A Proof in the ΠΕΡΙ ΙΔΕΩΝ

Author(s): G. E. L. Owen
Source: The Journal of Hellenic Studies , 1957, Vol. 77, Part 1 (1957), pp. 103-111
Published by: The Society for the Promotion of Hellenic Studies

Stable URL: https://www.jstor.org/stable/628641

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

The Society for the Promotion of Hellenic Studies is collaborating with JSTOR to digitize,
preserve and extend access to The Journal of Hellenic Studies

This content downloaded from

193.225.200.92 on Tue, 13 Oct 2020 09:31:16 UTC
All use subject to https://about.jstor.org/terms
 A PROOF IN THE HEPI IAEQN
IN his lost essay 7TEpt 1EC-ov Aristotle retailed and rebutted a number of Academic arguments
for the existence of Ideas. Several of these, together with Aristotle's objections to them, are
preserved in Alexander's commentary on A 9 of the Metaphysics. The first object of the following
discussion is to show the sense and the provenance of one, the most complex and puzzling, of these
surviving arguments. For several reasons it seems to deserve more consideration than it has yet
had.' i. Its length and technicality make it singularly fitted to illustrate the sort of material on
which Aristotle drew in his critique. 2. Moreover, Alexander reports it by way of amplifying
Aristotle's comment that, of the more precise arguments on Ideas, o( pv C - ov 7rpd6 07rC rotoi3Lv
1ag, Lv o005 falEv Etvat KaO' av7o yEvos (Met. 99obI5-17 = 1079aIi-I3); and the condensed and
allusive form of this remark and its immediate neighbours in the Metaphysics can be taken to show
that here Aristotle is epitomising parts of his 7rEpt 1IE&Ov that are independently known to us only
through his commentator. We shall not understand the objection if we misidentify its target;
and another purpose of this discussion is to show that the objection is not the disingenuous muddle
that one recent writer labours to make it. 3. But Alexander's report of the argument is a nest of
problems, and the same recent writer brands it as almost incredibly careless. To this extent,
the success of our explanation will be a vindication of the commentator. But on all the heads of
this discussion I am well aware that much more remains to be said.

THE PROOF

In the authoritative text of Alexanderz (which, with a minor emendation of Hayduck's,3

David Ross prints on pp. 124-5 of his Fragmenta Selecta Aristotelis) the specimen argument
produces 18E'as -rwv rpdos - is given as follows.

I. When the same predicate is asserted of several things not homonymously (tk0 d/1wviV'w
but so as to indicate a single character, it is true of them either (a) because they are stri
(KvUpwS) what the predicate signifies, e.g. when we call both Socrates and Plato 'a man
(b) because they are likenesses of things that are really so, e.g. when we predicate 'man' of m
in pictures (for what we are indicating in them is the likenesses of men, and so we signify a
identical character in each); or (c) because one of them is the model and the rest are likenesse
e.g. if we were to call both Socrates and the likenesses of Socrates 'men'.
II. Now when we predicate 'absolutely equal' (-r Ziov a;d') of things in this worl
we use the predicate homonymously. For (a) the same definition (Adyos) does not fit th
all; (b) nor are we referring to things that are really equal, since the dimensions of sens
things are fluctuating continuously and indeterminate. (c) Nor yet does the definition
'equal' apply without qualification (dKptPfls) to anything in this world.
III. But neither (can such things be called equal) in the sense that one is model
another is likeness, for none of them has more claim than another to be either model or likene
IV. And even if we allow that the likeness is not homonymous with the model, the c
clusion is always the same-that the equal things in this world are equal qua likenesses of what
is strictly and really equal.
V. If this is so, there is something absolutely and strictly equal (artL i- aVh'tauov K
KvpIwS) by relation to which things in this world, as being likenesses of it, become and
called equal. And this is an Idea. (Alexander, Met. 82. 11-83. 16 Hayduck.)
I shall refer to this report of the argument in the 7TEpt 13Ecv as P. Its gist, if not its detail
seems clear. What is allegedly proved, for the specimen predicate 'equal', is a doctrine fam
i It has been discussed by Robin (who first assignedtextit of our passage in a clumsy attempt to evade the
to the 7rept i1de&v), Theorie platonicienne des Idees et des
difficulties discussed infra, pp. io4-6. (But notice that,
Nombres, 19-21, 603-5, 607; Cherniss, Aristotle's Criticism
where A uses Socrates and Plato as examples, LF at first
uses Callias and Theaetetus, reverting then to those in
of Plato and the Academy, I, 229-33, esp. n. 137, and Wilpert,
A.) On Robin's attempt (1.c.) to assign LF equal autho-
Zwei aristotelische Friihschriften, 41-4, each of whom knew
only Robin's discussion; and Suzanne Mansion, 'La
rity with AM see Wilpert, n. 38, Cherniss, n. I37-
critique de la th6orie des Idles dans le i7ept i&deiv d'Aris- 3 Cf. p. I07, n. 26 infra.
4 'We': not of course the Platonists, who make no such
tote', Revue Philosophique de Louvain, xlvii (I949), i81-3, esp.
n. 42. I shall refer to these writings by the author's name.
error, but generally the unwary or unconverted to whom
2 The A of Bonitz and later edd. The version of the argument is addressed. The objector envisaged at
the commentary in L and F excerpted in Hayduck's Phaedo 74b6-7, and Hippias (Hipp. Maj. 288a and 289d),
apparatus is later in origin (Hayduck, Alexandri insee
Met.
no objection to using aw3r6 -r6 ~aov and a?i6d td KaAdv
Commentaria, pref. viii-ix and ix, n. 2). It modifies
of the
sensible things.

This content downloaded from

193.225.200.92 on Tue, 13 Oct 2020 09:31:16 UTC
All use subject to https://about.jstor.org/terms
 104 G. E. L. OWEN
from several Platonic dialogues: things in
virtue of resembling a Paradigm that c
73c-75d is especially obvious. Both arg
and prove that what it describes is no ph
of our argument is clear. As we shall see
takes pains to sharpen the logical issues t
must be intended as an exhaustive anal
ambiguity.5 Now it is Alexander's repo
seems plausible to say that the author
illustrated in I(c) as non-homonymous,
the other hand that he cannot have reg
in the sense of that expression required
non-homonymous must be a confusio
understanding the argument and establis

CRITERIA OF SYNONYMY IN ARISTOTLE AND PLATO

The difficulty in I(c) seems both logical and historical. We may say 'That is a man' without
ambiguity when pointing to each of two flesh-and-blood men. Or (in a very different case) we may
say it when pointing to each of two pictures, and what we say has the same sense of both pictures:
in that respect we are still speaking unambiguously. But we are inclined to add that now we are
not using the predicate in the same sense as in the first case: otherwise we should be mistaking
paint and canvas for flesh and blood. Moreover this is Aristotle's view, and his examples suggest
that he has our argument in mind.6 Yet, as it stands, I(c) says just the opposite. The analysis
seems to have distinguished cases (a) and (b) in order to assert with all emphasis that a combination
of them in (c) imports no ambiguity at all.
The later version of the scholium (supra, n. 2) takes a short way with the difficulty, reclassifying
I(c) as a case of homonymy. Robin (n. I) tried to wrest this sense from the original text; Wilpert
(l.c.) rejected the attempt but regretted the anomaly. Yet the problem is fictitious. The logical
issue can only be touched on here. The fact is that, although the difference between I(a) and
I(b) predication does show an ambiguity of an important type, this is not the sort of ambiguity
that can be exhibited by the methods of Aristotle and the Academy.7 It no more proves that the
predicate-word has two paraphrasable meanings than the fact that I can point to a portrait and
say 'That is Socrates' proves that Socrates had an ambiguous name. This is true, but it is doubtful
whether it is the point that our author is making. For the wording of I(b) suggests that in its
derivative use the predicate is to be paraphrased otherwise than in its primary use (i.e. in terms
of 'likeness'), though this difference of paraphrase does not constitute an ambiguity. Similarly
we shall find (infra, I o9-I o) that the argument of II can be construed as allowing, with one proviso,
that a predicate can be used unambiguously of several things even when the Aoyor of that predicate
differs in the different cases; the proviso is that the different Ao'yo shall have a common factor.
(In the cases distinguished in I this factor is the primary definition of 'man', and in II it is the
definition of ~' 't'aov a~dr.) If this interpretation is correct our specimen of Academic argument
containsare
which anin
obvious
a senseparallel to Aristotle's
synonymous admission of a class
(Met. ioo3a33-Ioo3bI5, cf. of rrp01
Eth. Eud.Ev1236ai5-20,
Ka ~ ltav va'
andYan.7 37
tlV AEydOlEva
infra) .
But Aristotelian parallels are irrelevant to showing the reliability of P. What matters is that
the analysis in I would misrepresent its Platonic sources if I(c) were not a type of unequivocal pre-
dication. This is implied by the reference in Republic 596-7 to a bed in a picture, a wooden bed
and the Paradigm Bed as 7-p-i-al KAZvaL (even when, as in P.I, only one of these is 'really' what
the predicate signifies); and more generally it is implied by such dicta as that nothing can be just
pls 6ut d y)upwy in the Aristotelian sense but not, as we Phys. H. 248bi2-21: H. neglects such passages in detect-
shall see, using Aristotelian criteria. Some will detect ing a book of Speusippus behind Topics A 15). More-
the influence of Speusippus in P.I, noticing that in
it the vehicles of homonymy and its opposite seem to be over
that in
weP.can
III say
the is6djwbvvtia are things,
that P reflects not words.
a general All
academic
not things but words, and that this is held to be charac- usage.
teristic of Speusippus by contrast with Aristotle (Ham- 6 De Part. An. 64ob35-641a3, De An. 412b2o-22, and
bruch, Logische Regeln der plat. Schule, 27-9, followed by on the traditional interpretation Cat. IaI-6 (cf. Porphyry,
other scholars including Lang, Speusippus, 25-6. Ham- Cat. 66.23-28, followed by later commentators, and see
bruch contrasts Aristotle, Cat. IaI-i2, with Boethus's earlier Chrysippus fr. 143 (von Arnim). But Ccoov, the
account of Speusippus in Simplicius, Cat. 38.19). Quite predicate cited, is ambiguous in a more ordinary sense:
apart from doubts about the tradition represented by LS8 s.v. II).
Boethus, it is clear that Aristotle's usage is far from being 7 For a connected discussion I can refer now to P. T.
as rigid as Hambruch supposes (see e.g. An. Post. 99a7, 12, Geach in Philosophical Review, LXV (1956), 74-

This content downloaded from

193.225.200.92 on Tue, 13 Oct 2020 09:31:16 UTC
All use subject to https://about.jstor.org/terms
 A PROOF IN THE HEPI IAEQN 105
or holy or beautiful if the corresponding Form is not so.8 These utteran
the predicate applies without difference of meaning to model and liken
integral to the doctrine that things in this world resemble the Forms. T
found the latter doctrine in his chief source (Phaedo 73c-4e) and remarked th
by the relation between Simmias and Simmias yEypappdE'vos (73e), and in
do no more than put his original into precise logical shape. We recall J
Aristotle did this very service to Plato in the Eudemus. But we had better de
the authorship of our proof.

To tov avrod AND 7r; 'toov

A second puzzle turns on the three occurrences in P of the key-word
tinguishes three possible cases in which a predicate can be used jpq dcowv
paraphrase to mean 'not ambiguously'. But P.II then seems to contend th
is used wdCovi5kwsg of things in this world, although the explicit conclusion o
of the dialogues on which P is based prove that such predication would
Lastly, P.IV puts the case that the likenesses carry the predicate non-ho
model, which squares with I but seems incompatible with II. In fact P.
again the later version in LF takes the short way, replacing the 4uowv`t4ws
odV KvUplwS 8 so as to bring the predication in question clearly under I(b). R
argument (l.c.), which covertly reduces it to a petitio principii and contradic
has been criticised by Cherniss (l.c.). Mlle Mansion (n. 42) has seized th
P.II is concerned not with iT 'toov but with iT 'toov a'7Td, but I have not un
the argument is a reductio ad absurdum and I do not agree that IV is an inte
not considered the problem.
Cherniss has propounded a singular solution (n. I37). He holds that oy
used in the same sense throughout P; and accordingly he claims that in II
warning in a Platonic sense, such that the Platonic d[twvaotwso is compatible
[4-j tkUwvtkUws in I (which he at once denounces as a 'careless summary' by
The Platonic sense is identified as 'having the common name and nature
effect is exactly that of the verbal change in LF. But he is then faced with
On his interpretation this cannot contradict the other occurrences of the ex
plausibly let himself say that it is a return to the 'Aristotelian' sense 'in the
Consequently he has to provide a different Platonic sense, equally unad
whereby t cuj 6xcoivvov in IV signifies that 'the image is not of the same clas
in order that the use of ctowvit[ws in the first 'Platonic' sense shall be comp
p7) uotvvoov in the second 'Platonic' sense and both of these compatible with
in the original 'Aristotelian' sense. In face of this it is easy to sympathis
the t in the third occurrence must be an interpolation.
On the canons of this interpretation I have something more to say,
reviewed the problem. A closer reading of the text seems sufficient to
maintained in II is that iT 'tov ai3'T would be predicated homonymously
and - 'toov ad-o' is expanded in V into advtdoov xKa Kvplws (sc. Kvplws '
Al s otoov). Thus the question broached by II is just whether 'aoov can b
in this world, i.e. as a case of the non-derivative predication illustrated i
that, except by a sheer ambiguity, it cannot be so used.Io But this concl
patible with the conclusion in IV and V that 'oov without this qualifica

8 See e.g. the instances cited by Vlastos,

wePhilosophical
are speaking of the standard yardstick and
Review, LXIII (I954), 337-8. But Vlastos areobscures
speaking theof other things (Geach, l.c.), but t
point by saying 'any Form can be predicated of that
not entail itself ...has two meanings. Aristotle
'yard'
F-ness is itself F'. The very fact that Plato could assume
commonly treats the Forms as avvd)'vvya with their images
without question that awTrd T6 o d'ye6Oo is big (e.g.
Phaedo Io2e5, cf. Parmenides x50a7-bx and (cf. de Lin. Insec. 968ag9-o, 7 6' i6 a rpd5r xCv 1avvw~juwv).
The i31d),
objectionwhereas
considered in Physics H 4, that avvd)vvuya
in English such an assumption about bigness makes no
sense, should give us qualms at renderingneed
tothe not be avlflhrAld,
title
safeguard of thesis
this may well
the from stem from
the 'Third Man'. the attempt
Form conventionally in such contexts by an abstract
Io Instead of asking in set terms whether 'equal' can,
noun (Vlastos' 'F-ness'). V.'s formula misleads him into
without ambiguity, be predicated strictly of such things, II
assimilating the two regresses in Parmenides
seems 132-3. If the
to introduce the compound predicate 'strictly
first can (but with reservations) be construed as confusing
equal' and ask whether this can, without ambiguity, be
bigness with what is big, the second requires
predicatedonly that
of such things. This comes to the same thing
the Form should have the character it represents. If the is too hard-edged for the Greek),
(in fact the distinction
first forces a choice between two possiblebut functions of a the author of LF into the absurd
it helped to seduce
Form, the second reduces one of these to absurdity.
notion
9 This is unaffected by the fact that properly that
the Forms theincompound
be used,are
predicate ardthov could
a derivative sense, of earthly things.
standards. 'That is a yard long' has a different use when

This content downloaded from

193.225.200.92 on Tue, 13 Oct 2020 09:31:16 UTC
All use subject to https://about.jstor.org/terms
 So6 G. E. L. OWEN
unambiguously of a group including phy
by the derivative sort of predication sh
prove that, if 'equal' keeps its proper sen
III proves the corollary, that no group o
mixed, I(c) predication (which would enta
is not even considered in II and III is wheth
as in I(b).
Now IV is concessive in form,Iz and what it concedes is just this third possibility. (Its form
does not of course mean that it is surrendering any part of the argument. It is concessive because
it forestalls an objection: the objection that the talk of ambiguity in II is misleading and may be
taken to apply to 'C'rov, not r 't~uov ai3vT.) And, in fact, I(b) predication is the only possibility
still open to us if we are to keep any unity of sense in our everyday ascriptions of equality. But
copies entail models, and this conclusion requires that i- toov is predicated Kvpl'g of something
not in this world, of which this world's instances of equality are likenesses.
But, finally, IV is only a concessive parenthesis, and it implies (dEtl FrErat) that the same result
would follow from II and III alone. So it does: for II maintains that when we talk of what is
KUVpLWS 'OV, what we are referring to (unless the expression is being used ambiguously)
be anything in this world. It follows that, unless we call everyday things equal in som
unconnected with the first, they must be so called derivatively. And since this conclus
explicitly drawn in V, II, III and V form a complete argument.
So the form of P is clear and its use of the terminology introduced at the start is, as we m
expect, consistent. But it is worth noticing two other considerations which are jointly f
Cherniss's account. The gross carelessness of which he accuses Alexander is out of charact
has not remarked that, when the commentator does introduce d1kvvkos in the non-Arist
sense, he takes pains to explain the ambiguity.13 Moreover, apart from all particular que
of interpretation (but see nn. 15, I9), the evidence adduced by Cherniss for the existence
'Platonic' senses of otkwvvto'14 has no tendency to prove his point; and the reason for this is
emphasis. Plato does use 0r`[Wvvkosg fairly frequently. It seems clear that he does not use it i
technical Aristotelian sense of 'equivocal'. Sometimes (as at Tim. 52a, Parm. I33d, Phdo.
it is applied to cases of what Aristotle would doubtless call synonymy. But it does not for a m
follow that the expression meant for Plato what is meant by Aristotle's uvvovv1tLos, any mor
it follows that because 'soldier' can be applied to all bombardiers, 'soldier' means 'bomba
Elsewhere the same word is used of things that plainly do not have the same Ao'yos -ri o
This should entail for Cherniss that Plato's use of the word was ruinously ambiguous, but of c
it was not. As Plato uses it, what it means, its correct translation, is 'having the same nam
the argument never requires more than this of it (cf., for instance, the versions of Cornford
mistake recurs in Cherniss's further comment that 'for Plato kvvkosg when used of the relation-
ship of particulars and ideas meant not merely "synonymous" in Aristotle's sense. The particular
is o1uvpCLoV 7Ck ' EL&t, not vice versa, because it has its name and nature derivatively from the idea'.
Yet elsewhere the word is used of an ancestorfrom whom the name is derived'6 and elsewhere again
where there is no derivation either way.'7 Nor does Plato reserve any special meaning for the
metaphysical contexts Cherniss has in mind.'8 The fact is that when he thinks it necessary to say
that particulars are like the Form in nature as well as name he says so explicitly ( yvv"uov o"otodv
rE, Tim. 52a5) and when he wants to say that they derive their names from the Forms he says that
too (Phdo. Io2b, Io3b, Parm. I3oe'9). The second 'Platonic sense' of the word rests on the same
basis. 9a
it It may be said (I owe the objection to Mr. D. J. 16 Republic 33ob, Parmenides I26c.
Furley) that the argument in II(a) is designed to rule 17 Protagoras 31 b, cf. n. 19 infra.
out I(b) predication as well as I(a), since even I(b) would 1s That Aristotle, who certainly knew that particulars
presumably, require an identical Ao'yog in the various were 'called after the Ideas' (Met. 987b8-9), did not
subjects. But in that case the conclusion of II would recognise a sense of 6dycdvvyogo in these contexts such that
contradict V, as well as being a thesis foreign to Plato and
never attacked by Aristotle; moreover the difference ofmust 'the particular is 6d1cdvvvyov xc> eldet and not vice versa'
be proved for Cherniss by Met. A 99ob6, which
Adyot reports that the Form is 6ydc)vvtyov with its particulars:
they alldoes
havenot entail ambiguity
a common factor (p.since, as we shall see,
og infra). here Cherniss is ready to find 'Plato's sense of the word'
I2 Cf. Alexander, Met. 86.I 1-12 Hayduck. (n. 1o2).
I3 Alexander, Met. 51.11-15, 77.I2-I3. Cf. ps.-Alex. 19 Not however Parm. I33c-d, which Cherniss has
Met. 500.12-35, 786.15. misread (l.c.): it is not the Ideas that are referred to as
14 Cherniss, n. Io2, citing Taylor, Commentary on Plato's
Timaeus 52a4-5. Ajv 'likenesses-or-what-you-may-call-them'
the 7siCyt vexovTex elvat gKaO-ra ~87ovoyaCdtPeOa but
in this world.
s5 Laws 757b; cf. Phil. 57b, which Cherniss (l.c.) mis-
construes as saying that 'the different mathematics, if Since
to thethe particulars
Forms, are nevertheless
this sentence alone, ifsaid
he to betakes
still 6dtdvvoza
it as
seriously, explodes his thesis.
61idavylov, are a single xiXvq' when the point is that
although they are dOJVvvLya it would be wrong to infer 19a And a misreading of the text cited, Phil. 57b: cf. n.
that they are one x6Xvq (57d6-8). 15 supra.

This content downloaded from

193.225.200.92 on Tue, 13 Oct 2020 09:31:16 UTC
All use subject to https://about.jstor.org/terms
 A PROOF IN THE HEPI IdAEQN I07
But why labour this point? Because the thesis in question seems a particu
tion of one general principle of interpretation, and this principle underlies a we
the 'unity' (in the sense rather of fixity) of Plato's thought, to which Professor
tinguished heir. It is often observed that arguments for this theory assume
one context must carry a special sense determined by its application in quit
And no doubt some of the things to be said in this paper do not square well with

KaLO al AND TrpoS -rL

So far, P keeps our confidence. It remains to discuss it as a digest of Plat
a target of Aristotle's criticism.
On the face of it, P distinguishes two sorts of predicate: those such as 'm
predicated Kvpiws of things in this world [I(a)], and those such as 'equal', wh
are used unequivocally of such things can be predicated of them only deri
all appearance it seeks to provide forms for predicates of the second class by con
those of the first; and we shall see this impression confirmed by other evidence
of the argument. This distinction Cherniss tacitly suppresses in his precis
accordingly able to find 'no reason to suppose that the argument ... was not also
the existence of Ideas in the case of all common predicates'.22 He suggests
rewriting, unless it is (what is in any case no justification) that the similar argu
is said to apply to all things o t E'rctrpaytgdfEla r% av)To 'orf t' (75c-d). But
includes all predicates whatever is to beg the same question. The predicates
as examples-Lroov,
to 'HLEtCOv
which the argument ,Aarrov,
of P applies; inKaAdCV, J'yaodv,
the relevant &'Kacov,
respect ootov-are
they are all of
all, as we sh
congeners of 'equal' and not of 'man'.23 Moreover, the same distinction, whic
argument of P and its sources, is the basis of Aristotle's criticism of these argum
gives the rest of our discussion its starting-point and conclusion.
It has come to be agreed that Aristotle's objection to the arguments whi
relatives' (Met. A 9, 99bI16-17, cf. p. 103 supra) is not of the same form as those
context. He is not arguing that such proofs as that reported in P can be use
that were explicitly rejected by the Platonists. He is saying that their concl
logical principle accepted by the Academy; and the commentary of Alexander
to identify the principle in question. (But Sir David Ross is one scholar who w
this identification (Aristotle's Metaphysics, ad loc.), and in this he is followed by
Aristotle in this and the following sentence of his critique is turning against th
dichotomy of KaO6' 5do' and rrpos k-:24 a dichotomy inherited from Plato and e
not only exclusive but exhaustive, since the school of Xenocrates maintained it a
elaboration of Aristotle's own categories.25 Aristotle is objecting that such
a 'non-relative
such class. class of relatives', a KaOG a5-r yEvos 706V Tpdo TL, and that 'we
The first thing to remark is the wide sense carried by the Academic srpdos ko when measured
by more familiar Aristotelian standards. This seems to have eluded Alexander: hence, perhap
his reference to P as proving 18'&a Ka 7rcdv lrpdc 7- where Aristotle says only 18'as 7wy ^u pdS' L :2
20 I can refer now to Vlastos, op. cit. 337, n. 31; ment
cf. valid for every negatively defined predicate (Met.
Robinson, Plato's Earlier Dialectic (2nd edn.), 2-3- 99obI3: cf. Alexander and Ross ad loc.). Readers other
zi Cherniss, p. 23o. To do this he omits the illustra-than those aoaaCTciat xoi_ 6"Aov are likely to find the
tions of the three types of predication in P.I. Yet (a) comment of D. J. Allan in Mind LV (1946), 270-I,
without the illustrations the analysis is merely formalsound
and and to the point.
without explanatory force; (b) that the predicate cited in24 Alexander, Met. 83.24-26, 86.13-20. The rele-
the first paragraph of Alexander's source was not loaov vance of this dichotomy was pointed out by D. G. Ritchie
and was not a 'relative' term is implied by Alexander's against Henry Jackson: cf. J. Watson, Aristotle's Criticisms
remark that at any rate the proof goes on to deal with qf Plato, 32.
waov, which is relative (83.23-4); and (c) in any case thez5 Sophist 255c-d, Philebus 5Ic, cf. Republic 438b-d,
illustration from portraits cannot be excised since it comes
Charmides I68b-c, Theaetetus I6ob. Xenocrates, fr. 12
from the Platonic source (supra, p. I05). This in addition
(Heinze) = Simplicius, Cat. 63.21-4. I am not con-
to the considerations adduced in the following pages. cerned here with the development and supplementation
22 Cherniss, n. I86. of this dichotomy in the early Academy, which has been
23 Similarly those given to illustrate similar formulae
the subject of recent studies. The subsequent conflation
at Phaedo 76d, 78d, Rep. 479a-d. The one passageof inthe Platonic 'categories' with the Aristotelian, e.g. in
which Plato seems unequivocally to require a Form for Albinus (Witt, Albinus, 62-7), may derive from Aristotle
every predicate (Rep. 596a) cannot be ingenuously cited himself (E.N. Io96a 19-21).
by any critic wedded to the 'unity of Plato's thought' 26 Alexander, Met. 83.I7, 22, 85-7. But the text of
since (even if Parmenides 130 is brushed aside) taken
82.11 (6d /~V 8K TCV 7pd( TZt KaTaCKevdKCov ta Ca6 dyOO)
literally it contradicts Politicus 262a-3e and incidentally
should not be amended, for this comes from the rsept iei5r
and not from Alexander.
leaves Aristotle's criticism of the Ev 1l 7oAA,,v argu-

This content downloaded from

193.225.200.92 on Tue, 13 Oct 2020 09:31:16 UTC
All use subject to https://about.jstor.org/terms
 Io8 G. E. L. OWEN
he seems to have seen that the proof applies, n
outside the Aristotelian category. (He reassur
to prove it his own, that 'anyhow the examp
sense: Met. 83. 23-4.) In any case he is betray
Metaphysics Aristotle argues from the priority
rrpdc 1et (99obi9-2I). Here Alexander repor
a&ptOtuo -rVdT EdgETL, Met. 86. 5-6; cf. Aristo
are, inter alia, Jora ac-ra 7 aEP Er7v -EdpwY EC
was an Academic premiss: Alexander, Met. 7
interpret the anomaly away (86. I I-13);27 an
argument in Metaphysics M Io79a15-I7, whic
here to subsume number under rd irrpds rt as a
Nor are the sources of such a classificatio
numbers are classed with such characteristics a
our senses can never discover any of them K
they are inseparable from their opposites.28
what is one of something is any number of
say, for convenience, that 'one' as we ordinarily
accordingly as we complete it in this way or th
applied. Now the same is true, or Plato talks
Republic and earlier works supply him with
the logical-mathematical and moral-aesthetic pr
postulates Forms in Parmenides 13ob-d. In th
right or pious, is so in some respect or relation
other.29 As large is mixed with small (Rep.
commerce with bodies and actions3o have co
earlier context Plato argues that such seeming
different respects or relations in which the
how various such specifications will be: som
('large') or can be forced into this mould ('bea
overtly relational ('equal'32), some are neither
it is a certain number of, what it is equal to
that even of physical things some can be KaAa
what is said of pleasure at Rep. 584d seems a
in the Republic.
Notice, too, that Plato's treatment of these in
idea of physical mutability, often though that
it is with the compresence and not the successi
With these predicates Plato contrasts other

27 Pace Wilpert, Io9, who cannot think that Alexander

would allow himself such an interjection. But see 152d7
32 Yet,and,aswith
many due reserve,
have rTept
said, for dpyabl
Plato lpZptKI?)
at this time Xv.
Mansion, n. 79, Cherniss, 301-2. equality and other relations are attributes of the indi-
z8 vavria, in a sense that includes any prima facievidual. (It is worth recalling that oaov could be used
incompatibles (e.g. different numbers). to mean 'of middle size' and in this use is not overtly
29 With Republic 479a-b cf. 331c and 538d-e and relational.) Geach's conviction (op. cit., 76) that Plato must
Shorey, Republic, vol. i, 530, n. a. have thought of any case of equality, including the Form,
30 'actions': but Plato seems to have in mind types of as a pair of related terms cannot be justified by the bare
action (refs. in last note; cf. Jtaaol Ao'ot 3.2-12). The acrd d itaa of Phaedo 74c I. Geach writes that the
Symposium I8oe-Ia makes the necessary distinction but Form 'has to consist of two equals, or there wouldn't be
here, as elsewhere, seems a step beyond the Republic. equality at all'; Aristotle in the iepti iMeov, discussing
31 The debate on this passage has doubtless lived toothe same line of thought in Plato, said 'What is equal
long, but the natural sense is surely that given above. must be equal to something, so the a3xzdutov must be
equal to a second adrr6utov' (Alexander, Met. 83.26-8),
The Kolvowvia of the opposites with each other is a charac-
teristic of those 'manifestations' in the physical worldand whatever we think of Aristotle's methods of polemic
which seem to make a plurality of the Form; this is the this would have been absurd if Geach were right. See
only sort of pluralisation in question in the passage (cf.infra, I I0.
476b, 479a-b), and any attempt to read back the KOtVWLVia 33 The argument of Phaedo 74b-c is probably better
cOiv yevov of the Sophist into this text simply fits the construed on these lines, taking the T pi'v . . . 36 6' o0 of
argument too loosely. Plato is talking in terms of pairs 74b8-9 (despite the then misleading dative in 74 i) as
of opposites-the unity of a Form is proved by contrasting neuter and governed by faa. This at any rate seems
it with its opposite, and the same Adyo, is said to hold to be the sense that the argument in P makes of its chief
good of the rest (476a)-but the corresponding pluralisa-source (infra, Io09). Otherwise it turns directly on rela-
tion that is marked by the reconciling in one object of tivity to different observers (cf. Symp. 2I Ia4-5).
such a pair of opposites has nothing to do with the Sophist. 34 "'iua, Republic 524e2, 525a4, 523ci and d5, del, 479b8,
Good and bad cannot 'communicate' in the Sophist sense au3xd 6'vTa, cf. Phaedo 74b8 with Parm. 129b6 and Phaedo
I o2b-c.
(Soph. 252d). Cf. rather the Kpdrtgo 7rpd A aAAqAa of Tht.

This content downloaded from

193.225.200.92 on Tue, 13 Oct 2020 09:31:16 UTC
All use subject to https://about.jstor.org/terms
 A PROOF IN THE HEPI IAEQ2N I09
seen KaO' a4d -: sight never reports it to be at the same time not a fin
then, breeds no contradictions that have to be resolved by specify
evidently true of 'man', and of 'fire' and 'mud': all those predicates
is unready to admit Forms.35 That something is a finger is a matte
to pronounce (523b, 524d), and it is characteristic of the sorts of th
Ideas that they are just what we see them to be (Parm. 13od). The Phae
(263a: cf. Alcibiades I, I II-12) when it argues that men disagree not
but on that of 'good' or 'right'-or, we can add, on that of 'one' or
puzzles, like the moral antinomies of his successors, were built on such
the Parmenides of itself would suffice to show that these two classes of
earlier theorising. If we hope to resolve such disagreements by refe
standard, we shall find that the world which contains unambiguou
contains no comparable cases of goodness or similarity or equality
unambiguous Paradigms must be located elsewhere, in a voyrv'ds rdTroI.
Plainly, the exclusion of Forms of such non-relative predicates a
of later dialogues nor even of the last book of the Republic. A greater
(as in the Timaeus) would naturally suggest that in a further sense all
their earthly applications, for all apply at one time and not at an
expressly made in a dialogue marked by that preoccupation, the Sy
principle which could suggest it is already enunciated in the Rep
argument of P, which ignores this extension of the theory, isolates on
which in his earlier work at least he took small care and had small mo
to reconcile with others. The same is true of other arguments colle
what seems beyond serious question is that the earlier accounts o
preoccupation with incomplete predicates, in the narrower sense given

Man, fire and water seem to have remained stock Academic instances
by contrast with -r irp g'ETEpov or -ra 7rp6s -L,36 and there is small d
sketched above between complete and incomplete predicates in Pl
Academic dichotomy as well as of some major arguments for Ide
Aristoteleae preserved by Diogenes Laertius define Ta Ka Eawra dv AEy
JLUG7EVo, 7TpooraSETL and r 7Tpd? s hEyd0LEva accordingly as o a lr
Mutschmann). Now it seems plain that the same distinction unde
this explanation of ra rrpod TL recalls the argument of II(c) that the d
apply without further specification, JdKptfl9g,37 to anything in t
thing is called equal (and here again we have to note that equality is
individual thing) is to specify another with whose dimensions tho
seems only the other face of this coin, for different cases of equality
completed in different ways.38 (II(b) seems to add the rider that, since
things are constantly fluctuating, even to say 'having the same siz
without fixed meaning.) But even in Alexander's possibly condense
and II(c) are not duplicates and that their sequence is important.
the specification of various correlates can be no part of the meanin
ambiguous, and the point of II(c) is that when the common core
accretions it no longer characterises anything in this world.
Such arguments apply only to predicates which in their everyda
sense, relative. They follow Plato in deducing the existence of Ideas fr
of 'equal' (or mutatis mutandis of 'beautiful' or 'good') when this is
plexing expressions as 'man'. To this II(b) alone might seem an ex
imply (what it certainly does not say) that phenomenal things ar
35 Parmenides I3oc-d. Parmenides' explanation of neglect a parallel of thought and language in the Eudemian
Socrates' choice, that he rejects Ideas of yeAoTa, is applied
Ethics. In the discussion of three types of friendship in
only to mud, hair and dirt (I30c5). In any case it is a E.E. VII. 2 it is said that one Adyo'o does not fit all the
diagnosis of motive and not a characterisation of the cases (I236a26), but the Aodyog of friendship in the primary
reasons that Socrates could have offered. sense (Kvpio?g) is an element in the Ad'ot of the rest
36 Hermodorus apud Simpl., Phys. 247-30 ff., Diogenes(1236a 20-22: 'the rest' are here of course species and
Laertius III. io8, Sextus Empiricus adv. Math. X. 263. not, as in P, individuals). For whereas friendship in the
37 = dTA6,, opposed to KGTc a rpdrOeatv: cf. An. strict
Post. sense is to choose and love a thing because it is good
87a34-7, Met. 982a25-8 and Io78a9-13, E.N. I148a1I1.
38 Or the sense may be that different cases involve and pleasant d7TA5g, friendship in its derivative senses is
to do this because it is good po'pd -rt or pleasant rtvt.
specifying different measurements; but this would leave
In other words a definition that fits primary friendship
the senses of Ao'yog in II(a) and II(c) unconnected. And
II(c) may mean just that nothing is equal withoutneeds without
beingto bequalification
completed to (d67A,-
give the = -dKptflo
AOdo' of the in P. II(c))
derivative
unequal too. But, besides robbing Aristotle's reply of So in P: the similarity of language is very striking.
cases.
its immediate point (infra, Ilo), these interpretations

This content downloaded from

193.225.200.92 on Tue, 13 Oct 2020 09:31:16 UTC
All use subject to https://about.jstor.org/terms
 Iio G. E. L. OWEN
respects and so not KvpJ'S the subjects
death of P. It would contradict P.I, a
special arguments of II(a) and II(c) wo
it would leave Aristotle's identifica
unaccountable. For it seems to be true
they produce such Ideas, inter alia,39
argument concerned directly with 'r 7r

A NON-RELATIVE CLASS OF RELATIVES

The author of our proof is substantially faithful to the class of Platonic arguments he represents,
but here again he is anxious to sharpen a logical issue. What the dialogues describe as an appeal
to an intelligible Paradigm is seen, in practice, to be the application of a correct definition (e.g
Euthyphro 6e). It is in terms of definitions that P is framed. To say that nothing on earth affords
an unexceptionable Paradigm of equality is re-phrased as saying that to nothing on earth can th
definition of 'equal' be applied, pared of irrelevant accretions. Now this re-phrasing brings out
more clearly than Plato's words, the crucial point at which Aristotle directs his objection-and an
success in explaining his reply must stand in favour of our interpretation of the argument. Where
a Paradigm is required for a predicate that is incomplete in its ordinary use it must indeed be (a
the argument of P faithfully shows) a Standard Case, exhibiting rather than being the characte
it represents. But more: it seems that the Form, and the Form alone, must carry its predicate
Kac' at6rod in the sense given by the dichotomy. a3rd -rd ;'ov is indeed equal, but how can we without
absurdity ask to what it is equal? It cannot be equal to everything or to nothing (both would
engender paradoxes), and it cannot be equal to some things but not others (which would re-import
just the compresence of opposites that the Form was invented to avoid: Parm. I29b-I30oa). Th
incompleteness which so embarrassingly characterises 'equal' in its ordinary applications cannot
it seems, characterise it when it designates the Form. This is the natural sense of Socrates' warning
that the 'equal' he is to discuss is not 'stick equal to stick or stone equal to stone but just equal'
(Phdo. 74a), and it is the main point of the argument in P that unless 'equal' is merely ambiguou
the core of meaning common to all its uses must apply to something JKpitflc or, as Aristotle puts
it in the Metaphysics,
absurdities in a similar KaO' a-7o. One
treatment aim of
of 'one'. the
It is second
the partcase
extreme of of
theGreek
Parmenides, I take it,
mistreatment of is to find
'relative'
terms in the attempt to assimilate them to simple adjectives.4o
This is the point on which Aristotle fastens, and his rejoinder is not the simple deception that
Cherniss reads into it.41 It is developed in more than one place. In the Metaphysics he is conten
to observe that such arguments construct a 'non-relative class of relatives', i.e. a class of non-relative
instances of relatives. They require that any essentially incomplete predicate shall in one applic
tion behave as though it were complete-yet the Academy's use of the familiar dichotomy recognises
no such exceptions (see the Sophist 255c-d). Alexander reports what is in effect the same objection:
nothing can be equal that is not equal to something; but this entails that 7~ a'vror'ov is equal t
another at-do'Lov, and thus the Form is duplicated (Met. 83. 26-8). But even without thi
corroboration we could be sure of Aristotle's sense. In chapter 31 of the de sophisticis elenchis h
says: 'We must not allow that predications of relative terms (rcwv rpds -r AEyo1z'vwv) mean anything
when taken out of relation (KaO' ara-ds), e.g. that "double" means something apart from "doubl
of half" merely because it is a distinguishable element in that phrase. . . . We may say that by
itself "double" means nothing at all; or, if anything, certainly not what it means in context'-an
this rebuts the treatment of 'equal' in P and its sources as applying synonymously to earthly things
and to the Form. If 'equal' does not behave as tractably as 'man' in this world, that does not
entail that there is another world in which it does: the use of 'equal' is irreducibly different from
that of 'man'.
The consequence attacked by Aristotle is, I think, implied by the Platonic arguments on which
the proof in P relies. But did Plato clearly contemplate the consequence in framing the arguments?
That is surely doubtful. It would be easy to overlook it in the case of an asymmetrical relatio
such as double-of-half, where the absurdity of having to give the Form a twin in order to supply it
with its appropriate correlate does not arise. And Plato's very use of Kad' a-d-, by contrast wit
the Academic usage that grew out of it, shows the weakness; for in characterising a case of X
KaO' av-7
which he entry
gives evidently means
to an rather
opposite to exclude
(Parm. thei29d,
I28e and opposite
Rep. of X than
524d: to exclude
notice that thethe relativit
solution of
39 Met. 99obiI-I 7. The proofs Kadr -d r6v 817 i7tO7AAI v40 Cf. Cornford, Plato and Parmenides, 78, n. I, and for a
and Ka T d ovoev t qSO 0apg'vro do so because they later
are parallel R. M. Martin, Phil. and Phen. Research. XIV,
logically unrestricted in scope. For the 1Ayo't K 65 v
211.

iTtarrvItiv see Alexander, Met. 79.13-15- 41 Cherniss, 279-85.

This content downloaded from

193.225.200.92 on Tue, 13 Oct 2020 09:31:16 UTC
All use subject to https://about.jstor.org/terms
 A PROOF IN THE 17EPI IzAEQN III
contradictions by specifying rrpd& r and Ka7r T- 1is broached in quit
Republic). Nor is the latter exclusion the only means to the former
or covertly a comparative it can as well be represented as superlat
everything; so that here the predicate would retain its 'relative' ch
Idea. Between these alternatives the treatment of aVTr 0' KaAdv in Sy
be ambiguous. But 'equal' and 'one' are not so amenable: their purit
them, in strict analogy, equal to or one of everything. The pr
mistaken about the implications of its source.
Yet it brings out those implications with a new clarity, and in doin
Aristotle's hands. This fact, and the obvious concern of its author
suggest that here at least we should be incautious in treating our r
source of fresh information on Academic arguments about the Ide
may be responsible for the representative proof that he produces for
wholly plausible, for by characterising such proofs as iKptflE'CUEpoL (M
means to commend his opponents and not himself for the logical
developed. And the argument of P is not a mere (even disingenuou
arguments, but a new structure of argument in its own right. But is
the suspicion?
G. E. L. OWEN.
Corpus Christi College, Oxford.

This content downloaded from

193.225.200.92 on Tue, 13 Oct 2020 09:31:16 UTC
All use subject to https://about.jstor.org/terms