MICHAEL OTTE

COMPLEMENTARITY, SETS AND NUMBERS

ABSTRACT. Niels Bohr’s term ‘complementarity’ has been used by several authors to
capture the essential aspects of the cognitive and epistemological development of scientific
and mathematical concepts. In this paper we will conceive of complementarity in terms
of the dual notions of extension and intension of mathematical terms. A complementarist
approach is induced by the impossibility to define mathematical reality independently from
cognitive activity itself. R. Thom, in his lecture to the Exeter International Congress on
Mathematics Education in 1972, stated “the real problem which confronts mathematics
teaching is not that of rigor, but the problem of the development of ‘meaning’, of the
‘existence’ of mathematical objects”. Student’s insistence on absolute ‘meaning questions’,
however, becomes highly counter-productive in some cases and leads to the drying up
of all creativity. Mathematics is, first of all, an activity, which, since Cantor and Hilbert,
has increasingly liberated itself from metaphysical and ontological agendas. Perhaps more
than any other practice, mathematical practice requires a complementarist approach, if its
dynamics and meaning are to be properly understood. The paper has four parts. In the first
two parts we present some illustrations of the cognitive implications of complementarity.
In the third part, drawing on Boutroux’ profound analysis, we try to provide an histor-
ical explanation of complementarity in mathematics. In the final part we show how this
phenomenon interferes with the endeavor to explain the notion of number.

KEY WORDS: attributive and referential use of terms, complementarity, history and epi-
stemology of mathematics, hypostatic abstraction, set-theoretical explanation of number

1. C OMPLEMENTARITY IN EPISTEMOLOGY AND COGNITION

Niels Bohr’s term ‘complementarity’ has been used by several authors

to capture the essential aspects of the cognitive and epistemological de-
velopment of scientific and mathematical concepts (Otte and Steinbring,
1977; Kuyk, 1977; Otte, Keitel and Seeger, 1980; Otte, 1984, 1990, 1994;
Douady, 1991; Sfard, 1991; Jahnke, 1992).
The principle of complementarity was first formulated by Niels Bohr
around 1930. We need not here be concerned with all the subtleties of
the theory of atomic processes that led to Bohr’s principle. It suffices to
note that Bohr’s notion of complementarity has been motivated by the fact
that when observing atomic phenomena one has to conclude that an inde-
pendent reality in the ordinary physical sense can neither be ascribed to the
phenomena nor to the agencies of observation. Bohr believed in the general

Educational Studies in Mathematics 53: 203–228, 2003.

© 2003 Kluwer Academic Publishers. Printed in the Netherlands.
204 MICHAEL OTTE

epistemological and metaphysical significance of his principle. Mathemat-

ical practice, which has increasingly liberated itself from metaphysical and
ontological agendas since Cantor and Hilbert, requires a complementarist
approach – perhaps more than any other field of knowledge – in order to
be understood properly.
On the occasion of a lecture given by H. Wiener in 1891, Hilbert made
a remark “which contains the axiomatic standpoint in a nutshell: It must
be possible to replace in all geometric statements the words point, line,
plane by table, chair, mug” (Reid, 1970, p. 264). Hilbert’s remark is usually
interpreted as expressing the tendency towards a de-ontologization of mod-
ern axiomatized mathematics. This is not so. To the contrary, mathematics
distinguishes itself from logic exactly by the fact that it has objects. These
objects are, however, nothing but hypostatizations or substantializations
of mathematical activities and thoughts themselves. These processes of
substantialization, traditionally codified in set theoretical terms, have to
be constrained in order to avoid paradoxes. Whence the requirement of
descriptions by means of consistent systems of axioms. After the set the-
oretical paradoxes appeared, Zermelo had provided set theory itself with
an axiomatic foundation.
Modern axiomatized theories became, on the one hand, intensional the-
ories in the sense that the axioms as a set of postulates not only determine
the intensions of the theoretical terms, but also constitute the extensions
or referents. In Euclidean geometry, the objects about which the theory
speaks seem to be given by intuition, and independently of the theory. In
Hilbert’s geometry, the situation is quite different, as the above quotation
shows. To answer questions such as, what is a point? or, what is a number?
one provides the respective axiomatic descriptions of the relations or laws,
which govern these entities. On the other hand, as implied by Gödel’s in-
completeness theorem, all axiomatic descriptions must necessarily remain
incomplete. And Skolem demonstrated, in fact, the uneliminable existence
of nonstandard models of arithmetic, as well as of any other non-trivial
theory, already in 1933. Any formal theory has various intended applica-
tions or non-isomorphic models, and what the axioms describe are classes
of objects rather than particular objects themselves. In this respect math-
ematical axioms resemble natural laws. And like the latter, they have to
be supplemented by an indication of the domain of objects to which they
apply.
We shall conclude that mathematical terms, the senses or intensions of
which are given by the systems of axioms and law-like statements, can be
(and are) used ‘attributively’ as well as ‘referentially’. That is, the terms
occurring in the axioms of a theory can be regarded, on the one hand, as
 COMPLEMENTARITY, SETS AND NUMBERS 205

giving descriptions of their referents, to be applied to those and only those

entities with reference to which they are true. On the other hand, the terms
contained in the axioms or in mathematical discourse in general can be
used ‘referentially’, too. In this case, we do not regard the expressions of
the theory as referring to those objects, which satisfy the given denotation,
but as saying something (may be falsely) about objects fixed independently
of the given description. In group theory, for example, one uses axiomatic
presentations of groups together with group representations (in terms of
linear transformations or in terms of permutations or whatever). And in
elementary algebra and arithmetic already some unknown number may
be indicated by the famous Cartesian ‘x’ and only afterwards, after some
more or less prolonged calculations, does one gain a characterization or
description of that unknown number.
Anyway, it becomes obvious that the relations between the intensions
and extensions of mathematical terms are more complicated than the clas-
sical law of inverse range suggests. Intensions and extensions become
relatively independent from each other and are circularly connected or
complementary to each other. In this paper, we will conceive of com-
plementarity in terms of the dual notions of extension and intension of
mathematical terms. This complementarity becomes visible, and becomes
distinguishable from mere duality, only from a genetic perspective, which
concentrates on the evolutionary character of our (mathematical) know-
ledge. Only from this perspective the relation between subject and object,
rather than the object as such, comes into focus. The notion of comple-
mentarity is thus relevant, in particular, to any study of the epistemological
foundations of mathematics education.
Understanding mathematics from a genetic perspective implies abandon-
ing notions such as the unknowable ‘things in themselves’, that is math-
ematical objects, which are not represented, or ideas like ‘mathematics
as such’ or truth as separated from our possibilities of verification and
proof. Our view of mathematics, therefore, has a pragmatic flavor, and its
tradition goes back to Peirce, Kant, and finally Berkeley. Berkeley agreed
with Locke’s assumption that our knowledge is founded on ideas perceived
by sensation and, on this basis, he undertook to examine everything that
could exist. He concluded that anything that exists is either the perceived
or the perceiving, i.e. either our ideas or ourselves. How about matter?
According to Locke, matter was the unperceived cause of our ideas. How
then, asked Berkeley, there being no means to verify its existence, could
matter be said to exist?
Berkeley’s principle could also be expressed by saying that, in order
for something to exist for us, it must be represented. This was how Kant
206 MICHAEL OTTE

and, later, Peirce understood Berkeley, and this brings us to the idea of
complementarity, which arises because signs are at the same time used
referentially and attributively. Knowledge is an activity, rather than a mir-
ror image of some existing world, and what underlies the frequent talk of
existence in mathematics is the phenomenon of mathematical objectivity,
rather than objects in the concrete empirical sense. This ‘pragmatic’ view
of mathematics could be reformulated as follows: A mathematical concept,
such as the concept of number or function, does not exist independently of
the totality of its possible representations, but must not be confused with
any such representation either.
Now, a representational ‘equation’ A = B is commonly interpreted as
saying that A and B are different intensions of the same extension, or
different designations of the same object. Both terms A and B have the
same reference, while the sense or the mode of presentation is different.
In Frege’s famous essay on Sinn und Bedeutung, the author quotes some
examples from elementary geometry. Frege writes:
Let a, b, c be the lines connecting the vertices of a triangle with the midpoints
of the opposite sides. The point of intersection of a and b is then the same as the
point of intersection of b and c. So we have different designations for the same
point, and these names (‘point of intersection of a and b’; ‘point of intersection
of b and c’) likewise indicate the mode of presentation, and hence the statement
contains actual knowledge. (Frege, 1969, p. 40; my translation)
Every mathematical theorem can be established as an equality A = B. In
Frege’s example, A and B are names of descriptions of a certain point. The
abstraction underlying this procedure is very important for mathematics,
as Peirce was one of the first to observe.
One extremely important grade of thinking about thought, which my logical ana-
lyses have shown to be one of the chief, if not the chief, explanation of the power
of mathematical reasoning, is a stock topic of ridicule among the wits. This op-
eration is performed when something that one has thought about any subject is
itself made a subject of thought. (Peirce, NEM IV – 49)
In this way even the means and conditions of thought may become objects
of thought. A predicative or attributive use of some concept is transformed
into a referential use in order to incorporate the entity thus synthesized
into new relational structures. This implies that in mathematics the re-
lationship between the particular and the general, between objects and
concepts or relations, for instance, is of the utmost importance, rather than
the search for absolute objective foundations. While in empirical science
there is a quite natural distinction between facts and laws, or things and
relations, relations seem to be all-pervasive in mathematics. The distinc-
tion between objects and relations therefore becomes extremely relative.
Within mathematics there is no absolutely fundamental ontological level.
 COMPLEMENTARITY, SETS AND NUMBERS 207

Still, mathematics is neither an analytical science from concepts nor purely

descriptive knowledge based on the mere abstractive observation of pre-
given objects. Mathematics is neither completely intensional nor merely
extensional knowledge.
Now Frege claims that mathematicians define neither concepts, nor
their contents, but rather their extensions.

For the mathematician, it is no more correct and no more incorrect to define a

conic section as the circumference of the intersection of a plane and the surface
of a right circular cone than as a plane curve whose equation with respect to
rectangular co-ordinates is of degree 2. Which of these two definitions he chooses,
or whether he chooses another again, is guided solely by grounds of convenience,
although these expressions neither have the same sense nor evoke the same ideas.
(Frege, quoted after Dummett, 1991, p. 32)

Some remarks should be added at this point. Firstly, with respect to the
growth of knowledge, it seems very relevant indeed which definition is
chosen, which perspective is taken, or how a problem situation is represen-
ted. How a mathematician defines something, a conic section, for instance,
to take Frege’s example, is certainly important. Two concepts A and B are
not the same, contrary to extreme Platonism, even if contingently or neces-
sarily all A’s are B’s and vice versa, say, all rectangles and only these are
blue, or all equilateral triangles have three equal angles, because different
concepts help to establish different kinds of relationships and thus influ-
ence development in different ways. Two concepts could be extensionally
equivalent and yet could be different, and might function differently, within
a certain cognitive context and with respect to the growth of knowledge.
Then A = B might perhaps be more conveniently interpreted sometimes as
a transformation or a relation of reference.
Secondly, this extension, as in the examples of mathematical entities
or in case of theoretical terms, like energy – of which heat and motion
are different representations, for instance, – or the electro-magnetic field,
or the general triangle, is not necessarily given as such, like an empirical
object, but is rather a universal object or an invariant relation. This may
once more lead to questioning Frege’s interpretation of A = B, under-
standing it rather in terms of relations or functions. Let us consider, as
an elementary example, the equation 2 + 2 = 4. According to Frege ‘2 +
2’ and ‘4’ have the same meaning but distinct sense. This interpretation
presupposes the existence of numbers as objects in a quasi Platonic sense.
Such views met with strong objections by constructively minded mathem-
aticians, for whom mathematical existence claims make sense only relative
to a language or an axiomatic system.
208 MICHAEL OTTE

Lebesgue, for example, favors a treatment of arithmetic completely

within the boundaries of the decimal system of numeration. What could
be the reasons for declining such an approach – he asks –?
First of all our metaphysical habits. Is it not a blasphemy to call a number a
symbol since numbers once constituted the very essence of things? Here we have
a fear manifesting itself in the most varied forms. For instance, let us say we may
certainly use interchangeably the English word chair or the French word chaise
because they both refer to the same object, but what is the analogue of the object
chair in the use of the symbols 101 in the binary system and 5 in the decimal
system? Since there is no chair hidden under 5, we can avoid the difficulty by a
verbal pirouette and speak of the metaphysical entity 5, which will replace the
physical reality chair. This amounts to refusing to answer the question.
(Lebesgue, 1966, p. 16)
Lebesgue is able to distinguish between true and false propositions only
because he considers arithmetic as an applied discipline. Mathematics is,
according to him, nothing but an instrument for other sciences. And arith-
metic is not a theory of its own applicability, or, as Lebesgue said: “L’arith-
métique s’applique quand elle s’applique” (Arithmetic applies when it
applies; Lebesgue, 1974, p. 31). Hilbert, like Lebesgue, considered pure
mathematics to be incomplete in so far as its growth depends on applic-
ations of various kinds. This makes the intensional or instrumental view
popular in school.
Pupils, as a rule, however, have difficulties with equations because they
have interpreted and learned the equals sign in exactly this functional sense
of ‘yields’. This functional ‘input-output’ interpretation represents a direct
understanding of the equation. The concept of equation has not yet been
transformed into an object of mathematical reflection.
The functional view has a strong affinity to certain standardized situations of
application [which could be characterized as] producing a new object out of given
objects by executing given rules; or, more generally, the correct transformation of
a given initial state into a desired final state.
(Winter, 1982, p. 191, my translation)
This functional or purely instrumental understanding of equations is insuf-
ficient, however, already in problems such as 8 + x = 13. Even elementary
tasks require a different interpretation of an equation, an interpretation
that treats the equation as an independent concept, a metaphor, as it were.
Students, for instance, understand very well that the same can be done
to the two sides of an equation, but in general they fail to understand
that adding or subtracting an equation A = B is just as legitimate as it
is in the case of the equation A = A. These pupils know how to think
in terms of functions or procedures but have difficulties with straightfor-
ward predication. Mathematical cognition, however, is characterized by a
 COMPLEMENTARITY, SETS AND NUMBERS 209

complementarity between predicative or extensional and instrumental or

intensional interpretations.
This brings us to a third objection to the views of Frege. Frege’s uni-
versal conception of logic leads to an inadequate and formalistic view of
logic and mathematics, because it implies the belief in the ineffability
of semantics. Frege therefore cannot use existence as a predicate as it is
common in mathematics. The mathematician uses the word ‘exists’ as a
predicate, but uses it relatively to an intended universe of discourse. Does
the real number x, which satisfies the equation x2 = –1, or written differ-
ently, x = -1 /x, exist? If so, it must be a root of unity in the real number field
and must thus be equal to 1 or to –1 and this yields 1 = –1, a contradiction!
But the mathematician enlarges his universe and finds a new system of
numbers, complex numbers.

2. ATTRIBUTIVE AND REFERENTIAL USE OF SYMBOLS AND

CONCEPTS

The notions of intension and extension have already been defined in 1662
by Arnauld and Nicole in their famous Port Royal Logic. They say:
I shall call the contents of an idea the attributes which the idea encompasses and
which cannot be separated from it without destroying the idea, just as the content
of the idea of the triangle encompasses size, shape, three angles, and equality
of these angles with two right angles, etc. (Part I, Chapter VI, our translation;
although the term intension does not occur, the term ‘content’ can be regarded as
synonymous with ‘intension’ here)

This expresses an analytic conception of mathematics, according to which,

for example, the theorem about the sum of angles in a triangle is some-
thing immediately belonging to the idea of a triangle, while Kant, as is
well known, claimed this theorem to be synthetic or constructive (Kant,
Critique, B 742). And mathematical construction does not proceed from
concepts alone, but rather has to rely on particular instantiations of these,
as well, that is, on particular objects. Even in a deductive proof one might
argue, for example, that line A is parallel to line B, or intersects with it at
point C, etc. Kant took great pains to draw our attention to these aspects of
proving. As Hintikka comments:
Kant’s characterization of mathematics as based on the use of constructions has
to be taken to mean merely that, in mathematics, one is all the time introducing
particular representatives of general concepts and carrying out arguments in terms
of such particular representatives, arguments which cannot be carried out by the
sole means of general concepts. (Hintikka, 1992, p. 24)
210 MICHAEL OTTE

Peirce called such reasoning theorematic, in contrast to corollarial reason-

ing, which relies only on that which is enunciated in the premises. If, by
contrast, a proof is possible only by reference to other geometrical objects
not mentioned in the original statement, if we need, for example, auxiliary
constructions in order to be able to carry out an argument, such proof
is theorematic. Peirce claims that theorematic reasoning is necessary to
gain new insight and that it depends on a sort of iconic representation,
which he calls a diagram. An icon or ‘diagrammatic sign’ (Peirce, CP
1.369) is always required to introduce something new into thought. And
since the proof of a theorem is in all the nontrivial cases connected to a
generalization,

. . .mathematical reasoning consists in constructing a diagram according to a gen-

eral precept, in observing certain relations between parts of that diagram not
explicitly required by the precept, showing that these relations will hold for all
such diagrams, and in formulating this conclusion in general terms. All valid
necessary reasoning is in fact thus diagrammatic. (Peirce, CP 1.54)

Concerning extension, the same text of the Logic of Port Royal says that
an extension of an idea are “the subjects to which this idea is due, which
are also called those subordinate to a general word”. Number, in this sense,
is simply all that for which Peano’s axioms hold, that is, the extension of a
concept would mean all the possible applications of the axiomatized the-
ory. As we cannot give this totality of possible applications, mathematics
cannot be purely extensional either.
As it seems that theories must either be extensional, or intensional these
conclusions seem puzzling or even paradoxical. The problem is, in fact,
well known to everybody working in education, or in cognitive theory.
Just think of the notions of “advance organizer” (Ausubel 1960), or of
the “fundamental ideas” and the role Bruner (1961) attributed to them
with respect to learning and teaching. On the one hand, these ideas are
what the development of an entire theory is devoted to unraveling and to
explicating. In mathematics, to understand a concept means to develop a
theory, and vice versa, the theory as a whole is logically founded, if it can
be understood as an – original – idea, which has been developed, made
concrete, and unfolded. The most far-reaching unfolding of the theory sub-
stantiates the original concept, although it is founded on the latter. Hence,
these ideas are the goal of theory development. These ideas are, however,
at the same time its beginning and its base. This means that they have to
be intuitively impressive, must motivate activity and orient representation.
But then again, by intuition, something is only given to us rather than
apprehended. As long as an object is not in some way incorporated into
a conceptual system or theory, it is not really known. On the other hand
 COMPLEMENTARITY, SETS AND NUMBERS 211

this conceptual system must be established and enlarged and for this some
intuition or new idea is necessary in each case.
Fundamental concepts or basic ideas are self-referential, that is they
themselves have to organize the process of their own deployment and
articulation (Otte and Steinbring, 1977). If it were impossible that the
theoretical concept supplies the basis of its own development, the only
standard left would be to try and see whether the new ideas and the new
concepts are similar to the old, or not. That is, in principle, nothing new
could result, as the already and previously given remains the backdrop for
everything. To explain something new would mean to try to reduce it to
the already known (Otte, Keitel and Seeger, 1980, p. 73).
If, conversely, the new became exclusively the basis of the world, and
of thinking about this world, in an absolute sense, there is nothing but
incommensurability and discontinuity, a total and unmotivated change of
perspective on reality. This would transform the development of know-
ledge into a random process. If we wish to avoid this, intensions and
extensions of our concepts must be seen as complementary to each other
in that, on the one side, they function in relative independence from each
other and remain, on the other side, circularly connected. This descrip-
tion resembles the well-known discussion about the hermeneutic circle of
text-interpretation.
Let us come back to the notion of complementarity, by pointing out
once more that we use our symbols and concepts in a twofold sense, both
attributively and referentially. Bertrand Russell illustrates the point by means
of the distinction he draws between names and descriptions. We have, he
writes,
. . . two things to compare: a name, which is a simple symbol, directly designating
an individual which is its meaning (or referent), and having this meaning in its
own right independently of the meanings of all other words; a description, which
consists of several words, whose meanings are already fixed, and from which
results whatever is to be taken as the ‘meaning’ of the description.
(Russell, 1998, p. 174)

On account of this distinction between naming and describing Russell is

led to criticize and refine Frege’s interpretation of A = B or of A = A.
Frege treated the difference between these two forms of an equation by
his own distinction between sense and meaning, concluding that singular
descriptions function like designations, as one usually understands them
referentially. Russell considers this to be an error, for we cannot gain
knowledge by just giving things new names.
Thus so long as names are used as names, ‘Scott is Sir Walter” is the same trivial
proposition as “Scott is Scott”. (Russell, 1998, p. 175)
212 MICHAEL OTTE

And,
. . . a proposition containing a description is not identical with what that pro-
position becomes when a name is substituted, even if the name names the same
object as the description describes. “Scott is the author of Waverley” is obviously
a different proposition from “Scott is Scott”. (Russell, 1998, p. 174)
And by the very same token,
If ‘x’ is a name, x = x is not the same proposition as “the author of Waverley is
the author of Waverley.” . . . In fact, propositions of the form “the so-and-so is the
so-and-so” are not always true: it is necessary that the so-and-so should exist. It is
false that the present king of France is the present king of France, or that a round
square is a round square. (Russell, 1998, p. 176).
‘Unicorn’ would then be an abbreviated description and ‘v-1’ as well. For
these descriptions the affirmation ‘x exists’ makes sense, although it may
be false, whereas, according to Russell or Frege, ‘y exists’ is meaningless
if ‘y’ is a name, because ‘exists’ is not a predicate.
But the essential point is that both, indices (names) as well as icons
(predicates or descriptions) are essential although we may never be able
to separate them completely, as we always use our linguistic terms both
referentially and attributively. To illustrate the latter point let us discuss
the following example. Suppose an English tourist visiting Amazonia sees
a biggish animal near the shore of a lake and asks what kind of animal
this is. He is told that what is seen is a Capivara. As the tourist cannot
speak Brazilian Portuguese this is only an indexical or referential designa-
tion, which leaves him without any representation for the moment. If he is
offered, to relieve his frown, an anglicization in the form of ‘water hog’, his
face lights up and he says ‘aha’, actually believing to have understood what
it is, the fact being that he is able to link something meaningful with the
words of ‘water’ and ‘hog’. This is thus a case of some kind of descriptive
designation, which has the disadvantage, however, of creating completely
false notions. For the Capivara is no swine at all, but a grass-eating rodent.
The Amazonian is in the opposite situation, as for him the Indian name of
Capivara has the meaning of ‘grass-eater’, while the designation ‘water
hog’ tells him absolutely nothing.
Now such a referential use sometimes serves as the starting point of
further observations if a motive or curiosity results. After some time, the
tourist may observe some characteristics and habits of the Capivara, and
then will be able to say, “Capivaras are good swimmers and divers”, or
“the Capivara lives in family groups”, etc. Gradually, the use of the term
changes and it is transformed into a description. And indeed theories in
statu nascendi are mainly used ‘referentially’ by their exponents as well
as by their opponents, while having reached their zenith, they are used
 COMPLEMENTARITY, SETS AND NUMBERS 213

‘attributively’, until a new theory emerges and ascends to its zenith, when
the former theory is used ‘referentially’ again.
The key thing about a name or an index is that it has a direct connec-
tion with its object. In the case of the present example, this connection is
established by concrete ostentation. It indicates its objects without giving
any information about them. Therefore we are able to understand an index
as a sign only by means of some ‘collateral experience’, or contextual
acquaintance with what the sign denotes to make the interpretation work.
For instance, I point my finger to what I mean, but I can’t make my companion
know what I mean, if he can’t see it, or if seeing it, it does not, to his mind, separate
itself from the surrounding objects in the field of vision. It is useless to attempt
to discuss the genuineness and possession of a personality beneath the histrionic
presentation of Theodore Roosevelt with a person who recently has come from
Mars and never heard of Theodore before. (Peirce, CP 8.314)
The interdependence of attributive vs. referential uses of terms is much
more prominent with respect to mathematical concepts than in empirical
ones, because mathematical objects firstly do not exist independently of
any representation and secondly because their instrumental character is
much more pronounced. In pure number theory and arithmetic numbers
are the objects of study, in most number-theoretic propositions numbers
occur as nouns, whereas in applied mathematics number terms are used
predicatively or as adjectives. Numbers seem to have come into being as
adjectives. Frege also considered the adjectival strategy as the more natural
one (Frege, 1884, §46ff). There are three apples, for instance. “Anything
can be three”, a child would say in the New Math classroom. Frege himself
acknowledged the necessity of using number-words as nouns or consider-
ing numbers also as objects because the equality A = B of numbers must
be established by the correspondence of sets of equal cardinality (Frege,
1988 (1884), §62ff).
Sometimes it is more important to develop a feel for this complement-
arity. The same holds true with respect to linguistic behavior. Jakobson,
for example, classified linguistic behavior as referring to either code or
context, and has accordingly described the diverse forms of aphasia in
relation to disturbances of either of these references. For aphasics of the
first type context is the indispensable and decisive factor. Their behavior is
characterized by Jakobson as “a loss of meta-language” rendering them
incapable of uttering a predication that has not been stimulated by the
context at hand.
In the pathological cases under discussion an isolated word means actually noth-
ing but blab. As numerous tests have disclosed, for such patients two occurrences
of the same word in two different contexts are mere homonyms. [Such a person
may never utter] the word knife alone, but, according to its use and surroundings,
214 MICHAEL OTTE

alternately call the knife pencil-sharpener, apple-parer, bread-knife, . . . so that

the word knife was changed from a free form, capable of occurring alone, into a
bound form. . . . The patient was able to select the appropriate term bachelor when
it was supported by the context. . . but was incapable of utilizing the substitution
set bachelor = unmarried man as the topic of a sentence, because the ability for
autonomous selection and substitution had been affected. [These patients cannot]
be brought to understand the metaphoric use of words.
(Jakobson, 1956, p. 79–80)
In the second type of aphasia, described by Jakobson as contiguity dis-
order, the ability to construct contexts is impaired.
The syntactical rules of organizing words into higher units are lost.. . . [This type
of aphasia] tends to give rise to infantile one-sentence utterances and one-word
sentences. The patient confined to the substitution set deals with similarities,
and his approximate identifications are of a metaphoric nature, contrary to the
metonymic ones familiar to the opposite type of aphasics. (ibid., p. 85–86)
One could call one type of aphasia a loss of predication, or, using semiotic
terminology, a lack of iconicity and the other a loss of instrumental or
functional orientation.
The less a word depends grammatically on the context, the stronger is its tenacity
in the speech of aphasics with a contiguity disorder and the earlier it is dropped
by patients with a similarity disorder. (ibid., p. 86)
Mathematics, considered as semiotic activity, is to be characterized by
the related complementarity, or to say it differently, by the necessity of
establishing such a complementarity within the process and evolution of
cognitive activity.

3. A HISTORICAL EXPLANATION OF COMPLEMENTARITY IN

MATHEMATICS

Complementarity of mathematical concepts is a historical phenomenon, at

least in its full development. Therefore, for a better understanding of its
origin, we have to consult the history of mathematics. Among the works,
that try to outline the historical development of mathematics as a whole,
that by Boutroux (1920) is certainly one of the most interesting ones.
Boutroux, a son of a well-known philosopher of science and a nephew
of Henri Poincaré, was a mathematician himself, but turned to history of
science already in his early years.
Boutroux divides the history of mathematics since Antiquity into three
periods, which could roughly be indicated by the following names:
1. Plato/Euclid
2. Descartes/Leibniz
 COMPLEMENTARITY, SETS AND NUMBERS 215

3. Bolzano/Cantor.
Boutroux sees the essential revolution or break to occur between 2. and
3., while classifying 1. and 2. as both dedicated to a synthetical ideal of
mathematics and characterized by a pre-established harmony
. . . entre le but et la méthode de la science mathématique, entre les objets que
poursuit cette science et les procédés qui lui permettent d’atteindre ces objets.
(Boutroux, 1920, p. 193)
At the beginning of the 19th century, pure mathematics arises based on
proof analysis and the creation of ever more abstract concepts, and the
harmony between means and objects of mathematical activity begins to
break down. Pure mathematics is the child of an explosive growth of math-
ematical activity that occurred around 1800 and that, in its sources, may
be briefly characterized by stating that for the first time in the history
of mathematics a great number of connections between apparently very
different results and problems were detected. Descartes’ discovery of ana-
lytical geometry already initiated a process that really became dominant
since early nineteenth century.
[Descartes] inaugurated an uninterrupted series of reciprocal assimilations between
branches of mathematics till then heterogeneous, which seems to be the main
reason why pure mathematics became aware of itself.
(Beth and Piaget, 1966, p. 229)
Operative reasoning and a functionalist perspective introduced by this op-
erability were necessary to achieve the transition from classical mathem-
atics to the algebraic outlook beginning with Descartes. A complementary
aspect of this process, which was equally indispensable although it came
much later, may be called ‘geometrization’ or relational thinking. It be-
came dominant at the beginning of the nineteenth century, when algebra
was transformed from a ‘language’ into a science of structures. The oper-
ative conceptual schemata themselves, in a way, had to become an object
of thinking. Algebra is meta-algebra, it is ‘algebra on algebra’ as Sylvester
(1814–1874) once remarked with particular reference to the algebra of
determinants. It is not by accident, I think, that such a remark came up
within geometrical algebra.
Boutroux himself characterizes the break in the history of mathematics
that occurred at the beginning of the 19t h century by two events. First
the aforementioned harmony between means and objects of mathematical
activity disappears.
L’harmonie dont nous venons de parler a presque complètement disparu. Lor-
squ’on nous propose un problème, il nous est impossible de prévoir quels sont
les procédés – le plus souvent très indirects – qui permettront de le résoudre.
Inversement, quelque rompu qu’il soit au mécanisme de son art, le mathématicien
216 MICHAEL OTTE

ne voit pas toujours clairement quels sont les problèmes auxquels il doit appliquer
cet art. . . . En d’autres termes, un dualisme se manifeste au sein de Mathématiques
pures. (Boutroux, 1920, p. 194)
Secondly, Boutroux believes that mathematics now becomes an analytical
science based on conceptual thinking alone.
Le fait mathématique est indépendant du vêtement logique ou algébrique sous
lequel nous cherchons à le représenter: en effet, l’idée que nous en avons est plus
riche et plus pleine que toutes les définitions que nous pouvons en donner, que
toutes les formes ou combinaisons de signes ou de propositions par lequelles il
nous est possible de l’exprimer. (Boutroux, 1920, p. 203)
One might add, however, that the number of different ways of representing
that, which seems to be the same mathematical fact, also increases beyond
imagination. One might even claim, in fact, that only from experiencing
the advantage offered by a change of representational modalities arose the
ideal of an amodal, purely conceptual mathematical thinking. In this sense,
it is interesting to read Minkowski’s laudation, given at the occasion of
the centenary of Dirichlet’s birth, of the latter’s work “in which he freed
himself of wrong resources by drawing, instead, on the analytical rules
for determining a function’s minima, only on the original concept of min-
imum” and, it might be added, on that of function (cf. more extensively
Jahnke, 1990, pp. 308–332; on Dirichlet’s concept of function cf. Otte,
1994). Mathematics now becomes an activity with whatever means.
Having in mind the traditional historiography of mathematics, which
tends to explain mathematical development alongside the struggling with
the notion of infinity – isn’t mathematics the science of infinity, after all –
one might wonder why Boutroux puts Descartes and Leibniz into the same
boat, so to speak. Isn’t there an essential difference to be observed between
Descartes’ finite algebra and Leibniz’ algebra of the infinite, which did,
after all, bring about infinitesimal calculus?! Descartes and Leibniz’ sys-
tems
. . . se distinguent surtout par cette circonstance que l’un effectue sur des com-
binaisons infinies ce que l’autre fait sur le fini. Or, est-ce là, du point de vue
technique, une différence essentielle? (Boutroux 1920, p. 127)
Boutroux answers the question in the negative:
Le calcul de séries n’est pas . . . d’une autre nature que le calcul algébrique élé-
mentaire; seulement il ne nous conduit pas directement au but parce qu’il ne nous
donne ce que nous cherchons que d’une manière approchée. Or l’idée d’approxi-
mation . . . n’a rien à voir avec le dynamisme. A moins, toutefois, que l’on ne
veuille admettre que l’existence du fait mathématique obtenu par approximation
est le résultat de cette approximation même. Mais c’est là une vue que Leibniz
lui-même n’eût pas adoptée. (Boutroux, 1920, p. 128)
 COMPLEMENTARITY, SETS AND NUMBERS 217

The latter viewpoint came into being only when Cantor defined the real
numbers as convergent series of rational numbers. The idea is no more
to approximate an already given quantity, but to establish a new type of
number by means of a set of more elementary numbers. It is no longer
necessary to think of this construction in terms of measuring a pre-given
quantity, for example. The latter viewpoint would be the traditional one,
as a real number, a decimal fraction, for instance, would be nothing but an
approximate determination of the quantity. Constructivists, like Kronecker,
also tried to base the continuum of real numbers ultimately on the integers
and wanted to use decision methods, that is rules or laws, which effect-
ively determine every term in an infinite sequence to this end. But Cantor
argued that Kronecker’s supply of number symbols could never completely
describe the continuum (Cantor, 1980, p. 384).
The essential discontinuity or revolution thus came about, according to
Boutroux, together with the introduction of the actual infinite into mathem-
atical thinking. Kronecker’s numbers would all be ‘computable numbers’,
that is, would be numbers whose decimal development is given by some
kind of algorithm. All numbers known to us, in a way, as individuals, are
computable numbers. For instance e, which is neither rational nor algeb-
raic, is a computable number. We cannot, so to speak, know a specific
number that is not definitely describable, but no general number concept
thereby emerges. The contention that everything that can be explicitly
described is part of an enumerable set results in an enumeration of all
formally explicable definitions and of the cognitive objects constituted by
them. While being countable, the computable numbers, however, cannot be
effectively enumerated because applying Cantor’s diagonal method would
otherwise lead to a paradox, similar to “Richard’s paradox” (Minsky, 1967,
p. 161).
The essential point about Cantor’s notion of infinite set is exactly in this
transformation of a concept into an object operating with the extensions
of concepts. How bold this move was, becomes visible from Cantor’s
own philosophical considerations. On the one hand, his affirmation, that
“the essence of mathematics resides in its freedom” is frequently quoted
and well known. Less known, on the other hand, is the fact that Cantor
drew a clear distinction between pure and applied mathematics, consid-
ering applied mathematics as not free at all, but rather as dependent on
adequate metaphysics. Moreover, Cantor was a philosophical monist and
he believed, like Leibniz, in a sort of pre-established harmony. There are,
he writes,

. . . two senses in our talking about the reality or existence of the whole numbers
be they finite or infinite. On the one hand we may consider numbers to be real
218 MICHAEL OTTE

insofar as we have established them by means of definitions in our mind. . .. On

the other hand we may attribute reality to numbers, as they must be considered
the images or expressions of events and relations of an outer world that confronts
the intellect.. . . I have no doubt that these two kinds of reality will always come
together in the sense that a concept, which is real according to the first meaning of
the term, shall also always be real in innumerable ways according to the second
meaning, although it is one of the most difficult tasks of metaphysics to verify this
thesis. (Cantor, 1980, pp. 181ff; my translation)

The notion of set has a double nature. It represents a concept – collection-

as-one – as well as a set of objects – collection-as-many. The foundational
crisis resulted from trying to eliminate this complementarity and introdu-
cing an absolute difference between things on the one hand, and concepts
on the other. This difference, however, is obscured in modern set theoret-
ical mathematics. It becomes relative. The concept of set in Cantor’s sense
itself shows that. Mathematicians stress that an abstract set is a totality of
things, which have no further properties in themselves, and insofar it is it-
self a thing. This is demonstrated by the identification of sets on the basis of
the axioms of extensionality, resulting in the view of ‘collection as many’
(Potter, 1990). Infinite sets cannot, however, be given in extension. They
can only be presented intensionally by a conceptual description. We can
present the real numbers by an axiomatical description only. An infinite
set is therefore an abstraction.
An idea or an abstraction is, as Peirce puts it, an ens rationis whose
mode of existence depends on the existence of other fundamental things.
In this sense, a set is a hypostatic abstraction or an idea, which is based on
the existence of its own elements (cf. Peirce MS 142; partly reprinted in:
Peirce, NEM IV, XX-XXI). Elsewhere, in his ‘Lowell Lectures’ of 1903,
Peirce elaborates on the double nature of the notion of set, introducing two
specific concepts: ‘gath’ – to characterize the set as a thing – and ‘sam’
to represent it as an idea (Peirce, NEM III, 367 f.). He emphasizes that
while for every ‘gath’ there is a corresponding ‘sam’, this does not hold
vice versa, as no claims of existence are connected with our general ideas.
We may speak of the phoenix or of the unicorn without claiming a factual
existence of such creatures (ibid.: p. 371). Now Cantor’s views were, as
we have just seen, somewhat different with respect to the last point.
This complementarity of extensional and intensional understandings of
the notion of set triggers an infinite recursive process of abstraction. In
fact, Bochner (1966) considers the iteration of abstraction to have been the
distinctive feature of mathematics since the Scientific Revolution of the
seventeenth century.
In Greek mathematics, whatever its originality and reputation, symbolization . . .
did not advance beyond a first stage, namely, beyond the process of idealization,
 COMPLEMENTARITY, SETS AND NUMBERS 219

which is a process of abstraction from direct actuality.. . . However,. . . full-scale

symbolization is much more than mere idealization. It involves, in particular,
untrammeled escalation of abstraction, that is, abstraction from abstraction, ab-
straction from abstraction from abstraction, and so forth; and, all importantly, the
general abstract objects thus arising, if viewed as instances of symbols, must be
eligible for the exercise of certain productive manipulations and operations, if
they are mathematically meaningful. . . On the face of it, modern mathematics,
that is, mathematics of the 16th century and after, began to undertake abstractions
from possibility only in the 19th century; but effectively it did so from the outset.
(Bochner, 1966, p. 18, 57)
In the discussion of the so-called fundamental theorem of algebra, it is
said that Lagrange had tacitly and implicitly used the intermediate value
theorem for continuous functions to give a proof of this theorem. In ac-
tual fact, Lagrange only provided algorithms for obtaining approximate
solutions of algebraic equations. It was Cauchy who read the intermediate
value theorem for continuous functions into Lagrange’s argument, trying
to make this presupposition explicit. What Cauchy did was but a hypo-
statization or reification of algorithmic procedures, transforming them into
new abstract objects. He turned an algorithmic rule into a mathematical
function, and a converging process into the notion of limit.
In a vein similar to Bochner, Peirce writes:
One extremely important grade of thinking about thought, which my logical ana-
lyses have shown to be one of the chief, if not the chief, explanation of the power
of mathematical reasoning, is a stock topic of ridicule among the wits. This op-
eration is performed when something, that one has thought about any subject, is
itself made a subject of thought. (Peirce, NEM IV, 49)
In this way the means and conditions of thought become an object of
it. A predicative or attributive use of some concept is transformed into
a referential use in order to incorporate the entity thus synthesized into
new relational structures. The example of the introduction of the imaginary
numbers provides a case in point. At first, after having been introduced to
generalize certain algebraic operations, these ‘numbers’ seemed a paradig-
matic model of an artificial invention, whilst the subsequent history of the
theory of complex functions would tend to provide this invention with the
characteristics of something indubitably objective.
Abstraction from abstraction is certainly facilitated by the grammatical
structure of European languages, which makes it very simple for users of
those languages to speak of abstract entities as if they existed. Hypostatic
abstraction is thus achieved, for instance, by hypostazising a predicate or
a quality, thereby turning it into a subject capable of further predication.
We transform, for instance, the proposition, ‘honey is sweet’ into ‘honey
possesses sweetness’. This may sound trivial, although it facilitates such
thoughts as that the sweetness of honey is particularly cloying; that the
220 MICHAEL OTTE

sweetness of honey is something like the sweetness of a honeymoon, etc.

But language appears to be a flat game in this respect when compared to
mathematics or computer science. That abstractions of this kind are partic-
ularly congenial to mathematics is a fact exemplified particularly well by
Cantor’s set theory. Or think of the idea of a space of arbitrary dimension
invented by Grassmann in 1844: A point moves: it is by abstraction that
the geometer says that it ‘describes a line’. This line, though an abstraction,
itself moves; and this is regarded as generating a surface; and so on.
Again and again a construction or an algorithmic procedure is taken as
an object to be incorporated into another construction or procedure. But in
order to reify operational concepts it might be necessary to employ spatial
intuition, because mathematical intuition and activity do not operate on
singular objects but on ‘spaces’ of all kinds. We assume therefore that
mathematical meaning is to be conceived in terms of the complementarity
of extension and intension. Meaning has two objective components, one
of which refers to objects or indicates them; the other relating to linguistic
expressions or diagrammatic representations, which show the characterist-
ics of the object of activity (which in general is not the object named) and
which express how the characteristics hang together. The complementarity
is established by processes of generalization and verification. Castonguay,
in a similar vein, tried
. . . to lay bare two objective components of meaning, one of which refers to
objects, and which it is appropriate to name the extensional, or correspondence
component of meaning; the other relating to concepts or linguistic expressions,
and which it is suitable to call the intensional, or coherence, component, in that it
expresses, how a given concept or expression coheres, or hangs together, with its
fellows through relations of consequence. (Castonguay, 1972, p. 3)

4. H OW COMPLEMENTARITY INTERFERES WITH ATTEMPTS TO

EXPLAIN THE NOTION OF NUMBER

Sets have been, first of all, sets of numbers and later on attempts were made
to reduce numbers to sets. Moreover, the debate about the relation between
intensional and extensional views of mathematics has been particularly in-
tense with respect to the number concept. The intensional view, stressing
ordinality and axiomatic description, came first but suffered severe criti-
cism from others, who were interested primarily in mathematical applica-
tion. Let us have a closer look at Russell’s Introduction to Mathematical
Philosophy, published in 1919, which makes, despite all its deficiencies,
very exiting reading and which has sometimes and rightly been called “an
admirable exposition of the monumental work Principia Mathematica”.
 COMPLEMENTARITY, SETS AND NUMBERS 221

The book is even more than that, because it builds upon the results of all of
Russell’s fundamental work since about 1900. And it is at the same time
something else, as it provides a most original, lively and self-contained
introduction to the foundations of mathematics and epistemology.
The main object of Russell’s book is number and everything belonging
to number, to arithmetic, and to the logic of arithmetic. The foundations of
arithmetic had always remained the focus of Russell’s interest in logic and
mathematics. Russell accordingly begins his ‘Introduction’ with a chapter
on “the series of natural numbers”, in which the latter are introduced on
the basis of Peano’s axioms, the concept of ordinal number being the only
one to have a role here. There is no mention of cardinality in the beginning.
It might be suggested, Russell says at the end of this first chapter,
. . .that, instead of setting up ‘0’ and ‘number’ and ‘successor’ as terms of which
we know the meaning although we cannot define them, we might let them stand
for any three terms that verify Peano’s five axioms. They will no longer be terms,
which have a meaning, that is, definite though undefined: they will be variables.
(Russell, 1998, p. 9–10)
This is, indeed, the common understanding of the axiomatic approach. It
can also be expressed by saying that arithmetic is not about concretely
existing things, but rather about general relations or ideal objects. But this
point of view does not satisfy Russell. On two counts, Russell continues,
Peano’s approach
. . . fails to give an adequate basis for arithmetic. In the first place, it does not en-
able us to know whether there are any sets of terms verifying Peano’s axioms. . ..
In the second place. . . we want our numbers to be such as can be used for count-
ing common objects, and this requires that our numbers should have a definite
meaning, not merely that they should have certain formal properties.
(Russell, 1998, p. 10)

. . .if we start from Peano’s undefined ideas and initial propositions, arithmetic
and analysis are not concerned with definite logical objects called numbers, but
with the terms of any progression. We may call the terms of any progression 0,
1, 2, 3,. . ., in which case 0, 1, 2,. . . become ‘variables’. To make them constants,
we must choose some one definite progression; the natural one to choose is the
progression of finite cardinal numbers as defined by Frege. (Russell, 1954, p. 4)
Russell’s criticism was that the axiomatic characterization of number leads
to a situation where “every number-symbol becomes infinitely ambiguous”
(Russell, 1903, p. vi). Neither Peano nor Hilbert, according to Russell, are
really capable of defining what the number one is. Frege wrote in a similar
vein:
An arithmetic with no thought, as its content will also be without possibility of
application. Why can arithmetic equations be applied? Only because they express
thoughts. (Frege, 1980, section 91)
222 MICHAEL OTTE

According to Frege’s definition of number, which Russell, without know-

ing it beforehand, had rediscovered on his own after the Paris conference,
. . . primitive terms are replaced by logical structures, concerning which it is
necessary to prove that they satisfy Peano’s five primitive propositions. This pro-
cess is essential in connecting arithmetic with pure logic. We shall find that a
process similar in some respects, though very different in others, is required for
connecting physics with perception. (Russell, 1954, p. 4)
On this account, logic is interpreted in an utterly realistic way. A view also
expressed, for instance, by Russell’s statement that
. . . logic is concerned with the real world just as truly as zoology, though with its
more abstract and general features. To say that unicorns have an existence in her-
aldry, or in literature, or in imagination, is a most pitiful and paltry evasion. What
exists in heraldry is not an animal, made of flesh and blood, moving and breathing
of its own initiative. What exists is a picture, or a description in words. . .. There
is only one world, the ‘real’ world. . .. The sense of reality is vital for logic.
(Russell, 1998, pp. 169–170)
According to Russell, in order to provide the concept of number with some
extension, which is real, we have to understand “number as the number of
a quantity” and to provide an application for the concept thus defined by
demonstrating the existence of sets of arbitrary cardinality. Obviously, this
can only be done axiomatically. In doing so, however, the notion of ax-
iom must not be understood in the Peano-Hilbertian sense, instrumentally;
the term must rather be conceived of according to the classical Euclidean
tradition, that is, as an intuitively evident truth and as a precondition of
mathematics. This is why Russell introduces the “axiom of infinity”. We
have to ascertain or render plausible that there are in fact infinite collec-
tions or sets in the world to be able to found number (Russell, 1998, p. 77).
Infinite sets of arbitrary cardinality, Russell assumes, do exist and thus can
be referred to in our reasoning. Set theory must not be understood here as a
merely formal façon de parler, it should rather be seen as a representation
of reality.
In Russell’s account, arithmetical intuition is to be replaced by set-
theoretical intuition. This might appear strange, as the axiomatization of
arithmetic has been caused by the feeling that we may be unable to intuit
or fully grasp number and must therefore settle for the formal laws that
numbers satisfy. Now Russell apparently replaces number by the intuitive
concept of set as a foundation of these formal laws. Nearly half a century
later, mathematical education worldwide tried to repeat this move, with
little success. Mathematics is not a quasi-empirical science, which estab-
lishes its methods by means of the properties of its objects; rather, the
objects have to be constructed simultaneously with the rules and methods
of reasoning.
 COMPLEMENTARITY, SETS AND NUMBERS 223

Dedekind’s approach is illustrative here. Dedekind was not ready to

imagine a straightforward axiomatic definition of number, because, after
recognizing the essential characteristics of such a system, “the question
arises: does such a system exist at all in the realm of our ideas” (Dedekind
in his letter to Keferstein in 1890). Dedekind tried to provide an infinite
totality of things by ascribing to us, human subjects, the ability to infin-
itely repeat certain ideas or mental actions, such as adding another stroke.
Dedekind considered his thought experiment to be a logical existence proof
and he was not concerned, as Russell was, with the meaning of individual
number symbols. In contrast to Dedekind, Russell thought that one can
never attain infinite totalities by mere enumeration, and he considered it
an empirical fact “that the mind is not capable of endlessly repeating the
same act”. Russell says:
So the reader, if he has a robust sense of reality, will feel convinced that it is
impossible to manufacture an infinite collection out of a finite collection of indi-
viduals. (Russell, 1998, p. 135)

We cannot prove that infinite numbers or sets exist. And Russell himself
appeals to intuitive plausibility by means of the axiom of infinity.
Quite in contrast to Russell’s opinion about modern axiomatics one
might claim it is precisely in the interest of applicability that the manifold
interpretability of axiomatic systems makes for the strength and signific-
ance of the axiomatic method. In particular, it is difficult to understand
why the meaning of all concepts should be totally fixed prior to all applic-
ation. In the end, concepts would be no more than complete description of
individual entities. Mathematical axiomatics, in contrast, is often charac-
terized by Hilbert’s statement, according to which, in the axioms of plane
geometry, the terms ‘point’ and ‘line’ could just as well be substituted by
the terms ‘beer jug’ and ‘table’; ‘in other words: each and every theory
can always be applied to an infinite number of systems of basic elements’
(Hilbert in a letter to Frege dated Dec 29th , 1899).
When we have a special intended application in mind, there arises the
problem of how to construct the necessary correspondences. During the
entire 17th and 18th centuries, for instance, mathematics was understood as
the science of quantity. As the new phenomena of electricity and magnet-
ism were to be mathematized in the 19th century, it became necessary to
extend the concept of quantity to vector quantity (directed quantity) and to
generalize the quantity operations correspondingly (e.g. by introducing a
non-commutative vector product). As a rule, one cannot derive the axioms
from the ‘essence’ of the relata (from the meaning of the latter’s concepts).
All objective knowledge is, in fact, relational knowledge. One fixes some
assumptions or forms and searches to outline their consequences. Vec-
224 MICHAEL OTTE

tor calculus, for example, resulted from the mathematization of electrical

phenomena.
Russell, indeed, was not unaware of Hilbert’s axiomatic method. For
him, however, the interpretation of a deductive system had always been of
fundamental philosophical significance, as it would serve to clarify how
an empirical theory can be linked to perception, and how a mathematical
theory, like the theory of arithmetic, can be linked to logic. Despite all his
open-mindedness with regard to future developments of mathematics and
of the exact natural sciences, Russell repeatedly stressed his interest in the
absolutely invariant elements of mathematical concepts.
Russell deemed the axiomatic method to be incomplete, as unspecified
terms occur within the axioms. These uninterpreted terms must indeed be
specified in a way, which permits to establish a connection to the intended
application. An absolute or ultimate interpretation of mathematical con-
cepts, however, is generally neither possible nor desirable. The axiomatic
determination of mathematical concepts will always be incomplete, in so
far as one has to always take into account the possibility that a concept has
an empty extension (the axioms can be inconsistent), or that it is ambigu-
ous (a property desirable under the aspect of application). If one intends,
against that, to introduce all concepts by complete definitions, one must
necessarily make metaphysical and psychological assumptions about the
world, as it is in itself, which is a futile undertaking.
There is, in fact, no possibility of ultimately determining the meaning of
‘number’, not even within the framework of set theory. The widely known
and repeatedly reprinted collection on the philosophy of mathematics (cf.
Benacerraf and Putnam, 1983), where parts of Russell’s Introduction to
Mathematical Philosophy were reprinted as well, also contains an essay
by Paul Benacerraf entitled What Number Could Not Be. Benacerraf shows
that the number concept can be reduced to the concept of set in many very
different ways, without a chance to single out the correct set theoretical
interpretation from among all the possible ones uncovering the true identity
of the natural numbers in terms of sets. Benacerraf concludes that numbers
cannot be sets, or sets of sets, at all, as there exist very different accounts
of the meaning and reference of number words in terms of set theory.
And even Quine, who shared Russell’s dislike of the ‘disinterpretation of
mathematics’ and who pointed out that words like ‘two’ and ‘four’ do not
occur uninterpreted anywhere in our language, emphasizes that every set
theoretical interpretation of number words – Frege’s, von Neumann’s, or
Zermelo’s – is used “opportunistically to suit the job in hand, if the job
is one that calls for providing a version of number of all” (Quine, 1969,
p. 263).
 COMPLEMENTARITY, SETS AND NUMBERS 225

Even if it were possible to reduce the number concept unequivocally

to the notion of set, not much would be gained, as Russell discovered
with the set theoretical paradoxes. Thus Russell’s straightforward logical
realism brought about problems, which his theory of types was intended
to solve. What then is Russell’s theory of types about? In order to re-
pair certain paradoxes of logic and set theory, Russell introduced the rule
“Whatever involves all of a collection must not be one of the collection”
(Russell, 1971, p. 63). An all-concept referring to a totality cannot belong
to the totality. Now Dedekind, in his Was sind und was sollen die Zahlen?
(Dedekind, 1969, p. 14), founded the proof for the existence of infinite
sets on the antinomial set (of all things) “which can be an object of my
thought” (Russell’s set of sets that are not members of themselves certainly
is a possible object of thought). Cantor called Dedekind’s attention to the
inconsistency in the latter’s foundation of number as early as 1899. Later,
this led constructivism to abandoning the idea that the realm of number
could be conceived of as a completed totality.
Russell himself characterizes type classification as only a negative move,
as a principle of interdiction. In particular, certain limitations are to be
established with regard to the range of the quantified variables in pro-
positional functions. An expression like “all propositions are either true
or false”, for instance, makes no sense any more. On the one hand, type
theoretical distinctions are quite self-evident and common. The totality
of chairs is no chair, and the class of red things is not a red thing, but
rather a hypostatic abstraction, like redness, or a propositional function,
like “x is red”, or whatever. While concept and object, menu and meal,
map and territory are easily distinguished, this distinction, on the other
hand, becomes a relative one from the perspective of cognitive activity and
its dynamical development. We have pointed out indeed how essential the
iteration of abstraction is as a distinctive feature of the mathematics of
modernity.
As has been said, Russell’s type theory implies that sets cannot simply
be what common sense imagines them to be and what everybody has
learned who has been to school in the times of the ‘New Math’ reform
movement. Although Russell makes great efforts over extensive parts of
the book to reduce the concept of number to the concept of set, treating the
latter as a fundamental concept in doing so, the way he develops his own
argument shows that such conception must be revised. Hence, the last but
one chapter is about sets, and there Russell’s concern is to

. . . realize why classes cannot be regarded as part of the ultimate furniture of the
world. . .. We cannot take classes in the pure extensional way as simply heaps or
conglomerations. (Russell, 1998, pp. 182/183)
226 MICHAEL OTTE

If we conceived of sets extensionally as objects, Russell thinks, it would

be impossible for us to understand
. . . how there can be such a class as the null-class, which has no members at all
and cannot be regarded as a ‘heap’; we should also find it very hard to understand
how it comes about that a class which has only one member is not identical with
that one member. (Russell, 1998, p. 183)

Gödel considered this as an overreaction to the paradoxes, saying that

Russell’s argumentation shows, at best, “that the null class and the unit
classes (as distinct from their only element) are fictions, not that all the
classes are fictions” (Gödel, 1944: 141). The two particular classes men-
tioned above could be treated, Gödel continues, like points at infinity in
geometry, and similar mathematical generalizations. This, however, would
require that we generalize and create new axiomatic structures, in terms
of pure relational thinking and awareness of which relationships might be
essential and truthful, rather than being mere contingent facts. With this,
however, we should be back to the axiomatic view and then would have
moved in a circle and would again have to bother about Russell’s problem
of determining the consistency and applicability of an axiomatized theory.
From relations and functions (of axiomatics) to objects (sets) and from
there back again to relations or functions (propositional functions): this to
and fro movement demonstrates the logician’s desperate quest for a thing
absolutely given and existent, in order to make things ever more precise.

R EFERENCES

Ausubel, D.P.: 1960, ‘The use of advance organizers in the learning and retention of
meaningful verbal learning’, Journal of Educational Psychology 51, 267–272.
Benacerraf, P. and Putnam, H. (eds.): 1983, Philosophy of Mathematics, Cambridge
University Press, Cambridge.
Beth, E. and Piaget, J.: 1966, Mathematical Epistemology and Psychology, Reidel,
Dordrecht.
Bochner, S.: 1966, The Role of Mathematics in the Rise of Science, Princeton University
Press, Princeton, NJ.
Boutroux, P.: 1920, L’Idéal Scientifique Des Mathématiciens, F. Alcan, Paris.
Bruner, J.S.: 1961, ‘The act of discovery’, Harvard Educational Review 31(1), 21–32.
Cantor, G.: 1883/1980, Gesammelte Abhandlungen, Springer Verlag, Heidelberg.
Castonguay, C.: 1972, Meaning and Existence in Mathematics, Springer Verlag, Wien.
Dedekind, R.: 1969, Was sind und was sollen die Zahlen? Vieweg, Braunschweig.
Douady, R.: 1991, ‘Tool, object, setting, window,’ in A. Bishop, J. van Dormolen and S.
Mellin-Olsen (eds.), Mathematical Knowledge: Its Growth Through Teaching, Reidel,
Dordrecht, pp. 109–132.
Dummett, M.: 1991, Frege, Harvard University Press, Cambridge, MA.
 COMPLEMENTARITY, SETS AND NUMBERS 227

Frege, G.: 1988, Die Grundlagen der Arithmetik: eine logisch mathematische Unter-
suchung über den Begriff der Zahl / auf d. Grundlage d. Centenarausg. hrsg. von
Christian Thiel, Meiner, Hamburg.
Frege, G.: 1969, Funktion, Begriff, Bedeutung, Vandenhoeck & Ruprecht, Göttingen.
Frege, G.: 1980, The Foundations of Arithmetic, Blackwell, Oxford.
Gödel, K.: 1944, ‘Russell’s Mathematical Logic’, in P.A. Schilpp (ed.), The Philosophy of
Bertrand Russell, Open Court, La Salle, pp. 123–154.
Hintikka, J.: 1992, ‘Kant on the Mathematical Method,’ in C. Posy (ed.), Kant’s Philosophy
of Mathematics, Kluwer, Dordrecht.
Jahnke, H.N.: 1990, Mathematik und Bildung in der Humboldtschen Reform, Vandenhoeck
& Ruprecht, Göttingen.
Jahnke, H.N.: 1992, ‘Beweisbare Widersprüche – Komplementarität in der Mathematik’,
in E.P. Fischer, H.S. Herzka and K.H. Reich (eds.), Widersprüchliche Wirklichkeit, Piper,
München, Zürich, pp. 98–130.
Jakobson R. and Halle, M.: 1956, Fundamentals of Language, Mouton, The Hague.
Kuyk, W.: 1977, Complementarity in Mathematics, Reidel, Dordrecht.
Lebesgue, H.: 1966, Measure and the integral, Holden-Day, San Francisco.
Lebesgue, H.: 1974, Message d’un Mathématicien, Blanchard, Paris.
Minsky, M.L.: 1967, Computation: finite and infinite machines, Prentice-Hall, London.
Otte, M. and Steinbring, H.: 1977, ‘Probleme der Begriffsentwicklung – zum Stetigkeits-
begriff’, Didaktik der Mathematik 1, 16–25.
Otte, M., Keitel, C. and Seeger, F.: 1980, ‘Text, Wissen, Tätigkeit,’ Königstein/ Ts.:
Scriptor.
Otte, M.: 1984, ‘Komplementarität’, Dialektik 8, 60–75.
Otte, M.: 1990, ‘Arithmetic and Geometry – Some Remarks on the Concept of Comple-
mentarity’, Studies in Philosophy and Education 10, 37–62.
Otte, M.: 1994, Das Formale, das Soziale und das Subjektive, Suhrkamp Verlag, Frank-
furt/M..
Otte, M.: 1994a, ‘Intuition and logic in mathematics’, in D.E. Robitaille, D.H. Wheeler
and C. Kieran (eds.), Selected Lectures from the Seventh International Congress on
Mathematical Education, Les Presses de l’Université Laval, Sainte-Foy, pp. 271–284.
Peirce CP:= Collected Papers of Charles Sanders Peirce, Volumes I-VI, edited by C.
Hartshorne and P. Weiß, 1931–1935; Volumes VII-VIII, edited by A.W. Burks, 1958,
Cambridge, Mass.: Harvard University Press (reference by volume and paragraph
numbers)
Peirce MS:= Robin, R.S: 1967, Annotated Catalogue of the Papers of Charles S. Peirce,
The University of Massachusetts Press, Cambridge, MA.
Peirce NEM:= Eisele, C. (ed.): 1967, The New Elements of Mathematics by Charles S.
Peirce, Vol. I-IV, Mouton / Humanities Press, The Hague-Paris/Atlantic Highlands, N.J.
Port-Royal Logic / [by Antoine Arnauld and Pierre Nicole]. Transl. from the French with
introd. notes and appendix by Thomas Spencer Baynes, 2nd edition, Sutherland and
Knox, Edinburgh.
Potter, M.: 1990, Sets, Oxford University Press, Oxford.
Quine, W.V.: 1969, Ontological Relativity and other Essays, Columbia University Press,
New York, NY.
Reid, C.: 1970, Hilbert, Springer Verlag, Heidelberg.
Russell, B.: 1903/1937, The Principles of Mathematics, Routledge, London.
Russell, B.: 1954, The Analysis of Matter, Allan & Unwin, London.
228 MICHAEL OTTE

Russell, B.: 1971, in R.C. Marsh (ed.), Logic and Knowledge: Essays 1901–1950, Allen &
Unwin, London.
Russell, B.: 1998, Introduction to Mathematical Philosophy, Routledge, London.
Sfard, A.: 1991, ‘On the dual nature of mathematical conceptions’, Educational Studies in
Mathematics 22, 1–36.
Thom, R.: 1973, ‘Modern mathematics: Does it exist?, in A.G. Howson (ed.), De-
velopments in Mathematical Education, Cambridge University Press, Cambridge, pp.
194–212.
Winter, H.: 1982, ‘Das Gleichheitszeichen im Mathematikunterricht’, Mathematica Di-
dactica 5, 185–211.

Postfach 100131,
Universität Bielefeld,
D-33615 Bielefeld