The Changes of Language in the Development of Mathematics

Ladislav Kvasz, Faculty of Mathematics and Physics, Comenius University,

Mlynská dolina, 84215 Bratislava, Slovak Republic

ABSTRACT
The nature of changes in mathematics was discussed recently in a collection of papers Revolutions in
Mathematics (Gillies 1992). Although the question of the nature of changes has many philosophical aspects,
the discussion was dominated by historical and sociological arguments. An obstacle for a philosophical
analysis of this question is a deep discrepancy between our approach to formulas and to pictures in
contemporary philosophy of mathematics. While the formulas are understood as fully legitimate constituents
of mathematical theories, pictures are viewed only as kind of heuristic tools. For this reason the pictorial side
of mathematics can be treated only in the historical or sociological context. A philosophical analysis of
geometrical pictures comparable to the analysis of the formulas of arithmetic, which was started by Frege,
Peano and Russell, is still lacking.
The basic idea of this paper is to consider the pictures, contained in mathematical texts, as
expressions of a specific language. Thus mathematics contains two formal languages - the symbolic language
of arithmetic (algebra, calculus...) and the iconic language of geometry (synthetic, analytic or fractal). In this
way we get both - formulas and pictures - into one basic framework and so we will be able to analyse their
interplay in the course of history. The paper contains an analysis of the development of the symbolic and
iconic representations in the history of mathematics. We try to show how the expressive and the logical power
of mathematics grew due to the interplay of these two ways of representation.

1. Introduction
The roles of geometry and of arithmetic in contemporary philosophy of mathematics are
rather asymmetric. While arithmetic plays a central role in foundational approaches and therefore its
logical structure is thoroughly studied and well understood, geometry is the central topic of the
antifoundational approaches where the study is focused primarily on the patterns of discovery and
plausible reasoning. This of course does not mean, that there are no foundational studies of geometry.
It is sufficient to mention Hilbert or Tarski. Nevertheless, in these cases geometry is just another
illustration of the methods developed for the analysis of arithmetic. The visual aspect of geometry, the
very fact that geometry has something to do with space and spatial intuition is in these studies totally
ignored. On the other hand there are also some antifoundational studies of arithmetic. These studies
are just exceptions and they do not change the basic difference, that the philosophy of arithmetic is
dominated by the foundational approach, while philosophy of geometry is mainly antifoundational.
The reason for this asymmetry lies in the different attitude to the languages of these two main
parts of mathematics. Since the works of Frege, Peano, and Russell the language of arithmetic is fully
formalised and so the formulas of arithmetic are considered to be a constitutive part of the theory. On
the other hand in geometry the geometrical pictures are considered only as heuristic aids, which can
help us to understand the theory, but strictly speaking, they do not belong to the theory itself. Since
Hilbert the content of a geometrical theory is independent of any pictures and is given by the set of its

1
axioms. Thus in arithmetic in the course of its formalisation the specific symbols as "+" or "≤ " as
well as the rules which they obey are considered a part of the language. In geometry the process of
formalisation took rather the opposite direction and all the special symbols of geometry, like "." or
"" were excluded from the language. An interesting analysis of the reasons for this exclusion of
diagrams and of diagrammatic reasoning from the foundations of mathematics is given in Greaves
1997.
Although there were good reasons for such development of the foundations of mathematics,
we believe it might be interesting to try to bridge the gap between the philosophy of arithmetic and
the philosophy of geometry. In order to overcome the separation of geometry from arithmetic, it is
necessary to do the same in geometry what Frege did in arithmetic. First of all this means to formalise
its language. In this way the pictures turn from mere heuristic aids into an integral part of the theories
themselves. A picture is not just the physical object formed by spots of graphite on a more or less
smooth surface of the paper. We understand the picture as an expression of the iconic language with
its own meaning and reference. We follow here an analogy with arithmetic or algebra, where a
formula is understood not as a physical object, i.e. not as spots of ink on a sheet of paper. For our
epistemological purposes it is enough to give a short characterisation of the iconic language of
geometry. We interpret a geometrical picture as an expression (a term) of the iconic language. So a
geometrical construction becomes a generating sequence of the resulting expression (picture). In this
way the Euclidean postulates become formation rules of this language, analogous to the Fregean rules
for symbolic languages, which prescribe, how from an n-ary functional symbol F and n terms t1,

t2, ...tn a new term F(t1, t2, ...tn) is formed. A picture is called a well-formed expression, if each
construction step is performed in accordance with the formation rules (axioms).
We could give an inductive definition of a well-formed expression of the language of
geometry:
1. The empty picture is a term of the language.
2. If t is a term, then a picture, which we form by adding a point or a line segment, in any position to
t, is a term of the language.
3. If t is a term, which contains two different points A and B, then the picture, which we get by
joining the points A and B by a straight line, is a term of the language.
If we rewrite in this manner the Euclidean postulates 1, 2, 3, and 5 (see Euclid, p. 154), we
obtain a general description of the language. The questions, when two terms are equal, or how can we
introduce predicates into the pictorial language, and how do its propositions look like, are subtle
questions, which we don’t want to rise now. They are more logical than epistemological in nature, and
would require more detailed investigations, which would lead us rather far from the subject of this
paper. For our present purposes it is sufficient to realise, that the iconic language of geometry can be
treated with similar strength and precision, as that which Frege introduced for arithmetic. Seen from
this position, mathematics for us will no longer consist of an exact symbolic language supplemented
with some heuristic pictures, but rather of two languages of the same rank. One of them is symbolic

2
and the other is iconic.
Of course the pictures of Euclidean (synthetic) geometry are not the only pictures used in
geometry. There are pictures also in analytic (algebraic or differential) geometry as well as in iterative
(fractal) geometry. The situation is here similar to Frege’s. He also had to supplement the formulas of
elementary arithmetic with the formal accomplishments of algebra and calculus. Frege described the
evolution of the symbolic language from elementary arithmetic through algebra and mathematical
analysis to predicate calculus in the closing part of his paper Funktion und Begriff:
„If we look back from here over the development of arithmetic, we discern an advance from
level to level. At first people did calculations with individual numbers, 1, 3, etc.
2+3=5 2.3 = 6
are theorems of this sort. Then they went on to more general laws that hold good for all numbers.
What corresponds to this in symbolism is the transition to the literal notation. A theorem of this sort
is
(a + b).c = a.c + b.c.
At this stage they had got to the point of dealing with individual functions; but were not yet using the
word, in its mathematical sense, and had not yet formed the conception of what it now stands for.
The next higher level was the recognition of general laws about functions, accompanied by the
coinage of the technical term „function“. What corresponds to this in symbolism is the introduction
of letters like f, F, to indicate functions indefinitely. A theorem of this sort is
dF ( x ). f ( x ) df ( x ) dF ( x )
= F( x ). + f ( x ).
dx dx dx
Now at this point people had particular second-level functions, but lacked the conception of what we
have called second-level functions. By forming that, we make the next step forward.“ (Frege 1891,
p.30; English translation p. 40)
Our interpretation of this development of the symbolic language will differ from Frege’s in
two respects. First is terminological - we will not subsume algebra or mathematical analysis under the
term "arithmetic", but will rather consider them as independent languages. More important, however,
is the fact that we will show, how this „development of arithmetic“ described by Frege, interplayed
with the geometrical intuition. In order to achieve this we need to complement Frege’s analysis of the
„development of arithmetic“ with a similar „development of geometry“. Frege identified the main
events in the development of the symbolic language as the introduction of the concept of „individual
functions“ in algebra, then of the „particular second-level functions“ in the calculus and finally the
general concept of „second-level functions“ of the predicate calculus. In a similar way we will try to
identify the crucial events of the development of the iconic language of geometry. It will turn out that
the events parallel to those described by Frege are the creation of analytical geometry, fractal
geometry and set theory.
We will describe an interesting periodical motion in the history of mathematics, consisting in
alternation of its symbolic and geometrical periods. This phenomenon has not been sufficiently

3
understood yet. Nobody has tried to undertake a serious epistemological analysis of it. The reason
could be that this alternation of symbolic and geometrical periods in mathematics has a vague nature,
what dissuades people from undertaking its serious analysis. Nevertheless, we are convinced, this
vagueness is only a result of insufficient understanding of the nature of geometrical pictures. As long
as pictures and the whole of geometrical intuition in general remains something vague and
psychological, the alternation of this intuition with the formal languages, which is clearly visible in
the history of mathematics, must have vague contours. Thus, interpreting pictures as iconic language,
we create a framework, which makes it possible to understand the relations between the symbolic and
iconic periods.

2. History of mathematics seen as development of its language

In the analysis of the alternation of the geometrical and symbolic periods in the history of
mathematics we will follow an analogy with the ideas of Frege. In the creation of the predicate
calculus (Frege 1879) Frege took arithmetic as his starting point. This was correct in his times, i.e.
during the deepening of the crisis of the foundations of mathematics. But to follow Frege’s footsteps
means to follow the spirit and not the letter of his work. So we try to do the same with the language of
geometry what Frege did with the language of arithmetic. Detailed elaboration of this task exceeds the
present paper. It exceeds its epistemological horizon towards logic. So in some sense the whole of this
paper can be understood as a preliminary epistemological inquiry, plotting the starting points of
logical investigations of the pictorial language of geometry which are to follow.
The unification of the symbolic and iconic languages enables us to consider the development
of mathematics as the evolution of its language and in this way to provide a basis for the
understanding of the alternation of the symbolic and geometrical periods in history. We will study the
development of the language of mathematics from the following six aspects:
1. logical power - how complex formulas can be proven in the language
2. expressive power - what new things, which were inexpressible on the previous stages,
can the language express
3. explanatory power - how the language can explain the failures which occurred on the
previous stages
4. integrative power - what sort of unity and order the language enables us to conceive there,
where we perceived just unrelated particular cases on the previous stages
5. logical boundaries - marked by occurrences of unexpected paradoxical expressions
6. expressive boundaries - marked by failures of the language to describe some complex
situations
The evolution of the language of mathematics consists in the growth of its logical, expressive,
explanatory, and integrative power as well as in the overcoming of its logical and expressive
boundaries. Every language has its own particular logical, expressive, explanatory, and integrative
power as well as its own logical and expressive boundaries. In this paper we will try to characterise

4
each language from these six aspects. Nevertheless, it may happen, that by a particular language we
are not able to describe some of these aspects due to their complexity or technical difficulties. We will
leave them open for further investigation. We would like to stress that this is a paper aiming to present
a new philosophical approach to mathematics. The analysis of each of the eight described languages
would require a separate paper of approximately the same length, and many of the six aspects would
require special historical as well as philosophical research. Nevertheless, we believe it is useful to
have an overall picture presenting a global perspective and aims of research, not to lose our way in the
course of further investigations.
Mathematics has a tendency to improve its languages additionally. So, for instance, we are
used to introducing the concept of variable into the language of arithmetic (enabling us to write
equations in this language) and often we choose the field of real numbers as a base (so that the
language is closed with respect to limits). This is very convenient from the pragmatic point of view,
because it offers us a strong language, in which we can move freely, without any constraints. But, on
the other hand, it makes us insensitive to the historically existing languages. The old languages do not
appear to us as independent systems with their own logical and expressive powers. They appear only
as some fragments of our powerful language. As the aim of this paper is the epistemological analysis
of the language of mathematics, we try to characterise every language possibly close to the level on
which it was created. We ignore the later emendations, consisting in incorporation of achievements of
the later development (for instance of the concept of variable in algebra) into the former languages (of
arithmetic). In this paper the language of elementary arithmetic will be a language without variables.
We think, such stratification of the language of mathematics into the different historical layers will be
interesting also for the logical investigations, enabling to see the order in which different logical tools
appeared.
Before starting our investigation of the development of mathematics, we would like to say
that this is not a historical but a philosophical paper. That means that our aim is not to present some
new historical facts or interpretations. For the purpose of our investigation, standard presentation of
the history of mathematics as it can be found in the classical textbooks of this subject, say Boyer and
Merzbach 1989, is fully sufficient. We admit that historical research has overcome these
interpretations in many cases. Wilbur Knorr speaks in this context about „sifting history from legend“
(Knorr 1986). Nevertheless, we see confrontation with the historical research as a further stage in the
development of our theory. It will lead to refinement in many technical aspects and perhaps also to
improvements in some details, but we don’t believe, it is necessary to do it now.
The philosophical approach always represents a reduction in comparison with the historical
one. A historian studies the results of particular mathematician, their connections with the results of
his predecessors, contemporaries, and successors. In the philosophical perspective only those of the
particular results are interesting, which became part of the knowledge shared by the community. The
philosophical analysis is always a bit too simplistic from the historical point of view. Nevertheless, as
the quotation from Frege shows, it can bring interesting insights. Our aim is to identify the basic

5
events in the development of the language of mathematics. We do not mean these events in the
historical sense. We are not (at this stage) interested exactly who and when brought what basic
innovation. We consider these innovations as constituents of the language, shared by (smaller or
larger) community. Usually this shared language is based not precisely on the original version of the
first discoverer, but more often it uses some later modifications. Our aim is just to identify the basic
innovations in the development language of mathematics. To connect these innovations to particular
historical figures is a further question, which we will not rise now.

2.1 Elementary Arithmetic

Even if the main purpose of this paper is to bring the iconic language of geometry into the
centre of philosophical consideration, we will start our analysis with the symbolic language of
arithmetic. The reason is very simple - we follow the historical order and arithmetic was the first
mathematical theory at all. The language of elementary arithmetic is the simplest symbolic language.
It is based on manipulations with numerical symbols. There are many variants of this language, the
most common contains ten symbols for numerals 1, ..., 9, 0 and the symbols +, −, x, : and =. A basic
feature of this language is, that it has no symbol for variable. For this reason it is impossible to
express in this language any general statement or write a general formula. So for instance the rules for
division or for multiplication, as they are general statements, are inexpressible in this language. They
cannot be expressed in the language, but only shown.

a. logical power - verification of singular statements

Typical statements of elementary arithmetic are singular statements such as:
72 + 39 = 111 or 24 x 8 = 192
The language of elementary arithmetic contains implicit rules, which with the help of formal
manipulation with symbols enable verifying such statements. In the history of mankind it was the first
formal system making it possible to decide on purely syntactical basis the truth of some statement.
One consequence of the fact, that the language of elementary arithmetic didn’t contain variables was
the necessity for formulating all problems with concrete numbers. Let us take for example a problem
from the Rhind papyrus:
„Find the volume of a cylindrical granary of diameter 10 and height 10.
Take away 1/9 of 10, namely 1 1/9; the remainder is 8 2/3 1/6 1/18. Multiply 8 2/3 1/6 1/18
times 8 2/3 1/6 1/18; it makes 79 1/108 1/324 times 10; it makes 790 1/18 1/27 1/54 1/81 cubed
cubits. Add 1/2 of it to it; it makes 1185 1/6 1/54, its contents in khar. 1/20 of this is 59 1/4 1/108. 59
1/4 1/108 times 100 hekat of grain will go into it.” (Fauvel and Gray 1987, p. 18)
The scribe instead of rewriting the problem into a form of equation and solving it in a general
way, as we would proceed today, has to take the numbers from the formulation and to make with
them particular arithmetical operations, until he gets what he needs. With the help of concrete
calculations the general method, which is inexpressible in the language, is shown. The language of
elementary arithmetic was the first formal language in the history, which made it possible to solve

6
problems by manipulation with symbols. Its logical power is restricted to verification of singular
statements.

b. expressive power - ability to express however large numbers

The language of elementary arithmetic is able to express however large natural number. This
may seem nothing special, as we are used to negative, irrational as well as complex numbers and so
the natural numbers seem to us as a rather poor and limited system, where it is not possible to distract
or divide without constraints. Nevertheless, if we leave out of consideration these results of later
developments, maybe we will be able to feel the fascination, which must have seized the ancient
Egyptian (Babylonian, Indian, Chinese) scribe when he realised, that with the help of numbers it is
possible to count everything. This is the basis of bureaucratic planning, which was one of the most
important discoveries of the ancient cultures. The universality of bureaucracy is based on expressive
power of the language of arithmetic. Rests of the fascination by the expressive power of the language
of arithmetic can be found in affinity to large or special numbers in mythology, in the Kabbalah, in
the Pythagoreans and even in Archimedes. His book The Sand-reckoner (Fauvel and Gray 1987, p.
150) is devoted to demonstration of the expressive power of the language of elementary arithmetic.

c. explanatory power - the language of elementary arithmetic is nonexplanatory

The explanatory power of a language is defined in connection with the previous stage of the
development. As the language of arithmetic is the first formal language of mathematics, it is
nonexplanatory. This is obvious also from the preserved mathematical texts, which have the form of
collections of recipes, comprising a sequence of instructions without any explanation. This character
of the language was noticed also by the historians. For instance, Gray speaks about „contradictory
and nonexplanatory results“ (Gray 1979, p. 3).

d. integrative power - the language of elementary arithmetic is nonintegrative

For the same reasons as in the previous section, the language of elementary arithmetic is
nonintegrative. It makes it not possible to conceive any kind of unity or order. The mathematical texts
are mere collections of unrelated particular cases. Ordering of the problems is based on their content
(problems on calculation of areas of fields, volumes of granaries, etc.) instead of on their form. This
means that the ordering principle stems from outside of the language.

e. logical boundaries - existence of incompatible procedures

The logical boundaries of the language of elementary arithmetic manifest themselves in the
existence of incompatible procedures to solve the same problem. As an example we can take the
procedure used by the Egyptians to „calculate the area of a quadrilateral by multiplying the half the
sum of two opposite sides times half the sum of the other two sides“ (Edwards 1979, p. 2). This
procedure is wrong, but not absolutely. For a large field it gives a large number, for a small field a
small number. If the fields are approximately of the same form, the procedure makes it possible to
order them in accordance with their magnitude and to determine just taxes. Of course, for fields with
very slanting sides, the magnitude of the field determined according this procedure differs

7
considerably from its area. But who would stake out fields of such a form? Thus for the practical
needs the procedure is fully sufficient, even if, strictly speaking, it is wrong. As another example we
can take the Babylonian procedure to calculate the volume of a frustum of a cone or pyramid
(Edwards 1979, p. 4).
These examples display a fundamental shortcoming of the language of elementary arithmetic -
its inability to avoid incorrect procedures. These incorrect procedures are not just mistakes. They are
not caused by some lack of skill or attention of a mathematician. They form rather a characteristic
feature of the language itself. The language of elementary arithmetic makes it not possible to
introduce any concept of correctness for procedures used to calculate areas or volumes. This
shortcoming of the language is caused by the fact, that there is no other way to determine an area or
volume, independent of the calculation. Therefore there is no reality, we could compare the calculated
result with, in order to determine its correctness. Thus the result obtained by the scribe in the process
of calculation, cannot be correct or incorrect. No other, from the calculation independent concept of
area or volume, does exist. Therefore arithmetic is „infallible“, or more precisely, the scribe cannot be
aware of its fallibility. He may make mistakes as any human being. He may forget a step of the
procedure or execute it wrongly. But the procedures themselves are infallible. He cannot adopt a
critical attitude towards these procedures, because the language does not make it possible. So the
feeling of infallibility of the Egyptian or Babylonian officials, as well as of some of their
contemporary descendants, is not the result of their pride or conceit. It is rather the consequence of the
characteristic features of the language of elementary arithmetic, which they were using. A language,
which did not make it possible to create a concept of reality, independent of the language itself, with
which the results of calculations could be compared.
The only possibility how a scribe could be confronted with the logical boundaries of his
language, was the incompatibility of his procedures, i.e. the existence of procedures, which for the
same problem give different results. This is the reason, why we did not characterise the logical
boundaries of the language of elementary arithmetic as „existence of incorrect procedures“. Such a
characterisation would be perhaps more natural from the present point of view, but the concept of
correctness is alien to this language. Therefore we chose rather the term „existence of incompatible
recipes“. The scribes protected themselves from these logical boundaries by a strict canonisation of
their recipes. If everybody uses the same (even if wrong) method, the incompatibility becomes
invisible.

f. expressive boundaries - incommensurability of the side and diagonal of a square

The Pythagoreans developed a qualitatively new kind of formal language. It was the iconic
language of geometry. Nevertheless, at the beginning they connected this new geometrical language
with an interesting kind of „arithmetical atomism“. The Pythagoreans supposed every quantity,
among others also the side and the diagonal of a square, comprise a finite number of units. So the
proportion of the lengths of the side and the diagonal of the square equals the proportion between the
numbers of units, from which they are composed. The discovery of the incommensurability of the

8
side and the diagonal of the square refuted the Pythagorean atomism. It shows, however, that the
language of geometry is more general than that of arithmetic. In arithmetic the side and diagonal of a
square cannot be included in one calculation. We can either choose a unit commensurable with the
side, but then it will be impossible to express the length of the diagonal by a number, or we can
choose a unit commensurable with the diagonal, but then we will be unable to express the length of
the side. So the incommensurability of the side and diagonal of the square reveals the boundaries of
the expressive power of the language of elementary arithmetic.

2.2 Synthetic geometry

The geometrical language, in contrast to the language of arithmetic, is an iconic and not a
symbolic language. Its expressions are pictures formed of a finite number of points, line segments and
circles, rather than formulas formed of linear sequences of arithmetical symbols. From this point of
view the Greeks moved from the symbolic language of arithmetic to the iconic language of synthetic
geometry. They had good reasons for this, because, as we will see, the geometrical language surpasses
the language of elementary arithmetic in logical as well as in expressive power.

a. logical power - ability to prove universal theorems

The new mathematical language was originally developed in close connection with
arithmetic. We have in mind the famous Pythagorean theory of figurate numbers (see Boyer and
Merzbach 1989, p. 62). Surely, its content is arithmetical it deals with numbers. But its form is quite
new. Using small dots in sand or pebbles (psefos) it represents numbers geometrically - as square
numbers (i.e. numbers the psefoi of which can be arranged into a square, like 4, 9, 16, ...), triangular
numbers (like 3, 6, 10, ...), and so on. With the help of this geometrical form arithmetical predicates
can be visualised. For instance an even number is a number, the psefoi of which can be ordered into a
double row. So the property of being odd or even becomes a property expressible in the new
geometrical language. This very fact, that different arithmetical properties became expressible in the
language makes it possible to prove universal theorems, (and not only particular statements, as it was
the case until then). For instance the theorem that the sum of two even numbers is even can be easily
proved using this Pythagorean language. It follows from the fact, that if we connect two double rows,
one to the end of the other, what we get, will be again a double row. Therefore the sum of any two
even numbers must be even. We see that the Pythagoreans introduced an important innovation into
the language of mathematics. With the help of their figurate numbers they were able to express
general arithmetical predicates (like evenness or squareness - which in the frame of the previous
arithmetic were inexpressible) and to prove general theorems.
The discovery of incommensurability led the Greeks to abandon the Pythagorean arithmetical
basis of their new geometrical language and to separate the geometrical forms from the arithmetical
content. In this process the new language - the iconic language of geometry was created. From the
logical point of view this new geometrical language is much stronger than the language of elementary
arithmetic. It makes it possible to prove universal statements. The language of geometry is able to do

9
this, thanks to an expression of a new kind - a segment of indefinite length. (In fact this was the
essence of the Pythagorean innovation - they were able to prove that the sum of any even numbers is
even, because the double row, which represented an even number could be of any length. The
geometrical form is independent of the particular arithmetical value, to which it is applied.) If we
prove some statement for such a segment, in fact we have proved the statement for a segment of any
length, which means that we have proved a general proposition. The segment of indefinite length is
not a variable, because it is an expression of the iconic and not of the symbolic language. That means
it does not refer to, but rather represents the particular objects (side of a triangle, radius of a circle,
etc.). Of course, any concrete segment drawn in the picture has a precise length, but this length is not
used in the proof, that means, the particular length is irrelevant,.

b. expressive power - ability to overcome incommensurability

The language of geometry is superior to that of elementary arithmetic also regarding its
expressive power. From the geometrical point of view, the side and diagonal of the square are
ordinary segments. Thus the geometrical language can thematize situations, inaccessible to the
language of elementary arithmetic.

c. explanatory power - ability to explain the non-existence of a solution of a problem

The language of synthetic geometry can explain a strange fact, probably noticed already by
the Babylonians, namely that some mathematical problems have no solution.
„One advantage of the appeal to geometry can now be mentioned to illustrate the gain in
explanatory power. Evidently there are no numbers x and y whose sum is 10 and product 40, and the
Babylonian scribes seem to have avoided discussing such questions. However, we can now see why
there are no such numbers. In the language of application of areas we have to put a rectangle of area
40 on a segment of the length 10 leaving a square behind.

C x

y x

a
The area xy of the large rectangle C varies with x (and therefore with y) but is greatest when
the rectangle is a square. In that case x = y = a/2, and the area is a2/4. Therefore we can solve the
problem provided a2/4 exceeds the specified area C. In our example 100/4 = 25 is not greater than
40, so no numbers can be found. The discussion of the feasibility of finding a solution is indeed to be
found in Euclid (Book VI, prop. 27) immediately preceding the solution of the quadratics
themselves.“ (Gray 1979, p. 24).
Thus the language of synthetic geometry enables to understand the conditions, under which
problems of elementary arithmetic have solution. In this way the language of geometry makes it

10
possible to express explicitly the boundaries of the language of arithmetic.

d. integrative power - ability to integrate the problems of the same complexity

The language of synthetic geometry makes it possible to unify different problems on the basis
of the complexity of methods necessary for their solution. The language of elementary arithmetic is
unable to do this. This is the reason why the collections of problems from Egypt or Mesopotamia are
so unsystematic. Their classification into linear, quadratic, ..., is the work of modern commentators.
The Egyptians or Babylonians could not use such classification, because they had no concept of
equation, and so there was nothing to be linear, quadratic, ...
The language of synthetic geometry makes it possible to represent the process of solution of
an arithmetic problem in form of geometrical construction. Such a geometrical construction usually
just copies the steps of the original calculation and so there seems to be no difference. Nevertheless,
there is one. In an arithmetical calculation if we execute the particular steps leading to the final result,
we obtain just one number in each step. Let us recall the problem from the Rhind papyrus.
„...Multiply 8 2/3 1/6 1/18 times 8 2/3 1/6 1/18; it makes 79 1/108 1/324.... Add 1/2 of it to it; it
makes 1185 1/6 1/54... 1/20 of this is 59 1/4 1/108...” Thus the result is always a single number.
Contrary to this, in geometry all the intermediate lines and circles, which were used during
the construction of the resulting line segment (representing the solution) remain as constituents on the
resulting picture. Thus the intermediate steps of the geometrical construction are not lost in the same
way as in arithmetic, were we loose the intermediate numbers and obtain the result as a single
number. The language of geometry makes it possible to classify (i.e. to integrate into classes)
problems on the basis of the complexity of the resulting picture. This resulting picture represents the
complexity of the actual steps needed in the process of the construction of the solution. In arithmetic
the final result does not reveal any complexity and therefore it cannot be used for any classification.

e. logical boundaries - existence of insoluble problems

There are some well known geometrical problems, formulated during the early development
of Greek mathematics, which turned out to be insoluble using just the elementary methods of ruler
and compass construction. These problems are: to trisect an angle, to duplicate a cube, and to
construct a square with the same area as a circle. The insolubility of these problems using the classical
geometrical methods of construction was proven in modern algebra and complex analysis, that is, in
languages of higher expressive and explanatory power than the language of synthetic geometry. From
the point of view of synthetic geometry the fact, that nobody succeeded to solve these problems
despite the efforts of the best mathematicians, must have been a paradox. It showed the boundaries of
the geometrical language.
Of course there were presented many solutions using curves or methods which were
characterised by some historians of mathematics as ad hoc (Gray 1979, p. 16). As an example we can
take the trisectrix, a curve invented by Hippias, or the method of neusis, invented by Archimedes (see
Boyer and Merzbach, 1989, p. 79 and p. 151). Nevertheless, these new curves did not belong to the

11
language, as characterised in the introduction. They contained points, which could not be constructed
using only ruler and compass, but were given with the help of some mechanical devices. So even if
these new methods of construction are important from the historical perspective and attracted the
attention of historians (see Knorr 1986), they fall outside the scope of our paper.

f. expressive boundaries - equations of higher degrees

The new geometrical language prevailed in Greek mathematics to such a degree, that Euclid,
when confronted with the problem of solving a quadratic equation, presented the solution as a
geometrical construction.
„The discovery of incommensurability and the impossibility of expressing the proportion of
any two segments as a proportion of two natural numbers led the Greeks to start to use proportions
between geometrical magnitudes instead of arithmetical proportions and with their help to express
general proportions between magnitudes...
In order to solve the equation cx = b2, the Greeks regarded the b2 as a given area, c as a
given line segment and x the unknown segment. They transformed thus the problem into the
construction of an oblong whose area and one side is known. Or as they themselves called it, „the
attachment of the area to a given line segment...“ (Kolman, 1961, p. 115).
Solving equations with the help of geometrical constructions avoids the incommensurability.
However, such methods are appropriate only for linear and quadratic equations. The cubic equations
represented real technical complications, as they deal with volumes. The equations of higher degree
were beyond the expressive boundaries of the language of synthetic geometry.

2.3 Algebra
In order to overcome the problems that arise, if we want to solve equations of higher degrees,
it is necessary to develop a new language. This language was developed gradually by the Italian and
German mathematicians during the 15th and 16th century (see Boyer and Merzbach 1989, pp. 312-
316). The main invention of these mathematicians was the introduction of variable, which they called
cosa, from the Italian word meaning thing. They called algebra regula della cosa, i.e. the rule of the
thing. In this way they introduced a symbolic language, in which they could manipulate letters, just as
we manipulate things. For instance, if we add to a thing an equally great thing, we obtain two things,
what they wrote as 2r (to indicate the thing in their symbolism they usually used the first letter of the
Latin word res). This new language is a return from geometrical construction back to symbolic
manipulations, from the iconic to the symbolic language.

a. logical power - ability to prove modal predicates

In comparison with the language of classical arithmetic, the new symbolic language has a
great innovation - it contains a symbol for the variable. Thus we can say, that the algebraists
succeeded in transferring the basic advantage of the language of geometry - its ability to prove general
propositions into the symbolic language. For instance, it is possible to prove that the sum of two even
numbers is even, by the formal calculation

12
2l + 2k = 2(l+k)
So the symbolic language reached the generality of the iconic language of synthetic geometry. This
generality was achieved in geometry by using line segments of indefinite lengths, in algebra with the
help of variables. Nevertheless, the new symbolic language of algebra superseded in logical power the
geometrical language. Let us take for instance the formula for the solution of the quadratic equation

x1,2 = −b ± b − 4 ac .
2

2a
In our characterisation of the integrative power of the language of synthetic geometry, we
indicated the importance of the fact, that the particular steps of geometrical construction are not lost
(as are lost the intermediate steps during a calculation). Each line or point used in the process of
construction remains as constituent of the resulting picture. Nevertheless, what gets lost in the process
of construction is the order of its particular steps. This is the reason why geometrical construction is
usually supplemented by a commentary written in the ordinary language, which indicates the precise
order of its steps. It is important to notice that the commentaries do not belong to the iconic language
of geometry. They are extralinguistic.
The language of algebra, on the other hand, is able to express the order of particular steps of
calculation inside the language. Thus we need no further commentary to the above formula similar to
the commentary we need to a geometrical construction. The formula represents the process of
calculation. It tells us that first we have to take the square of b, distract from it four times the product
of a and c, etc. So the process of solution becomes expressible in the language. The structure of
algebraic formula indicates the relative order of all steps, necessary for its calculation. Thanks to this
feature of the language of algebra modal predicates, as for instance insolubility, can be expressed
inside the language.
The language of geometry has no means to express or to prove that some problem is
insoluble. The process of solution is something, what the iconic language of geometry cannot express.
As we have shown in the discussion of the explanatory power of the language of geometry, geometry
is able to express the fact that a problem has no solution. Nevertheless, the problem of insolubility is a
more delicate one. There is no doubt, that to each angle there is an angle being just one third of the
first one, or that each equation of the fifth degree has five roots. The insolubility does not mean the
non-existence of the objects solving the particular problem. It means that these objects, even if they
exist, cannot be obtained using some standard methods. The language of algebra is the first language
that is able to prove insolubility. It is clear that from the introduction of the variable in the 15th
century it took a long way to the Galois theory, but this was just an internal development of the
language of algebra. The symbolic language of algebra was the first language, which made it possible
to prove the insolubility of a particular problem.

b. expressive power - ability to form powers of any degree

In geometry the unknown quantity is expressed as a line segment of indefinite length, the

13
second power of unknown quantity is expressed as a square constructed over a line segment and the
third power of the unknown is a cube. The three dimensional space does not let us go further in this
constructions to form the forth or fifth power of the unknown. We characterised this feature of the
language of geometry as its expressive boundaries. The language of algebra is able to transcend these
boundaries and form the fourth or fifth power of the unknown.
We can find traces of geometrical analogies on the algebraic terminology of the 15th and 16th
century, as for instance the third power of the unknown is called cubus. But nothing hindered the
algebraists from going further, also beyond this third degree, beyond which Euclid was not allowed
by the geometrical space. They called the second degree of the unknown zensus and denoted it z. That
is why they wrote the fourth degree as zz (zensus de zensus), for the fifth rzz, for the sixth zzz and so
on. In this way the symbolic language of algebra transcended the boundaries, placed on the language
of geometry by the nature of space. Of course, we are not able to say what does the fifteenth power of
the unknown mean, but this is not important. The language of algebra offers us formal rules for
manipulation with such expressions independently from any interpretation.
The turn from geometrical constructions to symbolic manipulations made it possible to
discover the method of solving the equations of third degree. The history of this discovery is rather
dramatic (see van der Waerden 1985, pp. 52-59) but we present just the result in contemporary
notation. The solution of the equation x3 = bx + c was given in a form equivalent to
2 3 2 3
c  c  b c  c  b (1)
x=3 +   −  +3 −   − 
2  2  3 2  2  3

Of course Cardano newer wrote such a formula. He formulated his rule verbally.
Nevertheless, we think, that this result illustrates clearly the expressive power of the language of
algebra. It is interesting to try to express this result in geometrical terms (as a sum of sides of two
cubes, the volume of each of which is given in the form...) in order to see the advantages of formal
manipulations with symbols over geometrical constructions.

c. explanatory power - ability to explain the insolubility of the trisection of an angle

The language of algebra makes it possible to understand why some geometrical problems, as
the problem of trisection of an angle, the duplicity of a cube and the construction of a regular
heptagon are insoluble with the ruler and compass. All problems solvable using just ruler and
compass, can be characterised as problems in which only line segments of lengths, belonging to some
finite succession of quadratic extensions of the field of rational numbers occur. Thus in order to show
the insolubility of the three mentioned problems it is sufficient to show, that their solution requires
line segments, whose length does not belong to any finite quadratic extension of the field of rational
numbers. This can be done easily (see Courant and Robbins 1941, pp. 134-139). The language of
geometry makes it not possible to understand why the three mentioned problems are insoluble. From
the algebraic point of view it is clear. The ruler and compass constructions take place in too simple
fields.

14
d. integrative power - ability to create universal analytic methods
Euclidean geometry is a collection of disconnected construction tricks. Each problem is
solved in a specific way. Thus the Greek geometry is also based on memorising. Instead of
memorising the complete recipes as the Egyptians, only the fundamental ideas and tricks are to be
remembered. However there is still a large number of them. Algebra replaces these tricks by universal
methods.
This innovation stems from Francois Viéte. Before him the mathematicians used different
letters for different powers of the unknown (r, z, c, zz, rzz, ...), and so they could write equations
having only one unknown, the different powers of which were indicated by all these letters. Viéte’s
idea was to indicate the different powers of the same quantity not with different letters, but to use the
same letter and to indicate its power by a word. Thus he used A planum, A quadratum, and A solidum
for the first three powers of the quantity A. In this way the letters expressed the identity of the
quantity, while the words indicated the particular power. In this way Viéte introduced the distinction
between a parameter and an unknown.
The algebraists before Viéte worked only with equations having numerical coefficients. This
was the consequence of the use of different letters for the powers of the unknown. The algebraists
were fully aware, that their methods were universal, fully independent of the particular values of the
coefficients. Nevertheless, they were not able to express this universality in the language itself. Viéte
liberated algebra from the necessity to calculate with numerical coefficients only. His idea was to
express the coefficients of an equation by letters as well. In order not to confuse the coefficients with
the unknowns, he used vowels (A, E, I, ...) to express the unknowns and consonants (B, C, D, ...) to
express the coefficients. So for instance the equation which we would write as 2 x 3 − 3by 2 = c Viéte
wrote as
A2 cubus − B latus in A3 quadratum equatur C solido.
It is important to notice that for Viéte also the coefficients had dimension (B latus, or C solido), so
that all terms of the equation had to be of the same dimension. So his symbolism was a rather
cumbersome one, and many simplifications were needed until it reached its standard form used at
present. Nevertheless, the basic gain, the existence of universal analytic methods is already present.
Viéte’s analytic art, as he called his method, was based on expressing the unknown quantities
and the parameters of a problem with letters. In this way the relations among these quantities could be
expressed in form of an equation containing letters for the unknown quantities as well as for the
parameters. Solving such an equation we obtain a general result, expressing the solution of all
problems of the same form. In this way generality becomes a constituent of the language. The
existence of universal methods for the solution of whole classes of problems is the fundamental
advantage of the language of algebra. The language of synthetic geometry does not know any
universal methods. Geometry can express universal facts (facts which are true for a whole class of
objects), but it operates with these universal facts using very particular methods. Algebra developed
universal analytic methods, which played a decisive role in the further development of mathematics.

15
e. logical boundaries - casus irreducibilis
Studying equations of the third degree Cardano discovered a strange thing. If we take the
equation x3 = 7x + 6, which has the solution x = 3 (the two further solutions x = −1 and x = −2 were not
considered, as the cosa cannot be less than nothing) and substitute b = 7 and c = 6 into the formula
(1), we obtain

100 3 100
x = 3 3+ − + 3− − .
27 27
Cardano called this case casus irreducibilis, the insoluble case. In many respects it resembles the
discovery of incommensurability. In both cases we are confronted with a situation, in which the
language fails. The attempts to express this situation in the language lead to paradoxes. And in both
cases the therapy consists in extending the basic realm of objects, with the help of which the language
operates. In case of incommensurability it was necessary to introduce the irrational numbers, in case
of casus irreducibilis the complex numbers. After such extensions of the number system the
incommensurability is no more paradoxical, it just indicates the fact that the diagonal of the unit
square has an irrational length. Similarly after the introduction of complex numbers the casus
irreducibilis is no more paradoxical, it just indicates, that the formula expresses the roots of the
equation in the form of a sum of two conjugate complex numbers.

f. expressive boundaries - transcendent numbers

Even if algebra was able to explain why nobody succeeded in solving the problem of
trisection of an angle, the problem of quadrature of a circle resisted the algebraic methods. Gradually
a suspicion arose that this problem is insoluble as well. Nevertheless, its insolubility has not algebraic
reasons. This suspicion found an exact expression in the distinction of algebraic and transcendent
numbers. The transcendent numbers are numbers that cannot be characterised using the language of
algebra. The first example of a transcendent number was given by Joseph Liouville in 1851. It is the
number:

10-1! + 10-2! + 10-3! + 10-4! + 10-5! + 10-6! + ... =

= 10-1 + 10-2 + 10-6 + 10-24 + 10-120 + 10-720 + ... =
= 0,1100010 ... (17 zeros) ... 010 ... (96 zeros) ... 010 ... (600 zeros) ... 010 ...

In the decimal expansion of this number the digit 1 is on the n!-th places. All other digits are zeros.
This means that the digit one is on the first, second, sixth, twenty-fourth, ... decimal place. Even
though this number is relatively easy to define, it does not fulfil any algebraic equation. This exactly
means that it is a transcendent number - it transcends the expressive power of the language of algebra.
(For details see Courant and Robbins 1941, p. 104-107.)

2.4 Analytical geometry

The polynomial, as introduced by the algebraists, was a purely symbolic object, without any
geometrical interpretation. In order to transcend the boundaries laid on the Euclidean geometry by the

16
three-dimensional space and to be able to form higher degrees of the unknown, the algebraists had to
resign from the possibility of any visual representation of their formalism. They knew how to
calculate with the polynomials, but they never associated any geometrical form with them. Even if the
algebraic results were interesting, the loss of any visual representation of the symbolic language was
regrettable. Fortunately this loss did not last long. In the 17th century the analytical geometry was
developed. In analytical geometry any polynomial, for instance x17 + 24x5 − 4x + 2, is an ordinary
curve. In this way all algebraic concepts such as root, degree, etc. acquire geometrical interpretation.
For instance the degree of a curve can be geometrically interpreted as the maximal number of its
intersections with a straight line. The idea to associate a curve to any algebraic polynomial resembles
in many aspects the Pythagorean idea of visualisation of the numbers, which associated geometrical
forms to arithmetical properties with the help of figurate numbers. In a way similar to the Pythagorean
visualisation of arithmetic, analytical geometry visualises algebra. In both cases we have to deal with
creation of a new iconic language, which incorporates some features of the particular symbolic
language.

a. logical power - reduction of geometry to algebra

Descartes published his Géométrie in 1637 as an appendix to his Discours de la méthode. We
base our exposition on the paper of Paolo Mancosu (Mancosu 1992). Descartes’ Géométrie comprises
three books. The first book Problems the construction of which requires only straight lines and
circles opens with a bold claim: „Any problem in geometry can easily be reduced to such a term that
a knowledge of the lengths of certain straight lines is sufficient for its construction.“ Descartes
showed that any ruler and compass construction is equivalent to a construction of the root of a second
degree equations. In order to show, how these roots can be constructed, he explains how the
arithmetical calculus is related to the operations of geometry. The main point of geometrical
interpretation of arithmetical operations is to overcome the problem of dimensionality, which limited
the previous geometrical work to a great extent. In ancient geometry, as well as in Viéte, the product
of two lines is interpreted as an area, and the product of three lines as volume. But there is no
corresponding interpretation for the product of four or more lines. The basic innovation introduced by
Descartes was to interpret the multiplication of line segments a and b not as the area of a rectangle
with the sides a and b, but as a line segment with the length equal to the arithmetical product ab. In
this way Descartes desubstantionalized the operation of multiplication. The product of two quantities,
which in Viéte gives rise to a quantity of higher order, lost its substantial meaning. The product is a
number and nothing more. Desubstantionalizing the multiplication Descartes defined the product of
any number of line segments, and so to transferred into iconic language the basic advantage of
symbolic language of algebra − its ability to form powers of any degree.
After this geometrical interpretation of arithmetical operations Descartes shows a general
strategy for solving all geometrical problems. It consists of three steps: naming, equating and
constructing. In the first step we assume that the problem is already solved, and give names to all the
lines which seem needed to solve the problem. In the second step we ignore the difference between

17
the known and the unknown, find the relation that holds between the lines, and express it in form of
algebraic equation. In the third step we construct the roots of the equation. Descartes concludes this
part of Géométrie asserting that all problems of classical geometry can be solved using this method.
Thus analytical geometry brought mathematical research in classical constructive geometry to an end.
It introduced universal analytical methods of algebra into geometry. An illustration of this can be
found in Courant and Robbins where they describe the construction of the decagon (Courant and
Robbins 1941, p. 122):

S x B

„Suppose that a regular decagon is inscribed in a circle with radius 1, and call its side x.
Since x will subtend an angle 36o at the centre of the circle, the other two angles of the large triangle
will each be 72o, and hence the dotted line which bisects the angle A divides triangle SAB into two
isosceles triangles, each with equal sides of length x. The radius of the circle is thus divided into two
segments, x and 1− x. Since SAB is similar to the smaller isosceles triangle, we have 1/x = x/(1 − x).
From this proportion we get the quadratic equation x2 + x − 1 = 0, the solution of which is x = ( 5 −
1)/2.“
To construct a line segment of this length is easy, as 5 is the diagonal of the rectangle with

the sides 1 and 2. Then we just mark off this length ten times as a chord of the circle and the decagon
is ready. We need not to memorise any construction trick. The constructive (geometrical) part of the
problem is trivial. The classical constructive geometry was difficult, because to construct any object a
specific procedure had to be remembered. This procedure was appropriate in this particular case only.
In analytical geometry we do not construct objects. Properties of an object are rewritten into algebraic
equations, these are solved via general algebraic methods, and only line segments with lengths
corresponding to the solutions of the equations, are constructed. For instance, instead of constructing
a regular decagon we obtain in this way the much easier problem of constructing a line segment the
length of which is ( 5 - 1)/2. Thus Descartes has brought the universal methods of algebra into
geometry.
Nevertheless, it is important to stress that this powerful method of reduction is a property of
the language. It is a general logical tool, later used also in the process of arithmetization of
mathematical analysis by Cauchy, Weierstrass and Dedekind. The particular form of reduction,
developed by Descartes in order to reduce geometry to algebra is therefore not so important. The
important thing is the reduction itself. In a similar way as the language of synthetic geometry has

18
brought the possibility to prove general theorems, and as the language of algebra brought the
possibility to prove modal predicates, the analytical geometry enriched the logical tools of
mathematics by the possibility to reduce one theory to another.

b. expressive power - ability to represent algebraic curves of any degree

Analytical geometry brought a new method of generating geometrical pictures. The
configuration is constructed point after point using a co-ordinate system. This is something
qualitatively new in comparison with Euclid. Euclid generated the picture (a term of the iconic
language of geometry) with ruler and compass. This means, Euclid has some basic „mechanical“
forms, which he locates on the paper. In contrast to this, analytical geometry breaks every
configuration into points and plots the independent points separately, point after point. A form can be
associated with every polynomial in this way. Descartes invented this new method of constructing
curves solving a problem stated by Pappus. He writes: „If then we should take successively an infinite
number of different values for the line y, we should obtain an infinite number of values for the line x,
and therefore an infinity of different points, such as C, by means of which the required curve could be
drawn.“ (Mancosu 1992, p. 89).
In this way, using the language of algebra, a much richer universe of forms is disclosed,
unknown to the Greeks. Looking back from the Cartesian point of view onto the Euclidean geometry
we can say that apart of some few exceptions (as the quadratrix of Hippias, the spiral of Archimedes,
the conchoid of Nicomedes, or the cisoid of Diocles (Heath 1921, p. 226, 230, 238, and 264)) the
whole of Euclidean geometry deals with quadratic curves only (i.e. curves, the equations of which are
of second degree). Problems, in which quantities of the third degree appeared, as for instance the
problem of the trisection of an angle or the problem of duplication of the cube, formed an obstacle for
the language of synthetic geometry. In order to overcome this obstacle, the mathematicians reached
for the above-mentioned curves, which strictly speaking, did not belong to the language of synthetic
geometry. So these problems represent the boundary of the expressive power of the language of
synthetic geometry.
From the analytical point of view the third roots are roots like others. There is nothing special
about them. The circumstance that they cannot be constructed with a ruler and compass is of
secondary interest for analytical geometry. The universe of the analytical geometry is qualitatively
richer, it contains qualitatively more curves, than the Euclidean universe. Nearly every important
mathematician of the 17th century came up with a new curve. Let us just mention Descartes’ folium,
Bernoulli’s lemniscate, the cardiod, the astroid, or the strophoid. So the expressive power of the new
language is higher. It is important to realise that inside of the analytical universe we can reconstruct a
region, which will correspond to the Euclidean. The Euclidean universe is the universe of quadratic
curves.
In algebra a formula represents the order of particular steps of a calculation. This corresponds
to construction of separate points of analytic curve. For each point we have to calculate the values of
its coordinates using algebraic formula. Nevertheless, analytical geometry goes further. By plotting

19
the independent points separately, point after point, a new form becomes visible. None of the separate
points itself gives rise to the form. Only if they are all together, we can see the form, in which we
recognise local maxima or minima, inflex points, convexity. The algebraic formula itself determines
only each single point, but putting them all together, this is the new step done by the analytical
geometry. If we gradually change one coordinate, and for each of its values we calculate the second
coordinate according to the algebraic formula, we obtain a curve. The curve discloses the dependence
between the two coordinates. This is not a functional dependence yet (i.e. dependence of a function
from its argument). The concept of function was introduced by Leibniz. In analytical geometry we
have just dependence between coordinates, what means the dependence is of geometrical nature.
Nevertheless, this geometrical way of representing dependence was an important step towards the
concept of the function itself.
Thus the geometrical visualisation of dependence as dependence between variables resembles
the Pythagorean visualisation of arithmetical properties using the figurate numbers. In a similar way
as the line segment of indefinite length, which the Pythagoreans used in their proofs, was a precursor
of the concept of variable, the dependence between variables, used in analytical geometry, is a
precursor of the concept of a function. The line segment of indefinite length as well as the dependence
between variables are part of the iconic languages, while the concepts of variable as well as function
are constituents of the symbolic language. Nevertheless, the role of the geometrical intermediate states
in the formation of the concepts of a variable or function is clearly visible.

c. explanatory power - ability to explain the casus irreducibilis

Analytical geometry explains why algebraic formulas lead to paradoxes in some cases. The
idea stems from Newton. If solving an algebraic equation means to determine the intersection points
of particular curve with the x-axis, the universal solvability of all equations would automatically
mean, that each curve would have to intersect this axis. This is clearly nonsense. Therefore there must
be a way, in which an algebraic equation does not give rise to intersection points. The appearance of
negative numbers under the sign of square root is a possibility to prevent an algebraic formula to have
a solution.
Thus the casus irreducibilis is not a failure of algebra or of the algebraist. On the contrary. As
the algebraic expressions determine analytical curves, the formulas giving the solutions of algebraic
equations must fail in some cases, to give the curves the necessary freedom. In this way the failure of
the formulas, which might look as a weakness of the algebraic language from a purely algebraic point
of view, is no weakness at all. Neither is it an exceptional case. It must be a systematic feature of all
algebraic formulas in order to give the analytical geometry the necessary freedom. Thus the language
of analytical geometry explains the failure of the language of algebra, which looked rather odd from a
purely algebraic point of view. The situation is not so simple in the case of Cardano’s casus
irreducibilis, as the particular cubic parabola has three intersection points with the x-axis. Thus here
we have to do with a more delicate question. Nevertheless, the basic fact, that negative numbers under
the sign of square root are something quite normal, is rather obvious from the point of view of

20
analytical geometry.
Here again we have to do with an explanation similar to the explanation of the insolubility of
some arithmetic problems, given by synthetic geometry. In both cases the geometrical language
disclosed the richness of possible situations responsible for the failure of particular symbolic
language. Thus these explanations are not examples of the skill of some mathematicians. They rather
disclose an epistemological feature of the language itself, namely its explanatory power. Who, when,
and under what circumstances discovered the explanatory power of particular language is a historical
question. But the explanatory power itself is an epistemological fact, requiring philosophical rather
than historical analysis.

d. integrative power - the mechanistic universe

Analytical geometry opens a possibility to reduce geometrical problems to algebraic ones and
to solve geometrical problems using the methods of algebra. Nevertheless, this reduction is just the
first step of a universal method. We can reduce physical problems to problems of mechanical motion,
the problems of mechanical motion can be further reduced to geometrical problems about the
trajectory of the motion, these geometrical problems can be reduced to problems of algebra and solved
using universal algebraic methods. Thus the language of analytical geometry made it possible to
develop a universal method of description of all phenomena of the universe - the mechanistic world-
view.
As it turned out, each step of this magnificent vision was wrong. As the theory of relativity
shows, it is impossible to reduce all physical phenomena to purely mechanical motion.
Electromagnetic phenomena cannot be explained in this way. But even if the mechanical reduction
would work, it would not help much, as the modern chaos theory shows us. For many simple
mechanical systems their motion is chaotic, and so their trajectory cannot be determined easily. But
even if this would work, we could not find algebraic formulas determining these curves, as most of
the motions in the universe have as their trajectory not algebraic but transcendent curves (exponential
or goniometric). But even if the reduction to algebra would work, it would not help much, as
according to Galois theory the solubility is rather an exceptional property among algebraic equations.
This is a fascinating feature of the Cartesian philosophy of science that each step of the
reduction failed. We believe that this failure shows the greatness of Descartes’ vision, because each
point of its failure taught us much about science and about the universe. For a philosophical vision of
the 17th century it is would be a distinction to fail on the Galois theory, on the deterministic chaos, or
on the theory of relativity. But to fail on all of them shows the greatness of Descartes’ discovery. If
we turn to contemporary theoretical physics, there still remains something from the Cartesian
reduction. The technical tools with the help of which Descartes wanted to realise his reduction (i.e.
curves and algebraic equations) were abandoned. Instead of them contemporary physics uses functions
and differential equations. Nevertheless, what remained unchanged is the reduction itself. From the
times of Descartes to do science means to reduce phenomena to linguistic expressions (curves or
functions, algebraic equations or differential ones). His vision of reduction remained the core of

21
science. This is perhaps the best example for what we mean by the integrative power of a language.
The role of mathematics in science is precisely this − to offer efficient tools for the integration of
different aspects of our experience.

e. logical boundaries - the quadrature of circle

The transcendence of π was demonstrated in 1882 by Lindemann. This proof can be found in
many textbooks (see Stewart 1989, p. 66). For our purposes it is sufficient to notice, that this result
means that ancient problem of the quadrature of the circle cannot be solved even with the methods of
analytical geometry.

f. expressive boundaries - transcendent curves

Soon after the discovery of analytical geometry it turned out, that the language of algebra is
too narrow to deal with all the phenomena, encountered in the world of analytical curves. First of all,
two kinds of curves are not algebraic, the exponential curves and the goniometric curves. They cannot
be given with the help of a polynomial. These curves transcend the language of the classical analytical
geometry in a similar way as the transcendent numbers transcend the language of algebra.

2.5 The differential and integral calculus

The language of algebra turned out to be unable to describe all the phenomena of the universe
of curves disclosed by Descartes. Therefore Leibniz and Newton independently developed a new
symbolic language. Many geometrical properties of analytical curves, such as length, curvature, or the
area surrounded by the curve, can be in a very elegant way expressed in this new language.

a. logical power - ability to solve the problem of quadrature

Even if analytical geometry brought a decisive progress in geometry, it was not able to solve
many geometrical problems. First of all two problems, the problem of tangents, consisting in the task
of finding a tangent to a given curve, and the problem of quadratures, consisting in the task of finding
the area surrounded by a given curve, gave birth to technical methods exceeding the language of
algebra. These new technical methods developed in the context of the problem of quadratures were
based on the idea of dividing the given object into infinitely many infinitely small parts. These parts
could be, after an appropriate transformation, put together again so that a new configuration would be
formed, the area or volume of which, could be determined more easily. Kepler, Cavalieri, and
Torricelli were great masters of these new infinitesimal methods. They found the areas or volumes of
many geometrical configurations. But for every configuration it was necessary to find a special trick
for dividing it into infinitesimal parts and then for summing these parts again. In some cases a regular
division led to success, in others the parts had to obey some special rule. Every trick worked only for
the particular object, or for some similar ones in the best case. But it lacked an universal language,
enabling discovery of more general techniques.
The new language was the well-known differential and integral calculus. The basic idea of
Leibniz was in many respects similar to the regula della cosa of the algebraists. He too created a

22
symbolic language, enabling to manipulate with letters (more precisely groups of letters, namely the
differentials dx, dy, etc.) as with infinitesimals. The differential and integral calculus is, like algebra
or arithmetic, a symbolic language. It is based on formal manipulations with linear strings of symbols,
and it gives in quick and elegant way answers to questions, arising in the universe of analytical
curves.
The central point of differential and integral calculus is the connection between the definite
and the indefinite integral discovered by Newton and Leibniz:
b

∫ f ( x )dx = F ( b) − F ( a )
a
where F ( x )′ = f ( x ) .

(1)
This formula makes it possible to transfer the difficult geometrical problem of quadrature (expressed
by the definite integral and consisting in the division of particular area beneath the curve f(x) into
infinitesimally small parts, rearrange them so that it will be possible to determine their sum) into a
much easier calculative problem of formal integration (i.e. to find to a given function f(x) its
primitive function F(x) such that F ( x ) ′ = f ( x ) ). Really, the formula (1) makes it possible instead

of calculating the area given by the definite integral ∫ f ( x )dx , first to find the primitive function
a

F(x) and then to calculate just the difference F ( b ) − F ( a ) . Therefore if we wish for instance to
calculate the area enclosed beneath the curve y = x3 between the boundaries x = 3 and x = 5, it is not
necessary to calculate the integral
5

∫x dx .
3

1 4
It is sufficient to take the function x , which is the primitive function to x3 and to calculate the
4

54 34 625 − 81
difference ( − ) = = 136. Thus instead of complicated infinitesimal techniques it
4 4 4
is sufficient to perform some elementary operations. In most cases to find the primitive function is not
so easy, nevertheless the whole calculation is even then much more simple than geometrical methods
developed by Kepler or Cavalieri (see Edwards 1979, pp. 99-109).
The basic epistemological question is what made this fundamental progress in the solution of

the problem of quadrature possible? The basic idea was to consider a definite integral ∫ f ( x )dx , in
a

which the upper boundary is not fixed, but it varies with time. Let us imagine the area beneath the
curve f(x), closed between x = a and x = t, and let the parameter t grow gradually. This means that the
right side of the considered figure will move. The question is what will be the velocity of growth of

23
its area. It is not difficult to see that this velocity is
t +∆t

∫ f ( x )dx ≈ f ( t ). ∆t
= f(t) .
t
∆t
∆t
(2)
That means that the velocity of growth of the area beneath the curve in a given moment t is equal to
the value of the function f(t) in this moment. Thus instead of calculating the area beneath the curve it
is sufficient to find a function such that its velocity of growth is precisely f(t). The velocity is given
by derivation. Thus it is sufficient to find a function F(t), the derivative of which will be f(t). This fact
is represented by the formula (1).
We see that the decisive step in discovery of the formula (1) was the new view on the area
beneath a curve. Newton looked onto this geometrical quantity as a function, namely a function of the
upper boundary of the figure, the area of which we calculate. It is important to notice, that already the
concept of area beneath a curve is given with the help of a function, namely the function f(x) which
determines the particular curve. This is the way, how functions are present in the iconic language of
analytical geometry. What is new in the calculus is that the area, which is already determined with the
help of the function f(x), is considered to be the function of the upper boundary. Thus we have to do
here with implicit concept of a function of a function, thus of a function of the second degree, as
Frege described it in the quoted passage. To be able to deal with the second order functions (as
integrals or derivatives), the ordinary functions had to become explicit. The introduction of an explicit
notation for functions is thus a characteristic feature of the symbolic language of the differential and
integral calculus, which is fully parallel to the introduction of explicit notation for variables, which
occurred on the previous stage of the symbolic language, in algebra.

b. expressive power - ability to represent transcendent functions

The differential and integral calculus is a language of higher expressive power, than was the
language of algebra on which the analytical geometry was based. Functions, like ln(x), cos(x), or
elliptic functions, are not polynomial. They transcend the boundaries of the language of algebra. For
the differential and integral calculus they present no problem. For instance the logarithmic function
can be expressed in a form of infinite series as
x 2 x3 x4 x5 x6 x7
ln ( 1 + x ) = x − + − + − + − +... ,
2 3 4 5 6 7

or of an integral
x
1
ln( 1 + x ) = ∫ dt .
0
1+t

The logarithmic function is a rather simple example. It would be possible to present some more
complicated examples as for instance Euler’s Γ-function, Bessel’s functions, hyperbolic functions,

24
Riemann’s ζ-function, elliptic functions as well as a number of special functions occurring in
physics or in technical applications. The differential and integral calculus is a symbolic language,
enabling to express many functions, absolutely inconceivable in the framework of the language of
algebra. Nevertheless, it is important to notice that the definitions of these new functions use infinite
series, derivatives, or integrals, that is, functions of the second degree.
The language of algebra can be embedded in the new language of the calculus. If in an infinite
series (expressing for instance ln(x)) we restrict ourselves only to a finite number of initial terms, we
obtain a polynomial. In the universe of polynomials derivation and integration can be defined by
explicit rules, and this universe is closed under these rules. We can consider these restricted
operations (prescribing derivative and integral to any polynomial) as new unary algebraic operations.
So we can embed the language of algebra into that of mathematical analysis.

c. explanatory power - ability to explain the insolubility of the quadrature of the circle
Algebra can explain why the problem of trisection of an angle is insoluble. Nevertheless,
algebra is not able to explain why nobody succeeded in solving the quadrature of the circle. As we
already mentioned, the reason for this consists in the transcendent nature of the number π. In 1873
Hermite proved the transcendence of the number e and in 1882 Lindemann succeeded, using the ideas
of Hermite, to prove the transcendence of π. These proofs are based on the language of the differential
and integral calculus.

d. integrative power - mathematical physics

Modern science and first of all modern theoretical physics is based on the language of
differential and integral calculus. Even if the idea of universal mathematical description of nature
stemmed from Descartes, his technical tools based on the language of analytical geometry did not
have sufficient integrative power to accomplish his project. The Cartesian polynomials had to be
replaced by functions and algebraic equations by differential equations to integrate all natural
phenomena into the universal picture.
Descartes expressed the unity of nature on a metaphysical level. One of the basic purposes of
his mechanistic world-view was to unify all natural phenomena. Thus his metaphysics had to fulfil the
function which his formal language could not fulfil - to integrate nature into a unified theory. Modern
mathematical physics does not require any special metaphysical position in order to see the unity of
nature. The unity of contemporary physics is fully formal. It is provided by the language and not by
metaphysics. Maxwell’s equations of electromagnetic field remained valid also after we abandoned
the theory of ether. Ether served for Maxwell as an ontological foundation of his theory, but later it
became obvious, that the theory can do well without these or any other ontological basis. This shows
that the unity of contemporary physics is a formal unity independent of any ontology. The source of it
is the integrative power of the language of differential and integral calculus.

e. logical boundaries - crisis of foundations

The first serious criticism of the foundations of the differential and integral calculus appeared

25
soon after its discovery. In his famous book The Analyst, or a discourse addressed to an Infidel
Mathematician, which appeared in Dublin in 1734, George Berkeley expressed the view, that the
whole calculus is based on a series of errors. He criticised the way of reasoning, typical in differential
and integral calculus, by which one makes calculations with some quantity assuming that its value is
different from zero (in order to be able to divide by it) and at the end one equates this quantity with
zero. Berkeley correctly stressed, that if a quantity is zero at the end of some calculation, it must have
been zero also at its beginning. Thus the whole reasoning is incorrect. According to Berkeley the
correct results of calculus are due to compensation of different errors.
Various attempts were presented to rebuild the foundations of the calculus to save it from
Berkeley’s criticism. The perhaps most important of them were Cauchy’s introduction of the concept
of limit in 1820’s and the ε−δ analysis developed by Weierstrass in 1860-ties. They both relied on the
intuitive concept of continuum and so the next step was to develop the foundations of the continuum.
This was done in 1870’s by Dedekind and Cantor. Nevertheless, the theories of Cantor and of
Dedekind were not absolutely sound. They were based on intuitive concept of natural numbers. Thus
in the 1890’s Frege, Dedekind, and Peano independently developed alternative theories of natural
numbers. For a short time it seemed, that the project of rigorisation of the calculus, in the direction
started by Cauchy, reached its definitive and successful end. But soon emerged the well-known
paradoxes and the whole edifice of the foundations of mathematics crushed. Russell informed Frege
about his discovery of a paradox in Frege’s theory in a letter in 1901. Frege realised immediately, that
the same paradox can be formulated also in the system of Dedekind, and it is not difficult to see, that
the system of Peano has a similar fault (see Gillies 1982). This shows that the paradox is not the
consequence of some mistake of the particular author. It is not probable, that Frege, Dedekind, and
Peano would make the same mistake. The conceptual foundations of their systems are so different,
that the only explanation of the parallel occurrence of the same paradox in all of them can be
explained only as a feature of the language itself. Thus the logical paradoxes are not individual
mistakes, but they rather represent the logical boundaries of the language of differential and integral
calculus. This language introduced functions of the second degree, and the paradoxes stem exactly
from this source. They are caused by the careless use of the second order functions and predicates. In
this respect these paradoxes are analogous to the paradoxes appearing in algebra. In algebra the main
innovation of the language was the introduction of the (implicit) first order functions (as square root,
etc.) enabling express the solution of a problem in form of a formula. The paradoxes in algebra (the
casus irreducibilis) are caused by the careless use of these first order functions.

f. expressive boundaries - fractals

The successes of the differential and integral calculus justified the belief, that all functions can
be described with the help of infinite series, integrals, derivatives, and differential equations.
Therefore it was a real surprise when the first „monstrous“ functions started to appear. Bolzano’s
function does not have a derivative at any point, Dirichlet’s function is discontinuous at each point,
and Peano’s function fills the unite square. These functions gave rise to a considerable refinement of

26
the basic concepts of differential and integral calculus. In the course of they study it turned out, that
the methods of the calculus can be applied only to a rather narrow class of „decent“ functions. The
rest of the functions lie beyond the expressive boundaries of the language of the differential and
integral calculus.

2.6 Iterative geometry

Differential and integral calculus were born originally in very close connection to analytical
geometry. Perhaps that was one of the reasons why mathematicians for a long time considered
Descartes’ way of generating curves (i.e. generating them point by point according to a formula) to be
the correct way of visualising the universe of mathematical analysis. They thought, that it would be
enough to widen the realm of formulas used in the process of generation, and to accept also infinite
series, integrals or perhaps other kinds of analytical expressions instead of the polynomials. But it
turned out, that it is not enough to change the realm of formulas, the whole way of generating curves
has to be changed.
Bernard Bolzano was the first, to come up with a strange function, which differed from
anything we were used to in the universe of curves. He found a curve, which did not have a tangent at
any point. At the beginning such curves were considered as „pathological cases“. Mathematicians
continued to regard the universe of analytical curves to be the embodiment of the spirit of
mathematical analysis. In the second half of the 19th century there appeared an endless string of such
„pathological functions“. Nearly every important mathematician came up with some strange curiosity.
Perhaps the most surprising is Peano’s curve filling the unit square.
When such strange examples had been produced in sufficient number, the study of them led to
a new mathematical discipline - the theory of functions of real variable. It turned out, that all these
„pathological cases“ have something in common. They are generated in a way very different from the
one introduced by analytical geometry. They are generated not point after point according a formula,
but in an infinite iterative process. That means, that in contrast to the analytic case, where a point,
after it was once generated, remained unchanged during the whole process of further generation, in
this new method in each step of iteration the whole curve undergoes change. There is no part of the
curve taken out from further iterations (with perhaps the exception of some isolated points). Thus the
generation of the curve is a second degree procedure. At each step the curve represents a function and
the process of iterations is a process, which changes these (first order) functions. Thus it works on the
second degree. Therefore in a similar way as we described analytical geometry as a visualisation of
the symbolic language of algebra, we can describe the iterative geometry as visualisation of the
symbolic language of differential and integral calculus.
A new rebirth of iterative geometry appeared in recent time, thanks to computers. Computers
can implement the iterative process, which generates the curve, and we can see beautiful patterns
known as fractals on the screen. If we just look onto the screen, it is clear, that to try to generate these
pictures using the Cartesian method is hopeless. We don’t need much fantasy in order to see that the

27
universe of the iteratively generated forms is much richer than the Cartesian. But again we can
discriminate the universe of analytical geometry inside the iterative universe, just as we discriminated
the Euclidean universe in the Cartesian. It is enough to restrict our iterative process to convergence,
which will be uniform together with enough derivatives. Thus the universe of analytical geometry is
the „smooth part“ of the universe of the iterative geometry in a similar way as the Euclidean universe
is the „quadratic part“ of the universe of the analytical geometry.
The curves generated with the help of the iterative processes have many common features,
which were discovered during the 20th century. One of them is that their Hausdorff dimension is not a
natural number as we are used in ordinary geometry (dimension 1 for curves, 2 for surfaces, and 3 for
solids) but a fraction. The dimensions of typical iterative objects are 0,6309 for Cantor’s
discontinuum, 1,2619 for Koch’s curve, 1,5850 for Sierpinski’s carpet (see Peitgen, Jürgens, and
Saupe 1992). Therefore they are called also fractals, i.e. objects with fractal dimension. In his book
Fractal Geometry of Nature Benoit Mandelbrot pointed out the fact that many natural shapes, as those
of mountains or coastlines, resemble fractals in many respects. It turned out, that fractals play an
important role in turbulence and chaotic motions. So iterative geometry stopped being a “freak of
mathematicians” or „pathological cases“ and started to be regarded as a separate universe of forms.
Thus beside the universe of synthetic geometry which combines its pictures from parts of straight
lines and arcs of circles, and the universe of analytical geometry, which generates its pictures point by
point according to some analytic expression, the universe of iterative geometry represents a third kind
of iconic language, describing form. The terms (well formed expressions) of this language are fractals.
To describe the logical, expressive, explanatory, and integrative power as well as logical and
expressive boundaries of this new language is a difficult question requiring further investigations.

2.7 Predicate calculus

The discovery of the first fractals in the 19th century was a real shock. It turned out, that
many theorems, which were „proven“ by the most outstanding mathematicians, were not true. Their
„proofs“ were not real proofs, because they relied on some intuitive assumptions, which turned out to
be erroneous. Gottlob Frege undertook the task to free mathematics from such unpleasant surprises
and to create a formal notation, which would make it possible to determine, whether a proof uses only
the assumptions, which were clearly stated, or it contains also some unnoticed implicit ones. This
notation is called the predicate calculus. At present the predicate calculus is the last symbolic
language. While arithmetic is based on manipulation with numbers, algebra on manipulation with
things (regula della cosa) and mathematical analysis on manipulations with infinitesimals
(differentials), Frege created a calculus which formally reconstructs our use of concepts. Frege called
his new language Begriffsschrift what means "conceptual script".
As it is the last language based on formal manipulation with symbols, we cannot characterise
its logical, expressive, explanatory, or integrative power. No other, stronger language is available,
with which we could determine the power of predicate calculus. All we can say from the point of view

28
of the present stage of mathematics the power of the predicate calculus is total, i.e. it is the language
in which contemporary mathematics formulates its theories. All other languages, as for instance
elementary arithmetic, algebra or mathematical analysis, are parts of this language. Precisely this
imbedding of all symbolic languages into the predicate calculus made it possible to compare their
logical and expressive power.

2.8 Set theory

We have arrived to the last language created during the long history of mathematics. This
language fully prevailed in mathematics. In set theory the mathematicians formulate the whole of
mathematics, from arithmetic to the theory of dynamic systems. But what is this set theory? It is not
easy to answer this question of basic epistemological importance.
Set theory, at least according to the intentions of its creator Georg Cantor, ought to be a
transfinite arithmetic, i.e. a theory of calculation with infinitely large „numbers“. If we slightly widen
this approach, we could say, that set theory is a universal theory of the infinite. On the other side, with
equal justification, set theory could be regarded as a branch of geometry. In the spirit of the Erlangen
program set theory can be defined as a study of invariants of bijective mappings. Of course, here we
violate slightly Klein’s conception, because the bijective transformations of any set don’t form a
continuous group. But nevertheless, this approach sheds light onto many aspects of the set theory. It
shows that beneath the group of diffeomorphisms, which are the basis of topology, there is a more
general system of transformations, which captures even more basic properties of space. Also this
direction of thoughts can be found in the works of Cantor. His discovery that the cardinality of the set
of points forming a square is equal to the cardinality of the set forming a side of this square shows
that the invariants of set theory are more fundamental than those of topology. So we can regard set
theory as a further step in the development of geometry, a step beyond topology. Similarly great is the
overlapping of set theory and the theory of real functions. It was the problem of convergence of the
Fourier series, which led Cantor to his fundamental discoveries. And the first variant of the set theory
was a theory of point sets, which is the basis of real analysis.
In this way we have found three answers to the question what is set theory. It is the universal
theory of infinity, the most basic layer of invariants of space, and a universal theory of the continuum.
Each of these answers is correct, at least to some degree, but none of them is fully satisfactory. So we
have to try to get behind these three answers, deepen our question in the epistemological direction,
and ask what enabled the set theory to become all that, which was said above. We think, that set
theory is first of all some kind of „visualisation“ of Frege’s predicate logic. That means, that set
theory is for predicate calculus, what iterative geometry is for mathematical analysis, analytical
geometry for algebra and synthetic geometry for arithmetic. Analytical geometry gave better insight
into the properties of polynomials with the help of their geometrical representation. In a similar way
set theory offered a strong tool in the form of set theoretical models, providing us better insight into
many properties of formal axiomatic theories. For instance the independence of axioms can be easily

29
shown with such models. Thus set theory is a „visualisation“ of logic. A set is not a „symbol“ - it
does not denotate, it is an „icon“ it represents. In a similar way as in the case of predicate calculus,
also for set theory the logical, expressive, explanatory, and integrative powers are total, and so they
defy definition.

3 Concluding remarks
Frege’s analysis quoted in the introduction of this paper is purely logical hovering in an
epistemological vacuum. To understand the internal dynamics of the above-mentioned transitions, to
give their epistemological and not only purely logical description, is, we are convinced, possible only
in this contraposition of symbolic and iconic language. So without understanding the logical structure
of the iconic language, we will be able to write the epistemology of algebra or arithmetic, but we will
not be able to explain the relation of these disciplines. It will not be possible, because this relation is
mediated through geometry. So Frege’s phrases, that „then they went on“, „the next higher level“, or
„the next step forward“, which, from the logical point of view are perhaps correct, are misleading
from the epistemological point of view. All the transitions in the development of the symbolic
language, described by Frege, were mediated through iconic intermediate stages. A logician can take
the liberty to ignore them, but an epistemologist should not.
The comparison with Frege explains more clearly the nature of changes described in our
paper. It is the development of generality of language, which from the logical point of view consists
of the order of variables used in the theory. Elementary arithmetic used only constants, algebra used
variables and implicitly also functions of the first order, analysis used functions of the second order
ELEMENTARY
and predicate calculus made it possible to use functions of any order. Nevertheless, we have seen that
ARITHMETIC
the intermediate geometrical stages use variables and functions in a different way. A segment of an
indefinite length is not a variable, because it does not denotate but rather represents. In a similar way
SYNTHETIC
an arbitrary curve is not a function (i.e. a variable of theGEOMETRY
second order). To understand better the
logical properties of these intermediate geometrical languages remains an open problem.
The following diagram represents the basic eight formal languages, which were formed in the
ALGEBRA
course of the development of mathematics. The transitions between the following ones represent
epistemic ruptures, which consisted in the disclosing of a new universe. Four of them (1, 3, 5 and 7)
ANALYTIC
consisted in the construction of a new iconic language, while the remaining three brought a new
GEOMETRY
symbolic language.

DIFF. AND INT. 1

CALCULUS

FRACTAL
GEOMETRY
2

PREDICATE
CALCULUS
30
SET
THEORY
3

It is an interesting fact, that five of the seven ruptures described in this paper were discussed
in the collection Revolutions in Mathematics (the Pythagoreans by J. Dauben, Descartes by P.
Mancosu, Leibniz by E. Grosholz, Frege by D. Gillies and Cantor by J. Dauben). The only two
missing cases were the discovery of algebra and of fractal geometry. This shows a remarkable
agreement. Not that the authors would agree, that all these changes are revolutionary. Nevertheless,
there is an implicit coherence that these are the most important cases, which should be discussed. The
analysis presented in our paper explains more precisely the common features of these changes and
thus also the implicit agreement of the historians.
We do not want to rise the question, whether these changes are revolutionary or not. Instead
we formulate an epistemological principle, which we would like to call the principle of
epistemological coherence. According to it, if we declare one of the seven above-mentioned ruptures
to be revolutionary, we should declare as revolutionary also the six others. We think Joseph Dauben
used, at least in an implicit way, this principle when he declared as revolutionary the Pythagorean as
well as Cantor. This principle should shift the discussions about the revolutionary nature of some
particular historical changes towards the effort to understand the general pattern of changes, which we
are prepared to declare as revolutionary.

ACKNOWLEDGEMENTS
I would like to express my thanks to Donald Gillies for critical reading of the manuscript and
for valuable suggestions and to Peter Bugár for help with the English language. The financial support
of the GRANT 1/4310/97 is acknowledged.

31
References
Boyer, C. and Merzbach, U.: (1989): A History of Mathematics, second edition, John Wiley, New
York
Courant, R.; Robbins, H. (1941): What is mathematics? New York, Oxford University Press, 1978
Dauben, J. (1979): Georg Cantor, His mathematics and Philosophy of the Infinite Princeton
University Press, Princeton, New Jersey
Dauben, J. (1984): Conceptual revolutions and the history of mathematics: two studies in the growth
of knowledge, in: Transformation and tradition in the sciences, Essays in honour of I.
Bernard Cohen, ed. E. Mendelsohn. Reprinted in Gillies 1992a
Edwards, C. H. (1979): The Historical Development of the Calculus, Springer Verlag, New York
Euclid (c. 300 B. C.): The Elements, Translated with introduction and commentary by Sir Thomas L.
Heath, Dover reprint of the second edition, New York
Fouvel, J. and Gray, J. (1987): The History of Mathematics: A Reader, Macmillan, London
Frege, G. (1879): Begriffsschrift. Hildesheim, Georg Olms Verlag, 1993
Frege, G. (1891): Funktion und Begriff. in Frege, G.: Funktion, Begriff, Bedeutung. Göttingen,
Vandenhoec & Ruprecht, 1989, pp. 17-39 English translation in: Translations from the
Philosophical Writings of Gottlob Frege, edited by P. Geach and M. Black, Basil Blackwell,
Oxford 1952, pp. 21-42
Gillies, D. (1982): Frege, Dedekind, and Peano on the Foundations of Arithmetic, Van Gorcum,
Assen
Gillies, D. (ed.) (1992a): Revolutions in Mathematics, Clarendon Press Oxford
Gillies, D. (1992b): The Fregean revolution in Logic, in: Gillies 1992a
Gray, J. (1979): Ideas of Space, Euclidean, Non-Euclidean, and Relativistic, Clarendon Press Oxford
Greaves, M., T. (1997): The Philosophical Status of Diagrams, PhD Thesis, Stanford University
Grosholz, E. (1992): Was Leibniz a mathematical revolutionary? in: Gillies 1992a
Heath, T. (1921): A History of Greek Mathematics, Clarendon Press, Oxford
Kline, M. (1972): Mathematical Thought from Ancient to Modern Times, Oxford University Press
Knorr, W. (1986): The Ancient Tradition of Geometric Problems, Birkhäuser, Boston
Kolman, A. (1961): Dìjiny matematiky ve starovìku. Praha, Academie 1968
Mancosu, P. (1992): Descartes’ Géometrie and revolutions in mathematics, in: Gillies 1992a
Mandelbrot, B. (1977): Die fraktale Geometrie der Natur. Basel, Birkhäuser Verlag 1991
Peitgen, H. O., Jürgens, H. a Saupe, J. (1992): Chaos and Fractals. Springer, New York.
Stewart, I. (1989): Galois theory, Chapman and Hall, London
van der Waerden, B. L. (1985): A History of Algebra, Springer, Berlin

32