The Genesis and Development of Set Theory

Author(s): Phillip E. Johnson

Source: The Two-Year College Mathematics Journal, Vol. 3, No. 1 (Spring, 1972), pp. 55-62
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/3026799
Accessed: 20-06-2016 10:23 UTC

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

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].

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access
to The Two-Year College Mathematics Journal

This content downloaded from 104.239.165.217 on Mon, 20 Jun 2016 10:23:00 UTC
All use subject to http://about.jstor.org/terms
 Historical

If one year can be specified as the time

when set theory started, that year should
probably be 1874, the year in which Georg
Cantor's paper [2] was published establish-
ing the countability of the set of real
algebraic numbers and the noncountability
of the set of real numbers. The proof that
the set of real numbers is not countable
used nested intervals instead of Cantor's
famous diagonal process, which appeared
in a later work [9]. According to Fraenkel,
Cantor himself had first thought that the
continuum could be put in one-to-one
correspondence with the set of natural
numbers [12, p. 237]. Cantor observed in
the introduction of his paper that com-
bining the two theorems gives a result first
proved by Liouville-that in each given
The Genesis interval there exist infinitely many tran-

and Development scendental (nonalgebraic) real numbers.

There was some difficulty surrounding
of Set Theory the publication of Cantor's next article [3]
in 1878. The work remained in the publish-
ing room of Crelle's Journal longer than

Phillip E. Johnson usual for that time, apparently because

Cantor's ideas were rejected by Leopold
Kronecker, who was on the editing staff of
the journal [12, p. 198]. The work was
Phillip E. Johnson is Assistant PrQfessor of
eventually published in the journal, but it
Mathematics and Coordinator of Mathe-
was the last article Cantor published there.
matics Education at the University of
North Carolina at Charlotte. Prior to 1971,
One-to-One Correspondence
he had taught at the University of Rich-
The concept of one-to-one correspon-
mond and at Vanderbilt University. This
dence between sets is the fundamental idea
article is adapted from his book, A History
in Cantor's 1878 paper. It contains some
of Set Theory, which will be published this
important theorems of this kind of relation
spring.
between various sets and some suggestions
of a classification of sets. At the beginning
of the work he introduced the idea of

U, equivalence and used it to explain the idea

of power in a more concrete manner than
he did later in his systematic presenta-
tion [10]. Cantor borrowed the term
"power" from Jakob Steiner, who used it
in a special, but allied, sense.

55

This content downloaded from 104.239.165.217 on Mon, 20 Jun 2016 10:23:00 UTC
All use subject to http://about.jstor.org/terms
 A subset (Bestandteil: now proper sub- terminology) infinity of dimensions. He
set) of a finite set always has a power less also conjectured that the two powers of
than that of the set itself. That this is not the rational numbers and the real numbers
the case with infinite sets was first noticed exhaust all possibilities for infinite subsets
by Bernard Bolzano several years earlier; of the continuum. Time has shown that he
this fact was used in a definition of infinite was overly optimistic in that it has now
by Richard Dedekind, independently of been shown in the realm of axiomatic set
Bolzano and Cantor, in 1887 [11, p. 29]. theory that the conjecture is neither
For example, it is easy to see, as Cantor provable nor disprovable.
pointed out, that the set of natural
numbers has the same power as that subset Theory of Point Sets

of it consisting of the even natural num- In the years 1879, 1880, 1882, 1883
bers. (two essays) and 1884, Cantor published a
The collection of sets having the small- remarkable six-part sequence of essays [4]
est infinite power is an extraordinarily rich which must be considered the foremost
and extensive system. The set of natural line of ancestry of his theory of point sets.
numbers has this power. It is easy to show Insofar as the theory of point sets specific-
that if M is a set having this power, then ally is concerned, it was preceded only by
any infinite subset of M has this same the introduction of a few less important
power. Cantor also presented the theorem concepts in his 1872 paper [1], which was
that if M', M", M"',... is a finite or concerned with trigonometric series. The
countably infinite sequence of sets each first introduction of limit points and deri-
having the power of the natural numbers, vatives of point sets was contained in his
then the set resulting from the union of 1872 publication [1]. In the essay
these sets has this same power [3, p. 243] . sequence [4], he also discussed such con-
Also included in the 1878 paper is Cantor's cepts as well-defined, well-ordered, similar
interesting and well-known proof that the (although not using this term), closed,
set of rational numbers has the smallest perfect, dense, connected, continuum, con-
infinite power [3, p. 250]. tent theory, and his theory of adherences
Cantor's correspondence with Dedekind and coherences.
shows that the idea of going over from In the first essay of the sequence [4],
the one-dimensional to the multidimen- Cantor submitted linear sets to a close
sional to get higher transfinite powers analysis by classifying them according to
had occupied Cantor as early as 1874 [12, certain principles, and in the third essay he
p. 237]. It was commonly assumed that extended the concepts developed for linear
points in two-space cannot be traced back sets to sets situated in continua of n
to one-space. Yet Cantor proceeded to dimensions. The general character of the
prove in his 1878 paper [3] the indepen- power concept is stressed in the third
dence of the power of the continuum from essay. He noted that the concept of power
its number of dimensions, a fact that may be considered an attribute of every
surprised even himself as is evident in well-defined set. Also in the third essay is
another of his letters to Dedekind prior to the first use of the word countable
publication of the proof [12, p. 237]. (denumerable, enumerable) to describe a
Cantor observed that the result can be set which can be put in a one-to-one
expanded to the case of a countable (later correspondence with the set of positive

56

This content downloaded from 104.239.165.217 on Mon, 20 Jun 2016 10:23:00 UTC
All use subject to http://about.jstor.org/terms
integers, although this concept had been tion and multiplication of transfinite
used earlier, as has been seen. Cantor numbers and mentioned some of their
mentioned in a letter to Philip E. B. basic properties. The development of
Jourdain that he first formed the idea of Cantor's new numbers and the distinction
countability in 1873 [I 1, p. 32]. between the ordinal concept and the power
Cantor's lengthy fifth essay [4] was concept is of significance for the theory of
also published separately, with extra notes, all (finite and infinite) arithmetic.
as Grundlagen einer allgemeinen Mannich- The Grundlagen gives a good account
faltigkeitslehre. In a note to the Grundla- of the slow and sure way in which the
gen, Cantor remarked that he meant by the transfinite numbers forced themselves on
term Mannichfaltigkeitslehre a doctrine the mind of Cantor and also shows to a
embracing very much which before he had considerable degree Cantor's philosophical
attempted to develop only in the special and mathematical traditions. Both here and
form of an arithmetical or geometrical in later works are discussions of opinions
theory of sets (Mengenlehre). At this point, on infinity held by mathematicians and
he understood a set to be generally any philosophers of all times. Such names as
totality of definite elements which can be Aristotle, Rene Descartes, Baruch (or Ben-
bound up into a whole by means of a law. edict) Spinoza, Thomas Hobbes, George
Cantor had an interesting discussion in Berkeley, John Locke, William Leibniz,
the Grundlagen of what he would later call Bernard Bolzano, and many others are
ordinal numbers. He pointed out that, found in his works. There is evidence of
while it would be contradictory to speak of deep erudition and painstaking search for
a greatest number of the set of positive new views on infinity to analyze. Many
integers, there is nothing objectionable in pages were devoted by Cantor to the
imagining a new number, A, which is to schoolmen and the fathers of the
express that the entire set of positive church [11, p. 55]. He gave in the
integers is given by its law in its natural Grundlagen an interesting discussion of the
order of succession. He went on to develop conditions under which the introduction
larger numbers of this type by what he into mathematics of a new concept, such as
called "generation and limitation princi- A, is to be regarded as justified. He said
ples." He observed that in the case of finite that mathematics is, in its development,
sets this conception of number coincides quite free; that it is subject only to the
with power, but that these two concepts self-evident condition that its concepts be
diverge in the case of infinite sets. He (1) free from contradiction in themselves
stressed the necessity of this extension of and (2) set in fixed relations, arranged by
the concept of number and pointed out definitions, to previously formed and
that the fact that these numbers do not tested concepts. In the introduction of new
have all the qualities of the finite numbers numbers, in particular, it is obligatory to
or have certain qualities which cannot go give only such definitions of them as will
with the finite numbers cannot be reason afford them such a definiteness and, under
enough to reject them. Each enlargement certain circumstances, such a relation to the
of a basic principle, he said, brings the loss older numbers as permits them to be
of certain characteristics; witness the distinguished from one another in given
example of the complex numbers as an cases. He said that a number, as soon as it
enlargement of the reals. He defined addi- satisfies all these conditions, can and must

57

This content downloaded from 104.239.165.217 on Mon, 20 Jun 2016 10:23:00 UTC
All use subject to http://about.jstor.org/terms
 be considered as existent and real in Cantor was of the opinion that closed
mathematics [11, pp. 67-68]. sets should be sufficient for general obser-
In the discussion of the continuum vation, and his beginnings of the con-
concept, Cantor briefly referred to discus- tinuum problem were essentially condi-
sions of the concept by Leucippus, tioned through their over-estimation of the
Democritus, Aristotle, Epicurus, Lucre- closed set. When Felix Bernstein's disserta-
tius, and Thomas Acquinas. He then set tion reduced the significance of these sets
out to give a mathematically exact defini- in a decisive manner, Cantor was very
tion of the continuum without drawing surprised and interrupted soon after certain
upon the concepts of time or space. power investigations with which he was
From his researches into ordinal busy at that time [12, p. 250]. After
numbers and powers, Cantor derived the conducting investigations into the power of
idea that there exists a well-ordering for perfect sets, Cantor felt that the power of
any set whatsoever, and this he stated with the continuum could be determined as the
a promise to return to the subject later. He second transfinite cardinal number by
also gave a proof for the following making use of some former theorems along
theorem: If M is any well-defined set of the with insights gained from his investigations.
second (infinite) power, M' is a subset of As is known from a letter to G. M.
M and M" is a subset of M', and M" has the Mittag-Leffler, Cantor considered as start-
same power as M, then M' has the same ing point of the proof a closed set of the
power as M and hence as M". Cantor power 81, whose construction, however,
remarked that this theorem has general he did not succeed in establishing [12, p.
validity and promised to return to it later, 251] .
but the proof was still lacking in the total
presentation [10]. Part of the mathematics Transfinite Ordinal and Cardinal Numbers

found in the Grundlagen was later treated The theory of transfinite cardinal and
far more completely and drawn up with far ordinal numbers underwent a remarkable
more attention to logical form. development by Cantor from 1883 to
Cantor's mathematically rich sixth 1895. His papers [6], [7], and [8], which
essay in the sequence [4] began with a speak especially to a philosophical circle of
general explanation of the interval nesting readers, give an account of that develop-
method. He observed that the interval ment from 1883 to 1888. A large part of
nesting method he had developed was, the discussion is concerned with philoso-
according to its basic ideas, very old; and phers' denials of the possibility of infinite
he mentioned some of the people who had numbers and with extracts from letters to
used it in more recent times: Joseph-Louis and from philosophers and theologians.
Lagrange, Adrien-Marie Legendre, P. G. Cantor contended that infinite numbers
Lejeune Dirichlet, Augustin-Louis Cauchy, must constitute quite a new kind of
Bernard Bolzano, and Karl W. T. Weier- number as opposed to the finite numbers,
strass. and that people err when they attribute to
Some of the proofs of assertions in the the numbers in question all the properties
fifth essay were given in the sixth essay. of finite numbers and proceed on this basis
Among the theorems concerning countable to show the impossibility of such numbers.
sets was the proof that a perfect set is not In 1883 Cantor had begun to lecture
of the first (infinite) power. on his view of cardinal numbers and types

58

This content downloaded from 104.239.165.217 on Mon, 20 Jun 2016 10:23:00 UTC
All use subject to http://about.jstor.org/terms
of order as general concepts which relate to In a letter of 1884 Cantor pointed out
sets and arise from these sets by abstracting that the cardinal number of a set M is the
from the nature of the elements. He said general concept under which fall all sets
that every set of distinct things can be equivalent to M. He said that one of the
regarded as a unitary thing in which the most important problems of the theory of
things first mentioned are constitutive ele- sets consists in determining the various
ments. By abstracting both from the nature powers of the sets in the whole of nature,
of the elements and from the order in insofar as can be known. He thought he
which they are given, one obtains the had solved this problem as to its principal
cardinal number, or power, of the set. The part in his Grundlagen by the development
cardinal number of the set is a general of the general concept of ordinal number
concept in which the elements, as so-called (another term was used there) [11, pp.
units, have so grown organically into one 75-76]. Both at the end of this letter and
another to make a unitary whole that no in a letter of 1886 Cantor discussed in
one of them ranks above the others. Two what sense w may be regarded as the limit
different sets have the same cardinal to which the variable finite whole number
number when and only when they are n tends. For this most interesting discus-
equivalent to one another. Cantor further sion, see [11, pp. 77-78]. Also in two
said that there is no contradiction when, as letters of 1866, Cantor went in detail into
often happens with infinite sets, two sets the distinction between the "potential"
of which one is a subset of the other have and "actual" infinite of which he had made
the same cardinal number. He regarded the a great point under other names in his
failure to recognize this fact as the princi- Grundlagen.
pal obstacle to the introduction of infinite By 1885, Cantor had developed to a
numbers. In dealing with ordered sets, if large extent his theory of cardinal numbers
the act of abstraction referred to is only and order types. In his paper [8] he laid
with respect to the nature of the elements, particular emphasis on the theory of order
so that the ordinal rank in which these types and entered into details which he had
elements stand to one another is kept in not published before as to the definition of
the general concept, the organic whole order type in general, of which ordinal
arising is what Cantor called order (ordinal) number is a particular case. Here he intro-
type, or in the special case of well-ordered duced the notation M to denote the car-
sets an ordinal number. Two ordered sets dinal number of a set M, thus indicating by
have the same order type if a one-to-one the double bar that a double act of
correspondence is possible between them abstraction is to be performed. The ordinal
such that if a and b are any two different number of M he denoted by M, thus
elements of one, and a' and b' are the two denoting that a single act of abstraction is
corresponding elements of the other, if a to be performed. It is characteristic of
precedes or follows b, then a' respectively Cantor's views that he distinguished very
precedes or follows b' (that is, if they are sharply between a set and its cardinal
similar). Cantor said that these are the number. The arithmetic of cardinal
roots from which develops with logical numbers and order types was treated much
necessity the organism of transfinite theory the same as in his memoir of 1895 [10 I],
of types and in particular of the transfinite which fortunately is available in transla-
ordinal numbers [11, pp. 74-75 ]. tion [11, pp. 85-136] . A part of the paper

59

This content downloaded from 104.239.165.217 on Mon, 20 Jun 2016 10:23:00 UTC
All use subject to http://about.jstor.org/terms
 dealt with finite sets and their cardinal tion [10] . In the terminology introduced
numbers. The notion of the n-ple ordered in [10], the memoir [9] contained a proof
set is also introduced in [8], on which that 2, when exponentiated by a transfinite
notion is based the idea of n-ple order cardinal number, gives rise to a cardinal
types; the definition of two n-ple ordered number which is greater than the cardinal
sets M and N being similar was also given. number first mentioned. In [9] can be seen
Cantor gave some interesting criticisms the origins of the concept of covering
on the number concept as advanced by H. (Belegung), which was defined in [10].
von Helmholtz and Leopold Kronecker. The introduction of this concept was pro-
Cantor had arrived at a very clear notion bably the most striking advance in the
that the essential part of the concept of principles of the theory of transfinite
number lay in the unitary concept. In numbers from 1885 to 1895.
essays of 1886 Helmholtz and Kronecker The great double treatise [10], pub-
started with the last and most unessential lished in 1895 and 1897, was the last work
feature in the treatment of ordinal of his own which Cantor published. It is
numbers: the words or other signs which free of the philosophical and critical
are used to represent these numbers [12, p. encumbrances of [8] and was destined for
253]. a mathematical public. Fraenkel pointed
The immediately preceding papers of particularly to [10] along with [4] as the
Cantor's which have been discussed dealt "two quite great and immortal works of
considerably with philosophical questions. Cantor." He mentioned that of -the "classi-
The conclusion of the treatises by Cantor cal" theorems of abstract sets, there was
to set theory are formed by two purely lacking in [10] only the equivalence
mathematical works. theorem. Also from [4] to [10] there was
a considerable movement from the observa-
Comparability of Powers tion of sets to that of numbers. There is
Cantor's paper [9] dealt with an im- notable progress to be noted in the direc-
portant question in the theory of trans- tion of clarification which makes [10]
finite numbers. Cantor stressed that by even today didactically very usable [12, p.
means of generalization of the diagonal 257]. Much of [10] was not actually new
procedure used to prove the noncount- but was a systematic, updated treatment of
ability of the reals, it was possible to form what Cantor had done before; [10 II] in
from a set another with greater power. This particular, dedicated to well-ordered sets,
proof of the existence of infinitely many gave in new systematic garb much that was
powers had already been accomplished in contained in the Grundlagen.
the Grundlagen in another (less straightfor- Among the highlights of [10] is the
ward) fashion. Cantor took as example for express remark by Cantor that compara-
the proof of the general theorem that the bility is neither self-understood nor prov-
set of all single-valued real functions de- able. He mentioned that if for two cardinal
fined on the closed interval 0 to 1 has a numbers a and b, either a = b, a < b, or
greater power than the continuum. In the a > b, then several theorems concerning
proof Cantor used comparability of equivalence would be easy to derive,
powers, apparently not being quite so among them the famous Bernstein equival-
conscious of the difficulties connected ence theorem (Cantor-Bernstein theorem,
with comparability as in his next publica- Schroeder-Bernstein theorem, Cantor-

60

This content downloaded from 104.239.165.217 on Mon, 20 Jun 2016 10:23:00 UTC
All use subject to http://about.jstor.org/terms
Bernstein-Schroeder theorem). This tions of set theory are the Zermelo-
theorem was proved by Felix Bernstein in Fraenkel and the von Neumann-Bernays-
Cantor's seminar in 1898 and also was Goedel versions. The various schools of
proved independently by Ernst Schroeder thought in mathematics also gave valuable
in 1896 without using comparability of suggestions as to the underlying causes of
cardinal numbers [13, p. 101] . Implicit use the paradoxes and what could be done to
of the axiom of choice in the proof that correct them. Cantor's initial stimulus of
every transfinite set contains a countably set theory was thus valuable not only in
infinite subset is also worthy of mention. A itself but also as a means for generating
number of mathematicians implicitly used further discussion and research into the
this axiom without being aware that such a whole area of the foundations of mathe-
principle was being used. The first explicit matics.
statement of the axiom was given by Ernst Cantor continued to be occupied with
Zermelo at the suggestion of Erhard set theory after his publications were
Schmidt in 1904 for the purpose of prov- finished. In his continuing correspondence
ing the famous and still controversial well- with his old friend Richard Dedekind he
ordering theorem. A number of other showed special concern about the con-
extremely important concepts and develop- tinuum problem. The continuum problem
ments are contained in [10] , but, since the had made a deep impression on Cantor and
works are available in English transla- even produced doubts in him whether set
tion [11, pp. 85-201, the only ones of his theory could be maintained as a scientific
works which are translated in their entirety build-up in the form in which he had
into English], they will not be belabored developed it. Kurt Goedel achieved some
here. progress on the continuum problem about
1940. He showed that the (generalized)
Paradoxes
continuum hypothesis is relatively consis-
Hardly had Cantor's work been com- tent with the other axioms of set theory.
pleted before paradoxes began to appear in This result does not, of course, constitute a
his theory. In 1897, Cesare Burali-Forti proof of the continuum hypothesis, only a
discovered a paradox in Cantor's theory of proof that it cannot be disproved. Goedel
ordinal numbers (purportedly discovered also proved that the axiom of choice is
earlier by Cantor), and shortly thereafter relatively consistent with the other axioms
other paradoxes of an even more basic of set theory. In 1963, Paul Cohen made a
nature began to appear in set theory. The major breakthrough by constructing an
Cantor paradox was discovered about example in which the axiom of choice does
1899, the Russell paradox about 1902, and not hold but the other axioms of set
the dam was broken. Considerable discus- theory do. Thus, to Goedel's proof that the
sion was generated, and important contri- axiom of choice cannot be disproved is
butions to the foundations of mathematics added the result that it cannot be proved.
were made because of the stimulus Cohen also added to Goedel's discovery
furnished by the paradoxes. The axiomati- that the continuum hypothesis is not dis-
zations in the early twentieth centry (first provable the fact that it is not provable.
by Ernst Zermelo in 1908) were aimed Thus two of the vexing questions of the
primarily at eliminating the paradoxes foundations have been solved within the
from set theory. The notable axiomatiza- realm of axiomatic set theory.

61

This content downloaded from 104.239.165.217 on Mon, 20 Jun 2016 10:23:00 UTC
All use subject to http://about.jstor.org/terms
 A listing of some of the important 6. , Ueber die verschiedenen Stand-
early advances in set theory since Cantor's punkte in Bezug auf das aktuale
last publication in 1897, leaving aside Unendliche, Zeitschrift fuer Philoso-
applications of the theory of transfinite phie und philosophische Kritick, 88
numbers to geometry and the theory of N.F.(1886) 224-233.
functions, is made by Jourdain in [11, pp.
7. , Ueber die verschiedenen An-
204-208]. The numerous developments
sichten in Bezug auf die aktualunend-
following Cantor's work and related to it
lichen Zahlen, Bihang till Kongliga
serve to show what a stimulus his research
Svenska Vetenskaps-Akademiens
was for further research. The continuing
Handlingar, 11 Nr. 19(1887) 1-10.
influence of his work is its own best
monument to one of the giants of mathe- 8. , Mitteilungen zur Lehre vom
matics, Georg Cantor. Transfiniten, I. Zeitschrift fuer Philos-
ophie und philosophische Kritik, 91

References N.F. (1887) 81-125 und 252-270; II.

Ibid, 92(1888) 240-265.
1. Georg Cantor, Ueber die Ausdehnung
eines Satzes aus der Theorie der trigo- 9. , Ueber eine elementare Frage der
nometrischen Reihen, Mathematische Mannigfaltigkeitslehre, Jahresbericht
Annalen, 5(1872) 123-132. der Deutschen Mathematiker Ver-
einigung, 1(1892) 75-78.
2. , Ueber eine Eigenschaft des
Inbegriffes aller reellen algebraischen 10. , Beitraege zur Begruendung der
Zahlen, Journal fuer die reine und transfiniten Mengenlehre, I. Mathe-
angewandte Mathematik, 77(1874) matische Annalen, 46(1895) 481-512;
258-262. II. Ibid, 49(1897) 207-246.

3. , Ein Beitrag zur Mannigfal- 11. , Contributions to the Founding

tigkeitslehre, Ibid., 84(1878) 242-258. of the Theory of Transfinite Numbers,
Paper [10] translated and provided
4. , Ueber unendliche, lineare
with an introduction (pp. 1-82) and
Punktmannichfaltigkeiten, I. Mathe-
notes (pp. 202-208), by Philip E. B.
matische Annalen, 15(1879) 1-7; II.
Jourdain. New York: Dover Publi-
Ibid., 17(1880) 355-358; III. Ibid.,
cations, Inc., n.d.
20(1882) 113-121; IV. Ibid., 21(1883)
51-58; V. Ibid., pp. 545-591; VI. Ibid., 12. A. A. Fraenkel, Georg Cantor, Jahres-
23(1884) 453488. bericht der Deutschen Mathematiker
Vereinigung 39(1930) 189-266
5. , Grundlagen einer allgemeinen
(Contains an apparently complete bib-
Mannichfaltigkeitslehre: Ein mathema-
liography of Cantor's works).
tisch-philosophischer Versuch in der
Lehre des Unendlichen (A separately 13. R. L. WilderIntroduction to the Foun-
appearing edition of 16V with fore- dations of Mathematics, New York:
word), Leipzig, 1883. John Wiley and Sons, Inc., 1952.

62

This content downloaded from 104.239.165.217 on Mon, 20 Jun 2016 10:23:00 UTC
All use subject to http://about.jstor.org/terms