ARI S T O T L E'S

S Y LL O G I S T I C
FROM THE STANDPOINT
MODERN FORMAL

OF

LOGIC

BY

J A N L U KASIEWIC Z

SECOND

E D I TI O N

ENLARGED

OXFORD
A T T HE C L A R E N D O N P R E S S

Oxford UniuersityPress,AmenHouse,LondonE.C.4
OLASOOW
BOTBAY
CAPE

NEW

YORK

CALCUTA
TOWN

TORONTO
MADRAS

EALTSBURY
KUALA

MELBOURNE
KARACHI

NAIROBI

LUMPUR

HONG

WELLINOTON

HBORE
IBADAN

DACCA

PR EF AC E T O
T HE SEC ON D ED IT ION

ACCRA

(ONG

'Isr first edition of this book did not contain an exposition

o[ Aristotle's modal syllogistic. I was not able to examine
Aristotle's ideas of necessityand possibility from the standpoint
of the known systemsof modal logic, as none of them was in my
opinion correct. In order to master this difficult subject I had
lo construct for myself a systemof modal logic. The first outlines
of this I developed in connexion with Aristotle's ideas in my
lcr;turesdeliveredin the Royal Irish Academy during I95r and
ilr the Queen's University of Belfast in 1952. The complete
ilystem I published in The Journal of ComputingSlstems,rg5g.
My system of modal logic is different from any other such
iystem, and from its standpoint I was able to explain the diffir:rrltiesand correct the errors of the Aristotelian modal syllogistic.
My book on Aistotle's Syllogistichas met with a favourable
rcception to my knowledge in more than thirty articles and
reviewspublished over the world in English, French, Gerrnan,
| fcbrew, Italian, and Spanish. I have ever since been anxious
lirr an opportunity to discusssomeof the critical remarksof my
reviewers,but in the preseritissueit has been possibleonly to
nrkl the chapterson modal logic (as the text of the first edition
wrn alrcady printed) . I am most grateful to the Clarendon Press
firr thc chanceto do so.
J.L.
t) t,l t LIN

11tt.Jwurg55
rl

FIRST EDITION
sEcoND
REPRINTED

EDrrroN

I95I

(ENLARcED)

LITHOGRAPHICALLY

By LATTMER, TREND

& co.,

PUBLISH ER 'S N OT E

t957

IN GREAT BRITAIN
LTD., wHrrsrABLE

FROM SHEETS OF THE SECOND EDrrrON

r 9 5 E, 1 9 6 3

I'norrrsoR Jan Lukasiewiczdied in Dublin on the l3th of Feband thus could not seehis book through the Press.
r unryi r r;5r{i,
'l'hir wls <loncby his former pupil, Dr. CzeslawLejewski,who
renrl tlre prrxrlsof the addedchaptersand extendedthe index.

PR EF AC E T O
T HE F IR ST ED IT ION
ftvJune r g3g I read a paper at the PolishAcademy of Sciences
in Cracow on Aristotle's syllogistic.A summary of this paper
was printed in the same year, but could not be published
llccauseof the war. It appeared after the war, but was dated
During the summerof r939 I prepared,in Polish,a more
"g3g'. monograph on the same subject, and I had already
rlctailed
r<:ceivedthe proof of its first part when in September the
printer'sofficewascompletelydestroyedby bombing and everything was lost. At the same time my whole library together
with my manuscriptswas bombed and burnt. It was impossible
to continue the work during the war.
Not till ten years later did I get a fresh opportunity to take
up my investigationsinto Aristotle's syllogistic, this time in
l)ublin, where since 1946 I have been lecturing on mathernatical logic at the Royal Irish Academy. At the invitation of
llniversity College, Dublin, I gave ten lectures on Aristotle's
syllogisticin rg4g, and the presentwork is the result of those
lcr;turcs.
'l'his work is confined to tJle non-modal or 'assertoric' syllogisms,sincethe theory of theseis the most important part of
tlrr: Aristotelian logic. A systernaticexpositionof this theory is
corrtainedin chapters I, 2, and 4-7 of Book I of the Prior
Awil2tics.These chapters in Th. Waitz's edition-now more
lhitn a century old-are the main source of my exposition.
I rr:grct that I could not use the new text of the Prior Analttics
c<lilcd with an introduction and a commentary by Sir David
Itoss;rnd publishedin 1949,sincethe historicalpart of my work
wls llrcady finished when this edition appeared.I could only
(r,r'l('ct my quotations from Aristotle by the text of Sir David
Itrrrs. lrr tlrc English version of the Greek texts of the Anafutics
| ;rtllrt'rc'd as far as possible to the Oxford translation of
AlirlotlrJs works. Besidesthe text of the Prior Anal2ticsI took
irrlo corrsirlt:ration
the ancient commentators,especiallyAlexrurrl('r'.
I rnaymentionherethat I owe to an anonymousancient

Viii

PR EFAC E

PREF ACE

commentator the solution of historical problems connected

with the alleged invention of the fourth syllogistical figure
by Galen.
The presentwork consistsofan historicalpart, ChaptersI-III,
and a systematicpart, Chapters IV and V. In the historical
part I have tried to expound the Aristotelian doctrines following the texts as closelyas possible,but everywhereI have been
anxious to explain them from the standpoint of modern formal
logic. In my opinion there does not exist today a trustworthy exposition of the Aristotelian syllogistic. Until now
all expositions have been written not by logicians but by
philosophers or philologists who either, like Prantl, could
not know or, like Maier, did not know modern formal logic.
All these expositionsare in my opinion wrong. I could not
find, for instance,a single author who realized that there is a
fundamental differencebetweenthe Aristotelian and the traditional syllogism.It seemsto me thereforethat my own exposition is entirely new. In the systematicpart I have tried to
explain some theoriesof modern formal logic necessaryto an
understanding of Aristotle's syllogistic, and have tried to complete this syllogistic on the lines laid down by Aristotle himself. I was again anxious to be as clear as possible,so that my
expositioncould be understoodby scholarsnot trained in symbolic or mathematical thinking. I hope thereforethat this part
of my work may be used as an introduction to modern formal
logii. The most important new resultsin this part I consider
pupil J. Slupecki, and
to be the proof of decision,given by
-y
the idea of rejectionintroduced by Aristotle
and applied by *yself to the theory of deduction.
I am sincerely grateful to the Royal Irish Academy, which,
by giving me a positionin Dublin, has enabled me to write this
book, and to University College,Dublin, for its kind invitation
to deliver lectures on Aristotle's logic. I am grateful to the
Professors
of University College,Dublin, FatherA. Gwynn, S.J.,
and MonsignorJ. Shine,who were kind enoughto lend me the
necessarybooks.I owe a debt to Sir David Ross,who read my
typescript and made some suggestionsI was glad to accept.
My special thanks are due to the late Father A. Little, S.J.,
who,,although already dangerouslyill, willingly corrected the
English of the first chapter, to Victor Meally in Dublin, and in

rx

particular to David Rees of Bangor, who read and corrected

the English of the whole work. I am also deeply indebted to the
officials of the Clarendon Pressfor their zeal and courtesy in
preparing my typescript for printing. The section on Galen is
dedicated to my friend ProfessorHeinrich Scholz of Mi.inster,
Westphalia, who was of great assistanceto myself and to my
wife during the war, and especiallyduring our stay in Mtinster
in tg44. The whole work I dedicateto my belovedwife, Regina
I-ukasiewicz nie Barwiirska, who has sacrificed herself that I
might live and work. Without her incessantcare during the war,
and without her continual encouragementand help in the lonelinessof our exile after it, I could never have brought the book
to an end.

J.r,.

D U BLIN

7 May rg5o

CO N T EN T S
C H A P TE R
E LE ME N T S

OF TH E

S Y S TE M

I r. Tlu haeform d tlu Arisntelian syllogi.sm .I

.3
and terms
$ z. Premisses
E g. Wb singular tnms wereornitted$t Aristotle .5
.7
$ 4. Variables
. .to
$ 5. Slilogistic necessit2
.1 2
6.
What
isformal
logb?
$
p 7. What i.sfornalism?
.1 5
C H A P TE R
TH E S E S

OF TH E

II
S Y S TE M

$ B. Tluns andrulcsof iqfnena.

$ g. Thes2llogistic
figuru
andminortems
$ rc. Tlu major,middle,
$ rr. Thchistoryof an enor
$ n, Theorfurof thepremtsscs
$ tg. Enorsof somemodcritcommentators
$ 14. ThefourGalenianfgures
C H A P TE R
THE

.2 ( )

.23
.28
.30
.32
.34

.38

III

S Y S TE M

syllogisms
$ ry. Perfectandimperfect
16.
The
logic
of
terms
and
tlu logicof propositiotu
$
$ 17. Theproofsfu conuersion
I 18. Thcproofsby reductioad impossibile
$ tg. TheprooJs
fu ecthesis
Thc
rcjcctedforms
zo,
$
problems
unsolaed
$ zr. Some

43
47
5r
54
59
67
72

CONT ENT S

Xii

CHAPT ER
ARISToTLE'S

SYST EM

C O N TE N TS
IV

IN SYM BOL IC

C H A P TE R
F OR M

$ zz. Explanationof thesymbolism

$ zg. Theorl of deduction
$ 24. fuantifers
of thesyllogistic
$ 25. Fundamentals
thcses
of
s2llogistic
Deduction
26.
$
expressions
$ 27. Axiomsand rulesfor rejected
rules
axioms
and
of
our
Insfficienc2
zB.
I

CHAPT ER
THE

PROBL EM

TH E

79
B3
BB
9o

9+
9B

roo
r03
ro6
III
I20

n6
r3 0

CHAPTER VI
M oDeL

Loc lc

oF PROPOSIT ION S

r33
$ 96. Introduction
r3 +
their
interrelations
Modalfunctioru
and
$ 97.
r3 5
$ 38. Bcsramodallogic
I38
gg.
Laws
extewionality
of
$
| 4o. Aristotle'sproof of theM-law of extensionality r40
r43
of propositions
connexions
$ 4r. Necessaqy
r46
$ 42. 'Matnial' or'strict' implication?
propositions
T48
Anal2tic
| 49.
paradox
I
5I
E U. Ao Aristotelian
r54
in Aristotle
$ 45. Contingenc2

V II

OF MOD A L

LOGIC

$ +6. Thematrixrnethod
$+2. TIu C-Nfi-system
$ +8.6-Defuitions
of modallogic
$ 4g. Thefour-ualuedsltstern
and
tlu
Neussity
$ so.
four-ualuedsystemof modal
logit .

| 5r. Twin possibilities.

temof modal
| 52. Contingerc)andthefour-aalued s.2ts
logic
.
problems
.
.
$ 59, Some
further

OF DECISION

expressions
$ zg. Thc numberof undecidable
go.
rule
of
rejection
Stupuki's
$
$ 3r, Defuctiueequiualence
expressions
$ 32. Reductionto clementar.y
of thc syllogistic
$ gg. Elemcntarl expressions
of thes2llogistie
arithmetical
interpretation
Ao
I Z+.
$ 95. Conclusion

A RIsro rLE's

S Y S TE M

xlu

I58
r6o
r63
r66
r69
r72
17+
r7B

C H A PTER VIII
A R r s r o r L E 's

M o D AL

sYL L o GISTIc

premisses
r8r
$ 54. Moodswith two apodeictic
$ 55. Moods with orc apodeicticand one assertoric
premiss
rB3
Rejuted
moods
with
one
apodeictic
and
one
$ 56.
premiss
assertoric
186
rBB
$ 57. Solutionof thecontrouers2
possible
premisses
Moods
with
rgr
$ 58.
propositi.ons ry4
of contingent
$ 59. Laws of conuersion
6o.
Rectfication
Aristotle's
of
mistakes
rg8
$
premisses
2or
$ 6r. Moodswith contingent
2o5
$ 62. Philosophicalimplicationsof modallogic
IN D E X

2Og

ER R A TA

Page &4,line 35 : for h read b

Page go, last line: exchangea and c
Page rzo, line zg: for *roox*6r. clb read *6rx
*too. blc
Page te9, line z4: for lcd read NIcd

ANCIE N T TE X TS A N D
COMME N TA R IE S
AristotclesGracce,qc recensione fmmanuelis Bekkeri, vol. i, Berolini,
r83I.
Arbtohlis Orga.rcnGruce, ed. Th. Wartz, vol. i, Lipsiae, 1844; vol. ii,
Lipsiae, 1846.
Aristotle'sPrior @d PostcriorArulytics. A Revised Text with fntroduction and Comrnentary by W. D. Ross, Oxford, 1949.
Ahxandi in Arisbtalis Adyticorum Priorum Librum I Commentarium,
ed. M. Wallies, Berolini, 1883.
immonii in Aristotalis Atul2ticorum Priorum Librum I Commentaium,
cd. M. Wallies, Berolini, 1899.
Iounis Philoponi in Aristotelis Analltica Priora Commentari,a,id. M.
Wallies, Berolini, rgo5.
The texts of Aristotle are quoted according to Bekker's edition.

zAn. pr. i, +, z1bgTmaru: Anal2tica

friora, Book f, chapter

pegc 25, column b, line 37. The tacts of the commentators are
according to the above editions of the Academy of Be.rlin.
z Alcxanda roo. rr mearurrpage roo, line rr.

CHAPTER

E L E M E NT S

OF T H E

SYST EM

$ r. The trueform of theAistotelian s2llogism

IN three recently published philosophical works the following is
given as an exampleof the Aristotelian syllogism:r
(t) All men are mortal,
Socratesis a man,
therefore
Socratesis mortal.
This example seemsto be very old. With a slight modificationis quoted already by Sextus
'animal' instead of 'mortal'-it
Empiricus as a 'Peripatetic' syllogism.2But a Peripatetic syllogism need not be an Aristotelian one. As a matter of fact the
example given above differs in two logically important points
from the Aristotelian syllogism.
First, the premiss 'Socratesis a man' is a singular proposition,
as its subject 'Socrates'is a singular term. Now Aristotle doesnot
introduce singular terms or premissesinto his system.The following syllogism would therefore be more Aristotelian:
(z) All men are mortal,
AII Greeksare men,
therpfore
All Greeksare mortal.3
This syllogism,however,is still not Aristotelian. It is an inference,
where from two premissesacceptedas true, 'All men are mortal'
and 'All Greeks are men', is drawn the conclusion 'All Greeks
are mortal'. The characteristic sign of an inference is the word
I See Ernst Kapp, GrcckFoundationsof Traditional Logic, New York (rg4z)' p. I t ;
Irrederick Copleston, 5.J., A History of Philosoph2,vol. iz Greeceand Rome (t946),
London (r946), p. ztB.
p. r77; Bertrand Russell, History of Wcstnn Philosop,fr2,
2 Sextus Empiricus, Hyp. Pyrrh. ii. 164 Euxyiqs
d.v9panos, n6s dl|potos (Qov,
8axpd.n1s dpa lQov. A few lines earlier Sextus says that he will speak about the
s<r-called categorical syllogisms, nepl rGw xatrlyopmiw xoAoupivuv ouAAoynp6l,
used chiefly by the Peripatetics, ols Tpdvrat pdAnra oi d.nd to0 llepndrou. See also
ibid. ii. r96, where the same syllogism is cited with the premisses transposed.
3 B. Russell, op. cit., p. zrg, gives form (z) immediately after form (I)' adding
in brackets the remark: 'Aristotle does not distinguish between tiese two formsl
this, as we shall see later, is a mistake.' Russell is right when he says that these two
frrrms must be distinguished, but his criticism should not be applied to Aristotle.

E LB ME N T S OF T H E S YS T EM

$r

'therefore' (dpa). Now, and this is the second difference, no syllogism is formulated by Aristotle primarily as an inference, but they
are all implications having the conjunction of the premisses as the
antecedent and the conclusion as the consequent.A true example
of an Aristotelian syllogism would be, therefore, the following
implication:
(S)
If all men are mortal
and all Greeksare men,
then all Greeksare mortal.
This implication is but a modern example oi an Aristotelian
syllogism and does not exist in the works of Aristotle. It would be
better, of course, to have as an example a syllogism given by
Aristotle himself. Unfortunately no syllogism with concrete terms
is to be found in the Prior Anafittics. But there are some passages
in the Posterior Anal2tics from which a few examples of such
syllogisms may be drawn. The simplest of them is this:
(+)

Ifall broadJeavedplants are deciduous

and all vines are broadJeavedplants,
then all vines are deciduous.t

All these syllogisms, whether Aristotelian or not, are only

examples of some logical forms, but do not belong to logic, because they contain terms not belonging to logic, such as 'man' or
'vine'. Logic is not a science about men or plants, it is simply
applicable to these objectsjust as to any others. In order to get a
syllpgism within the sphere of pure logic, we must remove from
the syllogism what may be called its matter, preserving only its
form. This was done by Aristotle, who introduced letters instead
of concrete subjects and predicates. Putting in (4) the letter ,4.
for 'deciduous', the letter ,B for 'broad-leaved plant', the letter C
for 'vine', and using, as Aristotle does, all these terms in the
singular, we get the followine syllogistic form:
(S)
If all B is ,4
and all C is B,
then all C is A.
I An. post.ii. I6,
98b5-ro Eota ydp rd $vMoppoetvd{' oI A, rd 6t n\arJ$uilov 2S'
o8 B, dpndtos Dt d{' o0 f. ei 6'}1rQ B indpTet rd A (n6v ydp nAail$vAAov QvU\oppoei),
rQ 3i I izdpyet rd B (n6oa yd.p dpnelos z)czri{utr)os), tQ I tndpTer rd A, xai r6oa
d.pndtos $v\Aoppoei. From this somewhat carelessly written passage-after tQ B, tQ
3i f, and tQ T, navri ought to be inserted-we get the following syllogism in concrete terms: ei ndv nAorJ$vtrlov {utr.loppoei rai n6oe d4ne\os nAarrj$dtAos, n6oe
dpzelos {uMoppoei.

T H E TR U E

t1 t

FOR M

'l'lris syllogismis one of the logical theoremsinvented by Aristotle,

lrrrt cvr:n it differs in style from the genuine Aristotelian syllogism.
lrr lirrrnulating syllogisms with the help of letters, Aristotle
,rlw:rys;ruts the predicate in the first place and the subject in the
,,r'r'orrrf
. [Ie never says 'All B is A', but uses instead the expresrirrrr ',4 is predicated of all B' or more often'A belongsto allB'.t
l,r'l rrs rrpply the first of these expressionsto form (S) ; *. get an
txrrcl translation of the most important Aristotelian syllogism,
l ,rl cl r:l rl l cd' B arbara' :
(6)

If ,4 is predicated of all B
and,B is predicated of all C,
then I is predicated of all C.2

St:rrting with the unauthentic example (r) we have reached

llrrrs lry a step-by-step transition the genuine Aristotelian syllop,irrrr((i). Let us now explain these steps and establish them on a
l rxl rr;tl basi s.
and tenns
\'t. l'rcnisses
l'lvr:ryAris;otelian syllogism consistsof three propositions called
lrrcrrrisscs.A premiss (trpdraoc) is a sentenceaffirming or denyirrg soructhing of something.3 In this sensethe conclusion is also
i np,lr.r.,6, because it states something about something.a The
lwo ck:rncnts involVed in a premiss are its subject and predicate.
Alislotk: calls them 'terms', defining a term (6pos) as that into
wlriclr thc premiss is resolved.sThe original meaning of the Greek
rlrrrrr,:rs well as of the Latin,terminus, is 'limit' or 'boundary'.
'l'lrr tcrrrrs of a premiss, its subject and predicate, are the limits
n| t lrr' prcmiss, its beginning and end. This is the very meaning of
Ilrr worrl ,ipos,and we should be careful not to identify this logical
w,'rrl witlr such psychological or metaphysical words as 'idea',
'rrr,tirrr', 'concept', or Begrif in German.6
I t,l A xcttlyopetrqt
xotd revtds
toO B or td A Jrdpyet navti rQ B. See also
r .l , [.
' ,ln ln. i. 4, 25b37 ci ydp td A xatd. nevrds toi B xoi ti B xord, rovtds ro0 F,
,|,''ly,tr; rd A xatd. rsvtds roA T xatqyopcio?ac.The
word, dldyxl
omitted in the
tr ,r r {l r l i on
w i l l bc c x pl ai ned l ate r.
I lfrrrl t, :r4ar6 rpiruots
rlvds
ltiv o$v doti )\6yos xara,$ertxds I dno$onrds
lr

, f l ,r r l i i r ,',,1'B r d 6i oupni pao p a t i x a t d . n v 6 s i o t ; .

. llrrrl r, r,,,r4l'r{i <ipov 6t xclci eis 6v Erc,\Jezct
fi np6raoc, otov td te Kanlyor,,i r,) r,r0' oi xortlyopeirou
t'ttl.r.r
'' \rrnlr,llc ;rlso uscs thc word Spos in the sense of dprcpds, i,e. 'definition'.

ELEMEN TS

O F THE SYSTEM

gz

In building up his logic Aristotle did not take notice either of

vidual or singular.+It is evident that the terms of universaland

particular premisses
must be universal.Aristotle certainly would
not acceptasmeaningfulexpressions
like 'All Calliasesare men,
or 'SomeCalliasesare men', if there were only one Callias.The
samemust be said about the termsof indefinitepremisses
: they,
tpo, are-universal.This followsboth from the name Aristotle has
chosentor them and from the exampleshe gives.A man who is

t^ An. p,r. i. r, 24^17 (continuation

of the text quoted in p. 3, n. 3) ofios Ei i
xa,06)tou
fi.iv pipet i didpnros. AiyaSi xa06Aovpiv 16 n"r"[i
prl6rrl'in,ipyerr, iv
piper 6i il
twi fi pi1 nauti ind.p7<w,d}dprcrov ii rd ind|7ew p|1 ii,ipyecv
T,yl i t:i
ff
dy^evroA xo06\ou i xard. plpos, otov rd tdv dvaniav itvse tilv eJiy loror4pq,
i ;6'
"i,
fi6ovi1vpi1 etvar dya06v.
z Dc i1L
7, rTagg,l/7o Et xa06Aoupiv 6 ini nAet6vavn!$uxe xarqyopetolaq xal'
-Exoorov
D26 pi, otov dv|poros piv r6tv xa06\ov, Kal,\ics Ei rAv xali'Exasrov,'
3 Ibid. r, tiar6 rpdydAe$os.
I Alexander roo. t r (d7d ydp aio|ryoi rci e'yds xat, d,p$p6v
oix!0' dpp6(et td
xord' navtds ori8t d 6ropro4rrisd,\os' <i ydp Srcptopdsrerv tpatd.oeov ini rd,v'xtb6Aov
7,ipav E1ec.rri Et dropa oi xaldAov. Cf. ibid. 65. e6.

I'I

P R E MIS S E S

AND

TE R MS

rrrrrk'r'iclcd
whether it is true to say 'No pleasureis good' or only
'S'rnr: pleasure is not good', may say without defining the
(lrl:rntity of the subject: 'Pleasure is not good.'But in this last
:i('nt('nce'pleasure'is still a universal term as it was in the two
l,rcvious sentences.Throughout the whole systematic exposition
,,1 lris syllogistic Aristotle in practice treats indefinite premisses
likt' lritrticulars without explicitly stating their equivalence.I This
w;rs tlonc only by Alexander.2
lrrrlefinite premisses are of no importance in the Aristotelian
riysl(:mof logic. No logical thesis, whether a law of conversion or
;r syllogism, is formulated by Aristotle with this kind of premiss.
lt wrts but right that they should be dropped by later logicians,
rvlro rctained only four kinds of premiss, well known to every
strr<lt:ntof traditional logic, viz. the universal affirmative, the
rrrrivcrsalnegative, the particular affirmative, and the particuiar
n('sirtive. In this fourfold division there is no place left for singular
l rlr:rnisses.3
\,:. Wb singular termswere lmitted b2 Aristotle
'['here is an interesting chapter in the Pior AnaQtics where
Alistotle divides all things into three classes.Some, he says, are
srrt:h that they cannot be predicated truly of anything at all,
likc Cleon and Callias and the individual and sensible,but other
llrinss may be predicated of them, e.g. man or animal. Some
otlrcr things, and these are the second class, are themselves
plcdicated of others but nothing prior is predicated of them. For
tlris class of things no example is given, but it is clear that Aristotlc means what is most universal, like being, rd dv.To the third
,:lrrssbelong those things that may be predicated of others and
otlrcrs of them, e.g. man of Callias and animal of man, and
;rs ir rule, concludes Aristotle, arguments and inquiries are conr r:rncd with this classof things.a
' See, for example, An. pr. i. 4, z6^z9 6 yip aitds Eorec ouDoyrcpds d\rcpinov te
iv p!pa \1$0!wos, ot 7, 2ga27 6flou 3d xoi 6rt 16 d\dprotov dvri roA xatqyoptxoA
^,ri
,,,i iv ptpet zt\lpcvov rdv qJrdv nouioet ovA)oycopdviv dnaot tois oyfipaou.
' Alcxander 3o. zg nepi 3t riw dlcoptoror (scil' rfis t<iv d\npiotov 'ivttotpo$fis)
,,i, Alyeq 6rc p16t ypfiorltot rpds ov&oyrcpo,is ciol eito4 xai 6tc ioov tots ini pipovs
Srivovrat.

I Arguments on behalf of the thesis that singular propositions mayie regarded

,,* f,rrming a sub-classof universals-seerforexamplerJ.N. Keynes, Formal Logic,
f ,rrrrfon (19o6), p. ro2-are in my opinion entirely wrong.
1 An. pr. i. z7r 4g^25-43 drdvtav 6\ ritv 6vrav td" piv iotc tonAra tiate xetd'
1.,16cudsdAAou xarlyopeioflot ,itr106s xaldAov (oiov l{Aiav rai Ka}lics xai td xa|'

BLEM ENTS

O F THE SYST E M

$s

II,I

W H Y S IN G ULAR TERM S WERE O M I TTED

7
Aristotle must have felt its weakness, yet it is not corroborated
lry lny philosophical argument borrowed from Plato.
'l'hcrc is, however, another remarkable point that may throw
rorrrc light on our problem. Aristotle emphasizes that a singular
It'rrn is not suited to be a predicate of a true proposition, as a
rrrostuniversal term is not suited to be a subject of such a proposition. The first assertion, as we have already seen, is not gener:rlly true, and the second also seems to be false. But it does not
rrr:rttcr whether these assertions are true of false. It suffices to
krrow that Aristotle regarded them as true and that he eliminated
lirrm his system just those kinds of terms which in his opinion
w(:rc not suited to be both subjects and predicates of true propositions. And here, as I seeit, lies the chief point ofour problem.
It is essential for the Aristotelian syllo.gistic that the same term
rrray be used as a subject and as a predicate without any restriction. In all three syllogistic figures known to Aristotle there
r:xists one term which occurs once as a subject and then again
rrs a predicate: in the first figure it is the middle term, in the
srx:ond figure the major term, and in the third figure the minor
t<:rm. fn the fourth figure all three terms occur at the same
tirne as subjects and as predicates. Syllogistic as conceived by
Aristotle requires terms to be homogeneous with respect to
their possible positions as subjects and predicates. This seems
to be the true reason why singular terms were omitted by
Aristotle.
\ 4. Vaiables
In Aristotle's systematic exposition of his syllogistic no examples
lrrc given of syllogismswith concrete terms. Only non-valid combinations of premissesare exemplified through such terms, which
lrc of course universal, like 'animal', 'man', 'horse'. In valid
syllogisms all terms are represented by letters, i.e. by variables,
c.g. 'If ,R belongs to all S and P belongs to some ,S,then P belongs
to some.R ' .r
The introduction of variables into logic is one of Aristotle's
grcatest inventions. It is almost incredible that till now, as far as I
know, no one philosopher or philologist has drawn attention to
t Ibid. i. 6, zBbT ei ydp td ptv P nevri rQ J zd 3t 17 rwi, dvd.yxtl rd II rui rQ
l' indpxcw. This is a mood of the third figure, called later Disamis, with transposed
l)r ('rnrsscs,

EL EM ENTS

O F THE SYSTEM

$+

this most important fact.r I venture to say that they must all
have been bad mathematicians,for every mathematiiian knows
that the introduction of variables into arithmetic began a new
epoch in that science. It seemsthat Aristotle regardld his in_
vention as entirely plain and requiring no explanation, for there
is nowhere in his logical works any mention ofvariables. rt was
Alexarider who first said explicitly that Aristotre presents his
doctrine in letters, orotyeta, in order to show that we get the
conclusionnot in consequence
of the matter of the premisJes,
but
in consequenceof their form and combination; the letters are
marks of universalityand show that sucha cqncrusionwill follow
alwaysand for any term we may choose.2
There is another commentator,John Philoponus,who is alsofully aware of the significance and importance of variables. He saysthat Aristotle, alfter
showing by exampleshow every premissmay be converted,stbtes
someuniversal rules of conversiontaking letters instead of terms.
For a universalsentenceis disprovedby one examplein which it
tr
9lr-.,. but is proved either by going througf, ail particulars
(which is an endlessand impossibleoperationl or by stating an
evident universal rule. such a rule is given here by Aristotie in
letters, and the reader is allowed to zubstitute (inopiltAew) for
the letters any concreteterms he wants.3
we know alreadythat only universartermsmay be substituted
^ the variables.
for
fn an example quoted above,l Aristotle performs such a substitution, saying: ,L.t ,4 be decidrrorrr,b_
broadJehved plant, c-vine.' This is the only kind of substitution we meet in the Prior Anal\tics. Aristotlenever substitutesfor a
variablez4anothervariable.B,althoughhe is perfectlyawarethat
the same syllogistic mood may be formulited with different
r r am glad to learn that
sir David Ross in his edition of the Analytics. D. 20.
emphasizes that by using variables Aristotre became the founder
tigi"i'
2 Al-exander
53. zB tni ototyeiav tlv bdaoxaAiav norcirac iilp "rroi-uf
ro6 ivtei(iolat
iptv,,6tr o& napd, rlp J\1v ylveroe td..-ouptepd.opara dAtrd.napd ra
"Xrtt "--.J-ri,
ToL-a-urnv
tav nporioeav oupn.Aox)1vx.al rdv rp6rov. oi yd.p ht
fi6e i1 JAt1,owdyerat
ouDoytottxtltg zri6e, tiD' 6rc zi ou(uyia tonitl.
td. o\v otoryeia tol'xa06i\ov xii &ei
xai tni navtds ro'-A1$,ivros roto,rov trrr,t:;;rr";rff::_yj;)r::;;;:rr;rTr;:or:r*
d.oroeyeie nopal,opBdvav dwi r6v 6pav . , .
ropd.6ecypa,,is fi61 eipqrac, xotaoxevd.{<t
, 6rep doriv d.nerpovxai d}Jvorcv, i i 6d.
\d, ritv oroq1iay Dr8or)sxdnq,'donep
lew dvri zitv oroqgelov oies &v BoJAryat
a See p. z, n,

$+

VAR IABL ES

v:rriables.The mood Disamis,for instance,cited at the beginning

ol'this section,is formulated with the lettersft, ,S,P; elsewhereit
is formulated with C, B, A.t It is evident that the validity of a
syllogismdoesnot depend on the shapeof the variablesused in
its formulation: Aristotle knows that without sayingit. It is again
Alcxanderwho statesthis fact explicitly.'
There is no passagein the Prior Anal2ticswhere two different
variablesare identified. Even where the sameterm is substituted
lirr two variables,thesetwo variablesare not identified. In Book
II of the Prior Anal2ticsAristotle discussesthe problem whether
rr syllogismcan be made out of oppositepremisses.This can be
tl<rne,he states,in the secondand third figure. Let B and C, he
r:ontinues,both stand for 'science'and A for 'medicine'. If one
lrssumesthat 'All medicine is science'and that 'No medicine is
sr:icnce',he has assumedthat '.8 belongsto all A' and 'C belongs
lo no A', so that 'Some scienceis not science'.3The syllogistic
rnoodto which this refersruns thus: 'If B belongsto all A and C
lrclongsto no 21,then Cdoesnot belongto someB.'a In order to get
liom this mood a syllogismwith oppositepremisses,
it sufficesto
irlcntify the variables.B and C, i.e. to substituteB for C. We get
lry this substitution; 'If B belongsto all A and B belongsto no l,
lhcn .B doesnot belong to someB.' The heavy roundabout way
Ity meansof concreteterms, such as 'science'and 'medicine', is
rluitc unnecessary.
It seemsthat the straightway in this problem,
i.c. the way by identifying variables,was not seenby Aristotle.
Aristotleknowsthat sentences
like 'Somescienceis not science'
t:itnnotbe true.sThe generalizationof suchsentences'SomeI is
rrolA' (i.e.'A doesnot belongto some,4')alsomust be false.It
is not very probable that Aristotle knew this formula; it is
I An. pr. ii.
7, 5gar7 ei ydp rd I navri rQ B, td Ei A rwi rd B, dldyq td A nvi tQ
l 'Jri p76w .
' Alexander 3Bo. e oJ ydp napd td i p& A sJtAv elvat rd * B i f i ouvoyuyti' t6
1,,\1tohd yivoaq xdv dllors dvri roJrav yptlodtpe9a.
I An. pr. ii. 15,64a23 Eoru ydp tznrfipl iS' o0 i B xai f, iatprx| 6' i$' oi A.
,i
\dpor n6oav iotpexlv intorrjpqv xai pl\epiav iorptxlv incotfiplv, td B navti rQ
"Jv
,4 ,iAq$e xoi td I oiievi, <itor' Eoror tc nrcrfip1 oix drrcripry,
r 'l-his syllogism is a mood of the third figure, called later Felapton, with translxrrcrl prcmisses.In the systematic exposition of the syllogistic it is formulated with
tlrc fcttors R, S, P. See ibid. i. 6, z8az6 6.v ti ptv P ravri rQt E, i 6i I/ pl6ei
i' py1, iorat ouAAo.yto
ltds 5zr zd II rcvi rQ P oiX Jnip{et i{ dvd.yxr1s.
","t
llri<l. ii. t5,64o7 $avepdvDt xaiStr ix $eu66v piv Eonv dAqflis cuAAoyioao|a4
.
, .'x 6i rdv dvtxetp&av olx Eotw' d.ei yd,p Avavrtos 6 ovAAoynpds yivetat rQ
il l'try l.(rT.,

to

ELEM ENTS

O F THE SYSTE M

5+

Alexander again who saw the falsity and applied this fact to
prove the law of conversion of the universal negative premiss.
The proof he givesproceedsby reductio
ad absurdum:If the premiss
'l belongsto no .B' is not convertible,let us suppose that B
belongs to some ,,4.From thesetwo premisseswe get by a syllogism of the first figure the absurdconclusion:'r4 doesnot belong
to some4." lt is obviousthat Alexanderhas in mind the mood
of the first figure called later Ferio : 'lf A belongsto no B and B
belongsto someC, then,4 doesnot belongto someC',2and that
in this mood he identifiesthe variables,4 and C, substituting,4
for C. This is perhaps the neatestexample of an argument by
substitutionderived from an ancient source.
$ 5. Syllogisticnecessity
The first Aristotelian syllogism,called later Barbara, may be
represented,
aswe havealreadyseenr3
in the form of the following
implication:
If ,4 is predicatedof all .B
and B is predicatedof all C,
then I is predicatedof all C.
But there is still a differencebetweenthis formulation and the
genuine Greek text. The premissesare the samein the English
versionasin the Greek,but the exacttranslationof the conclusion
would be 'r4 must be predicated of all C'. This word 'must'
(dv,iyrcq\is the sign of the so-called'syllogisticnecessity'.It is
usedbyAristotle in almostall implicationswhich containvariables
and representlogical laws, i.e. laws of conversionor syllogisms.a
There are, however,somesyllogismswherethis word is omitted ;
take, for instance,this Aristotelianform of the mood Barbara: 'If
z4belongsto all B and C belongsto all l, then C belongsto all.B.'s
Sinceit was possibleto omit the word iniome syllogisms,it must
be possibleto eliminateit entirely from all syllogisms.Let us see,
therefore,what the word meansand why it is usedby Aristotle.
I Alexander
34, 15 Eveon 6i xai Erri ovAAoywpoit,etfcr Erd ro0 nptitou o2qtiporos
y;oltivov, <is xai atlrds npooTpfiror tfi eis d\Jvarov dnoyuyfi. ei yd.p ns pl Aiyot
dvrrcrpiSew tiv xa06Aou8.ro$anxfii, xe/o|u rd A p46evi rQ B. ei 6t pl d.vnotp!$ec,
Eno ri B nvi tQ A' yiverot iv np,ittp oyrjpan td A rwi ,Q A pi itndpTov, 6nep
d,ronov.
2 An. pr. i.
4, z6^25 <i.i ptv A pq}evi tQ B indpTeq zd 3t B zvi tQ f, d"vdyxl d A
nv i r Q I p l J n d p l q e w. 3 Se e p .3 , n . 2 . + Se e p . 7 , n .; p . g, nn. r, 4; above,n. z.
s An. !r. ii. I I, 6r b34 ei yd.pi A revi tQ B xoi rd f rawi rQ A, rd T novti tQ B,

Ii..,

S Y LL OGISTIC

N EC ESSITY

tl

'Ihe problem appearssimple,and is settledimplicitly by Aristotle himselfincidentally in his treatment of the laws of conversion,when he says; 'lf A belongsto someB, it is necessary
that .B
shouldbelongto someI ; but if ,4 doesnot belongto some,B,it is
not necessary
that B shouldnot belongto someA.' For ifr4 stands
lirr 'man' and B for 'animal', it is true that some animal is not
man, but it is not true that someman is not animal, becauseall
men are animals.IWe seefrom this example that Aristotle uses
the sign of necessityin the consequentof a true implication in
order to emphasizethat the implication is true for all valuesof
variablesoccurring in the implication. We may thereforesay 'If
I belongsto someB, it is necessarythat .Bshould belong to some
l', becauseit is true that 'For all A and for all B,if A belongsto
some.8, then -Bbelongsto some,4'.But we cannot say 'If z4does
not belongto someB, it is necessary
that.B should not belong to
somel', becauseit is not true that'For all A andforallB,if A
doesnot belong to someB, then B doesnot belong to some,4'.
There exist, as we have seen,values for .r4and ,B that verify the
antecedentof the last implication, but do not veri$ its conselike 'for all ,4' or 'for
quent. In modern formal logic expressions
all B', where A and -B are variables, are called universal quantifiers.The Aristotelian sign of syllogisticnecessityrepresentsa universalquantifier and may be omitted, sincea universalquantifier
may be omitted when it standsat.the head of a true formula.
This, of course,is all known to studentsof modernformal logic,
but some fifty years ago it was,certainly not known to philosophers.It is not strange,therefore,that one of them, Heinrich
Maier, has chosenour problem as the basisof what is, in my
opinion, a bad philosophicalspeculation.He states:2'The conclusion follows from the premisseswith necessaryconsequence.
'fhis consequencearisesfrom the syllogistic principle and its
necessityrevealsvery properly the syntheticpower of the function ofreasoning.'I do not understandthis last sentence,because
' Ibid. i. z, z5azo4 ei ydp rd A rwi tQ B, xoi td B nvi rQ A dvdyxq JndpTe*...
A pi itd.p7ew, otov ci
ei 6( ye ri A tcvi tQ B pl 6n,ip76t, oix ,i^iyxr1 xei td B rwi
"Q
ti plv B doti (Qov, rd 6t A dv|punos. dv|panos ptv yd.poi novti (<!tu, (Qov 62 tavri
iv|pln<p $n,ipyct.
vol. ii 6, Ttibingen (rgoo), p. 236: 'Aus
' H. Maier, Die S2llogistikdesAristoteles,
rlcn Prilmissen fol$t mit notwendiger Konsequenz der SchluBsatz. Diese Konserlrrcnz cntspringt dem syllogistischen Prinzip, und die Notwendigkeit, die ihr
;rrrlraliet, bekundet recht eigentlich die synthetische Kraft der Schlu6funktion.'

12

$s
I cannot grasp the meaning of the words 'the synthetic power of
the function of reasoning'.Moreover, f am not sure what is
meant by 'the syllogisticprinciple', as I do not know whether any
such principle exists at all. 'On the ground of both premisses
[Maier continues his speculations'] which I think and express,I
must also think and expressthe conclusion by virtue of a compulsion lying in my thinking.' This sentenceI can certainly
understand, but it is manifestly false.You may easily seeits falsehood if you think and pronounce the premissesof a syllogism,
e.g. 'All A is C' and 'Some,B is not C', without pronouncingthe
conclusion which follows from them.
EL EM ENT S

OF

T HE

SYST EM

| 6. What isformal logic?

'It is usualto saythat logic is formal, in sofar asit is concerned
merely with the form of thought, that is with our manner of
thinking irrespective of the particular objects about which we
are thinking.' This is a quotation from the well-known text-book
of formal logic by Keynes.2And here is another quotation, from
the History of Philosophyby Father Copleston: 'The Aristotelian
Logic is often termed formal logic. Inasmuch as the Logic of
Aristotle is an analysis of the forms of thought-this is an apt
characterization.'3
In both quotations I read the expression'form of thought',
which I do not understand.Thought is a psychicalphenomenon
and psychicalphenomenahave no extension.What is meant by
the forrfr of an object which has no extension? The expression
'form of thought' is inexact and it seemsto me that this inexactitude arosefrom a wrong conceptionof logic. Ifyou believeindeed
that logic is the scienceof the laws of thought, you will be disposedto think that formal logic is an investigationof the forms of
thought.
It is not true, however, that logic is the scienceof the laws of
thought. It is not the object of logic to investigate how we are
thinking actually or how we ought to think. The first taskbelongs
to psychology, the secondto a practical art of a similar kind to
mnemonics. Logic has no more to do with thinking than mathematics has. You must think, of course,when you have to carry
I Op. cit., p. 237:
'Auf Grund der beiden Prlimissen, die ich denke und ausspreche, muB ich kraft eines in meinem Denken liegenden Zwangs auch den
SchluBsatz denken und aussprechen.'
2 Op. cit., p. e.
3 Op . cit., p. 277.

$(;

WH AT

IS FOR M AL

L OGIC ?

I3

out an inference or a proof, as you must think, too, when you

havc to solve a mathematical problem. But the laws of logic do
not concern your thoughts in a greater degree than do those of
rnathematics.What is called 'psychologism'in logic is a mark of
the decay of logic in modern philosophy. For this decay Aristotle
is by no meansresponsible.Throughout the whole Prior Analltics,
where the theory of the syllogismis systematicallyexposed,there
t:xistsnot one psychologicalterm. Aristotle knows with an intuitive surenesswhat belongsto logic, and among the logical problcms treated by him there is no problem connected with a
psychicalphenomenonsuch as thinking.
What is therefore, according to Aristotle, the object of logic,
and why is his logic called formal? The answer to this question
is not given by Aristotle himself but by his followers, the Peripatetics.
There was a disputeamongthe philosophicalschoolsofAncient
Greeceabout the relation of logic to philosophy. The Stoicscontended that logic was a part of philosophy, the Peripateticssaid
that it was onlv an instrument of philosophy, and the Platonists
were of the opinion that logic was equally a part and an instrument of philosophy. The dispute itself is of no great interest or
importance,becausethe solution of the disputedproblem seems
to be for the most part a matter of convention.But an argument
of the Peripatetics,preservedby Ammonius in his commentary
deservesour attention.
on the Prior Anafuti.cs,
Ammonius agreeswith the Platonists and says: If you take
syllogismswith concrete terms, as Plato does in proving syllogisticallythat the soul is immortal, then you treat logic as a part
of philosophy; but if you take syllogismsas pure rules stated in
lctters, e.g.''A is predicated of all B, B of all C, thereforez4 is
Jrredicatedof all C', as do the Peripateticsfollowing Aristotle,
then you treat logic as an instrument of philosophy.t
I Ammonius to.
96 xard. ydp IIAdrowo xol dv d.AqflfiA6yovoire p(pos ioilv (scil.
oite p6vos 6pyavov, <itsoi.
i1 Aoyui), ds oiZtatxoi $oow xairlvts rdv lllarawxtw,
tr roa llepndlov Sooiv, ril)d. rci yipos imiv xoi |pyavov $Aooo$les' tdv ptv ydp
yad" ri; npaypd"rov \d,Bps tois A6yous, pipos iotiv, idv 3t lrloris tois xavdvas d,veu
tdv npaypdtov, 6pyovov. <iore xa),6s of 2r roA llepmd.tou ri. zapd Apntoti)ct
i$opdwes 6pyarov tJrlv $eou' $Aois ydp xov6vas nopoii}oow, oi npdyparo )top'
pdvav izox<ipeyc rillti tots notyeiocs tois xav6vos i$appd{av otov 16 A xatd.
novrds roa B, td B xord, navtds roA f, rd A dpa xard, navtds roA i-. The syllogistic
prrxrf of the thesis that the soul is immortal is given a few lines farther on ( r I. ro) :
zofzo 6i dildvarov, il,ltvyif i.po d|dvarov,
\ ,ltuyl oiroxivqrov, roAro 6i d.eaxivtytov,

14

ELEM ENTS

O F THE SYSTEM

$6

_ It is important to learn from this passagethat accordingto the

Peripatetics,who followed Aristotle, only syllogisticlawJ stated
in variablesbelongto logic, and not their applicationsto concrete
terms. The concreteterms, i.e. the valuesof the variables,are
called the matter, J)u1,of the syllogism. If you remove all concrete terms from a syllogism,replacing them by letters, you have
removed the matter of the syllogism and what remains is called
its form. Let us seeof what elementsthis form consists.
To the form of the syllogismbelong, besidesthe number and

Aristotelian logic. These constantsrepresentrelations between

field of universal terms.

It is obviousthat such a theory has nothing more in common
with our thinking than, for instance,the theoryof the relationsof
greaterand lessin the field of numbers.There are, indeed,some
similaritiesbdtweenthesetwo theories.Compare,for example,the
syllogismBarbara:
If a belongsto all D
and Dbelongsto all c,
then a belongs to all c,

with the following arithmeticallaw:

Ifa is greaterthan D
and b is greater than c,
then a is greater than c.

There are, of course, differences between these two laws: the

range of variablesis not the same,and the relationsare different.
I i.ndpyew navri, ind.pyel oi6evi,
Jndpyew nvi, oilg irtipyew nvi:
Jnd,p1ewoi
zavri.
of Jndpyew Aristotle sometimes uses the verb xonlyopeiod",.--Syllo-Instead
gisms in concrete terms are formulated with eivar. See p, z, n.; p,
3,,r. ,, urrd th"
next section (7).

ri

w i l A 'l '

l s l ;( ) l r M Al ,

l .( ) ( ;l ( :1

r .r

llrrt lxrth rr:llttiorrs,irlthough di{lbrcnt altd o(:(rurring betwcen

rlilli'rr:rrt tcrrns, hirvc onc propcrty in common: they are both
I r;rrrsitivc,i.c. thcy arc particular casesof the formula:
If a has the relation R to 6
and D has the relation rRto c,
then a has the relation R to c.
I t is a curious thing that this very fact was observed by
tlrr: logicians of the later school of the Stoics. Arguments like 'the
lilsl is greater than the second, the second is greater than the
tlrild, therefore the first is greater than the third' were called by
tlrr: Stoics, as Alexander declares, 'non-methodically conclusive'
;rrrtl were not treated as syllogisms in the sense of their logic.
Nr:vr:rtheless,the Stoics regarded such arguments as similar
(iilt"rtot)to categorical syllogisms.I This observation of the Stoics,
wlrit:h Alexandir tries to confute without producing convincing
('oltnter-arguments, corroborates the supposition that the logic
ol'Aristotle was conceived as a theory of special relations, like a
rnathematical theory.
\ 7. What isformalism?
Itormal logic and formalistic logic are two different things.
'l'lrc Aristotelian logic is formal without being formalistic,
whcreas the logic of the Stoics is both formal and formalistic.
logic is meant by
l,ct us explain what in inodern'formal
'lirrmalism'.
Modern formal logic strives to attain the greatest possible
(:xirctness.This aim can be reached only by means of a precise
lirnguage built up of stable, visually perceptible signs. Such a
l;tnguage is indispensable for any science. Our own thoughts not
lirrmed in words are for ourselvesalmost inapprehensible and the
thoughts ofother people, when not bearing an external shape,
t ould be accessible only to a clairvoyant. Every scientific truth,
in order to be perceived and verified, must be put into an external
lirrm intelligible to everybody. All these statements seem inr:ontestably true. Modern formal logic gives therefore the utmost
t Alexander 2r. go oi d.pe06}as repotvovtes\6yot napd. tots Dta'ixots, otov'td
rptittov roA ievtipou pei{ov, td 6ri Derizeporroi tpitou, rd dpo npitov toA tpirov
p.i{or.' Ibid. g45. rZ torcArol elot xoi ois Aiyouocv of vetitepot (i.e' Etu'ixoi)
,ip<OdEosrepeivovtos.oig 6tt ptv Fi A(youot ovhoytotwits ouvdyew,JyedsAiyouot...
in 6l r)yofivtot ipoious airois clvet toes xetqyopcxois oviloyrcpoes...toA asn6s
r'ttQltoptdvouow.

16

ELE M ENTS

O F THE SYSTEM

$Z

attention to precision of language. What is called formalism is

the consequenceof this tendency. In order to understand what it
is, let us analyse the following example.
There existsin logic a rule of inferbnce, called formerly modus
ponensand now the rule of detachment. According to this rule, if
an implication of the form 'If c, then B ' is assertedand the antecedentof this implication is assertedtoo, we are allowed to assert
its consequentB. In order to be able to apply this rule we must
know that the proposition c, assertedseparately, expresses'the
same' thought as the antecedenta of the implication, sinceonly
in this case are we allowed to perform the inference. We can
state this only in the casewhere thesetwo a's have exactly the
same external form. For we cannot directly grasp the thoughts
expressedby thesea's, and a necessary,
although not sufficient,
condition for identifying two thoughtsis the external equality of
their expressions.When, for instance,assertingthe implication
'If all philo'sophersare men, then all philosophersare mortal'
you would also assertas second premissthe sentence'Every
philosopheris a man', you could not get from thesepremisses
the
conclusion 'All philosophersare mortal', becauseyou would
have no guarantee that the sentence'Every philosopher is a
man' representsthe same thought as the sentence'All philosophers are men'. It would be necessary
to confirm by meansof a
definition that 'Every A is B'meansthe sameas 'All ,4'sare -B's';
on_the ground of this definition replace the sentence'Every
philosopheris a man' by the sentence'All philosophersare men;,
and only then will it be possibleto get the conclusion.By this
exampleyou can easilycomprehendthe meaning of formalism.
Formalism requires that the same thought should always be
expressedby means of exactly the same seriesof words ordered
in exactly the samemanner. When a proof is formed according
to this principle, we are able to control its validity on the basisof
its external form onJy, without referring to the meaning of the
terms usedin the proof. In order to get the conclusionp from the
premisses'If a, then p ' and c, we'need not know either what a
or what p really means; it sufficesto notice that the two a's contained in the premisses
have the sameexternalform.
Aristotle and his followers, the Peripatetics,were not formalists. As we have already seen,Aristotle is not scrupulously exact
in formulating his theses.The most striking caseof this inexacti-

$7

WH A T

IS FOR M AL ISM ?

r7

tude is the structural discrepancybetween the abstract and conr:reteforms of the syllogisms.Take as an example the syllogism
with opposite premissesquoted above, in our section 4.r Let B
rrnd C be 'science'and A 'medicine'. Aristotle states:
I n variables :

If B belongsto all A
:rndC belongsto no l,
thenC doesnot belongto some8.2

In concreteterms:

If all medicineis science

and no medicineis science,
then somescienceis not science.

fhe difference of corresponding premisses,of which the two

syllogismsconsist,is evident. Take, for instance,the first premiss.
'l'o the formula 'B belongs to all ,4' would correspond the
scntence'Sciencebelongsto all medicine', and to the sentence
'All medicine is science'would correspondthe f,ormula 'All I is
/J'. The sentencein concreteterms, given by Aristotle, cannot be
rcgarded as a substitution of the abstract formula accepted by
him. What is the causeof this difference?
Alexander gives three explanationsof this problem:3 the first
may be omitted as unimportant, the last is a philosophical one
and is, in my opinion, wrong; only the second deservesour
ruttention.According to this explanation, in formulae with the
vcrb 'to be predicated of something' and, we may add, with the
vcrb 'to belong to something', the subject and the predicate are
bctter distinguishable(yvaptlt<hepoc)than, we may add again, in
lirrmulaewith the verb 'to be'. In fact, in formulae with 'to be'
the subjectas well as the predicateis usedin thenominative; in
Iirrmulae preferred by Aristotle only the predicate is in the
rrominative, and the subject is either in the genitive or in the
drrtive and therefore can be more easily distinguished from
lhc predicate.Very instructive, too, is the final remark of Alexan<ler,from which it follows that to say 'Virtue is predicated of all
.justice'instead of the customary 'All justice is virtue' was felt in
Ancient Greek to be as artificial as in modern languages.
I See p. g, n.
3.
I The conclusion in variables is dropped in the Greek text.
r Alexander
54 . zr ypfitat 6i tQ xari' nevrds xai rQ xold' pqlevds e'rrf Er8corclig,
rlt 6rd.toitav yvtitptltos ti ouvoyoyil tdv Adyav, xai 6tc o$tos Aeyopivov yvapcyti'
rtpos 6 re xenlyopoi'ltcvos xai d inoxelpevos, xai 6rt npbtov rff <$Joetd xerd' zawis
rr,0 r'v 5tro ai'tQ, <Is npoeiptyau fi pivtot ypqoc i ouA\oytotrxil 2v rfi oury|eiq d'v,intAv ETer oJ ydp i1 d.peril Aiyetot xatd ndols 6rcanooJw7s,&M' ,ivdnaAw n6oe
Itrxutoowll d,pet{. 66 rci 6ei rcr' dp$otipos zris dr{opris yulrd(eu iautoJs, ive rfi
t, yprloet napaxoAovfleiv6wipe0e rai zf Er8corc,\/g.

r8

ELEM ENTS

O F THE SYST E M

$Z

There are still more casesof inexactitude in Aristotelian logic.

Aristotle constantly usesdifferent phrasesfor the samethoughts.
I shall give only a few examples of this kind. He begins his
syllogisticwith the words 'r4 is predicated of all B', but shortly he
changesthesewords into the phrase ',4 belongs to all .B', which
seemsto be regular. The words 'is predicated' and 'belongs' are
frequently omitted, sometimeseven the important sign of the
quantity 'all'is dropped.Besidesthe form '24belongsto someB'
there are forms which may be translated '24belongs to some of
the.B's'. The premissesof the syllogismare combined by meansof
different conjunctions. Syllogistic necessityis expressedin different ways and is sometimes entirely omitted.r Although these
inexactitudes have no bad consequencesfor the system, they
contribute in no way to its clearnessor simplicity.
This procedure of Aristotle is probably not accidental, but
seemsto derivefrom somepreconceptions.Aristotle saysoccasionally that we ought to exchangeequivalent terms, words for words
and phrasesfor phrases.zCommenting on this passage,Alexander declaresthat the essenceof the syllogism dependsnot on
words but on their meanings.3This statement,which is manifestly
directed against the Stoics,can be understoodthus: the syllogism
doesnot changeits essence,
i.e. it remainsa syllogism,if someof
its expressionsare replaced by other equivalent expressions,e.g.
if the expression 'to be predicated of all' is replaced by the
equivalent expression 'to belong to all'. The Stoics were of a
directlt'opposite opinion. They would say that the essenceof the
syllogismdependson words, but not on their meanings.If therefore the words are changed, the syllogismceasesto exist. This is
r The phrase td A xod. neyt6s toi B (xatqyopehar is twice omitted) is used in
the mood Barbara (see p. 3, n. z), i A ravri rQ B (JrdpTet is altogether omitted)
is used in another formulation of the same mood (see p. Io, n. 5), The phrase zd
A twi tdv B appears in the laws of conversion I elsewhere, e.g. in the mood Disamis,
we have rd A twi rr! B (see p. 9, n. t). The logically important word zcyrlis
altogether omitted in a formuletion of the mood Barbara (see p, z, n.). The
conjunction 'and'is for the most part denoted by p6. .. Dd (see,forexample, p.7,
n. or p. Io, n. z), sometimes by xcl (see p. 3, n. 2 ; p. I o, n. 5). Syllogistic necessity
is as a rule expressed by dvdyq Jndpyew (see p. 7, n. or p. 9, n. r), in the mood
Felapton it is denoted by riz<ip(eti( dvdyxrls (seep. g, n. 4), In one case it is dropped
( s e ep . r o , n . 5 ) ,
2 An. pr. i.
39, 49b3 Eei 3t rai petaAapBdvew d rd crird Sivataq dvdpata dm'
dvopdtov xai iiyous dmi Adyav,
3 Alexander g7z. zg oix iv tais Ai(eow 6 aulAoynpds rd elvot Eyeq dM' dv tois
or7ltowoltivo ts,

tiz

WHAT

IS FOR M AL ISM

t9

illustratedby Alexander with an example.fromthe logic of the

Stoics.The rule of inferencecalled mlduspnnensl
Ifa, thenp;
but a:
thereforeB,
is the first 'indemonstrable' syllogism of the Stoics. Both the
Stoicsand the Peripateticsseemmistakenlyto regard the phrases
'lf c, thenp'and'c entailsB'as havingthe samemeaning.
llut if, in the syllogismgiven above, you replace the premiss 'If
,v,then F' by ' a entailsp ', saying:
c entailsB;
but a;
therefore B,

you get according to the Stoics a valid rule of inference, but not
a syllogism. The logic of the Stoics is formalistic.I
I Alexander
g7g, zB Apntori\qs
ltiv o$v oitas repi tCw rcrd tds A((ets perc)vitlteuv $(pcrat (see p. IB, n. z). oi 6i ve,itepot (i,e. oi Eto'txoi), rais Afeow i.noxoIou|oavtes orirlt
Ei rots olpalopivots,
oi rair6v
$aot yiveo|ar dv rcis eis ztis
ydp otlpetvovros toi'
ei ti A
Ioo}uvaporioos Ai(cts pen)r$,!eoc rtv
dpov' tairdv
rrl B' rQ 'dxo\ou|<i
tQ A i
B', ouMoyrctrrcdv ptv Adyov $ooiv etvot tonrhls
dild
Aq$Oeio1s rfis )ri{eus ' ei ti A rd B, d Ei l, zri dpo B ' , oix&t 6t ouiioyroardr
rtpavrxdv
rd ' dxo\ouflei tQ A fi B, zri 6t 14, i d.po B'.

bII

C H A PT E R

II

THES E S OF TFIE S Y ST E M
andrulesof inference
$8. Theses
Aristotelian
theory of the syllogismis a systemof true proTnn
positionsconcerningthe constantsA, E,1, and O. True propositions of a deductivesystemI call theses.Almost all thesesof the
Aristotelian logic are implications,i.e. propositionsof the form
'If ct,then p'. There are known only two thesesof thislogicnot
beginningwith 'if ', viz. the so-calledlaws of identity : 'A belongs
to all A' or 'All A is A' , and 'A belongsto some,4' or 'Some,4 is
z4'.Neither of theselaws was explicitly stated by Aristotle, but
they were known to the Pcripatetics.I
The implications belonging to the systemare either laws of
conversion(and laws of the squareof oppositionnot mentioned
in the Prior Anal'tics)or syllogisms.The laws of conversionare
simple implications,for instance: 'If A belongsto all B, then .B
belongsto some A.'2 The antecedentof this implication is the
premiss 'r4 belongsto al| B' , the consequentis '.8 belongsto some
.,4'. This implication is regarded as true for all values of the
variablesA and B.
All Aristgtelian syllogismsare implications of the type 'If cv
and p, thefi y', where crand p are the two premissesand y is the
conclusion.The conjunctionof the premisses'cu and p' is the
antecedent,the conclusion7 is the consequent.As an example
take the following formulation of the mood Barbara:
If ,4 belongsto all B
and B belongsto all C,
then I belongsto all C.

TH E S E S

A ND

R U LE S

OF

IN FE R E N C E

I t must be said emphatically that no syllogismis formulated

lry Aristotleas an inferencewith the word 'thercfore' (dpa), as is
rl,rrrt:in the traditional logic. Syllogismsof the form:
AllB is l;
all C is B;
therefore
all C is,4

In this example a meansthe premiss',4 belongsto all ,B', p the

premiss'-B belongsto all C', and y the conclusion',4 belongsto
all C'. This implication is also regardedas true for all valuesof
the variables A, B, and C.

:rrc not Aristotelian. We do not meet them until Alexander.r

'l'his transferenceof the Aristoteliansyllogismsfrom the implicational form into the inferentialis probably due to the influenceof
the Stoics.
f'he difference between the Aristotelian and the traditional
syllogismis fundamental. The Aristotelian syllogismas an implicationis a proposition,and asa propositionmust be either true
or false. The traditional syllogism is not a proposition, but a set
ol'propositionswhich are not unified so as to form one single
proposition.The two premisseswritten usually in two different
lincsare statedwithout a conjunction,and the connexionof these
lrxrscpremisseswith the conclusion by means of 'therefore' does
not give a new compound proposition. The famgus Cartesian
principle, 'Cogito, ergo sum', is not a true principle, becauseit is
not a proposition.It is an inference,or, accordingto a scholastic
not
tcrminology, a consequence.Inferencesand consequences,
lrcing propositions,are neither true nor false, as truth and falsity
bclongonly to propositions.They may be valid or not. The same
lrasto be saidof the traditional syllogism.Not being a proposition
the traditional syllogismis neither true nor false; it can be valid
or invalid. The traditional syllogismis either an inference,when
statedin concreteterms, or a rule of inference,when stated in
variables.The senseof such a rule may be explained by the
r:xirmplegiven above: When you put such values for A, B, and
(,'that the premisses',4 belongsto all B' and 'B belongsto all
(i'are true, then you must accept as true the conclusion'A
hlongs to all C'.
If you find a book or an article where no difference is made
lx:twccn the Aristotelian and the traditional syllogism'you may

t Cf. p.
9, n. 5, p. ro, n. t. In the passagequoted in the latternoteAlexander
says that the ploposition 'l does not belong to some ,4' is absurd. That means that
the contradictory proposition ',4 belongs to all :4' is true.
2 An, pr. i. z, z5ar7 ei 3i nqwi rd A tQ B, xsi rd B twi rQ A in,ip{e.

I In Alexander
(Qov
47. gwe find a syllogism in concrete terms with d'pa: ndv
,,ith dori, riv (Qov Ep,ltvy|v ion, ris dpa oiote Elrlu7a6stonv. At 38e. rB we have a
crrrrgrlcxsyllogism in four variable terms with dpo: i A novti rQ B, td B nowi tQ
I', ,i A oi}evi rQ A, t6 dpe A oi6oi tQ f.

22

THESES O F THE SYSTEM

gs

be sure that the author is either ignorant oflogic or has never

seenthe Greektext ofthe Organon.
Scholarslike Wiitz, the modern
editor and commentator of the Organaz,Trendelenburg, the compilir of the Elementalogicu Aristoteleae,prantl, the historian of
logic, all knew the Greek text of the Organonwell, but neverthe_
lessthey did not seethe difference betweenthe Aristotelian and
the traditional syllogism. only Maier seemsto have felt for a

Aristotle, and does not even say what differenceshe has seen.r
when we realize that the difference between a thesisand a rule
of inferenceis from the standpoint oflogic a fundamental one, we
must agree that an exposition of Aristotelian logic which disregards it cannot be sound. We have to this day no genuine
exposition of Aristotelian logic.

to convince ourselvesthat this transformation is correct. Supposing now that a and B are true premissesof a syllogis-, *"
get the conclusion y, applying the rule of detachment-twice to
the purely implicational form of the syllogism. I{, therefore, an
Aristotelian syllogismof the form .If c and p, then y' is true, the
corresp_onding
traditional mood of the form . c, B, therefore7' is
valid. But conversely,it seemsimpossible to deduce the corret Maier, op. cit.,
vol. iia, p.74rn. z: .Es ist vielleicht gestattet, hier und im
_
Folgenden die
Darsterrungsform der spiteretitogit,'ai"
.geldufigere
lerchter zu handhaben
ist, an die Stele der aristoterischen
,Jtzin.' rne".rgLi"h
irood
"u
Barbara is quoted ibid., p. 75, thus:

:iHE;:*

alles C ist A
where the strbke replaces the word .therefore'.

$B

T H E S E S A ND R U L ES OF IN FER EN C E

23

spondingAristotelian syllogismfrom a valid traditional mood by

known logical rules.
\ g. Thesltllogisticfigures
There are some controversial problems connected with the
Aristotelian logic that are of historical interest without having
any great logical importance. Among theseis the problem of the
syllogisticfigures. The division of the syllogismsinto figures has,
in my opinion, only a practical aim: we want to be sure that no
true syllogistic mood is omitted.
Aristotle divided the syllogistic moods into three figures. The
shortestand clearestdescription of these figures is to be found
not in the systematicpart of the Prior Anal2ticsbut in the later
chaptersof that work. If we want, Aristotle says,to prove A of B
syllogistically, we must take something common in relation to
both, and this is possiblein three ways: by predicating either z4
of C and C of B, or C of both, or both of C. Theseare the figuresof
which we have spoken,and it is clear that every syllogism must
be made in one or other of thesefigures.'
It follows from this that A is the predicate and .B the subject of
the conclusionwe have to prove syllogistically. z4is called, as we
shall seelater, the major term and .B the minor; C is the middle
term. The position of the middle term as subject or predicate of
the premissesis the principle by which Aristotle divides the
syllogistic moods into figures. Aristotle says explicitly that we
shall recognizethe figure by the position of the middle term.2 In
the first figure the middle term is the subject of the major term
and the predicate of the minor term, in the secondfigure it is the
predicate, and in the last figure the subject, of both the other
terms. Aristotle, however, is mistaken when he says that every
syllogism must be in one of thesethree figures. There is a fourth
possibility, viz. that the middle term is the predicate of the lajor
ierm and the subject of the minor term. Moods of this kind are
now spokenofas belonging to the fourth figure.
In the above passageAristotle has overlooked this fourth
I An. pr. i. 23,
4ob3o ei Er) Eior rri A rotd roa B ouiloyloaolor' i i'n'ipyov fi pfi
Jn,iplgor, dvdyxr1 l,"Brtu tt xcrd rrvos. 4tat3 ei olv d.vd'yxt1piv n AtBeiv ryd's d'p$axowdv, toiro 6' iv676tat rycX,Is (i ydp i A toA I xoi i | rcA B xottlyoprioavras, .ii
xar' dp,$otv,fi dp$<o xeilt toi T), roita 6' mi rri eiprlpiva oyiporo, $ovepdv
ri'l
6tt ndlre ouiloyrcpdv dvdyxl ylveo?ot 6d' roirov rwds ritv oyqpdtov.
yvt'tpco0pevtd o1fipo,
'z Ibid. 3r, 47b| 3 ri to0 poov 0!.oet

24

THESES O F THE SYSTEM

$g

possibility, although a few chaptersfarther on he himself gives a

proof by a syllogism in the fourth figure. It is the sameproblem
again: we have to prove A of E syllogistically, where I is the
major term and E the minor. Aristotle givespractical indications
how to solve this problem. We must construct a list of universal
propositionshaving the terms A and,E as subjectsor predicates.
In this list we shall have four types of universal affirmative
proposition (I omit the negative propositions), 'B belongs to all
A' , 'A belongsto all C', '{ belongsto all E' , and '.Ebelongsto all
H'. Each of the letters B, C, 4,, and H representsany term fulfilling the aboveconditions.When we find among the C's a term
identical with a term among the /'s, we get two premisseswith a
common term, say( :' A belongsto all (' and',( belongsto all E',
and the proposition ',4 belongsto all E'is proved in the mood
Barbara. Let us now supposethat we cannot prove the universal
proposition'r4 belongsto all -E', as the C's and {'s have no common term, but we want at leastto prove the particular proposition 'r4belongsto someE'. We can proveit in two differentways:
if there is a term among the C's identicalwith a term among the
f/'s, sayH,we getthe moodDarapti ofthe third figure : 'l belongs
to all H','E belongsto all H', therelore'l must belong to some
.E'. But there is still another way when we find among the f/'s a
term identical with a term among the -B's,sayB; we then get a
syllogismwith the premisses'E bel,ongsto all ,B' and.'B belonlgsto
all A' , from which we deducethe proposition'A belongsto some
E' by cogverting the conclusion 'E belongsto all A' obtained
from thesepremissesby the mood Barbara.I
This last syllogism: 'If E belongsto all .B and B belongsto all
l, then I belongsto some.E', is a mood neither of the first figure
nor of the secondor third. It is a syllogismwhere the middle term

T H E S Y L LOGISTIC

$o

FIGU R ES

25

1l is the predicateof the major term.r4and the subjectof the minor

tcrm E. It is the mood Bramantip of the fourth figure. Neverthelessit is as valid as any other Aristotelian mood. Aristotle calls it
becausehe
ir'converted syllogism' (d.vteorpolt
lt'ivos ouiloy,'opr.ris)
of
the mood
proves this mood by converting the conclusion
of
the
second
Carnestres
iJarbara. There are two other moods,

ing this conclusioninto 'P belongsto some.R' we get the proof of

Diiamis. Aristotle here applies the conversion to the conclusion
of the mood Darii, which gives another syllogism of the fourth
figure called Dimaris: 'If R belongs to all S and ,Sbelongs to
someP, then P belongsto someR.'I
All thesedeductionsare logically correct, and so are the moods
obtained by their means.Aristotle knows,indeed, that besidesthe
Iburteen moods of the first, second,and third figures established
by him systematicallyin the early chapters of the Pior Anal2tics
there are still other true syllogisms.Two of them are quoted by
him at the end of this systematicexposition.It is evident, he says,
that in all the figures,whenevera syllogismdoesnot result, if both
the terms are affirmative or negative nothing necessaryfollows at
tll, but if one is affirmative, the other negative, and if the negative is stated universally, a syllogismalways resultslinking the
rninor to the major term, e.g. if z4belongsto all or somep' and B
belongsto no C; for if the premissesare converted it is necessary
that C does not belong to some ,4.2 From the secondpremiss

glad to see that this reading is also accepted by Sir David Ross.
"Q

A pl i nd.pycl .

26

THESES O F THE SYSTEM

$s

'given here by Aristotle we get by conversion the proposition 'C

belongsto no.B', from the first premisswe get '.8 belongsto some
A', andfrom thesetwo propositionsresults,according to the mood
Ferio of the first figure, the conclusion 'C doesnot belong to some
A'. Two new syllogistic moods are thus proved, called later
Fesapoand Fresison:
If I belongsto all.B
and B belongsto no C,
then C doesnot belongto somel.

If I belongs to some B
and -B belongs to no C,

thenC doesnot belongto someA.

Aristotle calls the minor term C and the major term A becausehe
treats the premissesfrom the point of view of the first figure. He
says,therefore,that from the given premissesa conclusionresults
in which the minor term is predicated of the major.
Three other syllogismsbelonging to the fourth figure are mentioned by Aristotle at the beginning of Book II of the Prior
Analytics.Aristotle stateshere that all universal syllogisms (i.e.
syllogismswith a universal conclusion)give more than one result,
and of particular syllogismsthe affirmative yield more than one,
the negativeyield only one conclusion.For all premissesare convertible except the particular negative; and the conclusionstates
something about something. Consequentlyall syllogismsexcept
the particular negative yield more than one conclusion, e.g. if A
hasbeenprovedto belongto all or to some,B,then.Bmust belong
to someA; and if A has been proved to belong to no -8, then .B
belongs to no .r4.This is a different conclusion from the former.
But if I does not belong to some .8, it is not necessarythat B
should not belong to someA, for it may possiblybelong to all l.r
We seefrom this passagethat Aristotle knows the moodsof the
fourth figure, called later Bramantip, Camenes,and Dimaris,
and that he gets them by conversion of the conclusion of the
moods Barbara, Celarent, and Darii. The conclusion of a syllogism is a proposition stating something about something, i.e. a
premiss,and thereforethe laws ofconversion can be applied to it.
I An. pr. ii. l,
5ga4 oi, ltiv xa06Aou(scil. ou,\,\oyropoi) ndwes dei nAeia odloyi(ovran, rit
!' iv piper oi. plv xatlyoprxoi n\eto4 oi 6' d.no$attxoi td oupnlpaopa pdvov.
ai, ptu yd.p dX\ot npotd,oers dvrrcrpi<$ouor, fi 62 orepTnxi oix d"wcotp!$er-16 6i
oupnipeopo ti xotd, nvds iotl. ito|' oi. piv d),Aor ouiloywpo| r),eia ouL)oyi(ovtac,
otov ei td A 6i6ecner nav:/r rQ B i wi, xai rd B nvi rQ A dvoyxoiov 6n,ipyirr: *oi ei
pqlevi tQ B d A, oriDl rd BoriEeyi rQ A. niro 6' Etepov roi Eprpoo0ev.-ei'61 twi p)7
indp76r, oix ,ivdyny xai td B nvi tQ A pl $ndp7ew. 2vidytac ydp navti ind.pyec/. '

$s

T H E SYL L OGISTIC

FIGU R ES

27

It is importhnt that propositions of the type 'l belongsto no .B'

and 'B belongsto no A' are regarded by Aristotle as different.
It follows from thesefacts that Aristotle knows and acceptsall
the moods of the fourth figure. This must be emphasizedagainst
the opinion of some philosophersthat he rejected these moods.
Such a rejection would be a logical error which cannot be imputed to Aristotle. His only mistakeis the omissionof thesemoods
in the systematicdivision of the syllogisms.We do not know why
he did so. Philosophical reasons,as we shall see later, must be
excluded.The most probable explanation is given,inmyopinion,
by Bochedski,' who supposes that Book I, chapter 7 and
Book II, chapter I of the Prior Analytics,where thesenew moods
are mentioned, were composedby Aristotle later than the systematic exposition of chapters4-6 of Book I. This hypothesisseems
to me the more probable, as there are many other points in the
Prior Analytfcssuggestingthat the contents of this work grew
during its composition. Aristotle did not have time to draw up
systematicallyall the new discoverieshe had made, and left the
continuation of his logical work to his pupil Theophrastus.
Theophrastus,indeed, found for the moods of the fourth figure
which are 'homeless' in Aristotle's system a place among the
moods of the first figure.2 For this purpose he had to introduce
a slight modification into the Aristotelian definition of the first
figure. Instead of saying that in the first figure the middle term
is the subject of the major and the predicate of the minor,
as Aristotle does,3he said genbrally that in the first figure the
middle term is the subject of one premiss and the predicate of
another. Alexander repeatsthis definition, which probably comes
from Theophrastus,and seemsnot to seethat it differs from the
Aristotelian description of the first figure.a The correction of
I I. M. Bocheriski, O.P., In Logtquc fu TMophrastc, Collectanea Friburgensia,
Nouvelle Sdrie, fasc. rodi, Fribourg en Suisse (tg47), p. Sg.
2 Alexander 69, z7 Oed'$ponos 62 tpoori|rlorv d,\)ous rlvrc tois ttooapor rohors
oixitc rd\ehus oi6' dlato}eixrous 6wos, <Iv pvtlpoveiet xoi 6 Apntotitrqs, t6w piv dv
roitq tQ BTBAiEqoe\0tiv, tCw Et dv rQ perd toaro rQ \ewipa xat' ,i.py<is.Cf. ibid.
r ro. t2.
t C f.p. 23, n. I.
r (ad i. zg) fi 62 rct pioou oytoc np6s t6., <$vlopBdvetan ploov,
' Alexander z5B. 7
tpryits yivetot (i ydp t poq dAetan aitdv tQ p2v iroxeipeos oit6r ro0 6E xanTyopoJpevos,i ,ip$ottpov xotqyopeetan, fi dp$oipoc indxenar). Ibid. 349. 5 (ad i. 3e)
Ev piv ydp 6 pioos 2v dpSotipars itv tais npord.oeow oiros fi <is rcA piv Kernyopcio9et o;il6r rcit 6i rizoneioOaq npGnov orar o2gffpa.

TH ESES O F THE SYSTEM

$g

Theophrastus is as good a solution of the problem of the syllogistic figures as the addition of a new figure.
$ rc. The major,middle,and minortcrms
There is still another error committed by Aristotle in the Prior
Anafutics,with more seriousconsequences.
It concernsthe definition of the major, minor, and middle terms as given in his
characterization of the first figure. This begins with the words:
'Whenever three terms are so related to one another that the last
is containedin the middle and the middle is containedor not in
the first, the extremesmust form a perfect syllogism.' This is how
he begins; in the next sentencehe explains what he means
by the middle term: 'I call that term the middle which is itself
contained in another and contains another in itselfi, which by
position also becomesthe middle." Aristotle then investigatesthe
syllogistic forms of the first figure with universal premisseswithout using the expressions'major term' and 'minor term'. These
expressions
occur for the first time when he comesto the moods.
of the first figure with particular premisses.Here we find the
following explanations: 'I call that term the major in which the
middle term is contained and that term the minor which comes

In fact theseexplanations can be applied only to syllogismsof

the mood Barbara with concreteterms and true premisses,
e.g.:
(t)
If all birdsare animals
and all crowsare birds,
then all crowsare animals.
In this syllogismthere is a term, 'bird', which is itself contained
in another term, 'animal', and contains in itself a third term,
| An. pr, i.
4, z5bgz 6rav oiv 6pot tpees oiros iyuor zptis <i)hilous itore t6v
io1.orov
iv 6Aq elvot rQ pioE xoi ziv ptioov iv 6Aqt te np,it<p i eirot
fi pi1 ,lrar,
fiu d.xpav elvac oiJoycopdv r(Auov. xeAit Di ploov pEv6 xai oiti Ju'ailE
'i;y,itx1
""i
dAAotv roitE iotlv, 6 xai rff |ioer ylverar ploov,
2 Ibid., z6azt )\iyoDt pei{.ov p2v
d.xpovdv tl rd pfuov dotiv, Elorrov Dt rd rizd rd
poov 6v.
3 Maier, op. cit., vol. ii a, pp.
49, 55, really treats them as definitions valid for all
the moods of the 6rst figure.

$ ro

THE MAJOR,

MID D L E,

AN D M IN OR

TER M S

29

'crow'. According to the given explanation 'bird' would be the

middle term. Consequently'animal' would be the major term
and 'crow' the minor term. It is evident that the major term is
so called becauseit is the largest in extent, as the minor term is
the smallest.
We know, however, that syllogismswith concrete terms are
only applications of logical laws, but do not belong to logic
themselves.The mood Barbara as a logical law must be stated
with variables:
(z) If all B is I
and all C is .B,
then all C is A.
To this logical law the given explanations are not applicable,
because it is not possible to determine extensional relations
betweenvariables.It may be said that.B is the subjectin the first
premissand the predicate in the second,but it cannot be stated
that.B is containedin A or that it containsC; for the syllogism(z)
is true for all values of the variables A, B, and C, even for those
which do not verify its premisses.Take 'bird' for r4, 'crow' for B,
and 'animal'for C: you get a true syllogism:
(g)
If all crowsare birds
and all animalsare crows,
then all animalsare birds.
The extensionalrelationsof the terms 'crow', 'bird', and 'animal'
are of course independent of'syllogistic moods and remain- the
samein syllogism(3) as they were in (r). But the term 'bird' is
no longer the middle term in (3) as it was in (r); 'crow' is the
middle term in (3) becauseit occurs in both premisses,and
This is the
the middle term must be common to both premisses.
figures.r
for
all
definition of the middle term acceptedby Aristotle
explanathe
special
This general definition is incompatible with
tion given by Aristotle for the first figure. The specialexplanation
of the middle term is obviously wrong. It is evident.alsothat the
explanationsof the major and minor terms which Aristotle gives
for the first figure are wrong, too.
Aristotle does not give a definition of the major and minor
termsvalid for all figures; but practically he treatsthe predicate
I An. pr.i.3z,47a3B pioov 62 |eriov tCw dpov tdv iv dp$oripots rets npord.oeot
Aeydpevov'dv,iyxl yd.ptd p&ov 2v d.p$oipors JndpTew lv d.ncot tois o1giltoow.

go

THESES O F THE SYSTEM

$ ro

of the conclusion as the major term and the subject of the conclusion as the minor term. It is easy to seehow misleading this
terminology is: in syllogism (3) the major term 'bird' is smaller
in extension than the minor term 'animal'. If the reader feels
a difficulty in acceptingsyllogism (3) becauseof its false minor,
he may.read 'someanimals' insteadof 'all animals'. The syllogism:
(+) If all crowsare birds
and someanimalsare crows,
then someanimalsare birds
is a valid syllogism of the mood Darii with true premisses.And
here again, as in syllogism(3), the largestterm 'animal' is the
minor term; 'bird', middle in extension,is the major term; and
the smallestterm, 'crowl, is the middle term.
The difficulties we have already met are still greater when we
take asexamplessyllogismswith negativepremisses,e.g. the mood
Celarent:
If no.Bis I
and all C is .8,
then no C is A.
B is the middle term; but doesit fulfil the conditions laid down
by Aristotle for the middle term of the first figure? Certainly not.
And which of the terms, C or A, is the major and which is the
minor? How can we compare these terms with respect to their
exten5ion? There is no positive answer to theselast questions,as
they spring from a mistaken origin.r
$ rt. Tlu historyof an mor
The faulty definition of the major and the minor terms, given
by Aristotle for the first figure, and the misleading terminology
he adopts, were already in antiquity a source of difficulty. The
problem arosein the caseof the secondfigure. All the moods of
I We have no guaranteer,ts Keynes (op. cit., p. z86) justly remarks,
that the
major term will be the largest in extension and the minor the smallest, when one of
thc premisses is negative or particular. Thus, Keynes continues, 'the syllogismNo l14 is P, All S is .il,f, therefore, No S is P-yields
as one case [here there follows
a diagram representing three circles M, P, and.S, a large ,Sincluded in alatger M,
oueide of them a small P] where the major term may be the smallest in extent,
and the middle the largest;' Keynes forgets that it is not the same to draw a small
circle P outside of a large circle S and to maintain that the term P is smaller in
extent than the term .S.Terms can be compared with respect to their extent only
in the case when one of them is contained in the other.

$ rr

T H E H I S TOR Y

OF AN ER R OR

3r

this figure have a negative conclusion and the first two moods,
called later Cesare and Camestres,yield a universal negative
conclusion.From the premisses'M belongs to all "iV' and'M
6X
belongsto no X'follows the conclusion belongsto no.M', and
by conversionof this result we get a secondconclusion' '"lf belongs
to noX'. In both syllogismsMis the middle term; but how are we
to decide which of the two remaining terms, "lf and X, is the
major term and which is the minor? Do major and minor terms
exist 'by nature' ($ioet) or only 'by convention' (9bec)?r
Such problems, according to Alexander, were raised by the
later Peripatetics. They saw that in universal affirmative premissesthere can be a major term by nature, becausein such
premissesthe predicate is larger in extension (2d dtiov) than the
subject,but the sameis not true in universal negativePremisses.2
We cannot know, for instance,which of the terms 'bird' or 'man'
is major, becauseit is equally true that 'no bird is a man' and that
'no man is a bird'. Herminus, the teacher of Alexander, tried to
answerthis question by modifying the meaning of the expression
'major term'. He saysthat of two such terms, 'bird' and 'man',
that is the major which in a systematicclassificationof the animals
6animal'. In our example it is the
is nearer to the common genus
term 'bird'.3 Alexander is right when he rejects this theory and
its further elaboration given by Herminus, but he also rejectsthe
opinion that the major term is the predicate of the conclusion.
'I'he major term, he says,would not be fixed in this case,as the
universal negative premiss is convertible, and what till now has
been a major term instantly becomes a minor, and it would
depend upon us to make the same term major and minor.a His
own solution is basedon the assumptionthat when we are forming a syllogism we are choosing premissesfor a given problem
I Alexander
72. 17 {qteitoq ei $rioec 2v 6eurip,p o1li1tott pei(uv ds 2ott xai
l\i,ttav d.xpos, xai ivt oitos xpt|ioetau
I lbid.
pei(ov d xanlyopoipevos xa06Aou,
72. z4 lni p2v ydp t6v roto$anxin
-6tt
xrrl lni tAiov' 6ri. ,orirou Tdp oJEt ,i,mempi$ec' 6torc $Joet oJnQ ti peilovo elvot
d)10isoixitttoirc
JrlpTtt. dzi 6! triv xa06Aoud'ro$arxtl
'Eppivos oleran, dv iewipE oTipetc tdv pei(,ova dxpov clvet ' ' ,rdv
'' ill,i.l. z7
lyyJrrpuv rJO xocvoayivous aitiw (dv ydp ,iocv oi dxpot dpveov xai dv|potos, dyyu'
tlpo ro0 xowoi ylvous air6v, toi ((ou, rd ipveov rbi dv|p,lnou xai 2v rff npdrp
Ih,uplurq 6d xai pei(uv dxpos ti 6pveov).
llritl. 75. Io tiM'oriSt it).6s n,i\w iqthv peilovodv 2v tQ oupnepd'opanroa
'
ttA)\oytupol xanlyoporlpevov,<is Eorei now' oJ62 yd'p oitos Eflos' d,\,\oze7dp dlos
ltntt xtl oJy dptoltivos tQ dvttotp&$ecv tilv xofliltou ino$arufiv, xoi 6 titts pei(ov
,rrlllrr r',\,irrov, xui d$' i11tivimu tiv crirdv rci peilo xai iA,itro norctv.

g2

THESES O F THE SYST E M

$ rr

conceived as the conclusion. The predicate of this conclusion is

the major term, and it does not matter whether we afterwards
convert this conclusion or not: in the problem as first given the
major term was and remains the predicate.I Alexander forgets
that when we are forming a syllogismwe are not always choosing
premissesfor a given conclusion,but sometimeswe are deducing
new conclusionsfrom given premisses.
The problem was settled only after Alexander. What John
Philoponus writes on the subject deservesto be regarded as
classic.According to him we may define the major and the minor
term either for the first figure alone or for all the three figures
together. In the first figure the major term is the predicate of the
middle and the minor is the subject of the middle. Such a definition cannot be given for the other two figures becausethe relations of the extremesto the middle term are in the other figures
the same. We must therefore accept as a common rule for all
figures that the major term is the predicate of the conclusion and
the minor term is the subject of the conclusion.2That this rule is
only a convention follows from another passageof Philoponus,
where we read that the universal moodsof the secondfigure have
a major and a minor term only by convention, but not by
nature.3
$ n. The ordn of thepremisses
Alound the Aristotelian logic arosesome queer philosophical
prejridiceswhich cannot be explainedrationally. One of them is
directed against the fourth figure, disclosingsometimesa strange
aversion to it, another is the odd opinion that in all syllogisms
the major premiss should be stated first.
r Alexander
75. zG dv 6r) e'r zQ npoxetpivtp npoBAtipan eis r)v 6et(w Ro.rlyopoilroov toito 0etlov pei(ova' xai ydp ei ,ivrcorp&$errci Erd roOto ytverar d arirds xcl
&aoxeipevos,dM' Evye tQ fiptv eisrd Eeifcr zpor etptvr4 xonlyopotipevosfiv te xoi pveu
2 Philoponus 67, rg i}opev tpdtepov xoi ris tqc pei(av 6pos xoi tis iAd.nut.
zo&o 6i 6wet6v p2v *eL xoa6s 2ni rdv rp'r6l oyr1pd,rav 6topioao0ot xai i6ig itti to0
nptittou. xei i6iq. ptv dtl roA np<irov oxtiporos pet(av 6posAoriv 6 toA piaou xor7yopoipevos, iAdrtuv 62 6 tQ pioE iroxeipevos. xai to}ro p2v i6re(6wus ini roi ztptitov
Aiyopev, dzedri ri pioos 2v rQ npdrq rot piv xatqyopeiran zrp 6t rizdrerror. d,\I'
Itet|f xar' oi}itepov t6v iiilov oyqpdtov 6rd.$opovEyovocoyior oi d,xpot np6s riv
ltioovr ifiAov 6rt oixitt dpp&oerfipiv oitos 6 npoolropropis 2t' 2xeivotv.yprlo#ov o$v
xowQ xavdvt izi rdv tptiw o1glpd,tov toJt<p, fur peilcov ioriv 6pos 6 dv rQ oupn<pd,opan xonlyopoipevos, i\dttov Et d e'vr<p ovpnepd,opatt inoxeipevos.
3 lbid. 87. ro rd 6i peilor d.xpov2v toinqt rQ oyfpott nlv EJonpord.oeuv xo06Aov
oio6v oir Eon $ioer ril.lri 0/oer.

$ rz

T H E OR D ER OF TH E PR EM ISSES

gg

From the standpoint of logic the order of the premissesin the

Aristotelian syllogismsis arbitrary, becausethe premissesof the
syllogismform a conjunction and the membersof a conjunction
arc commutable. It is only a convention that the major premissis
statedfirst. Nevertheless,somephilosophers,like Waitz or Maier,
maintain that the order of the premissesis fixed. Waitz censures
Apuleius for having changed this order,r and Maier rejects
'Irendelenburg's opinion that Aristotle doesnot tie it down.z No
arguments are given in either case.
I do not know who is the author of the opinion that the order
of the premissesis fixed. Certainly it is not Aristotle. Although
Aristotle hasnot given a definition of the major and minor terms
valid for all the three figures,it is always easyto determine which
term and which premiss are regarded by him as the major and
which as the minor. Aristotle, in his systematicexposition of the
syllogistic,usesdifferent letters to denotedifferent terms; for each
figure heputs them in alphabeticalorder (04oc)and saysexplicitly
which term is denoted by a given letter. We have thus for the
first figure the letters A, B, C; / is the major term, B the middle,
and C the minor.3 For the secondfigure we have the letters r11,
"ltr,X, where Mis the middle term, -ltfthe major, and Xthe minor.4
For the third figure we have the letters P, R, S, where P is the
major term, R the minor, and ^Sthe middle.s
I Waitz, op. cit., vol. i, p.
3Bo: 'Appuleius in hunc errorem se induci passusest,
ut propositionum ordinem irnmutaverit.'
r Maier, op. cit., vol. ii a, p. 63,: 'Darnach is Trendelenburg's Auffassung,
dnu Aristoteles die Folge der Priimissen frei lasse, falsch. Die Folge der Priirnirren ist vielmehr festgelegt.' It is not clear to me what reasons he refers to by
damuh.
I This follows from the definition grven by Aristotle for the first figurel see
1r,tB, n. L Cf. Alexandet 54. rz Eoro ydp pet(ov plv dxposti A, p,ioos 6t 6pos rti B,
iluitrov El dxpos zd I-.
a An, pr, i.
5, zibg4 &av 6i rri cJrd tQ phv nomi tQ 6i pl26eri indpy1, fi ixuripq
aurl I pqtevi, ri piv oyfi1n rd rocoArov xo)Ct ieitepov, pioov 6t e obQ Aiy<o d
xutr1yopolpeov ,ip$oev, dxpa 6E xaO' 3v \iye.tot ro&to, pei(ou 6l d.xpovt6 tpds tQ
ylo<p xclpevov,ilcrrov Et rd noppotipot toA pioou. d|etat 6t rd p(oov E(o piv fiiv
ilxpuv, npinov 62 tfi 0loeu Cf. Alexander 78. t 1pfitar ydp oroq5iors oi tois A, B, f,
ols r'r rrp nyitE oyfipatr, ri})d zois M, N, E, p,ioov piv AepBd.vuvtd M rd dp$ottpov
xrrqyoporipevov xoi riy.nptitqv Elgovtdfw iv tfi xoraypaSfi,. pci(ovo 62 dxpov i N
J$tffls xttpcvov petd. tiv ploov, Eoyorov 6t xoi ilvinova rd 8.
. An. pr. i. 6, z8aro drir Et r<! ainQ i piv nawi rri Ei p76eli hdpxn,
i dpinwtl I pqievl, rd piv oyfi1to td torcStov xoAit rphov, pioov E' dv oirQ llyat xa|' oi
llp,fu rd xatqyopoJpeve, drpc 6t td, xatlyopotlpeva, petlov 6' d,xpovtd toppcirepov
ro0 pluou, f)crroy Et td iyyitepov. d0erar 6i td ptoov E[o plv t6w d.xpov, EaTatov
ll rt,l 0lott, Cf. Alexander 98. eo dzi roJtou toi oTfipatos ndAcv ypfitar oroqge(n6

THESES O F THE SYSTEM

$rz

Aristotle states the major premiss first in all the moods of

the first and the second figure, and in two moods of the third
figure, Darapti and Ferison.rIn the remaining moodsof the third
figure, Felapton, Disamis, Datisi, and Bocardo, the minor premiss is.stated first.2 The most conspicuousexample is the mood
Datisi. This mood is formulated in the same chapter twice; in
both formulations the letters are the same,but the premissesare
inverted. The first formulation runs: 'IfR belongsto someS, and
P to all S, P must belong to some .R.'3The first premiss of this
syllogismis the minor premiss,for it containsthe minor term.R.
The secondformulation reads: 'If P belongs to all $ and .R to
sorneS, then P will belong to someR.'+ The first premissof this
second syllogism is the major premiss, as it contains the major
term P. Attention must be called to the fact that this second
formulation is given only occasionally, while the standard formula of this mood, belonging to the systematic exposition, is
enunciated with transposedpremisses.
In Book II of the Pior Anafuticswe meet other moods with
transposedpremisses,as Darii,s Camestres,6Baroco.TEven Barbara, the main syllogism,is occasionallyquoted by Aristotle with
the minor premissfirst.8I can hardly understand,in view of these
examples,how somephilosophersknowing the Greek text of the
Organoncould have formed and maintained the opinion that the
order of the premissesis fixed and the major premiss must be
stated first. It seemsthat philosophical prejudices may sometimestestroy not only common sensebut alsothe faculty ofseeing
facts as they are.
I t3. Enors of somemoderncommentators
The story of the fourth figure may serveas another example to
tots II, P, E, xai otv airQ ro0 F2v Fet(ovosd.xpovolpavnxdv td If, zo0 Et dldrroyos
xoi d{ei\ovtos Jnoxeiolat dv rQ ywoltivo ouptepd.opan ri P, zo0 6l ptoou $ E.
I See, for instance, p.
3, n. 2 (Barbara) and p. ro, n. z (Ferio).
' See p. 9, n. 4 (Felapton),,and p. 7, n. (Disamis).
t
4l. F. i. 6, a8brz ei td piv P twi tQ 2 i 6i If nevti Jndpyeq dvdyx7 td II wi
.
I
t Q P $ndp 2 6 w .
r lbid. z8br6 ei ydp novti i II rQ E ird.pyeqzri 6i P nvi te 2, xo,i rd II
twl
tQ P indp(ct.
s Ibid. ii. r r, 6rb4r ei ydp i A nvi zQ B, rd
6t f navti te A, tli ze B td I
$nip(et.
6 lbid. ii. B, 6oa3 ei td A plievi rQ f, tQ 6i B navti, oiievi te I i B.
7 lbid. 6oz5 ci ydp td A nvi ,Q F pi imlpyeq ?4 Ei B
trt!,
rd B nvi te I oiy
t Se e p , r o , n .5 .
J ndp{eu

$ 13

E R R O R S O F S O M E M OD ER N

C OM M EN TATOR S

35

show how strange philosophical prejudices sometimesare. Carl

Prantl, the well-known historian of logic, beginshis consideration
of this figure with the following words: 'The question why silly
playthings,as,for instance,the so-calledGalenianfourth figure,
are not to be found in Aristotle, is one we do not put at all; it
plainly cannot be our task to declare at everystep of the Aristotelian logic that this or that nonsensedoesnot occur in it.'r Prantl
does not seethat Aristotle knows and acceptsthe moods of the
so-called Galenian fourth figure and that it would be a logical
error not to regard these moods as valid. But let us go farther.
Commenting upon the passagewhere Aristotle speaksof the two
moods later called Fesapoand Fresison,zPrantl first statesthese
moods as rules of inference:
All ^Bis I
N oC i sB

SomeB is,4
NoCis. B

Some I is not C

Some./ is not C

-he doesnot, of course,seethe differencebetweenthe Aristotelian

and the traditional syllogism-and then he says: 'By transposition of the major premiss and the minor it becomes possible
for the act of reasoningto begin'; and further: 'Such kinds of
reasoningare, of course,not properly valid, becausethepremisses
ordered as they were before the transpositionare simply nothing
for the syllogism.'r This passagereveals,in my opinion, Prantl's
entire ignorance of logic. He seems not to understand that
Aristotle proves the validity of these moods not by transposing
the premisses,i.e. by inverting their order, but by converting
them, i.e. by changing the placesof their subjectsand predicates.
I Carl Prantl, Gcschichtcder Logik in Abcndlande,
vol. i, p. z7z:,Die Frage aber,
warum einfiiltige Spielereien, wie z. B. die sog. Galcnische vierte Figur, sich bei
Aristoteles nicht finden, werfen wir natiirlich gar nicht auf;
wir kcinnen
selbstverstiindlicher Weise qricht die Aufgabe haben, bei jedem Schritte der
aristotelischen Logik,eigens anzugeben, dass diesel oder jener Unsinn sich bei
Aristoteles nicht finde.'
' See p. 25, n. 2.
3 Prantl, op. cit., vol. i, p. e76:
'Alles B ist A
EinigesB ist A
Kein C ist B
Kein C ist B
Einiges A ist nicht C
Einiges A ist nicht C
woselbst durch Vertauschung des lJntersatzes mit dem Obersatze es mriglich wird,
<lassdie Thiitigkeit des Schliessens beginne; . . . nattirlich aber sind solches keine
cigenen berechtigten Schlussweisen, denn in solcher Anordnung vor der Vornahme
tler Vertauschung sind die Priimissen eben einfach nichts fiir den Syllogismus.'

36

THESES O F THE SYSTEM

$ tg

Moreover, it is out of place to saythat, two premissesbeing given,

the act of reasoningbegins when one premissis stated first, but
no syllogism results when the other precedes.Frorn the standpoint of logic Prantl's work is useless.
The same mav be said of Heinrich Maier's work. His treatise
on the syllogisticfigixes generally and the fourth figure in particular is in my opinion one of the most obscure chapters of his
laborious but unfortunate book.t Maier writes that two opinions
of the criterion for the syllogistic figures stand opposed to each
other: one (especiallyUeberweg) seesthis criterion in the position of the middle term as subject or predicate, the other (especially Trendelenburg) seesit in the extensionalrelations of the
middle term to the extremes.It is not yet settled,Maier says,
which of theseopinionsis right.z He adoptsthe secondashisown,
relying on Aristotle'Scharacterizationof the first figure. We know
already that this characterizationis logically untenable. Maier
not only acceptsit, but modifiesthe Aristoteliancharacterizations
of the two other figures according to the first. Aristotle describes
the secondfigure somewhatcarelesslyas follows: 'Whenever the
sameterm belongsto all of one subject and to none of the other,
or to all ofeachsubject,or to none ofeither, I call such a figure
the second; by "middle term" in it I mean that which is predicated of both subjects,by "extremes" the terms of which this is
said.'3 Maier remarks: 'When we reflect that the expressions
".8 is included in .d", ".4 belongsto 8", and "A is predicated of
B" areninterchangeable,then we may put this characterization
according to the description of the first figure in the following
words.'aMaier commitshere his first error: it is not true that the
three expressionshe quotes can be exchanged for each other.
Aristotle statesexplicitly: 'To say that one term is included in
another is the sameas to say that the other is predicatedof all of
the first.'s The expression'-B is included in z4'means,therefore,
I See Maier, op. cit., vol. ila, 'Die drei Figuren', pp.
47-7t, and vol. ii6,
'Erg?inzung durch eine 4. Figur mit zwei Formen', pp. z6r-9.
" Op, cit., vol. iia, p. 48, n. r.
3 See the Greek text on p.
33, n. 4.
{ Op. cit., vol. ii a, p.
49: 'Erwiigt man nd,mlich, dassdie Ausdriicke "B liegt im
Umfang von A", "A kommt dem Begriff B zu" und "A wird von B ausgesagt" mit
einander vertauscht werden kcinnen, so lisst sich die Charakteristik der zweiten
Figur, welche der Beschreibung der ersten parallel gedacht ist, auch so fassen,'
5 An, pr. i. I, e4be6 td Et dv 6Aqt etvar Etepov &(p<p xai rd xard. navris xortlyopeio|at |otipou |dtcpov ralriv iorw.

$ ':

E R R O R S O F SOM E M OD ER N

C OM M EN TATOR S

37

the same as 'r4 is predicated of all ,B' or '.4 belongsto all .B', but
doesnot mean'A is predicatedof ,B'or'l belongsto,B'. With
this first error is connecteda second:Maier maintains that the
negative premiss also has the external form of subordination of
one term to another, like the affirmative universal premiss.t
What is heremeant by 'externalform'? When r4 belongsto all B,
then,Bis subordinatedtoA, and the externalform of this relation
is just the proposition ',4 belongs to all .B'. But in a negative
premiss,e.g.'A belongsto no .B',the subordinationof termsdoes
not exist, nor doesits form. Maier's assertionis logically nonsense.
Let us now quote Maier's description of the secondfigure. It
runs thus: 'Wheneverof two termsone is included,and the other
is not included, in the samethird term, or both are included in it,:a
or neither of them, we have the second figure before us. The
middle term is that which includes both remaining terms, and
the extremesare the terms which are included in the middle.',
This would-be characterization of the second figure is again
logically nonsense.Take the following example: Two premisses
are given: '24belongsto all .B' and 'C belongsto no A'. lf A
belongsto all B, then .B is included in r4, and if Cbelongs to no
l, it is not included in ,4. We have thereforetwo terms, B and,C,
one of which, ,8, is included, and the other, C, is not included in
the samethird term ,4.According to Maier's descriptionwe should
have the secondfigure before us. What we have, however, is not
the secondfigure, but only two premisses'1 belongsto all -B' and
'C belongsto no A', from which lte can get by the mood Celarent
of the first figure the conclusion'C belongsto no .B', and by the
mood Camenes of the fourth figure the conclusion '.8 belongs
to no C'.
The peak, however, of logical absurdity Maier attains by his
assertionthat there existsa fourth syllogistic figure consistingof
only two moods, Fesapoand Fresison.He supports this assertion
by the following argument:.'The Aristoteliandoctrine overlooks
one possibleposition of the middle term. This term may be less
' Op, cit., vol, ii a, p. 6o, n. r: 'auch der negative syllogistische Satz hat
wcnigstens die iussere Form der Subordination.' Cf. also ibid., p. 5o.
2 Ibid., p.
49: 'Wenn im Umfang eines und desselbenBegriffes der eine der
br:iden tibrigen Begriffe liegt, der andere nicht liegt, oder aber beide liegen oder
gndlich beide nicht liegen, so haben wir die zweite Figur vor uns. Mittelbegriff ist
flcrjenige Begrifl in dessen Umfang die beiden iibrigen, iiu8ere Begriffe aber diejcnigen, die im Umfang des mittleren liegen.'

38

THESES O F THE SYSTEM

$ rg

general than the major and more general than the minor, it may
secondly be more general, and thirdly less general, than the
extremes,but it may be also more general than the major term
and at the same time lessgeneral than the minor.'I When we
remind ourselves that according to Maier the major term is
alwaysmore generalthan the minor,2 and that the relation 'more
general than' is transitive, we cannot avoid the strange consequenceof his argument that the middle term of his fouith figure
should be at the sametime more and lessgeneral than the minor
term. From the standpointof logic Maier's work is useless.
I 14. Thefour Galenianfgures
In almosteVerytext-book oflogic you may find the remark that
the inventor of the fourth figure was Galen, a Greekphysician and
philosopherliving in Romein thesecondcenturye.o. The sourceof
this remark is suspect.We do not find it either in the extant works
of Galen or in the works of the Greek commentators (including
Philoponus).Accordingto Prantl the medievallogiciansreceived
the information from Averroes, who saysthat the fourth figure
wasmentioned by Galen.r To this vague information we may add
two late Greek fragments found in the nineteenth century, and
alsovery vague. One of them was published in rB44 by Mynas in
the preface to his edition of Galen's Introductionto Dialectic, and
republishedby Kalbfleisch in r8g7. This fragment of unknown
authorship tells us that somelater scholarstransformedthe moods
added by Theophrastus and Eudemus to the first figure into a
new fourth figure, referring to Galen asthe father of this doctrine.+
The other Greek fragment was found by Prantl in a logical work
t _ O - p . - " ] t . ,r - o ] . i i b ,p .z6 4 :
' Die a r isto te lisch eL e h r e lii0 t ei ne mri gl i cheS tel l ung
.
des Mittelbegriffs unbeachtet. Dieser kann specieller als der ober- und allgemeinei
als der unterbegriff, er kann ferner allgemeiner, er kann drittens spcciellir als die
beiden-iuBeren Begriffe: aber er kann auch allgemeiner als der Obei- und zugleich
specieller als der Unterbegriff sein.'
2 lbid., vol. ii a, p.
56: 'Oberbegriffist stets, wie in der r.Figur ausdri.icklich
festgestellt ist, der allgemeinere, Unterbegriff der weniger allgemiine.'
3 Prarrtl, i.
57r, n. gg, quotes Averroes in a Litin tianslation edited in
Venice ( r 5-!3): 'Et ex hoc planum, quod figura quarta, de qua meminit Galenus,
non est syllogismus super quem cadat naturaliter cogitatio.' cf. also prantl, ii.
39o, n. 322.
a K. Kalbfleisch, Aba
Galcns Einleitung in diz Logik,23. Supplementband der
_
Jahrbi.icher fiir klassischePhilologie, Lelpzig (t8SZj, p. 7o7:'@r6$paoros Dt rci
Ei-dr1po1xai rryos &tpas ou{vytes nopd ,ds ineeelios ,e Aprotoriltei npooreltixaoc
t! npdrq oyrilrorr . . ., ds *oi t(taptov d"note)\etvoyfi1tottiv'veuripov <inilrlotiv'rcves
<,15zpris rttipa t)1v 66{av tdv fo\lvdv dve$ipomei,'"

$r+

T H E F OU R GAL EN IAN

FIGU R ES

39

ofloannes Italus (eleventh century A.D.). This author sayssarcastically that Galen maintained the existenceof a fourth figure
in opposition to Aristotle, and, thinking that he would appear
cleverer than the old logical commentators, fell very far short.r
That is all. In view of such a weak basis of sources,Ueberweg
suspecteda misunderstandingin the matter, and Heinrich Scholz
writes inhis Histor2of Logicthat Galen is probably not responsible
for the fourth figure.2
For fifty years there has existed a Greek scholium in print
which clearsup the whole matter in an entirely unexpectedway.
Although printed, it seemsto be unknown. Maximilian Wallies,
one of the Berlin editors of the Greek commentarieson Aristotle,
published in IBgg the extant fragments of Ammonius' commentary on the Prior Anal2tics,and has inserted in the preface a
scholium of an unknown author found in the samecodex as that
in which the fragmentsofAmmonius are preserved.The scholium
is entitled 'On all the kinds of syllogism', and begins thus:
'Therearethreekindsofsyllogism: thecategorical,the hypothetical,
Of the categoricalthere are two
and the syllogismxdrd.rp6oh1$cv.
kinds: the simpleand the compound.Of the simplesyllogismthere
arethreekinds: the first, the second,and the third figure.Of the compound syllogismthere are four kinds: the first, the second,the third,
and the fourth figure. For Aristotle saysthat there are only three
figures,becausehe looksat the simplesyllogisms,consistingof three
that there are four figterms.Galen,however,saysin his Apodeictic
ures, becausehe looksat the compoundsyllogismsconsistingof four
terms, as he has found many suchsyllogismsin Plato'sdialogues.'3
The unknown scholiastfurther gives us someexplanations,from
t Prantl, ii.
ttirtc'd I'clTrds 6i rci
3oz, n. t tz: rri 6t oyfipata tdn oil,)oyopir
t(taprov dni rorhoc ESooxo clvoq lvantas rpds tiv Etayetpinlv $ep6pcvos,6s Aop'
na)an6r tis
i{qyoupeuv
rpdrepov dva$avfivec oiipoos rdv tiv Aoycxiy rpaypa*iov
toppurdta eJ9tus ixniffiuxe
" Fr. Ueberweg, Sysbmdzr Lagik, Bonn (rBBz),34t. Cf' also Kalbfleisch, op. cit.,
dn Logik, Berlin (r93t), p. 36.
p. 699; H. Scholz, Geschichta
r M. Wallies, Ammonii in Arist'otclis Analyticorum Priorum librum I Commcntaium,
Ilerlin (I89g), p. ix: Ilepi tCw et66w nd'vtav toi ouMoytopoi. tpk ei6q imi roA
[rirlo0] oulloTopoA' td xarqyopxdv, zd ino|etudv, ri xatd. npddtq$cv. n0 6i
xatqyopcxoi }Jo ioriv ei6q' dnAoav, oJv|erov. xal toA piv iillo0 rpio oiv ci6q'
apdrov o1gfipa,ieitepov o21fipo,tpitov ayfipo. toA 6t ouvilnv #ooapd, imrv eiErl'
rp<irov o7fipa, Ee,itepovoyfipo, rpirov, tbaptov oyfipa. Aprr.totiAqs plv yd,p qia
rd oyfiporri t$qow np6s zoJs dzlo0s ouDoyrcpois dnoB)tinov tois ix tpGw dpuv
uuyxerpivovs. J-clryds D' dv rfi oixeig Ano\erxnxf 5 rd oTfipara )iyet tpds tols
ow\(tous ov))oyrcpois ,inoB[1.l.av zor)s e'x 5 6pov ovyxetpiwus ro)lods totoJrovs
cipdv tv roes llAdtovis EraldTor5.

THESES O F THE SYSTEM

40

$ r4

which we can gather how Galen may have found these four
figures. Compound syllogismsconsisting of four terms may be
formed by combinations of the three figures I, II, and III of
simplesyllogismsin nine different ways: I to I, I to II, I to III,
II to II, II to I, II to III, III to III, III to I, III to II. Two of
thesecombinations,viz. II to II and III to III, do not give
syllogismsat all, and of the remaining combinations II to I gives
the samefigure as I to II, III to I the sameas I to III, and III
to II the sameas II to III. We get thus only four figures,I to I,
I to II, I to III, and II to III.r Examplesare given, of which
three are taken from Plato's dialogues,two from ttre Alcibi.ades,
and one from the Republic.
This precise and minute account must be explained and
examined. Compound syllogismsof four terms have three premissesandtwo middle terms,sayB and C, which form the premiss
B-C or C-B.Let us call this the middle premiss.,Bforms together
with A, the subject of the conclusion, the minor premiss,and C
forms togetherwith D, the predicate of the conclusion,the major
premiss.We thus obtain the following eight combinations (in all
the premissesthe first term is the subject, the secondthe predicate) :
Figure
FI
Fer

F3
F4
F5
F6
F7
FB

A_B
A_B
A-B
A-B
B_A
B_A
B-A
B-A

C_D
D-C
C-D
D_C
C-D
D.C
C-D
D.C

I
I
I
I
I
I
I
I

A-D
A-D
A_D
A-D
A-D
A.D
A-D
A_D

ItoI
ItoII
II to III
IItoI
III to I
III to II
I tO III
ItoI

If we adopt the principle of Theophrastus that in the first

I Wallies, op. cit., pp, ix-x! 6 xanlyoptxds ov\Aoytopds drlo0s, cis Apototi)o1s'
oyfipo A B f. oriv|etos, ds fa\r1vds A npds A, A rpds B, A npds f, B npds B, B
npds A, B npdsl, f npis f, I npds A, I npds'8.
A zp d s A, A n p 6 sB, A n p d sT, B npdsl .
. y_. o T. !..
o_urA\ -A- o
L No v' ABr A
dno<$anx6v
dovAAdyntov' B npdsB, I npdsf, (oi ydp ytvetar ovA\oyropdsoire ix 613o
oite dx 6Jo pepwtw)'
B npds A, T npds A, I npis B,
BT A
oi oJroi eiow tois oviloynpois tis inoyypattoc.

$ t4

T H E F OU R GAL EN IAN

FIGU R ES

4l

Aristotelian figure the middle term is the subjectof one premissit does not matter of which, the major or the minor-and the
predicate of another, and define by this principle which figure is
formed by the minor and middle premisseson the one hand, and
by the middle and major premisseson the other, we get the combinations of figures shown in the last column. Thus, for instance,
in the compound figure Fz the minor premisstogether with the
middle forms the figure I, as the middle term,B is the predicate
of the first premissand the subject of the second,and the middle
premisstogetherwith the major forms the figure II, as the middle
term C is the predicate of both premisses.This was probably how
Galen has got his four figures. Looking at the last column we see
at oncethat, as Galen held, the combinationsII to II and III to
III do not exist, not for the reason, as the scholiastmistakenly
says,that no conclusion results either from two negative or two
particular premisses,but because no term can occur in the
premissesthree times. It is obvious also that if we extend the
principle of Theophrastus to compound syllogismsand include
in the samefigure all the moods that from the samecombination
of premissesyield either the conclusion A-D or the conclusion
D-4, we get as Galen doesthe samefigure from the combination
^ I to II as from the combination II to I. For, interchanging in
figure F4 the letters ,B and C as well as the letters A and D, we
get the scheme:

F4

D-C

B.C

A_B

D_4,

and as the order of the premissesis irrelevant we see that the

conclusion D-r4 results in F4 from the same premissesas A-D
in Fz.. For the samereasonfigure Fr doesnot differ from figure
IrB, F3 from F6, or F5 from F7. It is possible,therefore, to divide
the compound syllogismsof four terms into four figures.
'fhe scholium edited by Wallies explainsall historical problems
r:onnected with the alleged invention of the fourth figure by
Galen. Galen divided syllogismsinto four figures, but thesewere
tltc compoundsyllogismsof four terms,not the simplesyllogisms
of Aristotle. The fourth figure of the Aristotelian syllogismswas
invcnted by someoneelse,probably very late, perhapsnot before
tlre sixth century a.n. This unknown scholar must have heard
sotnetltingabout the four figuresof Galen, but he either did not
tnrcl<:rst;rnrl
thcm or did not have Galen'stext at hand. Being in

42

THESES O F THE SYSTE M

$ I4

opposition to Aristotle and to the whole school of the Peripatetics, he eagerly seized the occasion to back up his opinion by
the authority of an illustrious name.
RBIrr,tnx. The problem of compound syllogismsraised by Galen has
considerable interest from the systematic point of view. fnvestigating
the number of valid moods of the syllogismsconsisting of three premisses,I have found that theri are forty-four valid moods, the figures
FI, Fz, F4, F5, F6, and F7 having six moods each, and figure F8
eight. Figure F3 is empty. It has no valid moods, for it is not possible
to find premissesof the form A-8, C-8, C-D such that a conclusion
of the form l-D would follow from them. This tesult, if known, would
certainly be startling for studentsof the traditional logic. Mr. C. A.
Meredith, who attended my lectures delivered on this subject in 1949
at University College, Dublin, has found some general formulae
concerning the number of figures and valid moods for syllogismsof z
terms, including expressionsof l and 2 terms. I publish theseformulae
here with his kind permission:
Number of terms
Number of figures
.
2n-l
Number of figures with valid moods . [(n2-n-lz)
Nrimber of valid moods
. n (3n-r)
For all n .+'rerynon-empty figure has 6 valid moods, except one that
has za valid moods.
Examples:
Number of terms
. r, 2, 3, 4r,,,, ro
Number of figures
r, 2, 4, 8,..,, 5rz
Number of figureswith valid moods . r, 2, 4, 7,..., 46
Number of valid moods
. 2, ro,2+, ++,.,., 2go
It is obvious that for large z's the number of figureswith valid moods
is comparatively small against the number of all figures. For z: ro
we have 46 against 5rz respectively,i.e.466 figures are empty.-For
n: I there is only r figure, A-A,with z valid moods,i.e. the laws of
identity. For z:2
there are z figures:
.
Premiss Conclusion
Fr "
A-B
A-B
Fz
B-A
A-B
with ro valid moods, 6 in Fr (viz. four substitutionsof the propositional law of identity, e.g. 'if all A is B, then all A is'B', and two laws
of subordination), and 4 moods in Fz (viz. four laws of conversion).

CHAPTER III

TH E S Y S TE M
I rg. Perfea and imperfectslllogisms
In the introductory chapter to the syllogisticAristotle divides all
syllogismsinto perfect and imperfect. 'f call that a perfect syllogism', he says, 'which needsnothing other than what has been
stated to make the necessityevident; a syllogism is imperfect, if
it needseither one or more componentswhich are necessaryby
the terms set down, but have not been stated by the premisses.'r
This passageneeds translation into logical terminology. Every
Aristotelian syllogism is a true implication, the antecedent of
which is the joint premissesand the consequentthe conclusion.
What Aristotle saysmeans,therefore, that in a perfect syllogism
the connexion between fhe antecedent and the consequent is
evident of itself without an additional ppoposition.Perfect syllogismsare self-evidentstatementswhich do not possess
and do not
need a demonstration; they are indemonstrable, dvan66etrctot.2
Indemonstrable true statementsof a deductive system are now
called axioms.The perfect syllogisms,therefore,are the axioms of
the syllogistic.On the other hand, the imperfect syllogismsare not
self-evident;they must be proved by meansof one or more propositionswhich result from the premisses,but are different from
them.
Aristotle knows that not all true propositions are demonstrable.3He saysthat a proposition of the form '24belongsto ,B'
is demonstrable if there exists a middle term, i.e. a term which
forms with A and.B true premissesof a valid syllogismhaving the
above proposition as the conclusion. If such a middle term does
I An. pr, i. t, z4bzz riAerou
ltiv o\v xaAit ou&oyopdv zdv plDads dDou rpoo8e&
pcvov nopd.rd eiArlppiva rpds td <fovfivot ri dvayxaiov, <irelf 6t z6v npooiedpevov i
tr6s i rAedvav, d ion piv d,veyxata 6rd tiv inoxeqtlvov 6pov, oJ piy eiAwtor 6rd
np6raoeav.
2 Commenting upon thE above passage Alexander
uses the expression 6.vendiecxros, 24. z: ivds ptv oiv npooi(ovrdr of dzelcis ouiloyrcpoi oi, ptis dlrrotpo$fis
6e6pcvot rpds rd ,ivoxOivec ets tuo tdv tv tQ npdtE oTtipan rCw te\el,uv xoi d,vano6clxtav, nAct6vqv 6t 6oo1 6rti 6,io
'iwrcrpo$dv eis dxeivav rwd, d.vdyovtau Cf. also
p. 27, n. 2r An. post. i.
3, 7zbr8 ripeis Di $apev oite n6oav inwtzipqv dnoie*ru<lv etvan,
<i,\,\<irr)r niv d.piouv d.van66eqtov.

THE SYSTEM

44

0 rS

not exist,the propositionis called 'immediate',d.y.eoos,

i.e. without a middle term. Immediate propositionsare indemonstrable;
they are basic truths, dpyai.t To thesestatementsof the Posterior
Analyticsmay be added a passageof the Prior Anafittrcswhich
states that every demonstration and every syllogism must be
formed by meansof the three syllogisticalfigures.2
This Aristotelian theory of proof has a fundamental flaw: it
supposesthat all problems can be expressedby the four kinds
of syllogistic premiss and that therefore the categorical syllogism is the only instrument of proof. Aristotle did not realize
that his own theory of the syllogism is an instance against this
conception. The syllogistic moods, being implications, are propositions of another kind than the syllogistic premisses, but
neverthelessthey are true propositions,and ifany of them is not
self-evidentand indemonstrableit requires a proof to establishits
truth. The proof, however, cannot be done by means of a categorical syllogism,becausean implication doesnot have either a
subjector a predicate,and it would be uselessto look for a middle
term between non-existent extremes.This is perhaps a subconsciouscauseof the special terminology Aristotle usesin the doctrine of the syllogistic figures. He does not speak of 'axioms' or
'basic truths' but of 'perfect syllogisms',and does not 'demonstrdte' or 'prove' the imperfect syllogisms but 'reduces' them
(d"vd.yet
or d.va)t,1e,.)
to the perfect. The effects of this improper
terminology"persist till today. Keynes devotes to this matter a
whole sectionof his FormalLogic,entitled 'Is Reduction an essential part of the Doctrine of the Syllogism?', and comes to the
conclusion 'that reduction is not a necessarypart of the doctrine
of the syllogism,so far as the establishmentof the validity of the
different moodsis concerned'.rThis conclusioncannot be applied
to the Aristotelian theory of the syllogism, as this theory is an
axiomatized deductive system, and the reduction of the other
syllogistic moods to those of,the first figure, i.e. their proof as
theoremsby meansof the axioms, is an indispensablepart of the
system.
Aristotle accepts as perfect syllogisms the moods of the first
I An. post.i. 23, B4bIg
$avepdv62 xai 6rq 6tav rd A rQ B inlpTp, ei piv Eort tc
lrioov, Eott 6et(at 6n rd A rQ B indpyec. . ., ei Et pri iorw, oixitt Eotw dzdEerfrs,
<il,\' { dri zris cipXds 666s aitq dcriv.
2 An. pr. i. z3,4tbr n6oav dnd6e$w xei n6lta ovAAoyrcpdvdvdyxl ytveoilar 6rd.
tP.@v tav

tPoLPnpcvttv

sxlpef@v.

3 Op, cit., pp.

325-7.

I tS

P E R F E C T A N D IM PER FEC T

SYL L OGISM S

+S

figure, called Barbara, Celarent, Darii, and Ferio.r Yet in the

last chapter of his systematicexposition he reducesthe third and
fourth moods to the first two, and takestherefore as axioms of his
theory the most clearly evident syllogisms,Barbara and Celarent,2This detail is ofno little interest. Modern formal logic tends
to reduce the number of axioms in a deductive theory to a
minimum, and this is a tendency which has its first exponent in
Aristotle.
-that
Aristotle is right when he says
only two syllogisms are
neededas axioms to build up the whole theory of the syllogism.
He forgets, however, that the laws of conversion,which he uses
to reduce the imperfect moods to the perfect ones,also belong to
his theory and cannot be proved by means of the syllogisms.
There are three laws of conversion mentioned in ttre Prior
Analltics: the conversionof the E-premiss,of the ,4-premiss,and
of the /-premiss. Aristotle proves the first of theselaws by what
he calls ecthesis,which requires, as we shall seelater, a logical
processlying outside the limits of the syllogistic. As it cannot be
proved otherwise,it must be stated as a new axiom of the system.
The conversionof the l-premiss is proved by a thesisbelonging
to the square of opposition of which there is no mention in the
Prior Analttics.We must therefore accept as a fourth axiom either
this law of conversion or the thesis of the square of opposition,
from which this law follows. Only the law of conversion of the
I-premissescan be proved without a new axiom.
There are still two thesesthat have to be taken into account,
although neither of them is explicitly stated by Aristotle, viz. the
laws of identity:'A belongsto all A' and'A belongsto somez4'.
The first of these laws is independent of all other thesesof the
syllogistic. ff we want to have this law in the system,we must
accept it axiomatically. The second law of identity can be
derived from the first.
Modern formal logic distinguishesin a deductive system not
only between primitive and derivative propositions, but also
between primitive and defined terms. The constants of the
(to
Aristotelian syllogisticare the four relations:
belong to all'
' At the end ofchapter 4, containing the moods of the first figure, Aristotle says,
An. Pr. i.4, z6be9 6f,\or 6t xri. 6n ndlres of dv ei'tQ ouiloywpoi riAeni eiow.
' Ibid. 7, zgbt Eon 62 xai dveyayeCvtdvras tois ouAloynporis eis zor)s dr zr.i
nyirq oyfi pen xq06)\ou outrA,oyroltoris.

THE SYSTEM

5rs
or A, 'to belong to none' or E, 'to belong to some' or d and 'to
not-belong to some' or O. Two of them may be defined by the
other two by means of propositional negation in the following
way:'A doesnot belongto someB'means the sameas 'It is not
true that./ belongsto all .B', and'A belongsto no.B' meansthe
same as 'It is not true that ,4 belongsto some.B'. In the same
manner.d could be definedby O, and lby E. Aristotle doesnot
introduce these definitions into his system, but he uses them
intuitively as arguments of his proofs. Let us quote as only one
example the proof of conversionof the .I-premiss.It runs as follows: 'If z4belongsto some.8, then .B must belong to somel.
For if .B should belong to no l, r4 would belong to no 8.'I It is
obvious that in this indirect proof Aristotle treats the negation of
IB belongsto someI' as equlvalent to '.8 belongsto no 7'. As to
the other pair, A and O, Alexander saysexplicitly that the phrases
'to not-belong to some' and 'to not-belong to all' are different
only in words, but have equivalent meanings.2
ff we accept as primitive terms of the systemthe relations ,4
and f, defining E and O by meansof them, we may, as-I stated
many years ago,3build up the whole theory of the Aristotelian
syllogism on the following four axioms:
r. .d belongsto all A,
z. I belongsto someL
3. lf A belongsto all .Band .B belongsto all C, then A
belongsto all C.
Barbara
4. lf A belongsto all .Band C belongsto some.8, then
I belongsto someC.
Datisi
46

It is impossible to reduce the number of these axioms. In

particular they cannot be derived from the so-called dicnm dc
omniet nullo. This principle is differently formulated in different
text-booksof logic, and alwaysvery vaguely. The classicformulation, 'quidquid de omnibusvalet, valet etiam de quibusdamet de
singulis' and 'quidquid de nullo valet, nec de quibusdam nec de
I An. pr. i. z, z5azo ei ydp 16 A nvi rQ B, xoi rd B r*i tQ A ,ivdyxy
indpyeu. el
yd.p pq}evi, oriEi rd .d ori8evir<! B. [Corr. by W. D. Ross.]
2 Alexander 84. 6 zd twi pi1 Srdpyer ioov iluvd.pevov
tQ tli novti xoti. tfiv A({w
6ta$ipeu
3
J. Lukasiewicz, Ehmmty logiki matemattcznej(Elements of Mathematical Logic),
edited by M. Presburger (mimeographed), Warsaw- (rgzg), p. t7z; 'Znaczenie
analizy logicznej dla poznania' (Importance of Logical Analysis for Knowledge),
Przcgl. Filoz. (PhilosophicalRcuicu), vol. xxxvii, Warsaw (tgg+), p. g7S,

$ rS

P E R F E C T AN D IM PER FEC T

SYL L OGISM S

+7

singulisvalet', cannot be strictly applied to the Aristotelian logic,

as singular terms and propositions do not belong to it. Besides,
I do not seehow it would be pOssibleto deduce from this principle the laws of identity and the mood Datisi, if anything at all
can be deducedfrom it. Moreover, it is evident that it is not one
single principle but two. It must be emphasizedthat Aristotle is
by no meansresponsiblefor this obscureprinciple. It is not true
that the dictumdeomniet nullo was given by Aristotle as the axiom
on which all syllogistic inference is'based, as Keynes asserts.rIt
is nowhere formulated in the Prior Ana['tics as a principle of
syllogistic. What is sometimesquoted as a formulation of this
principle is only an explanation of the words 'to be predicated of
all'and 'of none'.2
It is a vain attempt to look for the principle of the Aristotelian
logic, if 'principle' means the same as 'axiom'. If it has another
meanipg, I do not understand the problem at all. Maier, who
has devoted to this subject another obscurechapter ofhis book,3
spins out philosophic speculationsthat neither have a basis in
themselvesnor are supported by texts of the Prior Anal2tics.From
the standpoint of logic they are useless.
| fi. The logic of termsand the logic of propositions
To this day there existsno exact logical analysisof the proof!
Aristotle gives to reduce the imperfect syllogismsto the perfect.
The old historians of logic, like Prantl and Maier, were philosophersand knew only the 'philosopliical logic' which in the nineteenth century, with very few exceptions,was below a scientific
level. Prantl and Maier are now dead, but perhaps it would not
be impossible to persuade living philosophersthat they should
ceaseto write about logic or its history before having acquired a
solid knowledgeof what is called 'mathematicallogic'. It would
otherwisebe a wasteof tinie for them as well as for their readers.
It seemsto me that this point is of no small practical importance.
No one can fully understand Aristotle's proofs who does not
know that there exists besidesthe Aristotelian system another
systemof logic more fundamental than the theory of the syllogism.
I Op. cit., p.
3o.I.
z An. pr. i. r, z4br8 \iyopev 62 rd xotd nrrds xatqyopetoflaq Srav pq62vfi \aBe?v
froa Jnoxerpi.vou (secl. W. D. Ross)], xa9' oE |dtepov oi \ey?fiaetat' xqi td xotd
3 Op. cit., vol. ii D, p. t4g.
plEerds tioorhos.

THE SYSTEM

48

$ 16

I-_t! m" logic of propositions. Let us explain by an example the

difference between the logic of terms, of which the Arist-otelian
logic is only a part, and the logic of propositions. Besidesthe
Aristotelianlaw of identity '/4 belongsto all.,{'or ,All Ais A,,we
have still another law of identity of the form 'Ifp, theny''. Let us
comparethesetwo, which are the simplestlogical formulae:
All A is A

and

If p, then p.

They differ in their constants,which I call functors: in the first

formula the functor reads'all-is', in the second ,if-then'. Both
are functors of two arguments which are here identical. But the
main difference lies in the arguments. In both formulae the
argumentsare variables,but of a different kind: the valueswhich
may be substituted for the variable .r4are terms, like ,man' or
'plant'. From the first formula we get thus the propositions ,All
men are men'or 'All plantsare plants'.The valuesof the variable
p are not terms but propositions,like 'Dublin lies on the Liffey'
or 'Today is Friday'; we get, therefore,from the secondformula
the propositions: 'If Dublin lies on the Liffey, then Dublin lies
on the Liffey' or 'If today is Friday, then today is Friday'. This
difference between term-variables and proposition-variables is
the primary difference between the two formulae and consequently between the two systems-of logic, and, as propositions
and terms belong to different semanticalcategories,the difference
is a fundamentalone.
The drst systemof propositional logic was invented about half
a century after Aristotle: it was the logic of the Stoics.This logic
is not a systemof thesesbut of rules of inference. The so-called
mo&tsponens,now called the rule of detachment: ,If c, then p;
but cr; therefore B ' is one of the most important primitive rules
of the Stoic logic. The variables a and B are propositional
variables, as only propositionscan be significantly substitutedfor
them.r The modern systemofthe logic ofpropositions was created
only in r 879 by the great German logician Gottlob Frege.Another
outstanding logician of the nineteenth century, the American
Charles Sanders Peirce, made important contributions to this
logic by his discoveryof logical matrices (rBB5). The authors
of PrincipiaMathematica,Whitehead and Russell, later put this
I Cf.Lukasiewiczr'ZurGeschichtedesAussagenkalki.ils',Er,tanntnis,vol,v,I*ipzig
(t 93 S ) , p p . r r r - g r .

$ 16

LOGIC

O F TER M S

AN D OF PR OPOSITION S

4s

system of logic at the head of all mathematics under the title

'Theory of Deduction'. All this was entirely unknown to philosophersof the nineteenth century. To this day they seemto have
no idea of the logic of propositions. Maier says that the Stoic
logic, which in fact is a masterpieceequal to the logic of Aristotle,
yields a poor and barren picture of formalistic-grammatical unsteadinessand lack of principle, and adds in a footnote that the
unfavourablejudgement of Prantl and Zeller on this logic must
be maintained. r The ErcyclopaediaBitannica of r g r r saysbriefly of
the logic of the Stoicsthat 'their correctionsand fancied improvements of the Aristotelian logic are mostly uselessand pedantic'.z
It seemsthat Aristotle did not suspectthe existenceof another
system of logic besideshis theory of the syllogism. Yet he uses
intuitively the laws of propositional logic in his proofs of imperfect syllogisms,and even setsforth explicitly three statements
belonging to this logic in Book II of the Prior Analytics.The first
of theseis a law of transposition: 'When two things', he says,'are
so related to one another, that if the one is, the other necessarily
is, then if the latter is not, the former will not be either.'s That
means,in terms of modern logic, that whenever an implication
of the form 'If c, then B' is true, then there must also be true
another implication of the form 'Ifnot-p, then not-a'. The second
is the law of the hypothetical syllogism.Aristotle explainsit by an
example: 'Whenever if z4is white, then .B should be necessarily
great, and if ^Bis great, then C should not be white, then it is
necessaryif ,4 is white that C shoirld not be white.'+ That means:
whenevertwo implications of the form 'If cr,then p' and 'If p,
then y ' are true, then there must also be true a third implication
'If a, then y'. The third statementis an application of the two
foregoing laws to a new example and, curiously enough, it is
false. This very interesting passageruns thus:
by the
'It is impossiblethat the samething shouldbe necessitated
beingand by the not-beingof the samething. I mean,for example,
I Maier, op. cit., vol, ii D, p.
384r 'In derHauptsachejedochbietetdieLogik
der Stoiker. . . ein dtirftiges, 6des Bild fbrmalistisch-grammatischer Prinzip- und
Haltlosigkeit.' Ibid., n. I : 'In der Hauptsache wird es bei dem ungiinstigen IJrteiI,
das Prantl und Zeller iiber die stoische Logik fiillen, bleiben miissen,'
2 rrth ed., C ambri dge (r grr), v ol . x x v , p.946 (s .v .' S toi c s ' ).
!,4n. pr. ii.
4, 57br 6tav 6io E7l oha zpris dlllla ttote |otipou 6mos i{ dvdyrcqs
etvac|d.tepov, rcthou pi 6wos piv oi6i |dtepov i.otot.
1 lbid. 6 6rav yd.p rov6l 6vtos \euxoi roi A rc6i
piye etvar rd B, pey,lAou
'd.vd'yxty
Ddro0 B 6mos ri I pi1 levxdv, dvd'yxr1,ei td A Aeuxdv,d I pl elvat l\eux6v,

THE SYSTEM

$16

that it is impossiblethat B shouldnecessarily

be great if z4is white,
and that B shouldnecessarily
be greatif I is not white. For if-B is not
greatA cannotbe white. But i{, when I is not white, it is necessary
that B shouldbe great, it necessarily
resultsthat if ,B is not great, B
itself is great. But this is impossible.'t
Although the example chosen by Aristotle is unfortunate, the
senseof his argument is clear. In terms of modern logic it can be
stated thus: Two implications of the form 'If cv,then p ' and 'If
not-a, then B ' cannot be together true. For by the law of transposition we git from the first implication the premiss 'If not-B,
then not-a', and this premiss yields together with the second
implication the conclusion'If not-p, then F'by the law of the
hypothetical syllogism.According to Aristotle this conclusion is
impossible.
Aristotle's final remark is erroneous.The implication 'If not-p,
then p', the antecedentof which is the negation of the consequent, is not impossible;it may be true, and yields asconclusion
the consequentfl according to the law of the logic ofpropositions:
'If (if not-1, then p), then p.'z Commenting upon this passage,
Maier saysthat there would here result a connexion contrary to
the law of contradiction and therefore absurd.sThis comment
again revealsMaier's ignorance of logic. It is not the implication
'If not-p, then B ' that is contrary to the law of contradiction, but
only the conjunction 'p and not-p'.
A few years after Aristotle, the mathematician Euclid gave a
proof of a mathematical theorem which implies the thesis 'If (if
not-p, then p), then p'.+ He statesfirst that 'If the product of two
t An.
F. ii. 4, 57b3 to6 6' oitoi 6wos xai pi1 dnos, dirivetov i{ d.v,iyqs etvattd
oi+6. Ayo 6' otov rcA A 6yzos.leurof zd B elvat piya i( dvdyxls, xai pl 6wos Aeuxoi
toA A rd B elvet piye d( d,wiyxtls. Here follows the example of the hypothetical
syllogism quoted in p. 49, n. 4, and a second formulation of the law of transposition, The conclwion reads, I r ro0 Er)B pi16vtos peyd\ou i A oi1 ol6v re ),eurdv
cl,vat. roA Ei .d pr) 6nos \euxoi, ei d,v,iyxr1i B piyc etvoq ovpBoivet i{ dv,iyxrls toi B
peydAovpl 6wos ohd i B elvot 1y'yo. toAro 6' d6,ivetov.
2 See A. N. Whitehead and B. Russell, Principia
Mathematica, vol, i, Cambridge
( r 9 r o ) , p . r : 8 , t h e sis*2 .1 8 .
3 Op. cit., vol. ii c, p.33r: 'Es ergiibe sich also ein Zusammenhang, der dem
Gesetze des Widerspruchs entgegenstiinde und darum absurd wire.'
' See Scnlri di G, Vailati, Leipzig-Firenze, cxv. 'A proposito d'un passo dcl
Teeteto e di una dimostrazione di Euclide', pp.516-27;
cf. Lukasiewicz,
'Philosophische Bemerkungen zu mehrwertigen Systemen des Aussagenkalktls',
Compks Rmdus drs slancesdc la Sociltl des Scizncesct dts lzttrcs de Varsooic,xxiii (rg3o),
Cl. III, p. 67.

$ 16

LOGIC

O F TER M S

AN D OF PR OPOSITION S

5r

integers,a and b, is divisible by a prime number z' then if a is not

divisibleby n, b shouldbe divisibleby n'.Let us now suPposethat
a -- b and the product ax a (a2)is divisible by n. It results from
this supposition that 'If a is not divisible by z, then a is divisible
by z'. Here we have an example of a true implication the antecedent of which is the negation of the consequent. From this
implication Euclid derives the theorem: 'lf az is divisible by a
prime number z, then a is divisible by z.'
$ 17, Theproofs$t conansion
The proofs of imperfect syllogismsby conversionof a premiss
are both the simplest and those most frequently employed by
Aristotle. Let us analyse two examples.The proof of the mood
Festino of the secondfigure runs thus | 'lf M belongs to no .lf,
but to someX, then it is necessarythat "lVshould not belong to
some X. For since the negative premiss is convertible, "lf will
belong tono M;but Mwas admitted to belongto someX; therefore "ltfwill not belong to someX. The conclusion is reached by
meansof the fi'rstfigure.'I
The proof is basedon two premisses:one of them is the law of
conversionof the E-propositions:
(r) lf M belongsto no "lf, then "lf belongs to no M,
and the other is the mood Ferio of the first figure:
(z) If "lf belongs to no M and M belongs to some X, then "ilf
doesnot belong to someX.,
From thesepremisseswe have to derive the mood Festino:
(g) lf M belongs to no .lf and M belongs to someX, then "ltf
doesnot belong to someX.
Aristotle performs the proof intuitively. Analysing his intuitions
we find two thesesof the propositionalcalculus: one of them is
the above-mentioned law of the hypothetical syllogism, which
may be statedin the following form:
(+) If (if p, then 4), then [if (if a, then r), then (if p, then

r ) l;"

I An. pr. i.
5, z7a3z <i ydp i M tQ plv ff p46eri tri 6t E lri indpyeq dvd.yq td N
tvi rQ E pi Jrd.pyel. inei yd,pdnwtpi$ec ri mcpqt"x6v, oriEevizri' M Jndp(ec td N'
ti 6( ye M ,jn(xeno nvi tQ E indpxel' ,iote td N rwi tQ E oity t'ndp{et' yivetat ydp
o"Atroynpds 6td toA rpritov oytiparos.
*z'o6.
' See Principia Mathenatica, p. Io4, thesis

Sz

THE SYSTEM

$ rZ

The other thesisreads:

(5) If (ifp, then g), then (ifp and.r, then q and r).
This thesisis called in PrincipiaMathematica,
following Peano,the
principle of the factor. It showsthat we may 'multiply' both sides
of an implication by a common factor, i.e. we may add, by means
of the word 'and', top and to q anew propositionr.I
We start with thesis (S). Ar p, q, and / are propositional
variables,we may substitutefor them premissesof the Aristotelian
logic. Putting 'M belongsto no N' for p,'"lf belongsto no M' for
4, and 'M belongsto someX' for r, we get from the antecedentof
(5) the law of conversion(r), and we may detachthe consequent
of (5) as a new thesis.This new thesishas the form:
(6) If M belongs to no "ilf and M belongsto someX, then
"iV
belongsto no M and,M belongsto someX.
The consequentof this thesisis identical with the antecedentof
thesis(z). Therefore we may apply to (6) and (z) the law of the
hypothetical syllogism, substituting for / the conjunction 'M
belongsto no "ltfand M belongsto someX', for 4 the conjunction
'-lf belongs to no M and M belongs to someX', and for r the
proposition 'Jf doesnot belong to someX'. By applying the rule
of detachment twice we get from this new thesis the mood
Festino.
The secondexample I want to analyseis somewhatdifferent.
It is the above-mentionedproof of the mood Disamis.zWe have
to prdre the following imperfect syllogism:
(7) If n belongs to all ,S and P belongs to some S, then P
belongsto some-R.
The proof is basedon the mood Darii of the first figure:
(8) If n belongs to all S and S belongs to some P, then ft
belongsto someP,
and on the law of conversionof the /-propositions applied twice,
once in the form:
(g) If P belongsto some.S,then S belongsto some P,
and for the secondtime in the form:
(to) If ^Rbelongsto someP, then P belongsto some.R.
As auxiliary thesesof the propositional logic we have the law of
I See Prilrcipia Matlumalica, p. rr9, thesis *3.45,
The conjunction'p and r'is
2 See the Greek rext in p. 25, n. r.
calfed in thx Pincilia'logical
product'.

$ rZ

THE

PR OOFS BY C ON VER SION

53

the hypothetical syllogism, and the following thesis, which is

slightly different from thesis (5), but also may be called the
principle of the factor:
( r r ) If (if p, then 4), then (if r and /, then r and q).
The differencebetween (5) and (rr) consistsin this, that the
common factor r is not in the secondplace, as itt (S), but in the
first. As conjunction is commutable and 'p and r' is equivalent to
'r and p', this difference doesnot affect the validity of the thesis.
The proof given by Aristotle beginswith the conversionof the
premiss 'P belongs to some S'. Following this procedure, let us
substitutefor p in (lr) the premiss'P belongsto someS', for 4
the premiss'S belongsto someP', and for r the premiss'.Rbelongs
to all ^9'.By this substitutionwe get from the antecedentof (I r)
the law of conversion (g), and therefore we may detach the
consequentof (rr) which reads:
(rz) If .R bel<ingsto all S and P belongs to some S, then ^R
belongsto all S and S belongsto someP.
The consequentof (Iz) is identical with the antecedentof (B).
By applying the law of the hypothetical syllogism we can get
from (rz) and (B) the syllogism:
(r3) If R belongsto all ^Sand P belongsto some ,S,then R
belongsto someP.
This syllogism, however, is not the required mood Disamis, but
Datisi. Ofcourse, the mood Disamis could be derived from Datisi
by converting its consequent according to thesis (Io), i.e. by
applying the hypothetical syllogismto (t3) and (Io). It seems,
however, that Aristotle took another course: instead of deriving
Datisi and converting its conclusion, he converts the conclusion
of Darii, getting the syllogism:
(ra) If R belongs to all ^Sand S belongs to some P, then P
belongsto someR,
and then he applies intuitively the law of the hypothetical syllogismto (rz) and (I4). The syllogism(r4) is a mood of the fourth
figure called Dimaris. As we already know, Aristotle mentions
this mood at the beginning of Book II of the Pior Analytics.
In a similar way we could analyseall the other proofs by convcrsion. It follows from this analysisthat if we add to the perfect
syllogismsof the fint figure and to the laws of conversion three

THE SYSTEM

$ 17

laws of the logic of propositions,viz. the law of the hypothetical

syllogism and two laws of the factor, we get strictly formalized
proofs of all imperfect syllogisms except Baroco and Bocardo.
These two moods require other thesesof the propositional logic.
$ fi. Theproofsb2 reductio ad impossibile
The moods Baroco and Bocardo cannot be reduced to the first
figure .by conversion. The conversion of the l-premiss would
yield an /-proposition, from which together with the O-premiss
nothing results,and the O-premisscannot be converted.Aristotle
tries to prove these two moods by a reductioad impossibile,d"nayatyl
cis rd d6waroy. The proof of Baroco runs thus: 'If M belongsto
all "{ but not to someX, it is necessarythat "Mshould not belong
to someX; for if "iVbelongsto all X, and M is predicated also of
all N, M must belong to all X; but it was assumedthat M does
not belong to someX.'r This proof is very conciseand needsan
explanation.Usually it is explainedin the following way:2
We have to prove the syllogism:
Q) lf M belongsto all "M and M does not belong to some.l',
then "lf does not belong to some X.
It is admitted that the premisses'M belongs to all ,ltf' and 'M
doesnot belong to someX' are true; then the conclusion '.ltfdoes
not belong to someX' must also be true. For if it were false, its
contradictory, '"lVbelongsto all X', would be true. This last propositiotr is the starting-point of our reduction. As it is admitted
that the premiss 'M belongs to all "lV' is true, we get from this
premiss and the propostion 'JV belongs to all X' the conclusion
'M belongsto all X'by the mood Barbara.But this conclusionis
false,for it is admitted that its contradictory 'M doesnot belong
to someX'is true. Therefore the starting-point of our reduction,
'"lf belongs to all X', which leads to a false conclusion, must be
false, and its contradictory, '-l/ doesnot belong to someX', must
be true.
This argument is only apparently convincing; in fact it does
not prove the above syllogism. It can be applied only to the
traditional mood Baroco (I quote this mood in its usual form
I An. pr. i.
S,27"37 ei rQ p2v N nowi i M, tQ 6t E zrvi pfi indpyer, dv,iyq i N
nvi tQ E pl indpTew' ei yd,pnavzi Jtd.pyeq xarqyopeirat 6t xoi zri M navrds roi N,
dvdyq td M tovri tQ I Jnd.pTecv'Jn(xeno 6t zui pr) ind.p7ew.
2 Cf., for instance, Maier, op. cit., vol. ii a, p. 84.

$ 18

T H E P R O O F S BY R ED U C TIO

AD IM POSSIBIL E

55

with the verb 'to be', and not in the Aristotelian form with 'to
belong'):
(z)
All,rf is M'
SomeX is not M,
therefore
SomeX is not "lV.
This is a rule of inference and allows us to assertthe conclusion
provided the premissesare true. It does not say what happens
when the premissesare not true. This is irrelevant for a rule of
inference, as it is evident that an inference based on false premissescannot be valid. But Aristotelian syllogismsare not rules
of inference, they are propositions. The syllogism (r) is an implication which is true for all values of the variables M, N, and X,
and not only for thosevaluesthat verify the premisses.If we apply
this mood Barocoto the terms M:'bird', N:'animal', and X:
'owl', we get a true syllogism (I use forms with 'to be', as does
Aristotle in examples):

(g) *'siffii',?i"#Jti,u,,
"#someowlsare not animals.
then

This is an example of the mood Baroco, becauseit resultsfrom it

by substitution. The above argument, however,cannotbe applied
to this syllogism. We cannot admit that the premissesare true,
becausethe propositions 'All dnimals are birds' and 'Some owls
are not birds' are certainly false. We need not suPPosethat the
conclusionis false; it is falsewhether we suPposeits falsity or not.
But the main point is that the contradictory of the conclusion,i.e.
the proposition 'All owls are animals', yields together with the
first premiss 'All animals are birds' not a false conclusion, but a
true bne: 'All owls are birds'. The reduttioad inpossibileis in this
caseimpossible.
The proof given by Aristotle is neither sufficient nor a proof
by reductioad impoxibile.Aristotle describesindirect proof or the
by contrast with direct or ostensive
demonstrationper impossibile,
proof, as a proof that positswhat it wishesto refute, i.e. to refute
by reduction to a statementadmitted to befalse,whereasostensive
proof startsfrom propositionsadmitted to be true.r Accordingly,
I An. pr. ii. 14, 6zbzg inslpec 6' ri eis td d}Jvarov dad8erfrs rfis }etxnxffs tQ
rr\&at 6 Bo,JAetaid.rorpetr, drd.yovoe eis ipo\oyoJpevov ry'e0Eos'
fi 62 6etxnn\ dpyerut
if ipd\oyoupbav 0oeav (ctLfiAv).

56

THE SYSTEM

$ r8

if we have to prove a proposition by reductioad impossibile,we

must start from its negation and derive thence a statement
obviously false. The indirect proof of the mood Baroco should.
start from the negation of this mood, and not from the negation
ofits conclusion,and this negation should lead to un rrn"ond-itiotrally falsestatemeni, and not to a proposition that is admitted to
be false only under certain conditions. I shall here give a sketch
,M bilongs
to all
9{ sugh l proof. Let c denote the proposition
N', F 'N belongs to all X', and y 'M belongs to all X\ As the
negation of an z4-premissis an O-premiss,.not-B'r will have the
meaning'"lf doesnot belongto someX', and .not-y' ,Mdoes not
befong to someX'. According to the mood Baroco ihe implication
'If cvand not-7, then not-B ' is true, or in other words, c and not-y
are not true together with B. The negation, therefore, of this
proposition would mean that ' cuand B and not-y, are together
true. But from ' crand p ', ' y' resultsby the mood Barbai; *"
get therefore 'y and not-y', i.e. a proposition obviously false,
being a contradiction in forma. It can iasily be seen that this
genuine proof of the mood Baroco by redutio ad imposibih is
quite different from that given by Aristotle.
The mood Baroco can be proved from the mood Barbara by a
very simple ostensiveproof which requires one and only one
thesisof the propositional logic. It is thtfollowing compound law
of.transposition:
(+) If (ify' and 4, then r), then if p and it is not true that r,
then it is not true that q.z
p 'M belongsto all .lf ', for 4 ',tVbelongsto all X,
n:
,and for r
\!
'j/ belongsto all x'. By this substitution *. get in the antecedent
of (4) the mood Barbara, and therefore*e .in detach the consequent, which reads:
(S) If M belongs to all
"M and it is not true that M belongs
to all X, then it is not true that
-If belongs to all X.
rt:
is
the
negatibn
ofthe l-premiss, we may replace
$
Q-premiss
in_(5) the forms'it is not true that belongs.toall, by ;doesnot
belong to some', getting thus the mood Baioco.
There can be no doubt that Aristotle knew the law of transposition referred to in the above proof, This law is closely conr l-am using
'not-' as an abbreviation for the propositionar negation .it is not
2 See principia Matlumatia,p.
true that'.
I rg, thisis i3.g7.

$ rB

T H E P R O O F S BY R ED U C TIO

AD IM POSSIBIL E

s7

nected with the so-called 'conversion' of the syllogism, which he

investigated thoroughly.I To convert a syllogism means to take
the contrary or the contradictory (in prooGparinpossibilconly the
contradictory) of the conclusiontogetherwith one premiss,thereby destroying the other premiss. 'It is necessary,'Aristotle says,
'if the conclusion has been converted and one of the premisses
stands,that the other premissshould be destroyed.For if it should
stand, the conclusion must also stand.'z This is a description of
the compound law of transposition.Aristotle thereforeknows this
law; moreover, he applies it to obtain from the mood Barbara
the moodsBarocoand Bocardo. Investigating in the samechapter
the conversionof the moods of the first figure, he says: 'Let the
syllogismbe affirmative (i.e. Barbara), and let it be converted as
stated(i.e.bythecontradictorydenial).Then if z4doesnot belong
to all C, but to all B, B will not belong to all C. And if I doesnot
belong to all C, but B belongs to all C, /4 will not belong to all
,8.'3 The proofs of Baroco and Bocardo are here given in their
simplestform.
In the systematicexposition of the syllogistic thesevalid proofs
are replaced by insufficient demonstrationsper imposibile,The
reasonis, I suppose,that Aristotle doesnot recognizearguments
d( iro|ioeas as instruments of genuine proof. All demonstration
is for him proof by categorical syllogisms; he is anxious to show
that the proof per inpossibileis a genuine proof in so far as it contains at least a part that is a categorical syllogism.Analysing the
proof of the theorem that the sideof ztsquareis incommensurable
with its diagonal, he states explicitly: We know by a syllogism
that the contradictory of this theorem would lead to an absurd
consequence,v'rz. that odd numbers should be equal to evens,
but the theorem itself is proved by an hypothesis,since a falsehood results when it is denied.+ Of the same kind, Aristotle
I An.
!r. ii. 8-ro.
2 Ibid. B,
dmcotpo$imos xai tfis ripds
5gb3 dvdyq ydp rcO aupnepd.opdros

peoJoqs npord,aeos dvotpeio|at rlv Aonfiv' ci ydp Eoteq xai td ovpnipaopo otat.
C[ Top, viii. I4, t63eg4 ,ivriyxrl ydp, ei td oupnipoopo pi Aon, ptav nvo. d.vatpcto|et
t6v npord,oeov, einep naotw teletotirv dvdyxl fiv ti oupnipoopz etvac.
r An. p.ii.8,5gbz8
Eno ydp xanlyoprxds d ouiloywpis, xai dwtotpe$ioilu oinus
(i.e..dwnecp&as). oJrotu ei d A oi' noni tQ f, rQ 62 B nomi, td B oi nami tQ f'
xai ci td piv A pl novri nit f, td 62 B nam| d A oi navri tQ B.
1 Ibid. i. 29,
4ra2g tdmes ydp oi Erri zo0 d\uvdtov nepcivomes rd p2v $efilos
ovAAoyi(ovraq rd 6' d( ,ipyffs i( Jno9ioeus Setxviovocv,6tav d\Jvat6v tr ovpBalvTl tfis
dln$doeus te|eioqs, otov 6tr doJppetpos i1 id"petpos 6td td yiveo|er td, tepnrd, ioo

58

THE SYSTEM

Sr8

concludes,are all otherhypothetical arguments; for in every case

the syllogism leads to a proposition that is different from the
original thesis,and the original thesisis reachedby an admission
or someother hypothesis.rAll this is, ofcourse,not true; Aristotle
doesnot understand the nature of hypothetical arguments. The
proof of Baroco and Bocardo by the law of transposition is not
reachedby an admissionor someother hypothesis,but performed
by an evident logical law; besides,it is certainly a proof of one
categorical syllogismon the ground ofanother, but it is not performed by a categorical syllogism.
At the end ofBook I of the Pior AnalyticsAristotle remarks that
there are many hypothetical arguments that ought to be consideredand described,and promisesto do so in the sequel.zThis
promise he nowhere fulfils.3 It was reservedfor the Stoics to include the theory of hypothetical arguments in their system of
propositional logic, in which the compound law of transposition
found its proper place. On the occasionof an argument of Aenesidemus (which is irrelevant for our purpose) the Stoics analysed
the following rule of inference which correspondsto the compound law of transposition: 'If the first and the second,then the
third; but not the third, yet the first; thereforenot the second.'a
This rule is reduced to the second and third indemonstrable
syllogismsof the Stoic logic. We already know the first indemonstrable syllogism, it is the modusponens;the secondis the modus
tollens:'lf the first, then the second; but not the second; therefore
not the fint.' The third indemonstrable syllogism starts from a
denied conjunction and reads: 'Not (the first and the second);
but the first; therefore not the second.' According to Sextus
Empiricus the a4alysisruns thus: By the secondindemonstrable
syllogismwe get from the implication 'if the first and the second,
tots d,pttoc oultltirpou *|eior1s. rd piv olv ioe yiveo|ot td zepnrd tois d.pttors
ouiloyilerat, rd 6' doipperyov etvor riv }dpctpov 2( Jno0ioetos 6eiwuow, inel
Ite06osoupBaiver 6rd z)r d.vri$aocvo
I An. pr. i. zg,
4ta37 tioairos 8i rai oi dA\or nd"vresoi it ino|ioeus' 2v dnaot yd.p
6 ptv ouiloywlr6s ytveru npds td petaAapBavdpooy, zri E' d( d,pyfis nepaiverec 6t'
6poAoytos fi rcvos d)Aqs iio|ioeos.
oris inoxi' Ibid.44, 5oa3g zolloi Ei roi zcpor repelvowan i{ iro|ioeus,
piv oiv of 6n$opo,i roitav, xai tooayits
$ao|ar 6ci xai itdoqpffvat xa0tp6s, :.y';ves
ytverer ti d( ino|ioeos, iotepov dpoipev.
3 Alexander
3Bg. 32, commeqting on this passage says: l37er rai d).trouszol)oris
2( Jno|ioeas nepalveo0an, repi ,iv Srenri1erat piv tis ip,5v impeAimepov, oJ p)v
$ipeat a&toA olyypappa repi oiriw.
+ The Stoics denote proposition-variables by ordinal numbers.

$ 18

T H E P R O O F S BY R ED U C TIO

AD IM POSSIBIL E

59

then the third', and the negation of its consequent'not the third',
the negation of its antecedent 'not (the first and the second)'.
From this proposition, which is virtually contained in the premisses,but not explicitly expressedin words, together with the
premiss 'the first', there follows the conclusion 'not the second'
by the third indemonstrable syllogism.' This is one of the
neatestarguments we owe to the Stoics.We seethat competent
logicians reasoned2,ooo years ago in the same way as we are
doing today.
| ry. TheproofsQt ccthesis
The proofs by conversion and per impossibileare sufficient to
reduce all imperfect syllogismsto perfect ones. But there is still
a third kind of proof given by Aristotle, viz. the so-calledproofs
by exposition or trcileots.Although of little importance for the
system,they have an interest in themselves,and it is worth while
to study them carefully.
There are only three passagesin the Prior Anafi,ti.cswhere
Aristotle gives a short characterization of this kind of proof. The
first is connected with the proof of conversion of the E-premiss,
the secondis a proof of the mood Darapti, the third of the mood
Bocardo. The word ix|lo1at occurs only in the second passage,
but there can be no doubt that the other two passagesalso ar'e
meant as proofs by ecthesis.z
Let us begin with the first passage,which runs thus: 'If A
r Sextus Empirictrs (ed. Mutschmann), Ada. math. viii. 235-6 ouviotrlxe ydp 6
torcAtos lciTos (scil. d nopd.tQ AtvryDjp<p ipoq|cts) 2x \ewipou dvanotetxrou xal
tpitov, xaids ndpeon pa|civ 2x rfis dvaAfioeas, lns oe$cor$o p6.AAovyoioetot inl
toi tpdnou nonyoaptvav ilpil :rilv ElDcorcl/or, lgowosoitas' ' ei rd npDrcv xai td
ieirepov, td tpitov' ol11y'6i ye td rphov, d))d rci t6 npittov' oir dpo rd 6eftepov.'
inci ydp Eyop<vowqltltoov iv Q fiyeitat oupnenAeypivov (td) 'td tpitrov xoi td
tphov', E2ppev62 xai td d'wtxeipevov toA )\iJyovros td
ileJrcpov ', l{yer 6t (rd) 'i
' oi td rpirov', ouvaT9foeror i11tiv xai td dvtwelpevov toa iyoultivou i ' oix d.po
td npdttov xai td }etjrepov ' 6eut$q dvonodeixrtll' d fd Et) roAro oiti xard p2v iy
\Jvapw Eyxeuot tQ A6yqt, irei Eyopevtd. owomud oJroi Afippato, rcrri 6i rr)r,
nptirov'rl
rpo$opdn nopeeret. 6rcp td(awes perd toa Aenopivov Afipparos to6'rd
Z(opev ouvoyLpevov zi oupntpaopo zd 'oJx dpo td iethepov' rphE dvono|cixrqt.
f' roA np<irou codd., ro0 tp6nou Kochalsky, zo0 'rd npOtrov'scripsi. (tp6nos:
antecedent'
implication, iyorJpoov:
mood expressed in variables, ouvlltltivov:
: consequent, oupnenAcypivov : conjunction.)]
Affyov
'!
Th.r.arei w ootherpassages deal i ngw i thec thes i s , A n.pr.goe6-t4and3ob3r4o (I owe this remark to Sir David Ross), but both are related to the scheme of
modal syllogisms.

THE SYSTEM

$r9

belongsto no.B, neither will.B belong to any r4. For if it should

belong to some, say C, it would not be true that .24belongs to
no B; for C is someof the .B's.'I The conversionof the E-premiss
is here proved per impossibile,
blut this proofper impossibile
is based
on the conversionof the.I-premisswhich is proved by exposition.
The proof by exposition requires the introduction of a new term,
called the 'exposedterm'; here it is C. Owing to the obscurityof
the passagethe very meaning of this C and of the logical structure
of the proof can be reached only by conjecture. I shall try to
explain the matter on the ground of modern formal logic.
We have to prove the law of conversionof the .I-premiss: 'If
.B belongsto somer4, then ,4 belongsto someB.'Aristotle introducesfor this purpose a new term, C; it follows from his words
that C is included in ,Baswell asin ,4, sothat we get two premisses:
'.8 belongsto all C' and'A belongsto all C'. From thesepremisses
we can deduce syllogistically (by the mood Darapti) the conclusion 'r4 belongs to some B'. This is the first interpretation given
by Alexander.2 But it may be objected that this interpretation
presupposesthe mood Darapti which is not yet proved. Alexander prefers, therefore, another interpretation which is not
basedon a syllogism: he maintains that the term C is a singular
term given by perception, and the proof by expositionconsistsin
a sort of perceptual evidence.3 This explanation, however,
which is acceptedby Maier,ahas no supportin the text of the Pior
Analytics:Aristotle does not sav that C is an individual term.
Moieo,nEr, a proof by'perception is not a logical pt'oof. If we
I An. pr. i, z, z5ar5 et o$v p46oi rQ B rd A indpTcr, oi'62 tQ A oiicvi $nd.p(ertd
B. el ydp rrvr, ohv rQ f, oix d.trqflis Eoter i' pq}evi tQ B i A ind.pyew' rd yd.p I
rdv B il e'mrr, [Corr. W. D. Ross.]
2 Alexander gz. tz ei ydp td' B tcvi tQ A indpTer . . . Jnapyha tQ l'Ema
ydp
roAro ri toA A, t$ $ndpyet td B. Eotec 6l d f iv 6A,prQ B xei ti airo6, xei td B xotd
rovrds toi f' taitdv yd.p rd e 6ltp xai xard nevr6s. dll' {v zd I ri roA A' iv 6Aq)
dpa xai rQ A rd | 2otiv'ci Et e'v,i,\(+r,xard, navt6s airoa it1|$oerac td A, fiv 3t i I
i rc6 B' xai td A d,po xatd. rwds 1$AB xatqyopl0ioerac.
r Ibid.
3z I dpeudv torc xai oixecdtotov rois Aeyop(voc rd 6c' ix|ioeas xti
aio9qtrxds \(yecv r\v 6ei(w yeyovlvor, dl)d p?) tdt etplplvov rpdtov pq}t ouA\oyronxits. d 7dp 6rri rfs tx|loeos rpdnos 6r' aio|zioear yivetar xai oi ovMoynrtxits'
toroirov ydp tt \apBdvetot rd I td ixrfiipevov, 6 aio|rydv 6v p6pc6vAon toi A' ei yd.p
xatd, popiou to6 A 6vros toA I aio|ltoA twos xai xa0' Exeore Aiyono 16 B, eiq &v
xei toA B pdprov rd aird f 6v ye iv aJtQ' riste rd I efu Ev dp$oipuv pdprov xoi dv
d.p4o4poc airoes.
a Op. cit., vol. ii a, p. zo: 'Die Argumentation bedient sich also nicht eines
Syllogismus, sondern des Hinweises auf den Augenschein.'

$ rs

T H E PR OOFS BY EC TH ESIS

6r

want to prove logically that the premiss 'B belongs to some ,4'
may be converted, and the proof is to be performed by meansof
a third term C, we must find a thesis that connects the above
premiss with a proposition containing C.
It would not, of course,be true to say simply that if B belongs
to some.r4,then.B belongsto all C arrdA belongsto all C; but a
little modification of the consequent of this implication easily
solves our problem. We must put before the consequent an
existential quantifier, the words 'there exists', binding the variable C. Forlf .B belongsto some/, there always existsa term C
such that B belongsto all C and A belongsto all C. C may be the

state, therefore, the following thesis:

(I) If ,B belongsto some.,4,then there existsa C such that,B
belongsto all C and A belongsto all C.
This thesisis evident. But also the converseof (r) is evident. If
there existsa common part of A and .B' .B must belong to some
A.We get, therefore:
(z) If thereexistsa C suchthat,B belongsto all C and .,4belongs
to all C, then -B belongsto somel.
It is probable that Aristotle intuitively felt the truth of these
theseswithout being able to forinulate them explicitly, and that

known to Aristotle:
and q, then q andP.
$)lt!
It is the commutative law of conjunction.r Applying this law to
the premisses'.8 belongsto all C' and'A belongsto all C', we get :
(a) If ^Bbelongsto all C and A belongsto all C, then.,4 belongs
to all C and.B belongsto all C.
I See Pincipia Matlvmatica, p' I 16, thesis r3'zz-

6z

THE SYSTEM

$ rg

To this thesis I shall apply the rules of existential quantifiers.

There urretwo such rules; both are stated with respect to a true
implication. The first rule reads: It is permissibleto put before a
consequentof a true implication an existentialquantifier, binding
a free variable occurring in the consequent.It results from this
iule that:
(5) If ,B belongs to all C and A belongs to all C, then there
exists a C such that A belongs to all C and B belongs to
all C.
The secondrule reads: It is permissibleto put beforethe antecedent of a true implication an existential quantifier, binding
a free variable occurring in the antecedent,provided that this
variable does not occur as a free variable in the consequent.fn
(S) Cir already bound in the consequent;thereforeaccordingto
this rule we may bind C in the antecedent, thus getting the
,formula:
(6) If there existsa Csuch that B belongsto all Cand I belongs
to all C, then there existsa C such that A belongsto all C
and.B belongsto all C.
The antecedentofthis formula is identical with the consequentof
thesis (r) I it results, therefore, by the law of the hypothetical
syllogismthat:
Q) lf B belongs to some A, then there existsa C such that A
belongr to all C and B belongs to all C.
From (c) by interchanging.B and,A we get the thesis:
(B) If there existsa Csuch that A belongsto all Cand.B belongs
to all C, then 't4 belongs to some.B,
and from (7) and (B) we may deduceby the hypotheticalsyllogism the law of conversionof the /-premiss:
(9) If B belongsto someA, then I belongsto some,8.
We seefrom the above that the true reason of the converti*
bility of the .I-premissis the commutability of the conjrinction.
The perception of an individual term belonging to both A and B
may intuitively convince us of the convertibility of this premiss,
but is not sufficient for a logical proof. There is no need to assume
C as a singular term given by perception.

$ rg

T HE PR OOFS BY EC TH ESIS

63

The proof of the mood Darapti by exposition can now be

easily understood. Aristotle reducesthis mood to the first figure
by conversion,and then he says: 'It is possibleto demonstrate
this also per impossibileand by exposition. For if both P and ,R
belong to all $ should some of the S's, e.g. Jf, be taken, both P
and .R will belong to this, and then P will belong to some ft.'r
Alexander's commentary on this passagedeservesour attention.
It begins with a critical remark. If "ff were a universal term included in ,S,we should get as premisses'P belongsto all -lf'and
'R belongs to all "iV'. But this is just the same combination of
premisses,ou(uyia, as 'P belongs to all S' and '.R belongs to all
S', and the problem remains the same as before. Therefore,
Alexander continues, Jf cannot be a universal term; it is a
singular term given by perception, a term evidently existing in P
as well as in 8, and the whole proof by ecthesisis a proof by
perception.2We have already met this opinion above. In support
of it Alexander adducesthree arguments: First, if his explanation
were rejected,we should have no proof at all; secondly,Aristotle
doesnot say that P and R belong to all "ll, but simply to "M; thirdly,
he does not convert the propositions with Jf.3 None of these
arguments is convincing: in our example there is no need of
conversion; Aristotle often omits the mark of universality where
it should be used,aand as to the first argument, we know already
that there existsanother and a better,explanation.
The mood Darapti:
(ro) IfP belongsto all S and R belongsto all $ then P belongs
to some-R,
I An. pr. i.
5, z$zzz imr El rci 6d toO d\wdtoq xoi rQt ix01o0a" aocetv tlv dn6De$w' ei ydp dp$u (scrl. II xai P) ravri rQ 2 ind.pyt, Ev \$0fi n rlv 2, olol rri N,
toJrqt xai rd II xei i P i'f,,ip(<q tTote nvi rQ P il' Ilindp(et.
2 Alexander gg. zB ti yd.p En$ipet tQ 2 indpyew laBctv nwi td re II xoi rd P
xai p(pet nvi rofi 2 r? N; d yd.p alti xai ini roA N lryliwos piver fi yd.p oitil
ov(q'i.a iuiv, dv te xotd toi N nowds dxei.vovxdtepov, dv te xatd. roi 2 xanlyopi
tau fi oi totainl i1 Eeifrs, f ypfftor' 6 ydp 6i 2x$ioeos tpdnos 6t' aie|joeus yivetot.
oi yd,pivo rocoitiv n toa 2 Ad,Bope, xo|' oA iq|tioera, nawis xsi ri II xai d P,
)tiyec , , . ril,l' ive tt ttl Jn' oto1rlow nntdvrov, 6 $avepdv dmry 6v xei 2v tQ II xoi
dv tQ P.
3 Ibid. roo.
7 hc ydp ais9ryl { 6rri zfs 2x0ioeos 6ee(cs,otlltciov npdrov pEvtd
ei pi oitos AopBdvono, pq}epi,ev yiveo|et 6ei(cv' Enene 6t rci rd a&tdv ptlxht
1qpfioao|a, hi to0 N, 6 fiv n ro0 2, tQ navti eJtQ izdpyecv t6 rc II rci tri P, d)l'
dn\6s 9etvat td ind.p26l' dl,ld rci rd pq}e:/pav dvrrcrpi$ar.
' See, for instance, p. 2, n.

64

THE SYSTEM

$ Ig

results from a substitution of thesis (z)-take P for B, and R

for A:
(rr) If there existsa C such that P belongsto all C and.R
belongsto all C, then P belongsto sometR,
and from the thesis:
(ra) If P belongsto all S and.R belongsto all 'S,then there
exists a C such that P belongsto all C and .R belongs to
all C.
Thesis (rz) we may prove by applying to the identity:
(r3) IfPbelongs to all Cand R belongsto all C, then Pbelongs
to all C and R belongsto all C,
the secondrule of existential quantifiers, getting thus :
(r4) If Pbelongs to all Cand,R belongsto all C, then there
exists a C such that P belongs to all C and ,R belongs to
all C,
and substitutingin (I4) the letter,S for the free variable C, i.e.
performing the substitution in the antecedentonly, as it is not
permissibleto substitute anything for a bound variable.
From (rz) and (rr) the mood Darapti results by the hypothetical syllogism. We see again that the exposed term C is a
of course,to
universalterm like A or B.It is of no consequence,
denote this term by "lf rather than by C.
Of greater importance seemsto be the third passage,containing the proof by exposition of the mood Bocardo. This Passage
reads: 'If R belongsto all ,S,but P doesnot belong to some'9' it is
necessarythat P should not belong to someR. For if P belongsto
all R, and ,R belongsto all ^S,then P will belong to all ^S;but we
assumedthat it did not. Proof is possiblealso without reduction
ad imposibile, if some of the .S'sbe taken to which P does not
belong.'t I shall analysethis proof in the sameway as the other
proofs by exposition. '5
Let us denote the part of ^Sto which P doesnot belong by C;
we get two propositions: '^Sbelongs to all C' and 'P belongs to
no C'. From the first of these propositions and the premiss 'R
t An. pr. i. 6, zBbIT ei ydp i P r.ovti rQ 2, i Et I/ zrvi 1ti1itipyeq dl,iyq i II
nvi tQ P pi1indpyew, ei ydp navti, xoi td P navziltQ 2, xai rd II nami tQ E i'n'ipfet
Et xci dveu tfis drayoyffs,2d'v \$0fi t rttv E Q td n p+1
dM' oril iiflpyru.ieixvvtat
Jtdpyeu

$19

T H E PR OOFS BY EC TH ESIS

65

belongs to all ,S' we get by the mood Barbara the consequence

'ft belongs to all C', which yields together with the second
proposition ? belongs to no C' the required conclusion 'P does
not belong to some,R'by the mood Felapton. The problem is how
we can get the propositionswith C from the original premisses'-R
belongsto all S' and 'P does not belong to some .9'. The first of
thesepremissesis uselessfor our purpose asit doesnot contain P;
from the secondpremiss we cannot get our propositions in the
ordinary way, since it is particular, and. our propositions are
universal. But if we introduce the existential quantifier we can
get them, for the following thesisis true:
(t5) If P doesnot belong to someS, then there existsa C such
that S belongsto all C and P belongsto no C.
The truth of this thesis will be obvious if we realize that the
required condition for C is always fulfilled by that part of S to
which P doesnot belong.
Starting from thesis(15) we can prove the mood Bocardo on
the basisof the moods Barbara and Felapton by means of some
laws ofpropositional logic and the secondrule ofexistential quantifiers. As the proof is rather long, I shall give here only a sketch.
We take as premisses,besides.(r5), the mood Barbara with
transposedpremisses:
(tG) IfS belongsto all Cand ft belongsto all S, then R belongs
to all C,
and the mood Felapton, also with transposedpremisses:
(t7) IfR belongsto all C and P belongsto no C, then P does
not belong to someR.
To thesepremisseswe may apply a complicated thesisof propositional logic which, curiously enough, was known to the Peripatetics and is ascribed by Alexander to Aristotle himself. It is
called the 'synthetic theorem', ouv1err.rcdv
|edpr1y.a,and runs
thus : 'If a and p imply y, and y together with 6 implies e, then c
and B together with E imply e.'I Take for u, p, and y the first
I Alexander e74. I9 Er' 6r Ei,\e7er v0v, inqpd$ecfipiv
$avepdrcpov td Aeydpeov
' ow|enxdv 9e<ipr1pa', oi alrds 2orw cipctis. imr Di ri neprcy) altoa toratjfll' ' 6tav
Ex nvov owdyqtoi z, rd 6i ouvaydp,evovperd, nvds i nvdv owdy71 tq xal td ouw
axtlxd o}rrc,apel' o6 i 1x0' 6v ouvdyctanlxeivo, rai aJrd rd aitd owd.(eu' The following example is given ibid. 16 jrci ydp td ' niv iixetov dya|dv' ouvaydpevov&rd
t6v ' adv iixatov xo,Adv,tdv xdtdr dye06v'' owdyet petd to6' nilv ,iy&dv oup$ipov'
5807

T H E SY ST E M

!rg

premiss, the second premiss, and the conclusion respectivcly-of

Barbara, for 6 and e the second premiss and the conclusion
respectivelyof Felapton; we get the formula:
(rB) If S belongsto all C and R belongsto all.9 and P belongs
to no C, then P doesnot belong to someR.
This fofmula may be transformedby another law ofpropositional
logic into the following:
(r9) If .Sbelong:sto all C and P belongs to no C, then if .R
belongsto all 'S,P doesnot belong to someR.
To this formula may be applied the second rule of existential
quantifiers. For C is a free variable occurring in the antecedent
o1(r9), but not in the consequent.According to this rule we get
the thesis:
(zo) If there ocists a C such that ^Sbelongs to all C and P
belongs to no C, then if R belongs to all 'S, P does not
belong to someR.
From premiss(r5) and thesis(zo) there resultsby the hypothetical syllogism the consquence:
(zr) If P doesnot belong to someS, then if ^Rbelongsto all S,
P doesnot belong to someR,
and this is the implicational form of the mood Bocardo.
It is, of course, highly improbable that Aristotle saw all the
steps of this deduction I but it is important to know that his
intuitions with regard to the proof by ecthesiswere right. Alexander's commentary on this proof of the mood Bocardo is worthy
of quotation. 'It is possible',he says,'to prove this mood without
assumingsome .Sgiven by perception and singular, but taking
such an S, to none of which P would belong. For P will belong to
none of this $ and R to all, and this combination of premisses
yields as conclusion that P doesnot belong to someft.'r Here at
last Alexander concedesthat the exposedterm may be univenal.
The proofs by exposition have no importance for Aristotle's
td 'nilv itxetov oup{lpov', xo,f,td' 'nilv Edrcrol rcldr, nilv xeAdv dyaidv' 6we
iya06v ovy$ipov' ouvdlet ti
ouvaxtrxd toi 'nilv }txarov ,iyofliv' petd' toi'n6v
' ndv iltxanov ovp$ipov'.
-ro4.
! Alexander
g irlvetot 6' iri fs oulvylos taritrls iewv6veq xoi ei p)1 ato0ry'6v
n roi E \appdvono xoi xai' Exonc, d)tAd' toainov, oS rcrd plEevds xttqyopl0fioetu
td IL Eorot ydp td piv Il xct' ori8evris eito6, td 6i P rcrd nawds' fi 6' o&teos Zxouoe
ovlvyta ou),Aoytozrtis Ed6errtct ouvdyovoz rd twi tQ P td II pl iadpTew.

0 rg

T H B PR OOFS BY EC TH ESIS

67

syllogistic as a system.All theorems provcd by ecthesiscan be

proved by conversion or pn tnpossibilz. But they are highly
important in themselves,as they contain a new logical element
the meaning of which was not entirely clear for Aristotle. This
was perhaps the reason why he dropped this kind of proof in his
final chapter (7) of Book I of the Pior Anafitics,where he sums
up his systematicinvestigation of syllogistic.t Nobody after him
understood theseproofs. It was reserved for modern formal logic
to explain them by the.idea of the existential quantifier.
$zo. The rcjutcdforms
Aristotle in his systematicinvestigation of syllogisticforms not
only proves the true ones but also shows that all the others are
false, and must be rejected. Let us seeby means of an example
how Aristotle proceeds to reject false syllogistic forms. The
following two premissesare given: ,4 belongs to all B and B
belongstonoC. It is the first figure: z{ is the fint or the major
term, I is the middle, and Cis the last or the minor term. Aristotle
writes:
'Ifthe first term belongsto all the middle,but the middle to noneof
the last,therewill be no syllogismof the extremes;for nothing necessaryfollowsfrom the termsbeingso related;for it is possiblethat the
first shouldbelongto all aswell as to noneof the last, so that neither
a particular nor a universalconclusionis necessary.
But if there is no
necessaryconsequenceby meansof thesepremisses,there cannot be
a syllogism.Termsof belongingto all: animal,man, horse; to none:
animal, man, stone.t2
In contrast to the shortnessand obscurity of the proofs by
ecthesis,the above passageis rather full and clear. Nevertheless
I am afraid it has not been properly understood by the commentators. According to Alexander, Aristotle showsin this passagethat from the same combination of premissesthere can be
I Cf, the com.Erent of Alexander, who rnaintains to the end hir idea of the
perceptual character of proofs by ecthesis, rr2. ggt dz 6i ri 6t' ixiioeos Dcifrs rfr
pvtlpove,iew eitfis
aiolqnxil xai oi ouAAoyrcttxi, ifftrov rai dr zo0 v6v oitdv plxitt
<,isErri ouDoT:opoA rwos ywopm1s.
2 An. pr. i.
4, z6'z ei 6i rd pir tpittov temi rQ ploE dxoAou0ei,'ri 62 pioov pq6ei
tQ ioyirqt &t6.pyeq oix Sotat oa)d.oywpis tir dxpov'ouD& ydp dveyxaiov ovpBoivet
tQ toina elvar xai yd.p aowi roi p16evi dvti26tat ri tyitov tQ ioy,irE ftadpTcr,
dne oite td xatd pipos oi* ti xc06)ou yiverot dvayxaior pq}evis 62 6mos dvayxaiou
6td tothotv oix Emat ouhoytoltis, 6pot rc6 noni Jtdpyeu lQov, dv0poaos,ittros' to0
p16cvi lQov, dvOpurog, Aieos.

68

THE SYSTEM

$zo
(6uvdp.evov
for
ouvdyeoila,.)
some
concrete
terms
a
derived
universal affirmative conclusion, and for someother concrete terms
a universal negative conclusion. This is, Alexander asserts,the
most obvious sign that such a combination of fremisses has no
syllogistic force, since opposite and contradictory propositions
which destroy each other are proved by it (Eeiruuzor).rWhat
Alexander says is certainly misleading, for nothing can be formally derived from an asyllogisticcombination of premisses,and
nothing can be proved by it. Besides,propositionswith different
concrete subjects and predicates are neither opposite to each
other nor contradictory. Maier again puts the terms pointed out
by Aristotle into a syllogistical form:
all men are animals
no horse is a man
all horsesare animals

all men are animals

no stone is a man
no stone is an animal

(the premissesareunderlined by him, as in a syllogism), and says

that there results (ugibt sdci)from logically equivalent premisses
a universal affirmative proposition as well as a universal negative.2 We shall seebelow that the terms given by Aristotle are
not intended to be put into the form of a syllogism, and that
nothing results formally from the premissesof the would-be
syllogismsquoted by Maier. In view of thesemisunderstandings
a logical analysis of the matter seemsto be necessary.
ff we want to piove that the following syllogistic form :
(r) If l'belongs to all B and B belongsto no C, then r4 does
not belong to some C,
is not a syllogism, and consequentlynot a true logical theorem,
we must show that there exist such values of the variables l, .8,
and C as verify the premisseswithout verifying the conclusion.
For an implication containing variables is true only when all the
t Alexander gg, 22 xai yip xa06Aov xoto$anxdv lni
nvos J,\r;s 8eifer Duvd.pevov
owdyeo0or xai rdAu e'z' d,\,\r1s xad\ov dno$orcxdv, 6 dvapyiuatov oqpeiov roi
pliepiav iTew tlv ou(wykv toirttlv ioTiv oviloynttxfiv,
ei ye td. te dvemia xoi td,
dnueipcvo v eitfi ieiwuroq 6wa d)Jrfiiuv d"vacpetxd,i
z Op. cit., vol. ii. a, p.
76: 'Es handelt sich also um folgende Kombinationen:
aller Mensch ist Lebewesen
aller Mensch ist Lebewesen
kein Pferd ist Mensch
kein Stein ist Mensch ..
alles Pferd ist Lebewesen
kein Stein ist Lebewesen
So wird an Beispielen gezeigt, dass bei der in Frage stehenden priimissenzusamvon logisch vcillig gleichen Vordersiitzen aus sowohl ein allgemein
Tgnlteltgng
bejahendcr, als ein allgemein verneinender Satz sich ergeben k6nne.'

$zo

T H E R EJEC TED

FOR M S

69

values of variables that verify the antecedent verity the consequent also. The easiestlvay of showing this is to find concrete
tlrms veri$ing the premisses'z{ belongsto all.B' and '.B belongp
to no C', but not verifying the conclusion '24doesnot belong to
some C'. Aristotle found such terrns: take 'animal' for A, 'man'
for Br'horse' for C. The Premisses'Animal belongsto all man'
or 'All men are animals', and 'Man belongsto no horse' or 'No
horsesare men', are verified; but the conclusion 'Animal does
not belong to some horse' or 'Some horsesare not animals' is
false. Formula (I), therefore, is not a syllogism. For the same
reason neither will the following form:
(z) If Abelongs to all B and B belongsto no C, then z{ belongs
to no C,
be a syllogism, becausethe premissesare verified for the same
concrete tirms as before, but the conclusion 'Animal belongs to
no horse' or 'No horsesare animals' is false. It follows from the
falsity of (l) and (z) that no negative conclusion c3n be drawn
from the given premisses.
Nor cari an affirmative conclusion be drawn from them. Take
the next syllogistical form:
(3) If .rabelongsto all.B and B belongsto no C, then / belongs
to some C.
There exist values for A, B,and C, i.e. concreteterms, that veriff
the premisseswithout verifying the conclusion. Aristotle again
tanimal' for Ar''mant for B, 'stonet for C'
gives such terms: take
premissesare verified, for it is true that 'All men are animals'
the ''No
stone is a man" but the conclusion 'Some stone is an
and
animal' is obviously false. Formul" (g), therefore, is not a syllogism. Neither can the last form:
no C, then z4belongp
@) lf Abelongs to all B and.B belongsto
C,
all
to
are verified
since for the given termsthe premissessyllogism,
be a
is not verianimals'
are
stones
'All
conclusion
the
b.fore,lut
a,
fied. It resultsfrom the above that no conclusionwhatever can be
derived from the combination of premisses'/ belong-sto all Bf
and '.8 belongs to no C', where ,4 is the predicate and B is the
subjectof theionclusion. This combination of premises is useless
for syllogistic.

7o

THE SYSTEM

$ zo

The main point of this processof, rejecgionis to find a true

universal affirmative proposition (like .All horsesare animals')
and a true universal negative proposition (like ,No stone is an
animal'), both compatible with the premisses.It is not sufficient
to find, for instance,for some terms a true universal affirmative
statement, and for some other terms a true particular negative
statement. This opinion was put forward by Alexander's teicher
Herminus and'some older Peripatetics,and was rightly refuted
by Alexander.I This is again a proof that Aristoi'le's ideas of
rejection have not been properly understood.
. Th. syllogistic forms (r)-14) are rejected by Aristotle on the
basis of some concrete terms that veiify the premisseswithout
veriffing the conclusion. Aristotle, howiver, knows yet another
kind of proo_ffor rejection. Investigating the syllogisiic forms of
-geneially
"in
the
figure, Aristotle states
tirat
this figure
-second
neither two affirmative nor two negitive primissesyield a t J"rrsary conclusion,and then continuesthus:
'!.e.tM belongto no ,i!y',
and not to someX. It is possiblethen for
"lly'
to belongeither to all X or to no X. Terms of bllonging to none:
black,snow,animal. Termsof belongingto all cannot5e 6und, if .r1,1
belongsto someX, and
{oesnot belonglo someX. For if ,tf belonged
to all X-,.and M t9 no JV,then M would belongto no X; but ii is
assumedthat it b3l_ongs
to
X. In thisway, tLen,it is not possible
-some
to taketerms,and the proofmust
startfrom the indefinitenatureof the
particularpremiss.For sinceit is true that M doesnot belongto some
evenif it belongsto no X, and sinceif it belongsto no Xaiyllogism
f,
is not possible,clearlyit will not be possibleeitf,er.,z
Aristotle here^beginsthe proof of rejection by giving concrete
terms, as in the first example. But then he breaksonni, proo{, as
he cannot find concrete terms that would verifr the p..-irr",
words of Herminus are quoted gq. aa:
r:i1vdw!{aow rroa ouvo}opr1, a".iio,,
w. )\iyew fis iv S rd, ivowia ouvdyetac

6povs,ix Et ro0 dEropdozo

v Eerxriov. dneiyr
8 rci ei p16evi ind,pleq pq}evi 6l in,ipyot

5 zo

T H E R EJEC TED

FOR M S

Zt

'M belongsto no "lf 'and 'M doesnot belong to someX', without

verifying the proposition '.itfdoesnot belong to someX', provided
M, which doesnot belong to someX, belongsat the sametime to
some (other) X. The reasonis that from thc premisses'M belongs
to no "lf' and'M belongsto sgmeX'the proposition '"lf does
not belong to someX'follows by the mood Festino. But it is not
necessarythat M should belong to some X, when it does not
belong to some (other) X; M might belong to no X. Concrete
terrnsverifiing the premisses'M belongsto no "iV'and 'M belongs
to no X', and not verifying the proposition '"lf doesnot belong to
someX', can easily be chosen,and in fact Aristotle found.them,
rejecting the syllogistic form of the secondfigure with universal
negative premissesl the required terms arcz M-'line',
.iV'animal', N-31nv11'.tThe same terms may be used to disprove
the syllogistic form:
$) lf M belongsto no "trfand M does not belong to someX,
then "lf doesnot belong to someX.
For the premiss 'No animal is a line' is true, and the second
premiss'Some man is not a line' is also true, as it is true that 'No
man is a line', but the conclusion 'Some man is not an animal' is
false. Aristotle, howevei, does not finish his proof in this way,z
becausehe seesanother possibility: if the form with universal
negativepremisses:
(6) If M belongsto no "iVand M belongsto no X, then "Mdoes
not belong to someX,
is rejected,(5) must be rejectedtoo. For if (5) stands,(6), having
a strongerpremissthan (5), must also stand.
Modern formal logic, as far as I know, doesnot use 'rejection'
as an operation opposedto Frege's'assertion'.The rules of rejection are not yet known. On the ground of the above proof of
Aristotle we may state the following rule:
(c) If the implication 'ff a, then p ' is asserted,but its consequent p is rejected,then its antecedentd must be rejectedtoo.
I Ibid. 27a2s ori6' (scil. Eotot ouiloyopis) &av pfite rca N pfite rc6 E p48ezds
xenryopfitac rd M. dpor tofr irdpTew yp.l,l"|, lQov, dv|patnos, toA pi1 indpyew
ypot"t"i, (Qov, Aiflos.
'2 Alexander completed this proof, 88. rz: toA nolti td N tQ E Jndpgw 6pot
ypapp| i M, lQov rd N, dvilpanos rd E' fi plv ydp yp"ttt i oriDei ldg xoi tri oJ1
ind.pyeui'v0p<inq dzei rci pq6evi, lQov Ei zrqvTi'iv0p6nE,

THE SYSTEM

0eo

This rule can be applied not only to reject (5) if (0) is rejected,
but also to reject (z) if (l) is rejected. For from an E-premiss
an O-premissfollows, and if (z) is true, then (r) must be true.
But if (I) is rejected,so must (z) be rejected.
The rule (c) for rejection correspondsto the rule ofdetachment
for assertion.We may accept another rule for rejection corresponding to the rule of substitution for assertion.It can be formulated thus:
(d) If ctis a substitution for p, and c is rejected,then B must be
rejccted too.
Example: supposethat'A doesnot belongto someI' is rejected;
then 'r4 doesnot belong to some-B' must be rejectedtoo, sin0e,if
the secondexpressionwere asserted,we should obtain from it by
substitution the fint expression,which is rejected.
The first of theserules was anticipated by Aristotle, the second
was unknown to him..Both enable us to reject someforms, provided that someother forms have already been rejected.Aristotle
rejectssomeforms by meansofconcrete terms, as 'man', 'animal',
'stone'. This procedure is correct, but it introduces into logic
terms and propositions not germane to it. 'Man' and 'animal'
are not logical terms, and the proposition 'All men are animals'
is not a logical thesis.Logic cannot dependon concreteterms and
statements.If we want to avoid this difficulty, we must reject
someforms axiomatically. I have found that if we reject the two
following brms of the secondfigure axiomatically:
Q) lt A belongsto all.B and A belongsto all C, then.B belongs
to some C, and
(B) If / belongsto no.B and A belongsto no C, then I belongs
to some C,
all the other forms may be rejectedby the rules (r) and (d).
poblems ''
I zr. Someunsolaed
The Aristotelian systemof non-modal syllogismsis a theory of
four constantswhich may be denoted by 'All - is', 'Ir[e - i5',
tSome
- is', and 'Some - is nott. These constantsare functors
of two arguments which are representedby variables having as
values only concrete universal terms. Singular, empty, and also
negative terms are excluded as values. The constants together

$zt

S O M E U N SOL VED

PR OBL EM S

73

with their arguments form-four kinds of proposition called premisses,viz. 'All A is B', 'No .,4is B', 'Some A is B', and 'Some
z{ is not B'. The systemmay be called 'formal logic', as concrete
terms, like 'man'or 'animal', belong not to it but only to its
applications. The systemis not a theory of the forms of thought,
nor is it dependenton psychology; it is similar to a mathematical
theory of the relation 'greater than', as was rightly observed
by the Stoics.
The four kinds of premissform thesesof the systemby meansof
two functors 'if- then' and 'and'. Thesefunctors belong to propositional logic, which is an auxiliary theory of the system. In
some proofs we meet a third propositional functor, viz. the
propositional negation 'It is not true that', denoted shortly by
'not'. The four Aristotelianconstants'All - i5', '\[e - is', 'Some
- is' and 'Some- i5 nei', together with the three propositional
constants'if- thent, 'and', and 'not', are the soleelementsof the
syllogistic.
All thesesof the systemare propositionsregarded as true for all
values of the variables that occur in them. No Aristotelian syllogism is formulated as a rule of inferencewith the word 'therefore',
as is done in the traditional logic. The traditional logic is a
systemdifferent from the Aristotelian syllogistic, and should not
be mixed up with the genuine logic of Aristotle. Aristotle divided
syllogismsinto three figures, but he knew and accepted all the
syllogistic moods of the fourth figure. The division of syllogisms
into figures is of no logical importance and has only a practical
aim: we want to be sure that no valid syllogistical mood is
omitted.
The system'isaxiomatized. As axioms Aristotle takes the two
first moods of the first figure, Barbara and Celarent. To these
two axioms we have to add two laws of conversion, as these cannot be proved syllogistically. If we wish to have the law of
identity, 'All I is l,' in the systemwe havc to assumeitaxiomatically. The simplestbasiswe can get is to take the constants'All - is'
and 'Some- i5' 3s primitive terms, to define the rwo other constants by means of those terms with the help of propositional
negation, and to assumeas axioms four theses,viz. the two laws
of identity and the moods-Barbara and Datisi,_or Barbara and
Dimaris. It is not possibleto build up the systemon one axiom
only. To look for the principle of the Aristotelian syllogistic is a

74

THE SYSTEM

5zr

vain attempt, if 'principle' means the same as 'axiom'. The socalled dirtum dc otnni et nulh cannot be the principle of syllogistic in this sense,and was never stated to be such by Aristotle
himself.
Aristotle reducesthe so-calledimperfect syllogismsto the perfect, i.e. to the axioms. Reduction here meansproof or deduction
of a theorem from the axioms. He usesthree kinds of proof : by
conversion, by rcdurtio ad inpossibilc, and by ecthesis. Logical
analysis showsthat in all the proofs of the first two kinds there are
involved thesesof the most elementary part of propositional logic,
the theory of deduction. Aristotle usesthem intuitively, but soon
after him the Stoics, who were the inventors of the first system of
propositional logic, stated some of them explicitly-the compound law of transpositionand the so-called'synthetic theorem',
which is ascribed to Aristotle but does not exist in his extant
logical worls..A new logical element seemsto be implied by the
proofs by ecthesis: they can be explained with the help of
existential quantifiers. The systematic introduction of quantifiers into the syllogistic would completely change this system: the
primitive term 'Some- is' could be defined by the term 'All is', and many new theseswould arise not known to Aristotle. As
Aristotle himself has dropped the proofs by ecthesisin his final
summary ofthe syllogistic,there is no need to introduce them into
his system.
Angther new logical element is contained in Aristotle's investigation of the inconclusive syllogistic forms: it is rejection.
Aristotle rejects invalid forms by exemplification through concrete terms. This procedure is logically correct, but it introduces
into the systemterms and propositionsnot germane to it. There
are, however, caseswhere he applies a more logical procedure,
reducing one invalid form to another already rejected. On the
basisofthis remark a rule ofrejection could be statedcorresponding to the rule of detachment by assertion; this can be regarded
as.the commencementof a new field of logical inquiries and of
new problems that have to be solved.
Aristotle does not systematically investigate the so-called
polysyllogisms,i.e. syllogismswith more than three terms and
two premisses.As we have seen,Galen studied compound syllogisms consisting of four terms and three premisses.It is an old
error to ascribe to Galen the authorship of the fourth figure:

$er

S O M E U N S OL VED

PR OBL EM S

75

Galen divided the compound syllogismsof four terms into four

figures, but not the simple ones known to us by their medieval
names.His investigationswere entirely forgotten. But compound
syllogismsalso belong to the syllogisticand have to be taken into
account, and here is another problem that has to be studied
systematically.An essentialcontribution to this problem is the
set of formulae given by C. A. Meredith, and mentioned above
at the end of section 14.
There still remains one problem not seenby Aristotle, but of
the utmost importance for his whole system: it is the problem of
decision. The number of significant expressionsof the syllogistic
is infinite; most of them are certainly false,but someof them may
be trug like valid polysyllogismsof n terms where z is any integer
whatever. Can we be sure that our axioms togetherwith our rules
of inference are sufficient to prove all the true expressionsof the
syllogistic? And similarly, can we be sure that our rules of rejection, formulated at the end of section 2c, are sufficient to reject
all the falseexpressions,provided that a finite number of them is
rejected axiomatically? I raised these problems in rg38 in my
Seminar on Mathematical Logic at the University of Warsaw.
One of my former pupils, now Professorof Logic and Methodology at the University ofWroclaw,J. Slupecki,found the solution
to both problems. His answer to the first question was positive,
to the secondnegative.Accordihg to Slupecki it is not possibleto
reject all the false expressions,ofthe syllogistic by means of the
rules (c) and (d) quoted in section zo, provided a finite number of them is rejected axiomatically. However many false expressionswe may reject axiomatically, there always exist other
false expressionsthat cannot be rejected otherwise than axiomatically. But it is impossibleto establishan infinite setofaxioms.
A new rule of rejection must be added to the systemto complete
the insufficient characterizationof the Aristotelian logic given by
the four axioms. This rule was found by Slupecki.
Slupecki's rule of rejection peculiar to Aristotle's syllogistic
can be formulated in the following way: Let c and p denote
negative premissesof the Aristotelian logic, i.e. premissesof the
type 'No A is B' or 'SomeI is not B', and let y denoteeither a
simple premiss (of any [cind) or an implication the consequentof
which is a simple premiss and the antecedent a conjunction of
suchpremisses:if the expressions
'If c, then y' and 'IfB, then y'

76

THE SYSTEM

$ zr

are rejected,then the expression'If c and p, then y' must be

rejected too.r This rule, together with the rules of rejection (c)
and (d) and the axiomatically rejected expression'If all C is .B
and all A is B, then someA is C',enables us to reject any false
expressionof the system.Besides,we supposeas given the four
assertedaxioms of the syllogistic, the definitions of the E- and
the O-premiss,the rules of inferencefor assertedexpressions,and
the theory of deduction as an auxiliary system.In this way the
problem of decision finds its solution: for any given significant
expressionof the systemwe can decide whether it is true and may
be assertedor whether it is false and must be rejected.
By the solution of this problem the main investigations on
Aristotle's syllogistic are brought to an end. There remains only
one problem, or rather one mysterious point waiting for an
explanation: in order to reject all the false orpressionsof the
systemit is necessaryand sufficient to reject axiomatically only
one false expression,viz. the syllogistic form of the secondfigure
with universal affirmative premissesand a particular affirmative
conclusion. There exists no other expressionsuitable for this
purpose.The explanation of this curious logical fact may perhaps
lead to new discoveriesin the field oflogic
t
J. Slupecki, 'Z badari nad sylogistyk4 Arystotelesa' (Investigation on Aristode's
Syllogistic), Tranaux tlc la Socitt dcs Scittucs ct des ltttrcs de Wroclaw, 56r. B, No. g,
Wroclaw (rgr48). See chapter v, devoted to the problcm of decision.

CHAPTER

A RI S T O T L E ' S

IV

SYST EM IN SYM BOL IC

F OR M

$ zz. Explanationd thc s2mbolism

Trus chapter doesnot belong to the history of logic. Its purpose is
to set out the systemof non-modal syllogismsaccording to the
requirementsof r,nodernformal logic, but in closeconnexion with
the ideas set forth by Aristotle himself.
Modern formal logic is strictly formalistic. In order to get an
exactly formalized thmry it is more convenient to employ a
symbolisminvented for this purposethan to make useof ordinary
language which has is own grammatical laws. I have therefore
to start from the explanation of such a symbolism. As the
Aristotelian syllogistic involves the rnost elementary part of the
propositional logic called theory of deduction, I shall explain
the symbolic notation of both thesetheories.
In both theoriesthere occur variables and corstants. Variables
are denoted by small Latin letters, constantsby Latin capitals.
By the initial letters of the alphabet a, b, c, d, ..., I denote termvariables of the Aristotelian logic. These term-variables have as
values universal terms, as 'man' or 'animal'. For the constantsof
this logic I employ the capital letters l, E, I, and O, usedalready
in this senseby the medieval logicians. By means of these two
kinds of letters I form the four functions of the Aristotelian logic,
writing the constantsbefore the variables:.
Aab meansAll a is b
or D belongsto all a,
Eab ,, No a is 6
,, Dbelongsto no 4,
Iab
,, Somea is 6
,, 6 belongsto some4,
Oab

,,

Some a is not 6 ,, D does not belong to some 4.

The constants A, E, I, and O are called functors, a and 6 their

arguments. All Aristotelian syllogismsare composed of these
four types of function connectedwith each other by means of
the words 'if'and 'and'. These words also denote functors, but
of a different kind from the Aristotelian constants:their arguments are not term-expressions,i.e. concrete terms or termvariables, but propositional expressions,i.e. propositions like

78

ARISTO TLE' S

SYSTEM I N SYM B O L I C

FORM

$ze

'All men are animals', propositional functions like 'Aab', or

propositional variables. I denote propositional variables by y',
g, t, sr..., the functor 'if' by C, the functor 'and' by ff. The
expressionCpq means'ifp, then g' ('then' may be omitted) and
is called 'implication' with p as the antecedent and q as the
consequent. C does not belong to the antecedent,it only combines the antecedent with the consequent. The expressionKpq
means 'p and 4' and is called 'conjunction'. We shall meet in
some prooG a third functor of propositional logic, propositional
negation. This is a functor of one argument and is denoted by
"iV.It is difficult to render the function ")Vpeither in English or
in any other modern language, as there exists no single word
for the propositional negation.t We have to say by circumlocution 'it-is-not-true-that p' or 'it-is-not-the-case-thatp'. For the
sake of brevity I shall use the expression'not-1D'.
The principle of my notation is to write the functors before
the arguments. In this way I can avoid brackets. This symbolism without brackets, which I invented and have employed in
my logical papers since r9eg,2can be applied to mathematics as
well as to logic. The associativelaw of addition runs in the
ordinary notation thus:
(a-lb)lc : al(b*c),
and cannot be stated without brackets. If you write, however,
the functor,* before its arguments, you get:
(a*b)*c:
and al(blc):
lafbc.
**abc
The law of associationcan be now written without brackets:
**abc:

{albc.

Now I shall explain some expressionswritten down in this

symbolic notation. The symbolic expression of a syllogism is
easyto understand.Take, for instance,the mood Barbara:
If all 6 is c and all a is 6, then all a is c.
It reads in symbols:
CKAbcAabAac.
t The Stoics used for propositional negation the single word orill.
2 See, for instance, Lukasiewicz and Tarski, 'Untersuchungen iiber den Aussagenkalkiil', Comphs Rcndus desslanas de la SociCtCdrs Scimccset det Lcttrcs dz Varsottie,
rxiii (r93o), Cl. III, pp. gr-4.

lzz

E X P L A N ATION

OF TH E SYM BOL ISM

79

The conjunction of the premissesAbc and Aab, viz. KAbcAab,is

the antecedent of the formula, the conclusion Aac is its consequent.
Some expressionsof the theory of deduction are more complicated. Take the symbolic expression of the hypothetical
syllogism:
If (ifp, then q), then [if (if g, then r), then (ifp, then r)].
It reads:
CCpqCCqrCpr.
In order to understand the construction of this formula you
must remember that C is a functor of two propositional arguments which follow immediately after C, forming together with
C a new compound propositional expression.Of this kind are
the expressionsCpq, Cqr, and Cpr contained in the formula.
Draw brackets around each of them; you will get the cxpression:
C(Cpq)C(Cqr)(Cpr).
Now you can easily seethat (Cpq)is the antecedentof the whole
formula, and the rest,i.e. C(Cqr)(Cpr),is the consequent,having
(Cqr) as its antecedent and (Ct ) * its consequent.
In the same way we may analyseall the other expressions,
for instance the following, which contains "lf and rf besidesC:
CCKpqrCKNrqNp.
Remember that K, like C, is a funotor of two arguments, and
that "ltfis a functor of one argument. By using different kinds of
bracketswe get the expression:

c lc (Kpq)rl{clK(Nr)ql(Np)}.
[C(Kpq)r] is here the antecedent of the whole formula while
ir its consequent, having the conjunction
{C$Wr)ql(/fp)}
as
its
antecedent
and the negation (Np) as its conWWr)q)
sequent.
$ zg. Theorltof deduction
The most fundamental logical systemon which all the other
logical systemsare built up is the theory of deduction. As every
logician is bound to know this system,I shall here describeit in
brief.

8o

ARISTOTLE' S

SYSTEM r N SYM BO LT C F O R M

$zg

The theoryofdeduction can be axiomatizedin severaldifferent

ways, according to which functors are chosenasprimitive terms.
The simplest way is to follow Frege, who takes as primitive
terms the functors of implication and negation,in our symbolism C and "lf. There exist many sets oi axioms of tfre C_,it/system; the simplest of them and the one almost universallv
acceptedwas discoveredby myself before r92g.r It consistsof
three axioms:
Tt. CCpqCCqrCpr

Tz. CCNppp
Ts. CpCNpq.

The first axiom is the law of the hypotheticarsyllogismalready

explained in the foregoing section. The seconduiio*, which
words 'If (if^not-y', then p), then p,, was applied by
:.uq: j"
Euclid to the proof of a mathemalical theorem., t iitt it ttre
law-ofclavius, as clavius (a learnedJesuit riving in the second
half of the sixteenth century, orie of the constructorsof the
Gregorian calendar) first drew attention to this law in his com-

$ r:

T H EOR Y

I Fint published
in Polish: 'o znaczeniu i potrzebach logiki matematvcznei,
(on the rmportance and Requirements of Mithematical
L"ogic), n^k"'i;i;i;,
vol. x, Warsaw (r92g), pp. 6ro-rz. Cf. also the German contribution quoted
in
p. 28, n. r: Satz 6, p. 35.
2 See above, section
16.
3 -Cf. my paper quoted in p.
48, n.

8r

significant expression;(c) Cap is a significant expressionprovided a and p are significant expressions.

The rule of detachment is the modusplnens of the Stolcs
referred to above: if a proposition of the type CcB is asserted
and its antecedenta is assertedtoo, it is permissible to assert
its consequentB, and detach it from the implication as a new
thesis.
By meansof thesetwo rules we can deduce from our set of
axioms all the true thesesof the C-,t!y'-system.
If we want to
have in the systemother functors besidesC and ylf, e.g. K, we
must introduce them by definitions. This can be done in two
different ways, as I shall show on the example of ff. The conjunction 'p and q' means the same as .it-is-not-true-that (if p,
then not-q)'. This connexion between Kpq and NCpNq
b.
expressedby the formula:
-uy
KPq:

NCPNq,

where the sign : corrspondsto the words ,means the same

as'. This kind of definition requires a special rule of inference
allowing us to replace the defniensby the definiendum
and vice
ve1sa.Or we may expressthe connexion between Kpq and
NCpNq by an equivalence,and as equivalenceis not a primitive
term of our system,by two implications converseto each other:
CKpqNCpNq

There belong to the systemtwo rules of inference:the rule of

substitution and the rule of detachment.
The rule of substitution allows us to deducenew thesesfrom
a thesisassertedin the systemby writing insteadof a variable a
significant expression,everywherethe same for the same variable. significant. expressionsare defined inductivery in the following
(a) any propositional variable is a significant
-way:(b)
expression;
N" is a significant expressionprovided a is a

OF D ED U C TION

and

CNCpNqKpq.

In this casea special definition-nlle is not needed. I shall use

definitions of the first kind.
Let us now seeby an examplehow new thesescan be derived
from the.axiomsby the help of rules of inference.I shall deduce
frrrm Tr-T3 the law of identity Cpp. The deduction requires
two irpplicationsof the rule of substitutionand two applications
ol'tlrr: rule of detachment;it runs thus:
Tr. qlCNpqxCT3-74
'l'4.. (iCCNpqrCpr

'l'+.qlp,rlpxCTz-T5
'l'5.t:pp.
'l'he lilst lirre is called the derivational line. It consistsof two
plrts s(:l)irrutcrlfrom each other by the sign x . The first part,
'l't. qlcNl)q, rrr"airsthat in Tr CNpq has to be substitut.a foi

8e

ARISTO TLE' S

SYSTEM I N SY, M B O L I C F O R M

$zg

q. The thesisproduced by this substitutionis omitted in order to

savespace.It would be of the following form:
(I) CCqCNp qCCCNpqrCpr.
The second part, CT3-T4, shows how this omitted thesis is
constructed;makingit obviousthat the rule ofdetachmentmay be
applied to it. Thesis (I) beginswith C, and then there follow axiom
T3 asantecedentand thesisT4 as consequent.We can therefore
detach T4 as a new thesis. The derivational line before T5 has
a similar explanation. The stroke (/) is the sign of substitution
and the short rule (-) the sign of detachment. Almost all subsequentdeductionsare performed in the samemanner.
One must be very expert in performing such proofs if one
wants to deducefrom the axioms Tr-T3 the law of commutation CCpCqrCqCpror even the law of simplification CltCqp. I
shall therefore explain an easy method of verifying expressions
of our systemwithout deducing them from the axioms. This
method, invented by the American logician Charles S. Peirce
about rBB5, is based on the so-calledprinciple of bivalence,
which statesthat every proposition is either true or false, i.e.
that it has one and only one bf two possibletruth-values: truth
and falsity. This principle must not be mixed up with the law of
the excluded middle, according to which of two contradictory
propositionsone must be true. It was stated as the basisof logic
by the Stoics,in particular by Chrysippus.I
All functions of the theory of deduction are truth-functions,
i.e. their truth and falsity depend only upon the truth and
falsity of their arguments.Let us denote a constant false proposition by o, and a constant true proposition by r. We may
define negation iri the following way:
and
No:r
Nt:o.
This means: the negation of a falsepropositionmeansthe same
as a true proposition (or;:'shortly,is true) and the negation of
a true proposition is false. For implication we have the following four definitions:
Coo :

t,

Cot

It

CIo

o,

Ctt

t.

I Cicero, Acad. pr. ii.

95 'Fundamentum dialecticae cst, quidquid enuntietur
(id autem appellant ,i,{iapo) aut verum esse aut falsum'; De fato zr 'Itaque
contendit omnes nervos Chrysippus ut persuadeat omne d.(iape aut verum esseaut
falsum.' In the Stoic terminology df/opc means 'proposition', not 'axiom'.

013

TH EOR Y

OF D ED U C TION

83

'Ihis means: an implication is falseonly when its antecedentis

true and its consequentfalse; in all the other casesit is true.
This is the oldestdefinition of implication, stated by Philon of
Megara and adopted by the Stoics.t For conjunction we have
the four evident eqpalities:
KIo : ot
Ktt : t.
Koo : o,
KoI : o,
A conjunction is true only when both its arguments are true;
in all the other casesit is false.
Now ifwe want to verify a significant expressionof the theory
of deduction containing all or someof the functors C, N, and K
we haveto substitutefor the variablesoccurring in the expression
the symbolso and.r in 4ll possiblepermutations,and reducethe
formulae thus obtainedon the basisofthe equalitiesgiven above.
If after the reduction all the formulae give r as the final result,
the expressionis true or a thesis; if any one of them giveso as
the final result, the expressionis false.Let us take as an example
of the first kind the law of transposition CCltqCNqNp;we get:
For pfo, qlo: CCooCNoNo: CrCrr : Crr : r,
,, !lo, qlr: CCotCNrNo : CrCor : CII : I,
,, llI, qlo: CCroCNoNr : CoCro : Coo : r,
,, llI, qlr: CCrrCNrNr : CrCoo: CIr : r.
As for all substitutionsthe final result is r, the law of transpositionis a thesisof our system.,Letus now take as an example
of the secondkind the expressionCKpNqq.It sufficesto try only
one substitution:
plt, qlo: CKrNoo : CKrro : CIo : o.
This substitution gives o as the final result, and therefore the
expressionCKpNqqis false. In the same way we may check all
the thesesof the theory of deduction employed as auxiliary
premissesin Aristotle's syllogistic.
\24. Quantifurs
Aristotle had no clear idea of quantifiersand did not usethem
in his works; consequentlywe cannot introduce them into his
syllogistic.But, as we have already seen,there are two points in
his systemwhich we can understand better if we explain them
r Sextus Empiricus, Ada. math. viii. rI3 6 y2v iDiAav iAeyev dA402syiveo|et td
ouv11t1tivov,6rav pi1 dpyqrot d.n' dAqflois xoi Arjy'g dzi ,y'e0Eos,titore rpq6ts plv
yiveo|ot xat' ait6v ,iLq9is owtlpplvov, xa|' tve 6t rp6nov ,ltea}os.

84

ARISTOTL E' S

SYSTEM I N SYM BO LI C

FORM

$e+

by employing quantifiers. Universal quantifiers are connected

with the so-called'syllogisticnecessity',existentialor particular
quantifierswith the proofsby ecthesis.I shall now translateinto
symbols the proofs with existential quantifiers set down in
section rg, and then the argument dependent on universal
quantifiers mentioned in section5.
I denote quantifiers by Greek capitals,the universal quantifier by II, and the particular or existential quantifier by X. I/
may be read 'for all', and ^t 'for some' or 'there exists'; e.g,
DcKAcbAcameansin words: 'There existsa c such that all c is
b and all c is a', or more briefly: 'For somec, all c is D and all c
is a.' Every quantified expression,for instanceDcKAcbAca,consists of three parts: part one, in our example J, is always a
quantifier; part two, here c, is always a variable bound by the
preceding quantifier; part three, here KAcbAca,is always a
propositionalexpressioncontaining the variable just bound by
the quantifier as a free variable. It is by putting Xc before
KAcbAcathat the free variable c in this last formula becomes
bound. We may put it briefly: X (part one) binds c (part two)
in KAcbAca(part.three).
The rules of existentialquantifiers have already been set out
in section r9. In derivational lines I denote by ^Xr the rule allowing us to put .E before the antecedent,and by 2z the rule
allowing us to put it before the consequentof a true implication. T[e following deductionswill be easilyunderstood,asthey
are translationsof the deductions given in words in section 19,
the correspondingthesesbearing the samerunning number and
havingcorrespondingsmalllettersasvariablesinsteadof capitals.
of the l-premiss
Proof of conuersion
Thesesassumedas true without proof:
(t) CIabEcKAcbAca
(z) CEcKAcbAcaIab
''
Theses(r) and (z) can be usedasa definition of the l-premiss.
(commutative law of conjunction)
(g) CKpqKqp
$) plAch,qlAcax(4)
@) CKAcbAcaKAcaAcb
@) Ezcx (5)
CKAcbAcaEcKAcaAcb
$)

$r+

QU AN TIFIER S

85

(5) Irc x (6)

(6) CEcKAcbAcaDcK AcaAcb
Tt. CCpqCCqrCpr (law of the hypothetical syllogism)
T t. p I lab, qIEcKAcbAca,r f2cKAcaAcbx C(r) -C(0) -( Z)
CIabEcKAcaAcb
fi)
(z) bla,albx(B)
(B) CEcKAcaAcblba
Tr. pllab, qlzcKAcaAcb,rllba x C(7)-C(B)-(9)
(g) CIabIba
The derivational lines show that (4) and (B) result from other
thesesby substitution only, and (7) and (g) by substitution and
two detachments.Upon this pattern the reader himself may try
to construct the proof of the mood Darapti, which is easy.
Proof of themoodBocardo
(The variables P, R, and ,S used in section rg must be reIettered,asthe correspondingsmall lettersp, r, ands are reserved
to denotepropositionalvariables:write dforP,afor R,and Dfor.S.)
Thesis assumedwithout proof:
(t) CObdEcKAcbEcd
Two syllogismstake-nas premisses:
(16) CKAcbAbaAca (Barbara)
$7) CKAcaEcdOad (Felapton)
T6. CCKpqrCCKrstCKKpost
This is the 'synthetic theorem' ascribedto Aristotle.
T6. plAcb,qlAba,rfAca,slEcd,tlOadxC(r6)-C(r 7)('8)
(tB) CKKAcbAbaEcdOad
T7. CCKKpqrsCKprCqs (auxiliarythesis)
T7. plAcb,qlAba,rlEcd,slOadx C(r8)-(r9)
CKAcbEcdCAbaOad
Q$
(tg) Etc x (zo)
(zo) CZcKAcbEcdCAbaO
ad
Tr. CCpqCCqrCpr
T r. p I Obd, q| 2cKAcbEcd,rI CAbaOad x C(t ) -C (zo)(zr)
(zt) CObdCAbaOad

86

ARISTOTL E' S

SYSTEM I N SYM BO LIC

FORM

$z+

This is the implicational form of the mood Bocardo. If we wish

to have the usual conjunctional form of this mood, we must
apply to (zr) the so-calledlaw of importation:
T8. CCpCqrCKpqr.
We get:
TB. p I Obd,qI Aba,rI Oadx C(zt)-(zz)
(zz)CKObdAbaOad

(Bocardo).

By the so-calledlaw of exportation,

Tg. CCKpqrCpCqr,
which is the converseof the law of importation, we can get
the implicational form of the mood Bocardo back from its
conjunctionalform.
The rules of universal quantifiers are similar to the rules of
particular quantifiers set out in section rg. The universal
quantifier can be put before the antecedentof a true implication unconditionally, binding a free variable occurring in the
antecedent,and before the consequentof a true implication
only under the condition that the variable which is to be bound
in the consequentdoes not occur in the antecedentas a free
variable. f denote the first of these rules by IIr, the second
by IIz.
Two derived rules result from the above primitive rules of
universal quantifiers: first, it is permissible(by rule IIz and the
law of simplification) to put universal quantifiers in front of a
true expressionbinding free variablesoccurring in it; secondly,
it is permissible (by rule Zr and the propositional law of
identity) to drop universalquantifiersstandingin front of a true
expression.How theserules may be derived I shall explain by
the example of the law of conversionof the /-premiss.
From the law ofconvrsion
(g) CIabIba
there follows the quantified expression
(zG) IIaIIbClablba,
and from the quantified expression(26) there follows again the
unquantified law ofconversion (9).

q U AN TIFIER S

$24

87

First: from (9) follows (26).

(law of simplification)
Trc. CpCqp
Trc. plClablbaxC($-(4)
(4) CqCIabIba
To this thesiswe apply rule I/: binding b, and then a, as neither
b nor a occursin the antecedent:
(zg) IIzbx(z+)
(z$ CqIIbCIabIba
(24) Ilzax (zS)
(2il CqIIaIIbCIablba
(25) qlcpcqp x CTro-(26)
(26) IIaIIbCIabIba
Secondly:from (26) follows (9).
(law of identity)
TS. Cpp
T5. plClablbax (27)
(27) CClabIbaClablba
To this thesiswe apply rule I/r binding b, and,then a:
(27) Ihb x (zB)
(zB) CII\C labIbaCIablba
(zB) IIru x (zs)
(zg) CII aII bCIabIb aCIabIb a

(zg)x c(26)-(e)
(g) CIabIba
Aristotle asserts:'If some a is b, it is necessarythat some 6
should be a.' The expression'it is necessarythat' can have, in
my opinion, only this meaning: it is impossibleto find such
values of the variables a and D as would verifr the antecedent
without verifying the consequent.That means' in other words:
'For all a, and for all D,if somea is D, then someD is a.'This is
our quantified thesis(26). It has been proved that this thesisis
equivalent to the unquantified law of conversion'If somea is 6,
then someb is a', which doesnot contain the sign of necessity.
Since the syllogistic necessityis equivalent to a universal
quantifier it may be omitted, as a universal quantifier may be
omitted at the head of a true formula.

88

ARISTOTL E' S

SYSTEM I N SYM BO LI C

FORM

$rs

oJ thes2llogistic
| 25. Fundamentals
Every axiomatized deductive systemis basedon three fundamental elements:primitive terms,axioms,and rulesof inference.
I start from thefundamentals for assertedexpressions,the fundamental elementsfor the rejected ones being given later.
As primitive terms I take the constantsI and d defining by
them the two other constants,E and O:
Df t. Eab : NIab
Df z. Oab : NAab.
In order to abbreviate the proofs I shall employ instead of the
above definitionsthe two following rules of inference:
Rule RE: NI may be everywhere replaced by ,E and conversely.
Rule RO: NA may be everywherereplaced by O and con.
versely.
The four thesesof the system axiomatically assertedare the
two laws of identiry and the moodsBarbara and Datisi:
t. Aan
z. Iaa
3. CKAbcAabAac (Barbara)
4.CKAbcIbaIac (Datisi).
Besidesthe rules RE and RO I acceptthe two following rules
of inferencefor the assertedexpressions:
(a) Rule of substitution: If cris an assertedexpressionof the
system, then any expressionproduced from a by a valid
substitutionis also an assertedexpression.The only valid
substitution is to put for term-variables a, b, c other termvariables,e.g. b for a.
(D) Rule of detachment: If Cup and d are assertedexpressions
of the system, then B i6 an assertedexpression.
As an auxiliary theory I assumethe C-"ly'-system
of the theory
of deduction with K as a defined functor. For propositional
variables propositional expressionsof the syllogistic may be
substituted,like Aab, Iac, KEbcAa6,etc. In all subsequentproofs
(and also for rejected expressions)I shall employ only the
following fourteen thesesdenoted by roman numerals:

$zs

F U N D A M E N T AL S

OF TH E SYL L OGISTIC

89

I. CpCqp
law of simplification)
II. CCqrCCpqCpr aw ofhypothetical syllogism, znd form)
lll. CCpCqrCqCpr aw of commutation)
aw of Duns Scotus)
IV. CpCNpq
aw of Clavius)
v. CCNppp
aw of transposition)
YI. CCpqCNqNp
YII. CCKpqrCpCqr aw of exportation)
YIII. CpCCKpqrCqr
lX. CCspCCKpqrCI{sqr
X. CCKpqrCCsqCKpsr
XI. CCrsCCKpqrCKqps
XII. CCKpqrCKpNrNq
XJII. CCKpqrCKNrqNp
){lV. CCKpNqNrCKprq
ThesisVIII is a form of the law of exportation, thesesIX-XI
are compound laws of hypothetical syllogism, and XII-XIV
are compound laws of transposition.All of thesecan be easily
verified by the o-r method explained in section 23. ThesesIV
but
and V give togetherwith II and III the whole C-.M-system,
IV and V are needed only in proofs for rejected expressions.
The systemof a;<iomsI-4 is consistent,i.e. non-contradictory.
The easiestproof of non-contradiction is effectedby regarding
term-variables as proposition-variables,and by defining the
functions A and.I as always true, i.e. by putting Aab : Iab :
KCaaCbb.The axioms r-4 are thep true as thesesof the theory
of deduction, and as it is known that the theory of deduction is
non-contradictory,the syllogisticis non-contradictorytoo.
All the axioms of our systemare independentof each other.
The proofs of this may be given by interpretation in the field of
the theory of deduction. In the subsequentinterpretations the
term-variablesare treated as propositional variables.
Independenceof axiom r : Take Kfor A, and Cfor L Axiom r
is not verified,for Aaa : Kaa, and Kaa giveso for afo. The other
axiomsare verified, as can be seenby the o-r method.
Independenceof axiom z: Take C for A, and fffor .L Axiom z
is not verified., for Iaa : Kaa, The other axioms are verified.
Independenceof axiom 4: Take C for A and -L Axiom 4 is
not verified, for CKAbcIbaIac: CKCbcCbaCacgives o for bfo,
alr, cfo. The rest are verified.

90

ARISTOTLE'S

SYSTEM I N SYM BO LI C

FORM

$zS

Independenceof axiom 3: it is impossibleto prove the independenceofthis axiom on the ground ofa theory ofdeduction
with only two truth-values,o and r. We must introduce a third
truth-value, let us say :, which may be regarded as another
symbol for truth, i.e. for r. To the equivalencesgiven for C, Jf,
and K in section 23, we have to add the following formulae:
COZ:Cr2:CZt:C2Z:
r,
Czo:O,
Ko2 : Kzo: ot
Ktz : Kzt : Kzz : t.

N2:o,

It can easilybe shown that under theseconditionsall the theses

of the C-"lf-systemare verified. Let us now define lab as afunction always true, i.e. Iab : r for all values of a and b, and Aab
as a function with the values
Aaa: t, Aot : Atz: r, and Aoz: o (the rest is irrelevant).
Axionis r, 2, and 4 are verified, but from 3 we get by the substitutions bfr, ef z, alo: CKAtzAorAoz : CKrrc : Cro : o.
It is also possibleto give proofs of independenceby interpretation in the field of natural numbers. If we want, for instance,to prove that axiom 3 is independentof the remaining
axioms,we can defineAabasa+I + b, and Iab asa*b : b*a.
Iab is always true, and therefore axioms z and 4 are verified.
Axiom r is also verified, for alt is alwaysdifferent from a. But
axiom 3, i.e. 'If b*t * c and af t t' b, then aIt # c', is not
verifiediTake 3 for a,2 for D, and 4for c: the premisseswill be
true and the conclusionfalse.
It resultsfrom the above proofs of independencethat there
existsno single axiom or 'principle' of the syllogistic.The four
axioms r-+ rr.ay be mechanicallyconjoined by the word 'and'
into one proposition,but they remain distinct in this inorganic
conjunction without representingone singleidea.
$ 26, Deductionof s2llogisriitheses
From axioms I-4 we can derive all the thesesof the Aristotelian logic by meansof our rules of inferenceand by the help of
the theory of deduction. I hope that the subsequentproofs will
be quite intelligible after the explanationsgiven in the foregoing sections.In all syllogistical moods the major term is
denoted by a, the middle term by b, and the minor term by c.

$26

DEDUCTION

OF SYL L OGISTIC

TH ESES

9r

T'he major premissis stated first, so that it is easyto compare

thc formulae with the traditional namesof the moods.r
A. Tne Lnws or CorqvrRsroN
Yll.'plAbc, qflba, rf lacxC4-5
5. CAbcClbalac
S. bla, cfa, afb x Cr-6
(law of conversionof the .I-premiss)
6. CIabIba
IlI. plAbc, qllba, rllacxCS-7
7. ClbaCAbclac
7. bla,clb xCz-B
(law. of subordination for affirmative preB. CAabIab
mrssesJ
ll. qllab, rlIbaxC6-g
g. CCplabCpIba
g. plAabx CB-ro
(law of conversionof the ,4-premiss)
rc. CAabIba
6. alb,blax rt
t r. CIbaIab
Yl. p I lba, qllab x Ct r-tz
n. CNIabNIb&
rz. RE x r3
(law of conversionof the E-premiss)
ry. CEabEba
pIAab,
qlIabxCB-r'4
YL
t4. CNIabNAab
r4. RE, RO x r5
r 5. CEabOaD (law ofsubordinationfor negativepremisses)
B. Tnr Aprrnuarrvn
X. ltlAbc, qllba, rllac x C4-r6
fi. CCsIbaCKAbcslac
fi. sllabxC6-I7
r7. CKAbcIabIac

Moons

(Darii)

t In my Polish text-book, Elcmcnts of Matlumatieal Zogdc,published in rgzg (see

p. 46, n. 3), I showed for the first time how the known theses of the syllogistic may
be formally deduced from axioms r-4 (pp. l8o-go). The method erpounded in
the above text-book is accepted with sorne modifications by L M. Bocheiski, O.P.,
in his contribution: On thc CategoricalS2llogism,Dominican Studies, vol. i, Oxford

('g+B).

ARISTOT LE' S

SYSTEM I N SYM BO L I C

16.slAabx Cro-IB
fi. CKAbcAabIac
B. alb,blax rg
ry. CAbaIba
fi. slAbaxCrg-zo
zo. CKAbcAbaIac
Xl. rllba, sllab xCtr-zt
zr. CCKpqIbaCKqplab
4. cla,alcx22
zz. CKAbalbclca
zr. plAba,qlIbc,blcxCzz-23
zg. CKIbcAbaIac
t7. cfa, alcxz4
24. CKAbaIcblca
2r. plAba, qllcb,blcxCz4-25
25. CKIcbAbaIac
rB.cla,alcxz6
c6. CKAbaAcblca
zt. plAba,qlAcb,blcxCz6-27
27. CKAcbAbaIac
C. TnE Nncerrvn

FORM

(Barbari)

(DaraPti)

(Disamis)

(Dimaris)

(BramantiP)
Mooos

, XIII. pllbc,qlAba,rflacxC4-zB
zB. CKNIacAbaNIbc
zB. RE x z9
zg. CKEacAbaEbc
zg. alb, blax3o
(Celarent)
go.CKEbcAabEac
plEba
Cr3-3t
x
IX. slEab,
y:CCKEbaqrCKEabE
gt. afc, qlAab,rlEatxC3o-32
(Cesare)
gz. CKEcbAabEac
g-Zg
Xl. rIEab,sIEbax Ct
93.CCKpqEabCKqpEba
gz.cla,alcx34
CKEabAcbEca
94.

S z6

$26

D E D U C T I ON

OF SYL L OGISTIC

TH ESES

3g.PlEab,qlAcb,afc,blaxCS+-yS
(Camestres)
CtiitibotEai
35.
3o.cla,alcxgG
36.CKEbaAcbEca
33.PlEba,qlAcb,alc,blaxC3617
(Camenes)
g7.ctiitimonoi
lI. qIEab,rI OabxCr5-38
38.CCPEabCPOab
blcx C3o-39
35.PlKEbcAab,
(Celaront)
gg.cinOiAaboac
bIcx C3z-4o
38.pIKEcbAab,
(Cesaro)
CKEcbAaboac
4o.
blc x C35-4r
38.pIKAcbEab,
(CamestroP)
CKAcbEabOac
4t.
bfcxCZT-42
35.PlKAcbEba,
(CamenoP)
ci,lcibtaoac
42.
XII'I. PlAbc,qllba,rflacxC+-+g
CKNIacIhaNAbc
43.
43,RE, ROx44
44.CKEacIbaObc
+ 4 . e lbb, lax 4 5
(Ferio)
criaciaioac
4,,.
3r. afc,qllab,rlOatx'C45-46
(Festino)
(:IiEcbiabbiac
46.
X. PlEbc,qllab,rlOacxC+S-+7
47. (|CsIabCKEbcsOac
4.7.sllbaxCrr-48
()ft'libclbaoac
4tt.
3r. afc, qllba, rlOacxC4B-49
4y. (|h'li:blbaoac

(Ferison)

ro. afb, blax5o

,ro.(i,'llnlah
47. tl,lbaxC5o-5r
qr. (;A'lihtlhuooc
'.1r.ttfc, qlAba,rlOa"cxC5r-52
,,r. t:i:ltt,)lh,i)uc

(Felapton)

(Fresison)

(FesaPo)

ARISTOI|LE'SSYSTEMIN SYMBOLICFORM
$26
s4
As a result of all these deductions one remarkable fact deservesour attention: it was possibleto deduce twenty syllogistic moods without employing axiom 3, the mood Barbara.
Even Barbari could be proved without Barbara. Axiom 3 is the
most important thesisof the syllogistic,for it is the only syllogism that yields a universal affirmative conclusion,but in the
systemof simple syllogismsit has an inferior rank, being necessary to prove only two syllogisticmoods,Baroco and Bocardo.
Here are thesetwo proofs:
XIl. plAbc,qlAalt,rfAacxCZ-SS
CKAbcNAacNAab
59.
53. RO x 54

54.CKAbcOacOab
5+.blc,clbx55
(Baroco)
55.CKAcbOabOac
XllI' plAbc,qfAab,rfAacx Cg-S6
CKNAacAabNAbc
56.
56.RO x 57
57.CKOacAabObc
blax 58
57.'alb,
(Bocardo)
CKObcAba.Oac
58.

expressions
$ 27. Axiomsand rulesfor rejected
Of two intellectual acts, to asserta proposition and to reject
it,I only the first has been taken into account in modern formal
logic. Gottlob Frege introduced into logic the idea of assertion,
and the sign of assertion(F), acceptedafterwardsby the authors
The idea of rejection, however, so far
of PrincipiaMathematica.
as I know, has been neglectedup to the presentday.
We asserttrue propositionsand reject false ones. Only true
propositionscan be asserted,for it would be an error to assert
a proposition that was not true. An analogousproperty cannot
be assertedof rejection: it is not o4ly false propositions that
have to be rejected.It is true, ofcourse, that every proposition
is either true or false, but there exist propositional expressions
that are neither true nor false. Of this kind are the so-called
propositionalfunctions,i.e. expressions
containingfree variables
t f owe this distinction to Franz Brentano, who describes the acts of believing as
ancrkmnen and vcrwerfen.

$ a7

R E J E C TED

EXPR ESSION S

95

syllogisticform therefore:
(i) CKAbcEablac
is not assertedby him as a valid syllogism, but rejected'
Aristotle himself gives concrete terms disproving the above
lorm: take for 6 'min', for c 'animal', and for a 'stone'.But.there
are other valuesfor which the formula (i) can be verified: by
identifying the variables a and t we get a true implication
(,'X.'AbaEablaa,
for its antecedentis falseand its consequenttrue'
of
1'he negation the formula (i) :
(j) NCKAbcEablac
must therefori be rejectedtoo, becausefor cfa it is false.
By introducing quantifiersinto the systemwe could dispense
with rejection. Instead of rejecting the form (i) we could assert
tlre thesis:
lac.
(k) ZaZbEcNCKAbcEab
'f 'lrismeans:there exist terms a, b, antcJ.
c that verify the negation
of'(i) . The form (f), thercfore,is not true for all a, b, and c, and
t:annotbe a valid syllogism.In the sarneway insteadof rejecting
the expression(.r) we might assertthe thesis:
(l) DaZbEcCKAbcEablac.
Itut Aristotle knows nothing of quantifiers; insteadof adding to
his systemnew theseswith quantifiershe usesrejection' As
rrjcction seemsto be ersimpler idea than quantification,let us
lirllowin Aristotle'sstt:ps.

96

ARISTOTL E'S

SYSTEM I N SYM BO LI C

FORM

$eZ

Aristotle rejectsmost invalid syllogisticforms by exemplification through concrete terms. This is the only point where we
cannot follow him, becausewe cannot introduce into logic such
concrete terms as 'man' or 'animal'. Some forms must be
rejected axiomatically. I have foundt that if we reject axiomatically the two following forms of the secondfigure:
CKAcbAablac
CKEcbEablac,
all the other invalid syllogisticforms may be rejectedby means
of two rules of rejection:
(r) Rule of rejection by detachment: if the implication 'If cr,
but the consequentp is rejected,then
then p'is asserted,
the antecedentd must be rejectedtoo.
(d) Rule of rejection by substitution: if p is a substitutionof
cr,?nd p is rejected,then a must be rejectedtoo.
Both rules are perfectly evident.
The number of syllogisticforms is 4 x 43 : z16i z4 forms are
valid syllogisms,2 forms are rejectedaxiomatically. It would be
tedious to prove that the remaining z3o invalid forms may be
rejected by meansof our axioms and rules. I shall only show,
by the example of the forms of the first figure with premisses
Abc and Eab, how our rules of rejection work on the basisof
the first axiom ofrejection.
RejectedexpressionsI denote by an asteriskput before their
serialnumber. Thus we have:
*59. CKAcbAabIac
(Axiom)
*59a. CKEcbEabIac
'
l. pllac, qlKAcbAabx6o
6o. CIacCKAcbAabIac
6o x C*6r-*59
*6t. Iac.
Here for the first time is applied the rule of rejection by
detachment. The assertedimplication 6o has a rejected consequent,*59; thereforeitsantecedent,*6r, must be rejectedtoo.
*64, *67. *7r,
In this sameway I get the rejectedexpressions
*74" and +77.
I Sce section zo.

! ez

REJECTED

EXPR ESSION S

97

Y. pllacx6z
62. CCNIacIacIar
62. RE x 63
63. CCEacIacIac
63 x C*64-*61
r64. CEacIac
r. alcx65
65. Acc
YIII. plAcc, 8lEac,rf lacxC65-66
66. CCKAccEacIatC Eaclac
66 x C*67-*64
r67. CKAccEaclac
*67 x*68. blc
rGB. CKAbcEabIac
the rule ofrejection by substitutionis applied. Expression
^I.r.
*_68
mustberejected,because
by the substitutionof Afo. cin * 68we
get the rejectedexpression+67.Thesameruleisusedtoget*
75.
ll. qlAab, rllab xCB-69
69. CCpAabCpIab
69.plKAbcEab,blcxTo
7o. CCK AbcEabAafi K A bcEablac
ToxC*7v*68
*7t. CKAbcEabAac
XlY. p I Acb,I I lac, rI Aabx 7z
7z. CCKAcbNIacNAabCKAcbAablar
72. RE, RO x 73
7g. CCKAcbEacOabCKAcbAablac

Tgxc*74- *sg

*74. CKAcbEacOab
*74x*75. blc,clb
*75. CKAbcEabOac
38.plKAbcEab,blcx76
76. CCKAbcEabEacCKAbcEabOac
76 xC*77_*75
*77. CKAbcEabEac
'fhe rejectedexpressions*68, t7r, *75, and *77 are the four

98

ARISTO TLE' S

SYSTEM I N SYM B O L I C

FORM

$zZ

possible forms of the first figure having as premissesAbc and

Eab. From thesepremissesno valid conclusioncan be drawn in
the first figure. We can prove in the same way on the basis of
the two axiomatically rejected forms that all the other invalid
syllogistic forms in all the four figures must be rejected too.
$ zB. Insuficiencyof our axiomsand rules
Although it is possible to prove all the known thesesof the
Aristotelian logic by mearu of our axiomsand rules of assertion,
and to disprove all the invalid syllogistic forms by means of our
axioms and rules of rejection, the result is far from being satisfactory. The reason is that besides the syllogistic forms there
exist many other significant expressionsin the Aristotelian logic,
indeed an infinity of them, so that we cannot be sure whether
from our system of axioms and rules all the true expressionsof
the syllogistic can be deduced or not, and whether all the false
expressionscan be rejected or not. In fact, it is easyto find false
expressionsthat cannot be rejected by means of our axioms and
rules of rejection. Such, for instance,is the expression:
(Fr) CIabCNAabAba.
ft means: 'If somec is D, then if it is not true that all a is b, all
b is a.' This expressionis not true in the Aristotelian logic, and
cannot be proved by the axiomsof assertion,but it is consistent
with them and added to the axioms doesnot entail any invalid
syllogistieform. It is worth while to considerthi systemof the
syllogistic as thus extended.
From the laws of the Aristotelian logic:
B. CAabIab
and
CAbaIab
5o.
and the law of ihe theory of deduction:
(n) CCprCCqrCCNpqr
we can derive the followirl'$new thesis78:
(n) plAab, qlAba, rllab xCB-Cp-7B
CCNAabAbaIab.
78.
This thesisis a converseimplication with regard to (Fr), and
together with (Fr) givesan equivalence.On the.ground of this
equivalencewe may define the functor 1by the functor ,4:
(Fz) Iab : CNAabAba.

$ z8

INSUFFICIENCY

OF OU R AXIOM S

AN D R U L ES

gg

This definition reads: ' "Some a is b" meansthe same as "If it

is not true that all a is 6, then all 6 is d".' As the expression'If
not-1, then 4' is equivalent to the alternation 'Eitherp or q',we
can alsosay: t "Some a is b" meansthe sameas "Either all a is
D or all b is a".' ft is now easyto find an interpretation of this
extended systemin the so-called Eulerian circles. The terms a,
b, c are represented by circles, as in the usual interpretation,
but on the confition that no two circles shall intersect each
other. Axioms r-4are verified, and theforms *59 CKAcbAabIac
and.*5ga CKEcbEablac
are rejected,becauseit is possibleto draw
two circles lying outside each other and included in a third
circle, which refutes the form CKAcbAabIac,and to draw three
circles each excluding the two others, which refutes the form
CKEcbEabIac.Coruequently all the laws of the Aristotelian
logic are verified, and all the invalid syllogistic forms are rejected. The system,however, is different from the Aristotelian
syllogistic,becausethe formula (FI) is false,as we can seefrom
the following example: it is true that 'Some even numbers are
divisible.by 31,but it is true neither that 'All evennumbersare
divisible by 3'nor that 'All numbersdivisible by 3 are even'.
It results from this consideration that our system of axioms
and rules is not categorical,i.e. not all interpretations of our
system veri$ and falsify the same formulae or are isomorphic.
The interpretation just expounded verifies the formula (Ft)
which is not verified by the Aristotelian logic. The systemof our
axioms and rules, therefore,is not sufficient to give a full and
exact description of the Aristotelian syllogistic.
In order to remove this difficulty we could reject the expression (Fr) axiomatically. But it is doubtful whether this remedy
would be effective; there may be other formulae of the same
kind as (Fr), perhapseven an infinite number of suchformulae.
'Ihe problem is to find a systemof axioms and rules for the
Aristotelian syllogisticon which we could decide whether any
given significant expressionofthis systemhas to be assertedor
rejected.To this most important problem of decisionthe next
r:hapteris devoted.

0zg

CHAPTER

THE PRO B LE M

OF DECISION

expressions
$ zg. The numberof undecidable
I tere,as the basis of my present investigation the following
fundamental elementsof the syllogistic:
(t) The four assertedaxioms r-4.
(z) The rule (a) of substitutionand the rule (6) of detachment
for the assertedexpressions.
(3) The two rejectedaxioms*59 and *59a.
(a) The rule (c) of detachmentand the rule (d) ofsubstitution
for the rejectedexpressions.
To this systemof axiomsand rules the theory of deduction must
be added as the auxiliary theory. From the axioms and rules of
assertion there can be derived all the known thesesof the
Aristotelian logic, i.e. the laws of the square of opposition,
the laws of conversion,and all the valid syllogistic moods;
on the basisof the axioms and rules of rejection all the invalid
syllogisticalforms can be rejected.But, as we have already seen,
this systemof axioms and rules does not sufficeto describethe
Aristotelian syllogistic adequately, becausethere exist signififor instanceCIabCNAabAba,whichcan neither
cant expres$ions,
be proved by our axioms and rules of assertionnor disproved
by our axioms and rules of rejection. I call such expressions
undecidablewith respectto our basis.Undecidable expressions
may be either true in the Aristotelian logic or false.The expression CIabCNAabAbais, of course,false.
There are two questionswe have to settle on this basis in
order to solve the problem of decision.The first question is, Is
finite or not? If it is
the number of undecidable',expressions
finite, the problem of decisionis easily solved: we may accept
as new assertedaxioms,and reject falseexprestrue expressions
sions axiomatically. This method, however, is not practicable
if the number ofundecidable expressions
is not finite. We cannot
assertor reject an infinity of axioms. A secondquestion arises
in this case:Is it possibleto complete our systemof axioms and
rules so that we could decidewhether a given expressionhad to

THE NUMBER

OF U N D EC ID ABL E

EXPR ESSION S

to t

be assertedor rejected? Both these questionswere solved by

Slupecki: the first negatively by showing that the number of
undecidable expressionson our basis is not finite, the second
affirmatively by the addition of a new rule of rejection.r
I begin with the first question. Every student of the traditional logic is familiar with the interpretation of syllogisms
by meansof Eulerian circles: according to this interpretation
the term-variables a, b, r are represented by circles, the
premiss Aab being true when and only when the circle a is
either identical with the circle 6 or is included in 6, and the
premiss lab being true when and only when the circles a and b
have a common area. Consequentlythe premiss Eab, as the
negation of lab, is true when and only when the circles a
and b have no common area, i.e. when they exclude each
other. If, therefore,a and D are identical,Iab is true and Eab is
false.
I shall now investigate various suppositions concerning the
number of circles assumedas our 'universe of discourse',i.e.
as the field of our interpretation. It is obvious that the rules of
our basisremain valid throughout all the interpretations.If our
universe of discourseconsistsof three circles or more, the four
axioms of assertionare of courseverified, and the axiomatically
rejected expression
*59. CKAcbAablac
is rejected,as it is possibleto draw two circlesc and,a excluding
each other and both included in the third circle 6. The premisses
Acb and Aab are then true. and the conclusion .Iacis false. The
expression
*59a. CKEcbEabIac
alsois rejected,as we can draw three circleseach excluding the
two others,so that the premissesEcb and Eab are true and the
conclusion 1ac is false. This interpretation therefore satisfies
the conditions of our basis,and so do all our other interpretations.
Let us now supposethat our universeof discourseconsistsof
I See the paper of Slupecki quoted inp.
76, n. I have tried to simplify the author's
arguments in order to make them comprehensible to readers not trained in mathematical thinking. I am, of course, alone responsible for the following exposition
of Slupecki's ideas.

t o2

THE PRO BLEM

O F DECI SI O N

$eg

only three circles, but no more, and let us consider the following
expression:
(F g) CE abCEacCEadCEbcCEbdIcd.

that only three circlescan be drawn, the expression(F3) is true

and cannot be disproved by our axioms and rules oi rejection.
If we suppose,however, that our universeof discourse'consists
of more than three circles,we can drawfourcircles,eachof them
excluding the three others, and (F3) becomesfalse. (F3) therefore cannot be proved by our axioms and ruresof asseiiion.As
(F3) can neither be proved nor disproved by the systemofour
axioms and rules, it is an undecidableexpression.'
Let us now consideran expressionofthe form
(F4) CurC,a2Car...Canp,
containing z different variables:
a b 4 2 r 4 g , ..., An ,

becomesfalse or the consequentis true. But if our universeof

!rg

THE NUMBER

O F U N D EC ID ABL E

EXPR ESSION S

to g

Such undecidable expressionsare infinite in number, as z

may be any integer whatever. It is obvious that they are all
false in the Aristotelian logic, and must be rejected, for we cannot restrict the Aristotelian logic to a finite number of terms,
and expressionsof the form (F4) are disproved when the number
of terms is infinite. This infinite number of undecidableexpressionscannot be rejected otherwise than axiomatically, as results
from the following consideration: (F3) cannot be disproved by
the system of our axioms and rules, and therefore must be
rejectedaxiomatically. The next undecidableexpressionof the
form (F4) containing five different terms cannot be disproved
by our system of axioms and rules together with the already
rejected expression(F3), and must again be rejected axiomatically. The sameargument may be repeated with respect to every
other undecidable expressionof the form (F4). Since it is impossibleto reject axiomatically an infinity of expressions,we
must look for another device if we want to solvethe problem of
decision affirmatively.
$ go. Slupuki's rule of rcjntion
I start from two terminological remarks: Expressionsof the
type Aab, Iab, Eab, and Oab I call simple expressions;the first
two are simple affirmative expressions,and the third and fourth
simple negative expressions.Simple expressionsas well as expressionsof the type:
CctlCc'sCo,s...Cun-1an,

where all the c's are simple expressions, I call elementary

expressions.With the help of this terminolory Slupecki's rule
of rejection'may be formulated as follows:
If a and p are simple negative expressionsand y is an elementary expression,then if Cuy and CBy are rejected, CuCpy
must be rejected too.
Slupecki's rule of rejection has a close corurexion with the
following metalogical principle of traditional logic: 'utraque si
praemissaneget, nil inde sequetur.' This principle, however,
is not general enough, as it refers only to simple syllogismsof
three terms. Another formulation of the same principle, 'ex
mere negativis nihil sequitur', is apparently more general, but
it is falsewhen applied not only to syllogismsbut also to other

ro4

THE PRO BLEM

O F DECI SI O N

$So

expressions of the syllogistic. Such theses as CEabEba or

CEabOabshow clearly that something does follow from merely
negative premisses.Slupecki's rule is a general rule.and avoids
the awkwardnessof traditional formulations.
Let us explain this point more fully in order to make Slupecki's rule clear. The proposition Aac does not follow either
from the premissAab or from the premissAbc; but when we
conjoin these premisses,saying 'Aab and Abc', we get the conclusion Aat by the mood Barbara. Eat doesnot follow from Ebc,
or from Aab' either: but from the conjunction of thesepremisses
'Ebc and Aab' we get the conclusionEac by the mood Celarent.
In both caseswe obtain from the conjunction of premissessome
new proposition which does not result from either of them
separately. If we have, however, two negative premisses,like
Ecb and,Eab, we can of course obtain from the fint the conclusion Ocband from the secondOab,but from the conjunction
of these premissesno new proposition can be drawn except
those that follow from each of them separately. This is the
meaning of Slupecki's rule of rejection: if 7 does not follow
either from c or from B, it cannot follow from their conjunction,
as nothing can be drawn from two negative premissesthat does
not follow from them separately. Slupecki's rule is as plain as
the corresponding principle of traditional logic.
I shall now show how this rule can be applied in the rejection
of undecidableexpressions.
For this purposeI use the rule in a
symbolic form, denoted by RS (Rule of Slupecki):
*CaCpy.
RS. *Ccy, *CFy
Here as everywhere I employ Greek letters to denote variable
expressions
satisfyingcertain conditions: thus, c and p must be
simple negative expressionsof the syllogistic, y must be an
elementary expression as explained above, and all three expressionsmust be such that.Ccy and,Cpy may be rejected. The
arrow (--+) means 'therefore'. I want to lay stresson the fact
that RS is a peculiar rule, valid only for negative expressions
a and p of the Aristotelian logic, and, as we have already seen,
cannot be applied to affirmative expressionsof the syllogistic.
Nor can it be applied to the theory of deduction. This results
from the following example: the expressions CJfCpqr and
CNCqpr are both not true and would be rejected, if rejection

0go

ST UPECKI'S

RULE

OF R EJEC TION

to 5

were introduced into this theory, but CNCpqCNCqpris a thesis.

Also in algebra the proposition 'a equals D' does not follow
either from the premiss 'a is not lessthan D' or from the premiss
'6 is not lessthan a', but it follows from the conjunction of these
premisses.
As the first application of the new rule I shall show that the
expression
*59a. CKEcbEablac,
which was rejected axiomatically, can now be disproved.This
resultsfrom the following deduction:
g. plEac,alc,blaxTg
CCEaclcaCEaclac
79.
79 x C*Bo-*64
*Bo. CEailca
*Bo x *Br. cla,blc,alc
*Br. CEcbIac
+64x*Bz. blc
*Bz. CEabIat
R.S.ulEcb,BlEab,yllatx *Bt, *Bz --* *83
*83. CEcbCEabIac.
The rule RS is here applied for the first time; a and p are
simple negative expressions,and,.y is also a simple expression.
From *83 we get by the law of exportationVII the formula *59a :
YII. plEcb, qlEab,rlla"c'x&4
84. CCK EcbEabIacCEcbCEablac
84 x C*59a-*83
*59a. CKEcbEablac.
It follows from the above that Slupecki'srule is stronger than
our axiomatically rejected expression*59a. Since *5ga has to
be cancelled,formula *59, i... CKAcbAabIac,remains the sole
expressionaxiomatically rejected.
In the secondplace I shall apply the rule RS repeatedly to
disprovethe formula (F3):
*64x *85. dlc, cla
*85. CEadIcd
*Bsx*86. bla
*86. CEbdlcd

to 6

THE

PRO BLEM

O F DECI S I O N

$go

F.S. ulEad, BlEbd, ylhdx *85, *86 --->*87

*87. CEadCEbdlcd
*Bo x *BB. bla, dla
*BB. CEbclcd
RS. c/Ebc,PlEbd,ylhdx *BB, *86 --- *89
*Bg. CEbcCEbdlcd
R.S.alEad, plEbc, ylCEbdlcdx *87, *Bg ---'*go
*go. CEadCEbcCEbdlcd
*BBx*gt. afb
*gr. CEacIcd
RS. c/Eac,plEbd, ylhdx *gr, *86 --- *92
*gz. CEacCEbdlcd
R.S.afEac,plEbc, ylCEbdhdx *gz, *Bg--* *93
*gg. CEacCEbcCEbdlcd
RS. a/Eac,plEad, ylCEbcCEbdhdx*g7, *9o --->*94
*
Ebdlcd
94. CEacCEadCEbcC
*B5x*95. bld
*95. CEabIcd
F.S.alEab,plEbd, yllcdx *95, *86--* *96
*96. CEabCEbdlcd
R.S.alEab, plEbc, ylCEbdlcdx *96, *Bg -- *97
*97. CEabCEbcCEbdlcd
R.S.alEab,BlEad, ylCEbcCEbdlcdx*g7,*go -+ *98
* gB. CEabCEadCEbcC
Ebdlcd
RS. alEab, plEac, ylCEadCEbcCEbdicdx*g8,*94 *
*99
* gg. CEabCEacCEadCEbcCEbdlcd
The rule RS is usedin this deduction ten times; a and p are
always simple negative expressions,and y is everywhere an
elementary expression.In the same manner we could disprove
other formulae of the form (F4), and also the formula (Fr) of
sectionzB. It is needless,however,to perform thesedeductions,
sincewe can now setforth the generalproblem of decision,
$ gr. Deductiueequiaalence
We needfor our proof of decisionthe conceptof deductiveor
inferential equivalence.Since there are, in my opinion, some

D E D U C TM

$3t

EQU M L EN C E

to 7

misunderstandingsin the treatment of this concept,its meaning

must be carefully defined. I shall do this on the basis of the
theory ofdeduction.
c ?nd p, are deductively
It is usually said that two expressions,
equivalent to each other when it is possibleto deduce B from c
if a is asserted,and converselyc from p if p is asserted.The rules
of inferenceare always supposedas given. But they are seldom
sufficient.They suffice,for instance,in the following example.
we can
From the assertedlaw of commutation CCpCqrCqCpr
deduce the thesisCqCCpCqrCpr:
(r) CCpCqrCqCpr
Q) plCpCqr,rlCprxC(r)-(z)
(z) CqCCpCqrCpr,
and again from this thesis we can deduce the law of commutation:
pls, rlt x C(z)-(3)
(z) qlCqCCpCqrCpr,
CCsCCqCCpCqrCprtCst
e)
(z) qlCpCqr,plq, rlCprx(4)
(Q CCpCqrCCqCCpCqrCprCqCpr
(g) slCpCqr,tlCqCprx C(a)-(r)
(r) CCpCqrCqCpr.'
But we cannot in this simple way deduce from the asserted
cxpressionCNpCpqthe law of Duns Scotus CpCNpq,because
from the first expressionwe can derive new propositionsonly
by substitution, and all the substitutionsof CNpCpqbegin with
Cil, none with Cp. To deduce one of those expressionsfrom
another we must have further assistance.Speaking generally,
the relation of deductive equivalenceis seldom absolute,but in
most casesit is relative to a certain basisof theses.In our case
this basisis the law of commutation. Starting from

$) cNpcPq
we get by commutation the law of Duns Scotus:

(r) plNp,qlp,rlcx C(s)-(6)

(6) cpcNpq,
and starting from (6) we get again by commutation (S):

0) qIl,{p,rI q x C(6)-(5)

6) cNpcpq.

I This neat deduction was given by A. Tarski in Warsaw.

ro 8

THE

PRO BLEM

O F DECI S I O N

$l'

I saythereforethat CNpCpqand CpCNpqare deductivelyequivalent with respectto the law of commutation, and I write:
CNpCpq CpCNpq
with respecrto (r).
The sign
denotes the relation of deductive equivalence.
This relation is different from the ordinary relation of equivalence,denoted here by Q, which is defined by the conjunction
of two implications each converseto the other,

Q-Pq: KCpqcqp,
and requiresno basis.If an ordinary equivalenceQcB is asserted,
and c, or a substitution of c, is assertedtoo, then we can assert
B, or the correspondingsubstitution of p, and conversely.An
assertedordinary equivalenceQ"F is thereforea sufficientbasis
for the deductive equivalencea
but it is not a necessary
- Fi
one. This is just the point where explanation is needed.
Not only assertedor true expressionsmay be deductively
equivalent, but alsofalseones.In order to solvethe problem of
decision for the C-"lf-systemwe have to transform an arbitrary
significant expressionc into the expressionCNuzr,where z is a
propositional variable not occurring in c. This can be done by
meansof two theses:
Sr. CpCNpq
Sz. CCNppp.
I say that a is deductively equivalent to CNan with respectto
Sr and 53, and I write :
I. a
CNatr
with respectto Sr and Se.
All goes easily when o is asserted.Take as example NNCpp.
This is a thesiseasilyverified by the o-r method. I stateaccording to formula I that
with respectto Sr and Se.
NNCpp - CNNNC4P4
Starting from

0) NNCPP
we get by S I :

sr. plNNCpp
x C(7)-(B)
(B)CNNNCppq,
and starting again from (B) we get by substitution and Sz:

(B)qlNNCppx(s)
cNNNCppNNCpp
G)

D E D U C TIVE

$ er

EQU M L EN C E

to g

Sz.plNJ,tCpp
x C(g)-(Z)
0) NNCPP.
But c is an arbitrary expression;it may be false, e.g. Cpq. ln
this caseformula I reads:
Cpq CNCpqr
with respectto Sr and Se.
Here the difficulty begins: we can get the thesisCCpqCNCpqr
from Sr by the substitutionplC|q, qfr, but we cannot derive
from this thesis the consequent CNCltqr, for Cpq is not a thesis
arrd cannot be asserted.Therefore CNCpqrcannot be detached.
A still greater difficulty arisesin the other direction: we can
get from Sz by the substitutionFlclq the thesisCCNCpqCpqCpq,
but CNCpqCpqis not asserted,nor can we get CNCpqCpqfrom
CNCpqr by substitution, becauseCNCpqr is not a thesis. We
cannot say: Supposethat Cpqbe asserted;then CNCpqrwould
follow. The assertionof a false expressionis an error, and we
cannot expectto prove anything by an error. It seemstherefore
that formula I is valid not for all expressionsbut only for those
that are asserted.
There exists, in my opinion, only one way to avoid these
difficulties: it is the introduction of rejection into the theory of
deduction. We reject axiomatically the variable p, and accept
the clear rules of rejection, (c) and (d). It can easilybe shown
on.this basisthat Cpq must be rejected. For we get from the
axlom

(*to) p
and the thesis

Qt) CCCpppp
by the rulesofrejection:
(rr) x C( * r z ) - ( * r o )
(* tz) CCppp

qlp
(*rz)x (*ril PlCfu,
(*til cPq.
Now we are able to prove that if Cpq is rejected, Clr[Cpqrmust
be rejectedtoo; and conversely,if CNCpqris rejected,Cpq must
be rejectedtoo. Starting from

(ril c4q

THE PRO BLEM

rro

O F DECI SI O N

$gt

we get by Sz and the rules of rejection:

Sz.plCpq
x (r+)
ccNcpqctqc?q
QQ
( t+)xC(*r5)-(*r3)
(*15) CNCpqCpq
( * r S)x (*16)rlCpq
(*rG) CNCpqr.
In the otherdirectionwe easilyget Cpqfrom (*16) by Sr:
Sr.plCpq,qlrx (r7)
Q7) CCpqCNCpqr
( ,2)xC(*r3)-(*r6)

(*ry)c?q.

Formula I is now fully justified. We have, however, to correct

our previous definition of deductive equivalence,saying:
Two expressionsare deductively equivalent to each other
with respectto certain theseswhen and only when we can
prove by meansof thesethesesand of the rules of inference
that if one of those expressionsis asserted,the other must
be assertedtoo, or if one of them is rejected,the other must
be rejected too.
It followsfrom this definition that ordinary equivalenceis not
a necessarybasisof deductive equivalence.If Qo,Bis a thesis,it
is true that a is deductively equivalentto p with respectto Qap;
but if c is deductively equivalent to p with respectto certain
theses,it is not always true that Qcp is a thesis.Take as example
the deductive equivalencejust considered:
with respectto Sr and Sz.
Cpq CNCpqr
The corresponding ordinary equivalence QCpqCNCpqris not
a thesis,for it is false for plr, qlo, rlr.
It is obvious that the relation of deductive equivalenceis
reflexive, symmetrical, and transitive. There are caseswhere
cr is deductively equivalent to two expressionsp and 7 with
respectto certain theses.That means: if cris asserted,then B is
asserted and y is asserted,and consequently their conjunction
'B and y' is asserted; and conversely, if both p and 7, or their
conjunction 'p and y', is asserted,'thena is assertedtoo. Again,
if c is rejected,then the conjunction 'p and y' must be rejected,

$tr

D E D U C TM

EQU M L EN C E

rrr

and in this caseit is sufficient that only one of them, F o, y,

should be rejected; and conversely, if only one of them is
rejected,a must be rejected too.
expessions
$ gz. Reductionto elementar2
Our proof of decision is basedon the following theorem:
(TA) E-very significant expression of the Aristotelian syllogistic can be reduced in a deductively equivalent way,
with respect to thesesof the theory of deduction, io
a set of elementary expressions,i.e. expressionsof the
form
CarCa2Cus,..Cun-1an,
where all the c's are simple expressionsof the syllogistic, i.e. expressions
of the type Aab, Iab, Eab, or Oab,
AII known thesesof the syllogistic either are elementary expressionsor can easily be transformed into elementary expressions. The laws of conversion, e.g. CIabIba or CAabIba, arc
elementary expressions. All the syllogisms are of the form

which is not a simple expression but an implication. There

exists,of course, an infinity of suchexpressions,and they must all
be taken into account in the proofofdecision.
Theorem (TA) can easily be proved on the basis of an
analogoustheorem for the theory of deduction:
(TB) Every significant expressionof the theory of deduction
with C and "/Vas primitive terms can be reduced in
a deductively equivalent way with respect to a finite
number of thesesto a set of elementary expressionsof
the form
c,.1ca2c.*s...co-o-1,-n,
where all the c's are simple expressions,i.e. either
variables or their negations.
The proof of this theorem is not easy,but sinceit is essential

tr2

THE PRO BLEM

O F DECI S I O N

0gz

for the problem of decision it cannot be omitted. The proof of

(TB) given below is intended for readers interested in formal
logic; those not trained in mathematical logic may take both
theorems,(TA) and (TB), for granted.
Let c be .an arbitrary significant expressionof the theory of
deduction other than a variable (which may, but need not, be
transformed): every such expressioncan be transformed,as we
already know, in a deductively equivalent way with respectto
the thesesSr and Sz;
Sr. CpCNpq

Sz.CCNppp
into the expressionCNan, where z is a variable not occurring
in a. We have thereforeas transformation I:
with respectto SI and Sz.
I. cr CNan
Transformation I allows us to reduce all significant exPressions to implications that have a variable as their last term.
Now we must try to transform "lfc, the antecedentof CNa4
into a variable or its negation. For this purposewe employ the
following three transformations:
with respect to 53 and 54,
lI. CNN,B - CoF
55 and 56,
lII. CNCupy CaCNpy
,,
,,
57, SB, and 59.
.,
lY. CCaPy - CNaY, CPY
"
The respectivethesesare: for transformationII:

fu. CCNNpqCpq
54.CCpqCNNpq;
for transformation III :
55. CCNCpqrCpCNqr
56. CCpCNqrCNCpqr;
for transformation IV:
57. CCCpqrCNpr
58. CCCpqrCqr
Sg. CCNprCCqrCCPqr.
Let us now explain how we can get by thesetransformations
a variable or its negation in the antecedent of CJfaz. The
expression ct (rccurring in CNar may, like every significant
expressionof the C-.1[-system,be either a variable, or a nega-

$gz REDUCTIONTO ELEMENTARYEXPRESSIONS rrg

tion, or an implication. If cris a variable, no transformationis
needed; if it is a negation, we get CNNap, and two negations
annul each other according to transformation II; if it is an
implication, we get from CNCapy the equivalent expression
C"CNFy, the antecedentof which, c, is simpler than the initial
antecedent NCuB. This new c: may again be a variable-no
transformation is then needed-or a negation-this case has
already been settled-or an implication. In this last casewe get
from CCapy two expressions,CJfaT and CBy,with simpler antecedents than the initial antecedent Cap. By repeated applications of II, III, and IV we must finally reach in the antecedent
a variable or its negation.
Let us now seeby exampleshow thesetransformationswork.
First example:.NNCpp.
by I;
NNCPP
- CNNNCPPq
CNNNCPPq CNCPPq
,, II;
CNCppq
CpCNpq
,, III.
NNCpp is thus reduced to the expression CpCNpq with the
variable p in the antecedent. CpCNpq is an elementary
cxpression.
Second example: CCCpqpp.
by I;
CCCPqPP - CNCCCPqPPT
CNCCCpqppT CCCpqpCNpr ,, III;
CCCpqpCNpr- CNCpqCNpr, CpCNpr by IV;
CpCNqCNpr by III.
CNCpqCNpr
CCCpqppis thus reduced to two expressions:CpCNqCNpr and
CpCNp4 both with the variable 1 in the antecedent;both are
clementary expressions.
Third example: CCCpqqCCqpp.
I;
CCCpqqCCqpp
CNCCCPqqCCqPPTby
III;
CNCCCpqqCCqppr CCCpqqCNCCqppr
,,
CCCp qqCNC Cqp.pr CNC! QCNC Cqppr, CqCNC Cqppr by IV;
CNCpqCNCCqppr CpCNqCNCCqpprby III.
CCCpqqCCqpp
is reduced to two expressionsCPCNqCNCCqppT
md CqCNCCqppr,
both with a variable in the first antecedent.
Ncither of them, however,is elementary,sincethe first has the
t'ompound expressionNCCqppas its third antecedentand the
I

rr 4

THE PRO BLEM

O F DECI SI O N

$gz

$gz

REDUCTION

TO EL EM EN TAR Y

EXPR ESSION S

t1 5

second has tlre same compound expressionas its second antecedent.

As we can see from this last example, our task is not yet
finished. By transformations I-IV we obtain implications with
a variable in the first antecedent, and also expressionsof the
form:
Co,lCurCus...Can-1un,

in fact, NCpNq means the same as Kpq, and Srz and Sr3 are
other forms of the laws of importation and exportation respectively. Now CJfCaNpy,like CKapy, has only one antecedent,
whereasthe equivalent expressionCuCByhas two antecedents.
I{, therefore, a compound expressionoccurs in the fourth place,
as E.in CuCpCyCDe,
we can move it to the third place, applying
VII and then VI:

but not all antecedentsof this form, apart from ar, need be
simple expressions.fn order to get rid of such compound antecedentswe need three further transformations:
CPCIY
with respect to Sro,
Y. CaCPY
Yl. CaCpCyS, CICyCPD
SI I,
,,
,,
Slz and Sr3.
YII. CaCBy
CNC"NFy
,,
,,
The respectivethesesare: for transformation V:
Src. CCpCqrCqCpr;

CuCpCyCDe - CNC"l,tpCyC6eby VII;

CNCuNpCyCbe CNCINBC\CyI ,, YI.
From this last expressionwe get by the converseapplication of
VII the formula:

for transformationVI:
Str. CCpCqCrsCpCrCqs;
for transformation VII :
Srz. CCpCqrCNCpNqr
Sry. CCNCpN.grCpCqr.
By Sro we can move a compound antecedentfrom the second
place to the first, and by Sr l from the third place to the
second. Applying these transformations to the expressions
CpCNqCNCCqpprand CqCNCCqpprof our third example we get:
(a) CpCNqCNCCqppr CICNCCqp?CNqiby YI;
CpCNCCqppCNqr CNCCqppCpCNqr ,, Y;
CNCCqppCpCNqr CCqpCNpCpCNqr ,, lll;
CCqpCNpCpCNqr CNqCNpCpCNqr, CpCNpCpCNqr by
IV.
(p) CqCNCCqppr CNCCqllCqr by Y;
CNCCqppCqr CCq?CNPCqr ,, lll;
CCqpCNpCqr CNqCNpCqr,CpCNpCqrby lY.
is thus reduced to four elementary expressions:
CCCpqqGCqpp
CN qCNpCpCN qr, CpCNpCpCN qr, CN qCNpCqr, and CpCNpCqr.
Transformation VII is usedin all thosecaseswhere the compound antecedentoccurs in the fourth place or farther. This
transformationallows us to reduce the number of antecedents;

CuCpC\Cyeby VII.
It is now easy to bring E to the first place by VI and V:
CNCdNpC\Cye

by VI,
CaCBCtCye CuCECpCye
CaCDCpCye C\CuCpCy ,, V.
Applying transformation VII repeatedly in both directions we
can move any antecedentfrom the zth place to the first, and
transform it, if it is compound, by II, III, and IV into a simple
expression.
The proof of theorem (TB) is thus completed. It is now easy
to show that this theorem entails the proof of decisionfor the
C-,Iy'-systemof the theory of deduction. If all the elementary
expressionsto which a given explessioncrhas been reduced are
true, i.e. if they have among their antecedentstwo expressions
of the typel and Np, then a is a thesisand must be asserted.On
the other hand, if among the elementaryexpressions
to which cr
has been reduced there exists at least one expressionsuch that
no two antecedentsin it are of the type p and Np, then c must
be rejected. In the first case we can prove a by means of the
thesesSr-S13, in the secondwe can disproveit, adding to the
lbovc thcsestwo new ones:
5r4.. CpCCpqq
St5. NNCpp,
and the axiom of rejection:
rSr6. p.
Two exampleswill clarify this.

tt6

THE PRO BLEM

O F DECI SI ON

0gz

First example: Proof of the thesisCpCCpqq.

This thesismust first be reduced to elementaryexpressions.
This is done by the following analysis(L):
CpCCpqq
CNCpCCpqqr by I;
CNCpCCpqqr CpCNCClqqr ,, III;
CpCNCCpqqr CNCCpqqCpr ,, V;
CNCCpqqCpr CCpqCNqCpr ,, III;
CCpqCNqCpr CNpCNqCpr,CqCNqCprby IV.
The elementary expressionsto which CpCCpqqis reduced are
CNpCNqCprand CqCNqCpr.Both,like all expressionsto which
transformation I has been applied, have as their last term a
variable not occurring in the antecedents.Such expressions
can
be true only on condition that they have two antecedentsof the
type p and Np, and any expressionof this kind can be reduced
by transformationsV, VI, or VII to a substitution of Sr from
which the proof of a thesis must always begin. Here are the
required deductions:
*. qlCNqrxQ)
(r) CpCNpCNqr
Sro. qlNp, rlCNqrx C(r)-(z)
(z) CNpCpCNqr
Srr. plNp, qlp, rlNq, slrxC(z)-(g)

(s) CNgCNqCpr
St. !lq, qlCprx(a)
@) CqCNqCpr.
Havinggot in (3) and (4) the sameelementary
expressions
as
we reachedat the end of our analysis(L), we now proceedfrom
them to their equivalents on the left, by applying theseson
which the successive
transformationswere based.Thus, step by
step,we get our original thesisby meansof Sg, 56, Sro, and Sa:

Ss.rICNqCpr
x C(3)-C(a)-(5)
6) CCpqCNqCpr
56.pICpq,rICprxc(s)-(0)

(6) CNCCpqqCpr
Sro.plNCCpqq,
qlpx C(6)-(7)
CpCNCCpqqr
0)

$gz

REDUCTION

TO EL EM EN TAR Y

EXPR ESSION S

tt7

s6. qlCCpqq
x C(7)-(B)
(B) CNCpCCpqqr

(B)rlcpccpqqx(s)
$) CNCpCCpqqCpCCpqq
Sz. p I CpCCpqq
x C(s)-(r o)
(ro) CpCCpqq.
Upon this model we can prove any thesiswe want.
Secondexample: Disproof of the expressionCCNlqq.
We first reduce this expressionto elementaryexpressions
on
the basisof the following analysis:
CCNlqq
CNCCNPqqrbyI;
CNCCNPqqT CCNpqCNqr ,, III1'
CCNpqCNqr CNNpCNqr, CqCNqr by IV;
CNNPCNqT CpCNqr
by II.
The expressionCCNpqqis thus reduced to two elementaryexpressions,CqCNqr and CpCNqr.The first of theseis a thesis,but
the secondis not true, for it has no two antecedentsof the type
p and Np. The expressionCCNpqqrherefore, which leads to this
not-true consequence,must be rejected.We begin the disproof
from the top, successively
applying accordingto the given transformationsthe thesesSl, 55, 57, and 53:
Sr. plCCNpII, Ilr x (r r)
(tr) CCCNpqqCNCCNpqqr

55.plcNpqx (rz)
(r z) CCNCCNpqqrCCNpqCNqr
Sl.plNp, rlCNqrx (rS)
( t c) CCCNqqCNqTCNNpCNqT
Sg. qlCNqrx (r+)
(r a) C;CNNpCNqrCpCNqr.
Now wt: must disprovethe expressionCpCNqr;we need for this
l)lrlposcthc ncw thesesSr4 and Sl5 and the axiom of rejection.
S r a. 1tI NNCpp, qlp xCS r 5-( r5)
(rs) (;t:NN(:1t1t1tp
(rr,);<(,'(*r6)-*5t6
(* r{t) t:NNCppp

t r8

THE PRO BLEM

O F DECI SI ON

$gz

qICNNCppp
x CSr-( r 7)
Sra.pICpCNpq,
(r 7) cccpcNpqcNNcpppcMMcppp
( t Z )x C (*rB )-(*r6 )
(*fi) ccpcNpqcNMcppp
(+rB)x (r'ril plcpcNpq,qlNCpp,
rlp
(*ry) CpCNqr
cannow successively
rejectits anteHaving rejectedCpCNqrrwe
cedentstill we reachthe original expressionCCNpqq.
( t+ )x C(*zo)-(*r9)
(*zo) CNNpCNqT
( r g )x C(*zr)-(*zo)
(*zr) CCNpqCNqr
( r z) xC(*zz)-(*zr)
(*zz) CNCCNpqqr
( r r ) x C(*z$-(*zz)
(*n) CCNpqq
In this way you can disproveany not-true expressionof the
All thesedeductionscouldhavebeenmadeshorter,
C-"lf-system.
but I wasanxiousto show the methodimplied in the proof of
decision.This method enablesus to decideeffectively,on the
basisof only fifteenfundamentaltheses,
Sl-S15,and the axiom
of rejection,whethera givensignificantexpression
of the C-"ltfor rejected.As all the otherfunctors
systemshouldbe asserted
of the tlieory of deductionmay be definedby C and "l{, all
of the theoryof deductionare decidable
significantexpressions
on an axiomatic basis. A system of axioms from which the
fifteen fundamental thesescan be drawn is complete in this
ofthe systemcan be deducedin
sense,that all true expressions
it. Of this kind is the systemof three axioms set out in section
23, and also the systemof those three axioms on which transformation IV is based,,'viz. CCCpqrCNpr, CCCpqrCqr, and,
CCNprCCqrCCpqr.
The proof of theorem (TA), accordirig to which every significant expression of the Aristotelian logic can be reduced to
elementary expressions,is implicitly contained in the proof of
the analogous theorem for the theory of deduction. If we take
instead.of the Greek letters used in our transformations I-VII
(except the final variable in transformation I) propositional

$gr

REDUCTION

TO EL EM EN TAR Y

EXPR ESSION S

r t9

expressionsof the Aristotelian logic, we can apply those transformations to them in the same way as to expressionsof the
theory of deduction. This can easily be seenin the example of
CCNAabAbaIab.Weget:
CCNAabAbaIa$ nCNCCNAabAbaIabpbyI;
CNCCNAabAbaIabp CCNAabAbaCNIabp,, III;
CAbaCNIabPby IV;
CCNAabAbaCNIabp- CI,TNAabCNIahP,
by II.
CAabcNIabF
cNNAabCNIabp
and,Eab instead of
Oab,
write
Instead of NAab we can always
convenient to
be
more
it
will
NIab.In whatfollows, however,
employ forms with "lf.
Both elementary expressions,CAabCNIabpand CAbaCNIabp'
to which CCNAabAbaIabhas been reduced, have a propositional variable as their last term. This variable is introduced by
transformation I. We can get rid of it by the following deductively equivalent transformations where z is a propositional
variable not occurring in either a or B:
with respectto Sr7 and Sr8,
YIII..CuCfur - CqNp
lX. CuCNBn C"P
,, ,, Srg and Szo.

Thesesfor transformationVIII:
S'7. CCpCqNqCpNq
SrB. CCpNqCPCqr.
Thesesfor transformationIX:
Srs. CCpCNqqCpq
Szo. CCpqCpCNqr.

',

r2 o

THE PRO BLEM

O F DECI S I O N

$sz

transformation VIII or IX. Take, for instance, the following

example:
CNIabCAcbCAdcCladp
CNCNIabNAcbCAdcCladp
by VII;
CNCNIabNAcbCAtuCIadp
CNCNCNIabNAcbNAdcCIadp
by VII;
CNCNCN labNAcbNAdcCladp
CNCNCN labNAcbNAdcNI ad
by VIII;
CNCNCNIabNAcbNAdcNIad
- CNCNIabNAcbCAdcNIad
by VII;
CNCNIabNAcbCAdcNIad
-CNIabCAcbCAdcNIad,,VII.
Theorem (TA) is now fully proved; we can proceed therefore
to our main subject, the proof of decision of the Aristotelian
syllogistic.
of the syllogistic
$ gg. Elementaryexpressions
According to theorem (TA), every significant expressionof
the Aristotelian syllogistic can be reduced in a deductively equivalent way to a set of elementary expressions,i.e. expressions
of the form
Cu1Cc'2Co,s...Can-ru*
where all the c's are simple expressionsof the syllogistic,i.e.
expressions of the type Aab, Iab, Eab or NIab, and Oab or

proofs of rejection of the other expressions:

*rooX *6r. clb
*too. Iab
rr
B X C* ror-* roo
*rot. Aab
IY. pl Aaa, qI lab x Ct-roz
toz. CNAaaIab
r02 XC*ro3-tloo
*ro3. NAaa

(8. CAabIab)

(Iv. CpcNpq)
(:

oaa)

933 ELEMENTARY

EXPR ESSION S OF TH E SYL L OGISTIC

Izt

tro3 x *to4. bla

*rc4. NAab
(: Oab)
lY. pllan, qllabxCz-rc5
rc5. CNIaaIab
ro5 x C*106-*too
*rc6. NIaa
(: Eaa)
*ro6 x *ro7. bla
*rc7. NIab
( : Eab) .
Turning now tocompound elementary exPressionsI shall successivelyinvestigate all the possible cases,omitting the formal
proofs where it is possible,and giving only hints how they could
be done. Six caseshave to be investigated.
First case:The consequent cr' is negative, and all the antecedentsare affirmative. Such expressionsare rejected.
Proof: By identifying all the variables occurring in the expressionwith a, all the antecedentsbecometrue, being laws of
identity Aaa or laa, and the consequent becomesfalse. We see
that for the solution ofthis casethe laws of identity are essential.
Sccondcase:The consequent is negative, and only one of the
antecedentsis negative. This case may be reduced to the case
with only affirmative elements, and such cases,as we shall see
later, are always decidable.
Proof: Expressionsof the form CuCNpNy are deductively
equivalent to expressionsof the form CaCyBwith respect to the
thesesCCpCNrNqCpCqrand CCpAqrCPCNrNq.This is true not
only for one affirmative antecedbntc, but for any number of
them.
Third case:The consequent is negative, and more than one
antecedent is negative. Expressionsof this kind can be reduced
to simpler expressions,and eventually to the second case. The
solution of this caserequiresSlupecki'srule of rejection.
Proof: Let us supposethat the original expressionis of the
form CNaCNBCy...Np. This supposition can always be made,
as any antecedentmay be moved to any place whatever. We
reduce this expressionto two simpler expressionsCNaCy..,Np
ancl L'NBCy...,lfp,omitting the second or the first antecedent
respectivcly. If these expressionshave more negative antecedents than one we repeat the same procedure till we get
formulac with only one negative antecedent.As such formulae

t22

THE PRO BLEM

O F DE C I S I O N

!gg

according to the second case are deductively equivalent to

decidable affirmative expressions,they are always either asserted
or rejected. If only one of them is asserted,the original expression must be assertedtoo, for by the law ofsimplification we can
add to this assertedformula all the other negative antecedents
which were previously omitted. I{, however, all the formulae
with one negative antecedentare rejected, we gather from them
by repeated application of Slupecki's rule of rejection that the
original expression must be rejected. Two examples will explain the matter thoroughly.
First example : CNAabCNAbcCNIbdCIbcNAcd,
a thesis.
We reduce this expressionto (r) and (z):
(r) CJtAabCNIbdCIbcNAcd, (z) CNAbcCNIbdCIbcNAcd.
In the sameway we reduce(r) to (3) and (4):
(g) CNAabCIbcNAcd,
@) CNIbdCIbcNAcd,
and (z) to (5) and (6):
$) CNAbcCIbcNAcd,

(6) CNIbdCIbcNAcd.

Now the last expressionis a thesis; it is the mood Ferison of the

third figure. Putting in CpCqp(6) forp, and,NAbc for g, we get
(z), and applying CpCqponce more by putting (z) for p, and
NAab for 4, we reach the original thesis.
Second example: CNAabCNAbcCNIcdCIbdNAad,
not a thesis.
We reduce,this expressionas in the foregoing example:
Q) CNAabCNIcdCIbdNAad, (z) CNAbcCNIcdCIbdNAad;
then wereduce (r) to (3) and (4), and (z) to (5) and (6):
(g) CNAabCIbdNAad,
Q) CNIcdCIbdNAad,
(6) CNhdCIbdNAad.
$) CNAbcCIbdNAad,
None of the above formulae with one negative antecedent is
a thesis, as can be proved by reducing them to the case with
orily affirmative elements. Expressions(3), (+), (S), and (6)
are rejected. Applying the rule of Slupecki, we gather from the
rejected expressions(5) and (6) that (z) must be rejected,and
from the rejected expressions(3) and (4) that (r) must be
rejected. But if (r) and (z) are rejected, then the original
expressionmust be rejected too:
Fourthcase:The consequentis affirmative, and some (or all)

$33 ELEMENTARY

EXPR ESSION S OF TH E SYL L OGISTIC

tz3

antecedents are negative. This case can be reduced to the

third.
Proof: Expressions of the form CaCNBy are deductively equivalent to expressions of the form CuCNpCNyNAaa
on the ground of the theses CCpCNqrCpCNqCNrNAaaand,
CCpCNqCNrNAaaCpCNqr,as NAaa is always false.
All the caseswith negative elementsare thus exhausted.
Fifth case:All the antecedentsare affirmative, and the consequent is a univenal affirmative proposition. Several sub-cases
have to be distinguished.
(a) The consequentis Aaa; this expressionis asserted,for its
consequentis true.
(D) The consequent is Aab, and,Aab is also one of the antecedents.The expressionis of course asserted
In what follows it is supposed that Aab does not occur as
antecedent.
(c) The consequentis Aab, but no antecedentis of the type
Aaf with;f different from a (and from D, of course). Such
expressionsare rejected.
Proof: By identifying all variables different from a and 6 with
6, we can only get the following antecedents:
Aaa, Aba, Abb, Iaa, Iab, Iba, Ibb.
(We cannot get Aab, for no anteqedentis of the type Aaf, f
being different from a.) PremissesAaa, Abb, Iaa, Ibb can be
omitted as true. (If there are no other premisses,the expression
is rejected, as in the first case.) If there is lDa besideslab, oneof
them may be omitted, as they are equivalent to each other. If
there is Aba, both lab and Iba may be omitted, as Aba implies
them both. After thesereductions only Aba or lab can remain as
antecedents.Now it can be shown that both implications,
CAbaAab

and

ClabAab,

are rejectedon the ground of our axiom of rejection:

X. plAcb,qlAba,rllac,slAabxCzT-roB
to9. CCAabAbaCKAcbAabIac(X. CCKpqrCCsqCKpsr;
roBx C* rog-*59
*tog. CAabAba
*ro9x *rrc. bla, alb
*rto. CAbaAab.

27. CKAcbAbalac)

124

THE PRO BLEM

O F DECI SI O N

$ss

If CAbaAabis rejected, then CIabAaDmust be rejected too, for

Iab is a weaker premissthan Aba.
(d) The consequentis Aab, and there are antecedentsof the
type Aaf with;f different from a. If there is a chain leading from
ato b, the expressionis assertedon the ground of axiom 3, the
mood Barbara; if there is no such chain, the expressionis
rejected.
Proof: By a chain leading from a to 6 I understandan ordered
seriesof universal affirmative premisses:
Aacr,AcLcz,..., Acn-rcn,Acnb,
where the first term of the serieshas a as its first argument, the
last term D as its secondargument, and the secondargument of
every other term is identical with the first argument of its successor.It is evident that from a seriesof such expressionsAab
resultsby fepeatedapplication of the mood Barbara. If, therefore, there is a chain leading from a to D, the expressionis
asserted;if there is no such chain, we can get rid of antecedents
of the type Aaf, identifying their secondargumqnt with a..The
expressionis reduced in this way to the sub-case(c), which was
rejected.
Sixthcase:All the antecedentsare affirmative, and the consequent is a particulai affirmative proposition.Here also we have
to distinguishseveralsub-cases.
(a) The consequentis laa; the expressionis asserted,for its
consequentis frue.
(6) The consequentis lab, and as antecedentoccurs either
Aab, or Aba, or lab, or lba; it is obvious that in all thesecases
the expressionmust be asserted.
In what followsit is supposedthat none of the above four premissesoccursas antecedent.
(c) The consequentis lab, and no antecedentis of the type
A"fa,f different from a, or of the type Agb,g different from D.The
expressionis rejected.
Proof: We identify all variables different ftom a and D with
c; then we get, besidestrue premissesof the type Accor lcc, only
the following antecedents
:
Aac, Abc,Iac, Ibc.
Aac implies lac, and lDc implies lbc. The strongestcombination

$33E LE ME N TA R Y D XPRESSI O NSO F THE SYLLO G I STI C I z5

of premisses is therefore Aac and Abc. From this combination,
however, Iab does not result, as the formula
CAacCAbcIab
is equivalent to our axiom of rejection.
(d) The consequent is lab, and among the antecedents there
are expressions of the type Afa (f ditrerent from a), but not of
the type Aeb (S different from 6). If there is Abe or lbe (Ieb), and.
a chain leading from e to c:
(a) Abe; Aeer,Aerer,..., Aena,
(p) Ibe; Aeer, Aerer, ..., Aena,
we get from (c) Abe and Aea, and therefore lab by the mood
Bramantip, and from (p) Ibe and Aea, and therefore lab by the
mood Dimaris. In both casesthe expression is asserted. If, however, the conditions (c) and (p) are not fulfilled, we can get rid
of antecedents of the type Afa by identifying their first arguments with a, and the expression must be rejected according to
sub-case (c).
(e) The consequent is lab, and among the antecedentsthere are
expressionsofthe type lgb (g different from D), but not of the type
Afa (f different from a) . This case can be reduced to sub-case (d),
as a and D are symmetrical with respect to the consequent lab.
(/) The consequent is lab, and among the antecedents there
are expressions of the type AJa (f ditrerent from a), and expressions of the type Agb (g different from, D). We may suppose that
the conditions (cr) and (p) are not fulfilled for A,fa, or the analogous conditions for lgD either; otherwise, as we already know,
the original expression would be asserted. Now, if there is Aca
and a chain Leading from c to b:
(y) Aca; Accr, Acrcr, ..., Acnb,
or Adb and a chain leading from d to a:
(6) Adb; Ad\, AdJz, ..., Adna,
we get from (7) Aca and Acb, from (6) Adb and Ada, and therefore in both caseslab by the mood Darapti. Further, if there is
an antecedent lcd (or ldc) and two chains, one leading from c
to a, and another from d to D:
Accr, Acrcr, ..., Acr,a,
,.,
\-) llcd; Addb Adiz, ..., Adnb,
\Icd;

t26

THD PRO BLEM

O F DECI SI O N

0Sg

we get by the first chain the premiss Aca, by the second chain
the premissAdb, and both premissesyield together with .Icdthe
conclusion Iab on the basis of the polysyllogism:
CIcdCAcaCAdblab.
We prove the polysyllogism by deducing ladfrom Icd and Acaby
the mood Disamis, and then lab from lad and Adb by the mood
Darii. In all thesecasesthe original expressionmust be asserted.
I{, however,none of the conditions (7), (E),or (e) is satisfied,we
can get rid of expressionsof the type Afa and'Agb by identifying
their first arguments with a or with 6 respectively,and the
original expressionmust be rejected according to sub-case(c).
.All possiblecasesare now exhausted,and it is proved that every
significant expression of the Aristotelian syllogistic is either
assertedor rejected on the basis of our axioms and rules of
inference.
of thes2llogistic
E g+. An arithmcticalinterpretation
In r679 Leibniz discoveredan arithmetical interpretation of
the Aristotelian syllogistic which deservesour attention from the
historical as well as from the systematicpoint of view.r It is an
isomorphic interpretation. Leibniz did not know that the Aristotelian syllogisticcould be axiomatized,and he knew nothing
about rejection and its rules. He only testedsomelaws of conversion and somesyllogisticmoods in order to be sure that his
interpref;ation was not wrong. It seems,therefore, to be a mere
coincidencethat his interpretation satisfiesour assertedaxioms
r-4, the axiom of rejection *59, and the rule of Slupecki.In any
caseit is strange that his philosophic intuitions, which guided
him in his research,yielded such a sound result.
' Leibniz's arithmetical interpretation is basedon a correlation
of variables of the syllogistic with ordered pairs .of natural
ntrmbers prime to each other. To the variable a, for instance,
correspondtwo numbeni, say a, and ar, prime to each other; to
the variable D correspond two other numbers, say b, and br,
alsoprime to eachother. The premissAab is true when and only
when a, is divisible by Dr,and a, is divisibleby br.Ifone of these
conditions is not satisfied, Aab is false, and therefore NAab is
t SeeL.Couturat, Opuscuhsetfragnmtsinld;tsdeltibniz,Patis (tgog),pp.77s.q.
Cf, also J, Lukasiewicz, 'O sylogistyce Arystotelesa' (On Aristotle's Syllogistic),
do Cracoab,xliv, No' 6 (t939)' p. ceo.
Comltcs Rendusdo l'Acad. dcs Scitttr,cs

A N A R I T H M ETIC AL

$g+

IN TER PR ETATION

r27

true. The premiss lab is ftue when and only when a, is prime
to 6r, and a, is prime to 6r. If one of these conditions is not
satisfied, Iab is false, and therefore NIab is true.
It can easilybe seenthat our assertedaxioms I-4 are verified.
Axiom r, Aaa, is verified, for every number is divisible by itself,
Axiom 2, Iaa, is verified, for it is supposedthat the two numbers
corresponding to a, a, and a2, are prime to each other. Axiom 3,
the mood Barbara CKAbcAabAac,is also verified, since the relation of divisibility is transitive. Axiom 4, the mood Datisi
CKAbcIbaIac,is verified too; for if Dt is divisible by cr, b, is
divisible by cr, b, is prime to ar, and 6, is prime to dr, then at
must be prime to cr, and ds must be prime to cr. For if ay and c,
had a common factor greater than r, a, and 6, would also have
the samecommon factor, since 6, contains cs. But this is against
the suppositionthat a, is prime to br.ln the sameway we prove
that a, must be prime to ar.
It is also easyto show that the axiom *59 CKAcbAabIacmust
be rejected.Take as examplesthe following numbers:
4t:

15, br:

az:

14,bz: 7, cz :

3 , c1 :

I2 e

35.

lcD is true, for c, is divisible by D, and c, is divisible by br; Aab

is also true, for a, is divisible by Dt and a2is divisible by Dr; but
the conclusion Iac is not true, for a1 and cz are not prime to
each other.
The verification of Slupecki'd rule of rejection is more complicated. I shall explain the matter with the help of an example.
Let us take as the rejected expressions,
(* r) CNAabCNIcdCIb dNAad and (* z) CNIbcCNh dCIbdNAad.
From them'we get, by the rule of Slupecki,
*CNuy, *CNFy --- *CNuCNpy,
a third rejected expression,
(* g) CNAabCN IbcCN IcdCI bdNAad.
Expression(r ) is disproved,for instanceby the following set of
numbers:
| ^\ IaL: 4, br: 7, cr: 3, dt: 4,
"' lar : 9, bz: 5, cz : 8, dz: 3.

TH E PRO BLEM

t 28

O F DECI SI O N

$g+

It can easilybe proved that accordingto this interpretation Aab

is false (since 4 is not divisible by 7), and therefore NAab is
true; Icd is false (sincer, is not prime to d1),and therefore NIcd
is true; Ibd is vue (for both pairs of numbers, D, and d2,brand
dr, are prime to each other) ; but NAad is 'false,becauseAad is
true (a, being divisible by dr, and arby dr). All the antecedents
are true, the consequentis false; therefore expression(r ) is
disproved.
The same set of numbers does not disprove expression(e),
because1Dcis true (as both pairs of numbers,D, and cr, and b,
and cr, are prime to each other), and therefore NIbc is false.
But if the antecedentof an implication is false,the implication
is true. In order to disproveexpression(z) we must take another
set of numbers, for instancethe following:
,.r, Ior: 9, br: 3, cr: B, dr:
: 2, bz: 2t cz- 5, dr:
'" lo,

3,
2.

According to this interpretation all the antecedentsof expression (z) are true, and the consequentis false; the expression
is thereforedisproved.But this secondset of numbers doesnot
disprove expression(r), becauseAab is true, and therefore
NAab is false,and a false antecedentyields a true implication.
Neither, therefore,of the sets (4) and (5) disprovesexpression
(3), which containsNAab as well as NIbc.
There, is a general method that enables us to disprove
expression(3) when expressions(r) and (z) are disproved.r
First, we write down all the prime numbers which make up
thesetsof numbersdisproving(r) and (z). We get for (r) the
series2, 3, 5, and 7, und for (zJ the seriesz, 3, und 5. Secondly,
we replace the numbers of the second seriesby new primes,
all different from the primes of the first series,for instance:
z by rI, 3 by 13, and 5 by 17.We get thus a new set of
numbers:
(6\ [ar t-'
to, :

13 . 1 3, Sr :
II,

bz :

13, c t :
tt, cz:

I L I'L rI, dL :
t 7,
dr:

13,
,r.

This set also disproves(z), sincethe relationsof divisibility and

primenessremain the sameasthey were beforethe replacement.
! This method was discovered by Slupecki, op. cit., pp. z8-3o.

$S+

A N A R I T H M ETIC AL

IN TER PR ETATION

r29

Thirdly, we multiply the numbers of correspondingvariables

occurring in the sets(4) and (6). We thus get a new set:
7.I3rcr: 3.Ir. II .rIrd!: 4.13,
frl [ar: 4. I3.13, \:
bz:5,rr)c2-8.17,
dr:!.tr.r
"' lor:9.rr,
This set disproves(3). For it is evident, first, that if to the premissAef or lef there correspondsthe set of numbers
0r, ezr.fr,.fz, e1prime to errfl prime toJ,
and there is another set of numbers
ti., ei,fi.,-fi,

e{ prime to ei,fi prime to;fj,

all of them composed of different primes from the numbers of

the first set, then the product of e1 and ei, i.e. er.ei, must be
prime to the product of e, and e'r,i.e. er.ei, andfr.r( prime to
fr.fi. Secondly, if Aef is verified by the first set, i.e. if e, is
divisible byfi, and erby.fz, and the sameis true of the second
then er.ei must be
set, so that ei is divisible by fi, and eiby
-fi,
and er.eLby .fr.fl. Again, if Ief is verified by
divisible by
"fr."fi,,
the first set, i.e. e, is prime to fr, and e, is prime tofi, and the
same is true of the second set, so that ci is prime tofi, and,ei
is prime tofi, then er.ei must be prime to
er.e'2prime
-fr..f;and
to A.-f;, since all the numbers of the secondset are prime to
the numbersof the first set. On'the contrary, if only one of the
conditions for divisibility or primeness is not satisfied, the respectivepremissesmust be false. It can be seenin our example
that Aad and lcd are verified by (l), for they are verified by (+)
and (6), and lbc is disprovedboth by (4) and (6), and therefore
also by (7). Aab is disproved only by (a) (but this sufficesto
disprove it by (7)), and lbc is disproved only by (6) (but this
also sufficesto disprove it by (7)). This procedure may be
applied to any caseof the kind, and thereforeSlupecki'srule is
verified by the Leibnizian interpretation.
Leibniz once said that scientific and philosophic controversiescould alwaysbe settledby a calculus.It seemsto me that
his famous 'calculemus' is connected with the above arithmetical interpretation of the syllogistic rather than with his
ideason mathematical logic.
' If there is a variable occurring in one of the disproved exprcssions but not in
the other, we simply take its corresponding numbers after eventual replacement'

tgo

THE PRO BLEM

O F DDCI SI O N

0gS

$ 95. Conclusion
The results we have reached on the basisof an historical and
systematic investigation of the Aristotelian syllogistic are at
more than one point different from the usual presentation.
Aristotle's logic was not only misrepresentedby logicians who
came from philosophy, since they wrongly identified it with the
traditional syllogistic, but also by logicians who came from
mathematics. In text-books of mathematical logic one can read
again and again that the law ofconversion of the l-premiss and
some syllogistical moods derived by this law, like Darapti or
Felapton, are wrong. This criticism is based on the mistaken
notion that the Aristotelian universal affirmative premiss 'All a
is 6' means the same as the quantified implication 'For all c, if c
is a, then c is D',where c is a singularterm, and that the particular
affirmative premiss'Somea is D' meansthe sameas the quantified conjunction 'For some ,, c is a and.c is b', where c is a-gaina
singular term. If one accepts such an interpretation, one can
say of course that the law CAabIbais wrong, becausea may be
an empty term, so that no c is a, and the above quantified
implication becomestrue (for its antecedent is false), and the
, above quantified conjunction becomesfalse (for one of its factors
is false). But all this is an imprecise misunder3tandingof the
Aristotelian logic. There is no passagein the Anal2ticsthat would
justi$' sgch an interpretation. Aristotle does not introduce into
his logic singular or empty terms or quantifiers. He applies his
logic only to universal terms, like'man' or'animal'. And even
these terms belong only to the application of the system,not to
the system itself, In the system we have only expressionswith
variable arguments, like Aab or lab, and their negations, and
two of these expressionsare primitive terms and cannot be
defined; they have only thosepropertiesthat are stated by the
axioms. For the same r,asonsuch a controversy as whether
the Aristotelian syllogistic is a theory of classesor not is in my
opinion futile. The syllogistic of Aristotle is a theory neither of
classesnor of predicates; it exists apart from other deductive
systems,having its own axiomatic and its own problems.
I have tried to set forth this systemfree from foreign elements.
I do not introduce into it singular, empty, or negativeterms, as
Aristotle has not introduced them. I do not introduce quanti-

$ gs

C ON C L U SION

rgr

fiers either; I have only tried to explain some ideas of Aristotle

by the help of quantifiers. In formal proofs I employ thesesof
the theory of deduction, sinceAristotle usesthem intuitively in
his prooft, and I employ rejection, because Aristotle himself
rejects some formulae and even states a rule of rejection.
Wherever in Aristotle's exposition there was something not
completely correct, I have been anxious to correct the flaws of
his exposition, e.g. some unsatisfactory proofs by reductioper
impossibile,or the rejection through concrete terms. It has been
my intention to build up the original systemof the Aristotelian
syllogistic on the lines laid down by the author himselfi,and in
accordancewith the requirements of modern formal logic. The
crown of the systemis the solution of the problem of decision,
and that was mpde possibleby Slupecki's rule of rejection, not
kfiown to Aristotle or to any other logician.
The syllogistic of Aristotle is a systemthe exactnessof which
surpasseseven the exactnessof a mathematical theory, and this
is its everlastingmerit. But it is a narrow systemand cannot be
applied to all kinds of reasoning,for instanceto mathematical
arguments. Perhaps Aristotle himself felt that his system was
not fitted for every purpose, for he added later to the theory of
assertoricsyllogismsa theory of modal syllogisms.rThis was
of coursean extensionof logic, but probably not in the right
direction. The logic of the Stoics, the inventors of the ancient
form of the propositional calculus, was much more important
than all the syllogisms of Aristotle. We realize today that the
theory of deduction and the theory of quantifiers are the most
fundamental branches of logic.
Aristotle is not responsible for the fact that for many centuries his syllogistic,or rather a corrupt form ofhis syllogistic,
was the sole logic known to philosophers. He is not responsible
either for the fact that the influence of his logic on philosophy
was, as it seemsto me, disastrous.At the bottom of this disastrous influence there lies, in my opinion, the prejudice that
cvcry proposition has a subject and,a predicate, like the premissesof Aristotelian logic. This prejudice, together with the
is the basis
criterion of truth known as adaequatio
rei et intelleetzs,
I I take it that the theory of modal syllogisms expounded by Aristotle in Chapters
8-zr of Book I of the Prior Anal2tics was inserted later, since Chapter e3 is obviously
an immcdiatc continuation of Chapter 7.

rg2

THE PRO BLEM

O F DECI SI O N

$gS

of somefamous but fantastic philosophicalspeculations.Kant

divided all propositions (he calls them Judgements') into analytic and synthetic according to the relation of the predicate of a
proposition to its subject. His Critiqueof PureReasonis chiefly an
attempt to explain the problem how true synthetic a priori
propo5itions are possible. Now some Peripatetics, for instance
Alexander, were apparently already aware that there exists
a large class of propositions having no subject and no predicate, such as implications, disjunctions, conjunctions, and so
on.r All thesemay be called functorial propositions, since in all
of them there occursa propositionalfunctor, like 'if-then', 'or',
'and'. Thesefunctorial propositionsare the main stock of every
scientific theory, and to them neither Kant's distinction of analytic and synthetic judgements nor the usual criterion of truth
is applicable, for propositions without a subject or predicate
cannot be immediatbly compared with facts. Kant's problem
loses its importance and must be replaced by a much more
important problem: How are true functorial propositions possible? It seemsto me that here lies the starting-point for a new
philosophy as well as for a new logic.
r In conns<ion with Aristotle's definition of the r,pdraots Alo<ander writes,
tr. 17: ;oi 6i oJror oi, 6pot npord,oeos oitdnls
ri.l,ld rfs ,inlfis rc xai xo)tovpdwls
xatrlyoprxffs' td ydp n xotd. nvos ETew xai td xa06Aou i iv pipet fi ditdptorov i6n
nhqr
fi ydp Jto9enxil oJx dv tQ tt xetd. tvos Aiyeoflat d),\' iv ciroloudig fi pdX71rd
ilqfiis fi rd /e08os 3xer,

CHAPTER

VI

A RI S T O T L E 'S M OD AL L OGIC
P R OPOSIT ION S

OF

$ 96. Introduction
THnnB are two reasonswhy Aristotle's modal logic is so little
known. The first is due to the author himself: in contrast to the
assertoricsyllogistic which is perfectly clear and nearly free of
errors, Aristotle's modal syllogistic is almost incomprehensible
becauseof its many faults and inconsistencies.He devoted to this
but the
subject some interesting chapters of De Interpretationa,
systemof his modal syllogistic is expounded in Book I, chapters
3 and &-zz of the Prior Anal2tics.Gohlker suggestedthat these
chapters were probably later insertions, because chapter z3
was obviously an immediate continuation of chapter 7. If he is
right, the modal syllogistic was Aristotle's last logical work and
should be regarded as a first version not finally elaborated by the
author. This would explain the faults of the systemas well as the
corrections of Theophrastus and Eudemus, made perhaps in
the light of hints given by the master himself.
The secondreasonis that modern logicianshave not asyet been
able to construct a universally acceptablesystem of modal logic
which would yield a solid basisfor the interpretation and appreciation of Aristotle's work. I have tried to construct such a
system,different from those hitherto known, and built up upon
Aristotle's ideas.2The present monograph on Aristotle's modal
logic is written from the standpoint of this system.
A modal logic of terms presupposesa modal logic of propositions. This was not clearly seenby Aristotle whose modal syllogistic is a logic of terms; neverthelessit is possibleto speakof an
Aristotelian modal logic of propositions, as someof his theorems
are generalenoughto compriseall kinds ofproposition, and some
others are expresslyformulated by him with propositional variables. I shall begin wittr Aristotle's modal logic of propositions,
I Paul Gohlke, Dic Entstzhmg der Aristntzliselun Ingik,Berlin (1936), pp. 8&94.
2
Jan Lukasiewicz, 'A System of Modal I-ogic', TIu Jotanal of ComputingSystctts,
vol. i, St. Paul (1953), pp. rIr-49. Asummaryof this paper appeared under the
same title in the ProccedingsoJ tfu Xhh Intcmational CongressoJ Philosoph2, vol. xiv,
Brussels (tgSg), pp. Be-87. A short description ofthe system is given below in $ 49.

tZ4

ARISTOTLE'S

M O DAL

LO G I C

O F PRO P O S I T I O N S

$ sZ

$g6

which is logically and philosophically far more important than

his modal syllogistic of terms.
$ 97. Modalfunctionsand their interrelations
There are four modal terms used by Aristotle i dvayrcaeov' necessary',&6ivarov-' impossible',6uvar 6v-' possible',and dv8ey6y.cvov-'contingent'. This last term is ambiguous: in the De
Interpretatione
it meansthe sameas 6uvar6v,in the Prior Analyticsit
has besidesa more complicated meaning which I shall discuss
later.
According to Aristotle, only propositions are necessary,impossible,possible,or contingent. Instead of saying: 'The proposition "1" is necessary',where "p" is the nameof the proposition
1D,I shall use the expression: 'It is necessarythat p', where p is a
proposition.So, for instance,insteadof saying: 'The proposition
"man is an animal" is necessary',
I shallsay: 'It is necessary
that
man shouldbe an animal.' I shall expressthe other modalitiesin
a similar way. Expressions
like : 'It is necessarythat p', denoted
hereby Lp, or 'It is possiblethatp', denotedby Mp,I call 'modal
functions' ; L and M, which respectivelycorrespondto the words
'it is necessary
that' and 'it is possiblethat', are 'modal functors',
I is their 'argument'. As modal functions are propositions, I say
that L and M are proposition-formingfunctorsofone propositional
argument. Propositionsbeginning with I or their equivalentsare
called,'apodeictic',thosebeginningwith M or their equivalents
'problematic'. Non-modal propositionsare called'assertoric'.
This modern terminology and symbolismwill help us to give.a
clear expositionof Aristotle'spropositionalmodal logic.
Two of the modal terms, 'necessary'and 'possible',and their
interrelations, are of fundamental importance. In the De Interpretatione
Aristotle mistakenly assertsthat possibility implies nonnecessity,i.e. in our terminology:
(a) If it is possiblethatp, it is not neccssar)
thatp.t He later sees
that this cannotbe right, becausehe acceptsthat necessity
implies
possibility,i.e.:
(b) If it is necessary
thatp, it is possiblethatp, and from (D) and
(a) there would follow by the hypothetical syllogism that
I De int. t3,22ar5
rQ piv yd.p }warir
etvat td dv6$6o0at etvat (d.xo\ou0ei),
ixeiva &vrrcrpi$eq
xai i
pi1 dlrjvatov etvat xai rd pi1 dvayxatov eivol.

toiro

xai

MODAL

F U N C TION S

AN D IN TER R EL ATION S

tg 5

(c) IJ it is necessary
thatp, which is abthatp, it is not necessar!
surd.r After a further examination of the problem Aristotle rightly
statesthat
(d) If it ispossiblethatp, it is notnecessaryt
thatnotp,2 but doesnot
This
correct-his former mistake in the text of De Interpretatione.
correction is given in the Prior AnaQticswhere the relation of
possibilityto necessityhas the form of an equivalence:
that
(c) It is possiblethat p-if and onl2 if-it it not necessarlt
notp.t
I gather from this that the other relation, that of necessityto
possibility, which is stated in the De Interpretatione
as an implication,+ fu also meant as an equivalence and should be given the
form:
(n n it necessar)
thatf-if andonfuif-it is notpossiblethat notp.
If we denote the functor 'if and only if ' by Q,,t putting it
before its arguments,and 'not' by "{ we can symbolically express
the relations(e) and (/) thus:
r. qMpNLNp, i.e. Mp-if and onlyif-NLNp,
z. QLpNMNp, i.e. Lp-if andonll if-NMNp.
The above formulae are fundamental to anv svstem of modal
logic.
$ 38. Basic modallogir
Two famous scholasticprinciples of modal logic: Ab oportere
ad
esseualetconsequentia,
were
and Ab essead posseualetconsequentia,
known to Aristotle without being formulated by him explicitly.
The first principle runs in our symbolic notation (C is the sign of
the functor'if-then') :
g. CLpp, i.e. If it is necessary
thatp, thenp.
The secondreads:
I lbid. zzbrr rd p2v ydp dveyxaiov etvat iwotdv el,vat . .. r4 <il)ri p)v tQ ye
}wotdv etvot td oJx d|rivotov etvecd.xoAou9ei,rorhq 62 rd pi dvayxaiov elvet' ,itote
ovppaiver rd dvayxdiov elvot pl dvayxatov elvat,6rcp dtonov.
2 [bid,. zzbzz Aeitetan rolvuv td oix d.vayxaeovp) ctvu &xoAovfleivtQ 6urar6, etvau
3 An. pr. i. 13,
Jndpyew' rci 'orir d}rivatov 6nd.pyeu' xai'oJx
32"25 td'iviiletor
dvdyxq pl [rdp76l', fitot taitd. Eotocfi dxo\ou|o0vrc dll{lor5.
a Dc int. tg,zz"zotQDt pil6warQ pi1etvet xei pj iv}e2qopilq p) elvetrd d.vayxoeov
etvat xei rd dilvorov pi eivet (<iroloudei).
5 I usually denote equivalence by E, but as this letter has already another
meaning in the syllogistic, I have introduced (p. to8) the letter Qfor equivalence.

,]

tg6

A RISTOTLE'S

M O DAL

LO G I C

O F PRO PO S T T I O N S

$g8

4. CpMp, i.e. If p, it is possiblethatp.

*5. CMpp, i.e. If it is possiblethatp, thm p-rejected.

The corresponding formulae for necessityare also stated by
Alexander who saysthat necessityimplies existence,i.e. Clpp,but
not conversely,i.e. CpLp shouldbe rejected.+We get thus another
rejected expression:
*6. CpLp,i.e. If p, it is nccessary
that p-rejected.
Formulae r-6 are acceptedby the traditional logic, and so far
as f know, by all the modern logicians. They are, however, insufficidnt to characterizeMp and Lp as modal functions, because
all the above formulae are satisfiedif we interpret Mp as always
true, i.e. as 'uentmof p', and Lp as alwaysfalse, i.e. as falnn
ofy''. With this interpretation a systembuilt up on the formulae
r-6 would ceaseto be a modal logic. We cannbt thereforeassert

*7. M!, i.e. It is possiblethat p-rejected, and

*8. NLp, i.e. It is not necessary
that p-rejected.
Both formulae may be called Aristotelian, as they are consequencesof the presumption admitted by Aristotle that there exist
' 4r. ry. i. 16, 36815 $avepdv6' 6r xoi ro6 tv6yeo0u pi1 ind.pTeu ylyvetcr ouMoynpds, eine-p xai rcO pl $td.pyeu. - iv6iTeo0at means here th;-.po$ible', not the
'contingent'.
I Alexander zog. z td piv yi,p indp1pv
xai ivieydpcvov ilqlis elmeiv,zd E' e'v6c1tipevov oi rdmus xoi Jndpyov.
3 Asserted enprcssions are marked throughout
the chapten vI-vIII
by arabic
numerals without asterisks.
Alcxander t gz. gi td ydp dvayxatov xoi &rdpyov, oixht 6i t6 Jrd.pyov dvoyxatov.
'

0g8

B A SIC M OD AL

L OGIC

r37

asserted apodeictic propositions. For, if Ic is asserted, then

LNNa must be assertedtoo, and from the principle of Duns
Scotus CpCNpq we get by substitution and detachment the
assertedformulae CNLup and, CNINNa|. As r' is rejected,
-lVZc
and NLNNa are rejected too, and consequently NLp and NLN4,
i.e. Mp, must be rejected.
I call a system 'basic modal logic' if and only if it satisfies
the formulae r-8. I have shown that basic modal logic can be
axiomatized on the basisof the classicalcalculus of propositions.r
Of the two modal functors, M and L, one may be taken as the
primitive term, and the other can be defined. Taking M as the
primitive term and formula z as the definition of Z, we get
the following independent set of axioms of the basicmodal logic :

4.CpMp *s.CMpp *7. Mp

s. QM|MNN\,

where 9 is deductively equivalent to formula r on the ground of

the definition z and the calculus of propositions.Taking Z as the
primitive term and formula r as the definition of M, we get a
correspondingset of axioms:
*6. CpLp *8. NLp
g.CLpp
rc. QI.1LNNq,
where ro is deductively equivalent to formula e on the ground
of the definition r and the calculus of propositions.The derived
formulae g and ro are indispensableas axioms.
Basic modal logic is the foundation of any system of modal
logic and must always be included in any such system.Formulae
r-B agree with Aristotle's intgitions and are at the roots of our
conceptsof necesity and possibility; but they do not exhaustthe
whole stockof acceptedmodal laws.For instance,we believethat
if a conjunction is possible,each of its factors should be possible,
i.e. in symbols:
rr. CMKpqMp
and
12. CMKpqMq,
and if a conjunction is necessary,each of its factors should be
necessary,
i.e. in symbols:
13. CLKpqLp and
14. CLKpqLq.
None of theseformulae can be deduced from the laws l-8. Basic
modal logic is an incomplete modal system and requires the
addition ofsome new axioms. Let us seehow it wassupplemented
by Aristotle himself.
I See pp. r r4-r7 of my paper on modal logic.

r3 B

ARISTOTL E' S

M O DAL

LO G I C

O F PR O P O S I T I O N S

$gs

$ 39. Laws of extercionalit2

Aristotle's most important and-as I seeit-most successful
attempt to go beyondbasicmodal logic consistedin his accepting
certain principleswhich may be called 'laws of extensionalityfor
modal functors'. These principles are to be found in Book I,
and are formulated in three
chapter 15 of the Prior Anal2t'i.cs,
passages.
We read at the beginning of the chapter:
'First it has to be said that if (if cris, p must be), then (if cuis
possible,p must be possibletoo).'t
A few lines further Aristotle saysreferring to his syllogisms:
'If one should denote the premissesby c, and the conclusion
by p, it would not only result that if a is necessary,then p is
necessary,
but also that if a is possible,then p is possible.'z
And at the end of the sectionhe repeats:
'It has been proved that if (if c is, p is), then (if c is possible,
then B is possible).'r
Let us first analysethesemodal laws beginningwith the second
passage,which refersto syllogisms.
All Aristotelian syllogisrrxiare implications of the form CcB
where cr is the conjunction of the two premissesand p the conclusion.Take as examplethe mood Barbara:
ry. CKAbaAcbAca.
aB
According to the secondpassagewe get two modal tfeorems,in
the form of implications taking Cup as the antecedentand CLILB
or CMuMp as the consequent,in symbols:
t7. CCapCMuMp.
16. CC"pCLaLp and
The lettersc and B stand here for the premisses
and the conclusion of an Aristotelian syllogism. As in the final passagethere is
I An. pr. i. r5,
34'5 npdrot 3i )exriov 6tr ei rcA A Swos d,vdyxq rd B elvoq xai
iwatoa 6vtos roa A \uvatdv Eorac xai td B d( dvdyxqs.
2 Ibid,
34822 ei tc |eiq rd piv A rd.s npordoets, rd 6i B rri oupntpoopa, ovpBoivot
6v oJ pdvov d,vayxaiou toA A 6vros dpa xai rd B elvar d.voyxaiov, dMd rci SweroA
3wat6v.
3 lbid.
34a29 6dDerrrcr 6tt ei toA A 6wos td B Eorr, xei iwaroi 6vros toA A Eotat
ri B 6wat6v.

L A WS OF EXTEN SION AL ITY

5 gg

r39

no referenceto syllogisms,we may treat thesetheoremsas special

casesof general principles which we get by replacing the Greek
lettersby propositionalvariables:
rB. CCpqCLpLq and
rg. CCpqCMpMq.
(laws
of extensionBoth formulaemay be called in a wider sense
ality', the first for I, the second for M. The words 'in a wider
sense'require an explanation.
The general law of extensionality, taken sensustricto, is a
formula of the classicalcalculusof propositionsenlargedby the
introduction of variable functors.and has the form:
zo. CQpqCEpDq.
This meansroughly speaking:Ifp is equivalentto q, then if 6 of
p, 6 of q, where 6 is any proposition-forming functor of one propositional argument, e.g. Jf. Accordingly, the strict laws of
extensionalityfor L and M will have the form:
zz. CfuqCMpMq.
zt. CQpqCLpLq and
Thesetwo formulae have strongerantecedentsthan formulae tB
and r9, and are easilydeduciblefrom them, zr from rB, and zz
from rg, by meansof the thesisCQpqCpqand the principle of the
hypotheticalsyllogism.It can be proved,however,on the ground
of the calculusofpropositionsand the basicmodal logic that converselyrB is deduciblefrom zr, and lg from zz. I give here the
full deduction of the l-formula:'
The premisses:
zg. CCQpqrCpCCpqr
24. CCpqCCqrCpr
25. CCpCqCprCqCpr

3.clpp.
The deduction:

zg. rlClpLqxCzvz6
cpccpqcLpLq
26.
x Cg-Cz6-27
24.pILp, qlp, rlCCpqCLpLq
27. CLpCCpqCLpLq
25.pllp, qlcpq,rlLqxCzT-tB
rB. CCpqCLpLq.
q

r4o

ARISTOTL E'S

M O DAL

LO G I C

O F PRO PO S I T I O N S

$ eS

In a similar way rg is deduciblefrom ez by meansof the premisses CCQpqrCNqCCpqr,CCpqCCqrCpr,CCNpCqCrpCqCrp,and,

the transposition CNMpNp of the modal thesis CqMP.
We seefrom the above that, given the calculus of propositions
and basic modal logic, formula rB is deductively equivalent to
the strict law of extensionality 2t, and formula rg to the strict law
ofextensionahtyzz. We are right, therefore,to call thoseformulae
'laws of extensionality in a wider sense'.Logically, of course, it
makes no difference whether we complete the Z-systemof basic
modal logic by the addition ofCCpqCLpLqor by the addition of
CQpqCLpLq; the sameholds for the alternative additions to the
.iVf-system
of CCpqCMpMqor CQpqCMpMq. Intuitively, however,
the differenceis great. Formulae IB and Ig are not so evident as
formulae zr and 22. lf p implies 4 but is not equivalent to it, it is
not always true that if 6 ofp, 5 of q; e.g. CNpNq doesnot follow
from Cpq.But ifp is equivalent to q, then always if 6 ofp, E of q,
i.e. if p is true, g is true, and if p is false, 4 is false; similarly ifp is
necessary,4 is necessary,and ifp is possible,4 is possible.This
seemsto be perfectly evident, unlessmodal functions are regarded
as intensional functions, i.e. as functions whose truth-values do
zol depend solely on the truth-values of their arguments. But
what in this casethe necessaryand the possiblewould mean, is
for me a mystery as yet.

S +o

A R I S T O T L E 'S

PR OOF OF TH E

M - L AW

t1 l

Alexander assimilatesthis Aristotelian definition ofcontingency to

that of possibility by omitting the words 'which is not necessary'.
He says 'that a B which is impossible cannot follow from an c
which is possible may also be proved from the definition of
possibility: that is possible,the supposedexistenceof which implies nothing impossible'.rThe words 'impossible'and 'nothing'
here require a cautious interpretation. We cannot interpret 'impossible'as 'not possible',becausethe definition would be circular; we must either take 'impossible'asa primitive tenn or, taking
'necessary'as primitive, define the expression 'impossible that
/' by 'necessarythat not p'. I prefer the secondway and shall
discussthe new definition on the ground of the Z-basic modal
logic. The word 'nothing' should be rendered by a universal
quantifier, as otherwise the definition would not be correct. We
get thus the equivalence:
zB. QMpIIqCCpqNLNq.
That means in words: 'It is possiblethat p-if and only if-for
all q, if (ifp, then 4), it is not necessarythat not q.' This equivalence has to be added to the Z-basic modal logic as the definition
of Mp instead of the equivalence r which must now be proved as
a theorem.
The equivalencez8 consistsof two implications:
zs. CMpIIqCCpqNLNq
go. CIIqCCpqNLNqMp.
'iO
From e9 we get by the theorem CIIqCCpqNINqCCpqNLNq and,
the hypothetical syllogism the consequence:
y. CMpCCpqNLNq,

$ 4o. Aristotb's proof of the M-Iaw of extensionalit2

In the last passagequoted above Aristotle says that he has
proved the law of extensionality for possibility. He argues in
substancethus: If cr is possibleand B impossible, then when cr
came to be, B would not come to be, and therefore a would be
without p, which is against the premiss that if c is, p is.' It is
difficult to recast this argument into a logical formula, as the
term 'to come to be' has an ontological rather than a logical
meaning. The comment,'.however,given on this argument by
Alexander deservesa careful examination.
Aristotle definesthe contingent as that which is not necessary.
and the supposedexistenceof which implies nothing.impossible.z

and from 3r there easilyresultsby the substitution qlp, Cpp,commutation and detachment the implication CMPNLN|. The converse implicanon CNWpMp which, when combined with the
original implication, would give the equivalence r, cannot be
proved otherwise than by means of the law of extensionality for
L: CCPqCL1L4.As this proof is rather complicated, I shall give
it in full.

I An,pr.i.I5,g4oSeioJvtip2v6uvatdv,SteDwatdvetvaqylvon'dvrtd6'd}ivarov,
6t' d,6ivorov, oix Ev y6ono, iipt 6' ei td A Dwcriv xoi t6 B ,i6ivotov, 2v6(yon' Ev ri
2 See below, p. r54, n.
A yelo1ar dvev ro0 B, ei 6A yevioilaq xel etvat.
3.

! Alexander
r77. tt ietxvtiotro 6' dv,6tt
neo|or td B, xai ix roi 8pcopoi to6 }weto0
ori6,iy <iElvczoy oupBaivet 6cd toAto.

pl

otdv re |uvarQ 6mt tQ A d|ivotov

inu,
oi 6note0tnos
elvot
. . .iwordv

r42

A R I STOTL E'S

MODAL

LO G I C

O F PRO PO SI TI O N S

0 +o

The premisses:
fi. CCpqCLpLq
24. CCpqCCqrCpr
so. CIIqCCpqNLNqMp
gz. CCpqCNqNp
gg. CCpCqrCqCpr.
The deduction:
tB. plNq,qlMpxg4 \
34. CCNqNqCLNqLNP
24.pICpc,qICNqNp,rI CLNqLNqx Cgz-Cg4-35
s5. CCpqCLNqLNp
32.pllNq, qlLNpx36
36. CCLNqLNqCNLN4NLNq
2+. pICpq,qlCLNqLNp,rI CNLNpNLNqx Cg5-C9617
s7. CCpqCNLNpNLNq
33.plcpq, qlNLNp, TINLNqxCSZ-gB
sB. CNLNpCCpqNLNq
gB.Ilzqxgg
CNLNpIIqCCpqNLNq
39.
za. pINLNp, qIII qCCpqNLNq,rIMp x Cgg-Cgo-4o
CNLN|M|.
4o.
for M, which was
We can now provethe law of extensionality
the purposeofAlexander'sargument.This law easilyresultsfrom
that the proof by
the equivalencer and thesis37.We seebesides
commeansof the definition with quantifiersis unnecessarily
plicated.It sufficesto retain definition r and to add to the Lsystemthe ZJaw of extensionalityin order to get the M-law of
extensionality. In the sameway we may get the L-law of extensionality, if we add the M-law of extensionalityto the M-system
and definition z. The Z-systemis deductively equivalent to the
M-system with the laws of extensionalityas well as without them.
It is, of course, highly improbable that an ancient logician
could haveinventedsuchan exactproofas that given above.But
the fact that the proof is correct throws an interesting light on
Aristotle's ideas of possibility. I supposethat he intuitively saw
what may be shortly expressedthus: what is possibletoday, say
a sea-fight,may becomeexistentor actual tomorrow; but what is

$4o

A R I S T O T L E 'S

PR OOF OF TH E

M - L AW

r +z

impossible,can never becomeactual. This idea seemsto lie at the

bottom of Aristotle's proof and of Alexander's.
connexions
of propoitions
$ 4r. Necessar2
The IJaw of extensionality was formulated by Aristotle only
once, together with the M-law, in the passagewhere he refers to
syllogisms.r
According to Aristotle there existsa necessaryconnexion between the premissescr of a valid syllogism and its conclusion p.
It would seemthereforethat the laws of extensionalityformulated
abovein the form:
16. CCupCLaLB and
17. CCuBCMuMB,
should be expressedwith necessaryantecedents:
and
4r. CLCapCLaLp
42. CLCuBCMaMB,
and the correspondinggenerallaws of extensionalityshouldrun:
and
$. CLCpqCLpLq
44. CLCpqCMpMq.
This is corroborated for the M-law by the first passagequoted
abovewherewe read : 'If (if cris, p must be), then (if c is possible,
p is possible).'
Formulae 43 and 44are weaker than the correspondingformulae with assertoricantecedents,rB and 19, and can be got from
them by the axiom CLpp and the hypothetical syllogism 24. It is
not, however, possibleto derive the strongerformulae conversely
from the weaker. The problem is whether we should reject the
strongerformulae rB and r9, and replace them by the weaker
formulae 43 and 44. To solve this problem we have to inquire
into the Aristotelian concept of necessity.
Aristotle accepts that some necessary,i.e. apodeictic, propositions are true and sfrould be asserted.Two kinds of asserted
apodeictic prbposition can be found in the Anajtics: to the one
kind there belong necessaryconnexions of propositions, to the
other necessaryconnexionsof terms. As example of the first kind
any valid syllogismmay be taken,for instance[he mood Barbara:
(g) If euer2b is an a, andeuerlc is a b, thenit is necessary
that euer2
c shouldbean aHere the 'necessary'does not mean that the conclusion is an
r Sec p. r38, n. e.

14+ ARISTOTLE'S

M O DAL

LO G I C

O F PRO P O S I T I O N S

$ +r

apodeictic proposition, but denotes a necessaryconnexion between the premissesof the syllogismand its assertoricconclusion.
This is the so called 'syllogisticnecessity'.Aristotle seesvery well
that there is a difference between syllogistic necessityand an
apodeictic conclusionwhen he says,discussinga syllogismwith an
assertoricconclusion,that this conclusionis not 'simply' (tiz)6's)
i.e. necessary
necessary,
in itself, but is necessary'on condition',
i.e. with respect to its premisses(zorizc.rv6nav)J There are
passages
where he puts two marks of necessityinto the conclusion
saying,for instance,that from the premisses:'It is necessarythat
every b should be an a, and somec is a b', there follows the conclusion: 'It is necessarythat somec should be necessarilyan a.'z
The first 'necessary'refersto the syllogisticconnexion,the second
denotesthat the conclusionis an apodeictic proposition.
By the way, a curiousmistakeof Aristotle should be noted: he
saysthat nothing followsnecessarilyfrom a singlepremiss,but ohly
from at leasttwo, as in the syllogism.rIn the Posterior
Ana\tics he
assertsthat this has been proved,+ but not even an attempt of
proof is given anywhere.On the contrary, Aristotle himself states
that 'If someDis an a, it is necessarythat somea should be a b',
drawing thus a necessaryconclusionfrom only one premiss.s
I have shown that syllogistic necessitycan be reduced to universal quantifiers.6When we say that in a valid syllogism the
conclusion necessarilyfollows from the premisses,we want to
state that the syllogismis valid for any matter, i.e. for all valuesof
the variablcs occurring in it. This explanation, as I have found
afterwards,is corroborated by Alexander who assertsthat: 'syllogistic combinations are those from which somethin$ necessarily
follows, and suchare thosein which for all matter the samecomes
to be'.7 Syllogistic necessityreduced to universal quantifiers cdn
I An. pr. i. ro,
3ob3z rd oupnipaope oix Esrrv d.vayxaiov dnArits, d)tAd nhov
dvtuv dvayxaiov.
2 lbid.
ptv A navrl rQ B $rapyira i{ dvdyqs, i 6i B nvl rQ I
9, goagT i
$nepTito pdvov' dldyxl 6l zd A rwi 7Q | $td.p1;ewi{ dvdyxls,
3 lbid, t5,
94"17 oi ydp Eotw d0Eiy e'f dvdyxr1s vds rwos 5vros, d,\,\d 6uoiv
il\oTtototv ,otov Stav ai npotd.oersoizcog EXoow ds t\iyirl xatd, tdv ouiloynp6v.
a An. post' i.
3, 73"7 &ds pir o$v xetpivou DdDerrzqo6tr oi6tnot' d,vdyxq rc etvat
Etepov (Aiyo6' iv6s,6noi7e 6pov vdsoite 1ioeos prf,s rc|eiot1s), dr 6Jo 6i ddoeoy
'tprirov
xai 2\oyistav tviyeteu
s An. pr. i. z, z5'zo ei ydp rd A rui tQ B, xai td B nvi rQ A ,&vdyq ind.pyeu.
6 Seeg5.
7 Alexander zo8. t6 ouA)oyiatuai62 at ou(vykt airot of i{ dvdyxqs n nvdyouoar.
rota0tac 6d, e cis 2d rd,otls iLls yiverat td oird.

N E C E S S A R Y CON N EXION S

$4r

OF PR OPOSITION S

r +s

be eliminated from syllogistic laws, as will appear from the following consideration.
The syllogism(g) correctly translated into symbolswould have
the form:
(h) LCKAbaAcbAca,
which meansin words:
(i) It is necessar)
that (if eueryb is an a, andeaeryc is a b, theneuerlt
c shouldbean a).
The sign of necessityin front of the syllogism shows that not
the conclusion, but the connexion between the premissesand
the conclusion is necessary.Aristotle would have asserted(r).
Formula
(j) CKAbaAcbLAca,
which literally correspondsto the verbal expression(g), is wrong.
Aristotle would have rejected it, as he rejects a formula with
stronger premisses,viz.
(k) CKAbaLAcbLAca,
i.e. 'If eaeryb is an a andit is neccssary
that eaer2c shouldbea b, it is
necessar)
that euerlc shouldbean a.'r
By the reduction of necessityto universal quantifiers formula
(h) can be transformed into the expression:
(l) IIaIIbIIcCKAbaAcbAca,
i.e. 'For all a, for all b, for all c (if ei'ery Dis an a and every c is a b,
then every c is an a).' This last expressionis equivalent to the
mood Barbara without quantifiers:
(m) CKAbaAcbAea,
since a universal quantifier may be omitted when it standsat t}re
head of an assertedformula.
Formulae (n) and (m) are not eqirivalent. It is obvious that (m)
can be deducedfrom (n) by the principle CLpp,but the converse
deduction is not possiblewithout the reduction of necessityto
universal quantifiers. This, however, cannot be done at all, if the
above formulae are applied to concrete terms. Put, for instance,
I An. pr. i. g,
3ore3 ei Et zd piv AB pl Enu dvayxatov, zd 6t Bl" dvayxatov, oir
*,:;r:,
oupnipaopo dvovxaeov.
L

t 46 ARISTOTL E'S

M O DAL

LO G I C O F PRO POS I T I O N S

0 +'

in (l) 'bird' for b,'crow' for a, and 'animal' for c; we get the
apodeictic proposition :
bird is a crowand eaerlanimal is
(n) It is necessar!
that (if eaer.y
a bird, theneaeryanimalshouldbea now).
From (z) resultsthe syllogism(o):
(o) If eaerybird is a crow and eueryanimal is a bird, theneuerl
animalis a crow,
but from (o) we cannot get (z) by the transformation of necessity
into quantifie$, as (z) doesnot contain variableswhich could be
quantified.
And here we meet the fint difficulty. It is easyto understand
the meaning of necessitywhen the functor I is attached to the
front of an assertedproposition containing free variables. In this
casewe have a general law, and we may say: this law we regard
asnecessary,becauseit is true of all objectsof a certain kind, and
doesnot allow of exception.But how shouldwe interpret necessity, when we have a necessaryproposition without free variables,
and in particular, when this proposition is an implication consisting of false antecedentsand of a false consequent,as in our
example (z) ? I seeonly one reasonableanswer: we could say that
whoever acceptsthe premissesof this syllogismis necessarilycompelled to acceptits conclusion.But this would be a kind ofpsychological necesity which is quite alien from logic. Besidesit is
extremqly doubtful that anybody would accept evidently false
propositionsas true.
I know no better remedy for removing this difficulty than to
drop everywherethe l-functor standing in front of an asserted
implication. This procedure waS already adopted by Aristotle
who sometimesomits the sign of necesity in valid syllogistical
moods.t
$ 42. 'Material'or'strict',implication?
According to Philo of Megara the implication 'If p, then q',
i.e. Cpq,is true if and only if it doesnot begin with a true antecedent and end with a false consequent.2This is the so-called
'material' implication now universally acceptedin the classical
calculus of propositions.'Strict' implication: 'It is necessarythat
t Scc p. ro, D.
5.

' See p. 83, n. r.

$ +z

'M A T E R I A L '

OR 'STR IC T'

IM PL IC ATION ?

t4 7

ifp, then q',i.e. LCpq,is a necessary

material implication and was
introduced into symbolic logic by C. I. Lewis. By means of this
terminology the problem we are discussingmay be stated thus:
Should we interpret the antecedent of the Aristotelian laws of
extensionalityasmaterial,or asstrict implication ? In other words,
shouldwe acceptthe strongerformulae rB and 19 (I call this the
'strong interpretation'), or should we reject them accepting the
weaker formulae 49 and,44 (weak interpretation)?
Aristotle was certainly not aware of the difference between
thesetwo interpretations and of their importance for modal logic.
He could not know Philo's definition of the material implication.
But his commentator Alexander was very well acquainted with
the logic of the Stoic-Megaric school and with the heated controversiesabout the meaning of the implication amidst the followers of this school. Let us then see his comments on our
problem.
Commenting on the Aristotelian passage'If (if cvis, B must
be), then (if cr is possible,p must be possible)'Alexander emphasizesthe necessarycharacter of the premiss 'ff c is, p must
be'. It seemstherefore that he would accept the weaker interpretation CLCapCMuM p and the weaker M-law of extensionality
CLCpqCMpMq.But what he meansby a necessaryimplication is
different from strict implication in the senseof Lewis. He says
that in a necessaryimplication the consequent should always,
i.e. at any time, follow from the antecedent, so that the proposition 'If Alexander is, he is so and so many yearsold' is not a
true implication, even if Alexander were in fact so many yeanr
old at the time when this propositionis uttered.'We rriay saythat
this proposition is not exactly expressed,and requiresthe addition
of a temporal qualification in order to be always true. A true
material implication must be, of course,always true, and if it
containsvariables,must be true for all valuesof the variables.
Alexander'scomment is not incompatiblewith the strong interpretation; it doesnot throw light on our problem.
Some more light is thrown on it, if we replace in Alexander's
proof of tll'e M-law of extensionality expounded in $ 4o the
I Alexander t76. z Eon 3l ,ivoyxoia d.xoAouflia
oiy fi npSoxaryos,ril)' dy .rf<iei zti
<iArlppivov Eneo|ar ton rQ d eihlppivov ,is iyoitrcvov etvau oJ yap d,Aqlts owtlFp&ov rd 'ei A\Qavdpos Eoru, AAl(avipos \nAiyerac', i 'ei AA!(avipos Zotc, rooiiie
dr<iv iort' , rci (ei) ei1, 6te Aiyerat I rp6taots, rooohav iritv.

t4 8

ARISTOTLE' S

M O DAL

LO G I C

OF PROPOSITIONS

$ +z

material implication Cpq by the strict implication LCpq. Transforming thus the formula
gr. CMpCCpqNLNq,
we get:
45. CMpCLCpqNLNq.
From 3r we can easily derive CM4NLNq by the substitution 4/p
getting CMpCCppNLNp,from which'our proposition results by
commutation and detachment, for Cppis an assertedimplication.
The sameprocedure, however, cannot be applied to 45. We get
CMpCLCppNLNp,but if we want to detach CMqNLNq we must
assertthe apodeictic implication LCpp. And here we encounter
the samedifficulty, as describedin the foregoingsection.What is
the meaning of LCpp? This expressionmay be interpreted as a
general law concerning all propositions, if we transform it into
npcpp; but such a transformation becomesimpossible, if we
apply LCpp to concreteterms, e.g. to the proposition 'Twice two
is five'. The assertoricimplication 'If twice two is five, then twice
two is five'is comprehensibleand true being a consequenceof the
law of identity Cpp;but what is the meaning of the apodeictic
implication 'It is necessarythat if twice two is five, then twice
two should be five'? This queer expressionis not a general law
concerningall numbers; it may be at most a consequenceof
an apodeictic law, but it is not true that a consequenceof an
apodeictic proposition must be apodoictic too. Cpp is a consequence of LCpp according to CLCppCpp,a substitution of CLpp,
but is not apodeictic.
It followsfrom the above that it is certainly simplbr to interpret
Alexander's proof by taking the word oupcpaivet
of his text in the
senseof material rather than strict implication. Neverthelessour
problem is not yet definitively solved.Let us thereforeturn to the
other kind of assertedapodeictic proposition accepted by .A,ristotle, that is to necessaf connexionsof terms.
$ 49. Anafuticpropositions
Aristotle assertsthe proposition: 'It is necessarythat man
should be an animal.'I He stateshere a necessaryconnexion
between the subject 'man' and the predicate 'animal', i.e. a
I An. pr. i.
9, 3oa3o (Qov piv ydp 6 dv|punos i{ dvdyxqs iori.

$+s

A N A L YTIC

PR OPOSITION S

r49

necessaryconnexion between terms. He apparently regards it as

obviousthat the proposition'Man is an animal', or better 'Every
man is an animal', must be an apodeicticone, becausehe defines
'man' as an 'animal', so that the predicate 'animal' is contained
in the subject 'man'. Propositionsin which the predicate is contained in the subjectare called 'analytic', and we shall probably
be right in supposing that Aristotle would have regarded all
analytic propositionsbasedon definitions as apodeictic, since he
saysin the PosteriorAnalyticsthat essentialpredicates belong to
things necessarily,rand essentialpredicatesresultfrom definitions.
The most conspicuousexamples of analytic propositions are
thosein which the subjectis identical with the predicate.If it is
necessarythat every man should be an anirnal, it is,'a fortiori,
necessarythat every man should be a man. The law of identity
'Every c is an a' isan analytic proposition,and consequentlyan
apodeicticone. We get thus the formula:
(p) LAaa, i.e. It is necessar)
that euerlta shouldbean a.
Aristotle doesnot state the law ofidentity Aaa as a principle of
his assertoricsyllogistic; there is only one passage,found by Ivo
Thomas, where in passinghe usesthis law in a demonstration.z
We cannot expect, therefore, that he has known the modal thesis
LAaa.
The Aristotelian law of identity Aaa, where r4 means 'every-is'
and a is a variable universal term, is different from the principle
of identity Jxx, where J means 'is identical with' and r is a
variable individual term. The latter principle belongs to the
theory of identity which can be established on the following
axioms:
@) J*r, i.e. x is identiculwith x,
(r) CJxTC$x$1,i.e. If x is identicalwith y, then if x satisfes$,
2 satisfes$,
where f is a variable proposition-forming functor of one individual argument. Now, if all analytic propositions are necessary,
so also is (4), and we get the apodeictic principle :
(s) LJxx, i.e. It is necessar)
that x shouldbeidenticalwith x.
' An. post,i, 6, 74b6 zri Ei rcd'adrd. tind.pyovro dvayxoio rois npdypoor.
2 lvo Thomas, O.P., 'Farrago Logica', DominieanStudies,vol. iv (1951), p.
7r.
The passage reads (ln. pr, ii, zz,68"19) xaqyopeiza Dt zo'B xei ait6 e$roi,

r5o

ARISTOTLE'S

M O DAL

LO G I C

O F PROP O S I T I O N S

$ +g

It has been observed by W.V. Qyine that the principle (s), if

asserted,leads to awkward consequences.IFor lf LJxx is asserted,
works
we can derive (t) from (r) by the substitution $lLJx'-LJx
here like a proposition-forming functor of one argument:
(4

CJx2CLJxxLJxl,

and by commutation
(u) CLJxxCJxlLJxy,
from which there follows the proposition:
(u) CJxyLJxy.
That means, any two individuals are necessarilyidentical, if they
are identical at all.
The relation of equality is usually treated by mathematicians
as identity and is bascd on the same axioms (q) and (r). We may
therefore interpret J as equality, x and2 as individual numbers
and say that equality holds necessarilyif it holds at all.
Formula (u) is obviously false. Q;rine gives an example to show
its falsity. Let.r denote the number of planets, andT the number
9. It is a factual truth that the number of (major) planets is equal
to g, but it is not necessarythat it should be equal to g. Qyine
tries to meet this difficulty by raising objections to the substitution
of such singular terms for the variables. In my opinion, however,
his objections are without foundation
There is another awkward consequenceof the formula (r.r)not
mentioried by Quine. From (a) we git by the definition of Z and
the law of transposition the consequence:

(w) CMNJx2NJx2
That means: 'If it is possiblethat.r is not equal to7, then r is
(actually) not equal to7.'The falsity of this consequence
may be
seenin the following example: Let us supposethat a number r
has beenthrown with a die. It is possiblethat the numberT next
thrown with the die will be different from r. But if it is possible
that .r will be different from2, i.e. not equal to7, then according
to (nr) r will actually be different from y. This consequenceis
obviously wrong, asit is possibleto throw the samenumber twice.
t W. V.
Quine, 'Three Grades of Modal Involvement', Proccedingsof thc XIth
Internalional Congressof Philosophy, vol. xiv, Brussels (1953). For the following
argumentation I am alone responsible.

I ,r:r

A N A L YTIC

PR OPOSITION S

r5l

'l'lrcn: is, in my opinion, only one way to solve the above diffir rrltics: w(: must not allow that formula ZJxr should be asserted,
i.r'. tlrrrt thc principle of identity Jxx is necessary. As Jxx is a
tylrir:ul unalytic proposition, and as there is no reason to treat this
prilrr:iplc in a different way from other analytic propositions, we
;rrr'<:ompclledto assumethat no analytic proposition is necessary.
llt'firrc dcaling with this important topic let us bring to an end
orrr invcstigation of Aristotle's concepts of modalities.
I U. A" Aristotelianparadox
'l'lrt:rc is a principle of necessityset forth by Aristotle which is
lriglrly controversial. He says in the De Interpretationethat'anytlring cxistent is necessary when it exists, and anything noncxistt:nt is impossible when it does not exist'. This does not mean'
lrr rulds, that whatever exists is necessary,and whatever does not
cxist is impossible: for it is not the same to say that anything
cxistr:nt is necessarywhen it does exist, and to say that it is simply
rrcccssary.rIt should be noted that the temporal 'when' (dzav) is
rrsctlin this passageinstead of the conditional 'if'. A similar thesis
is sct lorth by Theophrastus. He says,when defining the kinds of
tlrirrgs that are necessary,that the third kind (we do not know
wlr:rt the first two are) is 'the existent, for when it exists, then it is
irrrpossible that it should not exist'.2 Here again we find the
tcrnporal particles 'when' (dze) and 'then' (zdre). No doubt an
lur:tlogous principle occurs in medieval logic and scholars could
lirrrl it there. There is a formul4tion quoted by Leibniz in his
'l luodiceerunning thus: Unumquodque,quandoest, oportetessa.3
Note
;rg:rin irr this sentence the temporal quando.
What does this principle mean? It is, in my opinion, ambiguorrs. Its first meaning seems to be akin to syllogistic necessity,
wlrit'h is a necessaryconnexion not of terrns, but of propositions.
r\h'xirnder commenting on the Aristotelian distinction between
nirrrplc and cclnditional necessity,asays that Aristotle was himself
I I)e int
9, 19'23 i ptv o$v elvat td 3v,6tav fi, xai rd p) 6v pii elvat, &av pi1 fi,
,ir,!yr1 6l plv oire td 6v drov dvdyxq etvat oire td pi1 6v pi1 etvot' OJ ydp taltdv
,'rrr ni 6v d.nav elvat i( d.v&yx1s6te Eotq xoi rd dzl.i,s etvat d{ 'iv,iyxqs.
' Afcx:rndcr I56. zg 6 yoiv @e6spoorosiv tQ npittq rdv llporipuv &voAurcx,Tv
,\,ly,q, nqri tdv ind roa d.vayxalouoqltocvoltivov oirus yp<i$er'rphov td indplov
i't, 1,ip inipTet, t6re oiy otdv rc pl indp76w.'
I I'hiloufihivhe SchriJten,
ed. Gerhardt, vol. vi, p. I3r.
r S cc
1r. r4+, n. r.

t52

ARISTOTL E'S

M O DAL

LO G I C

O F PRO PO S I T I O N S

944

aware of this distinction, which was explicitly made by his

friends (that is, by Theophrastusand Eudemus),and quotesas
a further argument the passageof the De Interpretatione
above
referred to. He is aware that this passageis formulated by
Aristotle in connexion with singular propositions about future
events, and calls the necessityinvolved 'hypothetical necessity'
(dvayrcaCov i{ iro0be:os).|

This hypothetical necessitydoes not differ from conditional

necessity,except that it is applied not to syllogisms,but to singular
propositionsabout events.Such propositionsalways contain a
temporal qualification. But if we include this qualification in the
content of the proposition, we can replace the temporal particle
by the conditional.So,for instance,insteadofsayingindefinitely:
'It is necessary
that a sea-fightshouldbe, when it is', we may say:
'It is necessarythat a sea-fightshould be tomorrow, if it will be
tomorrow.' Keeping in mind that hypothetical necessityis a
necessaryconnexion of propositions,we may interpret this latter
implication as equivalent to the proposition: 'It is necessarythat
if a sea-fight will be tomorrow, it should be tomorrow' which is
a substitution of the formula LCpp.
The principle of necessitywe are discussingwould lead to no
controversy, if it had only the meaning explained above. But it
may have still another meaning: we may interpret the necessity
involved in it as a necessaryconnexion not of propositions,but of
terms. This other meaning seemsto be what Aristotle himself has
in mindg when he expoundsthe determinist argument that all
future eventsare necessary.
In this connexiona generalstatement
given by him deservesour attention. We read in the De Interpretatione:'If it is true to say that somethingis white or not white,
it is necessary
that it shouldbe white or not white.'z It seemsthat
here a necessaryconnexion is stated between a 'thing' as subject
and 'white' as predicate. Using a propositional variable instead
of the sentence'Somethingis white' we get the formula: 'If it is
I Alexander t4r. t Lpa 6t rai n)r toA &vayxaiou \taipeocv 6tc xai eirds oi}ev,
lv
6vtav'), fiv
oi batpot airoa renohlmcc, EeErilore 6cd rfis npoo9jxqs (scil.'toJtav'
$9doos fi6t1 xai dv rQ Ilepi tpplveias 8dEerxer,d.vots nepi rfis eis tdv piAAome xp6vov
)eyopivr1s dvre$doeas nepi tCw xafl' Ercaoroveip4pivav \ye"' 'i ptv oiv etvac rd 6v,
6tav fi, xai i pi1 6v pi1 elvoq 6tav pi1 fi, dvdyq'. i ydp i( ino|ioeos d.voyxatov
rooit6v iott.
2 Dc int. g, 18'39 ei yd.p dh102s eireiv 6rt lcuxdv 6tt oJ A<ux6v dotw, dvdyxrl
fi
etvat Aeuxdvi oi ievxdv.

{,l r

A N A R ISTOTEL TAN

PAR AD OX

r53

Ir rrc I lrat p, it is necessarythat p' .I do not know whether Aristotle

worrlrl lravc accepted this formula or not, but in any case it is
irrtcrcstingto draw some consequencesfrom it.
lrr two-valued logic any proposition is either true or false.
I lcrrt:r:thocxpression 'It is true thaty'' is equivalentto ,p,. Applyirrg tlris equivalence to our case we see that the formula ,If it is
Ir.rrr' th:rt1, it is necessary that p'would be equivalent to this
sirrrlrlt:r cxpression: 'If p, it is necessary that I' which reads in
syrrrlxrls: CpLp. We know, however, that this formula has been
rticr;tcd by Alexander, and certainly by Aristotle himself. It must
lrc rtjcctcd, for propositional modal logic would collapse, if it

1rlrrpositional modal logic.

llut it is possible to expressin symbolic form the idea implied
lry tlrr: formula 'If it is true thatp, it is necessarythat p': we need
rrrrfy rcplace the words 'It is true thatp'by the expression,a is
irrscrtcd'. These two expressionsdo not mean the same. We can
prrt lbrward for consideration not only true, but also false prolxrsitions without being in error. But it would be an error to aisert
ir l)roposition which was not true. It is therefore not sufficient to
s:ry '2 is true', if we want to impart the idea that p is really true;
,a
is
lt trrity bc false, and 'p is true' is,false with it. We must say
rrsscrtt:d'changing 'p' into 'u', as'p' being a substitution-variable
cirnrr<rtbc asserted, whereas 'u' riiray be interpreted as a true
lrrrrlxrsition. We can now state, not indeed a theorem, but a rule:
(x) u-, La.
4
lrr words: 'c, therefore it is necessarythat c'. The arrow means
'tlrcrcftrre', and the formula (r) is a rule of inference valid only
wlrcn rr is asserted. Such a rule restricted to .tautologous' pro'positions is accepted by some modern logicians.I
lirom rule (x) and the asserted principle of identity
Jxx there
lrrlfows thc asserted apodeictic formula LJxx which leads, as we
Ir;rvr:sr:cn,to awkward consequences.The rule seemsto be doubtlirl, r'vt:n if restricted to logical theorems or to analytic proposi||.

I Srr., c.g. G. H. von Wright, An

Essa2 in Modat Zogja, Amsterdam (lg5r),
t4 rr).

rs4

A RISTOTL E'S

M O DAL

LO G I C

O F PRO PO S I T I O N S

$ ++

tions. Without this restriction rule (r) would yield, as aPpears

from the example grven by Aristotle, apodeictic assertionsof
merely factual truths, a result contrary to intuition. For this
reason this Aristotelian principle fully deservesthe name of a
paradox.
$ 45. Contingenryin Aristotle
I have already mentioned that the Aristotelian term dvdeXciand sometimesin the
y.evovis ambiguous.In the De Interpretatione,
Prior Anal2tics,it meansthe sameas 6uvar6v,but sometimesit has
another more complicated meaning which following Sir David
Ross I shall translate by 'contingent'.I The merit of having
pointed out this ambiguity is due to A. Becker.2
Aristotle's definition of contingency runs thus: 'By "contingent" I mean that which is not necessaryand the supposed
existenceof which implies nothing impossible.'3We can see at
once that Alexander's definition of possibility results from Aristotle's definition of contingency by omissionof the words 'which
is not necessary'.If we add, therefore,the symbolsof thesewords
to our formula zB and denote the new functor by 'T', we get the
following definition:
46. QTqKNL4II qCCpqNLNq.
This definition can be abbreviated, asnqCCpqNLNqis equivalent
to NLNp. The implication:

s1j.cNtNpIIqccpqNLNq
has been already proved; the converseimplication
47. CnqCCpqNLNqNLNp
easily results from the thesis CIIqCCpqNLNqCCpqNLNqby the
substitution qIp, commutation, Cpp,and detachment. By putting
in 46 the simpler expressionNLNq for IIqCCpqNLNq we get:

aB.QrpKNLpNLNp.
This meansin words: 'It is contingent that p-if and only if-it
I W. D. Ross, loc. cit., p, eg6.
2 See A. Becker, Die Aristotelischc Tluorie der Miiglichluihschhisse,Berlin (rg33).
I agree with Sir David Ross (loc. cit., Preface) that Becker's book is 'very acute',
but I do not agree with Becker's conclusions.
t A n. p r . i . t 3 , 3 z o r 8 ) t i y a 6 ' e 6 $ g co 0 e txe itd d vie yd ye vo v,o ip i6 vrosdvoyxa!,ou,
rc|iwos 6' indpyew, oriD,ivioror 6rd zo0r' d.}ivarcv.

$ 4:,

C O N T I N GEN C Y

IN AR ISTOTL E

r55

is rrot ncct:ssarythat p and it is not necessarythat noty'.' As the

plrr:rsc 'not necessarythat notl' means the same as ,not impossifrlt' that f', w9 may say roughly speaking: ,something is ionlingcnt if and bnly if it is not tr.i.ssury-and not imfossible.'
Alr:xander shortly says: 'The contingent is neither neceisary nor
irnpossible."
Wc get another definition of Tp, if we transform NLNp acrrrrding to our definition r into Mp, and NLp jnto MNp:
or
4q. qTpKMNpMp
5o. QTpKMpMNp.
l'irrmula 5o reads: 'ft is contingent thatp-if
and only if-it is
lrossible that p and it is possible tJrat not 1.' This defines conting<:ncyas 'ambivalent possibility', i.e. as a possibility which can
irrtlccd be the case,but can also not be the case.We shall seethat
the consequences of this definition, together with other of
Aristotle's assertions about contingency, raise a new major
rl i l l i cul ty.
In a famous discussionabout future contingent events Aristotle
trics to defend the indeterministic point of view. He assumesthat
things which are not always in act have likewise the possibility of
lx:ing or not being. For instance, this gown may be cut into
picccs, and likewise it may not be cut.2 Similarly a sea-fight may
happcn tomorrow, and equally it may not happen. He says that
'()f two contradictory propositions about such things one must
lrr: true and the other false, but not this one or that one, only
whichever may chance (to,be fulfilled), one of them may be
rnorc true than the other, but neither ofthem is as yet true. or as
yt:t false.'r
'l'hese arguments, though not quite clearly expressed
or fully
thought out, contain an important and most fruitful idea. Let us
t:rkc the example of the sea-fight, and suppose thit nothing is
<lct:idcdtoday about this fight. I mean that there is nothing that
is rcal today and thatwould cause there to be asea-fight tomoirow,
n()r yet anything that would cause there not to be one. Hence, if
I Alexander
r5B. zo orire yd.p dvayxoCov oire ddtjvarov t6 dvieydpevov.
t I)c int.9, Iqng iozrv iv tois
p) d.ei ivepyoioc rd iuvordv etvat xal p) dpoius , . .
t2 otov 6rr rovti rd f.pd.rrcv }uvatdv ton Seerpq?fivaq , . , dpoiug Et,xci zri pr) brorp40itwt luvor6v.
| | lrir l. I q"1;6 to,huv yd.p (i.e.
dzi zois pr) d.ci oiow l pl d.ei pi oiocv) dvdynl
ptv
ll,i t t p,r' p,i1,nv rfis dmr$doeas d.A1|is ei"at
i $ei6os, oi pivtot zd6e f zd6e dM; 6n6rcp,
itt,yt, xtti p6),lov piv ,i\Ofi rlv dr(pav, oi pivrot
167 ,&\0fi fi rleudfi.

156 ARISTOTLE ' S

M O DAL

LO G I C O F PR O P O S I T I O N S

$ +s

truth restson conformity of thought with reality, the proposition

'The sea-fightwill happen tomorrow' is today neither true nor
false.It is in this sensethat I understandthe words 'not yet true
or false'in Aristotle.But this would lead to the conclusionthat it
is today neither necessary
nor impossiblethat there will be a seafight tomorrow; in other words, that the propositions 'It is
possiblethat there will be a sea-fight tomorrow' and 'It is
possiblethat there will not be a sea-fight tomorrow' are today
both true, and this future event is contingent.
It follows from the above that according to Aristotle there exist
true contingent propositions,i.e. that the formula Zp and its
equivalent KMpMNp are true for somevalue of p, say d. For example, if ct means 'There will be a sea-fighttomorrow', both
Mu and MJfa would be accepted by Aristotle as true, so that
he would have assertedthe conjunction:
(A) KMaMNa.
There exists, however, in the classicalcalculus of propositions
enlarged by the variable functor E, the following thesisdue to
Le6niewski'sprotothetic :
5t. C6pC6NpEq.
In words:'lf 6 of p, then if Eof notl, Eof 4', or roughly speaking:
'If something is true of the proposition p, and also true of the
negationof p,it is true of an arbitrary proposition4.' Thesis5l is
equivalent to
52. CK6pDNp6q
on the ground of the laws of importation and exportation
CCpCqrCKpqrand CCKpqrCpCqr.From (A) and 5z we get the
consequence:
52. 6lM, pl", qlpxC(A)-(B)
,r
(B) Mp.
Thus, if there is any contingent proposition that we accept as
true, we are bound to admit of any proposition whatever that
it is possible.But this would causea collapseof modal logic;
Mp must be rejected, and consequentlyKMaMNa cannot be ,
asserted.
We are at the end of our analysisof Aristotle's propositional

$41,

C ON TIN GEN C Y

IN AR ISTOTL E

t57

rnorl:rllogic. This analysishas led us to two major difficulties:

tlrr:first difficulty is connectedwith Aristotle'sacceptanceof true
irlxxlcictic propositions,the secondwith his acceptanceof true
<:ontingcntpropositions.Both difficultieswill reappear in Aristotlt:'s modal syllogistic, the first in his theory of syllogismswith
ont: assertoricand one apodeicticpremiss,the secondin his theory
ol'contingentsyllogisms.If we want to meet thesedifficultiesand
to r:xplainas well as to appreciatehis modal syllogistic,we must
lirst cstablisha secureand consequentsystemof modal logic.

r59

CHAPTER

VII

THE SYSTE M O F MOD A L

LOGIC

$ 46. The matrix method

Fon a full understanding of the system of modal logic expounded
in this chapter it is necessaryto be acquainted with the matrix
method. This method can be applied to all logical systems in
which truth-functions occur, i.e. functions wh6se truth-values

argument are put on the left, those of the second on the top,
and thg truth-values of C can be found in the square, *h.ii
the lines which we may imagine drawn from the -truth-values
on the margins of the square intersect one another. The matrix
of "M is easily comprehensible.
q

C
'{:

IO

N
o

II

IO

Mr

T H E M ATR IX M ETH OD
$+ 6
according to equalities stated in the matrix gives r as final resuJt,
the exprCssionis proved, but if not, it is disproved. For example,
CCpqCNpNq is disproved by Mr, since when ! : o and,q -: r,
*. nu't CCofiNoNr : CrCto: Cro: o. By contrast,CpCNpq,
one of "tour axioms of our C-N-p-system,' is proved by Ml,

because we have:
For p : I, 4 :
,, ! : r, 4 :
:
,, ! :o,4
,, ! : o, Q :

t: CrCNrr
o: CrCNrc
r : CoCNor
o:CoCNoo

:
:
:
:

CrCot :
CrCoo :
CoCr t :
CoCrc :

Ctr :
Ctr :
Cor :
Coo :

r,
r'
r'
r'

In the same way we can verify the other two axiotns of the
C-Jrfjrsyste m, CCp qCCqrCpr and CCNppp. As M r is so constructed
that the property of always yielding r is hereditary with rcspect
to the rules of substitution and detachment for asserted expressions, all assertedformulae of the C-M-P-systemcan be proved by
the matrix Ml. And as similarly the property of not always
yielding r is hereditary with respect to the rules of inference for
rejected expressions, all rejected formulae of the C-"1y'l-system
cin be disproved by Mr, ifp is axiomatically rejected. A matrix
which u.rifi.t all formulae of a system, i.e. proves the asserted
and disproves the rejected ones, is called 'adequate' fo1 thesystem. Mr is an adequate matrix of the classical calculus of
propositions.
Mr is not the only adequate ,matrix of the C-N-p-system. We
get another adequate matrix' M3, by 'multiplying' Ml by itself'
The processof getting M3 can be described as follows:
Fiist, we form ordered pairs of the values .r and o, viz.: Q, r),
new m at r ix.
Q,o), (o,i , (o,o) ;these ar e t he elem ent sof t he
Secondly, we determine the truth-values of C and "l/ by the
equalities:
: (Cac,Cbd),
0) C(o, b)(c, d)
:
(N a,Nb) .
(z) N (a,b)
Then we build up the matrix Me according to these equalities;
and finally we transform Mz into M3 by the abbreviations:
(r, | : r, (r,o) : z, ( o, I ) : 3, and ( o, o) : o'
t See p. 8o.

THE SYSTEM O F M O DAL

(r, r)
(r, o)
(o, r)
(o, q)

LO G I C

$+ 6

(t, t)

I23o

(r, r)
(t, r)

rr3 3

(t, t)

I2I2

I23o

(r, r)

IIII

Mz

M3

Symbol r in M3 again denotestruth, and o falsity. The new

symbolsz and 3 may be interpreted as further signsof truth and
falsity. This may be seenby identifying one of them, it doesnot

I I OO

NC

IOIO

IIOO

OI

IOIO

IIOO

oo

IIII

IIII

II

IOIO

IIII

IO

IIII

o
o

M4

M5

matter which, with r, and the otherwith o. Look atM4, where

2 : r, and3 : o. The second row of M4 is identical with its first
row, and the fourth row with its third; similarly the second
column of M4 is identical with its first column, and the fourth
column with its third. Cancelling the superfluous middle rows
and columns we get Ml. In the same *ay *e get Mr from M5
wh e re 2: oand3- r .
u four-valued matrix. By multiplying Mg by Mr we get
Yf.ir
an eight-valued matrix, byfurthermultiplication Uy M, asixteJnvalued matrix, and, in general, a ez-valued matrix. All these
matrices are adequate to the C-N-p-system, and continue-to be
adequate, if we extend the system by the introduction of variable
functors.
$ 47. The C-N-6-p-s7stem
We have already met two theseswith a variable functor 6: the

even to logicians.

q,t7

THE C-"'V-6-1-SYSTEM

r6r

'l'lrc introduction of variable functors into propositional logic

ix rlrrc to the Polish logician Lesniewski. By a modification of his
rrrl<'o['substitution for variable functors I was able to get simple
irrrrl r'lcgant proofs.r First, this rule must be explained.
I rlt:notc by 6 a variable functor ofone propositional argument,
urrtl I accept that 3P is a significant expression provided P is a
rigrrificant expression. Let us see what is the meaning of the
lirrrlllcst significant expression with a variable functor, i.e. 6p.
A variable is a single letter considered with respect to a range
ol valucs that may be substituted for it. To substitute means in
pr;rt:ticc to write instead of the variable one of its values, the same
v;rlrrt: for each occurrence of the same variable. In the C-N-p$yst(:m the range of values of propositional variables, such as p
.r' r/, consistsof all propositional expressions significant in the
flyst('m; besidesthese two constants may be introduced, r and o,
i.r'. rt constant true and a constant false proposition. What is the
r irrrgr:of values of the functorial variable 6 ?
It is obvious that for 6 we may substitute any value which gives
toliclhcr with p a significant expression of our system. Such are
rrot only constant functors of one propositional argument, as, e.g'
.M, lrrrt also complex expressions working like functors of one
;rfgrfmcnt, as Cq or CCNpp. By the substitution 6lCq we get from
li/ tf rt: cxpression Cq1t, and by \lCCNpp the expression CCNppp.
It is cvident, however, that this kind of substitution does not
crrvr:r all possible cases.We cannot get in this way either Cpq or
{|lt(:.Npq from 6p, because by no substitution for 3 can the p be
rtrrrovcd from its final position. Nevertheless there is no doubt
tlr:rt thc two last expressionsare as good substitutions of 6y', as
(,'t1f<>rCCNppp, since 6p, as I understand it, represents all sigrrilit'rrnt expressionswhich containl, including p and 6p itself.
I wrts able to overcome this difficulty by the following device
w'lrir'lr I shall first explain by examples. In order to get CPq ftom
l/ lry :r srrbstitution for 6 I write 6/C'q, and I perform the substitutiorr by dropping E and filling up the blank marked by.an
.r;rrrslr<)phc
by the argument of E, i.e. by p.ln the same way I get
lirrrrr 6/ thc cxpressionCpCNpq by the substitution \|C'CN'q.If
rrrrrlt' llurn one E occurs in an expression, as in C\pCSNpDq,and I
\\'.rnt to pcrform on this expression the substitution 6/C'r, I must
' Sce .l;rn Lukasiewicz, 'On Variable Functors of Propositional Arguments',
l'tunlinp oJ thc lloyal Irish Acadcm2,Dublin (tg5t)' 54 A z.

162

THE SYSTEM O F M O DAL

LO G I C

$ +Z

everywheredrop the E'sand write in their stead C'r filling up the

blanks by the respectiveargumentsof 6. I get thus from 6p-Cpr,
from 6"Mf-CNpr,from6q-Cqr, and from the whole expressionCCprCCNprCqr.From the same expression CEpC\NpDqthere
follows by the substitution E/C" the formula CCppCCNpNpCqq.
The substitution Ei' means that 6 should be omitted; by this
the principle of
substitution we get for instance from C6pC6NpDq
Duns Scotus CpCNpq.The substitution 6/E' is the 'identical'
substitutionand doesnot produce any change.Speakinggenerally, we get from an expressioncontaining E'sa new expressionby
a substitution for E, writing for E a significant expressionwith at
leastone blank, and filling up the blanks by the respectiveargumentsof the 6's.This is not a new rule of substitution,but merely
a descriptionhow the substitutionfor a variablefunctor shouldbe
performed.
The C-Jfi-fu-system can be built up on the single asserted
axiom known already to us:
5r. CEpCDNp\q,
to which the axiomaticallyrejectedexpression/shouldbe added
to yield all rejectedexpressions.
C. A. Meredith hasshown (in an
unpublished paper) that all assertedformulae of the C-N-Psystemmay be deducedfrom axiom 5r.t The rules of inference
are the usualrule of detachment,and the rulesof substitutionfor
propositional and functorial variables. To give an example how
theseruleswork I shall deducefrom axiom 5r the law of identity
Cpp. tompare this deduction with the proof of Cpp in the
C-N-p-system.z
5r. E/', ql px5Z

5s. CpCNpp
qlNpx CSS-S+
5t. 6lCpCNp',
5a. CCpCNpNpCpcNpNp
5r. 6/',qlNpx53

t C. A. Meredith has provJd in his paper 'On an Extended System of the Propositional Calculus', Proceedingsof the Royal Irish Academlt,Dublin (I95I), 54 A 3,
that the C-O-67-calculus, i.e. the calculus with C and O as primitive terms and
with functorial and propositional variables, may be completely built up from the
axiom CEEOEI.His method of proving completeness can be applied to the C-N-6-!system with CSpCENptqas axiom. In my paper on modal logic quoted P. I33, n. 2,
I deduce from axiom 5r the three asserted axioms of the C-"1\y'7-system, i.e.
CCpqCCqrCpr, Ccj,lppp, CpCNpq, and some important theses in which 6 occurs,
2 S ee p. B t.
among others the principle of exteruionality.

\ 't't

THE C-,IT-E-,-SYSTEM

r 63

,[t. OpCNpNp
c5. plcpcMpNpx c55-56
,t(;. ( t.NCpCNpNpNCpCNpNp
qlp x C54-C56-57
5r. 6/C",pICpCNpNp,
o!!.
!fl.
I slrould like to emphasize that the system based on axiom 5r is
rrrrrr'lr richer than the C-N-p-system. Among asserted conseI f rf('f rccscontaining 6 there are such logical law s as CCpqCCqpCBpD
q,
l,:h(,:lqC6p6q,C\CpqCpDq, all very important, but unknown to
,rlt nost all logicians. The first law, for instance, is the principle of
rxtcrrsionality, being equivalent to CQgqC\pDq, the second may
lrr'trrkcn as the sole axiom of the so-called'implicational'system,
tlrc tlrird as an axiom of the so-called 'positive' logic. All these
lirws <::rnbe verified by the matrix method according to a rule
gi vcn bcl ow .
Irr two-valued logic there exist four and only four different
firtrct<rrsof one argument, denoted here by V, S, N, and F (see
rrrrrl l i xM6).

VS NF

IIOO

IOIO

M6
lirr thc verificationof 8-expressions
the following practical rule
rlrrc in substanceto Lesniewskiis sufficient: Write for-6successively
tlrc lirnctors V, S, N, and d then drop,S,transform Vuinto Cpp,
;rrrrf/,)vinto NCpp.If you get in all casesa true C-Jf-formula,the
rxprtssion should be asserted,otherwiseit should be rejected.
l'lx;rrnJrlc:.
C6CpqC6p6q
must be asserted,
becausewe have:
()sCpqCSpSq: CCpcCpc,
CNCpqCNpNq,
{ : L',cpqcvpvq : ccppccppcpp, cFCpqcFpFq : cNCppcNCppNCpp.

(;{;ln1(l6p6q
must be rejected,for CCpqCNpNqis not a true C-Jfii,r'rrrrrla.
We seethus that all expressions
of the C-"AA-6-1-system
,rrc r':rsilyproved or disprovedby the matrix method.
\ ,1tt. E-l)efinitions
'l'lrc lirrrctor6 may be successfully
employedto expressdefiniti,rrs.'f 'lrr.:ruthorsof the PrincipiaMathematlca
expressdefinitions

164

THE SY STEM O F M O DAL

LO G I C

$ +8

by a special symbol consisting of the sign of equality':'that

connects the dcfnienswith the defniendum,and of the letters 'Df'
put after the definition. According to this method the definition
of alternation would run thus :

CNPq: gPq

Dl

andHpq ( 'eitherp
whereCNpq('If not1, then 4') is the defniens,

or 4') the defniendun.lThe symbol '.:. Df is associatedwith a

special rule of inference allowing the replacement of the defniens
by the defniendumand vice versa. This is the merit of this kind of
definition: the result is given immediately. But it has the defect of
increasing the number of primitive symbols as well as of rules
of inference which should be as small ag possible.
Lesniewski would write the same definition as an equivalence
thereby introducing into his system no new primitive term to
express definitions, because for this very purpose he chose
equivalence as the primitive term of his logic of propositions
enlarged by functorial variables and quantifiers, and called by
him 'protothetic'. This is the merit of his standpoint. On the
other hand he cannot immediately replace the defniens by the
defniendumor conversely, because equivalence has its own rules
which do permit such replacements.
In our C-Jf-57*ystem
equivalence is not a primitive term;
hence it must be defined, but cannot be defined by an equivalence
without a vicious circle. We shall see,however, that it is possible
to express definitions by C and 6 in a way which preserves the
merits of both standpoints without having their defects.
The purpose of a definition is to introduce a new term which as
a rule is an abbreviation of some complex expression consisting
of terms already known to us. Both parts of the definition, the
defniensas well as the defniendun must fulfil certain conditions in
order to yield a well-formed definition. The following four conditions are necessary and sufficient for definitions of new funcas well as the
tions introduced into 6ur system: (a) The definiens
defniendumshould be propositional expressions. (D) The defniens
should consist of primitive terms or of terms already defined by
them. (c) The defniendumshould contain the new term introduced
by the definition. (d) Any free variable occurring in the defniens
t I usually denote alternation by l,
meaning in my syllogistic.

but this letter has already got another

rr ,tl l

E -D E F IN ITION S

r65

sfrrrrrltl ttccur in the defniendum,and vice versa' It is easily seen

f lr:rt, r:.g. CNpq as defniensand,Hpq as defniendumcomply with the
lorrr :rlxlve conditions.
l,<'t rrs now denote by P and R two expressionsthat fulfil the
r rrrrrlitions(o)-(d),so that one of them, it does not mattcr which'
rrr;rylrt: t.t.r, at the defniens,and the other-asthe defniendum'ltis
that neither-of them contains 6' I say that the asserted
*,,1,1,,,r.,,1
,'*j,i'.,rriot CSPSRrepresents a definition' For instance:
,tll. C6CNI)q6HPq
thc definition of alternation' According to 58 any
r'('l)r'(:s(:nts
,'*'1,,.'rri.r. containing CNpq may be- immediately transformed
in which CNpq is re-placed by Hpq' As
irrt',r rrnother .*p..rril.
principle of Duns Scotus:
the
take
cxirrrrplcr". *uy
,,q. C)pCNpq,
then
lirrrrr wlrich we can get the law CpHpq,i'e' in words: 'Ifp,
citlrt'r'p or q', by the following deduction:

xC59-6o
58.6lCP'
{h. CpHpq.
Clavius:
tl'wt: want to apply our definition to the principle of

$r. oCllPPP,
wc rrtttstfirst put p fot q in 58 getting thus:

58.qlPx6z
ti'l. (:,\CNPP6HPP
62. 6lC'PxC6r-63
(;'t. (;llppp.
( l ' i rl rrrrrl :t(i ? states: ' If ei th er I
or axior
' pr i rrri ti vr'yl roposi ti ons'
They rigb
I'ttttrtltirtA'lallumatica.
nI l ,rrrt' l ogy' ,l rsi t statestha t t o
'lt rt lt', ts to s:ty simply '1'. The principle ot lJuns Dcotus' Ior
lrhl,rrr r', ts ll()r l tautology in any reasonablesense')
whigh enablesus
'l 1,, , ,,ttv.'rs,' inrplicatiln of 58 C6HpqDCN?q,
first' We can
the
together.with
given
t:Npqit
lty
f f f r'lll,ri r ll1u1
rules of substitutionand detachment
1,', ,"1',,,,,l,','.l,',,*i,tgtrnly thi
:
theorem
,
tl rr l r,l l ' \\' rtrl '11('
l l (' t' i tl

t66

TH E SYSTEM O F M O DAL

LOG I C

$ +8

(C) If P and R are any significant expressionsnot containing

E, and CEPDfiis asserted,then C6.RDP
must be aserted too.
The proof:
(D) CEP6R
(D) 6/CE'6Px(E)
(E) CC6P6PC6ft6P
(D) 6/CC6P6'CERDPx
(F)
(F) CCC5P6PC6ft6PCCEP6RC6^R6P

(F)xc(E)-c(D)-(G)
(G) C6ft6P.

If therefore P and R do not contain E, and one of them may be

interpreted as defniensand the other as defniendum,then it is clear
that any asserted expression of the form CEP6.R represents a
definition, as P may everywhere be replaced by R, and R by P,
and this is just the characteristic property of a definition.
$ ag. Thefour-ualued systemof modal logic
Every system of modal logic ought to include as a proper part
basic modal logic, i.e. ought to have among its theses both the
M-axioms CpMp, *CMpp, and *Mp, and the L-axioms CLpp,
*CpLp, and *NLp.It
is easily seen that both M and L are different from any of the four functors V, S, N, and F of the twovalued calculus. M cannot be V, for Mp is rejected-whereas
Vp : Cpp is asserted, it cannot be ,S, for CMpp is rejectedwhereas CSpp : Cp! is asserted, it cannot be either
"M or fl for
CpMp is asserted-whereas CpNp and CpFp : CpNCpp are rejected. The same is true for Z. The functors M and I have no
interpretation in two-valued logic. Hence any system of modal
logic must be many-valued.
There is yet another idea that leads to the same consequence.
If we accept with Aristotle that some future events, e.g. a seafight, are contingent, then a proposition about such events
enounced today can b'e neither true nor false, and therefore must
have a third truth-value different from r and o. On the basis of
this idea and by help of the matrix method with which I became
acquainted through Peirce and Schrcider I constructed in rgzo
a three-valued system of modal logic developed later in a paper
of l93o.I I see today that this system does not satisfy all our
!

Jan Lukasiewicz, 'O logice tr6jwarto6ciowej',

RuehFilozofuzn2, vol. v, Lw6w

SYSTEM OF M OD AL

t i 4 , r T H E F O U R - V A L U ED

L OGIC

t6 7

irrtrritions concerning modalities and should be replaced by the

$yst(:rndescribedbelow.
I rrrn of the opinion that in any modal logic the classicalcalculus
ol''ropositions should be preserved. Thil calculus has hitherto
rrlinifcisted solidity and usefulness,and should not be set aside
w i rfiou t weighty reasons. Fortunately enough the classical calculus
,,1'propositionshas not only a two-valued matrix, but also manyu,,i,,"d adequate matrices. I tried to apply to modal logic the
sirrrptcstmany-valued matrix adequate to the C-Jf-E-1-system,
i.r'. thc four-valued matrix, and succeeded in obtaining the
rk'sircd result.
As we have seen in $ 46, the matrix Mz whose elements are
prrirs of values r and o follows for "lVfrom the equality :
(z) N (a,b) :

(N a,Nb) .

'l'lrt: cxpression '(-1y'4,

general form
"MD)'is a particular caseof the
(,a, {6) where e and ( have as values the functors V, S, N, and F
,',1'thciwo-valued calculus. As each of the four values of e can be
t,rrnbined with each of the four values of (, we get 16 combinations. which define 16 functors of one argument of the fourvllrrccl calculus. I found among them two functors, either of
wf riclr may represent M. Here I shall define one of them, the
otht'r I shall discusslater.
(," ) M(a,b):

(S a,Vb) :

( a, Cbb) .

( )rr tlrc basisof (c) I got the m4trix M7 for M which I transformed
irrto the matrix MB by the same abbreviations as in $ 46, viz':
(,, r) ' - r, (I ,o) : z, ( o, r ) : 3, and ( o, o) : 6.
(r, , )
(t, o)
(o, ,)
(o, o)

( t, ,)
( t, t)
(o, t)
(o, t)

M7

MB

If :rvirrg tlrus got the matrix of M I chose C, N, and M as

( r ryro). .l;rn Lukasiewicz, 'Philosophische Bemerkungen zu mehrwertigen Systemen
,1,., ,\,rrs;rrl,.rrkalknls', ComplesRendusdzs Slanccsdc Ia Soci[E dcs Scicnecsct dcs ltttrcs
,h I unui c, vol . xxi i i , cl . 3 (I 93o).

168

T HE SY ST E M O F M O D AL L O GIC

$ +s
primitive terms, and based my system of modal logic on the
following four axioms:

5r. C6pC6NpDq 4.CpMp

"5. CMpp

*7. Mp.

The rules ofinference are the rules of substitution and detachment for assertedand rejected expressions.
Zy' is introduced by a E-definition:

64. C\NMNpBLp.
That means: 'NMNp' may be everywherereplaced by 'Lp', and
conversely'Lp'by'NMNp'.
The samesystemof modal logic can be establishedusing C,
"lf, and Z as primitive terms with the axioms:
*6. CpLp
*8. NLp,
5r. C6pC6Np6q g.CLpp
and the 8-definitionof M:
65. C6NLN/6M\.
M9 representsthe full adequatematrix of the system:
230
230

r33

2 r2

III

3
3

2
2
o
o

$,19 ' I' Il n FOU R -V A LUED SYSTEM O F M O DAL LO G I C

169

syst(:mmust L/admitted by anyone who accepts the axioms and

llrr: rulcs of inference. No serious objection can be maintained
rrr-{rrinst
this system. We shall see that this system refutes all false
infi'rcnccs drawn in connexion with modal logic, explains the
rli(Iicultics of the Aristotelian modal syllogistic, and reveals some
rrnr:xpcctedlogical facts which are of the greatest importance for
philosophy.
8 'ro. ,Macessitland thefour-ualued systemof modal logic
'I'wo major difficulties were stated at the end of Chapter VI :
thc first was connected with Aristotle's acceptance of asserted
:rpodeictic propositions, the secondwith his acceptance ofasserted
t:ontingent propositions. Let us solve the first difficulty.
ll'all analytic propositions are regarded as necessarily true,
thcn the most typical analytic proposition, the principle of
i<\<:ntityJxx, must also be regarded as necessarilytrue. This leads,
rrswe have seen,to the false consequencethat any two individuals
:rrc necessarilyidentical, if they are identical at all.
'I'his consequencecannot be derived from our system of modal
loeic, because it can be proved that in this system no apodeictic
proposition is true: As this proof is based on the law of extensionality CCpqCLpLq,we must first shor,vthat this law results from
our system.
A consequenceofaxiom 5r runs thus:

66. C6CpqC6pAq.
I hopd that after the explanations given above every reader will
be able to verify by this matrix any formula belonging to the
system, i.e. to prove assertedformulae, and to disprove rejected
ones.
It can be proved that the system is complete in the sensethat
every significant expression belonging to it is decidable, being
either asserted or rejected. It is also consistent, i.e. non-contradictory, in the sense fhat no significant expression is both
assertedand rejected. The set of axioms is independent.
I should like to emphasize that the axioms of the system are
perfectly evident. The axiom with 6 must be acknowledged by all
logicians who accept the classical calculus of propositions; the
axioms with M must also be accepted as true; the rules of inference are evident too. All correctly derived consequencesofthe

lirom 66 therefollowsby the substitution3lM'the formula:

67..CMCpqCMpMq,
;rrrclfrom 67 we get by CCpqMCpq,
a substitutionof axiom 4, and
lry tlre hypothetical syllogism the stronger M-law of extension:rlity:
te. CCpqCMpMq.
'f 'lrt: stronger L-law of extensionalityCCpqCLpZqis deducible
lirrm 19 by transposition.The problem left undecidedin $ 42,
whit:h interpretation of the Aristotelian laws of extensionality,
rlr(:strongeror the weakerone,shouldbe admitted,is thussolved
irr lavour of the stronger interpretation. The proof that no
:rporlcicticpropositionis true will now be given with full pre(:l sl on.

r70

THE SYSTE M O F M O DAL

LO G I C

$ So

N E C E S S I T Y A N D T H E FOU R - VAL U ED

SYSTEM

r7r

w(: !{ot LJxx : z, and CJx2CLJxxLJxT: CJryCzLJry. Jx)

The premisses:

*6. CpLp
rB. CCpqCLpLq
39. CCpCqrCqCpr
68. CCCpqrCqr.
The deduction:
68. rlCLpLqxCr8-69
69. CqCLpLq
n. plq, qllp, rlLqxC6g-7o
CLpCqLq
7o.
7o.!1", qlpxC*7r*6

*7t . L u .

The Greek variable ctrequires an explanation. The consequent of

7o, CqLq, which means the same as the rejected expression CpLp,
permits according to our rules the rejection of the antecedent Lp,
and any substitution ofZp. This, however, cannot be expressedby
*2r0, because from a rejected expression nothing can be got by
substitution; so, for instance, Mp is rejected, but MCpp-a
substitution of Mp-is
asserted. In order to express that the
antecedent of 7o is rejected for any argument of L, I employ Greek
letters calling them 'interpretation-variables' in opposition to
the 'substitution-variables' denoted by Latin letters. As the proposition d rr.ay be given any interpretation, *Zcr represents a
generdl law and means that any expression beginning with Z,
i.e, any apodeictic proposition, should be rejected.
This result, *Zcr, is confirmed by the matrix for Z which is
constructed from the matrices for "l/ and M according to the
definition of I. Anyone can recognize from a glance at M9 that
Z has only z and o as its truth-values, but never 1.
The problem of false consequencesresulting from the application of modal logic to the theory of identity is now easily solved.
As LJxx cannot be asserted,being an apodeictic proposition, it is
not possible to dcrive by detachment from the premiss:

Q) CJxyCLJxxLJry

$ 5,)

or

CLJxxCJx2LJx2

the consequence: (a) CJxltLJx2. It can be matrically proved

indeed that (l) must be asserted,giving always -r,but (u) should be
rejected. Since the principle of identity Jxx is true, i.e; Jxx : t,

rrury lrave one of

ll' ./x2 : r, then
,, .ifx2 : z, ,,
,, .7*) :3,
,,
,, .7*! : o, ,,

the four values, r, 2) 3, or

CJxTCzLJx2 : CfizLr :
CJx2CzLJxy : CzCzLz :
CJx2CzLJxy : C3CzLS :
CJx2CzLJx2 : CoCzI'o :

o:

CrCzz
CzCez
C3Czo
CoCzo

:
:
:
:

Crr
Czr
C33
CoJ

:
:
:
:

r,
r,
r,
r-

llr:ncc (t) is proved since the final result of its matrical reduction
is :rlw:rys r. On the contrary, (a) is disproved, because we have
lin' .'/xyt : r: CJxlLJxl : CtLt : Crz : z.
A pleasing and instructive example of the above difficulty has
lrccn given by W. V. Quine who asks what is wrong with the
li rllowing inference : I
(a) The Morning Star is necessarily identical with the
Morning Star;
(D) But the Evening Star is not necessarilyidentical with the
Morning Star (being merely identical with it in fact);
(c) But one and the same object cannot have contradictory
properties (cannot both be A and not be,4);
(d) Therefore the Morning Star and the Evening Star are
different objects.
(iivcn nry system the solution of this difficulty is very simple.
'l'lrc inlcrence is wrong, becausethe premisses(a) and (D) are not
I r rrc itncl cannot be asserted,so that the conclusion (/) cannot be
irrli'r'rr:clfrom (a) and (D) in spite of the fact that the implication
t;Qt){:(h)(d) is correct (the third premiss may be omitted being
tlrrr'). 'l'hc aforesaid implication can be proved in the followi rrg w :ty:
l,r'l r <l<:notethe Morning Star, andT the Evening Star; then
(,t) it l..'/xx, (r) ir NLl2x which is equivalent to NLJxy, as
irllrrtity is :t symmetriial relation, and (d) is NJx2. We get thus
f lrr li,r'rrrrrf:rOLJxxCNLJx2NJx2 which is a correct transformaItorr ol rl rt' tnrc thesi s(l ).
'l'lrr rxrrrnplt:given by
Q;rine can now be verified by our fourv' rfrrl rl trr;rl t' i xthus: i f ' x' ar 'd'7' have t he sam e m eaning as
,
f rt= frrrr'thcn
r; hence LJxx : LJr l:
Lr : 2,
.fxx : Jrl :
| | f,'rrrr,l tlrrr t'xlrn;>lc in the mimeographed, Logic.lfotes, 16o, edited by the
$
lh l"u lri rrt 'l l)lrilrsoplry of the Canterbury University College (Christchurch,
N /, ), rrr,l rrrt to rnc lry ProfessorA. N. Prior,

r72

THE SYS TEM O F M O DAL

LO G I C

$so

NLJxy : Nz : 3, and NJry : Nr : o, so that we have

according to CLJnxCNLJx2NJxT:CzC3o: Czz : r. The implication is true, but as not both its antecedentsare true, the conclusion may be false.
We shall seein the next chapterthat a similar difficulty was at
the bottom of a controversybetween Aristotle and his friends,
Theophrastusand Eudemus.The philosophicalimplications of
the important discovery that No apodeictic
proposition
is truewillbe
set forth in $ 62.
$ 5t. Twin possibilities
I mentionedi" $ +g that there are two functorseither of which
may representpossibility.One of them I denoted by M and
definedby the equality:
(a) M(a,b): (Sa,Vb): (a,Cbb),
the other I define by the equality:
(il W(a, b) : (Va,,Sr): (Caa,b),
denoting itby W which looks like an inverted M. According to
this definition the matrix of W is Mro, and can be abbreviated
to Mr l. Though I,1zis different from M it verifies axioms of the
same structure as M, becauseCpWp is proved by Mll, like
CpMp by MB, and *CWpp and *Wp are disprovedby Mrr, as
*CMpp and * Mp are by MB. I could have denoted the matrix of
Wby M.

Mro

Mrr

It can further be shown that the differencebetweenM and W

is not a real one, but merely resultsfrom a different notation. It
will be rememberedthat I got M3 from Mz by denotingthe pair
of values (r, o) by:, and (o, r) by 3. As this notation was quite
arbitrary, I could with equal justice denote Q, o) by 3, and
(o, r) by 2, or choose any other figures or signs. Let us then
exchangethe values z and 3 in Mg, writing everywhere3 for z,

r 73

TW IN PO SSI BI LI TI ES

\' ,r

rurrl l fior.3.We get from M9 the matrix Mtz, and by rearrangenr('nt ol- the middle rows and columns of Mtz, 'the matrix MI3.

r23o
I

r2 3o
rr3 3

2
z

o
o

I2I2

IIII

3
3

r23o

r32o

J-

r23o
II33

I2I2

o
o

IIII

r32o

I122

o
2

I3I3

IIII

3
o
3
o

Mr3

Mrz

I l'wc compare M9 with M r 3, we see that the matrices for C and

.M rt:main unchanged, but the matrices corresponding to M and
/, lrt:come different, so that I cannot denote them by M and L.
't'lrc matrix in Mr3 corresponding to.M in Mg is just the matrix
ol'tl/. NeverthelessMr3 is the same matrix as M9, merely written
irr ;rnother notation. I4l represents the same functor as M, and
rrrrrsthave the same properties as M. lf M denotespossibility, then
lll rlocs so too, and there can be no difference between these two
;rossibilities.
I n spite of their identity M and W behave differently when they
lxrttr occur in the same formula. They are like identical twins
wlro cannot be distinguished when met separately, but are
irrslantly recognized as two when seen together. To perceive this
fct rrs consider the expressionsMWp, WMp, MMp, and WWp,
It' M is identical wlth W, then those four expressions should be
irk'rrtical with each other too. But they are not identical. It can
lrr'proved by means of our matrices that the following formulae
;trc l tsscrted:
72. MW p

and

7Z. WM p,

li llrlt has as its truth-values only r or 2, and A[r as well as

lll t
r ; similarly Mp has as its truth-values only r or 3, and
lxth ll/r : r and W3 : r. On the other hand it can be proved
tl r:rt tl rt: l brmul ae:

t 74

THE SYSTEM O F M O DAL

74. CMMpMp

and

LO G I C

$sr

75. CWWpWp

are asserted, and as both Mp and Wp are rejected, MMp

WWp must be rejected too, so that we have:
*76. MMp
*77. WW!.
and

AN D TH E FOU R - VAL U ED

wc gct for ff the matrix

and

We cannot therefore, in 72 or 73, replace M by W or W by M,

becausewe should get a rejected formula from an assertedone.
The curious logical fact of twin possibilities (and of twin
necessitiesconnected with them), which hitherto has not been
observed by anybody, is another important discovery I owe to my
four-valued modal system. It is too subtle and requires too great
a development of formal logic to have been known to ancient
logicians. The existence of these twins will both account for
Aristotle's mistakes and difficulties in the theory of problematic
syllogisms, and justify his intuitive notions about contingency.
$ 52. Contingeru2and thefour-ualued systemof modal logic
We know already that the second major difficulty of Aristotle's
modal logic is connected with his supposing that some contingent
propositions were true. On the ground of the thesis:

52. CK6p6NpDq,
which is a transformationof our axiom 5r, we get the following
consequences:

52.6lM, plu, qlpxTB

78.CKMuMNaMp
7 8 'c* 79-*7
*79 . KMaMNa.
This means that 79 is rejected for any proposition cy,as d is here
an interpretation-variable. Consequently there exists no a that
would verify both of the propositions: 'It is possible that a' and
'It is possible that not a', i.e. there exists no true contingent proposition Tu, if Zp is deffned, with Aristotle, by the conjunction
of Mp and MNp, i.e. by:

Bo. CsKMpMNpDTp.
This result is confirmed by the matrix method. Accepting the
usual definition of Kpq:

u. CbNCpNqDKpq

g.;z CONTINGENCY

Forp:

Mt4,

r75

and we have:

tz KMpMNp:

,, f : 2:
r !: 3i
,, | : o:

SYSTEM

ss
,,
,,

KMrMNT : KtMo:

Kr3:3

: KMzMNz : KrM3: Kt3 : 3

: KM3MN3: K3Mz: K3r : 3
: KMoMNo: K3Mr : K3t : 3-

Wc see that the conjunction KMpMNp has the constant value 3,

rrrrd is therefore never true. Hence TP : 3, i.e. there exists no
t ruo contingent proposition in the sensegiven by definition Bo.
Aristotle, however, thinks that the propositions 'It is possible
th:rt there will be a sea-fight tomorrow' and 'It is possible that
tlrcre will not be a sea-fight tomorrow'may both be true today.
'['hus, according to his idea of contingency, there may be true
contingent propositions.
'fhcrc are two ways of avoiding this contradiction between
Aristotle's view and our system of modal logic: we must either
rlcny that any propositions are both contingent and true, or
rnodify the Aristotelian definition of contingency. I choose the
sr'<rondway, making use of the twin types of possibility discovered
:rbove.
'l'ossing a coin we may throw either a head or a tail; in other
words, it is possibleto throw a head, and it is possiblenot to throw
:r hcad. We are inclined to regard both propositions as true. But
tlrt:y cannot be both true, if the first'possible'is denoted by the
sarnc functor as the second. The first possibility is just the same
rrs tlrc sccond, but it does not follow that it should be denoted
irr tlrc same way. The possibility of throwing a head is different
li'orn thc possibility of not throwing a head. We may denote the
rrrrt'lry M, and the other by W. The proposition with the affirmativt' :rrgtrmcnt 'It is possible that p' may be translated by Mp, the
proposition with the negative argument'It is possiblethat not/'
lry ll'.N1t;or thc first by Wp, and the second by MNp. We get thus
tlv,r lirrrrtols ol'contingency, say X and I defined as follows :

||'t. t:tA'/lfltwMpSxp and

w. c\KwpMNfirp.

ll i:r irrrpossilrk:to translate thesedefinitions into words, as we

Ir,r\'r'rr,r n;rnr('slirr thc two kinds of possibility and contingency.
f ,r't rrs r,rlf tlrcrn 'M-possible' and 'W-possible', 'X-contingent'
.rrrrl ' l -r' orrti rrgt' rrt'.We m ay t hen r oughly say t hat 'p is X- con-

r76

T HE

SYST EM

OF M ODAL

L OGIC

$sz

means 'p is M-possible and Np is W-possible',and .r' is

lingent'
-2'-contingent'
means,p is W-possibleind Npis M_porriUte.'
From definitionsBz and 83 we can derive the matrices
ofx
and T. We get:
ForP:1i
Xt : 61e1r147*, :
Forp:21
Xz: KMzWNz:
ForP : j:
KM3WN3:
^X
F
oS
r p: : 6 ;
Xo : 6114o147yo:

KtWo:

Krz :

z; Tr :

KW:ITINT :

KrIVo :

Kry:

3.

KtW3 :

Kr r :

t; T z:

KWzM Nz:

K zM3:

K q:

o.

K3 Wz:

K3 z:

o; T3:

KW3 M N3:

K tMz:

K rt :

t.

KJW| :

K3r : 3; fo:

KWoMNo:

KzMt

Kzt :

z.

Irz (j ON TIN GE N CY AND THE FO UR_VALUED SYSTEM r 77,

tingt:nt. The negation of an X_contingent proposition
is a
?-lt:rntingent proposition, and .orrrr.rrJy th,
u
7'-..ntingent proposition is an X-contingent propoiition.
"lgutio"
"f
This
s'rrnds like a paradox, because we are accustomed
to think tJrat,
wlr^t is not-contingent is either impossibre or necessary,
r.tuii.rg
tlrr: impossible and the necessaryto the same kind
of iossibility.
lirt it is not true to say that, what is notx-contingent
is'eithe;/4irnpossibleor M-necessary; it should rather be
said that, what is
rrot X-contingent is
M-impossible or W_necessary,arrd
.either
th.t being either M-impossible
oi t4z_.r...rrary is equivui.rrt to
lrr.ing /-contingent.
'I'he same
misunderstanding lies at the bottom of the controvt:rsy about the thesis:

88. CKMpMqMKfq
whichis asserted
in our system.c. I. Lewisin someof hismodar

Mr5

84. XKWpWNp

and

(y) X! : rMp : Nrp'

systcmsacceptsthe formula:
TIq. CMKPqKMPMq,

85. f KMpMNp.

and

@) rp : XNp : NXp.

The lawsofcontradictionand of the excludedmiddreare

true for
Xp and Tp, i.e. we have:

86.NKXprp

and

87.Hxprp.

This means: no proposition can be both X-contingent

and f_con_
trngent, and any proposition is either X_contingent
or f-con_

lrrrt rejectsits converse,i.e. BB,by the following argument:r ,If

it
is pcrssible
thatp and 4 are both'true, thenp i-spo-ssible
and 4 is
p.ssible.This implication is not reversibte.po, .*u*pl.,
ii i,
;xrssiblethat the reader will seethis at once. It is also possibre
that he will not seeit at once.But it is not possiblethaihe
will
lxrth scc it at once and not seeit at once.'Th" pa.r,r^iu.rr.r,
of
llris argument is illusory. What meant by ,the reader,?
is
Iia;
i'<lividual reader,say l?, is meant, then R either will
seethis ai
on(:(',or R will not seethis at once.fn the first case
the first pre_
rtiss 'It is possiblethat R will seethis at once' is true;
but the
$r'r:orr<l
prcmiss is false, and how can a false proporiCo" l.
P.ssilrlytrrrc?In the secondcasethe secondp..*irri, ,*., U"i
lirst is falsc,and afalse proposition.urrrrotbe possibly
.tl.r.
i.ue.
'l'lrr' rw,
P*:missesof the formuia BBare not both irouuui., unJ
tlrc li rrrrrrrlrr
t::rnnotbe refutedin thiswav.
,rtr4.rirr
'thc reader'somereaderis meant, then the prelry
.ll
rtti*r's'I r is 1xrssiblc
that somereaderwill seethisat once,urrd,It
tr ;r,*ilrl. rlr:rrs.rnt:rcaderwiil not seethis at once'
may be both
It rrr', lrrrr irr rlris clrscthe concrusion'It is possibre
tirat some
| ( i l. f ,r.*,rr
lrrrl ( j. I I Langford, SymbolicLogic,New york and London
(r93r),

"''1"1;,

r78

THE SYST EM O F M O DAL

LO G I C
$sz
reader will see this at once and some reader will not see this at
once' is obviously also true. It is, of course, not the same reader
who will see this and not see this at once. The example given by
Lewis does not refute formula BB; on the contrary it supports its
correctness.
It seems, however, that this example has not been properly
chosen. By the addition of the words 'at once' the premisses have
lost the character of contingency. Saying that the reader will see
this, or not, 'at once', we refer to something which is decided at
the moment of seeing. The true contingent refcrs to undecided
events. Let us take the example with the coin which is of the same
sort as Aristotle's example with the sea-fight. Both examples concern events that are undecided at present, but will be decided in
the future. Hence the premisses 'It is possible to throw a head'
and 'It is possible not to throw a head' may at present be both
true, whereas the conclusion 'It is possible to throw a head and
not to throw a head' is never true. We know, however, that contingency cannot be defined by the conjunction of Mp and MNp,
but either by Mp and WNp or by Wp and MNp, so that the
example quoted above does not fall under the thesisBB. It cannot
therefore disprove it. This was not known to Lewis and the other
logicians, and on the basis of a wrong conception of contingency
they have rejected the discussedthesis.

| 59. Sonefurther poblcms

Although the axioms and the rules of inference of our fourvalued system of modal logic are perfectly evident, some consequences of the system may look paradoxical. We have already
met the paradoxical thesis tJrat the negation of a contingent
proposition is also contingent; as another thesisof this kind I may
quote the law of 'double contingency' according to which the
following formulae are true:

and ,o gr. QpTTp.

The problem is to find some interpretation of these formulae
which will be intuitively satisfactoryand will explain away their
apparent oddness.When the classicalcalculus of propositions
was only recentlyknown there was heatedoppositionto someof
its principles too, chiefly to CpCqpand CpCNpq,which embody
two logical laws known to medievallogiciansand formulatedby
9o. QpXXp

$:,:I

SOME FURTHER

PR OBL EM S

t79

tlrcm in the words: Verumsequiturad quodlibetand Ad falsum

scquiturquodlibet.
So far as I see,theseprinciples are now univcrsallyacknowledged.
At any rate our modal systemis not in a worsepositionin this
rospectthan other systemsof modal logic. Someof them contain
suchnon-intuitive formulae, as:
+sz.
QMNMpNMp
wherea problematicproposition'It is possiblethatl is impossible'
is cquivalentto an apodeicticproposition'It is impossiblethatp'.
Instead of this odd formula which has to be rejected we have in
our systemthe thesis:
y. QMNM1MNq which together with

s+. QMMpMp

r:nablesus to reduceall combinationsof modal functorsconsisting

of M and "lifto four irreducible combinationsknown to Aristotle,
viz. M : possible, NM : impossible, MN : non-necessary,
lncl NMN:
necessary.
'l'he secondproblem concernsthe extensionof the four-valued
modal logic into higher systems.The eight-valuedsystemmay
s(:rveas an example.We get the matrix Mr6 of this systemby
multiplying the matrix Mg by the matrix Mr. As elementsof the
rrcw matrix we form the pairs of values: Q, r) : r, (r,o) : z,
:5, $,o):6,(o,r):7,(o,o):
(r, r) : 3,Q,o):4i(3,I)
o,
irrrd thcn we determine the truth-values of C, N, and M accordirrg to the equalities(i, k), and'(cr).
M
C r2S456Zo

rz J 4 5 6 Z o
rr3 3 5 5 77
3 r2 r2 S 6 S6
4 rrrr5 5 5 5
r2 3 4 r2 34
6 rr3 3 rr3 3
I

7
o

5
4
3
q

I2I2I2I2
IIIIIIII

3
3
5
5
7

Mr6
lrigrn'rr rlcnotcs,asusually,truth;o falsity;and the otherfigures
iu'e irrtclrrrcrliirtt:values between truth and falsitv. If we

t 8o

THE

SYST EM

OF

M ODAL

L OGIC

53

attentively consider the matrix Mr6 we shall find that the second
row of C isidenticalwith the column of M. This row consequently
representsthe matrix of possibility. In the same way all the other
rows of C, except the first and the last, represent some kinds of
possibility. If we denote them by Mrto Mr, we can state that
Mrfor z ( i ( 7 satisfiesall the axioms of possibility, viz.

gS.CpMt?,

*96. CMnpp,

*g7. M#.

Among these different kinds of possibility there are some

'stronger' and 'weaker'; becausewe have, for instance, CMzpM4p
or CMrpMup,but not conversely. We may say therefore that in
eight-valued modal logic there exist possibilities of different
degrees.f have always thought that only two modal systems are
of possible philosophic and scientific importance: the simplest
modal system, in which possibility is regarded as havin[ no
degrees at all, that is our four-valued modal system,and the N,valued system in which there exist infinitely many degrees of
possibility. It would be interesring to investigate this problem
further, as we may find here a link between modal logic and the
theory of probability.

CHAPTER

A RI S T O T L E ' S

VIII

MOD AL

SYL L OGIST IC

Anrsrorrn's modal syllogistic has, in my opinion, lessimportance

irr r:omparison with his assertoric syllogistic or his contributions
to propositional modal logic. This system looks like a logical
cxcrcisc which in spite of its seeming subtlety is full of careless
rrrist:rkesand does not have any useful application to scientific
prrrlllcms. Nevertheless two controversial questions of this syllogistic are worth studying, chiefly for historical reasons: the
rlrrcstion of syllogisms with one assertoric and one apodeictic
prt:rniss,and the question ofsyllogisms with contingent premisses.
\ ,,4. Moods with two apodeicticpremisses
Aristotle deals with modal syllogisms after the pattern of his
;rssr:rtoricsyllogistic. The syllogisms are divided into figures and
rrurods,some moods are accepted as perfect and these need no
proof as being selfevident, the imperfect moods are prtlved by
corrvcrsion, reductioad absurdum,or by 'ecthesis', as it is called.
'l'lrr: invalid moods are rejected by interpretation through con( l'('tc terms. It is strange that with one exception Aristotle makes
n() usc of his theorems of propositional modal logic. We shall see
tlr;rt this would yield in several casesbetter and simpler proofs
llrrrrrthosegiven by him.
'l'lrr: laws of conversion for 'apodeictic propositions are anaLrgorrs to those for assertoric ones. The following theses are
;rt'r:orclinglytrue: 'If it is necessarythat no D should be an a, it is
rf('( ('ss:rrythat no a should be a b', in symbols :
\tl. OLIibaLEab,
rrrrrl ' l l ' i t i s nccessarythat e ver y D or som e D should be an a, it is
f rr'{('$s,u'ytlrlt some a should be a b', in symbols:
ry1. (;l,tlhul,lab
and
rco. CLIbaLIab.l
'l'lrr'

lrror'lsgivcn by Aristotle are not satisfactory.2He did not see

| ,ln
l) t,y tl
l l l l t,r
( ll
'

,r'iu2() ei piv ydp dvdyxrl ti A rQ B pqievi JndpTev, dvdyxq xai i

ln I t,
ytlh,ri i'n,i1yeo. -32
ci Di i( ivdyxls rd A ncvrii
nvi tQ B Jtdplgc, xoi
J t
,4 ,i t,i 1,x '1 i nr i py ew .
'pt
r \ l L,r hr r , l r x . c i t., p. go.

t 82

ARISTOT LE' S

M O DAL

SYLLO G I STI C

$s+

that the laws 98-roo may be immediately deduced from the

analogous laws of the assertoric syllogistic by means of the
theorem:
tB. CCpqCLpLq.
For instance,from r B, by putting Eba for p and Eab for q, we get
the assertoriclaw of conversionin the antecedent,hence we can
detach the consequent,i.e. law gB.
Syllogisms with two apodeictic premissesare, according to
Aristotle, identical with assertoricsyllogisms,except that the sign
of necessitymust be added to the premisses
aswell as to the conclusion.ITheformula for the moodBarbarawill accordinglyrun:
ror. CKLAbaLAcbLAca.
Aristotle tacitly acceptsthat the moodsof the first figure are perfect and need not be proved. The moods of the other figures,
which are imperfect, should be proved according to the proofs
of assertoricsyllogismsexcept Baroco and Bocardo, which are
proved in the assertoricsyllogistic by reductioad absurdum,and
should here be proved by ecthesis.zOnce again, for all these
proofsit would be easierto use theorem rB, as will appear from
the following example.
By meansof the laws of exportationand importation, CCKpqrCpCqrand CCpCqrCKpqr,
it can be shown that 15, the assertoric
mood Barbara, is equivalentto the formula:
tot CAbaCAcbAca.
This purely implicational form is more convenient for deriving
consequences
than the conjunctional form. According to the
thesis3 CLppwehave:
to3. CLAbaAba,
and from lo3 and ro2 we get by the hypotheticalsyllogism:
rc4. CLAbaCAcbAca.,',
On the other hand we have as substitutionof lB:
I An. pr, i. B, 29b35 ini piv oiv fiv dvayxalov oTe}dv ipotos iler xoi ini niv
JaapTivruv' <ioaltas yd.pn|epivav tdv 6pav Ev te rQ Jrd.pTev xai t,i i( dvdyxqs
indpycw l pl Jndppw Eotor te xai oix Eorat ou\Aoyopds, nAip Stoioct tQ npooxeio0ae rois 6poc td d! ,ivtiyxtls Sndpyew i pi1 i'ndpyu.
2 lbid. 3 o " 3 - r 4 .

$:,4

M O O D S WI T H

T WO APOD EIC TIC

PR EM ISSES

r83

t o5. CCAcbAcaCLAcbLAca,
irn<llrom ro4 and ro5 there follows the consequence:
ut6. CLAbaCLAcbLAca,
which is equivalent to ror. All the other syllogistic moods with
two apodeictic premissescan be proved in the same way without
rrr:wnxioms, laws ofconversion, reductioad absurdum,or arguments
lry ccthesis.
\ ,r,'. Moods with oneapodeiaic and oneassertoricpremissl
Syllogistic moods of the first figure with one apodeictic and one
rrsscrtoric premiss are treated by Aristotle differently according
to which premiss, the major or the minor, is apodeictic. He says
that when the major is apodeictic and the minor assertoric we
l{(:t :rn apodcictic conclusion, but when the minor is apodeictic and
thr: major assertoric we can have only an assertoric conclusion.2
'l'lris difl'crence will be made clear by the following examples
of
t lrr:mood Barbara. Aristotle assertsthe syllogism : 'If it is necessary
tlrrrt cvcry D should be an a, then if every c is a b, it is necessary
tlrrrt cvcry c should be an a.' He rejects, however, the syllogism:
'f f't:vcry D is an a, then if it is necessarythat every c should be a b,
it is ncccssary that every r should be an a.' In symbols:
(e) CLAbaCAcbLAca
(() CAbaCLAcbLAca

is asserted,
is rejected.

Aristotle considers the syllogism (e) as self-evident. He says:

'Sirr<:r'cvcry b is necessarily an a or not an a,, and c is one of the
/'s, it is cvident (Sav<pdv)that c too will be necessarily an a or
n()t ;rn a.'r For reasons that will be explained later it is difficult
lo slrow this by examples. But the following picture will perhaps
rrr;rkr' tlrr: syllogism (e) more acceptable to intuition. Let us
| ( ll,
,1. l,rrkasit:wicz, 'On a Controversial Problem of Aristotle's Modal Syllogi ;tr', l turi ni nn.l tudi es, vol . vi i (r 954), pp. rr4-z B .

t84

ARISTOTL E' S

M O DAL

SYLLO G I STI C

$ss

imagine that the"expression LAba means: 'Every D is connected

by a wire with an a.' Hence it is evident that also every c (since
every c is a 6) is connected by a wire with an a, i.e. LAca. For
whatever is true in some way of every 6, is also true in the same
way of every c, if every c is a D.The evidence of the last proposition
is beyond any doubt.
We know, however, from Alexander that the evidence of the
syllogism (e) which Aristotle asserted, was not convincing enough
for his friends who were pupils of Theophrastus and Eudemus.r
As opposed to Aristotle, they held the doctrine that if either premiss is assertoric the conclusion must be so, just as if either premiss is negative the conclusion must be so and if either premiss is
particular the conclusion must be so, according to a general rule
formulated later by the scholastics: Peioremsequitur semperconclusio partem.
This argument can be easily refuted. The syllogism (e) is
deductively equivalent to the problematic mood Bocardo of the
third figure: 'If it is possible that some c should not be an d, then
if every c is a b, it is possible that some l should not be an a.' ln
symbols:
Q) CMOcaCAcbMOba.
Syllogism (r7)is as evident as (e). Its evidence can be illustrated
by examples. Let us supposethat a box contains ballots numbered
from r to 9o, and let c mean 'number drawn from the box', b
'everl number drawn from the box', and a 'number divisible by
3'. We assumethat in a certain casefive even numbers have been
drawn from the box, so that the premiss: 'Every number drawn
from the box is an even number drawn from the box', i.e. Acb, is
factually true. From this we can safely infer that, if it is possible
in our case that some number drawn from the box should not be
divisible by 3, i.e. MOca, it is also possible in our case that some
eaennumber drawn frgm the box should not be divisible by 3,
i.e. MOba.
Aristotle accepts the syllogism (r7) and proves it by a reductio
I Commenting on the passagequoted in n, z, p, r83, Alexander says tz4. B o{nos
ltiv oitas Alyet, oi 3i ye itatpot tJroi oi *pi Ei61p6, te xoi 6e6$pootov oix oJras
A(yovor, dAAd$oov iv ndoats rais i( dvayxa(,agre xti itapyotiotls ovluyiats, idv
,ioc ovyxeipevat ovAAoytntxds, indpyov yivco|at rd ovpntpaopa . . . t7 tQ E\anov
elvat td irdpyov roi &vayxaiou.

$s:,

A P OD E IC TIC A ND ASSERTO RI C PREM I SSES

r 8s

ad absurdumfrom the syllogism (e).I He does not, however, deduce

(e) from ('7), though he certainly knew that this could be done.
Alcxander saw this point and explicitly proves (e) from ('r) by
"
rcductioad absurdumsaying that this argument should be held as
tlrc soundestproof in favour ofAristotle's doctrine.2 As according
to him Aristotle's friends accept the syllogism (21)which fulfils
p<:ioremrule, and (e) is deducible from (a), they cannot reject (e)
on thc ground of this rule, which becomes false when applied to
rnodalities.
We shall see in the next Section that there was yet another
argument raised by Theophrastus and Eudemus against syllogism
(e) which could not be refuted by Alexander, as it stands or falls
with an Aristotelian argument. In spite ofAlexander's talk about
tlrc 'soundest proof 'one feels that some doubt is left in his mind,
ftrr hc finally remarks after having presented several arguments
in support of Aristotle's opinion, of which the argument quoted
tbovc is the last, that he has shown with greater rigour in other
works which of those arguments are sound and which are not.3
Alcxander is referring here to his work 'On the Disagreement
concerning Mixed Moods between Aristotle and his Friends',
:rnrl to his 'Logical Scholia'.+ (Jnfortunately both works are lost.
Our times have seen a revival of this controversy. Sir David
I{oss, commenting on syllogism (<) and its proof from syllogism
(ri), statesdecidedly:s 'Yet Aristotle's doctrine is plainly wrong.
l'irr what he is seeking to show is that the premissesprove not only
lhat all C is A, but also that'it is necessarily A, just as all B is
I An. pr. i. z I,
yd.p i ptv B navri rit T, i 6i A iv6cydo0a wi
39b33-39 inapxiro
rQ I' pil in,ipycw'
6l td A iv6lyeo0at nvi rQ B pi in,ipyew. ei yd.p nawi rQ
'ivtiyxq
Il i A inip16c i( d.vdyxqs, rd 3i B novi rQ f xeher ind.pyeu, td A novri r.i f i(
,iliyxls
yd.p 6l6errrcr
npirepov.
d\l'
it&etto
twi dv6iyeo0oc pi1
indp{et'
roito
,,lrrip1q111,.
6t
' Afrx:rnder says, commenting on syllogism (e), tz7. g Zorr 62 nntcitoao|a4
r'l lty,ilrrruiv itd Aprcror(Aous riTtis iotr, pd.Ante Drti rfs eig ,i}rirotov d.ndyayfis tfis
ytt,,y,lr',1s ,'v rphq olqfpote . . . tz iv ydp tfi tonrhV ou{vyig tfi i, rptnp olgfipan xai
)lpor,,riMt
boxri rci rois dreipocs oitoA ini pipos dv\eT6poov d.no$anxdv yiveo|at
1'l s uy n/y r o1tu.
I Afcx;rrrrlcr tz7. t4 toooJro$
xai rorcdrors dy zrs yprioano tepeord.pcvos rfi
ntt,l t,t'rttr )lprcrotilous
6,!'ft1, 4 6i roirov
i"y6s i pi i1/.As Aiyeo|u Eoxei dr
rl\l,rr1 ri1r,',, ,is i,l,r1r, perd d.xptpei,oseipqrau
' I f re f rtle rrl tlrc first work reads (Alexander rz5. 3o): Ilepiofis xaad.tds pl(ec
,1,,,,/,,,1,,r
r if 1r,, r , ' r .',lous re xoi ritv dratpav ariroO. Cf. Alexander 249. 38-25o. e, whqre
8r,r,y'r,r'r',rrrr rrrr rl irrstr';r<lof Erc{opds, and the other work is cited as 27$trn Aoyuri.
" W . l r . l ( or r , k x . r i t., p, 43.

ARISTOTLE'SMODAL SYLLOGISTIC
$sS
necessarilyA, i.e. by a permanent necessityof its own nature;
while what they do show is only that so long as all C is B, it is A,
not by a permanentnecessityof its own nature, but by a temporary necessityarising from its temporary sharing in the nature
of B.'
This argument is a metaphysicalone, as the terms 'nature of
a thing' and 'permanentnecessityof its nature' belong to metaphysics. But behind this metaphysical terminology a logical
problem is hidden which can be solvedby our four-valuedmodal
logic. Let us now turn to the syllogismrejectedby Aristotle.
186

premiss
andlne assertoric
moodswith oneapodefutic
$ 56. Rejected
Syllogism(() is as evident as syllogism(e). It is strangethat
Aristotle rejectsthe syllogism
(() CAbaCLAcbLAca,
though it is clear that this syllogismis on the samefooting as the
assertedsyllogism(e). In order to showits evidencelet us employ
the samepicture asbefore. If LAcbmeansthat everyc is connected
by a wire with a D,and every Dis an a, i.e. Aba,it is evidentthat
everyc is connectedby a wire with an a, i.e. LAca. Speaking
generally, if every 6 is an a, then if every c is connectedwith a D
in any way whatever,it must be connectedwith an a injust the
sameway. This seemsto be obvious.
The"most convincing argument that syllogism (() is sound
results from its deductive equivalence with the problematic
mood Baroco of the secondfigure :
(0) CAbaCMOcaMOcb,in words:
'If every D is an a, then if it is possiblethat somet shouldnot be
an a,it is possiblethat somec shouldnot be a 6.'This can be
illustrated by an example.Let us turn to our box from which five
numbers have been dra'fon, and let us supposethat every even
number drawn from the box (6) is divisible by Z @), i.e. Aba.
From this factual truth we can safely infer that, if it is possible
that somenumber drawn from the box (c) should not be divisible
by 3, i.e. MOca, it is also possiblethat somenumber drawn from
the box should not be aneaennumber, i.e. MOcb.This syllogism
seemsto be perfectly evident. In spite of its seemingso Aristotle

$16

R E JEC TED M OOD S

r87

<lisprovessyllogism ((), fint by a purely logical argument which

will be considered later, and then by the following example:
f ,ct c mean 'man', b 'animal', and a 'being in movement'. He
.rcceptsthat the proposition 'Every man is an animal' is necessrrrily true, i.e. LAcb; but it is not necessarythat every animal
should be in movement, this may be only accepted as a factual
trtrth, i.e. Aba, and so it is not ncessarythat every man should
bc in movement, i.e. LAca is not true.r
Aristotle's example is not convincing enough, as we cannot
admit as a factual truth that every animal is in movement. A
bctter example is provided by our box. Let c mean 'number
drawn from the box and divisible by 4', b 'even number drawn
lrom the box', and a 'divisible by 3'. Aristotle would agree that
the proposition 'Every number drawn from the box and divisible
by 4 is an even number drawn from the box' is a necessary truth,
i.c. LAcb, while the premiss 'Every even number drawn from the
box is divisible by 3' can be only accepted as a factual truth, i.e.
Aba, and the conclusion 'Every numher drawn from the box and
divisible by 4 is divisible by 3' is also only a factual truth, i.e.
Aca, and not LAca. The 'nature' of a number drawn from the box
:rnd divisible by 4 does not involve any 'permanent necessity'for
it to be divisible by 3.
It would seem, therefore, that Aristotle is right in rejecting
syllogism ((). The matter, however, becomes complicated, for it
can be shown that just the same argument can be raised against

syllogism
(e) CLAbaCAcbLAca.
'fhis was seen by Theophrastus and Eudemus who refute (e)
using in another order the same terms which were applied by
Aristotle for disproving ((). Let b mean 'man', d-'animal', and
r-'being in movement'. They agree with Aristotle that the proposition 'Every man is an animal' is necessarily true, i.e. LAba,
ltnd they accept as factually true that 'Everything in movement
is :r man', i.e. Acb. The premissesof (e) are thus verified, but it is
obvious that the conclusion 'Everything in movement is an
:rnimal', i.e. Aca, is not necessarily true.z This example is as
I An.
ltr. i. g, 3o'28 En xai tx r6v 6pov Savepdv 6tc oix Eotan rd opnipaopa
,itayxatov, otov ei td piv A eitl xivqots, rd 6i B (Qov, d{'<! 6a rri I dv|panos' {Qov
ylv ydp 6 dv|panos i{ d.wiyxqs doti, xwettan 6i td {,iov oir 2( dvdyxls, oJ6' 6 dv0punos.
I Alt:xandcr
l z4.z I ril)ri xai ?zi rfs rills EewvJouor toAro EXovoinas , . . z4 td yi.p

rBB

A RI S T OT L E ' S MOD AL SY L L O GIS TIC

$S 6
unconvincing as the corresponding one in Aristotle, for we cannot admit that the premiss Acb is factually true.
We can give a better example from our box. Let 6 mean 'number divisible by 6', 4-'nqslbs1 divisible by 3', 4nd 6-'sysn
number drawn from the box'. Aristotle would accept that the
proposition 'Every number divisible by 6 is divisible by 3' is
necessarilytrue, i.e. LAba,but it can be only factually true that
'Every even number drawn from the box is divisible by 6', i.e.
Acb, and so it is only factually true that 'Every even number
drawn from the box is divisible by 3', i.e. Aca. The propositions
Acb and Aca are clearly equivalent to each other, and if one of
them is only factually true, then the other cannot be necessarily
true.
The controversy between Aristotle and Theophrastus about
moods with one apodeictic and one assertoric premiss has led us
to a paradoxical situation: there are apparently equally strong
arguments for and against the syllogisms (e) and ((). The controversy shown by the example of the mood Barbara can be
extended to all other moods of this kind. This points to an error
that lurks in the very foundations of modal logic, and has its
source in a false conception ofnecessity.
$ 57. Solutionof the controuers2
The paradoxical situation expounded above is quite analogous
to the.difficulties we have met in the application of modal logic
to the theory of identity. On the one hand, the syllogisms in
question are not only self-evident, but can be demonstrated in
our systemof modal logic. I give here a full proof of the syllogisms
(e) and ({) based among others on the stronger L-law of extensionality known to Aristotle.
The premisses:

3.c L p p

fi. CCpqCLpLq
24. CCpqCCqrCpr
gg. CCpCqrCqCpr
toz. CAbaCAcbAca.
(Qov novti d.v|p<iny i(
(Qov rovrL xcvovp(v<p i(

,ivriyxrls, 6 dv9ponos
dvdyxrls.

SOLUTION
5s7
'I'hc deduction:

OF TH E C ON TR OVER SY

r89

rB . pl A ba, qfAcaxr oT
rc7. CCAbaAcaCLAbaLAca
gg. plAba, qfAcb, rf Acax Ctoz-ro8
rc9. CAcbCAbaAca
24. pl Acb, qlCAbaAca,rlCLAbaLAcax CroB-Cro7-ro9
rcg. CAcbCLAbaLAca
33. plAcb, qlLAba, rlLAcax Crog-r ro
(.)
rrc. CLAbaCAcbLAca
rB . pl A cb, ql Acaxt r r
t r r. CCAcbAcaCLAcbLAca
24. pl A ba, ql CAcbAca,r lCLAcbLAcaxCr oz- Cr r r - r r 2
(().
tlz. CAbaCLAcbLAca
We see that the syllogisms (e) and (() denoted here by r ro and
r12) are assertedexpressionsof our modal logic.
On the other hand, we get the thesis r 13 from r ro by the substitution bfa, and the thesis r14 from rrz by the substitution blc
and commutation of the antecedents:
try. CLAaaCAcaLAca

rt4. CLAccCAcaLAca.

Both theseshave in the consequent the expression CAcaLAca,i.e.

the proposition 'If every c is an a, then it is necessarythat every c
should be an a'. If this proposition were asserted, all true universally-affirmative propositions would be necessarilytrue which
is contrary to intuition. Moreover, as CAcaLAcais equivalent to
CNLAcaNAca, and Aca rneans the same as NOca, we should have
C.NLNOcaNltrOca or CMOcaOca. This last proposition which
mcans 'If it is possible that some c should not be an a, then some
c is not an a' is not true, for it is certainly possible that a number
drawn from the box should not be even ; so that, if the proposition
is true, every set of drawings would contain an odd number:r rcsult plainly contrary to the facts.
The expression CAcaLAca must be therefore rejected, and
w c get:
* rr5. C A caLA ca,

navrl

xwoup(va

inepyira'

oJx,irc rd

from which there follows according to our rules for rejected

t:xJrrcssionsthe consequence:

r9o

ARTSTO TLE' S

M O DAL

SYLLO G I ST I C

$sz

rr3. xC*rrG-*rr5
*tr6. LAaa.
The apodeicticAristotelian law of identity must be rejected
like the apodeicticprinciple ofidentity LJxx.This is conformable
to our general view according to which no apodeictic proposition
is true. The consequentof t t 3, i.e. CAcaLAca,
cannotbe detached,
and the incompatibilitybetweenthe acceptanceof true apodeictic
propositions and the assertion of the stronger L-law of extensionalityis solvedin favour of the law of extensionality.I do not
believe that any other systemof modal logic could satisfactorily
solvethis ancient controversy.
I mentioned earlier that Aristotle tries to refute the syllogism
(() not only by examples,but alsoby a purely logical argument.
Asserting that the premissesAba and LAcb do not give an apodeictic conclusionhe says: 'If the conclusionwere necessary,
there would follow from it by a syllogismof the first or the third
figure that someD is necessarilyan a; but this is false,because
b may be such that possiblyno Dis an o." Aristotle refershere to
the apodeicticmoodsDarii and Darapti, sincefrom (() combined
with either of these moods we can derive the consequence
CAbaCLAcbLIba.The proof from Darapti runs :
r 17. CCpCqrCCrCqsCpCqs
rlz. CAbaCLAcbLAca (0
ttB. CLAcaCLAcbLIba (Darapti)
o
,r7. plAba,qlLAcb,rlLAia, slLlbaxCrn-CtrB-I rg
try. CAbaCLAcbLIba.
The proof from Darii gives the same consequence,but is more
complicated. Aristotle seemsto disregard the premissLAcb, and
interpretsthis consequence
as a simple implication:
*rzo. CAbaLIba,
which is obviously false,and must be rejected.Or perhaps he
thought that LAcb could be made true.by a suitable substitution
for c and dropped. If so he was wrong and his proof is a failure.
We seebesidesby this example how difficult it is to confirm the
validity of suchtheses,as I Ig, I I2, or r to, through termsyielding
I An. pr. i. g,
3ooz5 (continuation ofn. z, p. rB3) ei Trip Eott, oupBfioerar rd A
nvi rQ B indpyecv d( d,vd.yxt1s
6rd re toa rp<itrou xci 6rd ro0 rpirou oytipatos. toOro 5d
,!ei6os"iv6{6tat
yd.p rorcAtov elvor td B Qt lyTupet 16 A pqtev| indpyecv.

$sz

SOLUTION

OF TH E C ON TR OVER SY

rqr

some would-be true apodeictic premisses. As many logicians

believe that such propositions are really true, it is impossible to
convince them of the validity of those syllogisms by examples.
Concluding this discussion we may say that Aristotle is right
in asserting (e), but wrong in rejecting ({). Theophrastus and
Eudemus are wrong in both ways.
$ 58. Moods with possiblepremisses
The Aristotelian theory of problematic syllogisms displays a
very strange gap: moods with possible premisses are entirely
neglectcd in favour of moods with contingent premisses.According to Sir David Ross, 'Aristotle always takes iz6iXeracin apremiss
as meaning "is neither irnpossiblenor necessary"; where the only
valid conclusiozis one in which ?rEtyerat means "is notimpossible",
he is as a rule careful to point this out'.r Aristotle, indeed, seems
to be careful to distinguish the two meanings of du8dxeogcr,r.
when
he says,expounding for instance the moods with two problematic
premisses of the first figure, that tv}(yeodor in these moods should
bc understood according to the definition he has given, i.e. as
'contingent', and not in the senseof 'possible'. He adds, however,
that this is sometimes overlooked.z Who may have overlooked
this ? Aristotle himself, of course,or some of his pupils just because
of the ambiguity of the term iv6(yeo0au In the De Interpretatione
ivSeydy.evov
means the same as .6uvardv,3while in the Prior Anal2ticsit has two meanings. It is always dangerous to use the same
word in two meanings which may be unconsciously confused; as
also to use two different words with the same meaning. Aristotle
somctimes says tyyapei instead of iv}iyerat, and also uses the
Iattcr in two meahings.a We cannot be always sure what he
rn(:ans by tv}/yerat The ambiguity of this term probably contributcd to the controversies between himself and his friends
'l'lrcophrastus and Eirdemus. It is therefore a pity that he did not
trtrrt rnoods with possiblepremissesseparately before introducing
corrtingr:ncy. We shall supply this deficiency which has hitherto
rsr' ;r;rcrltl rc noti ce of scholar s.
| !V I). Iloss, loc. cit., p.
44; see also the table of the valid moods, facing
rr. 'rl l (i .
t .ln
lr i r4, 33bzI 3ei Et zd dv6(yeo0at\appdv<w pi tv tots ivayxaiocs, dlJd,
a,,r,i t,),, ,ipr11Li"ov
6roptop6v.dviozeEi Aqv|dvettd rocoirov. 3 See n. t, p. I34.
(
A n l tr. i .3, e5bro (r. r, p. r9z ) and i , gr 3oaz T(n. l , p. l go)
' l l . l i rr i rrst:ur<'<:
(n. r, p. r93 ).
w i tl r i . r.1,;1rr',1o

r 92

ARISTOTLE' S

M O DAL

SYLLO G I STI C

$SB

Let us first consider the laws of conversion. Aristotle begins the

exposition of these laws in Book I, chapter 3 of the Pior Anafi'tics with the statement that the term iv6/yeo0at has several
meanings. He then says, without explaining the various meanings
of this term, that the laws ofconversion ofaffirmative propositions
are the same for all kinds of 2v\i76o0aa, but those of negative
propositions differ. He states explicitly that the problematic
propositions 'Every b may be an a' and 'Some b may be an a' (I
use the word 'may' to cover both kinds of the problematic proposition) are convertible into the proposition 'Some amaybe a b'
which gives for possibility the formulae:
nr.

CMAbaMIab

and

tzz. CMIbaMIab.

The law of conversion for universally-negative propositions is

explained only by examples from which we may infer the formula :
ng. CMEbaMEab.

ry. CCpqCMpMq.
The same theorem, i.e. the stronger M-law of extensionalitn
enables us to establish the whole theory of syllogisms with possible premisses. By means of the classical calculus of propositions
we get from t9 the formulae:
CCpCqrCMpCMqMr

and

n5. CCpCqrCpCMqMr.

Formula lz4 yields moodp with two possible premisses and a

possible conclusion: we merely have to add the mark of possibility to the premisses and to the conclusion of valid assertbric
t An.
inec}il
F. i. g,25"37-bt4
xera$artxoCs dpoios E(et xetd. iy
tQ B iv}yete4 xei ti B nvi tQ A
<ioatjras, <ill' 6oa piv lv\iyeo|at
dvd.yxt1s pi1 Jndpxecv, dpo|as, otov.
d.v|punov iyyapei pq}evi izn<p, . ..

M O O D S WI T H

POSSIBL E PR EM ISSES

r93

moods. So, for instance,we get according to r24 from the assertoric mood Barbara by the substitution plAba, qlAcb, rlAca the
syllogism:
tz6. CMAbaCMAcbMAca.
Formula rz5 yields moods with one assertoricand one possible
premiss,it doesnot matter which, e.g.
tz7. CAbaCMAcbMAca

:zB. CMAbaCAcbMAca.

The systemis extremely rich. Any premissmay be strengthened

by replacing the assertoricor problematic prooosition by the
correspondingapodeicticproposition.Besides,there are.moods
with one problematic and one apodeictic premiss which yield
apodeictic conclusionsaccording to the formula:
tzg. CCpCqrCMpCLqLr.
Thus we have,for instance,the mood:

It is tacitly assumedthat particularly-negative possible propositions are not convertible.' We seefrom this that the laws of conversion of possible propositions are somewhat negligently treated
by Aristotle. He apparently does not attach any great importance
to the concept of possibility.
Formulae r2r-3 are correct and are easily deducible from the
analogous laws of conversion for assertoric propositions by means
of the theorem:

n4.

$s8

no\Aaytts \(yetar ti v|tyeo9eq . . . dv ptv tots

dvtntpo$iv
dv dnoow. ei yip td A navrl ff revi
oia
iv6l7octo dr. . . , (od dv Ei rois dno$anxois
Aiyetor ff rQ i( dvdyx4s itd.p16l
i
tti it
"Q
. . (bg).i, . , dviiyetat prl}evi dv|pdn,p lnrov, xol
(br3) ripoios Et xai dzi tfis 2v yipet d.no$armfis.

t3o. CMAbaCLAcbLAca
which is contrary to the peioremrule acceptedby Theophrastus
and Eudemus.
I think that Aristotle would have accepted-not, of course,
the last syllogisticmood-but the moodswith possiblepremisses,
in particular rz6 and re8. There is, indeed,in the PriorAnal2tics
an interestingintroductory remark to the theory of problematic
syllogismswhich, in my opinion, rrlay be applied to possibility as
well as to contingency.Aristotle saysthat the expression'Of anything, of which 6 is predicated, a may be predicated' has two
meaningsthe besttranslation of which seernsto be this: 'For all a,
if everycisa 6, then everyc may be ana',and 'For all c,if everyc
may be a 6, then every r may be an a'. Then he adds that the
cxpression'Of anything, of which D is predicated, a may be
predicated'meansthe sameas 'Every b may be an a'.I We have
thus two equivalences:'Every b rliraybe an a' meanseither ,For
all c, if every c is a b, then every c may be an a', or 'For all c, rf
cvery d may be a D, then every c may be an a'. If we interpret
'may' in the senseof possibility,we get the formulae:
I An. pr. i. 13,3zbz7 t6 ydp,'xe?'
o0 rd B, i A 2v}iyeoflaf toitav oqpaivet
0drcpov, ff 'xa|' o0 Atyetat i B' i 'xa|' o0 2v6lyetar Alyeolar' . rd D{ , xol, ot i B, td
A iv6(yeo0oi i'navri rQ B td A iyyapeiv'oJEtv 8rc{lper.

rg4

ARISTOTLE' S

M O DAL

SYLLO G I STI C

$sB

QMAbaITcCAcbMAcaand rg2. Q,MAbaIIcCMAcbMAca

which are true in our systemof modal logic, and from which the
moods IeB and 126 are easilydeducible.I[ however, 'may' is
interpreted in the senseof contingency which seemsto be the
intention of Aristotle, then the formulae given above become
false.
ty.

propoitiorc
of contingent
| 59. Laws of conaersion
Continuing his exposition of the laws of conversion of modal
propositionsAristotle saysat the beginning of the Prior AnaQtics
that universally-negativecontingent propositions are not convertible, whereasparticularly-negativeonesare.t
This curious statementdemandscareful examination.I shall
first discussit critically not from the point of view of my modal
system, but from that of the basic modal logic accepted by
Aristotle and all logicians.
According to Aristotle, contingency is that which is neither
necessarynor impossible. This meaning of the contingent is
clearly implicit in the somewhat clumsy definition of Aristotle,
and is expresslycorroborated by Alexander.2 Let us repeat in
order to ensurecomplete clearness:'/ is contingent-means the
sameas-1 is not necessaryandp is not impossible',or in symbols:
a8. QTpKNLpNLNp.
This formula is obviously equivalent to the expression:
5o. QTpKMpMNf,
i.e. the contingent is both capable of being and capable of not
being.
Formulae 48 and 50 are quite general and applicable to any
propositionp. Let us apply them to the universally-negativepropositionEba. We get from 5o:
t 33. QT EbaK M EbaA(J,IE ba.
As NEba is equivalent to lba, we also have:
I An. pr, i.
3, z5bl4 (continuation of the text quoted in n. I, p. r9z) 6oa 6i zr! <,is
inizd ro)'J xai tQ te$uxtvor \iyetu 2v3i76o0oc,. . . oJX ipoios (er iv teis ueplflxais dmoapo.rcis, dl)' fi piv xo06Aouorepqtu)1 rpdtao6 oix dmntpi$eq i Ed dv
pipet d,vttotp&fet2 See above,
$ +S, in particular nn. 3r p. I54 and tr p. t55.

$ sg

L A WS O F C ON VER SION

t9 5

ry4. QTEbaKMEbaMIba.
Now we can derive from the laws of conversion:
tzg. CMEbaMEab

and

nz.

CMIbaMIab

that MEba is equivalent to MEab, and MIba to MIab; hence we

have:
ry5. QKMEbaMIbaKMEabMIab.
The first part of this formula KMEbaMIba is equivalent to TEba,
the second KMEabMIab to TEab; so we get the result :
rg6. QTEbaTEab.
This means that contingent universally-negative propositions are
convertible.
How was it possible for Aristotle not to see this simple proo{,
when he had all its premissesat his disposal? Here we touch on
another infected portion of his modal logic, even more difficult to
cure than the wound which his ideas about necessity inflicted on
it. Let us see how he tries to disprove formula 136.
Aristotle states quite generally that contingent propositions
with opposite arguments are convertible with one another in
rcspect of their arguments. The following examples will explain
this not very clear formulation. 'It is contingent that D should be
an a' is convertible with 'It is contingent that Dshould not be an a' ;
'It is contingent that every b should be an a' is convertible with
'It is contingent that not every D s,houldbe an a'; and 'It is contingent that some 6 should be an a'is convertible with 'It is contingent that some D should not be an a'.r This kind of conversion f
shall call, following Sir David Ross,'complementary conversion'.2
Aristotle would assert accordingly that the proposition 'It is
contingent that every 6 should be an a' is convertible with the
proposition 'ft is contingent that no Dshould be an a',in symbols :
(r) QTAbaTEba
(assertedby Aristotle).
This is the starting-point

of his proof,, which is performed by

I An. pr. i. rg,

3zazg oupBaiver 6| zdocs tds xotd td iv6yeo0acnpord.oers dlnorpi$e w 'i*fi),ars. )/7o Et ori rri s xata$orwds raes d,ro$atuots, d))' 6oecxata$atudv
i2qovorrd oXfipa xatd. tlv dvtieeow, otov td v6iyeo0ot indpyew tQ ivityeoflor pi1
l,n,ip1gew,xsi td nevri di6tyeo0at rQ 2v6yeo0erprlEevi rci pi tavti, xoi ti twi tQ
1ti1tvl.
r W. D. Ross, loc, cit., p.
44,

r96

ARISTOTLE' S

M O DAL

SYLLO G I ST I C

$rs

reductioad abrurdam.He argues in substancethus: If TEba were

convertible with TEab, then TAba would be convertible with
TEab, and as TEab is convertible with TAab, we should get the
false consequence:
(x) QTAbaTAab
(rejectedby Aristotle).I
What shouldwe sayto this argument? It is quite obviousthat
the definition of contingency adopted by Aristotle entails the
convertibility of contingent universally-negative propositions.
Consequentlythe disproof of this convertibility muit bi wrong.
Since it is formally correct, the error must lie in the premisses,
and as there are two premisseson which the disproofis based,the
assertedformula (r.),and the rejected (rc),then either it is wrong
to assert(r) or it is wrong to reject (r). This,-however,cannot be
decided within basic modal logic.
Within those limits we can merely say that the truth of the
assertedformula (r) is not justified by the accepteddefinition of
contingency.From the definition:
5o. QTpKMpMNp
we get by the substitution plMp the formula QfNpKMl{pMNNp,
and as MNN| is equivalent to Mp according to thesisg of basic
modal logic, we have:

ry7. QTNpKMpMN\.
From 5o and r37 there resultsthe consequence:

ry1."WpTNp,
and applying this consequenceto the premissEbawe get:,
rgg. QTEbaTNEba

or

r4o. QTEbaTIba,
as NEba means the same as lba. We see that QTEbaTIba rs
justified by the definition of contingency,but that QTEbaTAbais
not. This last formula hasbeen acceptedby Aristotle by a mistake.
We shall understand tlds error better if we examineAristotle's
I An. pr. i. 17,
36b35 nptizov oiv 6erxr(ov 6tt oik dvrnrp!{er td 2v tQ iv6lyeo0et
orepqrrxdv, otov ei ti A iviiyetat p16evi rQ B, oix dl,iyq rai zd B iviiTeoiat
p1}evi tQ A. xeio0o yd.p toito, xoi iv}eyio9a i B pq}ev| tQ A ird.pyew. oJxofiv
inei dwtmpd$ovou of iv tQ iviiyoflot xaraSd.oec rais d,no$d,oeoqxat ai ivavtiac xoi
at dmueipeaq
rd 5t B tQ A iv}lyerac pl}evi Jmdpyew, $avepdv 6n xei nawi Ev
iviiyocro rQ A irdpyecv. ro0ro 6l ry'e06os.ori yd.p ei il6e tQ6e nawi ivilyctaq
xai
.rriEe zrp6e dvayxaiov. ,iot' oix ,iwtmp($et rd neprytmdv.

!ss

L A WS OF C ON VER SION

r97

refutation of an attempt to prove the law of conversion for TEba

by redrctioad abnrdam.This attempt reads: ifwe supposethat it is
contingent that no 6 should be an a, then it is contingent that no a
should be a b. For ifthe latter propositionwere false,then it would
be necessarythat somea should be a b, and henceit would be
necessary
that someDshould bean a which is contrary to our supposition.t In symbols: If TEba is supposedto be true, then TEab
also must be true. For from NTEab would result LIab, and consequentlyLlba,which is incompatiblewith the suppositionTEba.
Refuting this argument Aristotle rightly points out that LIab
doesnot follow from NTEab.z We have, indeed, according to 48
the equivalence:
r4r. QTEabKNLEabNLNEab
or
t 42. QTEabKNLEabNLIab.
Thus for"MZEa6,applying QNKNpNqHp4, i.e. one ofthe so-called
'De Morgan's laws',l we have the formula:
t43. QNTEabHLEabLIab.
It can be seenthat by meansof r43 and the thesisCCHpqrCqrwe
can derive NTEab from LIab, but the convenieimplication does
not hold, since from NTEab we can derive only the alternation
HLEabLIab from which, of course, LIab does not follow. The
attempted proof is wrong, but it doesnot follow that the conclusion which was to be proved is false.
One point in this reduction deservesour attention: it is
apparent that insteadof r43 Aristotle acceptsthe formula:
(A) QNTEabHLOabLIab
which is not justified by definition 48. Similarly for the caseof
NTAab he adoptsthe formula:+
I An. pr. i. t7, g7"g il)d. pilv o116'dr zori diwdrou 6eq10{oerotdrttorpifuv, otov el
tts d.(uiocuv, e'zei g[e0Dostd ivti26o0at td B rQ A p76evi $n&pyew, ,i\r102sil, gti1
A
lv6lyo0at pl}evi ($doc ydp xai 'in6$aors), ei El roir', d)t1flis i{ 'ivtiyxqs tlirQ
Jnd,pTtcv'tiote xai rd A tui tit B'toiro 6' ,i6rjvotov,
I Ilrid.37'r4 (continuationof theforegoingnote) oJT,ipeipl2v6iTerutpq}evi:d
B tQ A, dvd.yxrl rli indpyew, rd' ydp pi1 iviiTeoiat y76evi6qg6s Liyetat, td piv ei i!
d.vd.yxr1c
t*I Jnd.p3ec,rri 6' ei e'f dvdyxrls rwi pfi ind.pyeu
t 'l'hese should properly be called Ockham's Laws, for so far as we know,
Or:khrnr was the first to state them, See Ph. Boehner, 'Bemerkungen zur Geschichte
der De Morganschen Gesetze in der Scholastik', Archiufiir Philosophic(September
rg5r), p, rr5, n.
t An,
lr. i. 17, 97'24 tQ iv}ileoflar nawi $tdpyew d r' 2{ dvdynls nvi Jnd.pycl
dvtlxtrar ral rd if dvdyxls t'i p'l ird.pycl.
I

r98

ARISTOT LE' S

M O DAL

SYLLO G I STIC

$ sg

Qt) QhrTAabHLOabLIab
which, again, is not justified by 48, whereas the correct formula
ru n s:
r44. QNTAabHLOabLAab.
From ()) and (pc)Aristotle may have deduced the equivalence
QNTAabNTEab, and then (c), which is not justified by his
defi nition of contingency.
$ 6o. Rectifcation of Aristotle's mistakes
Aristotle's theory of contingent syllogisms is full of grave mistakes. He does not draw the right consequencesfrom his definition
of contingency, and denies the convertibility of universallynegative contingent propositions, though it is obviously admissible.
Neverthelesshis authority is still so strong that very able logicians
havc in the past failed to sec these mistakes. It is obvious that if
somebody, Albrecht Becker for example, accepts the definition

aB. QTpKNLpNLNp
with p as propositional variable, then he must also accept the
formula:
t 4r. QTEabKNLEabNLNEab
which is derived from 48 by the subsritution plEab. And
since Qy valid logical transformations formula r4t yields the
"
thesis
r43. QNTEabHLEabLIab,
he must also accept r43. Yet Becker rejects this thesisin favour of
'structural lbrmulae'-a product of his imagination.r
The remarks of the foregoing section were written from the
standpoint of basic modal logic which is an incomplete systcm.
Let us now discussour problem from t-he point of view of fourvalued modal logic.
From the Aristotelian definition of.contingency we obtained
the consequence r3B, QTpTNp, from which we may deduce
the implication:
r See A . B e c k e r , l o c , c i t., p . r 4 , wh e r e fo r m u la T r r :4 8
wri tten i n another
symbolisrn, but with the propositional variable p, is accepted, and p. e7 where
formula r43 is rcjected.

$6o

RECTIFICATION

OF AR ISTOTL E'S

M ISTAKES

I9 9

t45. CTpTNp.
Now we get from the premisses:
(axiom of the C-"1/-6-1-system)
5r. CDpC6Np6q
(principle of Frcge)
r46. CCpCqrCCpqCpr

the consequences:
5r. 3lT'x t47
r+7. CTpCTNpTq

r 46.pI Tp, CI TNp, rI Tq x Cr47-Ct 45-r48

r4B.CTpTq,
and as the converse implication CTqTp is also true, as may be
proved by the substitutions plq and Clp in r4B, we have the
equivalence:

rag. QTpTq.
From r49 we get by substitution first the law of conversign r36
QTEbaTEab, then formula (r) QTAbaTEba which Aristotle
asserts,and formula (x) QTAbaTAaD which he rejects. We can
now determine where the flaw in Aristotle's disproof of the law of
conversion is : Aristotle is wrong in rejecting (rc).
Formula Qfpfq shows that the truth-value of the function Tp
is independent of the argumentp, which means that Tp is a constant. We know, in fact, from $ 52 that KMpMNp which is the
defniensof Tp has the constant value j, and therefore Zp also has
tlrc constant value 3 and is never true. For this reason Tp is not
suitable to denote a contingent proposition in Aristotle's sense,
sincc he believes that some contingent propositions are true.
f1r must be replaced by Xp or Tp, i.e. by the function 'p is X-continscnt'or its twin 'p is T-contingent'. I shall take into consi<lt'r:rtionmerely X-contingency, as what is true ofX-contingency
rvill :rlsobe true of I-contingency.
lf ir sl, I should like to state that the convertibility of universallyrrlr' ;rI ivt' r:ontingentpropositionsis independent of any definition
r rl r r rnl ins<'n<:y.
As Eba is equivalent to Eab, we must accept the
l , rrrrrrrl ;r
t,tt. O,\l ,.ha6E ab
;rrr,rrlirrl; (o tht: principle of extensionaliryCqpqc6p64, which
rl srrl l s l i orrr orrr:rxi om 5r . Fr om r 5o we get a t r ue st at em entf or
:rrry r,:rl rrcol ' 8, Ircnccal so f or E/ X':

2 oo

ARISTOTLE' S

M O DAL

SYLLO G I STI C

06o

r5t. CXEbaXEab.
Alexander reports that Theophrastus and Eudemus, unlike
Aristotle, acceptedthe convertibility of universally-negativecontingent propositions,Ibut saysin another passagethat in proving
this law they used reductioad absurdum.z
This seemsdoubtful, for
the only correct thing Aristotle had done in this matter was to
refute the proof of convertibility by reductio,a refutation which
cannot have been unknown to his pupils. Redurtio
can be usedto
prove, from CLIbaLIab, the convertibility of universally-negative propositions when they are possible (that is, to prove
CMEbaMEaD),but not when they are contingent. Another proof
is given by Alexander, continuing the former passage,but he
scarcely formulates it clearly enough. We know that Theophrastusand Eudemusinterpreted universally-negativepremisses,
Eba as well as Eab, as denoting a symmetric relation of disconnexion betweenb and a,t and they may have argued accordingly
that if it is contingent for b to be disconnectedfrom a, it is also
contingent for a to be disconnectedfrom ,.4 This proof would
conform with the principle of extensionality.At any rate, Theophrastus and Eudemus have corrected the gravest mistake in
Aristotle's theory of contingency.
Secondly, it follows from the definition of X-contingency:
Bz. CDKMpI4/NqDX7
that the so-called 'complementary conversion'cannot be admitted. QTpfNp is true, but QXpXNp must be rejected,because
its negation,i.e.:

ry2. NQXpXNp
is assertedin our systemascan be verified by the matrix method.
It is therefore not right in our systemto convert the proposition
t Alexander zzo. g @e6$paotosplvtot xai Eiiqpos . . . ,iwtmpi$ew
Qaoi xai rlv
xs06\ov &roSatwilv (scil. e'vDeXopivry) ditfi, Sonep d.wlotpeSe xai fi itd.pyouoe
xai ri dvayxaia. 4
xo4d\ou izoSatcri
2 lbid. zz3. g 66{et noi 6ui' ye rffs eis &6ivatov &nayoyffs 6$voo0at 6etxwo0er i7
xa06\ov droSonxl
ivieyoplry dworpi{ouoe.
fi oJrff 6e/fer rci oi taipor o&toA
xiypqmau
t See ibid. gr.
4-ro.
a lbid. zeo. rz dzr 5t dmrcrpi$eq \euvAow oirus' ei i A tQ B iv}iycrat
p16cvt, xai i B rQ A 2v6i7ptat p16oi. daei yd.p 2vitTetar i A tQ B p16et, 6te
dviiyctet pr16evi,tdte dviiyerat dzefe0y9ecrd A ndltav t6w roi B'.ei 62 roAt' , Eorct
tdre xoi td B rc6 A drefevypfuov' ei Et rooro, xoi rd B tQ A dv6476rerpq}evl

5 6o

RECTIFICATION

OF AR ISTOTL E'S

M ISTAKES

2ol

'It is contingent that every b should be an a' into the proposition

'It is contingent that some , should not be arr a', or into the
proposition 'It is contingent that no 6 should be an a', conversions which Aristotle acceptswithout any justification.t I think
that Aristotle was led to a wrong conception of 'complementary
conversion' by the ambiguity of the term 'contingent' (du8eXrip.evov).He usesthis term in the De Interpretatione
as a synonym of
the term 'possible'(6uvardv)rz
and continuesto useit thus in the
Pior Aml2tics,although the phrase 'It is contingent that p' has
there got another meaning, viz. 'It is possible that p and it is
possible that not p'. If we replace in the last phrase the term
'possible'by the term 'contingent',as Aristotle apparently does,
we get the nonsensethat 'It is contingent that p' meansthe same
as 'It is contingent that p and it is contingent that not p'. So far
as I know, this nonsensehas hitherto not been observedby anybody.
Thirdly, it follows from definition Be tfrat Xp is stronger than
Mp, becausewe have the thesis:
t5g. CXpMp,
but not conversely.This thesisis important, becauseit enablesus
to retain, with a little correction, a large number of syllogisms
with contingent premisses,in spiteof the seriousmistakesmade by
Aristotle.
premisses
,
| 6r. Moodswith contingmt
There is no need to enter into a detailed description of the
syllogisticmoods with contingent premisses,as Aristotle's definition of contingency is wrong and his syllogistic should be rebuilt
accordingto the correctdefinition. This, however,doesnot seem
to be worth while, for it is very doubtful whether a syllogisticwith
rxrrrtingentpremisseswill ever find a useful application. I think
tlrnt tlrc following generalremarkswill be sufficient.
!'irst, it may be shown that all the Aristotelian moods with a
contingcntconclusionare wrong. Let us take as an examplethe
m<xrd llarblra with contingent premissesand conclusion, i.e.
the mood
rr
714,CXAbaCXAcbXAca.
I S een. r, p. rg5.

2 S een. t, p. I34.

2O 2

ARISTOT LE' S

M O DAL

SYLLO G I STI C

$6r

This mood though accepted by Aristotlel must be rejected. Take

Aba and Acb as false, and Aca as true. These conditions fulfil the
assertoric mood Barbara, but from r54, applying the matrices
M9 and Mr5, we get the following equations: CXoCXoXT :
C3C3z : C3z: z. Similarly mood
*r55. CXAbaCAcbXAca
also accepted by Aristotle2 must be rcjected, since, for Aba : o,
and Acb : Aca: r, we have: CXoCrXr: CJCrz: C3z: :. It
was just these two moods that I was referring to when I said at
the end of $ 58 that formulae I 3 r and l3e, which Aristotle asserts,
became false, if we interpreted iv\iyeo9at as 'contingent'. It may
be said too that formulae r54 and 155 become true, if forXis put
Z, but Z-contingency is a uselessconcept.
Secondly, all the moods got by complementary conversion
should be rejected. I shall show by an example how Aristotle
deals with this sort of mood. He applies to I54 the formula
*156.
QXAbaXEba
which should be rejected (take Aba : r, and Eba : o), and gets
the following moods:
*r57. CXAbaCXEcbXAca
*r58. CXEbaCXEcbXAm
which must be rejected too.3 To show this, it sufficesto choosethe
termq a, D, and c of r57 in such a way that Aba : Ecb : o, and
Aca : r, and those of r5B in such a way that Eba : Ecb : o,
and Aca : r. We then have in both cases: CXoCXoXT :
C3C3z : C3z :2.
It seemsthat Aristotlc does not put much trust in these moods,
I An. pr. i. I4,
rQ f,
xqi rd B rovi
32b38 Srav oA, rd A zavti tQ B tv}lytyar
ind.pyeLv. zo0ro 3i $evepd,
ovL\oyropds Eotar rlAercs 6rt rd A navtl ri f iv|iyrat
ydp irEixeo|ar navr|. Jn&pycw oirus
iAlyopev.
ix toi 6pnpo6'
"d
1 Ibid. r5,33br5
lappdvryer ritv rpordocav,
dtiv 6'fi pivi'n,ipyel'fi6'iv}(4eo|ai
nd.vres o;.
6rav piv i1 npds rd pet{ov d.xpov iv3(yeo9ar o1patvy1, tiAecoi r' ioorrst
xard, rdv eiprlltivov 3rcpcop6u.
ov\A,oytolroi xai toA iviiyeo9at
3 Ibid. I4, gg"5drovStrd
B ivliXtlror, zri Ei B iv6(pyor
pq}evi rQ
AtovrirQ
6i
npord.oeu, oJ}eis yi,vetat ovAAoyrcp6s, dlnorpa$etols
i-, E,,i ptu tdv eiAlppivav
dpoios Et rai ei
ytverar 6 e}tde 6onep rp6repov. -93"12
tfis Bl xotd rd ivdtl6o0at
6' otov
npds dp$oipas
tds rpordoers fi d.n6$aots r<0ei1 pet& toa iv}/yeo?ou )lyo
eiAqppivav
A iv6/7etoc pr1leri rQ B xei rd B pqievi t<i .f' 8cd ptv ydp tir
ei i
6onep
3t n<iA"v 6 aitds Enu
npord.oeav oileis yiveto" ovA),oyrop6s, dvrtorpe$opivuv
rcoLnpdrepov.

$6'

MOOD S

WITH

C ON TIN GE N T

P R E MIS S D S

2oq

because he does not call them syllogisms at all. He merely says

that they can be reduced to syllogisms by means of complementary conversion. But moods reduced by the ordinary conversion are called by him syllogisms; why does he make a
difference between ordinary and complementary conversion, if
both kinds of conversion are equally valid?
Light upon this question is thrown by Alexandcr who, commenting on this passage,refers to a very important remark of his
master on two ontological meanings of contingcncy: 'In one
('contingent"
sense
means "usual (tnl rdn o)ri) but not necessary"
or "natural", e.g. it is contingent that men should go grey; in
another senseit is used of the indefinite, which is capable of being
thus and ofnot being thus, or in general of that which is by chance.
In either sense contingent propositions are convertible with
rcspect to their contradictory arguments, but not for the same
reason: "natural" propositions becausethey do not expresssomething necessary,"indefinite" propositions becausethcre is not, in
their case, a greater tendency to be more thus than not thus.
About the indefinite there is no scienceor syllogistic demonstration, becausethe middle term is only accidentally connccted with
the extremes;only about the "natural" are there such things, and
most arguments and inquiries are concerned with what is contingent in this sense.'r
Alexander discussesthis passage:his idea seemsto bc that, if
wc take any scientifically useful syllogism the premisscs of which
arc contingent in the senseof 'usual' (tni rd toAJ) or even 'most
usual' (izi rd nAeCorov),then we get premissesand a conclusion
wlrich are indeed contingent but are very seldom (tn' tAarrov)
rcitlized: such a syllogism is useless(d.yprloros).Pcrhaps this is
wlry Aristotle refusesto call what is so obtained a syllogism.z
| .'l t
r q,gz b4- z t
r dtv D i y < o 9 a r x c rd S ri o A (y e t a rrp d ro v s , v a p i v rd t i s i n i i
l tr .i
a'tl) yitro|u
xoi 8lo).elreo rd dvayxoiov, otov rd noAnAo|at dv9purov. . ., i 6,\os zr)
t t ,l,rra r't1i'n,irfrrv . . ., d)loz Ei rd ,i6prcrov,6 xai olras xai p) oitas 3v"ar6v, . . . fi 6Aas
,,i ,i",i ,rjl'is yrvdpevov. -(btg)
dvrrcrp(Set piv oiv xoi xard. rds ,ivrtxecplvas
rs i *i n pov ri'v tv\eyop(vov, oi plv rdv oir6v yc tpdnov, dl)d tt) piv ne$uxde
n 1",
',i't'
, ly", t ,,, 1' ) i l. ,irriyxls in,ipy<w . . ., zd 3' d.6proro, rQ pr16iv pdlAov oJtus i ixe lvas.
,'n," r /11,,1ir) n,,l ,rr,,\,\o7ropds tinoSerxrrxds r6v ptv d.opiorov oix Eorc 6ad-td d.rexrov
,lr,rt ,,\ 1t,',t,',, r,i, ii rc,fuxdrav Eort, xai oye66v oi\6yot xei ei, oxirltets yivovrot rcpi
tr l ,f,

,,i ,( f,\

i r i \t

yr tl u 'r r o t

ivieToplvq i
in'
' ,\fr x;rrrrl.r rlr.1 r r6t ydp ,is ini rd n\eiorov dro$otcxQ
i ) ,r t r ,r . a,',,r ,1,,'r ,^i 'r ' i Lr tor p( 6a.
ro J ro u 6 i x e t l t (v o v o u \ l o y rc p d s p t v E o t a 4 o i
-5
.
1,,)y S1,ri,'r1t,iLtr iy,,tr',,,is .,Jr<is nputne. Dtd xai ipoApcv zorjrcs rds ov(vyios

2o4

ARISTO TLE' S

M O DAL

SYLLO G I S T I C

$6t

This point, more than any other, reveals a capital error in

Aristotle's syllogistic, viz. his disregard of singular propositions.
It is possiblethat an individual, {, should be going grey while
growing older, indeed this is probable, though not necessary,
sinceit is the natural tendencyto do so. It is also possible,though
rather improbable, that { should not be going grey. What
Alexander says about the different degreesof possibility is true
when applied to singular propositionsbut becomesfalse when
applied to universal or particular propositions.If there is no
general law that every old man should go grey, becausethis is
merely 'usual' and someold men do not go grey, then, of course,
the latter proposition is true and therefore possible,but the
former is simply false, and from our point of view a false proposition is neither possiblynor contingentlytrue.
Thirdly, from a valid mood with possiblepremisseswe can get
other valid moods by replacing a possiblepremissby the corresponding contingent one. This rule is basedon formula r53 which
states that Xp is stronger than Mp, and it is obvious that any
implication will remain true, if one or more of its antecedentsis
replaced by a stronger antecedent. So we get, for instance,from
le6. CMAbaCMAcbMAcathe mood ryg. CXAbaCXAcbMAca
and from
l:'8. CMAbaCAcbMAca the mood

16o. CXAbaCAcbMAca.

Comparing the rejected moods 154 and r55 with the asserted
moods r5g and 160,we seethat they differ only by the substitu- \)
tion of M for X in the conclusion. If we examine the table of
Aristotelian syllogistic moods with problematic premisses,given
by Sir David Ross,t we shall find it a useful rule that by this
small correctton, M in the conclusion, instead of X, all those
moods become valid. O-nly the moods obtained by complementary conversioncannbt be corrected,and must be definitively
rejected.
&yp/,mous te xai d,n),,Aoyiotous etvau -to ious Dt rci crizds toito i{optipevos clte
'i oi yivetac ouLAoyrr1tris'.Cf. W. D. Ross's paraphrase of this passage, loc. cit.,
"i
p.
- i326.
W. D. Ross, loc. cit., facing p. 286; in the conclusion the index c should cverywhere be replaced by P.

0 6e

T M P L I C A TION S

OF M OD AL

L OGIC

2o5

$ 62. Philosophicalimplicationsof modatlogic

It may seemthat the Aristotelian modal syllogistic,even when
corrected, has no useful application to scientific or philosophic
problems. But in reality, Aristotle's propositional modal logic is
historically and systematically of the greatest importance for
philosophy.All elementsrequired for a completesystemof modal
logic are to be found in his works: basic modal logic and the
theoremsof extensionality.But Aristotle wasnot able to combine
thoseelementsin the right way. He did not know the logic of
propositionswhich wascreatedafter him by the Stoics;he tacitly
iccepted the logical principle of bivalence,i.e. the principle thai
everypropositionis either true or false,whereasmodal logic cannot be a two-valued system.Discussingthe contingencyof a
future sea-fighthe comesvery near to the conceptionof a manyvalued logic,but he laysno stresson this greatidea, and for many
centurieshis suggestionremained fruitless. Owing to Aristotle I
was able to discover this idea in lgro and to construct the first
many-valued systemof logic in opposition to the logic, hitherto
known, which I called 'two-valued logic' thus introducing a
term now commonly acceptedby logicians.r
Under the influence of Plato's theory of ideas Aristotle developed a logic of universal terms and set forth views on necessity
which were, in my opinion, disastrousfor philosophy. Propositions which ascribe essentialproperties to objects are according
to him not only factually, but alsonecessarilytrue. This erroneous
distinction was the beginning of :{ long evolution which led to the
division of sciencesinto two groups : the apriori sciences
consisting
of apodeictic theorems,such as logic and mathematics, and the
a posteriorior empirical sciencesconsisting chiefly of assertoric
statementsbasedon experience.This distinctionis, in my opinion,
false.There are no true apodeicticpropositions,and from the
standpointof logic thereis no differencebetweena mathematical
and an empirical truth. Modal logic can be describedas an
extensionof the customarylogic by theintroductionofa 'stronger'
I See
(Two.valued Logic), Przegl4d
J. Lukasiewicz, 'Logika dwuwarto5ciowa'
Filozofuzn2,eg,Warszawa (rgzr). A passageof this paper concerning the principle
of bivalence was translated into French by W. Sierpiriski, 'Algibre des ensembles',
Monografu Mataragtcznr, 23, p. 2, Warszawa-Wroclaw ( r g5r ). An appendix of my
German paper quoted in n. r, p. 166, is devoted to the history of this principle
In antrqurty.

"\

20,6

ARISTO TLE' S

M O DAL

SYLLO GI S T I C

06e

and a 'weaker' affirmation; the apodeictic affirmation Lp is

stronger, and the problematic Mp weaker than the assertoric
affirmationp. If we use the non-committal expressions'stronger'
and 'weaker' insteadof 'necessary'and 'contingent',we get rid
of some dangerous associationsconnected with modal terms.
Necessityimplies compulsion, contingency implies chance. We
assertthe necessary,for we feel compelled to do so. But if Za is
merely a stronger affirmation than cr,and a is true, why should
we assertZcr?Truth is strong enough, there is no need to have
a 'supertruth' stronger than truth.
The Aristotelian a priori is analytic, basedon definitions,and
definitionsmay occur in any science.Aristotle'sexample 'Man
is necessarilyan animal', based on the definition of 'man' as a
'two-footed animal', belongs to an empirical science.Every
science,ofcourse, must have at its disposalan exactly constructed
language and for this purpose well-formed definitions are indispensable,as they explain the meaningof words,but they cannot
replace experience. The analytic statement 'I am an animal'
madeby aman-analytic because'animal'belongsto the essence
of man-conveys no useful information, and can be seen to be
silly by comparisonwith the empirical statement 'I was born the
ztrt D.".-Ler IBTB'.If we want to know what the 'essence'of
man is-if there is such a thing as 'essence'at all-we cannot rely
on the meaningsofwords but must investigatehuman individuals
themsdves, their anatomy, histology, physiology, psychology,
and so on, and this is an endlesstask. It is not a paradox to say
even today that man is an unknown being.
The same is true for the deductive sciences.No deduqtive
systemcan be basedon definitions as its ultimate fundamentals.
Every definition supposessomeprimitive terms, by which other
terms may be defined, but the meaning of primitive terms must
be explainedby examples,axiomsor rules,basedon experience.
The true a piori is alwdls synthetic. It doesnot arise, however,
from somemysteriousfaculty of the mind, but from very simple
experimentswhich can be repeated at any time. If I know by
inspection that a certain ballot box contains only white balls, I
can say a prioi that only a white ball will be drawn from it. And
if the box contains white and black balls, and two drawings are
made, f can foretell a prioi that only four combinations can possibly occur: white-white, white-black, black-white, and black-

$6e

I M P L I C A TION S

OF M OD AL

L OGIC

2o7

black. On such experiments the axioms of logic and mathematics

are based ; there is no fundamental difference between a priori and
a posteiori sciences.
While Aristotle's treatment of necessity is in my opinion a
failure, his concept of ambivalent possibility or contingency is an
important and fruitful idea. I think that it may be successfully
applied to refute determinism.
By determinism I understand a theory which states that if an
event -E happens at the moment l, then it is true at a\y moment
earlier than I that E happens at the moment l. The strongest
argument in defence of this theory is based on the law of causality
which states that every event has a cause in some earlier event.
If so, it seems to be evident that all future events have causes
which exist today, and existed from eternity, and therefore all are
predetermined.
The law of causality, however, understood in its full generality
should be regarded as merely a hypothesis. It is true, of course,
that astronomers, relying on some laws known to govern the
universe, are able to predict for years in advance the positions
and motions of heavenly bodies with considerable accuracy. Just
at the moment I finished writing the previous sentence a bee flew
humming past my ear. Am I to believe that this event too has
been predetermined from all eternity and by some unknown laws
governing the universe? To accept this would Iook more like
indulging in whimsical speculation than relying on scientifically
verifiable assertions.
But even if we accept the law of causality as generally true, the
argument given above is not conclusive. We may assume that
cvery event has a cause, and nothing happens by chance, yet the
chain of causes produci.ng a future event, though infinite, does
not reach the present moment. This can be explained by a mathematical analogy. Let us denote the present moment by o, the
moment of the future event by r, and the moments of its cansei
by fractions greater than |. As there exists no smallest fraction
grcater than !, every event has a cause in an earlier event, but
thc whole chain of these causes and effects has a limit at the
moment |, later than o.
Wc may therefore assume that the Aristotelian sea-fight of
tomorrow, though it will have a cause which itselfwill have cause
tncl so on, does not have a cause today. Similarly we may assume

208

ARISTOTL E'S

M O DAL

SYLLO G I STI C

562

that nothing exists today which would Prevent there being a seafight tomorrow. If truth consists in the conformity of thought to
reality, we may say that those propositions are true today which
conform with today's reality or with future reality in so far as that
is predetermined by causes existing today. As the sea-fight of
tomorrow is not real today, and its future existence or nonexistence has no real cause today, the proposition 'There will be
a sea-fight tomorrow'is today neither true nor false. We can only
say: 'There may be a sea-fight tomorrow' and 'There may not
be a sea-fight tomorrow'. Tomorrow's sea-fight is a contingent
evnt, and if there are such events, determinism is refuted.

INDEX
,4, constant functor, means 'all-is'or'belongs
to all',pp. r4,77.
Aaa, axiom, p, 88; syllogistic law of identity independent of other theses,p. 45;
compared with the propositional law of identity, p. 48 ; used by Aristotle in a
demonstration but not stated explicitly, p. r49, n. r.
Aab, rneans 'all a is b' or '6 belongs to all a', p. 77.
ab csscad posscaalet cowcqucntia,known to Aristotle but not formulated explicitly,
pp. 135-6, n. r.
ab oporbrcad csseualct consequcntia,
known to Aristotle but not formulated explicitly,
p'r35'
adfakum scquitur quodlibet,p. t79.
d\ivarov, impossible, p. r34.
A enesi demus,pp.58,59, n. r.
affirmation, 'stronger' and 'weaker', pp. eo5-6.
A l exander, on defi ni ti on of the prem i s s ,p.4,n.4i
on i ndefi ni te premi s s esp.5,
,
n. 2 ; on variables, p. 8, n. z ; validity of moods not dependent on the shape of
vari abl es,p.g, n. 2; hi s proofofc onv ers i on ofthe E -premi s s ,p. to, n. r; on
non-methodically conclusive arguments of the Stoics,p. l5 n. I on formulations
of the syllogismswith 'to belong' and 'to be', p. 17, n. 3 i on the formalism of
the Stoics, p. 19 n.; knows the law of identity Aaa, p. zo,n. r ; quotes syllogisms as rules of inference, p. 2 r n. ; on Theophrastus' addition of five moods
to the first figure, p. 27,n.2; his definition ofthe first figure different from the
Aristotelian, p. 27, n.4; does there exist in the second figure a major and a
minor term $rjoet?, p.3r, nn. t-2; his polemic against Herminus' definition
of the major term, p. 3 r, n. 3 ; his own definition of the major term, p. 32, n, I i

rejection, p. 68, n. l; his polemic dgainst Herminus on rejection, p. 70, n. r ;

on the difference ofthe categorical and hypothetical premisscs, p. r32 n.;
states a general rule that existence implies possibility but not conversely, p.
r 36, n, z ; saysthat necessityimplies existencebut not conversely, p. r 36, n. 4 ;
assimilatesAristotelian definition of contingency to that ofpossibility, p. l4 r n. ;
his definition ofpossibility discussedon the ground ofthe Z-basic modal logic,
p. r4r ; on syllogistic necessity,p. r44, n. 7; acquainted with the logic of the
Stoic-Megaric school, p. 147; his interpretation of the necessaryimplication,

I).2O,J,n.2.
rittttnts tptttors, rrr irnrnediate premiss.
Amrr,rrirrs, rrr r<.1:rri.nof logic to philosophy, p, r3 n.; scholium preserved with
l ri s l i agrrrcnts,p. j g.

2t O

INDEX

dvayxatov, necssary, p. I34.

dvd.yxr1,sccsyllogistic necessity.
analytic propositions, defined, p. t4g; cannot be regarded as necessary, p. t5t.
Anaftics, Prinr, a hypothesis of Bochedski, p. z7; modal syllogistic probably inserted later, p. r3r n.; a hypothesis ofGohlke, p. r33.
and, propositional functor denoting conjunction, pp. 14,77.
ontrkcnttn, Brentano's distinction of'anerkennen' and 'verwerfen', p. 94 n.
antecedent of an implication, p. 78.
,inayayl
eis ri dDJvarov, see rcdution ad imlossibih.
apodeictic principle ofidentity, its consequences, pp. r4g-5o.
apodeictic propositions, defined, p. I34.
a piori, the distinction between the a prioi sciences and the a postcriori sciences dEcussedand criticized, pp. 205-7.
Apuleius, censured by Waitz for changing the order of premisses, p. 33, n. t .
dpc, seatherefore.
arguments, by substitution, p. Ioi non-methodically conclusive of the Stoics,
p. 15 n.; i( iro9ioeas, p, 57.
Aristotle, formulates all syllogisms as implications , pp. 2, 2o-2r, r 38 ; his definition
of 'premiss', p. 3, n. 3; his definition of 'term', p. 3, n. 5 i dposdifferent from
Bcgrif and. definition (6prcpds), p. 3, n. 6 ; his division of premisses, p. 4, n. I i
his definition of universal and singular terms, p, 4, n. 2 i treats indefinite premisseslike particular, p, 5, n. I ; omits empty and singular terms in the syllogistic, p.4; why he omits singular terms, pp. 5-7; his division of things a
division of terms, p. 6; his logic not influenced by Plato's philosophy, p. 6;
introduces variables into logic, p. 7; his term for syllogistic necessity corresponds to a universal quantifier, pp. l r,87, 144-5; his logic formal logic, pp.
r2-r4; not infected by psychology, p. 13; not formalistic, p. 16; his formulations of syilogisms <rftru inexact, p. I I ; examples of inexactness, p. t 8, n. r ;
his division ofsyllogistic figurs, p. 23, n. r ; accepts as principle ofdivision the
position of the middle term in premisssr p. 23, n. z; omits in his diyision the
moods of the fourth figure, p. z3; knows and accepts all the moods of the
fourth figure, pp. 25, n. 2, 26 n.; gives practical indications how to find
premisses for a given conclusion, p. 24 n. i defines wrongly the middle, major,
and rninor terms in the first figure, p. 28, nn. r-2 ; gives a correct definition of
the middle term for all figures, p, 29 n. I does not fix the order of premisses,
pp. ZB, 3'4, nn. r-81 accepts the perfect moods of the first figure as axioms,
p. 44; does not state the dietun de omni ct nullo as the principle of syllogistic,
p. 47; reduces all imperfect moods to the universal moods of the first figure,
p . 4 5 , n . 2 ; t h i s r e d u ctio n m e a n s p r o o { , p .4 4 ; h is theory ofproofunsati sfactory, p. 44; uses laws of propositional logic intuitively in proving the
imperfect moods, p. 49; knows the law of transposition, p. 49, n. 3; and the
law of hypothetical syllogism, p, 49, n. 4; erroneously rejects a thesis of propositional logic, p. 50, n. r ; his proofs by conversion imply laws of propositional logic, pp. 5I-54; [is usually given proofs of Baroco and Bocardo
nnsatisfactory and not proofs by rcduttio ad impossibile, pp. 54-55 ; his characterization of the proofsby rcductio ad impossibib, p. 55 n. ; gives correct proofs of
Baroco and Bocardo implying laws ofpropositional logic, p. 57, n. 3; does not
understand arguments d{ iroilloeos, p. 58; gives proofs by ecthesis for the
conversion of the .I-premiss, p. 6o, n. r ; for Darapti, p. 63, n. r ; for Bocardo,
p. 64 n.; his proofs by ecthesis may be explained by existontial quantifiers,
pp. 6r-66; rejects invalid syllogistic forms by exemplification through concrete terms, p. 67, n. e; employs a rule of rejection, p. 70, n. z; his syllogistic
misrepresented by some mathematical logiciaru, p. r 3o ; why his modal logic

IN D E X

2rr

little known, p. t33; his modal syllogistic has many faults, p. t33; it presupposes a modal logic of propositions, p. 133 ; his four modal terms, p. t 34 ;
mistakenly assertsthat possibility implies non-necessity, p. r34 n.; accepts that
necessity implies possibility, p. r94; gives correctly the relation of possibility
to necessity, p. r 35, n. 3, and that ofnecessity to possibility, p. r 35, n. 4 ; knows
two scholastic principles of modal logic but does not formulate them, pp.
t35-6; presumes existence of asserted apodeictic propositions, pp. ry6-7, r43;
his laws ofextensionality for modal functors, p. r38, nn. 11; his proofofthe
M-law of extensionality, p. r 40 n. ; his definition of contingency, pp. I4o, r54,
n. 3; distinguishes between simple and conditional necessity, p. r44, n. r ;
mistakenly says that nothing follows necessarily from a single premiss, p. r44,
n.3; omits the sign of necessity in valid moods, p. 146; his doctrine concerning the necessary connexion between terms, pp, r48j;
his principle of
necessitn pp. r5r, n. r, r52, n. 2; his defence of indeterministic view, p. t55,
nn.2-3; two major difficulties in his propositional modal logic, p. I57; the
difficulties of his modal syllogistic can be explained on the basis of the fourvalued modal system, p. 169; his acceptance of asserted apodeictic propositions in the Iight of the four-valued system of modal logic, pp. 169-7o; his
acceptance of asserted contingent propositions in the light of the four-valued
system of modal logic, pp. r74-7;his modal syllogistic lessimportant than his
assertoric syllogistic, p, rBr ; states laws of conveision for apodeictic propositions, p. t8t, n. r; his syllogisms with two apodeictic premissesanalogous to
those with two assertoric ones, p. r82, n. r; his doctrine concerning moods
with one apodeictic and one assertoric premiss, pp. rB3-8, and its criticism by
Theophrastus and Eudemus, pp. 184-5, 187-8; his controversy with Theophrastus in the light of the accepted modal system, pp. r88-gr ; negleqts moods
with possible premisses, p. rgr; distinguishes two meanings of iv}ixeo0oq
p. rgr, n. a I treats laws ofconversion for possible propositions with negligence,
p. rg2; his introductory remark to the theory of problematic syllogisms,
p. t 93 n. ; denies convertibility ofuniversally-negative contingent propositions,
p. r94, n. r; his doctrine of 'complementary conversion', p, lg5, n. t; his
definition of contingency entails the,convertibility
of universally-negative
contingent propositions, p. t96; his doctrine concerning the convertibiliry of
contingent propositions criticized fronl the point of view of the basic modal
logic, pp. lg4-g; his moods with contingent premisses and conclusion are
wrong, pp. zor-z; his moods by'complementary conversion'should be rejected, pp. 2o2,2o4i erroneously disregardssingular propositions, p. eo4; his
propositional modal logic, in contradistinction to his modal syllogistic, important for phitosophy, p, 2o5; tacitly accepts the principle of bivalence,
p. 2o5; comes near to the conception ofa many-valued logic, p. zo5; his views
on necessity disastrous for philosophy, p. 2o5; his definition of contingency
wrong, p. zor, but his concept ofcontingency fruitful, p. zo7.
rri tlrrnctical interpretation of syllogistic, pp. r 26-9.
erithrrrctical laws, compared with syllogismsby the Stoics, p, r5.
rl pyrf,l ruri c truth, p.44.
nfierti()n, irrtrrxluced by Frege, accepted in Principia Mothcmatba, p, 94.
rrrrrlorii: llrupnsitions, defined, p. r34.
amrx'intive Inw of addition, without brackets, p. 78.
Averrorl, on (ialcn's fourth figure, p. 38.
lxiourr, of tlrc thcory of deduction, p. Bo; of thesyllogistic, p.88; of basic modal
k rgir',p. r 117; of tlre theory of identity, p. r 49 ; of the C-"1\f7.system,verified by a
rrretrix, p. t',(); ()l' rhc C-./V-6-1-system, p. 16z; of the C-O-ry-system,
p, t{ir rr,; ol thc four-valued system ofmodal logic, pp. 167-8.

212

INDEX

,i{!upa, Stoic term for proposition, p. 8z n.

p' 3;
Barbara, axiom, p. 88; perfect syllogism, pp' 44-45; formulated by Aristotle'
,rith trar,rpo..d premisses and without the sign of necessity, p' ro, n' 5; its
weaknessin the system, p.94; equivalent to a purely implicational formula'
p. r8z.
Barbari, thesis,P' 92.
thesis,p. 94; formulated by Aristotle with transposedpremisses,p'34,.n'7 i
-Baroco,
-is'rln.atiriaciory proof by tcduetioad impossibite,pp' 55-56 ; how Baroco shou.ld
be proved by rciuctio ai impossibite,p. 56; cortect Proof given. by Aristotle'
be proved by ecthesis'
p- iZ,
3; with two apo-deictic piemisses, should
p . r 8 z . ".
basic'modal logic, definition ol; p. r37 I axioms ot, p' r37; is an incomplete

IN D E X

213

ofverifying its expressions,p. 163; its single axiom, P' t62; its rule ofsubstitu'
tion, pp. r6I-z; its rule ofdefinitions, pp. I63-6'
how to verify its expressions by means of the matrix method,
C-.4f7+ysiim,
pp. tS8-S; sec also classical calculus of propositions'
C-O-07-system, its axiom, P. 162 n.
commutation, law of, PP. Bz, 89' Io7.
commutative law of conjunction, p.6l; formulated in symbols, p' 84'
the
compound law of trans-position,inown to Aristotle, pP' 55-57; proved by
Stoics as rule ofinference, p. 59' n. I.
by
compound syllogismsof four termsl investigated by Galen, p' 39, n' 3; divided

modal system, P. 137'

Iot.
b a s i s , o f s y l l o g i stic,p .r o o ;n o tsu fficie n twith o u tslu p e cki 'srul eofrej ecti on'P .
n
.
I9
B
n
.
z,
r
8
I'
n
.
p
p
.
t5
4
,
2
,
A
.
,
Becker,
Bekker, I., p. z4 n.
variables
belong, i,nipyerr, p. t4 n.; used by Aristotle in abstract syllogisms-w-ith
iistead of .iro. in concrete examples, p. r 7 ; explanation of this fact by Alexa n d e r , p . t 7 ,n ,3 .
on
bivatence, p.in.ipl. oI, p. Bz ; tacitly accepted by Aristotle, p' zo5; Lukasiewicz

B o e h n e r , P h . , p . I9 7 , n .3 .
brackets, notation without, pp. 78-79.
24 n''
B.a-aniip, thesis, p. 9e; caliid by Aristotle dvteotpopp(vos ou\)\oytop6s' PP'
n.
26
by
him,
proved
z5;
P.
p. 78; its two-valued matrix, p' I5B; its fourC,- sign of implication'if-then',
-valued
matrix, pp. 16o, 168; its eight-valued matrix, p' t79'
Camnes, thesis,p. 93; proved by Aristotle, p. e6 n'
Camenop, thesis,p. 93'
thesis, p. 93; formulated by Aristotle with transposed premisses' p' 34'
C.^"rt.ir,
n. 6.
Camestrop, thesis,P. 93.
'"ogito, ergo sum', not a principle but an inference' P' 2 I '
Crrte.ian'i.in"ipi.,
categorical system' P. 99.
Celarent, thesis,p. 9z; perfect syllogism' p' 44'
Celaront, thesis' p. 93.
Cesare, thesis,p' 92.
ir'
Cesaro, thesis,P. 93'
chain, p. Ie4.
Chrysippus, p. 8z n.
Cicero, p. 8e n.
logic' p' 167;
Jar.ical'"alcrrt,rs of propositions, should be preserved in any modal
t7B-9;
some of its principies oppo..d at first then universally accepted' PP'
sce also theorY of deduction.
pp' Bo' I65'
Clavius, commentator on Euclid, p. Bo; law or principle of,
p' I b3 ; mcmo<r
C-.ff-6-p-system, explained, pp. | 60-3 ; some of its important theses'

t8t. n. t.
t30'
corru.rrion of the,4-premiss, thesis,p.9r; mistakenly regardecl aslvrong' P'
p' t o'
conversion of the .E-piemiss, thesit, p. g i; proved syllogistically by Alexander'
p' 6o'
convenion q1 dr6 1-premiss, thesis, pi 9l-; proved by Aristotle by ecthesis'
n. r ; proof by i*i.tential q.tnttiifiit , pp' 6r-62 ; the latter proof in slurbols'
PP. 84-85.
conversion of the O-premiss, invalid, p' It' n. I'
conversion of the syllogism, P. 57.
C opl eston,Fr., S .J., P p. I' n. I, 12.
Couturat, L., p. rz6 n.
law of identity, different ftom Aaa, p' 48; deduced within the
Cee, pt.i*iii"nal
C-"lf-07-system, pp. 16z-3'
Cpg, implication means 'if p then C', P. 78.
values cxplained'
E, variable functor of one propositional variable, its range of
pp. r6t-2.
may be proved
nur.pti, thesis,p. gz; proved by Aristotle by ecthesis,P' 63, n' t ;
pP.
63-64.
quantifiers,
by cxistential
with transI)arii, rhcsis, p. 9r i perfect ryi"gir;-t, p. 44; formulated by Aristotle
poseclpremisses,p. 34, n. 5.
3.4',n 3i
l)utisi, axionr, p. BB;-formulated by Aristotle *ith-transposed Pt:Ttt:t:'p'
ol L and
E-tl.firiitions, explained, pp. r63-6; 8-definition of H, p' t64;8'definitions
M, p. r6{l; E-definitionsof X and T, p. t75'
ofdeduction'
<lecisi,,n, the pioblcm of, solved for the Cl.lf-flsystem of the theory
pp. l lr-it|; for the syllogistic, pp. rzo-6'
de<htction of syllogistical laws' pp. 9t-94.
r"l ati ue-i o.-om.-thttes , p' I07; defi ned' p' Ito; di fferent
deducti vc
"q,,i uo1",t.,.,
f.om c,rilina.y equivalence, p. I Io; requires rejection, PP' r09-ro'

214

I NDEX

definitiorx, two ways of defining functors, p, B r in the principia

Mathcmatica,
;
';; pp.
r 63-4 ; in Le3niewski's system, p. r 64r-in the C_,1r'_07-system
;i,
, ,64_6 .
6-definitions.
De Morgan, A., p. r97, n. 3.
deriv ati o n a l I i n e , p . B r .
detachment, rule of, modusponensof aheStoics. p. 16,
determinism, refutarion oll pp. zo7-8.
D-expressions,the method ofverifying, p. r63.
dictum de omni et nullo, not a principle of Jyllogistic, p.
46; not formulated by
Aristotle, p. 47.
Dimaris, thesis,p. gz; proved by Aristotle, p. 26 n.
Disamis, thesis, p. gz; formulated byAristotie with transposed
premisses,p.7 n.i
p:.".9 by him by conversion of the conclusion of Darii, pp.
Sz_SS.
_
uurr
scorus, Iaw or principle ol, pp. Bo, t37, 16z, 165; tis pii"ciptc
^.
is not a
tautology, p. r65.
6uvat6v,possible,p. 134.
.E, constant functor, means ,no-is' or ,belongs to no,, pp.
t4, 77.
Eab, means 'no a is 6, or ,6 belongs to no o',-p.
77.
ecthesis,explained by existential quanrifiers; p. 6i; p.oof.
by ecthesis,pp. 59_67;
perceptual character ascribed to them by Alexander, pp.
6o, ,r. g,'eg,
,_'gi
67, n . r .
"il,
.ElqclopaediaBritannica, r r th edition, on logic of the Stoics, p. 49.
d,v6(xeo0ar,
its ambiguous use in Aristotle,
i. ,gr, ,rn. *_4,
iv\ey6pevov,contingent, p, r34, seccontingency.
E1b ayd NI1b, p. BB; ditreint irom deductive equivalence, p. r r
o.
:::1::t.l"t,:{
,bucud, employs the law of Clavius, p.
5o.
n.4,
I5 2 , ,7 z, r d a n ., r 8 5 , r B7 , r 9 r , r 9 3,2oo, n. r.
Pult
T*,,Po.38,
. r . 3 3 .,
Durerran
dragrars, apphed to a non_Aristotclian systemofsyllogistic,
p. gg; to the
problem ofundecidable expressions,p. ror.
ex is t ent i a l q u a n t i f i e r s , - e xp la in e dp, p .6 r ,- g 4 ; r u le s o ll p .6 z;
used i n prooG by
ccthesis, pp. 6r-66.
cx mcrcncgatiaisnihil sequitur,not generally true, p. l03;
connected with Slupecki,s
rule ofrejection, p. r03.
exportarion, law of, pp. 86, 89, rBz.
exposition, seeecthesis.
expression,significan! p. Bo; elementary, p. r03; simple, p.
ro3.
extensiona.lity,laws of for modal functors,
bp. ,-SS,,,n. ,_3, r39, r43, t47; general
law ol p. tgg; M-law o[ proved by Aiistotie and Uy"af."*^"a.i,
piri.'rf,"1.fac.tor, principle of the, pp.
52-53.
.relapton,.thesis,p. 93 ; formulated by Aristotle with
transposedpremisses,p. g, n. 4.
Ferison, thesis, p. 93.
thesis, p. 93; proved by Aristotle, p. 25, n. 2.
{esa.no,
Fes t ino, t h e s i s ,p . 9 3 ; p r o ve d b y Ar isto r le ,
i- .
;.5 ;;".
figures of the syllogism, division into figures'ha-sa practical
aim, p. z3; description
of the three Aristotelian figures, p. 23, n. r j position
of the middle term in
premissesprinciple of division into
fig-ures,p.
z; Maier,s opinion criticized, pp. 3G-38.
"3, ".
form, of the. Aristotelian syllogism, pp. r-g; of thought,
p. rz; of syllogism as
opposed to its matter, p. r41 consistsolnumber
ind dirposiiio' or.,rl.i"utl.
and of logical constants, p. 14.
_
formalism, pp. r5-r6.

IN D EX

2r5

fourth figure, omitted by Aristotle, p. z7; is moods accepted byAristotle, p. e7;

not invented by Galen, p. 4r ; opinions of Prantl and Maier criticized, pp. 35,
Jl '

four-valued system of modal logic, its primitive terrrr, pp. 167-8; its axioms,
p. 168; is rulcs of inference, p, 168; its adequate matrix, p. 168; some of its
odd coruequences,p. r78; a method of extending it into higher sysrems,pp.
r 7gF8o.
Frege, G., founder of modern propositional logic, p. 48; introduced assertion into
logic, p. 94.
Fresison,thesis,p. 93; proved by Aristotle, p. 25, n. 2.
functorial propositions, have no subject or predicates, p. r32.
functors, of syllogistic, 77; modal, r34; variable, introduced into propositional
logic by Ledniewski, p. r6r; the meaning of the simplest expression with a
variable functor of one propositional argument, pp. 16l-2.
Galen, divided compound syllogisms of four terms into four figures, pp. 3B-4o.
Gerhardt, p. t5r, n. 3.
Gohlke, P., his hypothesis concerning the composition of the Prior Analytics, p.
r33, n. r.
Il, sign of alternation, 'either-or', its definition, p. r64; its E-definition, p. 165.
Hermintu, modiFes the Aristotelian definition of thc major term, p.3r, n.3; misunderstands rcjection, p. 70, n. r.
homogeneous term, required by the syllogistic, p. 7,
ritr1, matter of the syllogism as opposed to its form, p. r4.
iroBdilew, term used by Philoponus for substitution, p, B.
hypothetical syllogism, law oll known to Aristotle, p.49, n.4; formulated, p.5r;
in symbols, p. 79.
1, constant functor, 6g2ns '566s-is'or
'belongs to some', pp. r4,77.
Iaa, law of identity, axiom, p. BB.
Iab, means 'some a is 6' or 'D belongs to some a' , p, 77.
identity, Iaws of, syllogistic Aaa and laa, p. 88; propositional, p. 48; principle otl
p. t 49 ; apodeictic principle of r {g ; axioms of the theory of, p. r49; the law otl
analytic, p. r4g; the law of, used by Aristotle in a demonstration, p. r49, n. z.
imnrediate premiss, d.peoosnp6raots, without a middle term betrveen its subject
and predicate, p. 44.
impcrfect syllogisms,moods of the second and third 6gure, p. 43.
i rrrpfi cati on,'i f p, then C ',p.78; defi ned as truth func ti on by P hi l o of Megara,
pp. 83, I46, I 58 ; its relation to the corresponding rule of inference, p. 22.
i rnportati on, l aw ol , pp. B 6, l B l .
irrrfclirrite Jrrcmiss,pp. 4-5i treated as particular, p. 5, nn. r-2.
itrrferrrrrnstr:rlrlepropositions, dvan66e*roq p. 49.
i trrl errrorrrtrrrl rlsyl
c l ogi smsof the S toi c s ,fi rs t, p. l 9; s ec ondand thi rd, p.58.
rrr,leprrrrlr.rrrr., prrxrli r:f independence of the axioms of syllogistic, pp. B9-go.
rrrrxIr tnrrr, r,l A ri stotcl i an l i l rmul ati ons , p. t8, n. r.
Ittl rtrrrr r, rrot ;r ;l roJxrsi l i ()n,p. 2 t.
rrl uul rl y trr,rny-v;rl rr<rl
rnocl alsy s tem,p. rB o.
i l rl rt ptrl i rl ton rrrri ;rl rl cs,Jr. t 7O.
l ol rrn"l l l nl rrr, ;r, .l (rrn l .
A , rrgrt ol r orri rrrrrti orr ';rnrl ', p. 7t|; i ts f< rur-v al uedmatri x , r75.

INDEX

216

Kalbfleisch, K., P. 38.

K ant , I . , p . l 3 z .
K app, E . , p . t , n . r l c r i t i c ize sPr a n tl, p .3 , n .6 .
Keynes,J. N., on singular propositions, P. 5, n. 3i o-n the major and minor term,
'
p. g6 n. ; on reduction of syllogisms to the first 6gure, p' 44i on dictum de omni
ct nullo, P. 47.
Kochalsky, p. 59, n. I.
'p and q', P. 78; its definition by C and ff' P' Bt ; defined
fpa, conjunitioi,
-.uttt P. 83'
as truth function,
Z, constant functor, means'it

is necessarythat', P. I34; its matrix in the four-

I491
n ' z;itsa n a lyticch aracter,p'
A ris t o t l e b u t n o t s t a t e d e xp licitly,p .r 4 g ,
.double contingency'' p' I78; of contradiction and excluded middle for
of
X-contingency'and f-contingency, p. 176' , Leibniz, C. W., tti. arithmetical interpretation of the syllogistic, pp' I z6-9 ; quotes
a formulation of the principle of necessity, p' r 5 r '
Leiniewski, S., a thesis of his prbtothctic, p. I 56 ; introduces variable functors into
proporitiorral logic, p. Ibr; his rule for,verifying expressions.with -variable
ir,niao., of prop6sitional arguments, p' I63; his method of writing definitions,

P . 2o 5 n .

IN D E X

2r7

does not understand the implication 'if not-p, then p', p. 50; accepts AIexander's interpretation of proofs by ecthesis, p. 6o, n.4; does not understand
proofs ofrejection, p. 68.
majoi term, predicate of the conclusion, p. 3z ; wrongly dehned by Aristode, p. zB,
n. l; Aristotle's definition modified by Herminus, P. 3I, n. 3; Alexander's
opinion on this subject untenable, pP. 3t-32; classical definition given by
Philoponus, p. 32, n. 2.
material implication, defined by Philo of Megara, pp. t46-7.
matrix, two-valued, for C--lf7-system, p. r 58 ; four-valued, for same' p. t 6o I twovalued, for the four functors ofone argument, p. r63; four-valued, adequate,
for C, N, M, L, p. r 68 ; four-valued., for W, p. r 7z ; four-valued, for K, p', 75 ;
four-valued, for X and T, p. t76; eight-valued, for C, N, M, P- ,79.
matrix method, explained, pp. I58-6o; known to Lukasiewicz through Peirce and
Schrtider, p. r66; method of 'multiplying' matrices explained, pp. I5gF6o'
Meredith, C. A., on number of figures and moods for n terms, P' 42 ; on extended
systemsofthe propositional calculus, pP. t6o' I62 n'
middle term, wrongly defined by Aristotle for the first figure, p. 28, n. r ; rightly
defined for all figures, p. 29 n'
minor term, subject of the conclusion, p. 3a; wrongly defined by Aristode, p. e8'
n. z 1 classicaldefinition given by Philoponus, p. 32, n. 2.
M-law of extensionality, stronger, enables us to establish the theory of syllogisms
with possible premisses, p. rgz.
modal functions, p. rZ+.
modal firnctors, p. I 34 ; different from any of the four functors of the two-valued
calculus, p. I66; alt combinations of, reducible to four ineducible combina'
trons, p. I79.
modal logic, of propositions, presupposed by any modal logic of terms, p. 133 ; its
fundamental formulae, pp. t34-5; two scholastic principles of, pp. 135-6;
basic, p. I37; four-valued system ol, developed, pp. 166-9; three-valued
system oq unsatisfactory, pp. 166n., 167; eight-valued system of, outlined,
p. r79; infinitely many-valued system oll p. I8o'
modal syllogistic, less important than assertoric syllogistic, p. r8r; contains mistakes, p. r33; should be rebuilt, p. eoI.
modusponens,frrst indemonstrable of the Stoics, p. r 9; rule of detachment' pp. I6, 8I.
moods, with two apodeicticpremisses, pp. tBI-3; with one apodeictic and one
assertoric premiss, pp, 183-6; with possible premisses, neglected in favour of
moods with contingent premisses, P. r9l ; with one problematic and one
apodeictic premiss, yielding apodeictic conclusions, p' t93; with contingent
premisses, not likely to find a useful application, p. eot ; with problematic
premisses, a method of correcting them, p. zo4; obtained by complementary
conversion, must be rejected.
Mutsc*rmann, p. 5g, n. t.
.rtf,sign of ncgation 'it is not true that'or 'not', p. 78.
ne( cisilry connexions, of propositions, pp. I43-6 ; of terms, r4B-9.
rrcrcsity, its relation to possibility expressed symbolically, p. I35; simple and
r orxl i ti orral ,pp. r44, n. t, t5r-2; hy potheti c al ,p, l 5z ; A ri s totl e' spri nc i pl e of,
pl,. rr,l 4; prirrciplc of, interpreted as rule, pp. I5z-3; Aristotle's views on,
rlirrrrtrorn l'rrr plrilosophy, p. 2o5; Jz syllogistic necessity,
rregnti orr,prol xrri ti on:rl ,tl t:noted by oJ X ,i bythe S toi c s ,p.78, n. r.
ncgnti vr terrri , cxi hrrl crl by A ri s totl e from s y l l ogi s ti c ,p. 72.
rrrrrrrl rcr,l ryl krgi rti t' l i rrl ns :rnd v al i d moods , p. 96.
rrrrrrrllcrol rrrrrlctirlrrlrlc cxllrcssions,infinite without Slupecki's rule, p. ro3.

2rB

I NDEX

number ofvalid moods and figures for n terrns, p. 42.

O, constant functor, means 'some-is not' or 'does not belong to some', PP. r41 77.
Oab, rneans'some d is not b' or 'D does not belong to some 4', P. 77.
Ockham, his laws, p. Ig7, n. 3.
order ofpremiss$, pp. 32-34; not fixed by Aristotle, pp'32-34.
oriXi, propositional negation of the Stoics, p. 78' n' I.
Oxford Translation of Aristotle's works, p. vii.
particular, premiss, p. 4; quantifier, secquantifiers.
Peano, G., p. 52.
peioremsequitursemperconclusioparlem' Pp. lB4, lg3.
Peirce, C. S., invented a method of verifying theses of the theory of deduction,
pp. 8 z , I 6 6 .
perfect syllogisms, moods of the first figure' pp. 43-45.
Peripatetics, a syllogism used by them, p. r ; on relation of Iogic to philosophy,
p, I 3 n . ; n o t f o r m a l i sts,P. 1 6 .
Philo of Megara, defined implication as truth function, pP.83 n., I46-7' t58'
Philoponus,John, on importance ofvariables, P'B' n. 3; usesrlzopritr)ervto denote
substitution, p. 8; his definition of the major and the minor term, p. 32, n. 2;
the second figr:re has a major and minor term by convention, P. 32, n. 3.
Plato, his supposed influence on Aristotle's logic, pp. 6, eo5; examples of compound
syllogisms, p. 4o.
Platonists, on relation oflogic to philosophy, p. I3.
possibility, its relation to necessity expressedsymbolically, p. I35; in the fourpp. t67, t7z;
valued system of modal logic, represented by'twin'functors,
their four-valued matriccs, p. r72; their use for defining contingency, pp.
r 75-6.
Prantl, C., criticized by Kapp, p.3, n.6; does not distinguish the Aristotelian
syllogism from the traditional, pp. 22, 35; his mistaken opinion on the fourth
figure, p. 35, nn. I, 3 ; his ignorance of logic, pp. 35-36 ; quotes Averroes, p. 38.
predicate, together with subject matter of the syllogism, p. I4; put by Aristotle in
the first place in abstract syllogisms, p. 3; predicate of conclusion : major
tefm, p. 3z; prejudice that every proposition has a subject and a predicate,
P 'l3 I '
premiss, defined by Aristotle, p. 3; divided by him into universal, particular, and
indefinite, p. 4.
primitive terms, of the syllogistic, p. 45.
Principia Mathenatica, by A. N. Whitehead and B. Russell, pp. 48, 50, n. 2, 5I, n. 2)
52, n , I , 5 6 , n . z , 6 I n ., 1 6 3 , 1 6 5 .
principle, of division of syllogisms into figures, p. z3 ; of identity, apodeictic, must
be rejected, p. rgo; of tautology, p. 165.
P rior, A . N . , p . r 7 r n .
proof, Aristotle's theory of prbof unsatisfactory, p. 44; Proofs of syllogistic moods
by conversion, pp. 5r-54; by reductioadinpossibilc,pp. 54-59; by ecthesis,pp.
59-67; how proofs should be performed by reductioadimpossibile,p.56; proof
ofdecision for the theory ofdeduction, pp. I I z-r8 ; for the syllogistic, pp. I zo6 ; of IJaw of extensionality, p. r 39 ; proof of CNLNpMq, pp. I 4r-z ; proof of
Cpp in the C--iV-6-1-system, pp. 16z-3; proof that no apodeictic proposition is true, pp. r6917o; proofofmoods with one apodeictic and one assertoric
premiss, pp. t8B-9.
proposition, np|raoc ofthe Peripatetics, p. 3; d.{iupa of the Stoics, p. Bz n.;
Alexander on the difference of categorical and hlpothetical propositions,

IN D EX

219

p. rg2 n.; functorial propositions havc no subjects or predicates, p. r32;

apodeictic, p. r34; problematic,p. rZ4; assertoric,p. r34; analytic, definition
and examples ol; p. r49.
propositional function, pp, 94-95.
Q, sign of equivalence, p. lo8; means 'if and gnly if', is employed instead of the
usual '.8', p. r35, n. 5.
quantified expressions, explained, p. 84.
quantifiers, universal denoted by I/, existential or particular denoted by J, p. B+;
rules ofexistential quantifiers, p.6z; rules ofuniversal quantifiers, p. 86;
universal quantifiers correspond to the syllogistic necessity, pp. ll, 87;
existential quantifiers may explain proofs by ccthesis, pp. 6r-66; universal
quantifiers may be omitted at the head of an asserted formula, p. r45.
Quine, W. V., on consequencesof the apodeictic principle of identity, p. r50 n.,
his example of the difficulty resulting from the application of modal logic to
the theory of identity, p. r 7 r ; solution of the difficulty, pp. r 7 r-2,
Rd rule allowing to replace NI by E and, conversely, p. 88.
nduclio ad absurdum, see reductio ad impossibile.
rcductio ad impossibilc,characterized by Aristotle, p. 55 n.; proofs by, pp. 54-59;
unsatisfactory for Baroco and Bocardo, pp. 54-55, rBz.
reduction of a:tioms to a minimum, has a predecessor in Aristode, p. 45.
reduction of syllogistical moods to the first figure, means proof, p. 44; Keynes's
opinion criticized, p. 44.
reduction to elementary expressions, in the theory ofdeduction, pp. r r r-r5; in the
syllogistic, pp. r rB-zo.
rejected expressions, denoted by an asterisk, pp. 96, 136.
rejection, used by Aristotle by exemplification through concrete terrns, p. 67, n, z;
a rule of rejection stated by him, p. 70, n. e ; its meaning explained, p. 96; its
rules, pp. 7r-j2, 961' how these rules work, pp. 96-97; reasons for its introduction into the theory ofdeduction, p. ro9.
RO, rule allowing to replace NA by O and conversely, p. 88.
Ross, Sir David, pp. vii, viii, 8, n. t, 24n., 46, n. r, 41, n. 2, r54, nn. r-2, rB5,
n. 5, r9r, n. r, r95, n. 2, 2o3, n: 2, 2o4 n.
R.9, Slupecki's rule of rejection, p. r04.
rule, 'c, therefore it is necessary that c', acceptcd by some modern logicians, p. I 53.
rule for the verification of8-expressions, p. r63.
rule of detachment-modus poncnsof the Stoics, pp. 16, 19, Br,
rule ofSlupecki, formulated, pp. 75, r03; explained, p. lo4; employed, pp. ro5-6.
rule ofsubstitution for variable firnctors, explained, pp. r6r-2.
rulcs of infererre, different from propositions, p. 2r I for asserted expressions: by
substitution, pp. 8o, 88; by detachment, pp. Br, BB; for rejected expressions:
by substitution, pp. 72t 96; by detachment, pp. 7r, 96.
l{rrssr:ll,l}., p. r, n. r; wrongly criticizes Aristotle, p. r, n. 3; see also Principia
Mathunatica.
Sr lrolz, l l., p. ix; on Galcn's authorship of the fourth figure, p. 39.
S r l rl i i rl rr, l ,).,p. r66.
rc:rf i gl rl , l )l ). rr)2, 155r,r75, r78,207-8.
Sextrrs lirrrpirit:us, (luotcs a Peripatetic syllogism, p. r, n. 2; gives the Stoic proof of
tlrr <orrr;xrrrrrrllaw of transposition, p. 59, n. r ; guotes Philo's definition of
i rnpl i c:rti on,p. l i 3 n.
S i crpi rl ski ,W., p. l r5.

22O

INDEX

significant expression, defined inductively, p. Bo.

simple expressions of t}re syllogistic, rejected, pp. r2o-t.
simplification, law of, p. 89.
singular terms, defined by Aristotle, p. 4, n. z; why omitted in his syllogistic,
PP. 5-7.
Slupecki,J., proves that the number ofundecidable expressions ofthe syllogistic is
infinite, p. I o r ; states a new rule of rejection, p. I o3 ; shows that the Leibnizian arithmetical interpretation ofthe syllogistic verifies his rule, p. rzB n.; his
paper quoted, p. 76 n.
Solmsen, Fr., his view on conversion ofthe conclusion refuted, p. 25, n. t.
square of opposition, not mentioned in the Anal2tics,pP. 20, 45.
Stoics, on exchange of equivalent terms in syllogisms, pp. I8, l9 n.; their logic
formalistic, p. rg; their logic a iogic of propositions, pp. 48, zo5; a system of
rules of inference, p, 48; misunderstood by modcrn commentators' p. 49;
denote variables by ordinal numbers, p. 58, n. 4; use oJTi as propositional
nation, p. 28, n. r ; adopt Philo's definition of implication, p. 83; state the
principle of bivalence, p. 8z n. ; modusponerc, the first indemonstrable syllogism of the Stoics, p. Ig; the second and third indemonstrable syllogisms,
p. 58; their proof of the compound law of transposition; the logic of the
Stoic-Megaric school well known to Alexander, p. I47.
otoqetc, letters, variables, p. B.
strict implication, p. r47.
subject, together with predicate matter of the syllogism, p. t4; put by Aristotle in
the second place in abstract syllogisms, p. 3; subject cif the conchsion:
minor term, p. 3z; propositions without subject or predicate, pP.44, r3r.
substitution, an ancient argument by substitution, p. r o ; term used for substitution
by Philoponus, p. 8, n. 3 ; rule of substitution for asserted expressions, p. 8o;
for rejected expressions, pp. 72, 96; for E-expressions,pp. 16l-2.
substitution-variables, distinct from interpretation-variables' p. l70.
syllogism, a Peripatetic, p. r ; in concrete terms given by Aristotle, p. z ; form of the
Aristotelian syllogism, pp. I-3; different from the traditional logically and in
style, p. 3; differently formulated in variables and in concrete terms, P. t7;
compared by the Stoics with an arithmetical law, p' I5; in purely implicational{orm, pp. zz, l8c; in symbolic form, p. 78; modal syllogisms dealt with
by Aristotle after the pattern of his assertoric syllogisms, p. lBI.
syllogistic necessity,its sign sometimesomitted by Aristotle, P. Io, n.5; its meaning
explained on occasion of the invalid conversion of the O'premiss, p. r r ;
wrongly explained by Maier, pp. II-I2l corresponds to a universal quantifier, p. r r ; proof of this correspondence in symbolic form, pp. 86J7 ; can be
eliminated from syllogistic laws, pp. t44-5.
symbolic notation, without brackets, pp. 78-79.
synthetic theorem, ascribed by Alexander to Aristotle, p. 65 n.; in symbolic
form, p. 85.
f, constant functor, means 'it is contingent that', p. 154; not suitable for the purpose ofinterpreting contingency in Aristotle's 6ense' p. 199.
Tarski, A., pp. 78, n. 2, ro7 a.
tautology, principle of, p. 165.
term, part of a premiss, p. 3; universal, singular, empty' P' 4; different from
Begri.f,p.3, n. 6; a division of terms, pp. 5-6; syllogistic requires homogeneous
terms, p. 7; major, minor, and middle term, pp. 2B-3o'
Tluodicee,by Leibniz, p. I5r.
Theophrastus, adds the moods of the fourth figure to the first, pp. 27, n. 2, 38, n. 4 i

IN D EX

221

probably defined the first figure differently from Aristotle, p. z7; makes
conections to Aristotle's modal syllogistic, p. r 33 i on the meaning of necessity,
p, r5r, n. z1 makes explicit the distinction between simple and conditional
necessity, pp, r5I-2; his doctrine concerning moods with mixed premisses,
pp. I84 n., I85, IB7-8, rgr ;his peioremrule violated by a modal mood, p. r93;
acceptstheconvertibilityofuniversally-negative contingent propositions,p. zoo,
nn. I-4.
theorem of reduction, proved for the theory of deduction, pp. I I l-r5; for syllogistic, pp. r r8-zo.
theory ofdeduction, the most elementary part ofthe logic ofpropositions, pp. 49,
7gr83; invented by the Stoics as a system ofrules ofinference, p. 48; founded
in modern times by Frege, p. 48; placed at the head of mathematics in Principia Mathematica,p. 48 I reasonsfor introducing rejection into this theory, p. I o9.
theory of identity, axioms of, p. r49; difficulties resulting from the application of
modal logic to the theory ofidentity explained, pp. r70-r.
theory of probability, may have a link with modal logics, p. rBo.
therefore, sign of inference, pp. 2, 2I.
Oiors,order ofterms adopted by Aristotle for the three figures, p.33, nn. 3-5.
thesis, true proposition of a deductive system, p. zo; different from a rule of inference, p. zr; relation ofan implicational thesis to the corresponding rule of
inference, p. ez.
Thomas, Ivo, O.P ., p. r4 g, n. 2.
traditional syllogism, a rule of inference, pp. 2 r -23 ; different from the Aristotelian,
p. zr; neither true nor false, only valid or invalid, p. zt; weaker than the
Aristotelian syllogism, pp. 22-29.
transposition, law of, known to Aristotle, p.49, n.3; its symbolic form, p. Bg;
compound law of transposition, proved by the Stoics, p. 59, n. I.
Trendelenburg, F. A., does not distinguish the Aristotelian syllogism from the
,
n,2; on the pri nc i pl e of
tradi ti onal , p.22; on the order of premi s s esp.33,
division of syllogismsinto figures, p. 36.
twin contingencies, p. t76.
twin necessities,p. r7+,
twin possibilities,explained, pp. r72-4.
Ueberweg, Fr., pp. 36, 39.
undecidable expressions,p. rclo; infinite in number, p. ro3.
universal premiss, p. 4.
universal term, p. 4.
unumquodque,
quandocst, oporletcsse,a principle of necessity, p, I5t.
ulraquesi praemissanegetnil inde sequctur,connected with Slupecki's rule of rejection,
p.ro3'
V :ri l ati , C ., p. 50, n. 4.
v;rl i rl i ty, propcrty ofi nfere nc es and rul es ofi nferenc e, p. z I.
vurinlrlr.s,intrtxlucecl into logic by Aristotle, pp. 7-B; truth of syllogismsdoes not
rl eperrrlorr shap<'ofva ri abl es p.
, g, n. z ; i denti fi c ati onofv ari abl es not k now n
to A ri stotl r',p. ry; thei r ex tens i onalrel ati onsc annot be determi ned, p. 29.
vcti l i r,rti otr ol ;i -rxl )rcssi ()nsex
, pl ai ned, p. I63.
utrum vqnitur arl '1uillihl, p. r79.
vorr Wri gl rt, (l l l ,, l ). I.)1 l l .
l /, r orr:t.i l rt l i rrrrtrrr, i ts l l rrr-v ;rl rrc rlmatri x , p. r7e; i ts rel ati on to i ts tw i n func tor
Il , pp, r 72 .l i rl $ r,rl c i rr tl c l i ni ng c onti ngenc y ,pp. t75-6.

thetradisv'es1sm-rro1
*, .igJl?1lr,t:"*
iil,.,,*., p.vii;adocs
for changing the
Apuleius
censuretr
n'
p'
24
;
criticisi,
to<tuai
tional, p. ze ;
order of Premisse,P. 33' n' l'
Wallies, M., P. 39.
fi-,, seePrircifu Matlunatica'
i{nit.tt."a,-4.

x.constantfunctorritsfour-valuedmatrixrp.IT6;itsMefinition'p'r75;its
-' -i.r.tio"
to ia twin functor f ocplaincd' PP' 175-7'
f-constantfirnctor,itsfour-valuedmatrix'P'q')its&definition'p'r75'its
-' -*r"ti""
t iu trrin functor x ocplained' pp' t75-7'
Zcllcr.,E.'P.49.