... ,_ . jtmction. . This fr8ditlonal'.:J'oilception of function and definition OfJ

that since, for instance. the equation zy = 1P is true if uz" and "11' '-. . fanction is"entfrelK ra"',Let 'as comm.eno& by considering a simple

are numeral variables but is falae if "x" and ~'y" a.re variable • in:~ ,," '·c· ~~ .~~~., or t1a8 aJt.em.ating function. Alt x, wh08B & non.commutativ~ ··alllebra, like the, algebra of matrices,' ~u.er0l8.. .. ~ V&hui:i;iIi'e . . JterO and ~", The definition would take some "zy = yx" cannot ItseIfbe a mathematical proposition and what',~.,·:;·' .. ,such ibtul· as ~ :<:Alt . .t:-= OK~. is: .. and Alt x = 1 if x is odd.

belo~ to ~thematic8 is .some such sentence as "if. "z",lU1Cl.~'tI sre:"f':. Pbservefitsttthat;!~'d~ "$bathe.fonD of an implication; yet.

numeral vanables then zy = 1/%", This. of ·course, re.inttddilces tlut , .. '. '~.':: ,:J"to _saythat 4Jt".,; O'if~:ili,eveDJ.Js'. ~theequo.tion which results

view that the propositions or mathematics tab the form'''p impUes I)~ ,. ., .'. from ~~AIt·:t,bt·O'!jf C~~~J~ ~~lbY mtr~ilmIlber;· or if we,;

It is true that such an expression as "cr;y = F' ~ 'Iii ~tlon Ja "i· ~,. the ,fabctioDB 21'8Q4· 2f~ 1 ~. in.~ _ "AIt 21 = ~ -

nothing b~t a pattern or design. No exp~on is by, ita.fdl'Jp. atone a, '.' "', ..... ~.: ·~ly ,t~~,~·.t2e f~)¥l'\~ It'!iI.the&eelillQtloJia alone which .

mathematical statement, If the rulesofohess aN completitlyforgotteDi~". ,.. ~. expresa,'im,'~~CS":~IHnl~~.*~<.or::the ~n Alt x,·· If we. '

and all record of them lost, then the game of oheas is·lostteven thou,gh' ~ " '., "',,:. repJaoe ':r! by a ~ .~~Ikf!!9'l.!tbD byJDBaD8 of the equatioq '~

every chess-set is preserved in full; the characteriBtla shapes ot.·the f • ",", • 2, 4 ~ 1:= 9 ail4-th&-~e~ D (2C .+,1"1:) ·~.I ;ii traDst'orm "AIt 9" .

chess pieces do not contain the game of chess, and with tJu)<pieoes ma.Jq " . . {.~._ ,1';into.' Itt (~8igna "1~~.~cc2",,·,.£·:~, ..• ;,~,~·,,~ Mlbreviations fgr "0'" .

games could be played. Yet the ruIes of ohes8 do not assart that ira ' " . I ... ".. ... and "0:"',' ~" ~. by 1iD8~oDlt'f·.,~~.O'", .~ = 0',.;:. etc.): .-

ccrto.in piece is the king of chess tAen it has such and such Dioves, for wha\ .'..; '".. Before. h~W8~,.tha '.-.yaia 1# the daJDitiqu Of the function AIt x is r«:

constitutes a piece king if not its moves 1 So; too. how ca11' We say' -. '·~l~te. we m_ ~49i' ~ fUrther·crueitmn of the a.deqoaoy of the

that if "e" and "y" are numeral variables tMn zy:!= 1P; for.ut ma.kea~";.'; ~ de1biihlOD.', In,. the ·CDnventiona}:·fwmulation of the definition. the .

a sign a numeral variable if not the role which the sign playa in a aystmD " . adequacy 18 deemed to be a consequence or the theorem "every number' .

of signs 1 . One could as well demand that "zy = '!P" be ~ bv .~. is odd or evenlt'and thiS'.theorem in its tum a ccmsaquencc of the axiom

the condition "if the "xlt is truly an Xosign and the "," truly a y,sign'~1 ' . I ! ) ~ not 'P" (th8, ~ of the excluded middle) since the proposition

That a sign is a numeral will show itself in the sy!ltlem ; but ~ that 'c a .~ If! ~ til two. or iB DDt· diviBible by two" is an iDstance

is added to the sign itself, be it a star, or speci&l colouring, or.qua.l.ifyin& ~ ., .. 0£' this ~Jl»~: tGkiDf ,the Value .. ~ number Ja divisible by two",

"1 ?c, ,,:ii' :';::.":.0: "<.'·f~;,,:;~.·~::_.,~":il}~~'·~)~~t~:,f":'{7> ':/ .

I~;; .' " :1

22

OON8TRUO'l'IVB FORMAJ.JSM

with a c1a.sa of objects, for just as we- eannot enumerate an iD&Dit& oIasa we cannot ea.y just which are the signs that may 6ll a -oerta.iD .~ sign space. ,. .", l':'.' <;1··'~''';;

The definition of a variable. is not gitJen in.~ o,tia ~ .. ~ Bigns may. take '!'e p_lace of tM fJlJriable ~ tu' bt("~ ruIes]of 1M use' of the tHJnable 8lfPl dsell· " . ':;",:' ';).. '

.... ,~ "~~!~.?It>"'''~~. 'I}.';,."'. ,',.-w.,;,,-.;..," '""J" .. ' ;,.,

'-t

'( '.J,

NUMBER ..urn FUNCTION

23

,;. ~

, -,

24

('OXSTRrM'JYE FORl1ALt~l1

There 1U'l'. however, many difficulties in this line of reasoning. In the first place what is the meaning of tho term "number" in the seateuee "a number is divisible by two" I certainly not a specific number; for am' such sentence as "six is divisible bv two or not divisible bv two" is pri'wable without nny use (If tho principle of the excluded middle. And if we tako "number" to mean "number variable" wo obtain the propositional function "x is divisible by 2" .. or rather not propositional J/Illdio" but pl"opGsitiOlUlI.Junctwn.Jorm, for how can we tell that "x is divisible by 2" will transform into a sentence when "x" is replaced by II definite numeral? )IURt we nut. so to speak, know in advance that every number is odd or even, before we can take "x is divisible by 2" as a value of l' in the axiom "p or not p" ; fur to postulate the axiom "p or uot p" makes this axiom a dellning characteristic of propositions, and according til this defining chnrneterlstic "x ill divisible by 2" is not a sentencaIand therefore not a It'gitimate value of "p") unless we have, in some sense, prOl'ed that "x is divisible by 2" ill nccell8arily true, or necessarily false, for each vuluo taken by x; but if wo I.af}e proved this, the principle of excluded middle is simply superfluous-for to whut purpose shall we po.o;tuitlle what has already been prol'ed ~ The source of our conviction that everv number is odd or even is not to be found in IIny axiom but in the r;,ct that the successor of an evon number is an (wid number, 111111 the successor of an odd number, an even one. The mll·qulley or the definition Alt 2x = 0, Alt (2x + 1) = 1 iR demonstrated by the equations (2x), = 2x + 1 and (2% + 1)' = 2(x'). which express this fact.

Thc Introduction of definite function signs is prO{lruailJe. startinI with definite numerals. numeral-varinbles and the function x', new function signs art' introduced by means of function sigll8 already. Introduced. Tho well known elementary functions like ~e sum funotion Sum (x, y), the product function Prod (x, g) and the exponential function Exp (x, y) are tho first, functions introduced; their method of in~duction is by r«urait:e equations of introduction which take ~e fo~., ..

F (x, 0) = a(z), ~.:(

F (x, g') = b(x, g, F (x, 1/»;.

that is by equatinna oxpressing F (x, y') in terms of F. (:I:: tij:

Thus Snm (x, y), Prod (x, y) and Exp (x,y) are introducod by the equations Sum (x. 0) = x, Sum (x, y') = (Sum (x, y»'; Prod (x, 0) = 0, Prod (x, g') = Sum (Prod (x, y). x) and Exp (x. 0) = I,

Exp (x, y') = Prod (Exp (x, y), x) respectively.

• Xeitller til" brackets ( ) nor the comma between :r nnd y in "P (x, y)" ill a part oC the function sign: tlwy are used only for typographic convenience and could be dispen"'d witb entirely.

e I

NUIIBBR AND PUNartON

By means of these equations the function signs are transformed into definite numerals when the variable signs are replaced by definite

numerals. .".

. Of greater complexity than reoursive introductory equntions, but of fundaDlentally similar character" are multiply.recuraive introductorv

equations, of which an important type is exemplified by .

F (x, 0) = a(x), F (0, y') = b(y), F (x', y') = e(x, y, J' (x, !I».

Multiply recursive introductory equations are also of the (more general)

form "

F (x, 0) = a(x), F. (0, y') = 6(y),

F (x', y') = e(x, y, F tx, d(:I:, v. F (x', 1/», F (x', y».

We admit also the familiar explicit introductory equations like F (x) = a and F (x, y) = a(b(x, y), c(x, tI», whore a iR some definite numeral and a(x, y), b(x, y) and c(z, y) are functioll8 already introduced.

Closed elwin8. The second general method of introduction of definite function signs is by. means of elosed cAaina of function signs. A speoific list of function signs (all previously introduced) is snid to form a closed chain if the snccessor of any value of nny of the functions of the list is itself a. value of one of the functions, "0" boing a value of one of the functions, no two functions of the list having a value in common, and no function. taking the same value twice,

. . .

. If for a given set of valUeS of the variables, a function A takes tho . value a, and a. function B takes ~e value 6, then we call the pair a b

the ~alue. Qf the pair of ~atio~ J.f, B. '

. .·A list of pairs of functions is' ~~ ~ form a closed elwin oJ pairs if (1) no .r.ur, takes. the same valuc.twice, (u) no two pairs have a common

,. value, {w) one peir ~08 the valUe 0, 0, and (iv) for each pair A, B of the chain, each value of the pair of functions J.f'. B and each value of the pair:d, B~ is a Jvalue' at' afunotion pair. of the oliai&;

. Closed. chaina>Yof trilJ:S,~U~~ ~./"iDa.,. be:'Cormulated along

analogons lines. A closed chain expresaea: the concept "every number takes one, and only one, of suob and snob f~'~' .• .

Closed chains might also be defined as folloWs' :':':"'A set of functions form~ a closed chain if every value of x is a unique value of a unique function of the set. A set of pairs' Qf functions forms a closed chain of .pairs if ev?ry .value of the pair (x, 11) is a .~que value of a unique paIr o.f functions of the set, etc. This definition is equivalent to the foregolng, and helps to bring out the significance of the chain concept more readily, but is 1lO~ so convenient in application.

26

('ONSTRt'CTlVE FORMALISM

A function sign "F (z]" is introduced by means of n closed chain. by formuluting equations of the form J' (c(x)) = P (x) for each function c(x) of the closed ehain ; P (x) must be a function sign previously Introdueed. Similarly. It two-variable function sign "F (x, y)" is introduced bv 0. closed chain of pairs, by formulating equations of the form F (j(x, y). k(x. 1/» = P (x. y) for each function pair j(x, y), k(x, y) of tho closed chain of pairs. each P (x, y) having been previously introduced. In each introductory equation. every variable sign which occurs on the right hand side must occur also on the left.

Ezamplell. If Sq (x) = Prod (x, x) and DI (x) = Prod (0", x), then the function "non-square numbers" Nsq (x, y) is introduced by the t·quat.iolU~

Nsq (Dl [z], 0) = Sum (Sq (x'). Dl (xn, Nsq (DI (x)'. 0) = Sum (Sq (x'), Dl (x'».

Nsq (DI (x), y') = Sum (Sq (Sum (x', y'», DI (e)'), Nsq (Dl (x)', y') = Sum (Sq (Sum (x', g'), DI (x'».

The pairs DI (x), 0; DI (x)', 0; DI (x), y'; DI (x)'. y' form a ~losed chain of pairs, and so too Sq (x). Nsq (x, g) forms a closed chain.

,

t
\
1
~
I
i t. •
I ..
. ,
1
l
I
I
,
.J
~
~ 27

CIuPrBa tt.

THE EQUATION· CALCULUS.

In this chapter we begin the construction of a formal svstem composedentirely of mathematical signs and operated bv trllosforrDntinn rules independently of the signs and axioms uf the prepositional calculus, The system may be described 88 a free mriable tlunsfoN1U1lion calculus.

In its full development the field of tho system is tho rationalnumbPl'lI. ~ut .by mea~ of th? ~es of relative C07&t.'ergence and relative conhnutty and diJJerentiabilify the system serves for the expression of a

wide part of classical analysis. .

The signs of. the calculus fall into three principal categories: . numeral variablea a, b, e, x, y, z, etc., definite numerals 0, 0', 0" •...

and definite function signs uf the form Fac (z). Sum (x, y), etc. '

Each formula of the calculus takes the form "a.= b" and expresses the fact that "a" ma.y replace, or be replaced by "b", where "a" and "b" are function Bigos; "b" may be called a tluriaform of "a".

~e obj~ of the .. ~lculus is the rot;.mu~on .of proofs. A proof is a (finite) senes of equations, each of which 18 an mtroductory equation of a ~~on sign,. or 1m equ",tion of the fonn CIa = a" (where "a" is a function sign), or 18 a PHOVED equation. If "a = 6u is an equation of a proof, then the equation formed £rom an equation or a proof by replacing

ClaU by "b" at De ... 1ft""" r'ta . th ..

? . or more JHD"""" 0 I .occurrence m e equation, 18

a. proved· equation. The equation formed by replacing a. variable sign at ~h plsee of its OtlCUlT8llce in an equati~ of a proof. by anothe; variable 81gn, or bya 'deftnite numeral or function sign, is a proved equation. If two function sfgnlJ.~'au, "b" satisfy the same introductory equations, then. "a = b" is a proved equation. (This last rule expresses the f~ that. a ·~ction, is· detmmined uniquely by the introductory

- "equations, of Ita BI8Ilk1·~ ... -: . _ ~.~ .

.. ~ ": :." .. ""~:.:: :~: : .. ":":. :.~'" ::, ".~.:. ;,;:_.:' ,/ . " ; .'

Bmmple.!.'~'(l) To' piove~Um C.,: y) = (Sum 'Cx, y»':

',;;.' We haVe " "",;; .;;: ' ., ," ~, ".,.;: ',

.(1) Sum (x, Of.=:~(2) eum.'(»',.O) ~:r/.' (3) (Sum (x, 0»' = x', (4) Sum (~ tI) ~;~ (~,,I»', ,;". (5) Sum (x',' tI) = (Sum (x', y»', (6) (Sum {~,. (».#:7'" ~(8~ (~, :.1/»')' ~ , '('I) 'Sum (x', y) = (Sum (z, y»'.

'~:~ .... c-,·",,-~,-;,!;--·.rC'4 .~#~'.;.'~,.'.\ -- ,of:_

" Commentary. EqUation (2) deriveaf'rom the 'introductorY equation (1) by substituting "x''' for "XU. Equation (3) is obtained rn;m (1) "ith the ~elp of~e oquation x' = x', ~y replacing the first "x" by "Sum (x, 0)" 88 18 pemuttod by (I). EquatIOn (6) derives from tho introductory equation (4) by tho substitution of "x''' for "al', and (6) derives from (4)

with the help of the oquation .. .

{Sum (x, y'»' '= (Sum (x, V»'.

28

CONSTRUCTIVE FOlUlALISlI1

Equations (2). (!\). (5), lind (6) show that "(Sum (x', y»" and "(Sum (,,., y»)'" sntillfy the same introductory equations, so that (7) is I' proved equation,

(2) To prove Sum (x, y) = Sum (y, x).

Sum (:t., 0) = .I', Sum (x', 0) = x', Sum (x' 0) = (Sum (x, 0»'.

Sum (0, ;r') = (Sum (0, x»)" Sum (0, 0) = Sum .(0, 0),

Sum (0, :to) = Sum (x, 0), Sum (g', x) = (Sum (y, x»'.

Sum (x, y') = (Sum (x, g))'. '

The last tlll'l>~ oquutlons show that Sum (y, x) and Sum (x, g) sa.tisfy the same introductory equations, which completes the proof.

In I' similar way we can show that Prod (x, g) = Prod (g, x).·

The difference Iunetion d(x, y) is introduced with the aid of the 'predecessor' Iunetion g(x) by the equations

g(O) = 0, g(l") = .1', d(x, tl) = x, d(x, g') = g(d(x, g».

Tile principal properties of d(x, y) are expressed by the provable equations

d(,r', y') = d(.r, y), d(SlIm (.I', y), g) = x, d(x, Sum (x, y» = O.

We shall also need the primary functions a(l:), b(x) introduced by the equations

a(O) = 0, a(l") = 0' ; b(O) = 0', b(x') = O.

The equation b(.I') = d(O', .r) is provable, so too the equations a(x) = b(b(x)) and a(Prod (.r, g)) = Prod (a(x), aCYl). The sum, produet and difference functions will also be denoted by the abbreviations x + y. xy and x"':' y.

------ --------'

• We prove fll'st Sum (Sum (;t. gl. =) = Sum (Sum (x. 21). y). by the equatioDS Sum (Sum (.r. y). (1) = Slim (x. g). Sum (Sum (x, 0). Y),=Sum(z, 1/) Sum (Sum (x, y), :') = Sum (Sum (x. g); %)' , Sum (Sum (x, :'). y) = Sum (Sum (2:, z)),. tI) = Sum (Sam (z, z), ,))';

It follo\\'10 that ?"',. '

Sum (Slim (a, g), zo') = (SUDl. (Slim (a, g). xl)' = (Sum (Sum (a, z) g)' : "

= (Sum (Sum (a; z). tI))' = Sum (Sum (a, x)~ 1/),

Then we have

Prod (g', 0) = O. Sum (Prod (y, 0), 0) = 0 Prod (g'. x') = Slim (Prod (y', z). g')

811m (Prod (y. X'" x') = 811D\ (SIIIIl (Prod (g. xl. g), z')

= Slim. (Sum (Prod (g, .rl, x). g').

Prod (g'. x') = SUIl\ (Prod ts- x'), .1"). Prod (g' • .1') = Sum {Prod (g • .r). z) Prod (g' • .1') = Prod (x. g'l.

l'rod (ol, H) = 0,

Prod (II, x') = Prod (Il. x)

Prod (x', (I) = n = Prod (.1'. (I). so tbllt 1'r",1 (II, .1') = T'r,lOt (,t'. n). ",hieb completes the proof that I'r\,d (g, .r) = Prod (J', y),

proving ",lwlII'"

anti "" Furthermore

.. :; .. "

t ~

I

,

;

i

f

"

I

THE EQUATION CALCUI,US

29

Formula! of the propositional calculus may be trall>llatllli iuto the equation calculus with the help (If the primary functions lind the :;UII1, product and difference functions, With the utomie pl'OpulIitillll!l jl, £I, r, etc., we associate variables z, y, c, ote., tho value zero III' tho as. ... ueiuted variable x corresponding to the true proposition p, IItIlI 1I011-ZCI'1I values of x corresponding to the false proposition p. The negation of p, ji, it! associated with b(.r). sinee this function is zoro if .r i,; nut zero, i.o, if p is false, and takes the value unity if x is ZCI'C). Le. if p i,; tl'IIO. Tho disjunction p v q is associated with tho product xy, for XII i:; zero if. and only if, one or both of x, y is zero, i.e, if one or both of p, If is true, The conjunction p & q is associated with the function x + y; this function takes the value zeru if, and only if, x and y are both zero, i,e. if both p and q are true.

By means of the equivalence p _,.. q = P t' q wo see thut, tho function associated with p _ qis b(x)g.

The assertion of a. true formula is translated by tho equation which expresses the function associated with the formula IlS a transform (If zero. A specific formula is, or is not, a tautology aecordina us it:;

associated function is, or is not, It. transform of zoro. 0

, For·IDst~ncti~.the axiOm of excluded middle p v ji is translated by the provable equation e , b(x) = 0; the provable equations () . btU) = o. x' . 0 = 0, b(~') ,0, x'. b(x') = 0 show that x . b(x) satisfil'iJ tho introductory equations Z(O) O,Z(x') .:. 0, where Z(x) is introduced by the equation Z(x) = 0, 'Q.l1dso :r:. h(x) = Z(x). Le, x. b(x) = o.

" Similarly;i; we;'a~:':: sho~' ~ha.t 'any 'axiom' or tautology of the

propositional calculus may be translated by a provable equation.

, ,

,,' THE EQUATION ICAx.cur.'uS 'APToBD8US. TRBBBFORB, A MATBEIlIATICAL PROCESS ro~ i>ROv.piO;c OR ~ISPRO:vmo, FomroL1B'OF THE PROPOSITIONAL

',;, .. ' ",:' , ,~'.;".,~,;,' }~"." . , .. ,' '_,,' ",,;,

: • -s., \;' o.~!!fh B?!,\'!~~ :QJ'~~,~~OCI.ATBP,.FDl'CTION8 THE ELEMENTS

'. :OP THE PROPOIWlJONAL::~US: MAY, DB .:BLDDNATED FROM THE

.' ..e.... { • .

'. !'ROOll' o:r'~~'mATBlWATroA'L'ioBliui:;A. ',;" ':,', "J.; t_,I:. ' ,

~:",,~.~, ;~~.1!' d; " r' :)·~ .: ~faJ._'~i;';"\,;,, '.:,'_. ~,,!

, The equiv8.IeD.ce p=gf88asobiate<iWith the function b(x). y+x. b(y);

for if x and y'are zero, then b(x) • y + x . b(y). = 0 and if neither x nor y is zero then b(x) = bey) = 0 and so h(x)',y + III • h(y) = O. Furthermore, if x is zero and y is not, then h(x) .'11 + x,, bey) = y and similarly if y is zero and x is not then b(x) • y + x , bey) = x. The function associated with a formula is one of many possible functions, e.g. with p we might associate a(x) since this function is zero if, and only if. x is zero. Although the functions x and a(x) are unequal, the equations x = 0, a(x) = 0 are mutually derivable; for a(O) = 0, and if a(x) = 0, then from the provable equationx(l":" a(x» = 0 we derive Ill":'" O.

,

,

I:·,

o .~

.. :; .'

32

CONSTRUCTIVE FORMALISM

Though it is clear that Ax (P (x» is adequately expressed by the free variable equntion p(x) = 0, it is perhaps less obvious that the operator Ax is superfluous in such contexts as "Ax (P (x) vQ(x»U and

.. {Ax (P (X))} v { Ax (Q (x))}", since the omission of the operator sign renders these distinct propositions identical. TIllS confusion is. removed, simply, by the introduction of further variables; thus

.Ax {p (x) v Q (X)} is expressed by p(x)q(x) = 0, but

{ Ax (P (x))} v {Ax (Q (x))} by p(x)q(y) = O.

It can be shown=, by means of the rule that function signs are equivalent if they satisfy the same introductory equations, that any classical inductive proof may be translated into proof equations in the calculus, so that the equalising rule renders superJluoU8 the p08tulati01~ oJ the principle of mathematical induction.

The equation calculus we have described presents mathematics as a system of transforms in which neither axioms nor symbolic logic play any part. The system is purely formal, and at the same time finitist in the strictest sense of tho term. By the elimination of the propositional calculus from mathematics, all disputes concerning the validity of the axioms of the propositional calculus, like the axiom of excluded middle, are rendered totally irrelevant, Both formal and non-formal characterisations of the notion of finite and constructive proof are crystallised into the single criterion of demonstrability in the equation calculus.

• Vide "Funct,jon thenry in an axiom-free equation calculus", R. L. Goodstein, ProceediDl!s uf the London Math. Society, Ser. 2, Yol. 48, 1945, pp.407-8.

e

I

,

. '

:·:1

,
I
I;
1.
,t
~
,~
1 ~
J
I" 33

CHAPTER Ill.

i ,I ;

i

" ·t

,I

j

il

';

RELATIVE CONVERGENCE .

Pile fractional calculus. One of the many calculi with pairs of numbers. is the calculus of fractio7UJ. A fraction is 0. pair of natural numbers p/q, q > 0, subject to the following arithmetic (expressed in

abbreviated notation) :-.' .

(A~ pig > "/8

<:

if, and only if;'" > '

, 1'8 <gr.

Sum (p/g',1'/8)=p/g + 1'/8 = (p8 + qr)!q8. p/g ..:.. 1'/8 = (ps.!.. qr)!q8. pIg.r/8 = pr/gs.

. (P/g)/(1'/8)=p/g + 1'/8 == 1'8/qr.

The foregoing introductory equations of the sum, difference. product and quotient illustrate the general method of introduction of fracti~ Junction sig'IUI. A fraction function F (P/q, 1'/8) transforms into a pair of functions N (p, q, 1', 8)/D (p, '1, 1', 8) such ihat for any It > 0, A > 0,

D (lcp, Itq. A1', As) > 0., '( . , " ,

. ~'!' 'f, and .

N «(P, Icq, Ar; M)/D (ltp. Kq, M, M) = N (P. q, 1', sliD (p, q, r, 8)' for all p, 1', q > 0, 8 :> 0, with an analogous formulation for a function

.of more than two£ra.ctio~ va.riablea. "

, By means of definitiotf(A) forinullil of the fractional calculus transfon;n into formulm of the eqaation calculus, fractional function signs being eliminated by the introductory equations of the sign. Conversely, any formula a = b, of the equation calculus may be transformed into the equation a/I = bll in the' fraetiooal calculus.,.

',A fractional function containing a numer,ical variable 11. (whioh may be taken ,to be therationaln/l) is:ealled a sequence . (in .).

'?-'he fun:otio~ concepts "th~~qu~i~nt'~f p divided by s". ~'the q~otient of p diVIded by '1,. to 11. decimal plaoes" and "the absolute ·difference between p/q.andr/B" are expressed by the functions

, Qt (P. q). Dec (pfq. n)'and Pd (p/q,r/8)

introduced~ respectively, by the equations '. '.

Qt (c, 0) = 0, Qt (a(b + e(+ e, '(b+ e)') = a t Dec (P/g. 11.) = Qt (lO"p, ~)/IO" t

and Pd (PIg. 1'/8) = (P/g":"'r/8) + (r/8'!"p/g).

'" _-------,-....,..---------

t The two pima C. 0; a(b +0)' + e, (6 + c)' fol'lIlo Q closed chaiu.

",', ,t ,~. ~ 3. ~ .... 10 ere abbzeviaUolI8 for 0', 1', 2', ••• 9' respectively.

CONSTRUCTIVE FOlWALISM

We abbreviate the equation

Dec (Pd (Plq, r/~), n) '= ()

to p/q - r» = O(n).

By means of these functions the following non-formal account of the function theory of the fractional calculus may readily be translafAld into proof equatrons in the system. The variable function signs we employ must be replaced b.v definite function signs in the formal translation sineo the equation calculus contains only definite function signs. For fractional variables, in the informal account, we write x, g, %, etc.

Convergence. A sequence f (n, x) is said to be cont'eryenl if we can introduce R least function Nll (/;, x) such that

". ~ II ~ N~ (/:, x) - f (II·. x) - f (n, x) = 0(/:).

As an example of It formulisation of the convergence property we observe that the convergence (If the sequence (1/11). (plq) is expressed by the provable equatioll

(m-:-n) { n_:_Prod[l<;xp(lO, k). Qt(P. q)] }Dec{ Pd{(p/q)/m, (p/q)/1t).1: }=o.

Relalit'e conver!lellc~. A function f (m. n) is said to be convergent in tn, relative to n,t if we can introduce a least Nl (1:). with Nl (I: + 1) > N, (1;). and a least v (m, m*. k) such that

{m* ~ m ~ Nl(k)} & {n G v (m, 1/1*, k)} _/(m·, n) -/(m, n)=O(I:).

The reduced sequence. IT / (m, n) is convergent in n, so that n· ~ n ~ N~ (k. m) - f (m, n·) - / (m, n) = O(k), and convergent in m relative to n, so that

{m. ~ m ~ Nl(k)} & {n ~ v (m.m •• k)} _/(m·, n) -/(m,n)=O(k), then the sequence

R, (n) = f (Nl (n). N2 (n, Nl (n))) is called the reduced sequence of f (m. n).

t Convergence in m relative to n does not involvo the convergence oC J (m, tI) Cor IIny fixed ,,"lUll of ,,; Cor i1l>ltanre, tho C"nr-tion m In ill convergent in m. relative to n, though III /11 i. not convergent. Cor any p:h'en wlue oC n.

I

t

I

u,.,. = Max {Nil (n. Nl (n», Nil (n·, Nl (".)). v (Xl (11), x, (u*), fl)}! where n· > n, then

R, (n) -/ (Nl (n), .,.,.) = 0 (n). R, (n·) -/ (N 1 (n·), 1') = 0 (1&)

and / (Nl (n), ,.,.) -/ (Nl (n-), ,.,.) = 0 (n)

SO that R, (P) - R, (n') = 0 (n), for 'P ~ n', which proves that R, (1a) is convergent.

It fono~ also tlud

/ (p, q) - R, (n) =' 0 (n - 1)

for p ~ Nl (n) and q ~ Max (Nil (n. Nl (n)), v (p, :il (n). 11)).

The convergence of the reduced sequence is equivalent to Dedekind's theorem. Dedekind proved that a convergent sequence of real numbers converges to a real number limit; we translate the concept of a convergent sequence of real numbers (completed infinite process) by a . function / (m. n). convergent in n, and convergent in m relative to n, and derive from f (tn, n) a convergent sequence R, (n) which (roughly speaking) approaches / (m, n) as closely as we please.

We may express this by sa.ying that R, (n) trn.nslatos the traditional

lim (lim f (m, !l»' '

m_oo n_oo

Continuity. A (fractional) function / (x) is continuous for a ~ x ~ b

if we can introduce e(Ie) so that ' '

(a ~ ~ ~ z* ~ b) & (~. - ~ = 0 (I:(k))) ~ I (x·) -/ (~) = 0 (Ie).

Relalive continuilr/. A function I (n, z) is continuous for a ~ % ~ b, relative to n, if we can introduce elk) and 0(1:, ~. z*) such that

(a ~ ~ ~ z* ~ b) & (z* - z =' 0 (e(le))) & (n ~ O{k, x, x*»

. )-1 (., z*) ~ I (n. ~) = 0 (1:).

~ . ,. '" . .

·1r:.a,ionIila. Ifl (;'"=,) is'~~nt in' n,' and continuous in ~ relative to n, and if g(m,' z) is convergent in m, then I (n, g(m, x)) is convergent in n, and convergent in m relative to n.

For . I (n, :1:*) -I'(n; z)'O (i),

provided z* -'z == 0 (e(J,)~' n ~ q~1e, x, z*),

and g(m·, z)- g(".. z) = 0 (k)

for . m·:~ m ~ N(1e,z) '.

and 80, for m* ~ m ~ N{e(k), z) &nd n ~ O(k, g(m, ~), g(m., x)),

/ (n, g(m·, x» -/ (n, g(m,%» = 0 (Ie) (1).

t Max (II, b. e) is introduaod by \he oquat.ioDi '

Maz (0,' a) = 0, Max (a + 6', a) c::I a + b', ~ (a, a + b') = G + 6' AD4 Max (a. bte) ~ Max (lllax (a. 6). e).

36

CONSTRUCTIVE FORMALISM

Furthermore, from the convergence of J (n, x)

we hun" for II· ~ n ;:;; N:!.(!·, g(m, x»,

J (n·, gem. xl) - J (11, gem, x)) = 0 (k) .•••.•••..••.•.•.• (2).

Formula (2) shows that J (n, gem, x)) is convergent in n, and formula (1) that! (II, g(m. r)) is convergent in m, relative to n.

The reduced sequence of ! (n, g(m, x)) translates the traditional lim ! (11, lim g(m, x)).

,,_00 m-~oo

Tho interpretation we have given of lim J (n, lim gem, x)) shows

11_00 m_oo

that in the equation eah'ulu1l the values of functions for 'irrational' values of the arguments are constructed, lint defined. THE NOTION OF A FPSCTIOS OF A!'I IItRATJONAL VARIABI.E PLAYS so PART AT ALL. Such 11 function as •. e" is 1\ funetion of t 11'0 varinbles, the rational variable .r and an unexpressed numeral variable, and is introduced by the equation

c(n, p/q) = E(n, p, q)/Prod (Fac (n), Exp (q, n)) where K(n, ]I. q) is introduced by the equations

E(O, p. q) = 0', K(1~', p, q) = Sum (prod (Prod (q, n'), E(n, p, q)), Exp (p, n')) and Fac (n) by the equations FILe (0) = I, Fnc (n') = Prod (1t', }l'o.e (n)).

So too He" is not a number but the sequence e(n, 111). The function

,

"e;" is the reduced sequence of the function e(n, e(m, x)) and the 'value

of e' for an irrational v alue of the argument, like e, is the reduced sequence of the sequence e(n, e(m, 1)).

The integer calculus. An integer is a pair of natural numbers [p, q] such that

according as

, > ,

p+q =p +q. <

It follows that [p + T, q + r] = [p, q] and in particular that [p, q] = [p.!.q, 0] if p ~ q

= [0, q.!.p] if p < g.

[p, q] is called a positive or negative integer according as p > q or p < q. Wo observe that [p, q] is positive if and only if [p, q] > [0,.0] and negative if and only if Lp, q] < [0, 0].

lit )

'j I

i!

I

RBLATIVE CONVERGENOE

37

If we can introduce functions l(p, q, r, 8), J(p, q, r, 8) such that .1(p + k, g + k, r + A, 8 + A) + J(p, q, r, 8)

= l(p, g, r, 8) + J(p + Ie, g + k, T + A, 8 + A)

for all p, q, r, 8, k, A, then we may introduce an integer function J([P, q], [r, 8]) with arguments [p, q], [r, 8]) by the equation

J ([p, q], [r, 8]) = [1(P, 'I, T, 8), J(p, g, r, 8)],

with an analogous formulation for functions of other than two arguments. For instance

Sum ([P, q], [T, 8]) => [p +", g + 8] Diff ([P, q], [T, 8]) = [p + 8, q + r] Prod ([P, q], (T, 8)) = [pr + g8, p8 + qr)

Mod ([P, q]) = (p.!. q) + (q":" pl.

In virtue of the foregoing definitions [0, 0] may be denoted by a simple

zero without ambiguity.' . .

i t

1

t I

I "

l '

1

;

f

,

The rational calculus. Denoting integers by single letters, i, j, with or without affixes, a rational number is a pair of integers i/j, j -:: 0,

. "-.

snch that iIi = i'li' .

if, and only if, if' = "i. '

, >

We observe that [p, q]f[r,. 8] = [q, p]f[8, r). r,< 8, 80 that a rational

number may a.lways be expressed in the fo~ iii With i a. positit.e integer.

'('-. " ." , '

With j, i' , both positiVe" "; :. ' . t:

" 'ifi .~·i~/i

=s. .• <.,

accordiog 88

A rational function I (l/j, ,i'lj') ,is .. Pair 'of integer functions

such that

P(i;"], i~~ j'):/ Q(i,.i, .. ~,i').'

.... .:~..( -. . .

Q(i, s.« j')'~ 0, ,and fork ~ 0, ". ~ 0,

'P(!:i, k./, ee, IIi') Q(i, i. r, 1') = P(i,'i, e. j') Q(ki, I:j, e«, Ie'n for all integers i, i,' ", i' with i ~ 0, r ~ O.

For instance Sum (iii, i' If) :e iii + i' Ii' = {u' + i'j);jj', ete., Mod (a/j) =- llod i I llod j.

38

CONSTRUCTIVE FORM4LISlI

In the following informal account of funotion theory in the rational calculus we denote rational variables by x, y, %, eto., and rational functions by / (x, y, e), etc. As before we shall write the formula x - y = 0 (Ie)· for the equation Dec (Mod (Diff (x, y», Ie) = O. The definitions of convergence and continuity take the same form as in the fractional calculus. To illustrate the character of the rational calculus we shall describe in detail the fundamental theorems on continuity and relative continuity.

The classical theorem that a continuous function which takes both positive and negative values attains also the value zero, is true in the rational calculus for a continuous rational function, in the sense that if /(x) is continuous for a ~ x ~ band /(a) . /(b) < 0 then we can introduce a convergent sequence 8(n) and a numeral N(Ie) such that

/ (8(n» = 0 (Ie) for n ~ N(k).

IT for Borne function ~(n) we can introduce a ifJ(le) such that ~(n) = 0 (Ie) for n ~ ifJ (Ie),

we shall any that ~(n) = 0 relative to n. Thus we have to prove that / (8(n» = 0 relative to n.

Without loss of generality we may suppose / (a) < 0 and / (6) > O.

Since / (x) is continuous we can introduce a steadily increasing numeral function i (Ie) such that the numbers

a: = a + (6 - a)r/2ilA,

antiafy / (x) -: / (a:) = 0 (Ie), for 0 ~ r ~ 2ilA" a: ~ x ~ a:+1 ;

since /(a) < 0 and /(b) > 0 we'can introduce a least r, rs say, suoh that/(a~ +1) > 0 and/(a~) ~ 0, whence/ex) = O(Ie) fora: ~ x ~ a: +1'

A A ,II II

Furthermore, writing aA = aA+1 and a" = all+1 we can introduce a

'A',. ',,+1 'II

least r, '''+1 any, between 8" and tA suoh that / (al+1 +1) > 0 and '''+1

I (al+1 ) s 0 and therefore / (al+1 ) = 0 (Ie + 1).

'1+1 '1+1

The introduction of the function rk is recursive (for this informal account is readily translated into formal recursive introductory equations). Since i(1e + I) > i(Ie), therefore

ak+1 - aA+1 ::;;; (all - all) / 2

,+1 , - ,+1 ,

which proves that the sequence a" converges. '"

.~

I'

'i

I

( I

I

I

I

RF.T.ATIY'E CO!<lVERO ENCB

39

The /und4mental theorem for relative continuity. The behaviour of a relatively continuous function is much more complioated, of course, than the behavious of a rational continuous function, but there is an important anologue of the foregoing theorem, for relative continuity.

IT / (n, x) is convergent in n and continuous in (a, 6), relative to n, and if there are integers p., ", such that

It». a) < -1O-P., /(n, 6) > 10-1' for n ~ II

then we can introduce sequences 8(n), 8.(n) such that, for 8(1e) ~ x ~ 8· (Ie),

/(n,x) = O(~ - 1), for n ~ N(1e + 1, x).

Proo/. Since /(n, x) is convergent we can introduce N(k, z}, with N(1e + 1, x) > N(Ie, z], such that

/ (n, x) - / (N(Ie, x), x) = 0 (Ie), for n ~ N(Ie, x).

Let g(k, x) =1 (N(1e + 1, x), x) so that if

g(Ie, x) ~ 10-11 then g(1e + I, x) > 10-"-1 and similarly

if g(k, x) ~ -1()-A then g(1e + I, x) < _IQ-o'-I.

Since 1 (n, x) is continuous relative to n, we 'can introduce elk) and O(k, e, tI) 80 that if al(r) - ,a + (6 - a)r/y (Ie), where y (Ie) = I()'IA),

then / (n, x) -I (n, al(;'» = 0 (Ie + 1)

for ak(,) ~ x ~ at(r + 1), . n .~ O(1e + 1, al(,), x), 0 ~ r ~ r (Ie) - I.

Hence, for , \' v-.

n ~ Max { O(Ie+I):. GA(r), G~(i+m,.' N(k~I, all(r», N(Ie+l, a"("+I»} , each of.l (n, a"(r» -I (n, ~(r + 1», 1 (n, G1(r»...,.. g(/e, al(r»

g(le, al(8» - Dec (g(k, cill(s». Ie + I),' s = r, r + I, and /(n, GII(r + 1»:"" g(Ie, al(~ + 1»

is 0 (Ie + 1) and therefore

.- • <I

Dec (g(le, a"(r + 1», Ie + 1) - Dec (g (Ie, ~(r», Ie + 1) = 5.0 (Ie + 1),

whence Mod{ Dee (g(k, aA(r + 1»,' Ie) .; Dee (g(le, aA(~», Ie)} s l/IQk; therefore, if k(le, 7) = Dec (g(Al, :r:), Ie), the integors lOAh(Ie, al(8» are equal or conseoutive for COnsecutive values ofs. 'l'hus h(k, x) take8 every value n/IQk between any two 0/ ita tNJlues, whick i8 the e88ential properly that I (n, x) has in common with a rational C01IIi1&'U0U8 fufldiO'lt.

40

COSSTRUCTIVS FOR!lA~

Since /(11, a) < -10-1' and /(11, b) > 10-1' for n ~ v, therefore

h(k, a) < -10-", II(k, b) > 10"" for k > 1" For I; ~ I' + 1 we define

'A, SA, with ,,, < 811, recursively as follows;-

rl'+l = 0, 81'+1 = Y (p. + 1) 80 that

h(p. + I, aJ'+I(rl'+l» < _10-1'-1, h(p. + I, aJ'+1(81'+l» > 10-1'-1.

If for some /,; > I'

h(k, a"(r,,)) < -10-", h(k, a"(s.» > 10-' ••..••.. " ..•.••.••..• (1).

we can introduce a least r, I" say, such that

l«> T" and h(k, ak(lA)) ;?; lO-k and It(k, a"(s» < 10"",

for 'A ~ 8 ;5 1" - I, and therefore h(k, a"(s» ~ 0; similarly we can introduce a greatollt r, Pit, between 1'k and h, such that

/,(k, a"(pAn ~ _10-11 (i.e. a least C?A such that h(k, aA(I" ...:. aA» ~ -10--').

Since h(k, ah(I,,» ~ 10-", li(k, aA(I"...!... 1» s 0 and

h(k, a"(IA» - h(k, a"(/A...!... 1» ~ 10-A

therefore h(k, ak(l,,» = 10-" anti h(k, a"(lA...:. 1» = 0; similarly h(k, a"(p,,» = -10-". From the pair of equations

h(k, a"(lA» = 10"", h(k, a"(pll}) = -10 ....

it follows that

h(/': + 1, at(lk» > 10 .... -1, h(k + 1, a"(p.» < _10""'-1.

But a"(T) = aH1(ry(k + 1)/y(k» and 80 if we take

'11+1 = pAy(k + 1)/y(k), 8"+1 = l"y(k + 1)/y(k) then

h(1.: + I, ah+1(r" + 1) < _10--'-1, h(k + 1, a"+I(s"+1» > 10--'-1

which prnves that the inequalities (1) hold for all I; ~ 1'+ 1 and therefore

I h(k, x) I ~ 10-" for all x satisfying al(l" + 1) ~ x s ak(lk _:_ I}. Thus if .9 (k) = a" (iA + I) and s* (k) = aA(/A _:_ 1) then I h (I.:, x) I ~ 10-" for s(k} ~ x ~ ,~* (I~) and therefore for such values of x, / (n, x) = 0 (k - I)

for n ~ X(I; + 1. x).

By definition the sequence .9(k) is Rtea~Ii1y increa~ing and bounded above (by any s*(1.:» and s*(1.:) lS steadily decreasing and bounded below (hy any .s(I·».

s

I

.

RELATIVE CO~"VEROESCR

41

In classical analysis a steadily increasing (or decreasing) bounded sequence is convergent, but this is not true in the equation calculus. To determine a convergent sequence ten} such that if R,(n) is the reduced sequence of I (n, t(m» then

R, (n) = 0, relative to n,

we require more information about the function I (n, x) than is given by continuity relative to n. Either. of the following (additional) eonditions is sutBcient

(a) If we can introduce a constant q and a numeral function U' (x, y) such that

Pd (f(n, x), I(n, y» ~ q Pd (x, y), for n ~ w(x, y), a ~ x ~ y ~ b

then from hCk,8Ck» = h(k, s·(k})"= 0

we can deduce Pd C8(k) , 8*(k» < ljq·l()A

which suffices to prove that s(k) and 8*(k) are both convergent : whence, since h(k,8(k» = 0, the reduced function of / (n, sen»~ is zero relative to n.

(b) If we can introduce numeral functions g (x), hex} such that Mod (f(n, x» ~ I/I()S(·'}, for n ~ h(x), then it is determinate whether for any rational x, I (n, x) < 0 or I (n, x) ~ 0, relative to n, and we can determine the convergent sequence ten) exactly as in the case of a rational continuous function.

Dillerenliability. A function I (n, 21), convergent in n, is said to be differentiable relative I to n, for a ;::i x ~ b, if we can introduce functions 11 (n, ~), d(k) and D(k, x, y) such that

(a~x~b) & (a~y~b) & (x ~ y) & {,-x. 0 (d(k» }& (n E oo, x, y)

-- {/(n; y) -/(,.,. x) }jCy':'" x) ~Jl (n, x) = 0 (k).

The function J 1 (n, x), or atty iunction -equal to'it; relative to n, is called the derivative of J (n, x), x:e1ative to ft. . Both J (n, x) and 11 (n, x) are

continuouain e, relative'to .n. .. .~. "'.'

The fundamental theorem of the dift'erential calculus, the meanvalue theorem, is provable in the fQI'ID:- .

Ifl (n, x) is dift'erentiablefor a ~'x ~ ii, relative ton, with derivative f1 (n, x) then we can introduce c., a < CA < b, and M(k), sueh that

{/(n, b) -/(n, a) }/(b - a) =/1 (n, cAl + 0 (k)

for n ~ M(k). .

We ~ot ensure the convergence of Cj without introducing additional

restrictions on I (n, x). .. . : . . .' . .

l

CO!\:!TRUCTIVB FOlUlALISli

CHAPTER IV.

THE CONSISTENCY PROBLEM.

1. The reduction of mathematics to a formal calculus in which the prnbkmIJ of mathematics, solved nnd unsolved, could be expressed by a finite number of combinatlons of predetermined signs, led mathe· maticians to believe that a process could be determined by means of which the most general possible combination of signs could be trans. formed inti> some simple equation like 0 = 0 or 0 = 1 according as .the initial complex of signs represented a true or false proposition, The determinatlon of such u process would effect the solution, not only of individual problems, but IIf every possible problem and the 'search' for such n process became the central problem in mathematical philosophy, Amongst, the possible combinations of signs is that which expresses the absence of contradictions in the calculus, so that the discovery of a general process of solution of nil problems would also determine the question whether or not the formal calculus itself was free from contradiction, This aim made mathematics the object instead of the field of mathematical activity and led to the conception of meial1wthe. matica, a language in which is expressed sentences about mathematics, or rather about a. particular formal calculus. the predicate calculus. the elements of which are not only the mathematical signs Droper. numerals. function signs. variable signs. etc., but, also the signs )'&", "V", " ....... u_", "E", "A", more or less synonymous with the terms "and". "or". "not", "implies", "there is", "for all", respectively. Although the predicate calculus itself admits many disputed coueepts and processes it was hoped their validity might be established by a metamathematical proof of freedom from contradiction which utilised none of the doubtful processes. The foremost conditions which this proof of freedom from contradiction is required to satisfy are that the proof be finitist. and constructivist, that is to say the proof must admit no operations which cannot be carried through in a finite number of steps (thus excluding for instance testing the truth or falsehood of a propositional function for all values of a numeral variable argument) and no number or function may be introduced unless a construction is given for its determination. ~[etama.thematics, however, retains much of the machinery of the traditional calculus, and in particular treats linkage terms as belonging to the system; this error involves mathematical philosophy in an infinite hierarchy of mathematical systems, each forming the meta. mathematic of the system preceding it, like the earth supported on an elephant. the elephant on a tortoise, and 80 on without end.

I

I

THB CONSISTBNCY PROBLEM

43

~ .

, I I

i

1

,

j I ,

The fundamental distinction between mathematics and science is this: the truths of science are hypotheses, of mathematics, conventions. A f~nda~enta1. error in mathematical philosophy is to confuse mathematics with SCience and to suppose that a hypothesis, expressed bv means, of. mathematical terms, appertains to mathematics. In 'a formahsa.~lOn o! classical analysis conditional propositions and pseudohypothesIS are Introduce,d, particularly in the application of the method of proof known as reduai» ad absurdum, the hypotheses being formulated as additional axioms and rejected when the system thus formed is shown to contain a co~t~adicti~n, Consider, for instance, the classical proof that a mon.otoruc mcre~lI~g bou~ded seque~ce is eonvergent : Suppose that a(n) 18 monotonic mereasmg and divergent, that is to say. a(n + 1) ~ a(n) and for some function g(x), and some constant r a(71 + g(n)) - aeon) > 1/10' so that a(g(O)) > a(O) + 1/10'. a(g(O)-~ g(g(O))) > a(O) + 2/Up and, generally, if h(O) = g(O), hen + I) = hen) + g(h(n)), then a(h(n» > a(O) + (n + 1)/10'. Since nj 10' can be made greater than any A by choosing a sufficiently great n for such a value of n, a(h(n)) > A and so a(n) is not bounded by A. 'Thus far th? proof might be s~id to have introduced no hypothesis-ancl in fact this proof could be directly translated into the formal calculus we have constructed-for the formulation qua hypothesis of the conditions satisfied by a(n) is ?nlY,an expression?f what kind the variable sign a(n) is (since we c~n give mstances of divergence proofs of monotonic increasing func~l~ns, fro~ which the variable is derived). But if we add to tho conditlo,ns! which a(n) satisfies, that a(n) is less than A, we obtain the contradiction that from some values of n onwards a(n) is both greater than.A an~ less t~an A. from which it is concluded. in clasaical analysis that If a(n) 18 both lUcrea.smg and bounded it is not divergent· in classical analysis "not c;livergent" is equivalent to "convergent" 88' an instance of ,the axiom A(x)P(x) = Ez P(x).

t

f

I

I

j

;

, Oonst~ve mathe~iC8 has no use f~r the concept 'not convergent' m the sepat' t~at a sequence a(n) is said to' be not convergent if the hypothCSlB of ~ts ~onvergence leads to a contradiction, The negative of converg~ ?B dlvergenu and proofs of divergence may be as direct ~d c~nstructive ns ~ofs of convergence: ,The contention that reductl~ ad ,absurdum 18,' the proper method of proof of negative propositions ,18 totally false. To ~h«?w that ~he equation a(n + N) - a(~ < 1/ UP leads to a contradictIOn does not suffice to prove an equation of. t~e form a(n + g(n)) - a(n) ~ 1/10' for the derivation of the c?nt~ctlOn may not formulate the introductory equations of the function mgn "g(n)".Nor does tho conventional identity A:,; -'(P(x)) = EJ (x) (P (I (x))) introduce a/unction sign, for the 1 (x) on the right hand side of the equation is not transformable to a definite numeral

for any value of the argument x, "

CONSTRUCTIVE FOBMALIS!l1

That to which we object is not a formal rule but an inference from a contradiction. Suppose the function Stp (x) is introduced by the

conditions

Stp (n) = 0 if for each r ~ n, 2r is a sum of two primes,

Stp (n) = 1 if for some r ~ n, 2r is not a sum of two primes.

For each value of n, Stp (n) is transformable to a definite numeral and Stp (n) is monotonic increasing and bounded above by unity. Yet we oannot say thlLt Stp (n) is convergent for wo cannot specify It value of n for which tho differenoe Stp (n + X) - Stp (n) is less than unity. Classical mathematics affirms the convergence of Stp (n) on th~ grounds that if there is a number k 8uch that 2k is not a 8um of two pnmes then Stp (1&) is convergent since Stp (k + N) = I, and if 2r i8 a SU1n of two primesfor every r then Stp (n) is converg~nt since Stp (1&) = 0 for every'!'. Since tho only way open to us of finding out whether or not there IS a value 2k ,,:hich 'is not a sum of two primes is to test the numbers 2, 4, 6, 8, ... in succossion, tho 'hypothosis' if there is a number k such that 2k is not a SIL1n of two primes, says "If WH have completed the endless task of searchinu through all the O\'CII numbers and hn \'0 found one which is not a sum of two primes", III) that tho 'hyputhesis' ill self-contradictory and tho Inference to the convergence of Stp (u) is an inference from

a contradiotion.

The argument is in no way weakened by the assumption that the unsolved problem "Every oven number ill a sum of t\\:o primes" might one day bo solved, fill' it would be this (unknown) solution which proved the convergence of Stp (71,) and the foregoing proof would stiU be an invalid inference from a contradiction. In any case an inference from a possible futuro. 'solution' of a problem ill itself an inference from Ii contradiction, if we suppose tho solution to be one of the infinitely many pOl!Siblo derivutiolls in a given calculus. It might well be, however, that the 'solution' of the problem will consist in the construction of 0. new calculus just as Kummer's (partial) solution of Fermal'8Uut theorem eonsisted in the construction of a new calculus in which an analogue of the Fermat theorem ill proved. (Remember that Fermat's 'hypot~LCsis', ,x"+2 + y"H _ z .. -t-t I ~ I is. expre~d by mea~ls of numeral vaf1abl~. Kummer proves equatlons 1Il which the variables are complex, his contribution to the solution of Fermat's 'hypothesis' being essentially the invention of a new language in which a new problem is solved, Ii problem of which Fermat had no conception! (It has been well said that in mathematics it is the solution which creates the problem.)

If we no longer presuppose the interpretation. of the signs, ~he classical proof of convergence of It bounded m.onoto~lC .so.quence, ca.rI?-ed out within t1111 framework of u forlllllisystem like Principia Mathemataca, is of COUfIIC " valid proof inside the system. From this standpoint, the conclusion til be drawn from tho foregoing example is not that the

i

i

I

~

! 1 j j

it

• I

THB OONSISTENCY PROBLEM

formal system is self contradictory but that in such n formal svstem ~he ~Sl' of the (so-~all~d) univ~rsal and existential operators do~'S /lot correspoud to our intuitdvo notions of universallty and existcne«,

Wo arc completely free, as the formalist rightly affirms to set 1I a.;(tb"ls~'Stclll how w~ please, and the meaning of the signs of'tho SV8tCI~ \H I e !:'lven by t~e rules of transformation, nor by U1l\' ('xtl'ILueous exp anatlons . s~perImpused on the s.)'Stem. A constrm·ti"ist cunnot deny the validity of a formal svstem bv denyin g tho v lidltv f

. d I . .• a I I .. II Nome

preooncerve interpretation of its .. iama What fL'e hav b· d

to shou: here is that the cia' . -0'-'. e een COllCerllC

.1 • '--'it _.8 • ss.cal mterprelaho1t of the operators in terms

oJ. umvutstU y ana UtBtence, i8 invalid. .

2: .Just. as the Russell paradox first focussed attention 011 the contradietions III the .Frega-Russell Theory of Clas8es, 80 GOdel'lI great work on the construction of non-demonstrable propositions led to th b d ment ?f the Formalist programme, tho attempt to pro\'e 0 ~/?fi~'~ masns, dtho fr.eedom-fr?m-contradiction of the convention'al caicullus an to etermme a universal method or solution of all mathematical p~obl~lDB. ~ot. only have the Godel theorem and the RUlISolI paradox

!ustonoally SImilar roles but their logical positions are th TI .

unpo t f G"d l' h ."- e same. 10

th f:r ~ntche °t' °diree s t eorem, .like that of the Russell paradox, lies in

e !l.c. . a It eted attention to the unsoundness of a s 'stem bv

undermining tho foundations of the system from the inside. l ,)

The formal s¥stem \~hieh GOdel considers (we shall refer to it as language1,) eontalDB the SIgns, 0, N, =, - V & ~. = 11 Z e C ::::l and in addition an unlimited number of'si' • ti ,-, 'I"· ' propositional variables and definite functions ~ a :~i::~a blarI~~les, ar~ rules for s.ubstituting assigned numerals for numerai :~ria iere aXlom~ eonce~g the use of the signs in the above list d bli' according to which a proposition is said to be an immediat ,an ru es of one or more other pro "ti ( .. .. " • "e consequence of "p" and "p'~ .... ) . r °flJ!l e_ng_. !l, U! an unmediat6 consequence

h f y. proo 1ll·.11 IS a finite sequence of propositions

ere h 0 whioh. is an axio~ .or an immediate consequence of one or more o . ~ e Pbcediug proPOSItIOns j the last proposition of the sequence is 8a1 to e proved by the proof. The numerals are represented in :f by 0, N(O), N(N(O» •• " ., etc. and these are abbreviated to Z Z Z

d• " . so that N(N(, •• N(O) " •• ». with an auigtied number Ie ~f N's ~ enoted by Z ••

The language in which we describe the properties f jf (in this

ease, the English language) .we shall call language E. 0 j

The terms, propositions and proofs of f form a sequence and can therefo~ b~ correlated ~th a. system of numerals. Godel effects this

correlation III the following way:' The signs 0 N C::::l' th

order ill which they appear above are numbe~ fr~~' 1 t~ 13.m Th:

CONSTRUCTJ\TE FORMALISM

numerals 3r + 14 are assigned to propositional variables, 3r + 15 to numeral variables and 3r + 16 to definite function signs. To the sequence of integers kit k~ ••.. k« corresponds the number 2"'. 3A s, S-U . . . p!., where p .. is the nth prime; since propositions are sequences oC terms this enables us to number ofT propositions. and so too proof's, which are sequences of propositions. The correlation is readily seen to be unique.

If, for example, x is the number of a proof (this may be expressed purely as a. numerical property of the number z] and y is the number of a proposition, then the relationship which subsists between x and y, if the proof whose number is x is the proof of the proposition whose number is y, can be expressed (in E) without reference to the objects in ~ which they represent; in the present case y is the exponent of the greatest prime factor of x. We denote this relationship by "xPy". Furthermore, if x is the number of a function with one free variable (i.e. a variable not operated on by 1:, E or II) and if y is the number of a numeral'Zp', then Ji», y) denotes the number of tho expression obtained by substituting "Zp" for the free variable in the function of number x. GOdeI gives the explicit numerical relationship of J(x, y) to x and y.

Recursive functions play an important part, for if we call "0" and "N(x)" the expressions in 1. of "0" and "x'" in E, then it may be shown that any recursive function in E can be expressed by a formula in jf, such that, if F (u, v) is the expression in j1 of ! (x, y) in E then F(z" z,) = ZA is provable in 1f if / (r, 8) = k is true in E (in the proof offer, 8) = k replace "/" by "F", "x'" by "N(x)", "k" by "Z,,' and we obtain the proof of F(z" z.) = z,,), and - (F (z" z,) = ZJI) is provable in JT if tv. 8) is not equal to k.

A relation' R(x, y) is recursive if a recursive function! (x, y) is determined so that / (x, y) = 0 if R(x, y) is true and / (x, y) = 1 if B(x, y) is false; it follows that a recursive relation B(x, y) in E is expressed in f by a relation m(x, y) such that 1R(zJI, z,) is demonstrable in f if R(k, l) is true and -1R(zAI, z,) is demonstrable in JT if B{k, Z) is false.

GOdel shows that xPy and J(x, y) are recursive 80 that B(u, v) and 8(u, v) can be determined in 1 such that B(ZA' z,) and 8(ZA, Z,) are demonstrable in f whenever kPl and J(k, l) are true in E. Hence B(ZA, 8(ZI' z,» is demonstrable if kPJ(l, l) is false. Let p be the number of the proposition

II v( -B(v, 8(u, u)))

80 that by the definition of .1, J(p, p) is the number of the proposition II v( -B(v, S(zp, zp))). Denote this expression for brevity by II v(Q(v», or simply h.v Q. Q is Godel's formula. Godel's theorem asserts that if f is free from contradiction, neither Q nor - Q is demonstrable in :f ; the proof is as follows.

;

i

a I

THE OON8ISTENOY PBOBLlW

4;

• Suppose we find a proof of Q; let Ie be its number. Then k is the number of the proof of proposition number J(p, p). i.e. kPJ(p, p) is true so that B(z", 8(z" zp» is provable in jf. But it is a principle of proof in 1 that if II v(Q (v» is provable then Q(z,,) is provable. Thus if Q is provable both B(ZA, 8(zp, zp) and -B(zJI, 8(zp, zp» are provable in f ~d f contains a contradiction. If Q is not provable, so that - kPJ(p, p) 18 true for every k, then -B(ZA, 8(zp, zp)) is provable in :Jr for any k (although II v( <Bi», 8(zp, zp))) is not provable in f) and so if - Q is provable, f contains a (weak) contradiction for both - IIv(Q (v)) and Q(ZA), for any k, are provable.

If Neg Z • i~ the number of t~e proposition which is the negative of the proposition whose number 18 z, then the proposition that :Jr is n:ee from c?ntradic~ion. takes the form (~)~y)(z) .... (xpz & yP Ne~ z). Since Neg Z IS reeursrve it may be expressed m f by a recursive function v(w), and so .... (xPz & yP Neg z) is expressed in f by - (B (u, w) & B(v, v(w»); although (x)(y)(z) - (xPz & yP Neg z) is not recursive and is not ther~rore ~ynonymous with Il(u, v! w) - (B (u, w) & B(,., V(IO)), neverthelesa Godel mterprets the vague notion of "a proof in f of the ~om from contradiction of f" to mean "a proof in f of the pro. posltion II (u, v, w) - (B (u, w) & B(v, v (w)))". It can be shown that Il (u, v, .w). (B (u, w) & Be", v (w))) _ II v( , Be", 8(zp, zp))) is provable, so that if II (u, e, w).- (B (u, w). & B(v, v (w))) \~ provable in f tben II v( - B(v, S(z" zp))) 18 provable In f and so f contains a contradiction' whence GOdel concludes, if f is free from contradiction then this freedo~

from contradiction cannot be - proved in 1. '

, .... :

The proof of non-demonstrability applies, unchanged to the proposition Q(w) with tree variable w, and although the' formula 11 w(Q(w» bas .the adde ~terest that its negation is also non-demonstrable, from t~ &tan~~m~ -?f the f~n~ndati9ns of mathematics tho non-demonstrabill~y or Q(w), m appo81tlOJl to ~he demonstrability of Q(ZA) for any k,. 18 the more remarkable. This apposition bas been held. to show that the.notio~ of a finite ~1 is essentially non-formal; for .if mathematic81 indqation.. and the rule of substituting for II. free vanable, fail to secure that _Q(w) is -cIeQlo~ble when each of Q(ZA) is demonstrable, then the ~ __ ~b.le w does not ATTI"_ the concept of 'any h,ite cardiDal'-in tlie -£orin&l sysUma ;I.f_ -~-

This result may be sharpened &s-!follo~. We extend the avstem 1 by ~troducing Russell's d~ptive operator is by means of \vhieh the umque value oC x for which P(x) holds is represented as iJl(x)' formally P(i.P(~» is derivable from the twin conditions '

Xx(P(~» • (U x)(U y)(P(x) & P(y) _ x = y).

COSSTRtTCTIVE FORMALISM

Next we define the minimum. operator p.~ by the equation p."P(:r) = i~(Q(x) & R(x))

where Q(.l~) and R(x) are abbreviations for

lIz{ .-P(Z)} _ (x = 0)

and 1::(P(z» _ {P(x) & II u(P(u) _ (x = tt) V % < u)}

respectively; the formulie which express the fact that p."P(%) is the least x for which P(x) holds, namely

1:x(P(x» _ P(p.sP(%», Pea) - p.'p(x) ~ a

are provable in the extended system 1·, say.

It is rondilv shown thnt Godel's construction of non-demonstrable propositions nu;y be carried uut wit~in f·; in fact, by means of ~he operator I". given a function ,,(x, y) m f·, we may introduce a function f (x) such that the recursion formulai

f (0) = 0 f (n') = a(n, f (n»

are provable in :f. with a numeral variable n. whereas, in j, the most tha.t can be shown for a. corresponding rf,(u) is that the equations

",(0) = o. "'(z~+Il = a(z~, "'(Z/I»

hold for every z". Accordingly, there is a recursive term f (n) in j1* such that f (n) = 0 is non.demonstrable in f· but f (0) = 0, ! (1) = 0, •.. are all demonstrable, and an F(u) such that

F(O) = 0, F(u') = F(n) + (a.U (n'» ..:. F(n»

(where 0.(0) = 0, o.(n') = 0'). F(n) is demonstrably bounded a.nd non- decreasing in f·, so that (by the familiar convergence argument)

(1:y)(II x)(x ~ y _ F(%) = F(y»

is provable in r, and therefore if ~ is an abbreviation for the expression p.y II x(x ~ y _ F(x) = F(y»

11 x(x ~ ~ _ F(x) = F(~»

then

is provable in 1*·

Since F(x) is non.decreasing, we readily deduce that F(>.) = 0 _ II xU (x) = 0)

is provable in f·, and so (if :ff* is free from contradiction) the equation F(>') = 0

is not demonstrable in f*·

.'

TIlE CONSISTENCY PROBLEM

49

Since/(ZA) = 0 is demonstrable for any Zit, clearly F(zA) = 0 is also demonstrable and therefore the equation >. = z" is not demonstrabh, in j1* for any k, that is to say, contrary to the intention expressed in the definition of p.s, p..P(%) may be demtm8trably different from every fltItural number.

We have incidentally shown that F(n) is a monotone bounded reoursively defined sequence (of natural numbers) for which no C07I8. 'rudiv8 proof of convergence is possible.

3. Godel's proof that the truth of the formula Q(ZA) for allk necessarily implies that Q(w) is non-demonstrable is of the kind that we have called a derivation from a contradiction; on the assumption that there exists a k which is the number of a proof of Q(w), f is shown to CQIltain a contradiction, whence it is concluded that if ;Jr is free from contradiction then Q(w) is non- demonstrable (and all Q(z~) are true). Since the only way open to us of finding whether there is such a Ie is to scaroh for it indefinitely through the series of natural numbers, therefore the a.ssumption that such a k exists cannot be refuted (since we cannot complete the endless task oftesting every k); accordingly, the existence a.ssumption is meaningless and cannot serve as a premiss for a valid deduction.

. Some coDstruotivist writers maintain that although a reductio ad absurdum proof of existence is invalid, nevertheless, a 'reductio' proof of universality is acceptable. For, it is said, a valid proof of existence must oonstruct the entity whose existence is proved and this a 'reductio' proof does not accomplish, whereas to establish a universal property it sufIlces just to show that the hypothesis of the existence of a counter example leads to a contradiction. This distinction however appears to be mistaken. for if existence can be established only by providing a constru~on then the hypothesis that a counter example uist8 Is an hypoth68l8 that sneh an example is constructible and the opposite of this is, not ... that the universal proposition holds but tha.t thero is no conalrudible coun~ example.· Thus in the case of GMel's proof, since the suppositipn t~t a Ie exists which is the number of a proof of Q(w) introdu~ a contradiction into ;Jr. the conclusion to be drawn, if f is supposed free from contradiction, is not ~t Q(w) is non-demonstrable })ut rather that the existence of a proof of Q(w) cannot be established oo~tructjvely. Since, however, tlie hypothesis tha.t ;Sf is free from contradiction is 'i~e1f a uniV6l'f!llol proposition which is unproved-and perha.ps in some sense unprevable-s-a ' conclusion drawn from this 'hypothesis must be a conclusion drawn from a supposed term by term verification of a universal proposition, and so a conclusion from a '~mpleted infinite process'. Each of the sentence forms 'Q(w) is c;lemonstrable' and '1[ is contradictory' might be validated by a chance d,iscQvery, bu~ neither could be refuted by an enumeration of cases and, fQf w~t of. an alternative procesa.of disproof, neither can be accepted as a '.male1Ice, a suitable premiss for a valid inference. For want of a

50

CONSTBUCT1VB FOBMALISM

method of refutation these forms are 8ensele.B8. Failing. & chance discovery, one might search inde~te~y through the Benes of natural numbers without finding a k which IS tbe number of a. proof of Q(w), aad 110 without finding a. contradiction in 1, and yet be unable to 8&y

that no number is tbe number of a proof of Q(w). .

GodeI's analysis of formal systems supports this conclusion. For if 'deri ... ation from a contradiction' is a. vo.l~d proof pro~ess ~hen all cun-e'" lormalisations 01 mlllMmatiu must be rejected lor their lallure to ezpre88 the concept 01 a finite cardina.l .. We must either. o.b~don the method of derivation from u. conteadietion or the formalisation of the number

concept.

4. There is another aspect of .GroeI's work. which ~ of consid,!o.bl,e importance in mathematical philosop~y. It IS t~e. ~Iew tbat G~d~ls 'arithmetisation' of a formal system gives the posslbility of d.eterm1rung propositions which assert something of themselves, m partIcular th~t the Godel formula Q asserts in jl that Q is non. demonstrable. ThiS view is based on the following line of reasoning.

The proposition II v( _ B(v, S(:.p. :.p))) is tbe express}~n in f of the proposition (xl _ xPJ(p, p), which says t~~ the propositien of number J(p, p) is unpro\'able, and tbe pr~pOSltion of n~ber. J(p, .p) .~ II t' _ (B(~', S(zp, %:»» itself. In the' Tractatus Logleo.PhilO8Ophl.CUS , Wittgcnstein observed that sinc~ a se!1tence cannot. be a. part of Its~lf, a proposition cannot say something ~f Itself, for .tbe sign of th~t .of v.:hioh a proposition speaks is one of the SIgns by w~~h the pro~tl~n Itself is expressed; tbis is not to say tbat & propositaon oannot 8lgnificaotly affirm something of tbe physical signs b! whicb it is e~d, bu~ tbat a proposition can say nothing about. Its o~n mea.rung: Accordingly, tbe view that tM proposition Q says of Itsel~, l~ jl, that _Q IS not provable, is fallacious' we shall in fact show that It 111 as a. sign of a language different fro~ 1 that Q asserts that the proposition expressed in :If by

Q is not provable (in f)· -

Tbe one-to-one correspondence which is established in t~e language E, bet.ween propositions ubout sentences in f. an~ cer~lD arithmetical propositions. is in tbe nature of a code-a. highly. mgeD10US code. For example, tbe proposition (in E) .that a proof S (ID j?) is & proof of a proposi.ti.on T, of ;9.'. is expressed. m tbe code, 01 say, by an arithmetic propOSitIOn expre.mg the relation of the numbers 8 of S and t of T, in fact by the proposition aPt. As a sign ~f the la.ngu.age Cit sPI says that S is tbe proof of T; though 88 a SIgn of car~al arithmetic 8Pt says just tbat tis the exponent of the greatest pnme fnctor of 8. Next a code C2 is constructed, which uses tbe signs of language jl; in C2 the proposition "S is the proof of T_" is expressed by i'B(z., z,)", but as a sign of 1, "B(z., z,)" says that t 18 the exponent of the greater prime factor of 8. If 88 a sign of 1, B(z" z,) says that

., 6

I

.;. .

~-' . ',. ~'-.'~

THE CONlUSTESCY PROBI.EM

!il

S is a proof of T then "B(z., :,)" plays two roles in f, and f is ambiguous, for there is nothing in f to distinguish these roles; were this the case then every language could be rendered ambiguous simply by constructing a code. It is in the code O2 and not in tbe language f that the sign "Q" asserts that the proposition expressed in f by Q is 1I0t provable; Q does not affirm something of itself in f, but is a sign which plays

two entirely different roles in two ditl'erent languages. •

...

5. Gentzen'!I proof of the fr8edom from oontradiction of cardinal arithmetio (reine Zahlenlheorie) utilises TrGn8,finite Induction as a mini. mum deviation from the ·finite proof processes admitted in the construction of the GOde} formula. The objeot of the proof is to show that bv means of a number of reduction processes the axioms and derivation processes of the conventional ealeulus (including the axiom of mathematical . (finite) induction and the "Excluded Middle" axiom) can be transformed into the equation 0 = O. The reduction procesaes applied to a contradiction are shown to result in the. equation 0 = 1; whence Gentzen affirms tbe freedom from ciontradiction of cardinal arithmetic.

. The dUliculty in transf'onning tlie' axioms to the equation 0 = 0 is in showing tbat the transf'OI'JIU\tion is completed after a finite number of applications of the reduction processes. To this end transfinite indudion (over ordinals less than cOl) is introduced to prove tbat a decreasing sequence of ordinals neceasarUy terminates.

.~ .

. Transfinite induction is of the form: If P(O) is true and if tbe truth of P(a) for any transfinite ordinal a, less tban b, implies the truth of P(b), then P(b) is true for' all b. As such, transfinite induction apparently contains the crontradictory notion of a completed endless process, since to establish for instance P(w'a) we are requited to establish the tinlimited number of propositiOns P(G) for G < WA.

. . . .....

.. We bave seen that (&nite) induction ja' Ii~ 'axfum but a mere affirmation of the sufficiency of the recunive introduotory equatioll8. In an ana.logoll8 way a transfinite induction croWd. in certain trivial cases ~ said to a.ftirm. the sufJiciency of multiply recursive introductory equations such as /(~, 0)'= a(x),/(O, y) = b(g)

. . l(x + 1" y+ 1) = ~ {x, y, I (x, :crIl(x + I, y»}.

These equa~ons determine / (x + 1, y + 1) when / (x + I, y) is known and / (x, r) IS known for any assigned r. If we crorrelate the pair r 8 ~tb the. ordinal ~w + 8, then we can say that / (z, y) is dete~ed if I (r, 8) IS determined for every ordinal rw + 81('SS than :uu + y. Such an application of the principle of transfinite induction in classical ma~hematics might t>t: ~ in~ a proof. t~at two function signs satisfy the same multiply recursave mtroduetory equations. Gentzen's use of transfinite induction is not of this kind.

CONSTRUCTIVE FORMALISM

CHAPTER V.

TR~~SFINITE NUMBERS.

1. The formal calculi we described in the second and third chapters contain but a single 'infinity', the unlimited possibility of replacing a numeral variable by a definite numeral or by a definite function sign. Tho rejection of the metaphysical concept of a "completed infinite process" involves the rejection of all tho traditional 'infinities' which are not the infinity expressed by a sequence; that this is the case has not however been fully realised even by conBtructivist writers on the foundations of mathematics,. who 'admit', for instance, the famous Cantor ternaryl set. If we divide the interval (0, 1) into three equal parts, reject the middle part, divide the remaining intervals into three equal parts, again rejecting the middle parts, and so on, we are said to determine the ternary set, What in fact we determine is a sequence of intervals; to say that a set of points is obtained if we continue the process ad infinitum says nothing at all. We can, of course, select from the sequence of intervals, nests of intervals, and we may if we please call such nests 'points' of the Cantor set, but the process which determines the initial sequence of intervals (and this is all we are given) does not of itself determine a single nest. The construction of the sel seems more plausible if we describe the process in a different way. We form fractions between 0 and 1, taking 3 as the scale radix. First we reject fractions which begin with .1, except .1 itself; next we reject those whose second digit is 1, except .01 and .21, then we reject those whose third digit is 1, except .001, .021, .201 and .221, and so on. What have we determined? We might say that what has been deter. mined is the sequence of fractions in the scale of 3 formed only of the digits 0 and 2 apart from a possible terminal digit unity-and this of course is not the Cantor set. Could we not, however, consider that we ha vo determined more than this, that for example the recurring fractions ,I):! or .200:! or the endless fraction .020020002 , .. have been determined, since all the digits of these 'fractions' are either 0 or 2' When we 'reject fractions with first digit unity' are we really talking of fractions, or of the interval (.1, .2); if the former, then rejecting fractions whose first digit is unity, is equivalent to retaining p/39H if P is less than or equal to 3q or .greater than or equal to 2.39 (a .c0!lcept which is expressed by the function! (a, 39)/39+1 where! (x, y) 18 inttoduced by the equations f (Sum (y, '1'), y) = Sum ·(2y, '1'), ! (Ditr· (y, '1'), y) = Dilf (!I. T)) and the set obtained by 9nntor's process is a. set of 'finite' fractions, neither such numbers as .112 nor such as .020020002 .•• (l.e, the sequence of fractions .02, .02002, .020020002, .•. ) being included in tho set, but.if it is to intervals that the·procesa.is applied,

I

i

I

"

.,

then as we. have alreadr seen, all that tho prOCl'SS deterrnines is a sequence of intervals, B!,d If the Cl!ntor set is not this sequence of intervals, but selected nests of mtervals (Le, sequences of pairs a... hOI such that a .. < b .. , a ... ~ a"-!;,, h" ~ h"+h h~-a .. < 1/10', n ~ N(k» then the Cantor set 18 fin,Jte, but undetermined, containing only such nests as ~«: aetua,lly. specify, an~ t~e constructiv~ formulation of the process IS illuso!,) , smoe that which IS constructed 18 not that which we purport to construct.

The source of the widespread belief in the existencc of 'infinities' greater than that o~ a sequenc~ is to be found in Cantor's diagonal process.. (The fun~tlOn! (n, n) IS called tho diagonal function of the two variables functlo!,! (m, n». Cantor observed that, since the function t/J(n! =It», n) + 1 18 gr~ater than !(11, n), ~(n) cannot be equal to an) of the funetlons o.htamed from / (m, n). by replucing m by a definite numeral.(e.g. ! (3, n) IS not equal to ~(n) since ! (3, 3) is less than ~(3» f~om which he. concluded that no sequence of functions of one variable (i.e., no function of two variables) can contain all functions of one variable, and ~cco~dingly that the concept 'all functions of one variable' ~xprcsses an mfinity greater than that of a sequence. Cantor takes It for granted that the. concept "all functions of one variable" is well defi~ed (and defined moreover independently of the diagonal process) ye~ If ~"e attempt to formulate this concept (and ccrtainly some formulation IS wanted, for the meaning of "all" in "all functions" is the verv crux of t~e pro~lem) we ~ceive that what we require is no totality, but a vanable. Sign, no definite function sign, but the fWlction variable /(x)., O~ JDlght, however, suppose that just as the concept 'all fractiOns 18 expressed by a definite function (of three variables) so too the co~cept "a~ functions (of one variable)" could 00 exp~ by a defirut~ !unctlon of two. variables. Cantor's diagonal process {I~01t.S. that this IS not the case, that no definite (unction.of.two.vanables Sign 18· a transformation of the function variable sign "/ (x)" The ~~~ral ,~aria~!e. "x" i8 a definite function, sign, but the f~ction vartable J (x) IS not a definite function of two variables sign (i e th~ :'!. " iii "t (~)" does. not play the part or a numeral variabl~)'; this 18 m no way lDcompattble with the fact that the totality of definite fun~tion signs is finite· (but unlimited) and conaiats only of the totality

of angns that have heen introduced. . •

2: o:r:he classical· iJiterpretation of the 8O~called 'ransfinite ordinal Signs mvolves, the contradicto~ concept of a sign succeeding an endless sequence ~f 8lgn&, e.g. the ordinal w is deemed to be the successor of all t~e 'firute' ~rdinals 0, I, 2, 3, ... , which since the sequence 0, 1, 2, 3, ... IS endless, amounts to saying that we write down (IIr imagine) the

..... ., .. - I' h .,.

Sign . W . a .... r -eomp etmg t e endless task of writing down successive

'finite' o~. In addition to the ordinal aeries which arc supposed formable ~ the ~or .a sequence (in conventional terminology.

COSSTRUCTJVE FORMALISM

the series of ordinals less than W1) mathematical literature introduces ordinal series formed of terms from classes containing more terms than those of a soquence. The sole justification brought forward for the introdurtion of such 'classes' is that the class of all numeral functions affords an example of a class containing more than a sequence of terms; but, as we have observed, the cla8s of numeral functions is finite, if undetermined, and 80 the whole structure of the theory of transfinite elaases is seen to be founded upon an illusion. Yet though the theory of infinite classes, and the attendant theory of the Cantor Aleph numbers, is without substance or significanee, the kernel of Cantor's work, the ordinal theory, which he falsely suppoaed to concern the tra.nsfinite, is capable of interpretation in terms of sequence laws.

The fundamental idea in this interpretation of ordinal theory is tIle concept of the sequence of numerals expressed in a scale by means of a scale symbol lUlII a finite number of digits (for example, the arabic notation with seule symbol 10 and digits 0, I, 2, 3, 4, 5, 6, 7, 8, 9). We are accustomed, in a scale of nutation, to employ three operations, addition, multiplication and exponentiation, but in fact the possible number of operations is quite arbitrary. Using only two operations, addition and multiplication, the number "eleven" in the sca.le of 2, with digits, 0, 1 is "1.2.2.2 + 0.2.2 + 1.2 + 1"; if we introduce exponentiation this becomes "1.22+1 + 0.22 + 1.2 + I". IT further. we introduce the function sign (')3 by the equations (0)3 = I, (*+1)3 = 3")8

(so that, for instance (3)3 = 388 = 327) then the number "thirty-seven" is expressed in the scale of 3, using four operations, by "(2)3 + 32 + 1". Next we could introduce the function sign ',<.,a" by the equations

(o,a = 1, ("""1 ,a = «d')a, and so on, the variable in each new function sign formed representing the number of iterations of a base sign in the previous function; thus the "z" in "m" denotes the number of iterations of the sign "a" in the sum a + a + a + " + a, the "e" in "as" denotes the number of iterations of the sign "au in the produot "a.a,a, . . . a" and the "x" in "(')a" denotes the number of iterations

of the sign "a" in the exponentiation "a .. ·' .... etc, The pouibility of introducing new function signs on these lines has the same multiplicity as the possibility of constructing w·functions which fascina.ted Cantor. An ordinal theory can be constructed eorrespondingtto the representation of numerals in n. scale using any assigned number of operations. We shall illustrate this for the case of the three operations, addition, multi. plication and exponentiation,

First we introduce the digits I, 2, 3, ' .. R by the series of definitions 0' = 1, l' = 2, 2' = 3, ' . . and so on up to the definition of the assigned digit R; thus the digits are abbreviations for the definite numerals, Next we show how to express any number n by means of the digits 0, I, 2, .. , R, R!;; I, and the scale symbol 8. Divide n

·i

..

!

TRASSFlNITE SfT!I1BERS

repeat.edJy by the number R + 1 and let r ' .

successive quotients and remamders TI qt. 1 'b q"!.. r:: ... , he the Bteadil,., dec '. , It" 10 • num Pi'll ql, Q2. Q3, ' , ,

• OJ rea.se, e q'-!-I be the first quotIent which is zero, so that

gil IS less than R + I, Then n = 'I + r2(R + 1) + r3(R + 1)2 +

+ rll(R + ~"-I + gll(R + I)", Each of the numbers r1. r2, 's. :

'II, gil (which are necessarily less than R + 1) .

abbreviation and R + 1 ' laeed b " IS replal'Eld bv its

, 18 rep '"-"'U y 8, gl\'lng .

n = Cilia" + C"~Is"I-l+ • , • + C2BtJI + CI8"1 + C

0, <

wheres~20::::;;Co< 0<

- ,_ s, er, C2, • , ., CII < s, 0 < at < a; < a

and each ai is itself of this form This' 11 d h ~ . , . II,

with digits 0 1 2 R( _. 18 ea e t e representation of n

, , ",., _ e _ 1) and scale symb I If .I.

denotes the representation of n with scnie s then this ex u~. "'«bll)

defined recursively by the equation' pression may II

",.(n) = csr/J.(a) + ",.(n _ C8")

where a is the exponent of the test

exceed n, and C8" is the greatest =, I Pfo~er of 8 whi.ch does not ip e 0 ,,~not exceedmg n.

If 8b; a· denotes the expression obtained b lael

point of its occurrence in an exp-aal'on b y rep dacmfing x, at each

-, • .....0 a, y y, we e e

, , 'l": (n) = 86': "'m (n),

~: ~perator 86':. applying to the expression for which "'m (n) stands,

Just to t~e SIgn "'~(n) itself, and define the re8tn'cW1 ordiMla 0/ the ,flr~ type ~m Cantor s terminology, the ordinals less than E, the first ordlna! to satisfy E = cu~) to be the expressions P'" (n) for any m and .. m~2 F . ·w '.,

-' or inBt&nce PSw(lO~) is the ordinal cuCU+I + 2w2 + 2w + 1

(formed by expressing 108 in th seal f ' "

replacing "3" bv" tt. ~1.~ eo, 3 WIth digIts 0, I, 2 and then

~ cu,. ..: "-RIwn-, we write

S;Cn):= 86;"'s(n), fot 11 ~Ii> 1~

80 that for instan~. 8:(16) _I,~o, ' •. since 16.= 2r and 4" = po, The formula! rw(n) ~ not aUdiatin~, for ~ m' > m

T~(s:. (n}) = P:'(n),

For any nl, 1iJl, ml ~ fn2 > I we say that pC:1 (nl) is greater than, equal to, or lees than ~ (ns) according as nl is greater than, equal to or less

than 8':.. ( .. _) (this d finiti " '

, m. ::" e on IS In accordance with Cantor's definition

of me qualities between transfinite ordinals).

56

(,ONSTRI'CTIVE FORYALISM

A decreasing sequence of ordinals therefore takes the form

T~ (n1), T~ (11~, T~ (ns), ••• , T: (far), •••

where, for each r, m'+I ~ m, and n,+1 < S:~+l(n,). . (For every eo.nstructively given sequence of ordinals the sequence m, l8 general recursIve though not perhaps primitive recursive in each case). For a given

function m, we obtain the 'longest' sequence by takingn,+I=S::+I(n,)..!..I, for T~(n) = 0 if, and only if, n = 0, and for n < n"

s» (n) < S:;:, (n,).

m,tl "H

Accordingly, Cuntor's famous theorem that every decreasing seq~ence of transfinite ordinuls is finite may be expressed (for restricted ordinals)

in the form:

For allY non decreasing function P" po ~ 2, and Jor n, defined by tAe recursive equation n,.\-1 = S~, (n,)":'" I, we can Ihlermi1Ul a value of r

Y,,'-I

for tchich T~ (n,) = 0, and is therefore equivalent to the following

number theoretic proposition:

Given any non decreasing function Pro Po ~ 2, a number no, and the function n, defined by n'+1 = S~~+l(n,) ":"'1, the" there is a value oj r Jor

which 11, = O.

We shall call this proposition P (p" no); it is the general vaUdity of P (Pr> no) for any no and p, that. Gen~zen requires i~ his proo! of the freedom from contradiction of hiS relne .Za~knthe_one. In virtu? ~f Godel's theorem II. proof of P (p" no) inside reme Z~hknth~ IS impossible. For limited values of no. we can ?owevel' gIv:, a s~e~l~

finitist proof of the proposition, for instance If no < po"0 • Wnting Vo = Po, V"+I = pov. the position appears to be that P(p" 110) is demonstrable (in reine Zahkntheorie) for no < .VII Jor any .a88igned k, b~t 11'10 (P (p" no» is non Ihmonstrahk. This of course mvolves tb!,t m reine Zahlentheorie there can be no general ~ormula Gp,. (Ie), With .. a free variable k, specifying the number of term~ 10 a decreasmg seq~ence commencing with the number "k, but only speciile formula) for particular

values of k,

To extend the representation of ordinals by, functions ~f n~~r theorutil! variables, to ordinals of anr type w~ l1ltrodu~e, in addItIOn to definite numerals and numeral variables, maJorant vonabks a, w'. w" r ~ 1. A relation containing II. single majorant variable 8 is required

,

,

i

,

TRA.~SFlN1TB ~l'MnER"~

a;

~o hold, not necessarily for all non negative integral values of a, but for all values greater than some assigned constant. A relation R(w, WI, Wi, ••• , w ... ) between the majorant variables w, w" (1 ~ r ;;i m), holds if we can introduce a constant Co, and functions c, (no. n I, . . ., 11,-1) such that R (no. nlo Ri, • • ., 11m) holds for' all non-negative integers no. nl, ~ •. , 11m such tha.t no e;; Co and 1I,+! ~ C'I-I (no, 1'10 ••• , n,), o ;;i r ~ m - 1. For instance, if a, b, c are definite numerals then

rP > aab + e, since 1In > anb + 0 if n > max (a, b, c); and an example of a relation between two majorant variables is

WW WJw + 3w2 +" > (l}lw WI\! + 7w2 + 11, which holds since

nnN" +3n2 +" > n3nN2 + 7n2 + 11 when n ~ 12 and N ~ nS.

Majorant variables obey the same rules of addition, multiplication and exponentiation as numeral variables, for if a relation R(no, nl, 1&2, •• "" n ... )· holds for all non- negative integers 11" 0 ~ r ~ m, without restriction, then a Jortiol'i,R(w, WI. ••• , W ... ) holds.

If J (x) is any function of a numeral variable x (and possibly of other numeral or majorant variables as well], such that J (0) = 0, J (I) = 1 and J (x + 1) ~ / (x) + 1 for all z; we introduce the function III)

'P". a (n), for any numerals k ~ 1, n ~ 0 and (a numeral or majorant

variable) ~ > J (k), by.the following recursive equations:

. !fJIf$) (0) = 0

... ,. Ilia .'

~. '

. (i)

. "

where II is the exponent of ~e P,ateat p'ower of i + 1 not exceeding n,

and c(1e + 1)41 is the greatest mnltiple 'of (Ie + 1)<1 not exceeding tao For given k aind n, th~ equations (i), (il) 'deterniine the'~va1ue' of !fJ~:~(1I) 88 a fonctionm-'" and 1(0), /(1)"'.,0 ., I(i> .~hich we call the represent,.tionof n,~th ciigil8l(r),.~ .. ~ t; ~ i. and bCJBe,,~ . For instance

. • • 1

'. ~~,,(561):__ 4a" + 9aC+ 1. ,"

~en 1. oontains more than one argument the operative variable is wrItten in the last argument place, . e.g. the operative variable in 1(P, g, x) is :1:.)

.... ' ",

58

('O~lITRl'CTIVE FOR~IAr.IS:'l1

Nt'xt we introduce the functions 5,.. p ,(11. k), .0" (n, 1') by the recursive equations

8".", (0, k) = k. D" (0, k) = k

... ( D~I".,t' k)

.J.::'''' ,.. II," (/:) .0 ( + I k) .J. f" (. Wo=W

E,.. p' (n + I, l:) = '1'/>1",.,.',", .. 1 ' "n ,. = 'l'pjn,.w" ' ' ,

where pC"~)' p'(n) are any two sequences such that

pIn) ~ I and p'(n) ~ S",,,, (n, p(n» for all n.

For given k, nand p(r), 0 ~ r ~ n - I, the 'values' of the function D,. (n, k) are called transfinite ordinals 0/ type n. The values of D,. (n, k) are, of course, not all distinct; in fact it can be shown that

if q(r) ~ 5", q (r, p(T», 0 ~ r ~ n - I,

then Dp (n, k) = .oq (n, S,., q (n, k».

For example, taking p(O) = I, p(l) = 4 and q(O) = 2, q(l) = 99, we have D,. (2, lI6) = WWWlw + (w + l)wI + I = .of (2, 27.IOOS + 4.100 + I).

The fundamental theorem is that, for all n, >

D,. (n, pIn)) = Dq (n, q(n» <

>

according as Ep,,(n, p('l» < Sq,,(n, q(n»

where r(/:) = max{ s; (le, p(k», s; (k, q(le»}, n > k ~ O.

It follows from this that lor ordinals 0/ f!-ny type_ Ca.nl?r's t~eorem that a decreasing sequence of ordinals necessarily termmates 18 equivalent to a number theoretic proposition, .

For the sequence of ordinals .of> (i, p" (i)), n = 0, I, 2, ... is

steadily decreasing if ....

S" , (i, P"+I (i» < S" , r (l, p,.(.»

,,+1 r" • "

where } 0 ..... r. '

r (k):;:: rnax{SI' ' (k, p,,(k», S" " (k, p,,+I(k» ,n ~ 0, ~I& < t •

., - n' .. ,,+1 "

and so Cantor's theorem is equivalent to the' proposition:

If for all nand 0 ~ k < i, p .. (k) ~ 1 and pe(i) ~ 0, and if r,,(k) = max{ S,.". ,,,(k, p .. (k», SP"+l' ,,,(k, p"+1(k»}

and

E , (i, p"+l(i» < Sf> " (i, p .. (i» when p .. (i) > 0,

1',,+1' ,J IS n

then there is a value of n ,for which pe(i) = O.

'i

,j

A

I

50

CHAnD' VI. ' ,

THE INTERIOR OF A CLOSED CURVE.

The length of a line segment is not the measure of some totality of points contained in the line, as the theory of measurable sets suggests, but is a function solely of the pair of end p'0ints of the segment; the essence of the notion of length is just the Iteration of a standard pair of points. So, too, the area of a plane surface is the number of iterations, over the surface, of a standard square, i.e., a standard set offour points. Mathematicians have confused the mathematical notion of length, or area, with the psychological concept of conknJ, and likewise the mathematical curve with a string or wire.

The common 'curves' of function theory, the circle, parabola or catenary are only hypothetical end terms of sequences of curves for it is not i mathematical circle which a pair of eompaseee describes, nor does a heavy chain hang in a mathema~ical catenary.

The fundamental theorem on relative continuity replaces the traditional continuous limit of a convergent sequence of functions len. :1:) by a sequence of step-funetiona h,(k, :1:), where k,(k, :1:) 1rel1er misses a 8lep, that is to say, takes every value n/lO- between any two of its values. This suggests that from the constructivist standpoint the classical conception of a curve may be replaced by a sequence of rectilinear figures. based on networks of squares, and determined by 1\ sequence of ~ ,of step functions h/(k, t), h,(k, t). With a trivial modification 8uc'h a rectilinear figure may be represented by a matrix

(%0. %1, :I:t. • • •• X:) go. 1I1t gB, • • " 11

where e, y, arc integral multiples of the mesh. 1/10-, (i,e. length of side of a square) of the network and

111,+1 -1Ir1 ~ 1/10- (i)

and

(X'+1 - :I:,)(Y,H - y,) = 0' ,

. .. ."

•••.•••••• (ii)

60

CONSTRUCTIVE FORMALISM

The length of the figure

(XO Xl X2 ••• X:) is _,E=ll { I X,+! - x, I + Iy,+! __ y,j }. Yo Yl Y: ... y

If I e, - x'.1 +' I y, -lIs I > 0 for all r, s, 0 ~ r ~ n, 0 ~ S ~ n, the figure is said to be simple and open.

(XO Xl X.. • • • X .. ). • 1 and open and if the matrix

H Yo Yl Y; ... s» 18 sunp e

(XO Xl X2 ••• X .. XII) satisfies the conditions' (i), (ii], then t~e latter . Yo Yl Y2 • • • y. Yo

represents a simple dosed figure.

Only the cyclic order of the elements of a matrix representing a

simple closed figure is significant for _.

(X, X,+I ••• x .. Xo Xl ..• X,) represents the same closed Y, Y'·H •.. 11 .. Yo 111 ••• y,

figure as

(.1:0 .1'1 x:! .•. x" xo) liD III Y:l • • . Y" Yo •

THB INTBlUOR OF A CLOSED OUBVB

'j

61

The horizontal botuularri level8 of a closed figure arc the values of y, (= y.+ I) for which I x,+t - z, I > 0, and the vertical boundary levels

are the values of x,( = X,+l) for which J Y,+t - Y, J > o. '-

It is easily shown that, in a simple closed figure, there are an even number of horizontal boundary levels between any two conseeutive vertical lines of the network, and an even number of vertical boundary

levels between any two consecutive horizontal lines. .

A square of tho network which has an .odd number of horizontal boundary levels' above and below it (including any side which is a boundary level), in the "vertical strip' determined by the l"ertical sides of the square, is called an interior x-square of the figure, and a square which haa an even number of horizontal boundary levels above and below it is called an exterior x-square. Similarly, we define interior • and exterior y.squares in relation to the vertical boundary levels.

By purely algebraic arguments for the matrix representing a simple closed figure it can be shown that each interior x.square is an intorior v-square, and conversely, 80 that we may drop the prefixes and speak of the interior (or exterior) squares of. the figure. This provides an algebraic proof of the Jordan sep&ration theorem (or the simple figures

under consideration. .

A distinction between clockwise' and anti-clockwise matrices is

readily Introduced; The matrices .'

,(; ~ ·,t;·,;)~ e ;" ~ ~' ;)

where x' is the sum of.x !U1d tho mesh of the network, are called, respectively, the clockwise and the anti-clockwise paths round the square (x x')(y 1/).

We observe that 1n.othe anti-clockwise 'path

(1) ..., " ;">'(~:. (_.<~~.: ~'.·X) ... :

'. ' ... 1J.·cY' 1/ _11 .Y .. ~ ,

and in the clockwise path' . ~,::,c~t. ,._ . -, ,,'''.'; .'

_." ',.-. ",0 _.~,.' ... .-' .' _ •• 'f>' .

(2) e z,·, ~ :10, 'x) ",:_,-c:·.;(X~l~fH." ¥ 'Z)

y Y Y· Y· Yc,'""7:' c ~~lI-!r'" 'y* ,

. (where y. is y 1088 the mesh of. the ·ne~or~), the term x precedes x' at

the boundary level Y; but in t~~a~~-cloc.kwise ~th .

(3) e : ;.;~;) _ (;.:.;:~;' ; ;.)

and tho cloekwiae ., ..

(4)

.... ,. "

(xy' x -. Z x' Xy')

y . 1/ 1/ .~

s:

62

CONSTRUCTIVE FOBJULISH

the term z' precedes x at the level y. Thus if 11 is a boundary level of a simple closed figure C, lying in the vertical strip (x, z') then either the matrices (1) and (4) or the matrices (2) and (3) represent paths round

the interior square with side (; ~'); in either case in only one matrix will the order of z, z' at the level y be the same 88 in the matrix

representing C. Similarly, for an interior square with side e ;,) ; accordingly with each boundary level of C we have asaociated a unique path round an interior square and this path is said to be described in the same sense 88 O. It is readily proved that aU the paths round interior squares of 0 and described in the same sense 88 the figure are either all clockwise or all anti-elockwise ; in the former case, 0 is said to be clockwise and in the latter, anti-clockwise.

The matrix

(Xy" Xy" -1 • • • Xyl %yo) is called the reverse of the matrix (Xo Xyl • • • Xy:) ;

n .. -1 • " 1 0 \Yo 1· ••

if one is clockwise the other is anti-elockwise.

To illustrate the use or this simple constructive curve theory we outline a proof of Cauchy's fundamental theorem for a differentiable . complex function.

We take for granted a number of extensions of previous results to two variables, and the elements of a complex rational algebra.

If z = x + i y where e, y arc rationals such that z = 0 (k), 1/ = 0 (k), then we write z = 0 (k).

Continuity. A complex function V(n, e) = u(n, x, y) + i II(n, x, 1/) is said to be continuous relative to n, in a rectangle R, if we can determine recursive functions elk), O(k, x, y, X, Y) such that

V(n, Z) - V(n, z) = O(k)

for any z, Z in R satisfying Z - z = 0 (e(k», and for n ~ O(k, e, 1/, X, Y), where z = x + i y, Z = X + i Y.

Differentiability. V(n, z) is differentiable in R, relative to n, if there is a complex function VI(n, z) such that for any two diJ/erent z, Z satisfying Z - z = 0 (d(k»,

V(n, Z) - V(n, z)

- ---Z -------- - Vl(n, z) = 0 (k)

-z

for n ~ D(k, x, y, X, Y).

@!!

I

TIm INTE&10B OF A CLOSBD CUBVB

63

Inlegrolion. If I (n, x) is continuous for a ~ x &; b so that I(n·, x) -/(n, z) = 0 (k) for n· ~ n ~ N(I.:, z],

and continuous in (a. b) BO that there is a. subdivision (a~), 0 :s:: r ~ Ilk of the interval (a. b) for wblch - - •

I(n, X) -/(n, z) = 0 (k), n ~ O(k, x, X),

where a~ ~ z ~ X &; a~+I' 0 ~ r ~ nA - I, then we define

nA-l

I, (k, a, b) = E I(N (k, a~), a:)(a~+1 - a:) , if b > a

,=00

=O,ifb=a

= - I, (k, b, a) if b < a.

The sequence I, (1&, a, b) is convergent in k, and is called the integral

of I(n, x) for a ~ z S;; b. .

Complex integral. If r is a simple figure (xo ~1 ••• X~)

, \Yo 1/1 ..• y~

(on a network of mesh 10-) and if ,u, u, denote u(n, x" y), u(n. x, y,) respectively. and similarly for ,II, v" then we define

Iv (k, r) =,.

=~il 1Iu (le, Z,,%'+I) -1 e(k, y" 1/,+tl+il" (k, x" z,;11+i1 u(~.1/'.1/'+ll t

,=01 , , , , I.

:] v (le, 1) is called the integral of Vent z) along r.

I, ' .·,O::i." .,'.'

If F' is ~h8 m~ which represents the ~Pte figure r on a net. work of mesh smaller t.han 10-", the additive properties of the integral show that 1 v(k, r) and 1 v~k, r') are ~qual relative to k.

The properties of simple figures which we have enumerated suffice ~ prove that, if r .is a simple closed figure, the integral of V along r 18 the sum of the mtegrals of V along the paths round the interior squares of r.

H r n is the matrix which represents r on a network of mesh lO-n, n > m, then it readily follows, along the familiar lines of the classical proof, tha.t if V is diJ/erenlitWk (relative to 11.) the BUm 01 Ihe integrals I v (k, s) round the interror aquares 8, 01 l' OJ, converges to uro, in n, relatitl8 to k, AND SO Iv(k, r) = O. RELATIVE TO k, which proves Ca.uchy's theorem for the simple figure r.

CONSTRUCTIVE FORMALIS1l

To extend the theorem we define next the integral of V over a sequence of curves :r = :r (n, t), y = y (11, t). If r .. is .the rect~near figure on II. network of mesh 10'" determined by II. pall' of relatively continuous functions x(n. I), Y(II. t), then the ,educed ~eque~. of I,· (n, r ".) is the integral of r 1I\'('r the se9uen<'o of cur!e9. ThiS ~efinitlOn anticipates the theorem that I v (II. r ... ) IS convergent m m, relative to n. We prove this, for differentiable J' by appeal to the Cauchy theorem for a simple figure.

Let ,11 > 11t and let r'm be tho equivalent of r lit on the network of mesh lO-M, so that

I r (n. r m) = I I' (n. r'm), relative to n.

If (ao, bo), (ai, bl) are the end po!r~ts of r; and (ao. b'"o). (iii. bl) the end points of r M then by the detinition of r n.

a,-a,=O (m-I), b,-b,=O (m-l). ,=0. l.

Join (a" b,) to (ii" b',) by a simple figure 1", of lcn~~ I ii, - a, I + Ib,-b,l, on the network of mesh 10·-·\1., = 0, 1, then r tn, 1'1, -r M and -yo together form one or more simple closed figures (where -r M, -1'11 denote the reverses of r M. yo respectively), and so

Iv(n, r'm) - Iv{n, riot) = Iv(n, yo) -Idn, 1'1), relative to n.

But Iv(n, yo) and Iv(n, 1'1) are both 0 (Ie + 1) for a sufficiently large m, and so

I v (n, r ... ) - Il'(n, r M) = 0 (Ie), relative to n, which proves that I v (n. r.) is convergent in 'm, relative to n.

IT r til is a simple closed figure for each value of m, then it follows immediately that the integral of V over the sequence of 0W'!es :r = x (n. t). y = y (n, t) converges to zero, for I v (n, r .) = 0, relative to n, for any m, and so the reduced sequence of I v (n. rill) converges to zero.

e

!

65

CIIAPTBB ro.

THE GENESIS OF THE NillmER SIGNS.

Although a. hypothetical reconstruotion of the development of number signs, unsupported by external evidence, is of little historical ,

r value, it may yet be of importance in the foundations of mathematics, throwing light on present usage in the way tha.t a theory of physics lluminates the facts of observation and experiment. The logicial

j development of a branch of mathematics seldom follows the line of \

historical discovery, and in fact there is an obvious sense in which the J

1 iBtory of mathematics. is without significance.

1

,I

~! ; §

~

j

Both by form and usage a stick pattern. ancestry for the &rabio numerals is strongly suggested. The original pa.tterns are not known, and in any case the evolution of the cursive forms is unlikely to ha.ve occurred without a variety of transitionary steps. but from the stand.

I point of a logioa.l development one would expect the earliest forms to , exhibit an 'inductive' genera.tion process of the kind shown in the

I following array:- " .

I LCD

o 000

o 000

" I o

L 0-

o o

L

c.

o

Here the patterns are baaed upon squares and the iteration of the first four figures and each pa.tternis oontained in its successor. The 'inoomplete' figures all look to th~ ~d the resemb1&1lce to the arabic numerals is scarcely a.ppa.rent. If ~e turn the 'three' to the left ,!,nd. t~anspose one stiok in the 'five', 'six':_and 'nine', thereby knitting theindiVld~.pa.tterns more c,loaely togetJi9r, we ob~ the figures

4(' • ,t • ~ -:

5 5 ::J 0 ·9

L ~ 0 0 00

1284 s 6 789

~bioh' bear a striking sbnnarity"to the ~ursi~~ n~era.ls.

The development of stiek patterns from primitive stick tallies like

II

III

IIII

IIIII

IIIIII

IIIIIII

marks the invention of couiltingr"

66

CONSTRUCTIVE FO~SM

Making a tally of a collection is th~ same operati~? 8$ ma~ng a copy' a stick- tally of a flock of sheep IS no more an abatraction of

the ~lIll1ber concept from the flock", as has sometimes been s1!Ppo,sed, than "' child's drawing is an a bstr:tction, but a cOPY. ot the flock In shcks, ~ If the tally is used to guard againsf loss from a s~ngIC1loc~, then ilie collection of sticks serves !IS well as a more detailed drawing, When

a distinction is needed between the tally of a flock of sheep and the tally of a herd of cows, the two collections of sticks, by themselves. are

of course no longer adequate and some indic~ti?n ?f sheep and cow respectively must be added to the tally, The dietinctton could of course

be made by replacing the sticks by drawings of sheep and cows, ,but

I the interesting discovery was that it suffi~ed. no~ to change the sticks \ but to attach some distinguishing mark. like a ,5mgle sheep or C?W, to , the stick tallies, This discovery is reflected III our language III the

equivalence of the sentences "a sheep and a sheep and a she~p" a?d "one and one and one sheep", and in algebra by the Identity "x + x + z = (1 + 1 + l)x". We do not have to perceive that a sheep and a cow are both 'instances of the number 0!1e' befor~ w~ can represent either by a stick; a tally of a flock or a herd IS a tally I,n vI~tue \

I of its use, not its form, :llld the (list~nc~ion between a ta~ly kept III sticks and one in sheep heads IS no more significant than the difference between I a tally in sticks and It tally in stones,

The organisation of a primitive, sti~k tally, into II; num~er p~ttern

has the obvious advantage of substituting an immediately Identlfiab!e

" design for a mere collection of sticks; if the designs are named, It i suffices, finally, to record only the name lionel dispense with the pattern, I

The arrangement. of a stick tally in a ~umber pa~tern do~ n~t

itself constitute the full discovery of counting. CountIng oonatBta In the successive. trans/ormation 0/ one stick paUern into the next, so, that nothing but stick patterns are construc~d at each, stage and the introduction of each stick transforms one stick pattern into another, If we count a flock of a dozen sheep by, patterns, we construct in turn the

o

L. C, ' , . B, the end of the operation being the single

patterns I, o

pattern 8 (the use of stick copies of the sheep is of c?urse. irrelevant,

, since we could imagine the sheep ~hemselves arranged III this patte~n).

(Counting is therefore a trans/or?Wt!on process whe~eby one number SIgn \ (the flock itself or a copy of It) 18 transformed into another number.J<Esign (the stick pattern or numeral),

When we count in words, one. two, three, and 80 on, we appear at first sight simply to be naming so~g whioh we hg.ve perceiv~ -,--

,

THE OESF.STS OF TlIF. StTMlIER STO~~

67

and since we know we cannot tell at II glanee the number of n group like I II III (and in fact we count because such recognition is beyond us) it becomes hard to understand what it is we name when we are counting. Russell supposed that the object of counting was to establish a one-to-one correspondence between the objects counted and the numbers themselves, which we may indicate by the scheme

I 1

I 3

I 4

I 5

I' I 6 7

I 2

It is certainly true that oounting does effect such a correspondence but setting up the correspondence cannot be the object of counting, as a consideration of stick patterns shows us, for to what purpose shall we add to the construction of the patterns a correlation of the pattenis that have been superceded to the parts of the group that is being oounted 1 For instance, to count the group of sticks I I I I we form a succession of stick patterns leading up to 0, which alone remains; to

establish a correlation of the kind ',.,

I I' ' I

LCD

, we should be obliged to introduce ten extra sticks, If such a correlation.,

I were the object of counting, then we reach the absurd conclusion that. counting one group is not possible without the aid of another group, '

, ,

Counting by' stick patterns shows us that counting in words, one, two, thr" 66, and so on, involves neither an immediate perception of number " nor a correlation of number Bigns with the objects counted, but is an

enumerat.ion .'.?f. the, transform. ati0?l rules one (and one is) two, (two \ and one '18) tliiee, (three and one 18) four, and so on, this enumeration

taking the place of t~e successive tran8f~rmations of one stick pattern into another by the ,8.ddition of a stick. Thus counting in words, like counting ~, p~tterns, ,is ~. transformation o£.number signs.

, . f: ~ - ..

, !" ;~: .. ,.. ~'-

Stiok patterns are only;one of the many notations that mathematics has employed in the past. and are of no special importance apart from their connection with, the cursive numerals. The Greeks made wide use of dot patterns and the Hebrews and' the Greeks both found in their alphabets a ready made number notation. It is well known that we owe to dot patterns a variety of interesting theorems on figurate numbers, like "the sum of two conseoutive triangular numbers is a square number". Dot patterns serve as well as stick patterna to elucidate the :na.ture of counting. In the form

. . ...

.. ,

68

CONSTRUCTIVE FORMALISM

the dot patterns share the inductive character of the stio~ patterns. each pattern being incorporated in the succeeding one! but m e~>ntra.st to the stick patterns. the dot patterns are not so ~adily taken m at.a glance. do not lend themselves to a cursive adaptation, and completely fail to suggest the idea of a scale of notation.

The use of an alphabet for number signs jnmps directly from ~he tallv to tho equivalent of a cursive numeral but the rules for transforming

II. tally into an alphabet letter will probably introduce s~me equivalent

of the missing pattern. construction (for instance, countmg). or as we have seen we are obliged to assume an ability to recognise at a gl~nce "the number of 0. tallv", Such perception is not logically fmpossible, ' and aided by 0. relati;ely small scale of notation, may well have been h-

i eommon ; for instance, a tally of a flock of 1797 sheep recorded on an I \ abacus, with scale number five, in the form

may be read at a glance.

In the logistic theory of number. a nnmber is defined as a' class of one-to-one related classes, so that, for instance, the number two is the class of nll pairs; since the totality of pairs (trios. etc.) is undefined. the logistic theory defines in fact not number but the property "the same number" between two classes. Two classes have the same number if their elements are one-to-one related. To win the concept of number, itself we would be obliged to introduce a succession of standard classes with which other classes could be matched-a standard unit, a standard pair. and so on; then to say that a class has three te~ would sa!

that the class is one-to-one related to the standard tno. But this ,

definition makes the number concept -rest in turn upon the undefined .

notion of a one-to-one relation. We prefer to say that "a class of V

' objects has sueh- and-such a number". is ,~ynonym0lll! .with "a class of /1 '

1\ objects is such.and-such a number-SIgn , by defirution, so that for instance to say that the class "abc" has three numbers says that "abc" Ii:' and ":J" arc equivalent number signs. " f \

Two signs are equivalent if they have the same part to play (iIi

a language, formal calculus, game, etc.), that is, if o~e may be ~bsti-

\ tuted for thc other in virtue of some rule of transformat~on. Acc?rdingly, number is the part played in language by the nU~-81!1?"'; for lDB~O~, the number two is the part played by a two-slgn m virtue of which It

TlIF. OENRSIS OF TIlE "Xl'\IRF.R SIr.N~

69

'~

J H "I !

is a two-slgn, And if we are asked what this rule is, we can only allll\l'('r by oonstructing u formal arithmetic, whether it be an arithmetic of sticks or stones or ink marks on a sheet of paper. The part which a. character takes in a play is defined by the play as a whole, by the lines which he speaks and the movements he makes. and their relation to the acts and words of the other charaoters in the play.

: In saying that a collection of objects is a number sign we are stressing one particular use which may be 'made of a collection. not some common property of the elements of a collection. We use a collection as a number sign when we match it against another collection, for instance in making a tally. or transform it into LL numeral by counting

or pattern construction. .

Making a tally, for example a stick tally of a flock of sheep, sets up a one- to-one oorrespondence between the sticks and the sheep, so that the collection of stioks and the flock of sheep 'have the same number' in the logistic sense, but it makes DO sense to say. as Russell said, that such a one-to-one correspondence shows that the two collections "have the same number even though it is not known what that numbel ja", for in making a tally We are using, the collections as number. ~igns.

\' Certainly, making th, e tally is not the same thing as counting the flook, 'tl II but counting is not an act of discovery, only a transformation &om

~ one nnmber notation to anothee.. To !).ave the flock before yon is to have the nnmber of the flock: oounting only transforms that number

(into a.more convenient syu;t.bolism. This links up with the fact that \ the idea of a one,-to-one cornru.tion is not ~he foundation of the number_) concept. We may transt'orin two. collections of sticks into 'figure fives' and thereby determine that the colleotions have the same number witbout correlating the elements of the two collections-and, of course, countlng

is another meanS to the same 'end. " '

.'1:: :The' ~~'~~On ~\~h~ther'w~ necessarily obtaiJi'the

. Same result wben~e teat ,the equality of two collections of sticks, fIrilly by matching them one ag~:the . other, and then by transforming ~ch colleotioninto a,stick.pattern. 'The BlP.e problem is met in eoaneotion with counting rame. coUecti~ where repeated counts may give difFerent results. Lot US imagfn8 two 'p'Ues have been formed by breaking four sticks in:'hrur and placing one ! half of each stick on each pile. By definition:' the piles are' -one-to-one related. Suppose that when we transform the piles into stiok patterns. one pile forms a figure five and the other a flgure six. Shall we say that the' two piles have the same nnmber ot' sticks or not 1 We might of course account {or the discrepancy by 'saying that the second pRe has spontaneously generated another stick (or that one stick has disappeared from tho first), but what is important to discover is the source of our conviction that tho two piles have the same number of sticks because they are one·to-one related. IneJreot. what we are saying, is that two collections whioh are one-te-one related by one method of matching must be one.to-one

70

('ONSTRtTCTIVE FORlIlAI.lSM

related by any other method of matching, for pattern formation may also be regarded as setting up a one-to-one correspondence of the two piles, or to reduce the problem still further, if two piles are one-to-one related now, and if neither is disturbed in any way, then they "ill be one-to- one related a few moments hence. In this form the solution of the problem is immediately evident, for there is nothing (in mathematics) which can foretell the result of an experiment. The piles may match now, and yet not match at II. later moment however much care

\ is taken to see that neither pile is disturbed. Just as two colours may { match now and not match a moment later.

Q- A world- in which objects appeared and disappeared spontaneously would be a world in which our common arithmetic would find no J application, just as our language would find no part to play in a world

~hich disrupted the familiar coincidences of visual, audile and tactile experiences.

The process of counting II. collection necessarily selects the elements of the collection in some order. Since the object of counting is to compare two collections, or t. 0 _gunr(l;-~ainst 1088, the order in which

\' the elements are selected is (immateri~; that is to say, we are not II interested in the number patthms.iornfed in the course of counting, but only in the final pattern, which is called the cardinal number 0/ the colledion. (Whether or not the cardinal number is independent of the order of selection is a question of the same kind as that we considered

in the last paragraph, and docs not concern us here). In other contexts, however, the successive number-signs formed in the act of counting are no less important than the terminal sign, for instance, when we number the houses in a street.

The number .given to each item of a collection in the process of counting the collection (that is to say the number-sign which that item completes) is called the ordinal of that item in the counting. Thus, when a line of soldiers "numbers off", .the number each man calls is his ordinal; each man's ordinal is the cardinal of the segment of the line up to and including himself, or what is the same thing, the successor of the cardinal of the line of men before him.

Labelling each sheep of a flock with its ordinal number is no better safeguard against 1088 from the flock than recording the cardinal of the flock, but if a sheep is lost the ordinal labels will show which one is lost. The ordinal use of numbers is a logical nomenclature for the individuals of a collection.

In virtue of the definition of number, a language lac~ number words can express everything which can be said in a richer Janguage, for numerals are abbreviations, and every abbreviation is dispensable. For instance, since a tree and a tree and a tree is three trees, the sentence "there are three trees in my,garden" means "there is a tree and a tree and a tree in my garden". '.

e

!

THB OBNBSIS OF THR "~BER SIGNS

71

Owing to the maturity and 8exibility of language an, 1 thv great range of services the number words perform, the elimination of number words is not necessarily as trivial in every instance all the foregoing example suggests. The relation of ordinal to cardinal number shows that the elimination of ordinals involvcs the introduction of an ordering relation, and the elimination of a cardinal. Thus, "his is the third house" may be transfOlmed into "there are two houses before his house", and then to "there is a house and a house before his house".

Once, twice, thrice, and so on, are adverbial uses of number, qualifying phrases, 80 that "he paid his fare bice", means "he paid his (are and he paid his fare", not "he paid his fare and hia fare" (which is "he paid two fares"). Single, double, treble, and so on, are used both as adjectives

\ and verbs, The adjectival use. is generally just a literary variant of I ' the cardinal numbers; a double deck omnibus is an omnibus with two decks, but a double bed is a bed for two, As a verb "double" is the product function 2%. Terms like pair,trio, two-fold and triplex are

compound abbreviations.

The introduction of number wordS in a language is not a logical

necessity but a practical convenience, and this is true not only of numbers j I but of the whole of mathematical symbolism. The vast expanse of

I mathematical diseovenee is traversible by the single steps of a formal ,system. There is no special mathematical intuition which attains

\ points inaccessible to all but a gifted few, but without the seven league boots of symbolism the journey Is so long and arduous that no one can

accomplish it.. . .; .

,I

i- c.

.- '.";- _.

I.' ~;

... ', ,1..

- - .. _r~ •.•.. ~~.. ;:>~

:- ::. ;:-. ~. ~-'

; ,

~. ' ....

;' .-,j

• :1

, i' .•

-"

72

COSSTRUCTIVE FORlULISM

CHAPTER VIII.

LANGUAGE AND EXPERIENCE.

In the foregoing chapter we have attempted a resolution of the fnrmalist-finitist controversy in the foundations of mathematics by the construction of an axiom-free calculus which provides a purely formal criterion for the intuitive concept of a finitiat proof. But tho reduction of mathematics to n sign language does not, of itself, contribute towards 1\ solution of the deepest problems in the philosophy of mathematics, for these are problems about th" nature of sigus and the relation of language to reality. In thil! chapter we take up the consideration of these 'ultimate' problems, not simply in relation to a mathematical system, but ill their full epistemological generality.

Problems concerning the nature of signs and the relation of language to reality find expression in such questions as: "Is language no more t.han a system of signs? Has language a content, or does it float above t:eality like a bubble above the earth? Can language point to something outside itself, has it roots in some actuality or are the truths oflanguage independent of all experience 1 If language is a medium of communication (between human beings) then what is it that is communicated, and how is this communication effected 1" As a preliminary to an examination uf these questions, let us consider what we ordinarily sayand do-when we sock to decide whether IL proposition is true or false. Suppose the proposition ill "The entrance to the College is in University Road". To see whether or not this is true we might first seek the address of the College in a Street Directory, then we might ask various people the question "Where is the College?" receiving the answer "University College is in University Road", and lastly we might walk along University Road looking at each building in turn until we reach one bearing the name-plate "University College Leicester".

That the first two criteria are of the same character may be seen by supposing that the directory we consult is a machine which reproduces (from It record) the 1 u ldress of any institution when the name of the institution is spelt out on lL dial on the machine. (For the present _ we ignore the sort of doubt which might be expressed by saying "the man who is nsked the question does not just answer (liko the machine) but must think first before he can answer'") When however we turn our atteutiou to the third criterion we are inclined to think, not only that it is Iundamentally different from the other criteria, but that it is COD-

I

,

I

I

LANGUAGE AND EXPERmNCE

;3

I

1

clusiv~ in a. war the ot~ers could n~t be; th~ reference book might contain a mtspnnt, our mformants might be mistaken or even wilfully deceive us, but "we cannot doubt the evidence of our own eyes". To see u:hat this last criterion has in common with the preceding, we shall describe yet another form 1\ reference book might take. The book might contain photographs of streets so that to find whether the pro. position "the College is in University Road" is true we look at the photograph of University Road to see if it contains a picture of the College. Thus we might contrast the criteria by saying that in the one we look at the object itself and in the other at a photograph of the object. Since we can use the reference book which contains the sentence "Universfty College is in University Road" and the reference book which contains the picture of tho College in Unh'ersity Road in exactly the same way, the sentence "University College is in University Road" and the picture of the College in University Road, must stand to one another

in the relation of syntactically equivalent sentences, like a French and English sentence with the same meaniug. To translate from one word language to another we place side by side the words which may be "Titton one for the other; to translate from a word- to a picture-language - (like the directory containing photographs) we place side by side words and pictures. . Thus the correspondence between a word and a picture language is established in the same wort of way as the correspondence between two word languages is established. But if we can find a correspondence between the sentence "University College is in Univendty Road" and ~ picture of the College in University Road, how can we doubt the existence of a 'eorrespondenee between the sentence and that

of which the photograph is a photograph, namely, the College building standing inU~versity. ~ad _, Can we not translate from a picture language to a real object language t If we want 0. man to build us

a house can we give him only a picture of-what we want (like the word "house" or a photograph of a house) and can we not just point to a _ house 1 When a pure formoJist- maintains that there is no such proeess ,'_ 88 derlvin~the truth of a proposition from some non-verbal occurrence

in what sense is he using the term "verbal", He accepts as-a criterio~' . of the truth of-a sen~ce that the sentence form one of a certain list

of sentences •• He .~ 'the. possibility of deciding whether or not two :ows of 81gIl8 form the ~e Bentenee; for wnen he speaks of the consistency of two sentences,1t 18 the shape of the Bigos which comp036 the sentences to which· he refers~ and he speaks of the posaibility of changing from one notation to another,. translating from one language

to 'F0tber. But if one admits the possibility of translating from one wo~ language to another t~n one must admit the possibility of translatmg from a wo~d-. to a ploture-Ianguage, and must therefore accept the non-vu:bal crIterion of the truth of, a sentence that a sentence is true if it is- a translation of a picture ('sentence') that forms one of a certain collection of pictures. .ADd if one accepts this criterion why

~:~:~~ not ~~t ~ the Crite~on of translation from an 'object

{'OSilTRtTM'TVE FORlIIAI.lSM

Of course, the formalist, mav say that he meant no leas than this himself, that by "verbal" he meant 'anything that could be 'used as a word', but if this is so his statement loses its entire point; for if "verbal" no longer serves to distinguish words from other signs, then "non-verbal" has nn meaning left to it. and if "word" is being used in this new sense then to say that the world is the world of sentences is only to say that the world is the 'familiar' world of facts. We cannot deny the formalist the right to call University College standing in University Road a sentence; for some purposes this is a valuable form of expression, and harmless so long as we do not forget that it is a metaphor and not, as the formalists seem to imply, the expression of a now discovery about the world. the discovery that the world is a world of (what we used to cnll) sontences.

Talking to someone, sending someone a letter, communicating with someone, have the character of drawing someono's attention to something, holding something in front of someone, pointing to something. Pointing to a nut on a table and then pointing to the table may be called saying the sentence "a nut is Oil the table" in the 'real-object' language, a sentence of which "a. nut is on the table" is the translation in the English word language, Saying what OIlO sees, hears, feels, describing an experiment, recording an observation, are all translation processes, Percl'i"illg II relation, observing a difference, recognising a Iikeness, are akin to naming a relation, a difference, a likeness. Pointing to a pencil and saying "pencil" ill one of the ways in which we translate from an object. to a word-language. Were it necessary, as the formalist maintains, not just to point but to say some such sentence as ''That which you see is called Il. pencil", learning a word language would be impossible; when a child is taught to say "sugar" each time it is shown a lump of sugar, it does not first have to understand the phrase ''That which, ~'ou sec is called , , .", we just attract the child's attention to the lump of sUjlar and say "sugar", perhaps once, perhaps many times, and eventually the child says "sugar" when we show it a lump of sugar. We teach the child to use the word "sugar" as a token; if it wants a lump of sugar it must first give us the word "sugar" in payment, What we teach is an exchange of things, Failure to understand this is one of the sources of the formalist's confusion; he feels that a definition must be a definition in words and accordingly he interprets the ostensive definition as defining the equivalence of the object. word and the ostensive definition sentence "That which you see is called ••. ".

Bound up with the problem of the ostensive definition we meet one of the oldest of the problems of philosophy, the problem concerning the universal word, How is it possible, one might ask, for 1\ child to learn that the word "sugar" means, is a token for, any lump of sugar and not just some one particular lump 1 Must the child first perceive what various lumps of sugar have in common (whatever that may be) before it can learn to give the word "sugar" as a token for any lump of sugar 1 Certainly, we could make a slot machine that would take

,

LANGUAGE AND EXPEBlENOE

75

only one particular coin, and reject all others, but we can also make slot machines that will take any penny piece, rejecting only coins of other values, or coins that differ in some other way from penny pieces, A child does not perceive what various lumps of sugar have in common, but fails to perceive such differences as there may be, Overlooking some differences in objects, but not overlooking others, ill the fundamental operation in language, We regard a child's ability to learn languages quickly as a mark of intelligence, yet a too subtle and clisc(lrWng child might never learn to speak his mother-tongue,

Let us examine more closely the three criteria we described above

to decide the truth or falsehood of the sentence "University College is

in University Road", We have already observed that the third criterion seems to be necessarily decisive, whereas the first and second are liable

to error. Yet could \\'(1 not conceive of the poBSibility of error also in the third criterion 1 If I walk along University Road and perceive a ' bnilding bearing the name-plate "University College Leicester", may it not be that I am deluded and suffering from an hallucination 1 Is there

in fact any criterion which is quite conclusive 1 Can we not doubt the validity of any criterion whatever 1 But if there is 110 criterion or combination of criteria that We are prepared to accept and call decisive, then the sentence "University College is in University Road" is isolated from the language system and deprived of its function, like & currency that has lost its purchasing power. One might say that in choosing the criterion, the conventions according to which the sentence is true,

or false, one is choosing the language in which the given complex of signs operates as a sentence, Accordingly, if we, say that the reference book criterion for the truth of the sentence "University College is in University Road" may be doubted, and, furthermore, that any criterion

we conceive of may be doubted, then it is only in relation to ~JOme other criterion or group of criteria that a ~cul&r criterion may be said

to be doubtful.. H there were but a smgle ,criterion for the truth of a sentence, then it would ma.ke no sense to say that t~criterion is doubtCul ; though in faot we do not ordinarily accept one critmon, we might do so: ~ itself provides for the nossibiIity of ,doubt- 88 is shown by such words as "mistake",. "£_hoodt• and '1JalluoiDation", and the corres. ponding truth criterion Of'the 'assent of the majority. 'l'Aey an sa.Ul - to 'be delude4' who tIJI 1101 8t8 wAallAe mt.ij0rily lee, dNe world ia 1101 lAB

world 01 the fMjorilY,o/,me.n,. '., . ;"',. ,.' . "

,. • ...,.....' c, - c l!f ; ..

It may seem to us,it'or We~ve grown acoustomed to believing'so, that only a madman could plaCe tmyorit.erion above the criterion of experience, yet & few hundred years agO the criterion of the reference book (particularly the works of Aristotle 'and the Bible) was accepted in preference to the criterion of experience, That is not to say that men were then blind, ignorant or foolish (except according to our present criteria of knowledge). H what Aristotle said is.tAB criterion of truth and if Aristotle said that a large object falls more swiftly than & small one, then Galileo WQ8 deluded, and however mnny times we drop object.s

76

CONSTRUCTIVE FORMALISM

of different weights, in a vacuum, and observe that they fall with the same speed, we are "tricked by our senses" (and we might account for this trick in many different ways just as subtle as the theory of relativity). It is not II. fact which is in dispute but the choice of a. mode of expression. We do not dispute the fact that different objects are seen to fall with the same speed, the question is whether we shall use a language which says that the sentence "Different objects fall with the same speed" is true because we perceive that different objects fall with the same speed, or whether we shall say that it is fal'se and the perception a delusion.

The criterion which we called the criterion of experience, the criterion according to which the sentence "University College is in University Road" is true if we see a building bearing the name-plate "University College Leicester" as we walk along University Road, might be formulated as a rule of logic permitting the derivation of the sentence "p" from the sentence ','I see p". This formulation, however, raises the problem of the nature of such expressions as "I see a red patch", "I hear a ringing noise", "I imagine a red patch", etc. We feel that these expressions have the certainty of necessary truths, yet they are neither linguistic conventions nor demonstrable sentences. The expression "I see a red patch" has the conventional sentence form but does not playa sentence role in language; it makes no sense to ask "How do I know I see a red patch 1" for there is nothing which we should call the process of finding' out that I see a red patch. Saying "I 88e a red patch" is analogous, not to saying "I have a red shirt" but to painting a red patch; in other words, saying "I see a red patch", is like saying "red patch". The distinction which we express by "I see a red patch", "I imagine a red patch" is not a distinction between two activities but is rather the difference between painting 0. vivid red patch and painting a faint one.

Seeing a red patch and saying "red", as opposed to 'defining' the word "red" by pointing to a red patch, is a cause and effect phenomenon. Imagine what mechanism we may (association, the action of light on the eye, ete.), we cannot bridge the gulf between seeing the red patch and saying "red". It may help to make this clearer if we replace "red" by painting a red patch. If a man looks at a red patch and then paints a red patch, is there a logical connection between what he saw and what he painted 1 Does it make sense to ask how he knows that the patch he is looking at and the patch he painted have the same colourI It is not whether he may be in doubt that puzzles us, but rather how he can fail to be in doubt. Suppose that on a shelf stand a number of bottles, each 'bottle bearing a label of a particular colour. I draw from a bolt a coloured token, place the token against each label in tum, reject bottle after bottle and then take down from the shelf the bottle which bears a label of the same colour as the token. How do I know that just this label has the same colour as the token 1 It cannot be necessary that I know what the colours are (called) since I may be unable to speak

,

•

LANGUAGB AND BXPBBIBNCB

77

a word language, and if we say that I must perceive that the two colours are the same then in what does this perception consist but in taking down the bottle which I did take clown 1 It might be objected that I must have seen 8omething, else why did I choose just that bottle and no other but wh~t criterion have we for deciding thisl I may suy that I acted mecha~cally, that when, I placed my token against that particular label I lust reached (or the bottle, and this may well be what happened; but that is not to say that I acted mechardcally as opposed to consciously' for it is what we ordinarily call a conscious action that I am now tempted to call mechanical. We could in fact easily construct a machine which ~lected a h?ttle bearing a label of the same colour lIS II. token placed m the machine; what puzzles us . about the analogy with the machine is tha~ we feel that when we choose a bottle we are guided by the 8ensabon of 8eeing the colour8 maleh, whereas Ilo machine cannot have sensatioDl!. Yet to say we are guided by our sensations is onlv to off~r a hypothetical mechanism to account for our actions; for "Whatever sensation w!l experienced h~w could this sensation bridge the gulf between 8eelng the label and the token and reaching for the bottle 1

.rust as saying "I see :e", where "x" is an object word, is akin to saying "x", so saying "I see »". "I imagine p", "I believe p" etc; where"p" is a sentence, is akin to saying "p". Accordingly, th~ experience-criterion for the truth of a sentence may be expressed by saying that 0. sentence "p" is derivable from the sentence "A savs p". Of course, the sentence "A says p" may itself be derived from- other sentences of the form "B says that A sap 1''' and so on, and the choice of the initial sentence in the derivation process is quite arbitrary. Remember the legal convention that "1''' is derivable from "A says l' and B says l' and 0 says 1''', but not from "A says 1'''. Propositions for which the (accepted) criterion of truth is derivabllity from sentences ,~f ,the form "A says p'~ may ~,called ~ental propositions.

t' , __ ' • _', - .~~.-i'.!; '-.' '. _" "

The reference book criterion for the truth of a proPosition may be used in two essentially different ;Ways. We. might, for example, say that a sentence "d" is true if it is' oneef the sentences in Euclid's geometry or tha.t "d" is'trUe if it is onctof the Sentences in the first book on a certain shelf.' Suppose that a. b, e, d are th&' sentences in Euclid's geometry so that the expression "the sentenC88', in Euclid's geometry" is synonymous with the class of sentences' "a, b, e, d"; then "d is one of the sentences in Euclid's geometIjt'·is derivable from "d is a member of the class a, b, e, d" which is derivable from the linguistic convention "a, b, e, d is the class whose members are a and band c and d", so that "d" is true. In this case "d" does not express an experiential proposition ; for the sentence "a, b, e, d are tho sentences in Euclid's geometry" says "the sentences a, b, e, d are called Euclid's geometry" and accordingly the sentence expresses a linguistic convention about the use of the expression "Euclid's geometry" and is not a derivative of such 0. sentence as "X saya that a, b, c, d are the sentenccs of Euclid's geometry". But

78

CONSTRUCTIVE FORMALISM

if we sav that a, b, e, d are the sentences in the first book on a. ~ shelf, this is an expericntial proposit!on .and the ~tence "€f' which is derived from it expresses an experiential proposition.

The primary diffil'ulty connected with the u~e o!. the criterion of experience may perhaps bcs~ be expressed. by asking How do 1 know that you see what I see? l\bght not two ~!gnll see~ to you ~ have the same form and different forms to me? Is this a ques~lOn. about experience, or about reality, or about language? H. we mo.mta.m that it is impossible to know whetl~er ~ou see th~ same th!ns t?at I see and if we refuse to accept any criterion acc:ordin~ ~o. which "? should say that we see the same things, then the impossibility of which we s~ is u logical liot a physical impossibility. We say that no man can bft ten tons and accept the test of men trying to lift the weight, and failing, and ad~it that of course a man may sometime in the future lift the weight. Wc do not say that any cre~ture whic~ lif~ the weight s~all not be called a man. Inability to lift the weight IS not a defining characteristic of "mcn". But when we say tbat under no oircumstances is it possible for me to know what you see, then it is with the use of the words me and you that we. are concerned ~ld. not with t~e nature of experience. As Wittgenstelll bas observed, It IS not wlud 18 soon that is in doubt but the ohoieo of a language-whether we shall usc the same word for what you see as for what I see, i.e., whether we shall admit hoth the sentences" I see an X" and "you see an X", or whether we shall allow the use of the object word X only in the sentence "I see an X" and use some other word, Y say, in the sentence "you see a Y" to expre~s what is now expressed b~ "~ou see an X". But on.ce we make this change we perceive that It IS redundant; for the difference between the sentences "I see all X", "you see an X", is already clearly shown by the opposition of "I" to "you".

~other form which the difficulties associated with this problem take may be expressed by asking "How could I ever have learned the meaning of the sentence 'I see a chair:' for how co~d anyone else kn?W wltat my sensations arc when I say I sec a chair, and not. knowmg these sensations how could they have taught me to call Just thue senaations 'I see a chair' .. (or, "how could lover learn the meani~g of 'toothache' for how could anyone else know when I have the expenenee they call '~othache' 1 "). Don't 1 mean something difl'eren~hen I say "I see a chair" from what you mean when you say this ntence 1 I know what private experience l.callseeing a chair, but I don' know ~hat experience you call seeing a chair, nor even that you have any experience at aU. Yet are the sensations I experience when I say "I see II. chair" the meaning of this sentenceI If I behaved exactly as I now behave, brought you a chair when ~ou .asked for it, w~ed across the r?om without stumbling into the furruturo! sat on !I' chair when I was tired, took my place in a row when you pointed to ~t, ~:ut yet had ~,~ 0/ the 8€1l8tltiollS wllich at present accompany my tmy'ng 1 see a chair , would

em

!

LANGUAGE AND BXPERIENOB

79

you not still say that I understood the sentence' You might think tha.t I could not behave as I do unlesa I have the sensations and experiences which 1 now have, but this is only a hypothesis. Imagine, for example, that the sensations I experience when I look at a red object and the sensations I experience when I look at a green object are inter. changed, but 1 retain my present use of the words "red" and "green" ; that is to say, I continue to use "red" and "green" in the way other people use these words, I stop my car when I see a. red light, a red light still brings the word danger to my mind, and so on, even though the colour sensation I experience now is that which I experienced before on seeing a green light. I might notice the change myself in the sense in which I might notice that yesterday drinking cold water gave me a toothache, whereas to-day it does not, but the change would not be perceived in any other way. It could not have been my special incommunicable experience that I was taught to call red or I should. now be obliged to change my language with the change in that experience.

Can a man show all the 'outward' signs of unhappiness and yet be happy 1 If he weeps and moans and presses his hand against his heart, rejects his food and speaks in a piteous voice of his grief-can this man, nevertheless, really be happy? Is it not possible, I might ask, that ouly my unhappiness is genuine and that others only simulate unhappiness ; how can I know that another really feels as I feel when I say I am unhappy t But if I choose to say this and decide that the words "I am genuinely unhappy" only make sense in my mouth, and if I now say "he simulates unhappiness", where before 1 said "he is genuinely unhappy", then to what is "he is simulating" opposed 1 Consider the antithesis 1 formerly expressed by "he is genuinely unhappy", "he is simulating unhappiness". My friend ..4, on receiving a letter telling him that his father has died, shows all the familiar signs of grief and mourning and another friend B, on receipt of the same news, gives similar evidence of great grief; 1 know, however, that..4 has always spoken affectiona.telr

of his father, imitated him in many ways, showed contentment in his company and expressed great concern over his Wness, whereas B lost no opportunity to avoid his father's society, spoke disparagingly of him ~ and Impatiently awaited the inheritance his father's death would bring him. ·1 should say that B only simulated unhappiness but, 1 should say this from what I· know of the 'context' of his grief, not because I know of some private sensation which ..4. experienced but B did not.

. If we admit the possibility of doubting any truth· criterion, that is to say if we maintain that logic leaves us free· to choose any language we please, we seem to lose the conneotion between language and reality which we thought the concept of truth established, or rather if we maintain that connection, then reality loses its uniqueness and is set free to revolve alongside the turning wheel of language. The correspondence of language and reality 'takes on again the chara.oter of an illusion, for ~he correspondence subsists only 80 long as we build the

CONSTaucrIVB roR)L\LlSM

~ in tltte image of our language. "Was that", runs a. Chinese

. ~ _ ''j.&o-tse dreaming he ~as a butterfly or this a butterfly

~~I be is Lao-tso". I See a piece. of woo.d before me. I put out ~~~ and touch it, feel the contact In my fmger tips, s~e my fingers ~~: th wood I take a saw and saw through the pIece of wood, ~':~lhe dust failing, smell the fragr!l'nce of pine, observe t~e gx:ain, ~9-~~ the Ifurfa.ce \\oith my thumb. nail, place t~e sawn-off piece III a ~~~p. of wa.ter and watch it float. . Do these thi!lgs prov~ that. wood ~ i!al substance l Might 1 not lust be watching a private emema , B>'" or dreaming! And if others talk to me and tell me they. too

~tJ.O~i..e woo(l and smell the pine and hear the rasp of the saw, mIght ~ ~. 11 this !loo be part of my dream 1 Ca.n I ever be sure that I have ~o~ ~ realiJlI ~ Yet why do I use the. ~ord real? I know per~ec~ly fOo~ ~o\\' I distinguish between a real piece of wood and an artdiClal ~el1 of wood, how 1 distinguish between sawing a piece of wood and -eiece ~ing that 1 am sawing a piece of wood. I use the familiar words, dre8-1 ".ant to give them a new and special sense, and it is essol_ltial th~t b~t. flew sense be private to me alone, that I cannot communicate thiS t.bis for if I could communicate it I should have to draw a distinction sellse~en real and something else, and this I don't want to do.

lJet~

~ real object, one might say, has a length. What is the length

's bar of iron ~ I place my standard (millimetre) measuring rod of t~de the bar of iron and find that I can set it off on the bar between e.1og a.nd 121 times. Accordingly, I say that the length of the bar is 12 een 120 and 121 millimetrcs. You object and say that I omitted bt'1i;:ke into account the temperature. I repeat the operations (ol)ta.ining to Sllme result) and write alongside my result the temperature 30° tbe tigrade recorded by my thermometer. Next day, when the therC611Jl1()ter records 35° Centigrade, I find I can set off my measuring rod JI10 ~'een 121 and' 122 times along the bar. What is the real length bft~be bar of iron? We should answer, without hesitation, that there o 0 sense to the question; for the length of the rod depends upon is P temperature, unless we choose some particular temperature as a. tbendard and call the real length of the iron bar the length associated 8~~b this standard temperature. If we observed, ~owev~, .that . what\\ the variation in temperature (between some assigned limits perhaps) e~length of the bar was always between 120 and 122 millimetrea, we t ~bt call "between 120 and. 122" the real length of the bar and this JJl Uld be in effect a change of unit, for if our standard measuring rod 'f~ 2 millimetres long we should now record the length as between

:0 and 61 units.

Suppose, however, we said "keeping the temperature constant, rely the rod must have an exact length to which the measurement

~~t,,"t(!1l 120 and 121 millimetres' is only IUl approximation". H by this 3-eDlffiU only that \I'~th ~o ma~tcr Whllt. unit we carri~ out the.m_easure. ~ nl we could imagIlle It earned out WIth a smaller' umt, then It IS true;

lilt , . .

I.

I'

I

LAlfGUAGIil AND EXPEBDmClll

but if we mean that we shall call "the length of the bar" not the result obtained by any measurement but only the 'limit' towards which the successive measurements converge, then we have deprived the word "length" of its use; for however many measuremonts we carried out there would remain the unlimited possibility of carrying out further measurements and the length of the bar is now unattainable by definition.

If we accept some other criterion than measurement for the determination of the length of the bar, for instance some calculation b~!"d upon the ~?locitie~ of the end points of the bar, then we are gt~g. ~he word. length a new use and must be prepared for the poSSIbility of a different answer to the question "What is the length of the bar of iron". Thus when the relativitist says "That which you thought to be a bar of fixed length is really shorter when it lies in some positions tha.n others, only it is impossible for you to detect this difference • as your .measuring rod also changes its length and your physical organs change m such a way that you are not aware either of this change or of the changes in the lengths of objects", he is using the words "length" and "change. of length" in two different ways. Has the relativist shown that the ordinary man's use of "length" is wrong 1 Is the relatiVitist's answer to the question "what is the length of the iron bar 1" the true ~nswer an? the ~ns~er whic~ the measuring rod gives false 1 Was it lust a foolish prejudice, a. habit of thought, to think that we can measure lengths with a measuring rod, a prejudice from which the relutivitist sets us free 1 The reiativitist mistakenly expresses a new convention about the use of the word "length" in the form of a discovery about the nature of the world. It is not a new fact that the rolativitist records but a change ~f la;nguage. ,If we mean by ':length", as we ordinarily mean, thatlwhich IS dete~med by a measunng rod, then it makes no sense to talk of a change 1D the length of the measuring rod for the measuring rod (the standard of length) has no length. By ~hanging the meaning of the word "length" we appear to make the strange disoov~ry that that whioh we thought to be the instrument of measurement m some way has now been shown to measure itself.

The effect of radium on living tissues isa new faot, the Copernican aatronomy iii new language. In the history. of' the human race the ~overy of a new la~age ma.:y be of' greater importance than any di&oovery .of the "experm1ental 8C1ences, but ~or the philosopher a new ~gu,:,g~ 1S only a new l~gua.ge, yaiuable for purposes of oomparison, but heving no greater olaun on his attention than any other language The Copernican or ~he relativitist use of the word "motion" is just one of the U868 of the word. There is a common belief that the Copernican use of the term "motion" was forced upon us by the overwhelInin;g evidence of facts, ~ha~ in some sense we were driven to 'admit' the motion of the earth whioh ignoraaee had hidden from us· whereas what Copernicus discovered was no new fact of the Unive~ but tho.t tho paths of the heavenly bodies were more simply described if by motion wo meant motion round the SUD and not motion round the earth.

. i

81

.'.

82

CONSTRUOTIVB POlUULISM

Parallel to the question "what is the nature of reality 1" we might ask "what is the nature of experimental science?" or "what validity has the scientific method t" We photograph the motion of a planet across the sky during some interval of time; the photograph presents the motion of the planet as a strip oflight across a darkened background. We wish to prolong this strip of light, but of the unlimited possibilities of prolongation, there is no one particular prolongation that is logically necessary. If two photographs are taken, one for six months and the other for this and a further six months, can we, without looking at the second photograph fill in on the first 'unfinished' photograph the path that the second photograph is recording? That is the problem of the experimental sciences. Theoretical physics singles out some particular one of the unlimited possibilities of prolongation as the correct pro. longatiou, the 'reasons' given for this choice constituting the so called theorv of the science, and this prolongation is compared with the second photograph. If till' prolongation is a good fit, other experiments of a similar character (choosing IL prolongation according to the theory) aye carried out, and if no marked discrepancy between the 'tbeoreeie' prolongation and the second photograph is observed, the theory is said to be \n,1I founded and provisionally adopted.

If we look more eloselv at the 'reasons' given for a particular guess (theoretic prolongation) we"find a mixture of observational and linguistic sentences, Some reasons, like "the velocity of light is independent of the velocity of till' source of light" are lingu~ic statem~nts, ex~

in the form of observational statements, The velocity of light 18 independent of the velocity of the source of light" tells us how the expression "the velocity of light" is going to be used. Such a statcm~ lays down the way we ha ve decided to talk about our experiments, serves to choose between conflicting evidence. Just as a principle of law like "a man is innocent until he is proved guilty" tells us, not something about the character of human beings (that 0. man cannot have done wrong unless he is proved to have done so), but something of the way we are going to use the words "innocent" and "guilty" and serves

to guide our treatment of untried prisoners. Moreover, even as we are prepared to change a principle of law-i.e., to adopt 0. new legallangua.ge -we are prepared to change the reasons we give for our gueesee (apart from changing the guesses theD18Clv68).

One might say that theoretical ph~C8 is 8; ~ between lan~ge and experience, the so-called theory formmg a dictionary for translatmg observational sentences into mathematical equations and vice- versa. Consider the statement "Forces are added by the parallelogram law". Is this an observational statement? Certainly we can use this rule to build hridges and certainly bridges built according to it have 'stood the tellt of time'; but might not a bridge built to- morrow according to the rille, fall down when tested? Contrast the statement with "vectors arc added by the parallelogram law" which is the mathematical

1

I

~.f

@ I

LANOUAOB AND EXPBRJRNCt:

tkflnition oJ addititm for vectors. We can imagine a time when IIIt'n perceived how forces are added, and stated that forces add by the parallelogram law just as one might observe that twenty people pess one's window every day. At sueh a time the parallelogram law is II, statement of fact, not a generalisation from particular instances. not an induction, but a statement of (act. At some subsequent time the statement of the parallelogram law ceased to be used as a statement of fact and became a linguistic convention. Though certain experiences in the future may lead us to abandon the use of the word "force" given by this convention, we might retain the conventi?~ in spite, of a~y experience whatever. In the same way the prOPOSItIOns of arIthmetIc 'evolved' from observational truths to linguistio conventions. It might happen in the future that when we placed books on a shelf or apples in a bag or men in a room and counted them one by one we never found a total beyond five; that is to say, after counting the books and finding there are five on the shelf, and then placing further books on the shelf and oounting again, we still obtain the answer five, and similarly after counting and recounting apples and men and so on. In such a world we might lose our interest In our common arithmetic (which of course is not invalidated by such experiences any more than it is validated by our present experiences) and adopt instead an arithmetic in which the sum of five· and one is five. on the other hand, we might still retain our common arithmetic and say that there are really six or seven, etc., books on the shelf but some have coalesced 80 that we seem to have only five. Compare thia with the position in which present day physies finds itself. Shall we say there are material particles which behave like waves,. or waves which behave like material particles !

The microaoope shows a drop of water as a universe of active and brightly coloured creatures, of forma unfamiliar to the naked eye. Which is the real drop of wa.ter, that whioh the unaided eye shows UII or this myriad of tiny creatures 9 Does the microscope enlarge our knowledge of reality or does it destroy the possibility or beUef in any world of the senses! . If we answer that there are two worlds. the world of our (ordinary) vision and the microscope world, which of these is the real world 1 And perhaps the creatures or the microscope would likewise under some more powerful microscope reveal themselves just as organisa· tions of other (smaller) creatures. Related to these questions is the metaphysical problem "are not human beings just parts of some greater organism of which we can have no conception Y" and the alUed problema Buch as "have human beings an independent existence or do they just serve some higher organisms, call them nations or civilisations, even as we ~ve that the creatures which dwell in our blood live but to serve us l' Another form in which we C&Il expreas these questions is "are microbes real I" or "is there really a microbe in this drop of water , .. It is important to remember that if we say microbes can only be seen through a microacope. they are too small to be seen with the naked eye, we are making a hypothesis, not stating 8 lOgical necesaity of vision.

COXSTRUCTJVB FORMAt.IRM

It is a hypothesis that what we see (when we look through a microscope) depends upon till' microscope. We can imagine a world in which on some days our visual experiences are what they are now when 'we look around with the unaided eyo', and on other days are what we are now familiar with on looking through a microscope. What we have before our mind's eye when we ask if there are really microbes in & drop of water is an image of u clear drop of water beside an image of the same drop speckled with microbes; but looking at a drop of water and then look ing at the same drop through a microscope is more like seeing & dear drop of water transform into, not a speckled drop, but & vessel (If water filled with tiny creatures. That this vessel of water and the drop of water arc the -same thing is also only an hypthesis; that is to say the form of expression "this vessel of water you BOO is rea.lly the drop of water seen through a microscope" expresses a purely arbitrary convention, one which \YO could abandon without denying any fact. Instead of saying that the microscope shows us the microbes (hidden) in the water we could say that the microscope Ira1UJjorms the drop of water into an expanse of witter filled with swimming creatures i of course, this change of expression (remember it is not the facts we change) will entail other changes of expression. Instead of saying that a man has typhoid fever because the microscope reveals the presence of the typhoid bacillus in his blood, we should say that the microscope transformation of his blood contains the typhoid bacillus. And to the question "surely there must be something already in the blood (before the microscope transformed it), else why is the man ill 1" the answer is that to say the man is ill because of a microbe in his blood is only to make a hypothesis-a linguistic convention-which w.e can retain or abandon as we please. If the microscope transformatIon of my blood is found to contain the typhoid bacillus, I should unhesitatingly accept the treatment which I had observed in previous cases to be followed by & return to health, not because I had found out that "such and such a process cures typhoid" is more than a linguistic convention, but because

that is how human beings behave. .

It is sometimes said that just as there are microbes too small to be seen by the unaided eye, 80 too there are electrons too small to be seen oven by a microscope. This is rather like saying that something is seen which is not seen. The sentence "I see an electron" baa not, that relation to "I see a microbe" which this latter has to the sentence

"I see a drop of water". If we ask a physicist to show us an electron ,.,..he shows us pictures of white (or coloured) lines on & black background /' and calls these pictures electron tracks; the physicist will say that he cannot show us single electrons, only streams of electrons, and will in

fact show us not streams of anything, but just streams. It may be

true that nowhere in the world can we find a single locust, that loousts

arc found only in swarms, but a swarm of locusts is specifioally a swarm

of loeusu. The physicist tells us that he can distinguish one electron

from another and in support of this shows US pictures whieh he calla

;,1

"

Sa

"streams of a-particles", "streams of ,a. particles" and so on, hnt it is not tho contents of the streams by which these pictures differ, Again we are tempted to ask the familiar questions "are electrons the framework of reality, is the world of electrons the real world behind the illusory world of the senses 1" "Is the real chair that which we see and handle, buy and soil, cover with tapestry or chop up for fire\vood, or is the real chair an organisation of electrons 1" Is it a mark of ignorance to believe there are no electrons; does not the electric light prove the existence of electrons 1 Are there facts which we eannot describe without the use of the word eleetron (or a syntactically equivalent word) 1 If we say that the supposition that there arc electrons explains the phenomenal world and enables us to predict the future so accurately, that it is no longer possible to doubt the truth of the supposition, do we mean that because of this success (in foretelling the results of experiment, etc.) we have more confidence than before that we shall' one day 'isolate an electron' 1 or do we only mean that we lose any temptation we may have had to abandon electron-language 1 After reading in the paper one morning that a certain man has taken his life and that beside his body a letter was found saying he was going to shoot himself because he had lost all his money, I might write a play in which a man immigrates to this country, works hard, makes a fortune, loses it and shoots himself, and then find that I have described correctly in every detail the life of the man I read about in the paper. Suppose that I repeated the experiment many times, that each time I read of a suicide I write a play and then find I have correctly told the life story of the dead man; If I now say that I believe I have second-sight, will my suecess have proved the truth of my supposition 1 or course, if we agree that "I have second sight" is just a form of expression for describing what I did, then to say I have second sight is just to say that I did what I did; but conld we not dispute as to whether there is suoh a thing as second sight' And if someone says that nothing ~ convince him of the reality of second sight, is he denying the possibility of 0. certain experience, or just refusing to use a certain form of expression' On the one hand I certainly did not lmDw the lifo story of .. the man who took his life, yet I wrote that story correctly in every detail. The facts are not in dispnte. I did not know the story but what I wrote turned out to be written just as if I had known the story. (The scientist does not know when there will be an eclipse, but what he says turns out to be true just as if he had known). Shall we give a description of these facts which stresses the similarity of my writing to actual knowledge or shall we give a. description which minimises this similarity; that is the choice we make when we accept or reject the expression "second-sight". If we grant the man who says he believes in survival after death any experience that he may desire, even a body walking the earth alike in all respects to that of one who bas died, having a.1I the memories of the dead man, speaking with the same voice, behaving in all respect. like the dead man formerly behaved, must we say that the survivalist's belief' haa been proved true! Is there a fa~

86

CONSTRUCTIVE FORMALIRM

in dispute if we argue whether we shall say the dead man has returned to earth or whether we shall say that the man who died has nob returned but another exactly like him now walks in his place ,

The word "reality" has whatever significance we choose to give it.

To say that there is no correspondenoe of language with reaHty is to make the decision not to use the word "reality" and to express this decision with the air of milking a new discovery about the nature of the world. And if we ask whether there is not some Reality behind the phenomenal world, a reality of which the world of our II8Il8eS is but a shadow, we must answer that if we choose to change our language, and talk, not of tables and chairs, but of shadows. of-tables and shadowsof-chairs and call what we now term "shadow" instead "shadow of a shadow", then we perceive that "shadcw-of-a.chair" is now 'I. complex term, that for the word "chair" itself we have given no use and therefore "shadow-of-a-chair" is but a redundant form of expression for "chair".

(

• I

87

BIBLIOGRAPHY.

j

I

\ I BROUW~R, L. E. ,1:. Dber die ~eutung des Satzes vom ausgeechloesenen 1\ Drltten, J. fur Math., 1M, 1924.

Zur Begriindung der intuitioJiiSiischen Mathematik, JIath. Ann., 93, 95; 1925, and 96, ~

CANTOR, G. Beitrage zur Begriindung der transfiniten Mengenlehre, . Math. Ann., 46 and 49.

CARNAl', R. Logical Syntax of Language, London, 1937.

CHURCH, A. A set of Postulates for the Foundations of LoSic, Annal"

oj Math., Vola. 33' and 34. .

II Ctnmy, H. B.Formalisation of recursive arithmetio, American J. Math., Vol. 63, 1941, pp. 263-282.

DEDBKIND, R. Was sind und was sollen die Zablen, Brunawic1c, 1888. tStetigkeit und irrationale Zahlen, BrunBlDicle, 1892.

FREOE, G. Die Grundlagen der Arithmetik, Breslau; 1884.

Grundgesetze der Arithmetik, Jena, Vol. I, 1893; Vol. II, 1903.

GO'lZEN, G. Die Wiedel'spruchsfreiheit. dar reman Zablentheorie, MalA. Ann., 112, 1936.

GODBL, K. tiber formal :unentscheidbare SAtze, Manat8Iu.fte ,.

Math. U. PhY8., 38, 1931..

GOODSTlCIN, R. L. Function theory in an axiom.free equation calculus, Proe. LonOOn Math. Soc", Ser. 2, Vol. 48. 1946, pp. 401-34 .

Mean value theorems in recursive function theory. Part 1.

Differential mean value theorems, j·bid., Vol. 52. 1950, pp. 81-106.

The formal structure of a denumerable system, Trans. American Math. Soc .• Vol. 68, 2, 1950. pp. 174-182.

The Gauss test for relative convergence, Amt'rican J. Math., Vol. LXXII, 2, 1950, pp, 217-228.

On the restricted ordinal theorem, J. Symbolic Logic, Vol. 9, 2, . 1944, pp. 33-41.

Transfinite ordinals in recursive number theory, ibid., Vol. 12, 4, 1947, pp. 123-129.

The strong convergence of the exponential function, J. London Math. Soc., Vol. 22, 1947, pp. 200-205.

Mathematical Systems, Mind, Vol. XLVIII, N.S. 189, 1939, pp.58-73.

. '

88

BIBLIOGRAPHY

HILBERT, D., and BER!'<AYS, P. Grundlagen der lll\thematik, Vol. 1,Berlin, 1934; Vol. II, Berlin, 1939.

KI.EF.!'<E. S. C. General recursive functions of natural numbers,

Jlath. AmI .• \'01. 112. 19:16, pp. 727-742.

Recursive predicates and quantifiers, Tram. American Math. 8oc ..

Vol. 53, 1. 1943. pp. 41-73.

NELSO~, DAVID. Recursive functions and intuitionistic number theory, ibid .• Vol. 61, 2, 1947, pp. 307-368.,

PETER, R. Begriff der rekursiven funotionen, Matl,. Ann., 110, 1934-35.

Uber die mehrfaehe rekursion, ibid. 113, 1936-37.

Zusammenhaug der mehrfaehen und trunsflniten rekursion, J. Symbolic Logic. Vol. 15. 4. 1950. pp. 248-272.

ROSSER, BARKI.EY. l'~xtellsions of some theorems of GOtlel and Church, J. Symbolic Logic. Yol. 1. 1936, pp. 87-91.

RUSSELL, B. The Principles of llathematics, 2nd Ed., London, 1937. -(with WHITEHEAD, A. ~.). Principia Mathematiea, Oambridge.

Yol. I, 1910, 1925; Vol. II, 1912, 1927; Vol. III, 1913, 1927.

I' SItOLEM, Tn. Begriindung der elementaren arithmetik durch die

,I rekurrierende denkweise, Videmk. 81:r., Kristiania, 1923. No.6.

WITTGENSTEIN, L. Tractatus Logico Phllosopbicua, London, 1922. \ \

\

Abacus, 68.

Calculus : equation, 27. fractional, 33. integer, 36. rational, 37.

Cantor:

ternary set, 62. Aleph numbers, 54.

Cauchy's theorem, 62. Cells:

interior, 61. exterior, 61.

Chains, closed, 26. Class, 19. . Classification index, 19. Code, 50.

Commentary, 18. Copernicus, 81.

Constructivism, 11. 42, 43, 49. ~ontinuity, 35.

-- relative, 35, 62 ••

i @

I

89

INDEX

Contradirtion, derivation from;

44,49.

Convergenoo, :M. -- relative, 34.

Counting, definition of, 20, 65, 66,70.

Curve. 59.

Dedekind's theorem, 35. Definition :

reeurstve, 24.

explicit. 25.

osWnaive, 74.

, Derivative. relative, 41, 62.

Divergence, 43.

Excluded . Uuddle axiom, 29. Existence, 12, 49. Ex.istential operator:

bounded,' 30. unbounded, 31.

\\0

'I!'t.rmuf ~ In .. 4l t heorem, 44. ].'initism. 42. i2. ].'uromlism. 11. :!l. ';:!. j:t }'n·j.,rt>. G .• )!t.

Function. HI. :!:1.

Oentzl'n. G.. 51. Gtilll'l. K.. 45.

Hallucinatiull. 7ii.

Inference. 18.

Infinitude uf primes, 16. Irrationals. 35.

Integral:

real, 63. complex. 63.

Interpretation. 44. 45.

Language: content of. i:? picture. 73. legal. 82.

Linkage terms. 18. Levels. boundnry. 61. Logical conllecth·cs. 18. __ elimination of. 29.

llenn vnlue theorem, 41. lh.'asurenltmt. 80. )IinimUID operator, 31. ..

lSDBX

~\lmber :

definition of. 19, 6K. eardinal, 70.

ordinal. 70. transfiuite, 55.

Numerical term, 48.

Operators, bounded. 30.

Paradox:

Zono's. 13. Russell's, 45.

Parallelogram law. 82. Path:

simple. 59. clockwise, 61.

anti clockwise, 61.

Primes, 16.

Principia Mathematica, 12, 44. Proof, definition of, 27. Propositional function, 30. -- form, 24.

Reality, nature of, 79, 86. Recursion:

simple, 24.

multiple, 25.

Reduced sequence, 34.

Reductio ad absurdum, 13, 43, 49. Related classes, 19.

Relative continuity:

definition of, 35. fundamental theorem on, 39.

Relativ« convergence, definition of,34.

-- derivative, 41. Relativity. 81.

Science, experimental, 82. Signs, number, 20, 21, 65, 67.

Tally, 20, 66.

Tertium non datur, 29. Transfinite:

induction. 51. ordinals, 53, 55, 56.

INDEX

91

Transformation rules, 19. -- of number tligng. 66. Truth, erik-ria for. 72. 75.

Universal operator: bounded. :JO. unbounded, 31.

Variable:

numeral. 21. 22. qualified. 22. majorant, 56.

Wittgenatein. L., 50, 78.

Zono, 13.