The British Society for the Philosophy of Science

Category Theory and the Foundations of Mathematics

Author(s): J. L. Bell
Source: The British Journal for the Philosophy of Science, Vol. 32, No. 4 (Dec., 1981), pp.
349-358
Published by: Oxford University Press on behalf of The British Society for the
Philosophy of Science
Stable URL: http://www.jstor.org/stable/687305
Accessed: 21-03-2018 04:43 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 [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

Oxford University Press, The British Society for the Philosophy of Science are
collaborating with JSTOR to digitize, preserve and extend access to The British Journal for
the Philosophy of Science

This content downloaded from 128.196.130.121 on Wed, 21 Mar 2018 04:43:50 UTC
All use subject to http://about.jstor.org/terms
Brit. J. Phil. Sci. 32 (1981), 349-358 Printed in Great Britain 349

Category Theory and the

Foundations of Mathematics*

by J. L. BELL

i Background to the Emergence of Category Theory

2 Category Theory: a Foundation for Mathematics?
3 The Nature and Significance of Category Theory

In recent years the mathematical discipline of category theory has emerged to

play a growing role in the perennial debate about the foundations of
mathematics. Some of its proponents have suggested that it should replace
set theory as the 'official' foundation for mathematics. My purpose here is to
discuss the background to this claim and to examine its tenability.

I BACKGROUND TO THE EMERGENCE OF CATEGORY THEORY

With the creation and refinement of set theor

Zermelo, von Neumann and others, the problem of
ture for the elaborate developments in areas of ma
algebra, analysis and topology appeared to be so
(structured by membership) came to be regarde
(those of a constructive tendency such as the in
notable exceptions) as the source of 'raw material' for building the
structures required in all branches of mathematics. Thus, although the
notions of structure and operation on structures had come to play a
fundamental role in most mathematical disciplines, these notions were not
taken as primitive, but were themselves explicated by reduction to the more
fundamental notions of set and membership. On the other hand, the
perspicuity and apparent reliability of the set-theoretical framework
enabled mathematicians not primarily interested in set theory (i.e., the vast
majority) simply to take this reduction for granted and instead concentrate
on isolating and axiomatizing the basic mathematical features of these
structures. With the rise of abstract algebra in the 1930o it came to be
recognised that these features and the laws governing them (I am thinking
particularly of the notions of isomorphism, homomorphism, substructure,
etc.) had a kind of universality and even inevitability that was apparently
independent of their set-theoretical origin. Moreover, it was observed that
many of the basic notions of abstract algebra could be derived from the

* Paper delivered at the Conference of the British Society for the Philosophy of Science,
Chelsea College, September 1980. I am indebted to Solomon Feferman and Mike Hallett for
valuable criticisms and suggestions.

This content downloaded from 128.196.130.121 on Wed, 21 Mar 2018 04:43:50 UTC
All use subject to http://about.jstor.org/terms
350 J. L. Bell

single idea of structure-preserving function, or, as it is now customarily

known, morphism. Thus the attitude gradually emerged that the crucial
characteristic of mathematical structures is not their internal constitution as
set-theoretical entities but rather the relationships among them as embodied
in the network of morphisms.' This attitude, strikingly reminiscent of the
operational structuralism associated with linguistics and psychology, was
exemplified most strongly by the Bourbaki school in France, which had
proposed a 'structuralist' account of mathematics as far back as the I930s.
However, although the account of mathematics they gave in their 'Elements'
was manifestly structuralist in intention, in actuality they still defined
structures as sets of a certain kind, thereby failing to make them truly
independent of their 'internal constitution'. In fact, it was not until the early
1940s that an axiomatic framework emerged which gave better expression to
the idea of operational structuralism in mathematics: I am referring to
Eilenberg and MacLane's theory of categories andfunctors. Here for the first
time we have an axiomatic theory which takes the notions of structure and
morphism as primitive and is indifferent to any particular set-theoretic
constitution that structures may have.
It will be helpful to give the formal definitions of category and functor.2 A
category E consists of two classes, the members of the first of which-
denoted by the letters X, Y,... -are called objects3 (structures) and the
members of the second of which-denoted by the lettersf, g, ... -are called
arrows4 (morphisms). Each arrowf is assigned an object X as domain and an
object Y as codomain, indicated by writing f: X -* Y. If g is any arrow
g: Y -- Z with domain Y, the codomain of f, there is an arrow fg: X -- Z
called the composition of f and g. For each object Y there is an arrow idy:
Y -- Y called the identity arrow of Y. These notions are assumed to
satisfy the following identity and associativity axioms:

f idy = f, idy g = g, f(gh) = (fg)h

for any arrows f: X -- Y, g : Y -- Z, h : Z -- W.
Note that these axioms are readily formulated within an appropriate first-
order language: the resulting theory is first-order category theory.
Given two categories D and E, afunctor F from D to E consists of a pair of
functions (both denoted by F), one from the class of objects of D to that of E,
and the other from the class of arrows of D to that of E, such that

if f : X - Y in D, then F(f) : F(X) -- F( Y) in E;

F(idx) = idF(X) for all objects X of E,

1 In particular, it came to be seen that the notion of identity appropriate for structures is not set-
theoretic equality but isomorphism, an idea going back in essence to Dedekind and Klein.
2 For an account of category theory, see, e.g. MacLane [1971].
The use of the term 'object' here in place of 'structure' is intended to reinforce the idea that
the theory is not about particular structures, but mathematical structures in general.
4 The term 'arrow' here is a metonym and stems from the habit of depicting functions as
arrows.

This content downloaded from 128.196.130.121 on Wed, 21 Mar 2018 04:43:50 UTC
All use subject to http://about.jstor.org/terms
 Category Theory and the Foundations of Mathematics 3 5 I
and

F(fg) = F(f)F(g) for any composable arrows f, g of D.

A functor may be thought of as a morphism of categories.

Categories and functors abound in mathematics; as examples of the
former one may consider:

Set : objects, all sets; arrows, all (set) functions.

Grp: objects, all groups; arrows, all group homomorphisms.
Top: objects, all topological spaces; arrows, all continuous maps.
Prominent functors include:

the 'forgetful' functor from Grp or Top to Set which assigns to each group
or topological space its underlying set (this functor has the effect of
'forgetting' the structure!);
the 'free group' functor from Set to Grp which assigns to each set the free
group it generates;
the 'homology group' functors from Top to Grp which, for each natural
number n, assign to each topological space X its nth homology group H,(X).

This last example played a key role in the creation of category theory. In
fact, it was the problem of providing a smooth theory of homology groups,
and of the limiting processes involved in their construction, which first
suggested the idea of a functor to its creators. It is clear that, in any case,
Eilenberg and MacLane regard the idea of functor as being in some sense
more fundamental than that of category. As they remark in their original
paper: 1

It should be observed that the whole concept of a category is essentially an

auxiliary one. Our basic concepts are essentially those of a functor and of a
natural transformation.... The idea of a category is required only by the precept
that every functor should have a definite class as domain and a definite class as
range, for the categories are provided as the domains and ranges of functors.

Thus, as far as its creators were concerned, the notion of category was only
introduced in order to furnish the more basic notion of functor with set-
theoretic legitimacy. Indeed the essentially operational spirit of the
enterprise is revealed when they go on to say:

Thus one could drop the category concept altogether and adopt an even more
intuitive standpoint, in which a functor such as 'Hom' [i.e. the functor which
assigns to each pair of objects of a category the collection of arrows between
them] is not defined over the category of 'all' groups but for each particular pair
of groups which may be given. This standpoint would suffice for applications,
inasmuch as none of our developments will involve elaborate constructions on
the categories themselves.

However, the observation that categories were well-nigh ubiquitous in

1 Eilenberg and MacLane [19451.

This content downloaded from 128.196.130.121 on Wed, 21 Mar 2018 04:43:50 UTC
All use subject to http://about.jstor.org/terms
352 J. L. Bell
mathematics soon convinced the mathematicians concerned of the funda-
mental importance and autonomy of the notion of category. From a
philosophical point of view, a category came to be thought of as an
embodiment of the 'abstract structure' that all its constituent objects
exemplify, or even literally to be that 'abstract structure'. (Thus for example
the category Grp is 'group structure'.) So category theory itself came to be
viewed as a theory of (mathematical) Structure.
Another consequence of the ubiquity of categories in mathematics was
that they came to be regarded as mathematical objects in their own right and
thereby subject-as are all mathematical objects-to a whole range of
mathematical constructions. One of the most important of these is the
construction offunctor categories. Given two categories D and E, the functor
category ED has as objects all functors from D to E and as arrows the so-
called natural transformations (the definition of which we omit) between
such functors. But here the set-theoretic definition of category (involving, as
it does, the notion of class) causes a snag. For if the collection of objects or
arrows of D is a proper class in the sense of G6del-Bernays set theory (such a
category is called large; if both collections are sets the category is called
small), then so is each functor from D to E and consequently these functors
cannot be collected into a class. Thus the current set-theoretical framework
does not allow the formation of functor categories ED when D is large (e.g.
when D is any of the examples given above). And, for similar reasons, th
same is true of the putative category of all categories (small or large).
Now the restrictions on the formation of classes (Zermelo) or on the
operations to which these classes can be subjected (von Neumann) had of
course been imposed originally in order to resolve the set-theoretic
antinomies. These restrictions were not judged by the majority of mathema-
ticians to be of intolerable severity because in most areas of mathematical
practice the 'proscribed' collections such as the class of all sets simply did
not play any role. (In fact, much early work in the foundations of set theory
was devoted to eliminating references in mathematical arguments to these
proscribed classes.) But the natural tendency of category theory to form
categories consisting of all objects of a certain kind once again thrust these
proscribed classes, with their attendant difficulties, into the foreground; and
the restrictions on the formation and manipulation of these classes imposed
by the official set-theoretic framework came to be regarded by some
category theorists as an irksome and possibly even unnecessary curtailment
of their mathematical activity. This feeling, reinforced by the steady
incursion of category-theoretic techniques and ideas into many branches of
mathematics (regarded as 'subversion' by some of the more conservatively
minded) led certain category-theorists first to question the adequacy of the
current set-theoretic foundation, then its necessity, and finally to propose
category theory itself as a possible foundation for mathematics. I will argue
here that, although there are grounds for accepting that current set-
theoretical foundations are inadequate for 'full' category theory, and

This content downloaded from 128.196.130.121 on Wed, 21 Mar 2018 04:43:50 UTC
All use subject to http://about.jstor.org/terms
 Category Theory and the Foundations of Mathematics 353

although category theory unquestionably has foundational significance, its

very nature makes it unsuitable as an exclusive foundation for mathematics.

2 CATEGORY THEORY: A FOUNDATION FOR MATHEMATICS?

In what sense could category theory serve as a foundation for m

There seem to be (at least) two possible senses: first, a strong se
all mathematical concepts, including those of the current logico-
metatheoretical framework for mathematics, are explicable in category-
theoretic terms. And secondly, a weaker sense in which one only requires
category theory to serve as a (possibly superior) substitute for axiomatic set
theory in its present foundational role.
Now it seems to me implausible that category theory is, or could be,
foundationally adequate in the stronger sense. For consider the meta-
theoretical framework in which category theory (or any other first-order
theory) is embedded. This framework has two basic aspects: the com-
binatorial, which is concerned with the formal, finitely presented properties
of the inscriptions of the ambient formal language, and the semantical, which
is concerned with the interpretation and truth of the expressions of that
language. Neither one of these aspects is-at present-reducible to the
other. The former deals with intensional objects such as proofs and
constructions whose actual presentation is crucial, while the latter employs
extensional objects such as classes whose identity is determined in-
dependently of how they may be presented or defined. So if category theory
is to furnish a foundation for mathematics in the stronger sense, it must
provide convincing accounts of both of these aspects. But a category is
defined to be a class of a certain kind, and classes are extensional, while
combinatorial objects are generally not. Since there is no reason to suppose
that a satisfactory account of intensional (i.e. non-extensional) objects can be
given solely in terms of extensional ones, it seems to me that category theory
as currently formulated in terms of classes must fail to provide a faithful
account of the combinatorial aspect at least. (Of course for similar reasons
this 'weakness' is shared by set theory.)
As far as the semantical aspect is concerned, we recall that the interpre-
tation of an expression of a classical first-order language involves a reference
to classes or pluralities in an essential way (as the 'range' of the variables in
the expression). In particular, grasping the concept of logical truth for
sentences of classical first-order languages requires that one has already
grasped the concept of class. To put it another way, the concept of class is
epistemically prior to the concept of (classical) logical truth. So if category
theory is to serve as an autonomous basis for classical semantics, and in
particular give a satisfactory independent account of logical truth, it must be
possible to give an explication of the notion of class (at least in so far as it is
involved in deriving the concept of logical truth) solely in terms of the notion
of category, and without already having defined the latter notion in terms of

This content downloaded from 128.196.130.121 on Wed, 21 Mar 2018 04:43:50 UTC
All use subject to http://about.jstor.org/terms
354 J. L. Bell

classes. But this seems to me highly dubious, for it is surely the case that the
unstructured notion of class is epistemically prior to any more highly
structured notion such as category: in order to understand what a category
is, you first have to know what a class is.' This also applies, mutatis mutandis,
to the notion of functor whose explication involves grasping the epistemi-
cally prior notion of operation.
It seems to me that these considerations show that category theory
as currently conceived is not capable of serving as a foundation for
mathematics in the strong sense. Of course, this is hardly surprising
since it is widely recognised that no single foundational scheme is at present
capable of providing a convincing explication of both combinatorial and set-
theoretical objects. What we actually possess is an informal system of
'multiple' foundations, with distinct combinatorial and set-theoretical
components.
Let us turn now to the weaker sense in which category theory could serve
as a foundation for mathematics, namely as a substitute for axiomatic set
theory in its current foundational role. One possible means of achieving this
would be to construct a formal interpretation of some 'foundationally
adequate' first-order version of set theory (e.g. Zermelo-Fraenkel set theory
with choice, ZFC) in a suitable consistent extension T of first-order category
theory in such a way that the interpretation of any theorem of ZFC is
provable in T. Now this has in fact already been achieved:2 T may be take
to be the theory of elementary toposes (a finite extension of the first-order
theory of categories), augmented by certain other axioms, notably category-
theoretic versions of the axiom of choice and the axiom scheme of
replacement.
The interpretation of ZFC in T is performed as follows. First, one
observes that, by the well-known Mostowski collapsing lemma, the notion
transitive set essentially corresponds to the notion extensional well-founded
relation. Now a relation may be regarded as a function from its field to the
power set of its field, a fortiori as an arrow within the category of sets.
Furthermore, extensionality and well-foundedness of a relation can be
translated into purely categorical properties of the corresponding arrow.
These properties may be 'lifted' to any category which is a model of T (in
fact any topos). An arrow possessing these properties then provides a
category-theoretic formulation of the notion of transitive set and is
accordingly called a transitive set-arrow. In view of the fact that every set is a
subset of a transitive set, one defines a set-arrow to be a subarrow of a
transitive set-arrow, i.e. a pair (f, r) consisting of a transitive set-arrow r and
a monic arrow f such that the codomain off coincides with the domain of r.
One can then define the relation of membership between set-arrows and with
some difficulty show that, for any model E of T, the collection of set-arrows

For a similar conclusion, see Feferman [1977]. My argument here owes much to this article.
2 See Osius [19741.

This content downloaded from 128.196.130.121 on Wed, 21 Mar 2018 04:43:50 UTC
All use subject to http://about.jstor.org/terms
 Category Theory and the Foundations of Mathematics 355

in E, structured by this new relation of membership, is a model of ZFC.

Thus, if we interpret the notion of set as set-arrow we obtain a translation' of
ZFC into T.
We may conclude from all this that it would be technically possible to
a purely category-theoretic account of all mathematical notions expr
within axiomatic set theory, and so formally possible for category theor
serve as a foundation for mathematics insofar as axiomatic set theory do
On the other hand, as we have seen above, the actual translation of set t
into category theory is awkward and has (unlike the basic category-theor
notions themselves) a factitious character which renders it unsuitabl
means of formalizing those mathematical notions which are norma
expressed set-theoretically.
What this translation of set theory really amounts to is the replaceme
the notion 'mathematical object as set' by the notion 'mathematical objec
pair of arrows (of a certain kind) in a category'. It would seem, however,
a more convincing and natural formalization of mathematics within
category theory would be obtained if mathematical objects could be
construed as categories tout court. (This would also be more in keeping with
the structuralist view that mathematical objects are given as structures and
that categories provide an embodiment of the idea of structure.) To this end
Lawvere has formulated a first-order theory2 1 in the language of categories
which purports to capture the characteristic features of the category of all
small categories. Furthermore, he claims that all mathematics currently
formulable within axiomatic set theory can be expounded within 1.
Unfortunately, certain technical defects in E have come to light which make
it difficult to assess whether Lawvere's bold program can be carried out in its
original form. And even if we grant that these technical difficulties will be
overcome, it seems doubtful whether a system like I will ever be accepted as
an autonomous 'foundation' for mathematics in the sense that, say, ZFC is.
For, in developing the notions of workaday mathematics within I it seems to
be necessary to bring in the notion of set 'through the back door', so to speak,
in the form of 'trivial' or 'discrete' categories. Of course, this is not
technically an appeal to the notion of set because the definition of discrete
category is given entirely within 1. But the question automatically arises as
to exactly why, in introducing ordinary mathematical notions into the
theory, one must make a detour through the somewhat opaque notion of
discrete category. It is difficult to see how this can be explained except by
appeal to the notion of 'unstructured' category, i.e. set. This, it seems to me,
will inevitably make a system like E appear artificial as a 'foundation' for
mathematics, despite the beauty and naturalness of the category-theoretic
notions themselves.

1 There is also an obvious translation of T into ZFC: simply interpret the notion object as set
and arrow as mapping. Thus ZFC and T are actually formally equivalent.
2 Lawvere [1966].

This content downloaded from 128.196.130.121 on Wed, 21 Mar 2018 04:43:50 UTC
All use subject to http://about.jstor.org/terms
356 J. L. Bell

3 THE NATURE AND SIGNIFICANCE OF CATEGORY THEORY

In expressing these negative views about category theory a

foundation for mathematics, I do not mean to imply that c
theoretical foundations should be taken as adequate for eve
category-theorists want to do, nor do I wish to claim that catego
just another branch of mathematics, devoid of foundational cont
contrary, with regard to the former, I have already pointed
operations on large categories which appear so natural to categor
are not justified by current set-theoretical foundations and
demand an extension or reformulation of the set-theoretical framework to
accommodate them. In this connection, however, it should be noted that the
failure of set theory to justify the unlimited application of category-theoretic
operations is a consequence of its success in eschewing the overcomprehen-
sive collections which were originally deemed responsible for the paradoxes.
(This fact was clear to Eilenberg and MacLane, for they point out in their
original paper that in category theory 'no essentially new paradoxes [over
and above those of intuitive set theory] are apparently involved'.) In fact, set
theory's failure to embrace the notion of arbitrary category (or structure) is
really just another way of expressing its failure to capture completely the
notion of arbitrary property. This suggests the possibility that a suitable
framework for 'full' category theory could reasonably be sought within a
theory of such arbitrary properties. Although such a theory would certainly
have to transcend current set theory, it would however be under no obligation
to appear category-theoretic in nature. The notion of a category would in all
probability continue to be a derived notion and not a primitive one.1
Nevertheless, category theory is more than just another abstract mathe-
matical theory. Like set theory, it provides a general framework for dealing
with mathematical structures, and-again like set theory-it achieves this
by transcending the particularity of structures. But set theory and category
theory go about doing this in entirely different ways. Set theory strips away
structure from the ontology of mathematics leaving pluralities of structure-
less individuals open to the imposition of new structure. Category theory, on
the other hand, transcends particular structure, not by doing away with it,
but by generalising it, that is, by producing an axiomatic general theory of
structure. The success of category theory, and its significance for found-
ations is due to the ubiquity of structure in mathematics.
It may be said that category theory, while still dependent on set theory as
the ultimate source of mathematical entities, nonetheless frees mathematics
from the particularform imposed on it by having to regard these entities as
pluralities of elements. The power and fertility of the 'element-free'
formulation of mathematics provided by category theory is most strikingly
realised by applying it to set theory itself. This gives rise to a startling new

1 See Feferman [1977] for an 'intensional' theory of partial operations and properties in which
much of 'full' category theory can be formulated.

This content downloaded from 128.196.130.121 on Wed, 21 Mar 2018 04:43:50 UTC
All use subject to http://about.jstor.org/terms
 Category Theory and the Foundations of Mathematics 357

kind of 'model' of set theory called an (elementary) topos.' A topos is a

category E which has the following features in common with the category
Set of sets.
(i) E has a terminal object and all finite products.
(2) There is in E a 'truth-value' object Q which plays the same role in E as
the truth-value set 2 = {o, I} plays in Set, i.e. for each object X there is a
natural correspondence between subobjects of X and arrows from X to Q
('characteristic functions' on X).
(3) For each object X of E there is a 'power object' PX in E which plays
the formal role of a power set of X in E.
These conditions can all be formulated in the first-order language of
category theory: hence the use of the term 'elementary'. In addition to the
category of sets, examples of toposes include the category of sheaves of sets
on a topological space (a sheaf may be thought of as a set varying through
space) and the category of all diagrams X0 --+ X - X2 --+ ... of sets Xi (such
a diagram may be regarded as a set varying through discrete time). A topos
may thus be thought of as a generalized model of set theory in which the
'sets' are varying in some manner and are generally not determined by their
'points'.
Topos theory has striking connections with logic. For example, in any
topos one can give natural definitions of arrows (to be thought of as 'logical
operations' in E), -: Q --+; A, V, >: Q x Q --+ in such a way that, if we
regard these arrows as algebraic operations on Q, the resulting algebra is a
Heyting algebra, i.e. satisfies the laws of intuitionistic propositional logic. In
this sense intuitionistic logic is 'internalised' in a topos (although of course
there will be toposes in which classical logic is internalised, e.g. the category
of sets). With some justice, then, we may regard a topos as an instrument for
reducing logic to mathematics, the remarkable thing being that the logic
obtained is not (in general) classical, but intuitionistic.
Because the internal logic of a topos is intuitionistic, one would intuitively
expect mathematics done 'in a topos' to be in some sense constructive.
However, at present it is not clear just how this intuitive expectation is to be
explicated. Certainly there seem to be strong connections between in-
tuitionistic mathematics and the forms that mathematical arguments take
within a (general) topos. This applies both to arithmetic and to the analysis
of the continuum, which can be made to resemble the intuitionistic
continuum in that it will not be built up from 'points'. In fact one can
construct toposes in which the unit interval [o, I] fails to be compact and
in which every function from reals to reals is continuous. Topos theory
thus provides an entirely new framework for modelling intuitionistic
mathematics.

It is generally agreed that the most significant recent contribution t

mathematical logic was Paul Cohen's construction of models of set theory in

' For an introductory account of topos theory, see Goldblatt [i979].

This content downloaded from 128.196.130.121 on Wed, 21 Mar 2018 04:43:50 UTC
All use subject to http://about.jstor.org/terms
358 J. L. Bell

which the continuum hypothesis and the axiom of choice fail. The
techniques he invented have led to an enormous proliferation of essentially
different models of set theory and the rise of a 'relativistic' attitude toward
the set-theoretical foundations of mathematics. This attitude involves
abandoning (or, at least, reserving judgment about) the idea that mathem
tical constructions should be viewed as taking place within an 'absolu
universe of sets with fixed and predetermined properties. Instead, o
works in suitably chosen models of set theory having the properties req
to carry out the constructions in question. Now topos theory carrie
idea1 to its conclusion: it provides models of set theory in which eve
logic can be tailored to suit the construction. This fact may provide furt
evidence for logical pluralism within mathematics, as already exemplified
the existence of classical and constructive logics. Be this as it may,
paramount achievement of topos theory is to have identified the basic co
set theory in such a way that the set concept becomes manifest in conte
(such as algebraic geometry or constructive mathematics) where befo
presence was at most tacit. Thus category theory, far from being in
opposition to set theory, ultimately enables the set concept to achieve a new
universality.
London School of Economics

REFERENCES

EILENBERG, S. and MACLANE, S. [1945]: 'General Theory of Natural Equivalences',

Transactions of the American Mathematical Society, 58, pp. 239-94.
FEFERMAN, S. [1977]: 'Categorical Foundations and Foundations of Category Theory' in Butts
and Hintikka (eds.), Logic, Foundations of Mathematics and Computability Theory.
Dordrecht: D. Reidel, pp. 149-69.
GOLDBLATT', R. [1979]: Topoi, the Categorial Analysis of Logic. Amsterdam: North-Holland
Publishing Company.
LAWVERE, F. W. [1966]: 'The Category of Categories as a Foundation for Mathematics', in
Proceedings of the La Jolla Conference in Categorical Algebra, Springer-Verlag, pp. 1-20.
MACLANE, S. [1971]: Categories for the Working Mathematician. Berlin: Springer-Verlag.
OsIus, G. [ 1974]: 'Categorical Set Theory: A Characterization of the Category of Sets', Journal
of Pure and Applied Algebra 4, pp. 79-11I 9.

1 It is interesting to note that Lawvere and Tierney's formulation of the notion of topos was
partly motivated by the idea of analysing Cohen's proof of the independence of the continuum
hypothesis in category-theoretic terms.

This content downloaded from 128.196.130.121 on Wed, 21 Mar 2018 04:43:50 UTC
All use subject to http://about.jstor.org/terms