International Phenomenological Society Philosophy and Phenomenological Research

International Phenomenological Society
Manifold and Category

Author(s): Alexandre Koyre
Source: Philosophy and Phenomenological Research, Vol. 9, No. 1 (Sep., 1948), pp. 1-20
Published by: International Phenomenological Society
Stable URL: http://www.jstor.org/stable/2103848
Accessed: 21-06-2017 23:48 UTC
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
http://about.jstor.org/terms
International Phenomenological Society is collaborating with JSTOR to digitize, preserve and extend
access to Philosophy and Phenomenological Research
This content downloaded from 148.210.21.144 on Wed, 21 Jun 2017 23:48:45 UTC
All use subject to http://about.jstor.org/terms
PHILOSOPHY AND PHENOMENOLOGICAL RESEARCH
A Quarterly Journal
VOLUME IX, No. 1 AUGUST 1948
MANIFOLD AND CATEGORY*
In an article in Volume VI, Number 3 of this journal, I investigated the

logical structure of the famous "paradox" of Epimenides the Liar, and a
number of other connected "paradoxes." This analysis established the
fact that not one of these "paradoxes" was really paradoxical; that they
could be "solved" by the simple means of avoiding meaningless utterances,
-giving to the terms employed a precise and well-defined meaning, and
distinguishing between non-sense and counter-sense.
We now turn our attention to the much more serious problems raised by
the so-called Russellian paradoxes of the "impredicable" (not-predicable-
of-one's-self), and of the class of all classes that do not contain themselves
as their own elements.
I. THE NOT PREDCOABLE
Bertrand Russell1 tells us that concepts can be classified into two mutually
exclusive groups: those that can be applied to themselves, and those that
* Since the publication of my Epimenides the Liar my attention has been drawn
to the very remarkable paper of Professor Paul Weiss, "The Theory of Types,"
Mind, 1928. I am sorry that at the time of my writing of my own papers-they were
written in 1940-42 and reached Philosophy and Phenomenological Research in 1944-
it escaped my notice. This more so as I feel myself in nearly complete agreement
with Proffessor Weiss.
I B. Russell, Principles of Mathematics, pp. 97 sq. It is interesting to note that
from the very beginning Russell foresaw the principles of the solution of the "para-
doxes" which he had just discovered. Cf. Principles of Mathematics, p. 102:
(1) "If x be a predicate, x may, or may not be predicable by itself. Let us assume
that 'not-predicable of one's self' is a predicate. Then to suppose either that
this predicate is, or is not, predicable of itself, is self contradictory. The con-
clusion in this case seems obvious: 'not-predicable of oneself' is not a
predicate....
(2) " ... A class-concept may or may not be a term of its own extension. 'Class-
concept which is not a term of its own extension' appears to be a class-concept.
But if it is a term of its own extension, it is a class concept which is not a term
of its own extension, and vice versa. Thus we must conclude, against appear-
ances, that 'class-concept which is not a term of its own extension' is not a class-
concept.
(3) "A class as one may be a term of itself as many. Thus the class of all classes is
a class; the class of all terms that are not men is not a man; and so on. Do all
classes that have this property form a class? If so, is it as one a member of
itself as many, or not? If it is, then it is one of the classes which, as ones, are
2 PHILOSOPHY AND PHENOMENOLOGICAL RESEARCH
cannot. For example, the concept "abstract" is itself abstract, the concept
of "concreteness" is not concrete,-on the contrary, it is abstract. Let us
call concepts which apply to themselves "predicable by itself'"' and those
that do not, "not-predicable of one's self." Consider first the concept:
"not-predicable of one's self." Is it predicable of itself or not? It must
necessarily be one or the other; yet it is easy to show that it can be neither
the one nor the other.
Indeed to be predicable means to apply to one's self; the concept "Iot-
predicable of itself" is thus, if it is predicable of itself, not-predicable of
itself. Conversely, to be not-predicable of itself means-not to apply to
one's self. The concept- "not-predicable of itself"-if it is not predicable
by itself is thus "not-predicable of itself." Consequently, it is predicable
by itself.
* * *
Bertrand Russell has not only not confined himself to formulating the
paradoxes which bear his name, but he has also tried to present a solution
for them.2
This solution, in the case we are dealing with, consists, in short, in a
negation of the very possibility of classifying the concepts into "predicable
by itself" and "not-predicable of itself." No concept, in reality applies to
itself. When it does appear to do so, it is only at the price of an ambiguity.
The term remains the same; however, its meaning changes. The concept
"abstract" is not abstract in the same sense as ordinary abstract concepts.
It is, in a way, abstract to the second degree. The failure to distinguish
between these meanings would make the proposition impossible and, more
importantly, it would deprive it of meaning.3
not members' of themselves as many, and vice vera. Thus we must conclude
again that the classes which as ones are not members of themselves as many do
not form a class-or rather that they do not form a class as one, for the argument
cannot show that they do not form a class as many.
2 K. Grelling, Der Einfluss der Antinomien auf die Entwicklung der Logik im 20
Jahrhundert, Travaux du XIe Congres International de Philosophie (Congres Des-
cartes), fasc. VI, p. 11: "Bertrand Russell geh6rt der Ruhm, nicht nur die Tragweite
der Antinomien entdeckt, sondern auch den ersten wenigstens teilweise erfolgreichen
Versuch zu ihrer Auflosung. unternommen zu haben."
3 B. Russell, "Les paradoxes de la logique," RMM, XIV, (1906), p. 638; note:
"XSi le jugement: 'Il n'y a pas de propositions' dnongait lui-meme une proposition, il
se refuterait naturellementlui-m~me; mais suivant la theorie expose plus bas, un
tel jugement ou bier. n'a pas de sens, ou bien ne s'applique pas A lui-meme, et 1'as-
sertion qu'il se refute implique le sophisme du cercle vicieux." Ibid., p. 640: "Pour
eviter le sophisme du cercle vicieux on doit admettre le principle: 'Tout ce qui con-
cerne une variable apparente doit etre exclu des valeurs possibles de cette variable.'
NouS l'appellerons le principe du cercle vicieux. Le cas important de ce principe
peut etre dnoncd moins exactement comme suit: 'Tout ce qui enveloppe tous ne peut
MANIFOLD AND CATEGORY 3
Thus, all concepts are- "not-predicable." Nevertheless, the statement

that the concept "not-predicable of itself" is not-predicable of itself has no
meaning, unless a distinction is made between the two meanings of the
term "not-predicable of itself '-not-predicablel and not-predicable.2
The solution, known as the theory of types, is extremely elegant, in-
genious, and in part correct. It is certain that the concept "abstract" is
abstract in a different way than that of "color," for example, it is certain
that (just as was taught by medieval logic) it is necessary to distinguish
between "primary intentions" and "secondary intentions"; yes, even be-
tween "secondary" and "tertiary."
In spite of this the Russellian solution seems difficult to accept. Not to
speak of the general difficulties of the theory of types, which will be ex-
amined later, it appears hardly probable that the ambiguity in question is
not founded on a fundamental unity of meaning;4 that, for example, the
concept of "concept" and that of "space," while being concepts of very
different structure-and rank-do not, after all, have something in common
which makes them both concepts.
* * *
In order really to grasp the nature of the paradox of the "not-predicable

of oneself," it is useful, and instructive, to give it a-symbolic formulation.
We shall borrow such a symbolic formulation from Fraenkel and Carnap:5
"A certain property F will be said to be "not-predicable of oneself"-Imp
if it does not apply to itself; in symbols:
(1) Imp (F) = -F (F)
From this definition results, a

(efiniens and the definiendum:
(2) (F) Imp (F) =_ F (F).
etre un de ces tous.' Ainsi un jugement sur toutes les propositions ne peut etre qu'un
non sens, ou bien 1'6nonc6 de quelque chose qui n'est pas une proposition au sens
vis6.... Ce r6sultat peut etre atteint en d6cidant qu'un jugement portant sur tou,
... est reellement l'affirmation (indetermin6e) d'une de plusieurs propositions por-
tant sur des cas particuliers."
4Russell himself recognizes this implicitly, since he speaks of the "systematic
ambiguity" of certain expressions such as: truth, falsity, function, class, etc. (cf.
Principia Mathematica, I, 67 quoted below, n. 13). This "systematic ambiguity,"
according to us, covers a fundamental unity of meaning; it can even be sustained that
the "ambiguity" is created by Russellian symbolism, especially by' his theory of
types.
6 Cf. A. Fraenkel, "Le problem des antinomies et son ddveloppement recent,"
RMM (1939), p. 229.
The conjunction of the two following implications is thus obtained:
3 (a) (F). Imp (F) D) F. (F).

3 (b) (F). - F (F) D Imp (F).
Let us now substitute for F the particular value Imp.; we obtain:
4 (a) Imp (Imp) D- Imp (Imp).
4 (b) - Imp (Imp) D Imp (Imp);
The antinomy is evident."

There is no doubt, but it is just as evident that the antinomy arises
from the "substituion for F of the particular value Imp.," just as the para-
dox of the Liar results from the substitution of the particular value "I"
for the X of "X lies."
Is this substitution more legitimate in the case of Imp. than in that of
"I lie"? It certainly seems that it is not.
Thus Behmann6 notes: "An expression containing abbreviated signs is not
correct unless it allows a complete substituion, within the framework of
the symbolism employed, of the meanings for signs." Now, for Imp., this
is impossible to achieve; the sign, no matter what is done with it, remains
in the formula, and its meaning cannot be substituted for it. A definite
indication, according to us, that the "meaning" of Imp. remains strictly
undetermined;7 this necessarily implies its illegitimacy.8
Behmann's solution, in the long run, consists in requiring that every
6 H. Behmann, "Zu den Widerspruchen der Logik und der Mengenlehre," Jahr-
buch der Deutschen Mathematiker Vereinigung, Vol. XL, (1931), pp. 37, 48. Cf.
likewise H. Behmann, "The Paradoxes of Logic," Mind, Vol. XLVI (1937), p. 220:
"The starting point of my own analysis of the paradox is this: Whereas, in any
logically correct form of argument, it must be possible to replace the notions and
symbols introduced by definition by those complex terms for which they stand, this
process of translation back into the original terms fails for the symbolic formulation
of Russell's paradox, especially for the complex F (F). (F. P. Ramsay wrote me on
this point: 'This is a striking fact which I, at least, had never realized.') Thus the
definition in question turns out to be more essential to the argument of the paradox
than being a mere definition it ought to be."
7A symbol of undetermined meaning is a symbol which has none.
S It is interesting to note that the same impossibility of getting rid of a symbol
once it is introduced is pointed out by M. Langford (cf. C. I. Lewis and C. H. Lang-
ford, Symbolic Logic (New York, 1932), p. 438 sq.) in his analysis of the Liar. If we
wish to realise the meaning of "I am lying" = p = p is false, we obtain "an infinite
sequence of more and more complicate expressions, each of which requires explica-
tion before its import becomes definite; so that no one of the expressions can be
significant unless the sequence terminate which it does not do." (p. 440). Cf.
my article on the Epimenides the Liar, This journal, Volume VI, Number 3.
formula-even a symbolic one-have a well defined meaning;9 it has not

acquired many followers. "All the formulas cannot be examined," is the
objection commonly raised.'0 This is quite so. But this impossibility does
not relieve us of the necessity, and even the duty, of examining those which
we use; provided, of course, we do not want to use meaningless ones.
"In destroying the automatic character of the symbolism we would make
it lose its essential merit." This is possible. It remains to be seen whether
this merit is so great that we have to maintain this automatic character at
all costs. After all, automatism is nowhere realized in mathematics-for-
mulas do not free us from the necessity of thinking, and everybody knows
that there are values-such as 0, for example, that cannot be substituted,
indifferently, to the variably of any and every formula. Why should logis-
tics have this privilege?
Be that as it may, Fraenkel informs us that from the point of view of
logistics "if the use of any sign must be preceded by the demonstration of
its legitimacy, one may hesitate as to the advantage that this method may
have over the theory of types.""
In our view, hesitation is not possible, if only for the reason that, as we
shall see, the theory of types is untenable and self contradictory.
Before studying the theory of types, however, let us pause for another
moment and continue our examination of the antinomy of the "not-predi-
cable."
We might note, first of all, that it is rather misleading to speak, as Fraen-
kel and Russell do, of "properties" applying or not applying to themselves.
Generally speaking, a "property" never applies to itself except in two cases:
(a) that of negative "properties," such as not-red, not-extended, etc., and
(b) that of the "properties" of concepts. But "negative properties" are
not real properties. This can easily be seen if one takes into account the
fact that they are almost entirely lacking in determination and, as for con-
cepts, there are very few- with the exception of "negative concepts"-
Nvhich enjoy the remarkable property of being able to be applied to them-
selves. (The concept of "concept"; of "abstract," already mentioned, of
"thinkable," of "formal," and several others.) Such concepts are so limited
in number that the entire list could be enumerated-reflexiveness is a rare
9 Cf. H. Behmann, "The Paradoxes of Logic," Mind (1937), p. 220: " . . . from
the point of view of exact logic, the very essence of the problem of paradoxes is no more
nor less than the problem how to state and to apply symbolic definitions correctly, more
generally, how to decide whether a given expression can be symbolically substituted in a
given expression." (italics mine)
10 Cf. A. Fraenkel, op cit., p. 238.
11 Behmann's attempt does not seem satisfactory to us either, fundamentally for
the same reasons.
thing, even in the world of concepts.'2 All those which possess it seem
to belong to a sort of family: they are all categorial or formal concepts.'3
Concerning the concept of "not-predicable of oneself" one may wonder if
it really denotes a common property of certain concepts. Not-predicability
designates an absence, not a presence of something. An absence can hardly
be considered as a common element. Moreover, to say of a concept that it
is not "predicable of itself" (if one endeavors to decipher the meaning of
this assertion) is equivalent to saying that it is not red, extended, concrete.
Unless the field of negation is surreptitiously limited, it seems rather dif-
ficult to form a unity of meaning with all these negations. For "non-red"
can mean "blue" as well as "concrete" or "abstract," and "non-concrete"
can mean "abstract" as well as "material" or "immaterial." And even
"blue" or "red."
Accordingly, though it may be possible to classify concepts into "predi-
cable by themselves" and "not-predicable of themselves," it does not follow
therefrom that the two opposing terms have a well-defined meaning. But
where the terms are not definite, where the question does not have a univocal
meaning, the answer cannot have one either. This is exactly the case of
the "not-predicable."'4
* * *
II. THE CLASS OF ALL CLASSES
Consider now the paradox of the class of all classes . . ., the first-born, the
one which started the movement, which terrified Frege, Dedekind, and
Cantor himself; which, in a very special way, is called Russell's paradox.
12 It is these reflexive concepts which engender the "paradoxes, as Russell very

well recognizes. Cf. Principia Mathematica, I, 64, "In all the above contradictions
(which are merely selections from an indefinite number) there is a common charac-
teristic, which we may describe as self-reference or reflexiveness. The remark of
Epimenides must include itself in its own scope. If all classes, provided they are
not members of themselves, are member of n, this must also apply to n. In the cases
of names and definitions, the paradoxes result from considering non-nameability and
indefinability as elements in names and definitions.... In each contradiction some-
thing is said about all cases of some kind, and from what is said a new case seems to
be generated, which both is and is not of the same kind of which all were concerned
in what was said."
13 It is to these concepts that Russell, misunderstanding their logical and onto-
logical character, attributes a "systematic ambiguityof type" cf. Ibid., I, 67: "An
indefinite number of other contradictions, of similar nature to the above seven, can
easily be manufactured. In -all of them, the solution is of the same kind. In all
of them, the appearance of contradiction is produced by the presence of some word
which has systematic ambiguity of type, such as truth, falsehood, function, property,
class, relation, cardinal, ordinal, name, definition. Any such work, if its typical
ambiguity is overlooked, will apparently generate a totality containing members
defined in terms of itself, and will thus give rise to vicious circle fallacies."
14 The term "not-predicable of oneself" has no determined meaning. Thus it is
Now it is easy to understand the terror which it inspired at the beginning.

Here is a reasoning which seems unassailable and where "only the notion of
aggregate (class, collection) and the relation with two variables 'X is an
element of the aggregate Y,' that is to say, the fundamental notions of the
theory of aggregates are involved: those which, because of their extreme
generality, are of a more logical than mathematical character."'5 A reform
of logic, or a limitation of the Cantorian notion of aggregate, or both, seemed
to be imperative. In our opinion, nothing of the kind is needed. But let
us not anticipate.
Let us once more borrow the exposition of the antinomy from Fraenkel:
"This is what Russell's antinomy amounts to in its original form: let us call
'normal aggregate' an aggregate which does not contain itself as an ele-
ment, and let E be the aggregate which contains as elements all normal
aggregates and nothing more. Let us then suppose that E itself is a normal
aggregate; by virtue of its definition it should contain itself as an element,
which is in contradiction with the hypothesis. Then E is not a normal
aggregate. This conclusion in turn leads to the following contradiction:
the E aggregate, by virtue of its definition, no longer contains itself as an
element, while the property X is not a normal aggregate, means precisely
that X contains itself as an element. The contradiction obtained is evi-
impossible to know the meaning of the question: is "not predicable of oneself" pred-
icable or not? The same is true of the "paradox" discovered by Grelling concerning
the terms "autologisch" and "heterologisch." Cf. K. Grelling, Der Einfluss der
Antinomien, etc., p. 10. " Ein ganz gleichartiger Widerspruch lAsst sich auch anstatt
mit Klassen und Begriffen mit Worten konstruiren, wie ich 1908 gefunden habe. Man
kann nimlich die Worte danach einteilen, ob sie einen Begriff bezeichnen, unter den
sie selbst fallen oder nicht. Die erste Art von Worten nenne ich autologisch, die
zweite heterologisch. Das Wort "deutsch" z.B. ist autologisch, denn es ist selbst ein
deutsches Wort, ebenso sind 'francais,' 'English' u.s.w. autologisch. 'Franz6sisch'
dagegen ist heterologisch weil es nicht franz6sisch ist. Auch solche Worte die keine
Eigenschaft bezeichnen wie 'aber,' 'und' u.s.w. sind heterologisch. Man kann nun
fragen ob das Wort 'heterologisch' selbst heterologisch oder autologisch ist, und
kommt auf denselben Widerspruch wie in den beiden vorigen FAllen.'! It is rather
amusing and, in the same time, significant, that nobody seems to have tried to raise
the question: is the concept "predicable by itself" predicable by itself, or not?
(Similarly: is "autologish" autologish?) These questions do not lead to "para-
doxes," but, on the other hand, they reveal their perfect lack of meaning and the
impossibility to give to them any answer whatever. Even if the terms had a well
defined meaning (which they have not) it would be perfectly impossible to apply them
to themselves. As a matter of fact, the dychotomy they pretent to create is illusory,
as in all the cases when a classification is established respecting a certain term, or
property. In all these -cases the term or property in question is outside of the two
classes which it helps to define.
15 A. Fraenkel, op. cit., p. 226.
dently independent of the matter of knowing whether aggregates other than

normal aggregates exist.'6
We do not share this last opinion of Fraenkel's, which, in any event, is
in no way proper to him: it is also Russell's, and it is even the opinio corm
munis of the logisticians. For, on the contrary, it seems rather important
to know whether aggregates other than normal ones, that is to say, if ag-
gregates which contain themselves as elements, exist or not. Indeed, this is
a matter of prime importance. For if such aggregates did not exist, the
dichotomy on which the Russell paradox is founded would be entirely fac-
titious. The classification of aggregates into 'those which do not contain
themselves as elements,' and 'those which contain themselves as elements'
(perfectly analogous to the classification of numbers into 'those which are
not greater than themselves' and 'those which are greater than them-
selves')'7 would be equivalent to their classification into 'those which
exist' and 'those which do not exist.' It seems that such a classification
cannot reasonably be made: beings cannot be divided into those that exist
and those that do not exist.
Our opponents could retort, however, that our objection is pure
quibbling; and they could even add that the Russellian dichotomy is not at
all factitious; moreover, that aggregates containing themselves as elements
exist effectively as, for example, the aggregate of all aggregates.18 Thus the
acceptance of our objection would in no way save us from the paradox but
would immediately plunge us into it: in fact, if 'non-normal' aggregates
do not exist, the aggregate E is identical to the aggregate of all aggregates.
As such, it contains itself as an element and is therefore not 'normal.'
However, we shall maintain our objection. For the true problem seems
to us to be precisely that of non-normal aggregates in general; and that of
all aggregates in particular.
* * *
Can an aggregate contain itself as an element? Can a totality be a part

of itself? That, Bertrand Russell tells us, is the central problem with which
the existence of antinomies confronts us."9 Of those which we have ex-
16 Ibid., p. 226. Cf. B. Russell, Principles of Mathematics, Chap. X. It is well

known that this "paradox" was published for the first time by G. Frege in an Ap-
pendix to his Grundlagen der Arithmetik in 1902.
17 Cf. iiy article on Epimenides the Liar, op. cit.
18 The legitimacy of this " aggregate," as well as of that of "all objects" is main-
tained by Behmann, Grelling, etc.
19 Georg Cantor has seen it as early as 1897. He has pointed out to the existence
of inconsistent wholes already in his letter to Dedekind in Gesammelte Ab-
handlungen, Berlin (1932), pp. 447, 451. Unfortunately he did not investigate the
subject further.
amined, as well as of all others, for they all imply, or can be reduced to, the
constitution of such totalities, or what, according to Russell, is the same
thing, to the definition of an object in terms of the class to which it is sup-
posed to belong.
Let us rule out the use of such definitions. Let us, for example, lay down
as a principle the illegitimacy of 'non-normal' aggregates and, in order to
prevent their reappearance, let us take this principle as a basis of a system
of logic, and classify objects and propositions according to the 'type' to
which they belong. The 'type' of a proposition-or of a property-will
express the degree of its logical complexity. Thus individuals-and no-
tions concerning individuals-as logical objects of the simplest structure,
will be of type 0. Properties of individuals, propositions having bearing
on individuals, classes of individuals, being logical objects which presuppose
individuals and are founded on them, will be of type 1; propositions, having
bearing on propositions, classes of classes, etc., will be of type 2, and so on.
It is clear that any proposition-or any class-will be of a type superior to
its elements, that is to say, to the objects on which it has a bearing or, if we
are dealing with classes, to those which it contains. The 'theory of types'
declares illegitimate any judgment which infringes upon the hierarchical
rule of constitution which we have just exposed. Or, as Fraenkel says20:
"The principle which is . . . at the basis of the theory of types . . . can be
formulated in the following way: in a given proposition no determined prop
erties can be attributed to individuals or properties unless the type of the
attributed properties is superior by one to the type of the individuals or
properties subject to attribution." It is clear that all 'paradoxical' judg-
ments are thereby ruled out. They are ruled out not only as false, but,
what is far more serious, as meaningless.
* * *
The theory of types is not a logical expedient but a true logical theory.
To quote Fraenkel2' again: "Historically, the theory of types grew out of
the problem of antinomies, but it does not in any way have the character of
an ad hoc construction. It is derived rather from Russell's vicious circle
principle according to which no collection (no aggregate) can contain ele-
ments which are definable only in terms of that collection itself," or as Rus-
sell says, "nothing which involves the whole of a collection may be part of
this collection,' '22 and "if a collection of objects must contain members

21 Ibid.
22 Cf. B. Russell, Principles of Mathematics, App. B, p. 526; cf. Les paradoxes de la

logique, p. 634: "Je reconnais done que la clef des paradoxes doit se trouver dans
V'Wd&e du cercle vicieux, je reconnais, en outre, ceci de vrai dans objection que M.
Poincar6 fait 'a 'idge de totality, que tout ce qui concerne d'une manibre quelconque
definable only in terms of the collection itself, taken as a whole, then this
collection is not a whole." Again: "if the supposition that a collection
forms a whole implies that it possesses members which are definable only in
terms of this whole, then this collection does not form a whole."23
This discovery-or rediscovery24-of "illegitimate totalities" (as Jorgen-
sen calls them), or of "non-totalizable multiplicities" as we prefer to call
them, seems to us to be one of Bertrand Russell's greatest merits. It has, as
a matter of fact, a very great bearing; and it seems to us extremely regret-
table that Russell did not realize its full implications.
* * *
The vicious circle principle-and the prohibitions it implies, both equiva-

lent, as Jorgensen has already noticed, to the prohibition, by Henri Poin-
car6,25 of non-predicative definitions-is extremely plausible. In its sim-
plest forms-that of the impossibility for an aggregate or a whole of
containing itself as an element-it even seems perfectly evident. For,
after all, it seems clear that there is always something which can never be
put into a bag, as large as it may be: it is this bag itself; as a matter of fact,
the container must always be larger than the contained.
One may ever wonder whether it is worthwhile to make a "principle"
of this observation.
Be this as it may. It is, in any case on this "principle," that Russell has
founded his new logic, the logic based on the theory of types. Even if it is
true that nothing is more artificial than this logic (which Russell himself
tout ou quelque ou un quelconquc des membres d'une classe ne doit pas etre un membre
de la classe. Dans le langage de M. Peano, le principe que je soutiens peut s'6noncer
comme suit: 'Tout ce qui contient une variable apparente ne doit pas etre une valeur
possible de cette variable.' " Cf. Ibid., I, 65: "a class cannot, by the vicious circle
principle, significantly be the argument to its defining function, that is to say, if
we denote. by ' Z(Soz)' the class defined pzo, the symbol 'z(po 4 z) ~' must be meaningl
Hence a class neither satisfies nor does not satisfy its defining function, and therefore
is neither a member of itself nor not h member of itself.... Thus if a is a class, the
statement 'a is not a member of a' is always meaningless, and there is therefore no
sense in the phrase 'the class of those classes which are not members of themselves.'
23 To be exact, it should rather be said: then there is no collection. Cf., Principia
Mathematica, I, 39 sq.: "Whatever involves all of a collection must not be one of the
collection.... If, provided a certain collection had a total, it would have a member
only definable in terms of that total, then the said collection has no total." Cf.
ibid, pp. 24, 65. Cf. B. Russell, Mathematical Logic as based on the theory of types,
p. 262: "Every expression regarding a totality is of higher type than the members of
the totality . . . any expression containing an apparent variable is of higher type
than that variable"; p. 171: "according to the theory of types, a proposition can never
be about itself."
24 After Cantor. And, fundamentally, after Aristotle.
25 Cf. Jorgensen, Treatise of formal logic, Copenhagen (1931), vol. I, pp. 167 sq.
has never, moreover, presented as definitive), that it has been the object of
numerous criticisms,26 and that, in our opinion, it involves insuperable diffi-
culties, it remains none the less the only logical theory which has tried to get
to the bottom of the problem of antinomies and, "in the present state of
science the least that we can affirm is this: as far as the solution of logical
antinomies is concerned, the position adopted by Russell in the first ten
years of the century cannot be considered as having been transcended. In
particular, the theory of simple types retains today all its value, in spite of
the inconveniences which it entails; it is not simply a well founded logical
theory, it is also the simplest of all the theories which have been proposed
up to the present to solve .. . the problems relating to logical antinomies
and to the axiom of the infinite."27
* * . *
We have just said that the theory of types prohibits us from totalizing
certain multiplicities: precisely those which result in the paradox of self-
inclusion; and forbids certain judgments: those, fiamely, which lead to, or
imply, the possibility of self-application. Let us state immediately that we
believe it to be erroneous precisely in this point: namely, in the identifica-
tion of the impossibility of self-inclusion with that of self-application. The
theory of types which ruined the extensionalist interpretation of logic, suc-
cumbs, in fact to an extensionalism which it did not overcome.
The theory of types did not receive an enthusiastic welcome from logisti-
cians. This can be rather easily understood. It compels us, as a matter of
fact, to accept as 'axioms' propositions which are not only completely
lacking in evidence (which after all would be a lesser evil), such as the axiom
of infinity, but even propositions obviously formulated ad hoc, such as the
axiom of reduction.' It also leads to very serious difficulties for which it
has been violently and justly criticized.
"The theory of types involves as a logical consequence the existence of an
infinity of universal classes, ....2 since classes which contain all the objects
26 Cf. Jorgensen, op. cit., p. 169 sq.

27 Fraenkel, op. cit., p. 242. Cf. K. Grelling, Der Einfluss der Antinomien, etc.,
p. 12: according to the theory of types "Ein circulus viciosus ... liegt vor wenn man
einer Gesammtheit Elemente zuschreibt, die nur vermittelst dieser Gesammtheit
selbst definiert werden k6nnen."
28 Principia Mathematica, I, 30: "The class determined by a function which is al-
ways true is called the universal class and is represented by V; thus
V = x (x = x) Df.
Thus A is the negation of V. We have

(X) .x ? V.
i.e.,.' x is a member of V' is always true."
of a certain type are distinct from one another, and since elements of a dete
mined class must be of the same type.29 There also exist, in the theory,
and this almost borders on a paradox, an infinity of null classes. But what
is most unexpected and seem most improbable to most mathematicians is
the existence of a plurality of systems of [natural] numbers. Numbers are
defined as classes of classes, viz., the classes of all classes [of the same type]
equivalent to a given class, and each type thus implies the existence of a
system of numbers proper to it. This consideration is rather unpleasant,
first of all because of the painful impression that it gives, then, and above
all, because it requires the use of a complicated formalism, necessary for the
treatment of this plurality of number systems.30"
* * *
Do the unpleasant and awkward consequences of the theory of types

constitute sufficient arguments against it? As far as we can see they do.
Indeed it seems to us that the value of a logical theory which gives an
entirely questionable picture of thought; or that of a mathematical theory
which leads to the necessity of admitting a plurality of systems of whole
numbers, and which requires unparalleled complications for the execution of
the simplest operations, such as counting a certain group, or number, of num-
bers, is condemned by these facts.31 But logisticians will not agree to this.
In their opinion improbability is not a proof of falsity. Therefore we shall
try to add several other objections, even more serious than those that have
been raised thus far.
29 Principia Mathematica, I, 31: "There are no classes which contain objects of

more than one type. Accordingly there is a universal class and a null class proper
to each type of object. But these symbols need not be distinguished, since it will be
found that there is no possibility of confusion."
81 Cf. W. Ackermann, "Mengentheoretische Begrundung der Logik," Mathe-
matische Annalen, Vol. CXV, fasc. 1 (1937), p. 1 sq.: "Die in den Principia Mathe-
matica aufgestellte Forderung, dass zum Definitionsbereich eines Pradikates nur
Elemente ein- und derselben Typs gehoren, stellt sich als eine unnotige Einschran-
kung dar, vor allem, wenn man berucksichtigt, dass die Mengen denselben Be-
dingungen unterworfen werden. Eine Menge, z.B. deren Elemente teils ganze Zah-
len, teils Mengen von ganzen Zahlen sind, hat doch gewiss nichts Paradoxes in sich.
Wesentlich erscheint nur, dass der Typ der Elemente der Menge einen bestimmten
Typ nicht iubersteigt. Ferner fehlt im Stufenkalkill die Fortsetzung der Reihe der
Typen ins transfinite, entsprechend der Reihe der Ordinalzahlen. Eine derartige
Fortsetzung der Typenreihe, die zuerst von Hilbert gefordert wurde, erweisst sich
auch in Folge der von Goedel gezeigten Unfollstandigkeit des Systems der Principia
Mathematica als notwendig." Cf. D. Hilbert, "Ueber das Unendliche," Mathe-
matische Annalen, Vol. XCV (1925), p. 14-190; Goedel, "Ueber formal unentscheid-
bare Satze der Principia Mathematica und verwandter Systeme", Monatshefte far
Mathematik und Physik, 38 (1931), p. 173-198.
MANIFOILD AND CATEGORY 13
It should be observed, first of all, that to speak of a plurality of systems of

numbers is insufficient. It should be said that the theory of types leads to
the admission of the existence of an infinity of systems of numbers: this is
an immediate consequence of the infinity of universal classes. Then it
should be noted that this infinitization, necessarily and evidently, is not
by any means limited to the realm of numbers: it extends to logic. Just as
we shall have an infinity of arithmetics, so will we have an infinity of logics,
and the law of contradiction will have to be formulated an infinite number
of times.32 Finally, it should be pointed out that all the propositions we
have written and which state either the consequences of the theory of types,
or this theory itself, have, according to the theory of types, no meaning what-
ever.
As a matter of fact, if each proposition must be of a determined type,
superior by a unity to the type of objects to which it applies, propositions
such as: there is an infinite number of systems of numbers, or of universal
classes, or of types of propositions and objects, should itself be of a more
than infinite type, that is, could not belong to any type whatsoever.33 As
to the fundamental proposition of the theory of types-and that applies to
all the propositions of the theory-according to which "every proposition
must be of a type superior to its object," not only would it belong to no
type, but it would exemplify the paradox that the theory of types precisely
aimed to avoid. Thus, it cannot be true unless it is false, and it is false, if
it is true. For it is either true and therefore applies to itself, in which case
it is false, and even impossible, being devoid of meaning; or, not applying
to itself, it is false and therefore could be true. In other words, the theory
of types prohibits the statement of propositions applying to all propositions,
and thereby contains such prohibited propositions. Therefore, we cannot
even ask ourselves whether the theory of types is true or false; what, as a
matter of fact, would be the type of a judgment that would affirm this
truth or falsehood? Let us conclude: the theory of types realizes to perfec-
tion the very type of the prohibited 'vicious circle.'
* * *
That the theory of types has no meaning is an unavoidable conclusion

of the theory itself.34 But, for our part, we cannot accept it. And our
32 B. Russell, "Les paradoxes de la logique", RMM, 1906, p. 644. "Cette theorie

peut etre illustree encore en considerant la loi du milieu exclu, sous la forme "toute
proposition est vraie ou fausse". Si cela est vrai, la plupart des gens croiront l6gi-
time d'inf6rer que la loi du milieu exclu est elle-meme vraie ou fausse; pourtant c'est
1lA une inference precisement de l'espece qui engendre le paradoxe du Menteur."
33 It could be said that it should belong to a transfinite type.
34 It is true that the theory Ol{ types proihibits from drawing these conclusions:
we do not, however, believe that that can save it from the counter-sense it implies.
opposition is quite particularly stiffened by the fact that neither Russell

himself, nor the numerous logicians who have discussed and criticized him,
have realized that they were indulging in meaningless talk.
The theory of types states that self-application of a proposition is im-
possible. Now, our analysis of the Epimenides has already shown us that
this is not correct: self-application of a judgment is possible and the rea-
soning both of the Epimenides and of the theory of types have a meaning.
There again we find encouragement to maintain-against Russell and
the theory of types-our opinion and our interpretation of the Liar for a
very simple and even very commonplace reason: if it were otherwise some-
one would have noticed! Now, ever since skepticism or relativism has been
discussed-too seldom, in our opinion, has it been recognized that the skep-
tic's or the relativist's assertion has exactly the same logical structure as
the famous Cretan's-with the sole exception of Russell no one, among all
-those who have fought against these doctrines, from Aristotle and St.
Augustine to Husserl, has noticed that they were fighting against something
that has no meaning. On the contrary, they have always attempted to
show that these doctrines were false, because they implied an absurdity or
better said, a counter-sense, just as the theory of types.
* * *
The abandonment of the theory of types to which we have been led car-
ries with it non-negligible advantages: thus we shall no longer need to mul-
tiply series of numbers35 to infinity: one will suffice; and we shall likewise
need only one logic; we shall even be able to formulate the grammar of a
given language in that language itself, without being obliged to invent a
new one.36
But in rejecting the theory of types do we deprive ourselves of the possi-
bility of solving Russell's antinomies? As a result of this rejection, and of
our acceptance of the legitimacy of the self-application of propositions, do
we find ourselves in a more difficult and more paradoxical situation than
Russell? We do not think so.
35Cf. A. Koyr6, "Remarques sur les nombres de M. Russell," RMM, 1912.

36 K. Grelling, The Logical Paradoxes, p. 485: "David Hilbert was the first to
make a distinction between mathematics and metamathematics. By this term he
designates the theory dealing with mathematical deductions as its objects. This
distinction has proved to be very fruitful. Lucasievicz, the great Polish logician,
has generalized that idea by introducing what he calls metalogic, which has to or-
dinary logic the same relation as meta-mathematics to mathematics. This whole
complex of problems has been systematised by Carnap in his treatise Logische Syntax
der Sprache and some separate papers. Also Godel and Tarsky have contributed
much to the development of these problems. The labor of these logicians has led to
important results concerning the syntactical paradoxes. They can now be proved
to disappear if one carefully distinguishes between a given language L and the lan-
guage L1 in which one speaks about L. Carnap calls the latter syntax-language."
Let us summarize our position, by opposing it, to that of Russell. Rus-

sell reasons as follows: (1) The multiplicities which imply a self-inclusion
cannot form a total. (2) The assertions which imply a self-application also
imply a self-inclusion.
Whence the conclusion: (3) The assertions which imply a self-application
are meaningless and impossible.
As far as we are concerned, while agreeing with Russell that (1) The mul-
tiplicities which imply a self-inclusion are not totalizable, we are obliged to
ascertain that (2) The assertions implying a self-application are possible.
Whence we conclude: (3) The assertions which imply a self-application
do not necessarily imply a self-inclusion.
Russell's reasoning assumes that every legitimate assertion determines a
field of application forming a totalizable multiplicity, in other words, that
to every concept correspond a class or an aggregate.
From our findings, it follows that this is not so. As strange as that may
seem at first, we are forced to admit that there are notions of which the field
of application does not form a totalizable multiplicity; in other words, no-
tions or concepts of which the extension forms neither a class nor a whole.37
Now, it is the failure to recognize this fact which seems to us to be at the
basis of all the difficulties run into by the theory of aggregates. They are,
as a matter of fact, founded (a) on the identification of the aggregate with
the class and (b) on the idea, plausible no doubt, but nevertheless false, that
there is a class which corresponds to every concept, that is, a totalizable
multiplicity; thus Cantor considers that the possibility of deciding whether
or not a given concept applies to an object, is equivalent to the. possibility
of determining whether or not it belongs to a given class.
It is interesting to note that the dropping of the correlation between 'con-
cept' and 'class' enables us to keep unaltered Cantor's celebrated defihi-
tion of the aggregate. This definition, "union in one whole of determined
and distinct objects of our-intuition or our thought," does, in fact, bring
in the notion of the whole.38 It thus implies that multiplicities forming
aggregates must be totalizable. Cantor, in formulating his definition, did
87 Already in the Principles of Mathematics, Russell established the very important

distinction between class as many and class as one (pp. 97 sq.) by showing that the
two acceptations of the term are not congruent. Cf. above and Principles of Mathe-
matics, I, p. 104: "Perhaps the best way to state the suggested solution is to say, that
-if a collection of terms can only be defined by a variable propositional function,
then, though a class as many may be admitted, a class as one must, be denied....
We took it as axiomatic that the class-as one is to be found-wherever there is a class;
but this axiom need not be universally admitted, and appears to have been the source
of the contradiction. By denying it, therefore,. the whole difficulty will be over-
come.... The distinction of logical types is the key to the whole mystery.'
38 G. Cantor, Gesammelte Abhandlungen, p. 282,
not yet realize that not all of them are able to form a total but, by a sor
genius-like intuition, his definition foresees and wards off the difficulties
* * *
Let us now return to Russell's paradox of 'the (class) of all aggregates

(classes) which do not contain themselves as elements. . . .' The theory
of types prohibits us from admitting the existence of aggregates which
contain themselves as elements, of 'non-normal' aggregates, as Fraenkel
calls them. As we have shown above, this prohibition implies the identifi-
cation of the aggregate of all normal aggregates with that of all aggre-
gates. But this last is, in its turn, prohibited by a new application of
the theory: the multiplicity of aggregates is not totalizable and the expres-
sion 'all aggregates' has no meaning.39
Now, we do not accept the theory of types, no more than we admit the
existence of non-normal aggregates. On the other hand, with Russell, we
admit the existence of non-totalizable multiplicities, and we consider, with
him, that the 'aggregate of all aggregates' does not exist, since the multi-
plicity of aggregates is precisely not totalizable.
That will in no way prevent us from formulating judgments bearing upon
all aggregates, that is applying to everything that is an aggregate, and from
saying that Cantor's theorem, according to which 'the aggregate of sub-
aggregates of an aggregate is of a power superior to that of the latter' is
true for all aggregates just as Zermelo's theorem according to which 'every
aggregate can be well ordered' is true-or untrue-of all aggregates. For,
according to us, the possibility of pronouncing a judgment valid for all ob-
jects of a certain kind-that is, a judgment bearing upon each and every
one of the objects which fall under a concept, does not imply the necessity-
nor the possibility-of uniting them into a whole.41
39 The Principia Mathematica, based on the theory of types, maintain (Vol. I,

p. 24) that "a class (which is the same as a manifold or aggregate) is all the objects
satisfying some propositional function." Only classes themselves are excluded from
the field of application of the function; cf. ibid: "a class cannot be an object, suit-
able as an argument to any of its determining functions." That is why if V. repre-
sents a class and pX' one of its determining functions iso that a = z(,oz)I, it is not
sufficient that joa be a false proposition, it must be nonsense." It is easily seen that
the breach with extensionalism is far from being complete.
40 If this is not true of the aggregate of all aggregates it is because this aggregate
does not exist. It is perfectly useless, as does Behmann (The Paradoxes of Logic,
p. 220) to find fault with-Cantor's theorem.
41 A judgment having bearing on 'all A's' does not have the 'whole' as an object,
but the A's, as has been recognized from Aristotle on, and misunderstood from about
that time. In other words, the universal judgment applies to 'all of A,' 'each A,'
'everything that is an A' and not to the class of A's. The judgment is in compre-
hension and not in extension or, to be quite exact, in extension through comprehen-
sion. Cf. E. Meyerson, Le Cheminement de Ta penske (Paris, 1931).
But then a problem presents itself: why are the 'aggregate of all aggre-
gates' and analogous 'aggregates' illegitimate? In other words, what are
the curious notions to which correspond no 'classes,' concepts which are
not as others and which have no 'totalizable' extension?
The theory of types did not answer this question. It confined itself to
indicating a general criterion for 'illegitimate totalities,' and to providing
a means for avoiding their formation. It did not explain why the aggre-
gate of all things red is a legitimate aggregate whereas that of things not-red
is not.42
Now we believe that such an explanation is possible and that the 'essen-
tially paradoxical notions,' that is, notions which, by essence, possess the
curious property of applying to themselves in one way or another and of
not being able to determine totalizable multiplicities (of classes) form a very
special group of notions.43
An example, which, besides is more than a mere example, will enable us
to elucidate the problem.
We have already concerned ourselves with the aggregate of all aggregates,
and we ascertained that it could not be formed. In a certain sense, it
proved itself too large. Thus there was, so to speak, no room for it any-
where. It could not be found within itself, and, on the other hand, since it
embraced (or claimed that it embraced) the entire universe of aggregates, it
could not be pushed out of itself either.
But the 'aggregate of all aggregates' is not, one realizes, 'the greatest
aggregate.' The aggregate of all objects is still far more extensive. Not
the aggregate of all individual and real objects, of course, but that of all
objects whatever, of all the 'somethings,' of all the objects of thought. It
is quite clear that this 'aggregate'-the universal class of the logistics-
is essentially a 'paradox' and that it must embrace itself as a member,
since it embraces all, for it, in the strictest sense of the term, there is n
room outside of itself; just as for Aristotle's Universe there was no room
outside of it, where it could be located. Aristotle concluded-and justly
so-that his Universe was not in a place. His successors, however, had to
accept another solution, namely, that Aristotle's Universe does not exist.44
42 Once again, the theory of types has been vitiated by the point of view of exten-
sion.
43 These notions are the logical constants to which Russell attributes 'essential
ambiguity of meaning.'
44 The fact of the non-existence of paradoxical aggregates was recognized-or
postulated-almost immediately after their discovery. Poincar6, E. Borel (cf.
Introduction a la theorie des fonctions, p. 164 sq.), Russell himself (cf. Principles of
Mathematics pp. 97, 105, 525, etc.), Hadamard (cf. Borel, op. cit., p. 158), Hilbert,
Mirimanoff ("Les Antinomies de Russell et de Burali- Forti et le probleme fonda-
mental de la thborie des ensembles," Enseignement Mathtmatique (1917), pp. 38,
48, 49) all recognized that these 'aggregates' do not exist. But not one of them
We see ourselves forced to adopt an analogous solution: the aggregate of

all objects, of all the 'somethings,' the 'universal class' of the logistics,
does not exist; is not an object of thought. The multiplicity of the 'some-
things' is not totalizable because the notion of object cannot serve as a uni-
fying link; as a matter of fact, it is not opposed to anything, and does not
exclude anything from its field of application; it is so indetermined that the
multiplicity to which it applies is, in the strongest sense of terms, undefined
and unlimited.46 The impossibility of totalizing the multiplicity of beings
(objects of thought) is thus explained by the absolute emptiness of the
notion of object.46
All this, after all, is not particularly astonishing nor even new. As a
matter of fact, from Plato on it has been well known that every notion is a
peras opposed to and imposed on the apeiron, and from Aristotle on no one
has been unaware of the fact that the concept of being is not a concept like
others: it is a category, it is not a genus, not even the supreme genus.
Thus it is easily understood that the concept of being does not determine a
class. The expression the 'aggregate of beings' conveys no meaning.
The 'universal class' is an absurdity.47
This immediately explains why multiplicities 'determined' by negation
-we have already pointed it out-are always 'paradoxical,' that is, are
never totalizable. In other words, why a purely negative judgment-what
was formerly called an infinite judgment-never determines an aggregate
nor a class.
As a matter of fact, the non-totalizable multiplicity of beings is, as we
have just seen, undetermined and inexhaustible. Now, if every real concept
outlines, as it were, a limited field in this apeiron, what remains outside of
this peras obviously remains an apeiron. In other words, if the 'universal
class' is not a class, what remains of it, outside of a determined class, is not
one either.
Therefore, there exist no negative classes. There is a class of 'red' ob-
could show why it was impossible to form them, that is, why one did not have the
right to apply in certain cases modes of reasoning admissible and legitimate in others.
For it is not sufficient to show that, being contradictory, these 'objects' of thought
cannot exist-that is what constitutes the paradox; what we need is to show, with-
out having recourse to the consequences, that is, without reasoning with the conse-
quences as a point of departure, a priori and not a posteriori, the vice of the definition
or of the 'logical' process which engenders them. That is what we have tried to do.
45 Every concept, and every notion, which claims to embrace all means nothing.
Negation, opposition, is an element essential to thought as well as to being.
46 Hegel had already recognized the equivalence-in the void of meaning-of
pure Being and of Nothingness.
47The multiplication of 'universal classes' does not make this absurdity less
absurd.
MANIFOLD. AND CATEGORY :19
jets; but not a class of 'not-red' ones. Thus one can ask oneself-and
having done so, reply in the negative-whether the concept 'non-red' is
really a concept. At all events, it is certainly not a real concept. For if
the essential defect (or merit) of the concept of 'being' is that of being
absolutely undetermined, it is clear that the determination which the nega-
tion of 'red' brings to it is perfectly illusory.48
The notion of being is not the only concept inaccessible to limitation and
thereby possessing an absolutely undetermined 'extension.' Medieval
logic was familiar with several others, very closely linked to that of being,
and equipollent with it. These notions, which medieval logic designated as
transcendentals, extend as far as extends being and are 'convertible' with it.
Such are the esse, unum, bonum, verum, en8, res, etc.
This list of transcendentalls' that is, of primary notions or categories,
determining being without limiting it, is certainly not exhaustive nor cor-
rect. But the notion itself is a valuable one and must be retained, though
transposed. 'Transcendental' notions, those which in the theory of types
would be of a 'transfinite type,' are nothing less than the notions which
constitute the essential categories of general logic and ontology. Linked
with those of being, they are of the same power as the latter. Conversely-
and this can serve as an external mark of these notions-the notions possess-
ing the same power as that of being, extending as far as being extends itself,
are 'transcendental' notions.
Such, for example, are the notions of unity and multiplicity, the notions
of the aggregate and of number, the notions of concept, of proposition, of
relation. Everything, as a matter of fact, every 'something' is a unity
and, at the same time, an element of a multiplicity, or of an aggregate.
Everything is the object of a concept, of a judgment, of a relation.49
Now, all these notions-Husserl's 'empty' notions50 which constitute
the true 'constants' of thought and of being, of logic, and of ontology,
are 'formal' notions, and they all possess, in one way or another, the prop-
48 Determinatio est negatio; 8ed negatio non e8t determninatio, at least, non immediate
et per se.
49 That is why they are 'essentially ambiguous,' that is, they necessarily appear
in any logical symbolism.
50 E. Husserl, Formale und transzendentale Logik (Halle, 1929), ?24, p. 68: "Wenn
man die naturgemass weiteste Allgemeinheit der Begriffe Menge and Zahl erwagt und
die ihren Sinn bestimmenden.Begriffe Element bzw. Einheit, so erkennt man, dass
die Mengen-und Ansahlenlehre bezogen ist auf das Leeruniversum Gegenstand
uiberhaupt oder Etwa8 iuberhaupt, in einer formalen Allgemeinheit die jede Sachhaltige
Bestimmung von Gegenstanden prinzipiell ausser Betracht lasst; ferner dass diese
Disziplinen speziell fur gewisse Ableitungagestalten des Etwas iAberhaupt interes-
siert sind. Solche Ableitungen sind neben Menge und Anzahl (endliche und un-
endliche), Kombination, Relation, Reine, Verbindung, Ganzes und Teil, usw."
erty of reflexiveness, that is, of being able to become their own object.
Thus numbers can be numbered and propositions can be made objects of
judgment. And that is why, in the last analysis, we can make judgments
which bear on all judgments and on all aggregates, and why we cannot form
the totality of aggregates or the totality of judgments.
ALEXANDRE KOYR1.
ECOLE PRATIQUE DES HAUTEs ETUDES,
SORBONNE, PARIS.

International Phenomenological Society Philosophy and Phenomenological Research

Uploaded by

Copyright:

Available Formats

International Phenomenological Society Philosophy and Phenomenological Research

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

International Phenomenological Society Philosophy and Phenomenological Research

Uploaded by

Copyright:

Available Formats

International Phenomenological Society

Manifold and Category

VOLUME IX, No. 1 AUGUST 1948

MANIFOLD AND CATEGORY*

In an article in Volume VI, Number 3 of this journal, I investigated the

I. THE NOT PREDCOABLE

Thus, all concepts are- "not-predicable." Nevertheless, the statement

In order really to grasp the nature of the paradox of the "not-predicable

(1) Imp (F) = -F (F)

From this definition results, a

(2) (F) Imp (F) =_ F (F).

The conjunction of the two following implications is thus obtained:

3 (a) (F). Imp (F) D) F. (F).

Let us now substitute for F the particular value Imp.; we obtain:

4 (a) Imp (Imp) D- Imp (Imp).

4 (b) - Imp (Imp) D Imp (Imp);

The antinomy is evident."

formula-even a symbolic one-have a well defined meaning;9 it has not

II. THE CLASS OF ALL CLASSES

12 It is these reflexive concepts which engender the "paradoxes, as Russell very

Now it is easy to understand the terror which it inspired at the beginning.

dently independent of the matter of knowing whether aggregates other than

Can an aggregate contain itself as an element? Can a totality be a part

16 Ibid., p. 226. Cf. B. Russell, Principles of Mathematics, Chap. X. It is well

20 A. Fraenkel, op. cit., p. 232.

22 Cf. B. Russell, Principles of Mathematics, App. B, p. 526; cf. Les paradoxes de la

The vicious circle principle-and the prohibitions it implies, both equiva-

26 Cf. Jorgensen, op. cit., p. 169 sq.

Thus A is the negation of V. We have

Do the unpleasant and awkward consequences of the theory of types

29 Principia Mathematica, I, 31: "There are no classes which contain objects of

It should be observed, first of all, that to speak of a plurality of systems of

That the theory of types has no meaning is an unavoidable conclusion

32 B. Russell, "Les paradoxes de la logique", RMM, 1906, p. 644. "Cette theorie

opposition is quite particularly stiffened by the fact that neither Russell

35Cf. A. Koyr6, "Remarques sur les nombres de M. Russell," RMM, 1912.

Let us summarize our position, by opposing it, to that of Russell. Rus-

87 Already in the Principles of Mathematics, Russell established the very important

Let us now return to Russell's paradox of 'the (class) of all aggregates

39 The Principia Mathematica, based on the theory of types, maintain (Vol. I,

We see ourselves forced to adopt an analogous solution: the aggregate of

You might also like