On The Representation of Curves in Descartes' Géométrie

On the Representation of Curves in
Descartes' Gdomdtrie
H. J. M. Bos
Communicated by J. R. RAVE'rZ
1. Introduction. Curves and their representation in 17 th century mathematics
1.I From antiquity until the beginning of the seventeenth century the collec-
tion of plane curves known to mathematicians did not change. It consisted in
the conic sections, some higher algebraic curves such as the conchoid of NIC0-
MEDES and the cissoid of Dt0CLES, and a few transcendental curves the most
important of which were the Archimedean spiral and the quadratrix of DIN0-
STRArtrS. There was a marked change in this situation in the seventeenth century.
Over a short period mathematicians vastly expanded the realm of curves open to
mathematical treatment. Through the new analytic geometry o f FERMAT and
DESCARTES the collection of mathematical curves came to include all algebraic
curves, that is, all curves whose equation in rectilinear coordinates involves only
the algebraic operations + , --, × , - and ~ / - ( k > 1, integer). The collection
of transcendental curves, that is curves which do not admit an equation as above,
was enlarged as well. The cycloid appeared on the mathematical scene around
1630, and the logarithmic curve in the 1660's. After that mathematicians encoun-
tered many more curves that depended algebraically on these two fundamental
transcendentals. Most of these curves occurred as solutions of inverse tangent
problems°
The new curves, like the earlier ones, could play three different roles. They
could be an object of study, they could be a means to solve a problem and they
could themselves be the solution of a problem. Thus PASCAL, in his famous
challenge to mathematicians of 1658, proposed the cycloid as an object of study
and suggested that its area, the areas and centres of gravity of its segments and
the contents and centres of gravity of solids arising by rotation of its segments
be determined.
The "Cartesian parabola", introduced by DESCARTESin his Gdomdtrie, was
an example of a new curve that served as a means to solve a problem; the curve
was used in the geometrical construction of the roots of equations of 5th and 6 th
degrees (eft 5.2).
296 H.J.M. Bos
Finally, the curve with modern equation

_L
ae ~=a--y+x (I;1)
is an example of a (transcende_ntal) curve which originated as the solution of a
problem, namely the famous problem of DEBEAtmE (1638) ( c f note 31).
1.2 The enormous increase in the number of curves forced 17th-century mathe-
maticians to think about how new curves could be introduced, described or de-
fined. If the curves were to serve as an object of study, a means of solving a problem
or as a solution, they had to be or to become known. In the previous period this
was no problem since all curves were known and mathematicians could refer to
any of them by name (ellipse, conchoid, spiral etc.), adding~ if necessary, the values
of their parameters.
But when is a new curve sufficiently known ? Seventeenth-century mathemati-
cians did not have a uniform definition of the concept of curve (nor apparently
did they feel the need for such a definition) and therefore they had no standard
form for specifying the curves they had in mind. In fact, there were many ways
of specifying curves. One could, for instance, indicate how points on the curve
could be constructed, one could describe a machine by which the curve could be
traced, and (after analytic geometry had been introduced) one could give the
equation of the curve. Some of these ways of describing curves were considered
satisfactory, others less so, some not at all.
I shall use the term "representations of curves" to mean ways of specifying
curves which were thought to make the curves sufficiently known. This term was
not used in the 17t~ century with that meaning; there was no term with that mean-
ing then. Nevertheless, mathematicians did use the term "construction of curves"
which has almost the same meaning but is more restricted.
The different ways in which curves were specified in 17th century mathematics,
the preferences that mathematicians expressed for certain of them and the reason
given for these preferences form an important and interesting theme for historical
study° These ways and preferences influenced the direction in which mathematics
developed° Until now historians of mathematics have been hardly aware of this
theme, mainly because a too rapid translation of 17th-century mathematical argu-
ments into modern analytical symbolism has obscured these aspects of the treat-
ment of curves° The subject is also of more general interest because it touches
on a wider mathematical, or perhaps metamathematical, question, namely, when
is a mathematical entity "known" or when is a problem "solved"?
1o3 In this study I shall deal in particular with the representation of curves
in DESCARTES' Gdomdtrie 1. At one stroke the Gdomdtrie brought all the algebraic
i R. DESCARTES,La Ge~omEtrie, one of the essays in his Discours de la mdthode pour

bien conduire sa raison, et chercher la u~rit~ clans les sciences, phts la Dioptrique, les
Meteores et la Geometrie qui sont des essais de cete methode, Leiden, 1637. In references
to the G~om~trie I use the abbreviation G and I shall use the page numbers of the original
edition (pp. 297--413). The original text is easily accessible in The geometry o f Rend
Descartes with a facsimile o f the first edition (tr. and ed. by D. E. Sirra & M. L. LA-
Curves in Descartes' Gdomdtrie 297
curves into focus. But, as historians o f mathematics have remarked 2 (with some
surprise), DESCARTES did not consider the equation to be a sufficient representa-
tion o f the curve; he used other kinds of representation instead.
Moreover, DESCARTES introduced a sharp distinction between admissible and
inadmissible curves. The first he called "geometrical" the other "mechanical".
The "geometrical" curves are what we now call algebraic curves (although DES-
CARVES did not explicitly say as much in the Gdomdtrie, this can be inferred from
what he did state); the "mechanical" curves are those which are now termed
transcendental curves. But because DESCARTES did not consider the equation a
sufficient representation o f the curve, he could not establish any distinction be-
tween geometrical and non-geometrical curves on the basis o f their equations; he
had to reason about it on the basis o f representations of curves which he did
find acceptable. The acceptability of representations of curves is therefore a
crucial concept in the Gdom~trie.
DESCARTES' distinction between '"geometrical" and "mechanical'., curves
provided a serious issue in seventeenth-century matheniatics. The increasing
interest in transcendental curves (curves therefore that to DESCARTES were not
admissible in geometry) forced mathematicians to use methods other than those
expounded in the Gdomdtrie and to take up a position with respect to the question
of how far such curves could be considered "geometrical" or somehow admissible.
Again, this question could only be dealt with in terms o f the representation of
these curves, and several of the representations used in these debates had been used
in DESCARTES' Gdom~trie.
1.4 The complicated structure o f the G~omdtrie, the different roles o f curves
in it, and DESCARTES' different criteria for geometrical acceptability o f curves
have not yet been satisfactorily unravelled 3. A detailed analysis of the roles of
THAM), New York (Dover) 1954. In my translations of texts from the G~om~trie I have
taken the English of SMrrI-I & LAX'l-tAMas my starting point. However, their translation
is very free and often unreliable, so in many cases I have had to modify it. The edition
of the G~om~trie in the (Euvres de Descartes (eds. C. ADAM & P. TANNERY,Paris 1897-
1913) vol. 6, pp. 367-485 also indicates the page numbers of the original. I shall use
the abbreviation A.T. to refer to the Oeuvres.
2 See, for instance, C. B. BORER, History of analytic geometry (New York 1956)
p. 88 and p. 102; see also M. S. MAI-IOrCEY, "Descartes: mathematics and physics",
Dictionary of scientific biography (ed. C. C. GILLISPIX,New York 197010 vol. 4 (1971),
pp. 55-61, footnote 7.
3 There have been many studies on DESCARTES'G~omdtrie. Most of~these, however,
are unsatisfactory as far as the questions I discuss are concerned because they seek to
answer the unfruitful question as to whether DESCARTESdid or did not invent analytic
geometry. The best source for the actual contents of the G~om~trie is the G~om~trie
itself. The best summary of its intention, its development and its place within DES-
CAR~S' mathematics is still G. MILHAtlD, Descartes savant, Pads 1921. J, ITARD'S La
g~omdtrie de Descartes (Conferences du Palais de la D~couverte, s&ie D, nr 39) Paris
1956 is very penetrating but als0 very dense. I have found A. G. MOLLArCD'S"Shifting
the foundations, Descartes' transformation of ancient geometry", Hist. Math. 3 (1976),
pp. 21-49 very helpful since it discusses the history of the concepts of construction and
classification of curves in antiquity and DESCARTES'opinions on these classical ideas.
298 H..l.M. Bos
curves and their representation in the G~omdtrie sheds new light on the structure
of DESCARTES' book, on its underlying programme and on the earlier development
of DESCARTES' ideas about geometry. There is a conflict in the Gdomdtrie between
geometrical and algebraic, methods of definition and criteria of acceptability.
This conflict reflects a break in the development of DESCARTES' thought about
geometry. In an early phase DESCARTESconsidered that the aim of geometry was
to construct solutions of geometrical problems by means of curves traced by certain
instruments; the instruments served as acceptable generalizations of ruler and
compass. He tried to find new constructions in this way and to classify them.
About 1630 that plan seemed to stagnate and DESCARTESalso became fully aware
of the power of algebraic methods. He then changed his programme. Algebra
became the dominant tool, both for the solution of problems and for the classifi-
cation of curves. But DESCARTEScontinued to believe in the principle of geometrical
construction by means of curves traceable by instruments. As a result, there are
conflicting elements in the G~omdtrie. "
I shall show that it was impossible for DESCARTESto keep strictly to his earlier
programme which was based on the use of instruments, But it was also impossible
for him to work out a fully algebraic programme. If he had kept to his earlier
plans, he would have lost the advantages of algebra; if he had adopted a fully
algebraic approach, he could no longer have claimed that he was doing geometry.
The contradictions in the G~om~trie were indeed unavoidable. In order to under-
stand DESCARTES' great contribution to geometry and algebra it is necessary
to make these contradictions explicit and to explain how they influenced the struc-
~ r e of the G~om~trie.
The later synthesis of algebraic and geometrical methods into what is now
called analytic geometry was possible only because later mathematicians were
n o t aware of (or forgot) the programmatic problems with which DESCARTES
had struggled.
2o The problem of Pappus
2.1 In his Gdomdtrie DESCARTES expounded a new programme for dealing

with geometrical problems, and he used one problem as the key example: the
problem of PA~'PUS. I shall explain DESCARTES' programme in Section 3, but
first I shall discuss the problem of PAPPUS and DESCARTES' solution of it. That
discussion may illustrate the sort of geometrical problems which DESCARTES
had in mind and explain the roles of geometrical constructions, curves and al-
gebraic calculations in the Gdomdtrieo
As far as the concepts of curves and constructions in DESCARTES'G~omdtrie are concerned
one can compare J. VUILLEMIN)Mathdmatiques et mdtaphysique chez Descartes, Paris
1960 (in particular Ch. III "De la classification cartrsienne des courbes", pp° 77ff.),
G°-G. GRANGER, Essai d'une philosophie du style, Paris 1968 (in particular Ch. III
"Style Cart~sien, style Argue~sien" pp. 43-70), and .to DHOMBI~S, Nombre, inesure et
continu, dpist~mologie et histoire, Paris 1978 (in particular the section pp. 134-143).
My interpretation of the Gdom~trie differs in several points from the interpretations
given in these articles.
Curves in Descartes' G~om~trie 299
In explaining the problem of PAPPUS4 I shall use symbols for the lines, dis-
tances and numbers. DESCAR~S presented the problem in prose accompanied by
drawings. Let (see Figure 1) a number of lines L~ be given in the plane, and let
tp~ be fixed angles.
J
t~ J
/ d~ = y ~ "- - "~/~'~
;22.- -----'z ~ q~z
><
/J
Fig. 1
Let ~.denote the length of the line segment from point P to Li which makes
an angle ofq~iwith Li. (If~0i is 90 °, d~ is the distance to L~.) Let ~ : fl be a given ratio
and a a given line segment. It is required to find points with the following properties:
for three.lines:
(al. a2) : (a~) = ~, :/~, (2; 1)
for four lines:
(d~. d_,): (d3-&) = ~ : f l , (2; 2)
for an uneven number (2n -- 1) of lines, n > 2:
(d~... d,) : ( d n + l . . . d~_ 1 • a) = 0c:fl, (2; 3)
and for an even number (2n) of lines, n > 2:
(al .. a,) : (a,+~ ... az~) = ~ : ~ . (2; 4)
PA1,Fus gives the problem for three and four lines as well as its generalization to
more lines 5. What we have here is a locus problem; in each case there are infinitely
many points which satisfy the condition; these points form a locus in the plane;
4 DESCARTESquoted the problem ((7, pp. 304-306) in Latin from the edition by
COre~ANDn~O, Pappi Alexandrini Mathematicae Collectiones, Bologna 1588, pp. 164v-
165L For the Greek text see F. HULTSCH(ed.) Pappi Alexandrini Collectiones quae super-
sunt Berlin 1876-1878, vol. 2, pp. 677.
s Note however that (2; 3) is not a generalization of(2; 1). In fact (2; 1), the problem
in three lines, is an exceptional case which arises when two lines coincide in the problem
for four lines.
300 H . J . M . Bos
this locus is generally a curve. PAPPUS states that for three and four lines the locus
is a conic section and that for more than four lines nothing is known about the
form of the locus.
2.2 DESCARTESsketches the general solution of the problem at the end of the
first book of the Gdomdtrie (pp. 309-314). His method is as follows. He sets
d~ = y (2; 5)
and he takes x to be the distance along Lt from a fixed point A to the intersection
of d~ with L,. He then shows by simple geometrical arguments that all d~ can be
expressed linearly in x and y:
di = aix + biy -+- ci. (2; 6)
He notes that in the exceptional case when all lines are parallel, x does not occur
in the expressions for the di.
He then remarks that the products d~ ... d,, d,+l ... dz~ and dn+t ... dz,-1 • a
become expressions in x and y of degree at most n. The conditions (2; I)--(2; 4)
can therefore be rewritten as equations. Equation (2; 4) for instance becomes 6
y(a2 x + bz.v + cz) ... (a,,x + b,,y -at- c,,)
0¢
= "~'(an+Ix + bn+tY + cn+1)... (aznx -'}- bz,d' + e2,,). (2; 7)
For n lines the equation will be of degree at most n. F o r n -- 1 lines the choice
o f di = y and the occurrence of a in the second product of lines implies that the
highest power o f x is at most n -- 1. Thus the equation is of degree at most n,
but the highest power o f x is at most n -- 1. This statement does not apply to
the problem with only three lines, which makes that one exceptional. Finally for
2n and 2n -- 1 parallel lines the result is an equation in one unknown, namely
y, of degree at most n; the locus then consists in a number of lines parallel to the
given lines.
DESCARTES goes on to consider how the points satisfying the requirements
o f the problem (the points on the locus) can be constructed. He chooses arbitrary
D~SCARTESdid not discuss the fact that as a result of this rewriting the di may have
negative values, whereas the obvious interpretation of the original problem would re-
quire the d~ to remain positive. The effect of this is that DESCARTESfound only one
curve as locus, while the original interpretation would lead to a locus consisting in two
curves. For instance in the four line locus (taking-~- = 1) D~SCAR~S worked out
$
y(azx + b2y ÷ c2) = (a3x -'k bay -}- c3) (a~x ÷ b.y ÷ c#),
and found one conic section as the locus. But if the d i were taken to be positive, the
equation would become
lY[ [a2x ÷ b2y ÷ e21 = la3x + b3y ÷ c3l la,,x + b,y + e,,l
or
Y(a2x + b2y + c2) = 4- (a3x ÷ b3y + ca) (a~.x ÷ bay ÷ c.),
that is, two conics.
values for y and then constructs geometrically the corresponding values for xo
In this way one can construct as many points as one wishes on the locus. In
Section 6 I shall discuss this type of pointwise construction in more detail. DES-
CARTESre_marks that for any chosen value of y, the corresponding x's are the roots
of an equation the degree of which is, for 2n lines, at most n, and for 2n -- 1
lines at most n -- 1. For three lines the result in general is an equation of degree 2.
The exceptional case of 2n -- 1 parallel lines leads directly to an equation in y
of degree n.
Thus the problem is reduced to the geometrical construction of roots of equa-
tions. DESCARTESthen anticipates results which he will explain in the third book
of the Gdom~trie. These results are the following: The roots of an equation of
second degree can be constructed by ruler and compass. The roots of equations
of third and fourth degrees can be constructed by the intersection of conics, in
particular the intersection of a parabola and a circle. The roots of equations of
fifth anc~ sixth degrees generally cannot be constructed by the intersection of
conics; more complex curves have to be used for them° It is possible to construct
these roots by the intersection of a circle with a certain curve of third degree
namely the "Cartesian parabola". On the basis of these results DESCARTESgives
at the end of the first book the following classification of the cases of the problems
of PAPPUS (G pp. 313-14):
a) 3, 4 or 5 lines, but not 5 parallel lines:
the equation in x is of degree ___2 and therefore points on the locus can always
be constructed with ruler and compass.
b) 5 parallel lines, 6, 7, 8 or 9 lines, but not 9 parallel lines:
the equation in x (or for 5 parallel lines, in y) is of degree ~ 4 and therefore
points on the locus can always be constructed by means of intersections o f
conics; in some cases, construction by ruler and compass only may be possible
(namely if the equations happen to be of degree < 2 or if they are reducible
to such equations).
c) 9 parallel lines, 10, 11, 12, 13 lines but not 13 parallel lines:
the equation in x (or in y in the case of 9 parallel lines) is of degree ~ 6; the
construction by means of intersection of conic sections will in general not be
possible and a more complicated curve will have to be used.
d) etc.
2.3 This classification refers to the construction of the locus. DESCARTES

returns to the problem of PAPPUS in the second book. There he gives another
classification, now according to what he calls the "genre" of the locus; I shall
translate "genre" by "class". This relates to a classification of curves according
to the degree of their equations, a classification that DESCARTES explains in the
second book (G pp. 319-323). He argues there that all geometrical curves have
algebraic equations (cf. Section 9.1). The curves of the first class are those with
equations of the second degree: the circle, the parabola, the hyperbola and the
ellipse. The second class contains the curves with equations of the 3~d and 4th
degree; the third class those with equations of the 5th and 6 th degree, and so forth.
302 H.J.M. Bos
I shall return to this classification in Section 3. It leads to the following classifica*

tion of the cases of the problem of PAPPUS (G pp. 323-324):
a') 3 or 4 lines:
The equation of the locus is of degree at most 2; the locus is of the first class°
b') 5, 6, 7 or 8 lines:
The equation is of degree at most 4; the locus is of the second class or, in
exceptional cases, of the first.
c') 9, 10, II or 12 lines:
The equation is Of degree at most 6; the locus is of the third class or of a lower
class in exceptional cases.
d') etc.
In this connection DESCARTES states that all (algebraic) equations can occur
as equations of the locus of some problem of PAPPUS:
And because the position of the given lines can vary in all sorts of ways and
thereby change the given quantities and the signs + and -- of the equations
in all imaginable ways, it is evident that there is no curved line of the first class
which would not be of use in this problem if it is proposed in four straight
lines, nor one of the second which would not be of use if it is proposed in
eight, nor of the third when it is proposed in twelve, and likewise with the
others. Thus there is no curved line which is subject to calculation and which
can be accepted in geometry, which is not of use for some number of lines.
(G p. 324)°
The statement is incorrect 7. But it is important in DESCARTES' further classifica-
tion of curves; I shall return to it in Section 9.
DESCARTES then gives a complete solution (G pp. 324-334) of the problem
of PAl,PUS in three and four lines, calculating the equations explicitly and discussing
the positions of the resulting conics in the plane. This section is well knownS;
since it is not important for my present subject, I will not discuss it here.
Finally DESCARTEStreats two special cases of the locus of the problem of five
lines which will be discussed in Sections 5.2 and 8.1.
2.4 DESCARTES'solution of the problem of PAPPUS supplies a good illustration

of the two different roles that curves can play in the solution of locus problems:
a curve can occur as a locus; it can occur also as the means to construct points
on the locus. DESCARTEStreats the curves in totally different ways according to
their roles. Consider for instance the case where conic sections occur as loci.
This happens in problems with 3-1ines and 4-lines and DESCARTES states that
these problems are "plane". That implies that he considers the locus in this case
to be constructible by ruler and compass. But with ruler and compass one can
7 See the Appendix.

8 For an extensive discussion of the section see D. T. WH1TESrDE,"Patterns of mathe-
matical thought in the later seventeenth century", Arch. Hist. Ex. Sci. I (1960-1962),
pp. 179-388, especially pp. 290-295.
Curves in Descartes' G~omdtrie 303
only construct points on a conic; one cannot construct or trace the conic as a
whole. Nevertheless DESCARTES considers such a pointwise construction of a
conic sufficient when it occurs as a locus. But when a conic is used as means
constrtiction DESCARTES sets stronger requirements. This occurs in the problem
for 6, 7, 8 or 9 lines (Case b in Section 2.2), where points on the locus are-construct-
ed by means o f intersections of conics with circles and straight lines. Apparently
DESCAR~S does not consider the conics involved here constructible by ruler and
compass, because if that were so the whole construction could be performed by
ruler and compass, and that is what DESCAR~S denies. Hence pointwise construc-
tions o f conics are not acceptable if these conics themselves serve as means o f
construction; in that case their construction must satisfy stronger criteria. F o r
higher-order curves too, the construction has to satisfy different criteria according
to whether the curve occurs as a locus or serves as a means o f construction of
points on a locus.
The reason for a stronger requirement when ~he curve is used as a means of
construction is that it is then supposed that the intersections of the ~urve with
circles, straight lines or other conics can be found° But if a curve is given only
through a pointwise construction as in the locus case, its intersections with other
lines cannot be determined. To illustrate this let C~ and Cz be two conics (see
Figure 2) whose equations are given (or equivalently whose vertices and diameters
C~ /f
A B
/y
Fig. 2
are given) and whose intersections I and J we want to construct. Let AB be the
axis of the x's and Y the direction of the y's. Points on C1 and C2 can be construct-
ed by ruler and compass by taking arbitrary values for x and constructing the
corresponding y's. But I and J cannot be accurately constructed in this way;
we can approximate them but their exact position would only be found if acciden-
tally we started our construction with xt or xl.
I f a curve is used as means of construction, it must be possible to find its
intersection with other curves. A pointwise construction is not sufficient for that
purpose. Instead one obviously needs a method o f tracing the curve by a con-
tinuous motion, so that the intersections with other lines are actually marked.
We will see (Sections 4 and 5) that the requirement that curves be traceable by
continuous motion is crucial to DESCARTES' Gdomdtrie.
304 H.J.M. Bos
3. Descartes' Programmefor Geometry
3.1 In his Gdomdtrie DESCARTESpresented a new approach to the solution

of geometrical problems; it contrasted so strongly with earlier approaches that
one can speak of a new paradigm. Before giving his own opinion on the programme
of geometry DESCARTES explained what mathematicians before him~ especially
those from antiquity, had thought about this matter. He said (G pp. 315-317)
that, traditionally, geometrical problems had to be solved by ruler and compass.
However, classical mathematicians had already encountered problems which
could not be solved in this way. They had solved these using intersections of
conics or even more complicated curves such as the conchoid. But they called
these curves mechanical, thereby implying that they did not consider them to be
genuinely geometrical. DESCARTESwent on to speculate about the reasons the
ancients might have had for this, and he rejected these. His rendering of the
classical arguments was oversimplified, if not inaccurate 9, but it served very well
as an introduction and contrast to DESCARTES' own view.
DESCARTES' view can be summarized as follows: Construction of problems
by ruler and compass is certainly simpler than, and therefore preferable to, construc-
tion by means of the intersection of conics or more complex curves. In the con-
struction of problems one should always use the simplest possible curves. But
this does not imply that more complex curves are necessarily less geometrical
than the straight line and the circle, or that constructions by means of these curves
• are less geometrical than constructions by ruler and compass. There is a collection
of curves of ever increasing complexity (circles~ conics9 conchoids, etc.) which are
in principle acceptable in geometrical constructions. If a problem can be con-
structed by the intersection of two such curves and it cannot be constructed by
simpler curves, then that construction is the right one to choose and it is no less
geometrical a construction than one by ruler and compass.
This vision of the geometrical procedure of constructing problems deter-
mined a programme in three parts° First DESCARTEShad to determine which curves
were acceptable as genuinely geometrical means for the construction of problems.
Secondly, he had to make it clear on which criteria some curves would be consid-
ered simpler than others; this would lead to a classification in order of simplicity
within the collection of geometrically acceptable curves. Finally, a method had
to be devised for finding the simplest possible curves by which each problem could
be constructed. This is essentially the programme which DESCARTESworked out
in his Gdomdtrie.
3.2 The first point of the programme--differentiating between the curves

which are acceptable in geometry and those which are not--caused DESCARTES
(and his successors) the greatest number of conceptual problems. Basically DES-
CARTEStook as geometrical curves those "which can be described by some regular
motion" (G p. 369). But this is not a very clear criterion° Also DESCARTESwished
to include in the collection of geometrically acceptable curves all curves that may
occur as locus solutions of problems such as the problem of PAPPUS. This meant
9 See the article by 1V[OLLANDcited in Note 3.

that in fact--although DESCARTES never explicitly said so--he wanted to regard

all algebraic curves as geometrical. But to do so he would have to prove that all
algebraic curves could be traced by continuous and geometrically acceptable
motions, or that they could be traced by other means which were just as geo-
metrical as the tracing by continuous motion. In Sections 4-9 I shall discuss how
DESCARTES dealt with this very complex part of his programme.
Algebra, in the sense of the existence of an algebraic equation of the curve,
was the essential criterion in the first part of the programme. But the algebra
had to remain implicit. DESCARTEScould not simply take as "geometrical" all
curves that admit an algebraic equation, because obviously that is not a geometri-
cal criterion; if he were to adopt this criterion, DESCARTEScould no longer claim
that he was doing geometry.
3.3 In the second and the third part.of the programme algebra could be used
quite openly, and it formed the crucial tool. DEScARTEs used the degree of the
curve equations to classify the curves according to their simplicity. He divided
them into classes ("genres"); the first class consisted of the curves with equations
of 2nd degree; these are the conic sections. DESCARTESdid not incorporate straight
lines in his classification. Curves with equations of the 3ra and 4th degree were
of the second class; those of the 5th and 6th degree were of the third class, etc.
(G p. 319).
DESCARTESstressed elsewhere that in constructions we should always use curves
of the lowest possible class (G p. 371). He noted that within one class some curves
may be simpler than others, in the sense that one cannot construct such compli-
cated problems with them as one can with the others. For instance, the circle
is of the first class, but there are constructions that can be performed with the
other curves of that class (the conic sections) but not with the circle. DESCARTES
aisomentioned the conchoid as an exceptional curve within the second class
(G p. 323), but he did not discuss how to distinguish the exceptional curves from
the other curves within one class.
It is not quite clear why DESCARTES,after taking the conic sections as the first
class of curves, lumped all curves of the third and the fourth degrees together in
the second class, those of the fifth and sixth degrees in the third class, and so forth.
DESCARTESexplained that he did so because there is a general rule whereby fourth-
degree problems can be reduced to third-degree ones, and sixth-degree problems
to fifth-degree ones (G p. 323). It seems likely that for problems of fourth degree
and third degree he had in mind FERRA~'S rule for reducing equations of fourth
degree (in one unknown) to ones of third degree. But there is no such rule for
equations of sixth and fifth degrees, so in this case DESCARTESwas making rather
a~rash extrapolation. 1°
The classification may also have been connected with the methods used for
constructing roots of equations by the intersection of curves. The roots of equations
lo DESCARTEStended to underestimate the dangers of extrapolating mathematical

results, as is evident, for instance, from the penultimate sentence of the Gdoradtrie:
"For in the matter of mathematical progressions, once one has the first two or three
terms, it is not difficult to find the others" (G p. 413).
306 H . J . M . Bos
of the third and fourth degrees can be constructed by circle and parabola, those
o f fifth and sixth degrees by circle and Cartesian parabola. In other words, by
introducing one higher curve as a new means of construction one can construct
roots of equations o f two successive higher degrees. But there is a puzzling aspect
here. DESCARTES'classification was for curves serving as means o f construction,
whereas this argument would classify problems rather than devices for construc-
tion. In fact, the degree o f the constructing curves (parabola, degree 2, Cartesian
parabola, degree 3, etc.) rises by steps of one.
In DESCARTES' classification, and especially in his arguments about the sub-
division within one class, there is a contradiction between algebraic criteria o f
simplicity (the form o f the equation, in particular its degree) and geometrical
criteria of simplicity (the use of the curve as a device for construction). The special
role that the circle plays in the first class shows that the classification is not really
adequate for dinstinguishing the means of construction. I shall return to this
contradiction in cofinection with pointwise constructions in Section 10.
3.4 For the third part o f the p r o g r a m m e - - t o find the simplest geometrical
construction o f the solution of a given geometrical problem - the crucial tool
again was algebra. A problem should be reduced to an equation in one unknown
(G pp. 300-302). Then the roots of this equation should be constructed geometric-
ally by the intersection of certain curves, which should be as "simple" as possible,
that is, of the lowest possible class. The simplicity of the curves by which a prob-
lem could be solved determined the class to which the problem belonged° Here
DEscAR~s followed classical usage and called problems plane if they could be
solved by circles and straight lines, and solid if they also required a conic section.
DESCARTES devoted most o f the third book o f the Gdomdtrie to this point of the
programme. He proved there that every equation of third or f o u r t h degree could
be constructed by the intersection of a circle and a parabola, and every equation
o f fifth or sixth degree equation by the intersection of a circle and a Cartesian
parabola? x
3.5 At the beginning o f the third book of the Gdomdtrie DESCARTESgave a suc-
cinct formulation o f t h e p r o g r a m m e which I have been describing:
Although all curved lines which can be described by some regular movement
must be admitted in geometry, this is not to say that for the construction of
~ For details of this construction see G pp. 402--411 and WnrrEsmE'S note in The
mathematical papers of lsaac Newton (ed. D. T. WHrlmSZDE, Cambridge, 1967-) vol. 1
(1967), p. 495, note 15. DESCARTES wrote the equation as y6 _ p y S -k qy~ - - r y a +
sy z -- ty + v = 0, where he intendedp, q, r, s, t, and v to be positive. He mentioned
as a condition for his construction that pZ be smaller than 4q. He stated that if this
condition is not satisfied, a substitution y ~ y + c with c large enough will yield an equa-
tion which satisfies the condition. The substitution and its inverse correspond to straight-
forward geometrical constructions~ so the construction of the roots of the resulting
equation yields also the roots of the original equation. In fact DESCARTES'construction
tacitly assumes some further conditions for the coefficients (such as for instance the
arrangement with positive factors and alternating signs), but an these conditions can
be satisfied by performing the substitution y ~ y + e with c large enough. The construc-
tion is therefore general and can be used to find the roots of any equation of sixth degree.
Curves in Descartes' G3om3trie 307
any problem we may use indifferently the first one that occurs. We must al-
ways take care to choose the simplest through which the solution is possible.
And it should be noted that by simplest curves one should understand not
only those which can most easily be described, nor those which make the con-
struction or the proof of the proposed problem easier, but primarily those
of the simplest class which can be used to determine the required quantity.
(G p. 369'-370)
Thereupon came an example of the construction of two mean proportionals
between two given line segments by means of curves traced by a certain machine
(which I shall discuss in Section 5.1). DESCARTESremarked that this construction
might well be the easiest possible construction and might provide the clearest
proof, but it used curves of a higher class than necessary and therefore
... it would be a mistake in Geometry not to use them [namely curves of a
simpler genre]. On the other hand it is als0 a mistake to try in vain to construct
a problem by a simpler class of lines than the nature of the problem allows.
(G p. 371)
4. The representation of curves in Descartes' G~omdtrie
4.1 DESCARTES dealt with the fundamental question of his programme at

the beginning of the second book of the G~om~trie. He formulated that question
in the marginal title as: "which are the curved lines that can be accepted in geo-
metry'" (G p. 315). He criticised the classical mathematicians for having called
certain curves used in geometrical constructions "mechanical" rather than "geo-
metrical". DESCARTES said that the fact that "'mechanical" curves are described
by certain machines does not make them less geometrical than the straight line
and the circle, which, after all, are also traced by machines, namely by ruler and
compass. DESCARTESdid not wish to impose such strict requirements for geometri-
cal curves; he accepted many more curves as geometrical:
To trace all the curved lines which I wish to introduce here, nothing else need
be supposed than that two or several lines can be moved one by the other,
and that their intersections mark other lines . . . . (G p. 316)
Such curves may be very complicated, but that need not make them less geometri-
cal:
It seems very clear to me that if we consider (as is customary) as geometrical
that which is precise and exact, and as mechanical that which is not, and if
we consider geometry as the science which furnishes a general knowledge
of the measures of all bodies, we have no more fight to exclude the more com-
posite lines than the simpler ones, provided that we can imagine them as de-
scribed by a continuous motion, or by several motions following each other,
the last of which are completely regulated by those which precede. For in this
way one can always have an exact knowledge of their measure. (G p. 316)
DESCARTES' criterion, then, to accept curves as geometrical was that they
could be traced by continuous motion. The tracing of the curve is basic for an
308 H.J.M. Bos
understanding of its nature; significantly, DESCARTES combined the word

"tracing" with understanding and conceiving; he used expressions such as:
"ways to trace and conceive curved lines" (G p. 319) and "to know and trace the
line" (G p. 307).
4.2 Since DESCARTES considered curves primarily as traced by continuous

motions generated by certain machines, he faced a number of difficult conceptual
problems which are summarised below:
a) There are certain curves, such as the spiral and the quadratrix, which DES-
CARTESdid not accept as geometrical but considered to be mechanical, in the sense
that they were imprecise and inexact. These curves, however, can be traced by
continuous motion (see Section 7.2). DVSCARTEShad therefore to specify which
types of motion he accepted and which he rejected.
b) In the course of his studies DESCARTEScame across several curves which he
could not, or would not, present as traced by some continuous motion/Instead
he presented them as constructed pointwise or as traced by machinery involving
string. He had therefore to argue that such constructions or methods of tracing
are just as acceptable in geometry as tracing by continuous motion.
c) Pointwise constructions and tracing machinery involving string can a/so be
devised for curves which DESCARTES did not accept in geometry. Therefore he
had to specify which pointwise constructions and which methods of tracing
with string were acceptable.
d) Algebra was the crucial tool in DESCARTES'new programme for geometry, and
the new curves he wished to introduce had to be amenable to algebraic treatment,
that is, they had to have an algebraic equation. Thus DESCARTEShad to consider
whether his new curves had such equations, and conversely, whether equations
resulting from the use of the algebraical methods would always correspond to
geometrically acceptable curves.
The representation of curves is fundamental to these problems. The arguments
about the acceptability of curves can only be formulated in terms of the represen-
tations of the curves, and the discussion is complex because three different methods
of representation are involved: representation by specifying the continuous mo-
tion which traces the curve, representation by the method of constructing points
on the curve, and representation by specifying the tracing machinery involving
string. To these three one may add the fourth method--the representation of a
curve by its equation. However, all these problems in fact arise because DFSCARTES
did not consider it acceptable in Geometry to represent a curve by its equation.
In the following sections I shall summarise DESCARTES"arguments with regard
to the four problems mentioned above.
5. Curves described by continuous motion
5.1 DESCARTESsaid that to trace the curves that are acceptable in geometry,
"'nothing else need be supposed than that two or several lines can be moved one
by the other and that their intersections mark other lines" (G p. 316). In the second
book of the G~omdtrie he gave two examples to illustrate the kinds of motion
he had in mind.
The first example concerned the famous instrument of Figure 3 (G p. 318 and
p. 370).
/
! /
/ / ,Q"
Fig. 3
It is a system of linked rulers. The rulers YX and Y Z are connected in Y by a

pivot. Ruler BC is fixed perpendicular to YX in B. The rulers CD, E F and GH
are made in such a way that they can slide along Y Z and still remain perpendicular
to it. Similarly the rulers DE and FG slide along Y X while remaining perpendicu-
lar to it. At the beginning of the motion of the instrument angle X Y Z is assumed
to be zero and all the rulers coincide in point A. Now the angle X Y Z is opened
by keeping Y Z fixed and rotating YX. Ruler BC pushes CD outwards, CD pushes
DE, DE pushes EF, etc. Point B (fixed on X Y ) describes a circle; points D, F and
H, sliding along YX, describe other curves (dotted in the figure). DESCARTES
argued that these curves, although described by more and more complicated
combinations of motions, should all be accepted in geometry:
But I do not see what could prevent us from conceiving the description of the
first [i.e. the curve described by D] as clearly and distinctly as that of the circle,
or at least as that of the conic sections, nor what could prevent us from con-
ceiving the second one and the third one and all the others, which one can
describe equally well as the first one; nor therefore what could prevent us
from accepting all these curves in the same manner, to serve the speculations
of geometry. (G pp. 318-19)
The instrument of Figure 3 is found even in DESCARTES' very- early studies;
I shall discuss its use and origin in Section 10. It is interesting to note that DES-
CARTES' discussion of the instrument in this passage did not serve primarily to
explain which motions he had in mind for the tracing of acceptable curves; he
added another example which explained this better. Rather, the instrument served
to show that, however composite a motion is, the resulting curve can be con-
310 H.J.M. Bos
ceived in a clear and distinct way, and is therefore acceptable in geometry. The
instrument was a precise illustration of the description, given above, of acceptable
tracing motions:
a continuous motion, or(-) several motions following each other the last of
which are completely regulated by those which precede. (G p. 316)
Here the first motion is the rotating motion of the rulers YX and_ BC, the subse-
quent motions are those of the rulers CD, DE, EF, etc. ;BC regulates the motion
of CD, CD that of DE and so forth.
The text, and especially the use of the key words, clear and distinct ("nette-
merit", "distinctement", G p. 318), show that DESCARTESsaw a paraIlet between
the series of interdependent motions in the machine, all regulated by the first
motion, and the "long chains of reasoning" in mathematics, discussed in the Dis-
cours de la Methode, which, provided each step in the argument is clear, yield
results as clear and certain as their starting point. ~2
5.2 But the example of the instrument of Figure 3 did not cover all the com-
binations of motions which DESCAaTm had in mind, because it involved only
straight lines as moving parts. When DESCARTES wrote "nothing else need be
supposed than that two or several lines can be moved the one by the other, and
that their intersections mark other lines", he also had moving curved lines in
mind. This becomes clear in a series of further examples. These concerned a tracing
process whereby new, more complicated curves are generated from the motion
of simpler curves and straight lines. DESCARTESproceeded as follows (G pp. 319 if;
see Figure 4):
. . , ....... ...
j,4 ~ , "
,*®q C °
/J A
Fig. 4
1: DESCAgTES~Euvres (see Note 1) vol. 6, p. 19.

A ruler G L pivots at G. It is linked at L with a device N K L which can be moved

along the vertical axis while the direction of the line K N is kept constant. When L
is moved along the vertical axis, the ruler turns around G and the line K N is moved
downwards remaining parallel to itself. The intersection C of these two moving
straight lines describes the curve G C E . DESCARTESderived the equation of this
curve
c
y : = cy - - --~ x y q- a y - - ac (5; 1)
(where G A = a, K L = b, N L = c, C B = y and A B --- x) and concluded that it

was a curve of the first class; he added that it was in fact a hyperbola. Thus the
straight line K N in the machine produced a curve of the first class.
Next DESCARTES asserted (G p. 322) that if the straight line in the machine
were replaced by a curve of the first class, the resulting curve would be of the
second class. He mentioned the case where K N i s a circle with centre L; the resulting
curve will then be the conchoid of NICOr~EDES.(See Figure 5; on all the lines through
a fixed point (G), the intercepts between the conchoid and an axis ( K A ) are equal.)
The conchoid is of a higher class than the conics.
! IK
/ I ~
Fig. 5
Then DESCARTESreplaced the circle by a parabola (see Figure 6), and stated
that the resulting curve would be the "first and simplest curve for the problem
of PAPPUS if there are only five lines given in position" (G p. 322). The curve
played a central role in DESCARTES" Gdombtrie; it became later known as the
II
!
"~'\ / l
I
\ / I
\\ / I
,, / I
I
Fig. 6
312 H. L M. Bos
"Cartesian parabola". 13 Further on in the second book (G pp. 335-337) he showed

that this curve was indeed the solution of a special simple case of the five line locus
problem 1'~, and he gave the equation o f the curve:
y a _ 2ayZ _ a 2 y + 2 a 3 = a x y . (5; 2)
la See G. LORIA, Spezielle algebraische u n d transzendente ebene Kurven, Leipzig

(2d ed.) 1910-1911, vol. 1, pp. 51-52. Other names for the curve are "trident" (N~WTON)
and "parabolic conchoid"°
14 Namely the problem ( c f Section 2.1)
d~ " dz " d3 = d~. " ds " a
in the case where L~, Lz, L , and L3 (in that order) are equidistant and parallel, and Ls
is perpendicular to the other lines; a is the distance between Lz and Lz, and all the
distances di are take n perpendicular to Lt. The equatio~i of the curve is:
y a _ 2ayZ _ aZy + 2a 3 = a x y .
Compare the figure (from G~ p. 336). The lines L i are GF, E D , I H , A B a n d G A respectively.
The distances are taken perpendicular to the lines, d l = CF~ d2-----CD, d3 = CH,
d,, -~ C B ~-- y~ ds = C M = x. G L is the ruler moving around G. C K N is the parabola
moving vertically along its axis ~lB. G E C is the branch of the "Cartesian parabola"
described by C~ N I O is the branch described by the other intersection N , cGc and o l n
are the branches of the other "Cartesian parabola" which occurs if the distance to Ls
is taken to be positive in the other direction.
Figure for Footnote 14

In the third book DESCARTESexplained how this curve can be used for finding
the roots of a sixth degree equation (G p. 403); he discussed there in more detail
how the curve was traced by the combined motions of a ruler and a parabola.
In the passage from the second book where he introduced this curve DESCARTES
said that it was o f the second class. Furthermore he claimed that if a curve o f the
second class was used in the tracing, the resulting curve would be of the third class
etc. (G p. 322), b u t h e did not prove this. 15
5.3 The linkage machines and the device of a moving curve whose intersection
with a ruler traces new curves are the examples which DESCARTESgave to illustrate
his concept o f tracing curves by combination o f motions. It is a fundamental
concept because DESCARTES stated that he would introduce new curves only if
they were traceable in this way. This means that there were other curves which
he would not accept as geometrical because they could not be traced in this way.
As examples DESCARTES mentioned the spiral and the quadratrix. However, both
the spiral and the quadratrix can be traced by a combination of continuous mo-
tions; they were in fact defined in such a way. DESCAR~.S had therefore to specify
a further requirement for the motions in order to rule out these curves. Before
discussing this requirement I shall indicate the ways in which the two curves men-
tioned can be traced by continuous motions.
The Archimedean spiral 16 (see Figure 7) is described by two motions, one
rotatory motion of a ruler OR which turns uniformly around O, and one recti-
linear motion of a point P which moves uniformly along the ruler OR. The point
P traces the spiral.
Fig. 7
t5 DESCARTESclaimed that it is easy to prove this by actual calculation. But this

is not true. A counter example arises when one takes ya = c2z for the moving curve
(taking z as vertical coordinate measured from K, and y as horizontal coordinate).
This curve has degree 3 and therefore belongs to the second class. The new curve described
by the intersection of this curve and the ruler GK has equation ( a - y) y a =
c:(xy -- by + ab) (where x is the vertical coordinate taken from A) a ----GA, b = KL).
This curve has degree 4 and therefore also belongs to the second class. In general, if
the equation of the moving curve is linear in z, the degree of the resulting curve will
be only one higher than the degree of the moving curve (comme il est fort ays6 a connoistre
par le calcul), and hence in that case the Tesulting curve may belong to the same class as
the original curve.
~s See ARCHIMEDES, On Spirals, definition 1, in The works of Archimedes (ed.
T. HEATH, New York, Dover reprint), p. 165.
314 H. L M. Bos
0 C
Fig. 8
The quadratrix 17 (see Figure 8) can be described by a combination of separate

motions, namely by one rotatory motion of a ruler OA which turns uniformly
around A, from position OA to position OC, and one rectilinear motion of a
ruler PQ which moves uniformly downwards from position AB to position OC,
during the same time that OR turns from OA to OCo The intersection S of both
rulers traces the quadratrix ASD. The construction implies that during the motion
always
A P : A R = AO AC. (5; 3)
DESCARTES said about the spiral, the quadratrix and similar curves, that
"they are conceived as described by two separate movements, between which
there is no relation ("raport") that can be measured exactly", and that for that
reason they "really only belong to Mechanics"° (G po 317) The absence of a mea-
surable "raport" is the essential point here, for in both cases the two movements
could in principle be linked in such a way that the one determines the other,
namely by a string mechanism which I shall discuss in Section 7. In view of the
methods of tracing the quadratrix and the spiral we may conclude that when
DESCARTES spoke about the measures of the motions, he meant their velocities.
Indeed these measures have no exactly measurable "raport", as the comparison
of the velocities involves the comparison of the lengths of straight and curved
lines, in particular the ratio AO : AC of the radius of a circle to the quarter arc
of that circle° The argument returns some pages later in the G~om~trie in connec-
tion with the tracing of curves by machines involving strings (see Section 7).
There DESCARTESwrote:
.,o the proportion between straight lines and curves is not known and I even
believe that it can never be known by man, (G p. 340).
Thus the separation between geometrical and non-geometrical curves, which
was fundamental in DESCARTES' vision of geometry, rested ultimately on his
conviction that proportions between curved and straight lengths cannot be found
exactly. This, in fact, was an old doctrine, going back to ARISTOTLE.Is The central
role of the incomparability of straight and curved in DESCAR~S' geometry ex-
x7 Often called the quadratrix of DINOSTRATU$,although the names of HiPPt~s and

NICOMEDF.Sare also connected with the curve. PnpPus discusses the curve in his Mathe-
matical collections° See I. BtrLMERTHOMAS,"Dinostratus", Dictionary of scientific bio-
graphy (ed. C° C. GmLiSPm, New York 1970ff.) vol. 4, pp. 103-105.
18 See T. L. HEATH, Mathematics in Aristotle (Oxford 1949), pp. 140-142.
plains why the first rectifications of algebraic (i.e. for DESCARTESgeometrical)

curves 19 in the late 1650's were so revolutionary: they undermined a cornerstone
of the edifice of DESCARTES' geometry.
5.4 These, then, were DESCARTES' arguments about tracing curves by con-
tinuous motion. A curve could be accepted as geometrical if there was an acceptable
way of tracing it. Obviously, this criterion was connected with the use of the curve
as a means of construction (see Sections 2.4 and 10); the intersections of the curve
with other lines could be considered constructable only if the curve was actually
traced. But in the Gdom~trie DESCARTES also accepted other ways of representing
curves. I shall discuss these, and their relation to tracing by continuous motion,
in the following sections.
6. Pointwise construction of curves
6.1 As we have seen (Section 2.2), DESCARTES solved the problem of PAPPUS
by constructing arbitrarily many points on the locus. The method was as follows:
first derive the equation of the locus in indeterminates x and y; then choose an
arbitrary value ~7for y and form the equation in one unknown for the correspond-
ing value or values of x; then solve this equation geometrically, that is construct
the root or roots ~; and finally construct the point or points with coordinates ~:,r/
on the locus. By repeating this process~ taking other values for y, one can find
arbitrarily many points on the locus. However, it is not at all obvious that this
construction can be regarded as a satisfactory construction for the whole curve
which forms the locus. It is not a construction by continuous motion. The process
yields only a finite number of points on the curve. And generally it is not possible
to use this construction for determining the intersection of the locus with a given
curve (cf. Section 2.4).
In his discussion of the problem of PAPPUS in the first book of the Gdomdtrie
DESCARTES did not say whether this pointwise construction could be considered
as a construction of the locus as a curve. In the case of the three-line and four-line
problems, where the locus is a conic, DESCARTESdid not stop after giving the point-
wise construction; he also indicated how in each case the position of the vertices,
axes, latus rectum and latus transversum z° could be found, thus giving a represen-
tat.ion of the locus curve by naming it (ellipse, hyperbola etc.) and giving its basic
parameters (G pp. 327-332). However, later on in the second book DESCARTES
returned to pointwise constructions of curves and stated that, in certain cases,
curves constructed pointwise should be accepted in geometry.
t9 Rectifications of algebraic curves were found around 1658, independently, by
VAN HEURAET, NEILE and FERMAT. See for instance M. E. BARON, The origins o f the
infinitesimal calculus (Oxford 1969), pp. 223-228.
zo Latus rectum and latus transversum are the classical terms for certain line segments
occurring in the defining properties of conic sections. If the vertex of the conic section
is taken as origin and the X-axis is along the diameter, then the latus rectum a and the
latus transversum b occur in the analytical formulas for the conics in the following way:
y2 _ ax (parabola); y~ = ax -- ~b x2 (ellipse); y z = ax q- ~ x z (hyperbola).
316 H . J . M . Bos
He made this stament when dealing with the five-line locus problem
dl°dzoaa=a,.as.a. (6; 1)
DESCAR~S solved the problem for two special cases, both involving four equi-
distant parallel lines and the fifth line perpendicular to the others. He took the
position of the lines as in Figure 9a and 9b, respectively. In the first ease (G
pp. 335-339; cfi section 5.2 and note 14) he found that the locus was the "Carte-
sian parabola" which he had introduced earlier as the curve described by a com-
bined motion o f a ruler and a parabola. F o r the second case DESCARTES gave
only a property o f the locus (formulated in a rather obscure wayZ~), from which
at best a pointwise construction could be derived (G p. 339); he did not explain
how that locus could be traced by a continuous motion. He then decided not to
give any more details because he had already indicated in the first book how
points on the locus could generally be constructed:
As to the lines serving in the other cases, I shall not bother to distinguish thena
into different kinds, for I have not undertaken to say everything. And now
zx DESCARTES'cryptic description of the curve is as follows: The curve is such that

if all the straight lines applied consecutively to its diameter [i.e. the ordinates] are
taken equal to those of a conic section, then the segments of the diameter between
the vertex and these lines [i.e. the abscissae] have the same ratio to a given line as
that line has to the segments of the diameter of the conic section to which these
lines are applied consecutively. (G p. 339)
Following C. RABUEL, Commentaires sur la Gdomdtrie de Me Descartes (Lyon 1730)
po 271, we may interpret the passage as follows. If we take the origin in the centre of the
figure (cf. Figure 9b),
dl " dz " dz -= d,~ " d5 " a
leads to
x(Yq-~a)(y--3a)=(Yq---~)(Y--'~) a
as the equation for the required curve° Taking
1 a- 1 (yZ - ' T 9 a z ) ,
W~ T !
we find
w:a =a:(x-a).
If we now take the "vertex" in DESCARTES'text to be the point V ( x = a, y = 0), and draw
the parabola
2aw = y2 - - - ~9- a ,
with w taken along the X-axis from V, then the required curve and the parabola are
related in such a way that for points (x, y) and (z, y) on either curve with equal ordinates
y, the abscissae x - - a and w (taken from IF) satisfy
w:a=a:(x --a).
This corresponds to what DESCARTESsays, but he does not specify that in this ease the
conic section is a parabola. If the conic section and the position of the vertex are given,
DESCARTES' description implies a pointwise construction of the curve.
L~ Lz L~ L~ LZ L~ L5 L;
(7 6, 6' 6,
L5
a b
Fig. 9
that I have explained the way to find an infinity of points through which they
pass, I think that I have sufficiently explained the way to describe them.
(G p. 339)
Thus DESCARTrS stated that pointwise constructions are sufficient to describe the
curves.
(i.2 Dr:.SC~TES said more about the acceptability of these constructions in

the next section, the margin title of which is:
Which are the curved lines that one describes by finding many of their points
and that can be accepted in geometry. (G p. 340)
As the title indicates, DESCART~ accepted pointwise constructions under certain
conditions as sufficient constructions for curves. As was the case in the tracing
of curves by continuous motion, these conditions must exclude curves such as the
spiral and the quadratrix. However, as DESCART~ said, there are pointwise con-
structions for these curves as well. DESCARTESprobably had in mind here the follow-
ing pointwise construction for the quadratrix 22 (see Figure 10). Divide arc A C
0 D C
Fig. 10
in 2, 4, 8, 1"6 etc. parts (this can be done with ruler and compass) and do the same
with the radius OA. Then draw radii such as OR to the points of division on A R C
and draw horizontals such as TS through the points of division of OA. The
intersections, such as S, of corresponding radii and horizontals are on the quadra-
trixo In this way arbitrarily many points on the quadratrix, lying arbitrarily close
z2 This pointwise construction follows immediately from the description of the qua-
dratrix by continuous motion; see Note 17. T. L. H~TI-I mentions the construction in
his A history of Greek mathematics (2 vols, Oxford 1921), vol. 1, p. 230, as a means to
find points on the curve near D (Figure 9) and thus to approximate D. But he gives
no reference to classical sources containing this construction.
318 Ho L M. Bos
to each other can be geometrically constructed. There is a similar construction for

the spiral.
Since DESCARTES had to exclude these pointwise constructions, he had to
explain the difference between these unacceptable constructions and the acceptable
pointwise constructions of, for instance, the loci for the problem of PAPPUS.
According to DESCARTESthe difference lay in the fact that for curves such as the
quadratrix the constructable points were special points. For the quadratrix
k
as constructed above, they are the points with ordinates ~-# OA; generally one
finds only those points of the quadratrix which correspond to a division of the
angle which is possible by Euclidean constructions. In the case of acceptable
pointwise constructions every point is in principIe constructable because the con-
struction may start from any given value of one of the coordinates. DESCARTES
. explained this as follows:
It is worthy of note that there is a great difference between this method of
finding several points to trace a curved line, and that used for the spiral and
similar curves° For with the latter one does not find indifferently all points of
the required curve, but only those points which can be determined by a simpler
measure than is required for the composition of the curve° Therefore, str/ctly
speaking, one does not find any one of its points, that is, not one of those
which are so properly points of the curve that they cannot be found except
by means of it. On the other hand there is no point on the curves which are
of use for the proposed problem [the PAPPUS problem] that could not
occur among those which are determined by the method explained above.
• And because this method of tracing a curved line by finding a number of its
points taken at random is. only applicable to curves that can also be described
_ by a regular and continuous motion, one may not exclude it entirely from geo-
metry. (G pp. 339-340)
Thus DESCARTES stated firmly, but without any attempt at proof, that curves
admitting a pointwise construction in which every point on them can, in principle,
be constructed, can also be traced by continuous motion and are therefore geo-
metrical. The passage suggests that DESCARTES saw a correspondence between the
complete arbitrariness of the constructed points on the curve and the continuity
of the motion.
6.3 After this passage DESCARTES repeatedly used pointwise construction

to represent curves. For instance he introduced the famous ovals z3, which are
curves with certain optical properties, by giving a pointwise construction. As an
illustration of such a representation by pointwise constructions I summarize
DESCARTES' introduction of the first oval (G p. 352, "this is how I describe them"):
2a The ovals which DESCARTESdiscusses on pp. 352-368 of the G~om~trie are curves
whose surfaces of revolution provide shapes of lenses with the property that light rays
coming from one point converge, after passing through the lens, to another point (and
variants of this property). DESCAXT~Sexplains how these ovals can be constructed when
the positions of the light source and the converging point, and the refractive index of
the lens material, are given.
L R
Fig. 11
Let two lines (see Figure 11) be given, intersecting in A at a given angle. A lies
between the points F and G on the one line; the ratio of A F to AG is given. R
lies on the other line, AG = AR. To construct points on the oval, take an arbi-
trary point K on AG. Draw a circle with centre F and radius FK. Draw KL per-
pendicular to AR. Draw a circle with centre G and radius RL. The two inter-
sections of the two circles lie on the oval. By repeating this construction starting
from other points K on AG, arbitrarily many points on the oval can be found.
The construction yields a geometrical curve, because the choice of Kis completely
arbitrary.
7. Curve construction using string
7.1 The passage in the second book on the geometrical acceptability of curves
given by pointwise constructions was followed by a passage about a third way of
representing curves, namely tracing them with machines involving strings. The
title in the margin of that section is:
And which curves that one describes by means of a string can be accepted.
(G p. 340)
DESCARTES then referred to his Dioptrique z4, in which he had given constructions
by strings for the ellipse and the hyperbola. The construction for the ellipse is
the well known "gardener's construction" (see Figure 12): A string is fixed in
the points A and B. It is stretched by a tracing pin T which is moved around A
and B, the strings being kept straight. It then traces an ellipse with loci A and B
(A.T. 6, p. 166).
Fig. 12
z,, La Dioptrique, one of the three essays of the Discours; in DESCARTES'¢~Euvres
(see Note I), vol. 6, pp. 79-228.
320 H. 3"oM. Bos
Fig. 13
For the hyperbola (see Figure 13) a ruler AR pivots at A; a string is fixed at B
and at point R on the ruler. The string is stretched by a tracing pin T which is
kept against the ruler. When the ruler is turned around A with T kept fixed to
the ruler and A T stretched, T describes one arm of a hyperbol~i with loci A and B
(A.To 6~ p. 176).
It should be noted that in the Dioptrique DESCARTEScalled this construction
of the ellipse "rather rough and not very exact" (A.T., 6~ p. 166) but thought that
it was a better means for understanding the nature of an ellipse than the section
of a cone or a cylinder. In discussing the construction of the hyperbola DES-
CARTESpictures "a gardener who uses it to mark offthe border of some flower bed"
(A.T. 6, p. 176). Nevertheless, in the Gdomdtrie DESCARTESaccepted these construc-
tions as genuinely geometrical representations of curves. This shows that he was
more concerned that his constructions should be clear and comprehensible in
principle than that they should be accurate in practice.
7.2 But constructions by strings could be used also to trace curves which
DESCARTES did not accept as geometrical. DESCARTES mentioned this but did
not give examples. He may have had in mind a method similar to the one which
HLIYGENSin 1650 suggested for tracing the spiral 25 (see Figure 14, which is HoY-
GENS' sketch): A ruler AB pivots in Bo Around B there is a circular disk E H
I
t
%
Fig. 14
2~ See C. HUYGENS, (~uvre$ Complbtes (22 vols. The Hague 1888-1950) vol. 11,
p. 216; a note from 1650.
Curves in Descartes' G~omOtrie 321
fixed to the plane. A string EAD is fixed to the rim of the disk in E, slung around a
small pulley at A and led along the ruler to the centre. A drawing pin is fixed at
the end D of the string. If the ruler is moved around counter-clockwise, the string
winds up round the disk, the pin D is drawn along the ruler and describes an
Archimedean spiral on the plane. DESCARTESwas certainly able to devise machinery
of this kind for tracing the quadratrix.
DESCARTES had to exclude this way of using string in tracing curves. He did
so by excluding the cases where the string is partly curved and partly straight and
where during the motion curved parts change into straight ones or vice versa.
His reason for excluding these cases was, as I have discussed above (Section 5.3),
his conviction that ratios between straight and curved lines cannot be given
exactly° DESCARTESargued as follows:
Nor should we reject the method in which a string or a loop of thread is used
to determine the equality of or the difference between two or more straight
lines which can be drawn from each point of the required curve to certain other
points or towards certain other lines at certain angles. We have used this
method in the Dioptrique to explain the ellipse and the hyperbola. It is true,
though, that one cannot accept in geometry any lines which are like strings,
that is, which are sometimes straight and sometimes curved, because the pro-
portion between straight lines and curved lines is not known and I even believe
that it can never be known by man, so one cannot conclude anything exact
and certain from it. Nevertheless, because in these constructions one uses
strings only to determine straight lines whose lengths are perfectly known, this
should not be a reason for rejecting them. (G pp. 340-341)
7.3 Further on in the Gdomdtrie DESCARTESused string constructions as an

alternative to pointwise constructions to represent ovals. By way of illustration I
summarise the string construction of the first oval (see Figure 15): FE is a ruler
that pivots at F. A string is fixed at E on the ruler and at G on the axis FAG.
It is slung around a pin K on the axis and it is kept straight by a tracing pin at
C against the ruler. Thus the string is kept to E - C - K - C - G . Now the ruler is
°°.~., .°o.'°
Fig. 15
322 H. J. M. Bos
turned around F and in that motion the tracing pin C traces the oval. The points
F, A, Kand G on the axis can be chosen such as to give the oval the required optical
properties.
8. Equations of curves in the G~om~trie
8.1 We have seen that DESCARTESused three different kinds of representation

of curves: tracing machines, pointwise constructions and tracing machines in-
volving strings. In each case further conditions (which exclude the transcendental
curves) have to be satisfied if the resulting curves are to be acceptable in geometry.
In the first and the third kind of representations these conditions have tO do with the
axiom of incommensurability of the straight and the curved, and in the second with
the randomness of the constructible points on the curve. It is noteworthy that DES-
CARTES did not try to connect these two types of condition. In fact they relate
to different aspects of curve tracing by continuous motion. Incommensurability
of the straight and the curved relates to the condition that the combined motions
which trace the curve regulate each other in a measurable way (cfi Section 4.1).
The randomness of the constructable points relates to the continuity of the tracing
motions (el. Section 6.2).
We must now consider the role of equations as representations of curves:
to what extent did DESCARTESconsider the equation to be a sufficient representa-
tion of a curve ? DESCARTESwas convinced that the equation of a curve incorporates
M1 information on its properties; He wrote:
Now if one knows the relation that all points of a curved line bear to all
points of a straight line in the way I have explained [i.e. as soon as the equation
is known], it is also easy to find their relation to all the other given points
and lines; and subsequently to find the diameters, axes, centres and other lines
or points to which each curve has some special relation, or a more simple rela-
tion than to others, and in that way to conceive various ways of describing
the curves, and to choose the easiest° (G p. 341)
The passage suggests that finding the description of the curve from its equation
still requires an effort, so the equation itself is not an appropriate representation
of the curve.
This is in keeping with the fact that nowhere in the Gdomdtrie did DESCARTES
use an equation to introduce or represent a curve. In several cases he treated
curves without giving their equations; in other cases he gave the equation almost
casually in the course of his arguments. The solution of the problem of PAPPUS
with five lines, four of which are parallel, equidistant and perpendicular to the
fifth (cf. Section 6.1 and Note 21), was given in Book II by a prose description
of a defining property of the locus. The description could have been translated
into an equation, and would certainly have been more informative if it had been.
Equations of the curves traced by the machine discussed in Section 5.1 were not
given in the Gdomdtrie, nor did DESCARTES present the equations for the ovals.
The Cartesian parabola, so fundamental to the Gdomdtrie, was introduced in
Book II as the curve traced by the intersection of a parabola and a ruler. Its equa-
tion was given afterwards, and clearly not as a representation of the curve b u t
as a means of proving that the curve solves the five-line locus problem (G p. 337),
or as a means to determine its tangents (G p. 344). For readers to whom the de-
scription of the curve by ruler and parabola "seems difficult" DESCARTESadded
as an alternative representation a pointwise construction, but not the equation
(G p. 407).
The conclusion from these facts must be that for DESCARTES.the equation of
a curve was primarily a tool and not a means of definition or representation.
It was part of a whole collection of algebraic tools which in the Gdomdtrie he showed
to be useful for the study of geometrical problems. The most important use of the
equation was in classifying curves into classes and in determining normals to
curves. Here the equation must actually be written out. In many other cases
DESCARTES could get through his calculations about problems without writing
down the equation of the curve explicitly.
9. Geometrical curves
9.1 Within his programme for geometry, DESCARTESdid not, and could not,
simply state that geometrical curves are those which admit algebraic equations.
But how did DESCARTESsee the class of geometrical curves ? Did he really consider
this class to be the same as the class of curves admitting algebraic equations, and
did he think that every such equation could occur as the equation for a geometric-
al curve ? And was he aware of the extension of the class of curves which he decided
to banish from geometry? I shall deal with these questions in this section.
DESCARTES stated firmly that all geometrical curves have equations. After
explaining the curve-tracing machine discussed in Section 5.1 he wrote:
I could give here several other ways of tracing and conceiving curved lines,
which would be more and more complicated by degrees to infinity. But to
understand the totality of all curves that are in nature and to distinguish
and order them in certain classes, I do not know a better way than to say that
aU points of those that can be called geometrical, that is those which admit
some precise and exact measure, necessarily have some relation to all points
of a straight line, which can be expressed by some equation, the same equation
for all points. (G p. 319)
He went on to explain how these equations can be found for the curves traced by
the machines discussed in Section 5.2.
The converse question, namely whether all algebraic equations describe geo-
metrical curves, is a much more difficult one and DESCARTESdid not answer it
explicitly. Taken in its strict sense the question is whether for every algebraic
equation a tracing machine, or a combination of continuous motions in the sense
explained in Section 5, can be found which describes the curve having that equa,
tion. DESCARTES did not deal explicitly with that question anywhere. However,
it is such a fundamental question in the whole Cartesian programme of geometry
that it seems very unlikely that DESCARTESwas unaware of it. His silence on this
324 H.J.M. Bos
question must be due to his inability to answer it. It is not surprising that DES-
CARTES could not answer it; a proof that the answer to the question is positive
was found only in the 19th century. 26
Implicitly, DESCARTES' answer to the question was positive. An equation of
a curve implies a pointwise construction; one takes successive fixed values for one
of the variables, say for y, and constructs geometrically the corresponding values
for x as the roots of the resulting equation in x. DESCARTESwas convinced that
this could always be done. In the third book of the Gdomdtrie, he showed that the
roots of equations (in one unknown) up to the sixth degree can be found by the
intersection of geometrically acceptable curves, and he claimed that the same can
be done for equations of higher degree (G po 413; c f Section 3°4)° This is how
DESCARTESsolved the problem of PAPPUS and he even claimed that every equation
can arise as the equation for the locus in a problem of PAPPUS in some number
of lines. Hence algebraic equations yield pointwise constructions for the curves
they describe and these constructions are acceptable in geometry because one has
a complefely free choice of starting point for the construction of the points (namely,
the choice of the y; see Section 6.2). Moreover, DESCARTES claimed that such
pointwise constructions of curves are equivalent to tracing by continuous motion,
and hence, implicitly, he claimed that all algebraic curves are geometrical in the
sense of being traceable by continuous motion.
It is clear that the crucial step in this argument is the equivalence of pointwise
constructions and constructions by continuous motion. Through this equivalence,
curves described by equations acquire a status in geometry equal to that of curves
traced by continuous motion. But we have seen that DESCARTES' arguments for
the equivalence were weak (Section 6.2). He must therefore have had strong
reasons for incorporating the equivalence in his geometry. In Section 10 I shall
say something more about his reasons and about some conclusions which may be
drawn from these concerning the formation of DESCARTES' ideas in geometry
in the years before the publication of the G~om~trie.
9.2 It is noteworthy that DESCARTES'basic argument in rejecting the transcen-

dental curves was the incommensurability of straight and curved lines° This
argument applied only to transcendental curves depending on the quadrature of
the circle, such as the quadratrix and the spiral, which were the only ones DES-
CARTES mentioned explicitly. But how many transcendental curves did he know,
and, what is more important, did he know curves depending on logarithmic rela-
tions, and which arguments did he use to exclude these from geometry ?
The idea that curves generated by motions not mutually subordinate are to
be rejected from geometry occurred earlier in DESCARTES' letter to BEECKMAN
of 26 March 161927 (see Section 10.2); he mentioned the quadratrix as an example.
By that time he had also hit upon the logarithmic relation in connection with
the problem "de reditu redituum'" (income on income, i.e. compound interest).
He considered two axes, one divided in equal parts, the other in proportional
26 A. B. I~MPE, "On a general method of describing plane curves of the n th degree

by linkwork", Proc. London Math. Soc. 7 (1876) pp. 213-216.
z7 A.T. 10, pp. 154-158.
parts. In one study 2s the idea of a curve representing the relation between corre-
sponding parts on each axis seemed to underlie his argument. In another study 29
he actually drew that curve, called it the linea proportionum and recognized it as
belonging to the same class as the quadratrix:
The line of proportions is io be put in the same class as the quadratrix for it
is generated by two motions, one circular and one straight, which are not
subject to each other. (A.T. 10 pp. 222--223)
It is not clear how DESCARTES got the (wrong) idea that the line of proportions
is generated by a combination of a straight motion and a circular motion; perhaps
he was thinking only of the quadratrix when he mentioned these motions. The
figure in the published text suggests that he had no clear idea about the form of
the curve.
There:is no evidence that DESCARTESbefore 1637 actively studied transcendental
curves other than the quadratrix and the spiral. But shortly after the publication
of the Gdom~trie we find DESCARTESdiscussing the logarithmic spiral in a letter to
MERSENNE3° and another logarithmic curve in connection with one of the prob-
lems set by DEBEAUNEal. Around this time he also studied the cycloid. 3z In
the case of DEBEAUNE'S problem DESCARTESdid not explicitly recognize that the
curve was connected with logarithms~ although he may well have seen the link.
He worked out two motions which together describe that curve and he found that
these two motions
are so incommensurable that they cannot be regulated by each other in an

exact way; and therefore that this line belongs to those which I have rejected
from my Geometry as being only mechanical. (A.T. 2, p. 517)
F r o m the little information we have, then, it seems that before the publication
of the Gdorndtrie DESCARTESmay have had the idea that by rejecting the quadratrix,
the spiral " a n d the like" (G p. 317) he was not really rejecting any interesting curves
but only those originating from motions which involve the relation between
curved and straight lines. Shortly after 1637 he came upon several other "non-
geometrical" curves; some of them obviously did not depend on the relation
between curved and straight, and some of them were indeed quite interesting.
28 A.T. 10, pp. 77-78; the study dates from before December 1618.
29 A.T. 10, p. 222-223, from 1619-1621.
30 DESCARTES to MERSENr~ 12-9-1638; A.T. 2, pp. 352-362, in particular p. 360.
3t For DESCARTES' solution of DEBEAUNE'Sproblem see his letter to DEBEAUNEof
20-2-1639; A.T. 2, pp. 510-519, and C. J. SCRIBA, "Zur L6sung des 2. Debeauneschen
Problems durch Descartes", Arch. Hist. Ex. ScL 1 (1960-1962), pp. 406--419.
32 M E R S E N N E mentioned the cycloid and ROBERVAL'Sstudies on the quadrature
in his letter to DESCAaT~S of 28-4-1638 (A.T. 2 pp. 116-122). In his answer of 27-5-1638
(A.T. 2 pp. 134-153) DESCARa'~S said that he had never thought of the curve before
(po 135). He discussed the curve and its properties in several subsequent letters to MER-
SENNE.
326 H.J.M. Bos
10. Once more: Descartes' programme
10.1 W.e have seen that in DESCARTES'programme for geometry as expounded

in the Gdomdtrie there was a contradiction in the criteria for the geometrical
acceptability of curves. On the one hand DESCARTES claimed that he accepted
curves as geometrical only if they could be traced by certain continuous motions.
This requirement was to ensure that intersections with other curves could be found,
and it was induced by the use of the curve as means of construction in geometry.
On the other hand DESCARTES stated that, under certain conditions, curves re-
presented by pointwise constructions were truly geometrical. Pointwise construC-
tions were related to curve equations in the sense that an equation for a curve di-
rectly implied its pointwise construction. Pointwise construction was used pri-
marily for curves that occurred as solutions to locus problems.
The link between the two criteria is DESCARTES' argument that pointwise
constructible curves can be traced by continuous motion. We have seen that that
argument, a n d hence also the link, is very weak (cf. Section 9.1). This makes
the criteria themselves all the more interesting. DESCARTESneeded both criteria;
he could not restrict his attention to one of them. On the one hand he could not
keep strictly to the criterion'of continuous motion because he wished to regard
all loci for problems of PAPPUS as geometrical curves. In studying the five-line
locus he came upon such a locus for which he could not, or would not, give a
construction by continuous motion (cf. Section 6.1 and Note 21). Certainly he
could not prove in general that all curves with algebraic equations can be traced
by continuous motion. On the other hand~ DESCARTESalso could not simply state
that a curve is geometrical when it can be constructed pointwise. Pointwise con-
struction presupposes means of construction and in DESCARTES' system these
means can only be curves traced by continuous motion. Hence if he would adopt
pointwise constructibility as a criterion for a curve to be geometrical, DESCARTES
would still have to show that the necessary curves for performing this pointwise
construction can be traced by continuous motion. This he also could not do in
general°
Why then did DESCARTES not cut this Gordian knot in the most obvious
way~ namely by defining geometrical curves as those which admit algebraic
equations 9 Why did he not simply state that all such curves are acceptable means
of construction and that the degrees of their equations determine their order of
simplicity? That principle would have removed the contradictions mentioned
above. But DESCARTESdid not accept the principle. In order to understand why
we have to look at the development of DESCARTES' ideas on geometry.
10.2 Even in 1619 DESCARTEShad a programme for his geometrical research.

We know this from what he wrote in his letter to BEECKMANof 26 March 1619:
I hope to prove (-) that certain problems can be solved with straight and
circular lines only; that others can only be solved with other curved lines
which originate in one single motion, and which therefore can be traced by
the new compasses, which I do not think are less certain and geometrical
than the ordinary ones with which circles are drawn; and that finally other
(problems) can be solved only by curved lines originating from different

motions that are not subordinate to each other and that certainly are only
imaginary (imaginariae); such a curve is the well known quadratrix. I think
that one cannot imagine problems that cannot be solved by at least these lines;
but I hope to be able to demonstrate which questions can be solved by the
first or the second method but not by the third; so that in geometry nothing
remains to be found. (A.T. 10 p. 157)
The passage shows that by 1619 DESCARTEShad formed the conception of geometry
which he adhered to all his life. He did not consider geometry primarily as an
axiomatic, deductively ordered corpus of knowledge about points, lines etc.,
but as the science of solving geometrical problems. Once all such problems could
be solved there was nothing further to do in geometry.
The compasses DESCARTEShad in mind here were two linkage machines 33
for solving problems or tracing curves° One of these we have met in the Gdornd-
trie; it is the machine illustrated in Figure 3. It was designed to find mean pro-
portionals between two given lines segments. The similarity of the triangles in-
volved immediately yields
YB:YC= YC:YD= YD:YE= YE:YF= YF:YG= YG:YH. (10;1)
Hence to find, for instance, two mean proportionals between YA and some given
line 2, one opens the compass until YE = 2; YC and YD are then the required
proportionals. Alternatively, one can first trace the curve AD by continuously
opening the compass. Then when2 is given, one intersects AD with the circle with
diameter YE = 2o The point of intersection is D; YC (the abscissa of D) and YB
are the required proportionals. In the same way the curve A F traced by F serves
to determine four mean proportionals.
The other compass was not mentioned in the Gdomdtrie. Just as the first com-
pass was based on a very simple geometric device to find proportionals, this one
employed a very simple device for dividing an angle into any given number of
equal parts. For the trisection of the angle the machine is as in Figure 16. There
are four rulers AB, AC, AD and AE, all pivoted at A. On each of them, at fixed
Fig. 16
33 A.T. 10, p. 234 and p. 240.

328 H.J.M. Bos
and equal distances from A, there are adjusted links FG, 11t, KG, and LtL all
equal in length° FGand KG are joined such that G can move along ruler AC; I H and
L H are joined such that H can move along ruler AD. If a given angle o~has to be
trisected the compass is opened until ~ BAE is equal to o~; then ,sY.BAC is equal
to ~- a. Alternatively (see Figure 17) let the point G trace a curve MG by opening
the compass while leaving ruler AE fixed. If angle o~has to be trisected one draws
the angle at A with one arm along AE, chooses a point F" on the other arm such
that AF' = A F and draws a circle around F" with radius FG. Through the inter-
section G' of the circle and the curve one draws AG'. Then ~ F A G ' = ~ ~ F'AE
~_~_1~.
Fig. 17
Around this time DESCARTEShad experimented with other compasses -similar

to these, and in his letter to BEECKMANhe mentioned that he had devised compasses
for the construction of all types of cubic equations. It was nothing new to use in-
struments such as DESCARTES' compasses for various special constructions;
they were used in classical mathematics, and PAPPUS~ Collections mention several
such machines. Also the idea of considering the curves traced by these machines
was not new; the eonchoid of NtCOMED~ for instance, is a curve traced by a
special kind of instrument for certain constructions, the so-called neusis construc-
tions 34. It should be noted~ however, that in formulating his programme DES-
CARTESconsidered the curves themselves rather than the compasses as the means of
construction.
10.3 If we compare the programme which DESCARTES outlined in his letter

of 1619 to BEECKMANwith the programme of the Gdom~trie we find significant
differences. These differences concern the role of algebra and pointwise construc-
tions. We see that by 1619 DESCARTES'programme contained the following ideas:
Geometry is the science of solving or constructing geometrical problems° Construc-
tion by means more complex than the circle and the straight line (the compass
34 See the chapter "On the problems known as neuseis" in The works of Archimedes
(ed. T. Lo HEATH,Dover edition), pp. c-exxii, in particular po evil. For the conchoid see
also Figure 5.
and the ruler) need not be less geometrical. Curves that can be traced by one single
continuous motion such as that provided by the compasses are acceptable means f o r
geometrical construction. There are also problems that can be constructed only
by curves traced by a combination of motions that are not subordinate to each
other. Such curves are "imaginary"; an example is the quadratrix. All problems
can be solved with such curves, but DESCARTES wishes to classify the problems
that can be solved by acceptable geometrical curves. These elements of the pro-
gramme of 1619 were all to be found in the Gdomdtrie as well. But several points
of the programme of the Gdomdtrie were still lacking in 1619. Most notable was
the absence of algebra. It is true that DESCARTESenvisaged geometrical solutions
of algebraic equations by means of compasses, but algebra did not yet play a role
in the classification of geometrical means of construction according to their sim-
plicity, nor in a method for finding the simplest possible constructions. It seems
likely.that by 1619 DESCARTES envisaged classifying the constructing curves
according to the simplicity of the compasses used to draw them. A trace o f this
is found in the G~omdtrie where DESCARTES, in discussing the compass for mean
proportionals, said
I do not believe that there could be an easier method to find as many mean
proportionals as one wishes, nor one whose p r o o f would be more evident,
than to use the curved lines traced by the instrument X Y Z . . . . (G p. 370)
But DESCARTESwent on to say that the curves traced by that instrument are o f a
higher class than necessary and that therefore they should not be used in a truly
geometrical solution o f the problem to find mean proportionals (G p. 371). We
may conclude that by 1637 DESCARTES' algebraic criterion for the simplicity of
curves, namely their class, defined via the degree of the equation, had replaced,
and indeed was in conflict with an earlier criterion f o r simplicity, namely the sim-
plicity of the compass and of the resulting p r o o f of the construction.
The other element not in the programme of 1619 concerned loci and pointwise
constructions. In 1619 DESCARTESdid not wish to introduce new curves in geometry
for purposes other than constructions. The problems r o b e constructed had one
solution or a finite number of solutions° In 1619 DESCARTES did not consider the
case where the solutions are infinite in number forming a locus which, by the
nature of the process of solving the problem, is constructed pointwise. Hence he
was not faced with the problem of whether or not such curves should be accepted
in geometry and according to which criteria.
10.4 Evidently, therefore, DESCARTES' programme of geometry changed be-

tween 1619 and 1637. The stages o f this change are in fact fairly well known, a5
32 See G. MILHAUD,Descartes savant (Paris 1921), Chs. 1, 3 and 6. JOHN SCHUSTER

has recently published a detailed study of the development of DESCARTES'ideas about
universal mathematics in relation to his metaphysics and his programmes for philosophy;
see J. A. SCHUSTER, "Descartes' mathesis unversalis; 1619-1628" in Descartes; philo-
sophy, mathematics and physics (ed. S. Wo GAUKROGER, Hassocks (Sussex) (Harvester
Press), 1980), pp. 41-96. He argues that one of the reasons why DESCARTESafter 1628
abandoned his programme for universal mathematics as formulated in the Regulae
ad directionem ingenii was that he encountered difficulties in working out the geometrical
330 H . J . M . Bos
It was probably shortly after his letter to BEECKMAN of March 1619 and before
November 1620 that DESCARTES studied the construction o f problems through
the intersection o f conics and found the solution of all equations of third and fourth
degree through the intersection of a parabola and a circle. This must have given
him the idea that the conics are the class o f constructing curves immediately
following the circle and the straight line.
This idea may have led him to search for a construction of all equations of
fifth and sixth degrees through the intersections o f a circle with one special curve
more complicated than the conics. He succeeded in finding this construction;
the curve is the Cartes/an parabola; the construction is explained at the end of
Book 3 of the G~om~trie (G pp. 402-411 ; cf. Section 3.4)° But we do not know
the date of this discovery36o These results must have induced DESCARTESto con-
sider that the degree o f the equation of the curve, rather than the simplicity of
the tracing machine, was the criterion for the geometrical simplicity of curves
used in constructions. ~
The other new aspect, loci "and pointwise constructions, probably was in-
corporated into DESCARTES' programme in 1631 when GOLIUS suggested that he
might try his hand at the problem of PAPPUS. We know that Dw:SCARTESsolved
the problem in a number of weeks and that the solution appearing in the G~o-
m~trie is essentially the one he sent to GOLIUS in January 163237° This study must
have turned DESCARTES' attention more to algebra, to the equation as embodying
all the information about the curve, to the need to incorporate all curves admitting
algebraic equations in g e o m e t r y a n d to the need to admit pomtwise construction
for curves.
10.S However, more important than the chronology of the changes in D~S-
CARTES' geometrical ideas is the fact that these changes explain the basic contra-
diction in DESCARTES' programme in the G(omdtrie. The programme of 1619
theory of equations. This fits in well with the chronology of DESCARTES'changing ideas
about geometry. Also SCt-IUSTER'Sstudy provides an illustration of DESCARTES'attitude
to contradiction and failure in his programmes (i.e. the programme of universal mathe-
matics) which seems to correspond well with what I find about DESC~TES' attitude to-
ward the failure of his earlier programme for geometry°
as It seems likely that this discovery was made later than 1628, for we have a note by
BEret:MAN about his interview with DESC~a~T~S in October 1628. D~SC~atTEShad ex-
plained to BEI~CKMANthe construction of the roots of any equation of the fourth degree by
the intersection of a circle and a parabola. BEECrdaAN noted that "M. DESCARTESmade
so much of this invention that he confessed never to have found anything superior himself
and even that nobody else had ever found anything better" (A.T. 10, p. 346)° It is not
likely that DESCARTESwould have made this kind of comment if by that time he had al-
ready known the general construction of the roots of equations of 5th and 6th degrees.
37 See DESCARTES'letter to GOLXUSof January 1632, A.T. 1, pp. 232-236. Dr SCART~S
refers to an "rcrit" sent earlier to GoLms; this "rcrit" has been lost. In the letter DEs-
CARTESadds a definition of classes ("genres") of curves. The definition is not clear and
the terminology is quite different from that used in the Gdom~trie. Still, from the further
indications in the letter, it seems likely that DESCARTEShad by that time found the essen-
tial elements of the solution of the problem of PAPPUS as it appeared in the G~ora~trie
and that the "~crit" sent to GoL~s contained a condensed version of this solution.
may have been impracticable, but it was consistent. It provided a demarcation

between geometrical andonon-geometrical constructions, and the criterion used
in that demarcation was a geometrical one: the constructions had to be performed
with machines that were generalizations of the ruler and the compass.
In the programme of 1637 algebra had become dominant. DESCARTES now
classified curves according to the degree of their equations and a large part of
the Gdomdtrie (especially the third book) is devoted to algebraical techniques
relating to the roots and coefficients of equations (reduction of equations, sign
rule, removal of terms from the equation, change of negative roots into positive
ones, etc.). But despite all this algebra, what had remained was DESCARTES'
conception that geometry was the science of solving geometrical problems by
the construction of points through t h e intersection of curves. Therefore the
main aim of the third book was the construction of roots of equations through
the intersection of curves.
This aim determined the structure of the third book and the nature of the
algebraic techniques presented in it. The reduction of'equations to other equa-
tions of lower degree was necessary for finding the construction by the simplest
possible constructing curves. The techniques relating to the roots and coefficients
of the equation served to reduce the equations to standard forms, for which DES-
CARTESthen gave standard constructions. For equations of third and fourth degrees
this was the construction by the intersection of circle and parabola; for equations
of fifth and sixth degrees the construction by the intersection of circle and Carte-
sian parabola.
Thus, although algebra occupied a dominant position in DESCARTES' pro-
gramme of 1637, it was the geometrical aim of the work that determined its struc-
ture and provided the motivation.
10.6 We now have the answer to the question raised in Section 10.1, namely
why DESCARTESkept the criterion of tracing by continuous motion for the geo-
metrical curves, and why he did not simply define geometric curves as those which
have algebraic equations. As we have seen, the whole structure of his G~om~trie
depended on the conception of construction by the intersection of geometrical
curves. For DESCARTES, these intersections were actually found or constructed
only if the curves could be traced by continuous motion. In that case one can
conceive clearly and distinctly that the intersections are found. If he were to re-
nounce his criterion of tracing by continuous motion and at the same time keep
to his programme of construction by the intersection of curves, he would have
to state as an axiom that for all curves having an algebraic equation the inter-
sections are given or constructible.
It is evident that DESCARTEScould not do this. An axiom which states that the
intersections of curves are constructible is by no means clearly and distinctly
evident, so it would not satisfy DESCARTES' criterion for accepting a statement as
a basis for further argument.
Moreover, by adopting this approach DESCARTEScould no longer claim that
he was doing geometry; he would be doing some kind of algebra. But that would
mean giving up the principal aim of his work; to bring order into the science of
geometry.
332 H. Jo M. Bos
Finally, the whole structure of the Gdomdtrie, which was based on finding
the simplest constructing curves for a given problem, would lose much of its
meaning. If the intersections of all algebraic curves are by axiom constructible,
there is no evident reason for findifig the simplest curves for a given problem~
and hence there is not much point in finding constructions for roots of equations.
The roots of an e q u a t i o n x ~ + ax ~-~ -+- . . . . 0 are the intersections of the curve
y = x n q- ax ~-~ ÷ .o. with the straight line y -----0; thus they are already given
as intersections of curves with algebraic equations.
We see that DESCARTEScould not give up his definition of geometrical curves
by continuous motion because then he would have lost the claim of doing geometry
and hence the rationale of the whole structure of his work would have been
destroyed.
11. Conclusion
11.1 As I hope I have shown, the representation of curves is the key to under-
standing the structure of DESCARTES' Gdomdtrie and its underlying programme.
Although there were contradictions in the structure and the programme, there
was an underlying unity of vision. DESCARTES had this vision as early as 1619,
but it did not find its clear expression until the Gdomdtrie of 1637o According
to this vision geometry can and should be structured, and the bewildering jumble
of problems, methods and solutions, in which it is impossible to know where the
problems end and the solutions begin, can and should be cleared up. DESCARTES'
view, in short, was that geometry concerns a surveyable, "orderable" collection
of well defined problems, well defined also in the sense that there are clear criteria
of adequacy for their solutions. DESCARTESleft it to his successors to work out the
programme, to find its limitations and to come to terms with its contradictions.
Appendix
On the curves that occur as loci o f problems o f PAPPUS

In the passage quoted in Section 2.3 DESCARTES asserts that every curve can
arise as the locus of some problem of PAPPUS. In this Appendix I shall discuss
the possible interpretations of this assertion and show that in each case the asser-
tion is wrong° First of all I shall derive some properties of polynomials occurring
in the equations of loci of problems of PAPPUS.
Let V~ be the (6n + 1)-dimensional real vector space consisting of vectors
v = (al, bl, cl, a2, b2, c2 . . . . . az~, bz~, cz~, c~).
Consider the mapping P~ defined by
P~(v) (x, y) = (aix -t- biy -I- ci) -- ~ [ I (aix + biY -t- ci).
i~l i~n+l
P~ maps V~ into the space of polynomials in two variables with real coefficients.
If none of the factors aix + biy + ci is constant (i.e. if for all i, as ~ 0 or b~ ~ 0),
then
t'.(v) (x, y) = o
is the equation of the locus of a problem of PAPPUS in 2n lines. Let these lines be
called L~; they have as equations
aix + biy + ci = O.
Taking into account the possibility that some of the factors may be constant, we
see that
en(v) (x, y) = 0
is the equation of the locus of the following generalized problem of PAPPUS:
Given k + l lines, with k _<__n and l ~ n, find the locus of points in the plane
such that the ratio of the product of their (oriented) distances to the first k lines
and the product of their (oriented) distances to the last l lines is constant.
This is a natural generalization of the original problem of PAPPUS. All 2m-line
and ( 2 m - O-line problems with m =< n occur among the P~(v)(x, y). I shall
call a polynomial that can be written as P~(v) for some n and some v E V~, a Pappus
polynomial. PAX'PUS polynomials are those polynomials which can be written
as the difference between two polynomials each of which can be decomposed
into linear factors. DESCAR~S' assertion therefore is about whether or not every
polynomial can be written as such a difference.
I shall need a number of properties of PAPPUS polynomials, derived in the
following Theorems 1-3.
Theorem 1. Let v E V~ and let the lines L~ corresponding to the nonconstant
factors a~x + b,y + c~ be not all parallel to each other. Then there are points
(x, y) in the plane such that P~(v) (x, y) = O.
Proof. Because not all the lines are parallel, there is a pair of lines L , Lj with
i ~ n and j > n, which intersect each other° Let (xo, Yo) be their point of inter-
section. Then
a~xo + b~yo + c~ = ajx0 + byo + cj = 0
and therefore
P,(v) (xo, Yo) ----"0.
If we are dealing with a proper 2n-line problem, i.e. if all the factors in Pn(v)
are non-constant, the degree of Pn(v) will generally be equal to n. But because
Pn(v) is the difference of two polynomials it may happen that the terms of high
degree cancel each other, so that a 2n-line problem may lead to an equation o f
degree less than n. An example is the 8-line problem
( y + 1) (y --1) (x + l) (x -- 1) -- (y + 2) (y -- 2) (x + 2) (x -- 2) = O
which yields a quadratic equation
X2 + y2 ~__.5.
334 H. Jo M. Bos
Conversely, this means that if we are to find all PAPPUS polynomials of degree
n, we may not restrict attention to all P#(v), because such polynomials may also
arise as Pro(v) for some m > n. This possibility will be important in interpreting
and checking DESCARTES'assertion. In particular we shall need the following theo-
rem about the degree of Pn(v):
Theorem 2. Let v E Vn such that all factors aix + biy + ct in P#(v) are non-
constant and such that not all the lines L i are parallel. Then
n
- ~ =< degree P~(v) ~ n.
Proof. It is obvious that degree P~(v) =< n. To prove the other inequality I call
L~ . . . . . L~ the first set of lines and L~+x . . . . . L2~ the second set of lines, and I con-
sider first the case where the two sets have no line in common. If there is a line
in the first set which is not parallel to any of the lines in the second set, I take that
line as L I and proceed by taking k -----0 in the argument below. If there is no such
line in the first set, I choose (renumbering if necessary) LI . . . . . Lk from the first
set of lines and Ln+I . . . . . L~+k from the second set such that
LI///-2//o..//-L~,//L,+I//...//L,+k,
and such that either in the first set of lines or in the second set there are no more
lines parallel to LI. I assume that in the second set of of lines there are no more
than k lines parallel to L1 (in the other case I can switch the sets). Because not all
the lines are parallel there is such a set of lines LI . . . . . Lk, L,+I . . . . . L#+k with
n
0__<k<- T,
n
because if k were > -~-, a smaller set with the same properties could be chosen
from the remaining lines. If b~ ~ O, I substitute
~aIX ~ C~
Y= bl
in
n 2n
P.(,O (x, y) = H (a,x + b,y + c,) - ~, 11 (a~x + O# + c.O.
i=l i=n+l
The first product then becomes 0 (because a l x + bly + c~ ---- 0), and in the second
product the factors become
aibi -- albi blci -- bicl
x+
bl bl
For n + 1 --< i ~ n + k, the lines L~ are parallel to Lt but not equal to Lx ;
hence in that case
a~bI -- alb l = 0 and bxci -- btc t =i=O,
so the first k factors of the second product become constants e~ =~=0. The lines
L.+k+I . . . . . ~ are not parallel to L1, hence for n q- k ,-+- 1 _< i _< 2n
aibl -- alb ~ ~ O,
so the last n - - k factors o f the second product can be written as
Ax + g. £ o.
Hence
x, "bl = FI e, (f,x + g,),

iffin+! l~n+k+l
a polynomial in x of degree n - - k . But this means that

degree Pn(v) (x, y) >= n -- k.
Also
n
n--k >=-~,
gO
n
degree P.(v) (x, y) =>
I f b l = O, Jt can be proved analogously (substitution x = -~Zt~) that P. (v) (--~-, y )

is a polynomial in y o f degree n -- k, so that also in that case
n
degree P.(v) (x, y) >= n -- k >=-~
Finally, if some of the lines o f the first and second sets coincide, let these lines be
L1 . . . . . La, that is, L i = L.+l, I-a = L.+2 . . . . . La = L.+d. Then, for 1 __< i _< d,
aix + bff + e~ = ~i(a,,+ix + b.+iy + c,,+i), ~i =~ O.
We then have
d
P.(v) (x. y) = 1-[ (a~x "k- biy -k- ci) "
iffil
• (alx q- bty Jr" c,) -- ~x 1-[ ~[-I l-I (aix + biy q- ci) •
i 1 i~l i~n+d+l
The factor between brackets is a PAPPUS polynomial P._d(w) for which the first
n--d
and second sets of lines have no lines in common. Hence its degree is _--_ ,
so
n--d n
degree P,(v) >= d +
2 =2"
This completes the proof o f Theorem 2.
DESCARTES noted that a case where all lines are parallel is exceptional. Theo-
rems 1 and 2 do not apply in that case. I shall therefore discuss it separately.
336 H . J . M . Bos
Let v E V, be such that none of the factors a~x + b~y -q- c i is constant and that
the 2n lines L~ are all parallel. Then the equations of the lines L i can all be written
as
with
z=alx--kbly.
Hence
fi 2n
P.(v) (x, y) = (o,: + fit) -- o, I-[ (~,,z + ill) = F(z),
i=1 i=n+l
where F(z) is a polynomial in z formed as the difference between two n th degree
polynomials in z, each having n real roots. But it may happen that F(z) itself has
n
no real roots or that its degree is < ~--. Hence the analogues of Theoremso 1 and 2
do not apply in this case. We note that in this case o f 2n parallel lines, the surface
in R a
w - - e.(v) (x, y) = e(z)
is a ruled surface.
As to the degree of F we prove:
Theorem 3. Let v E Vn be such that none of the factors aix + biy + ci is constant
and such that the 2n lines Li are all parallel. If P,(v) has degree m < n, then there
is a w E I'm, such that
Pn(v) = Pm(W)
and the lines corresponding to w are all parallel.
(In other words, if a problem of PAPPU$ in 2n parallel lines yields an equation
of degree m, then there is a problem of PAPPUS in 2m parallel lines which yields
the same equation.)
Proof. Because
n 2n
?.(v) (x, y) = I-[ (~,z + 8,) - ~ I-I (~,z + ¢,) = F(z),

1=1 l=n+l
we have to prove that every polynomial F(z) o f m th degree in one variable can be
written as the difference between two polynomials G(z) and H(z) of m th degree,
each having only real roots. Consider a segment [a, b], and let IF(z)[ < K for
z E [a, b]. Choose points in [a, b] as follows:
a = do < cl < dl < c2 < d2 < .o. < Cm-l < dm-t < cm < d., -----b
and consider
m
G(z) = II ( z - c,).
i=l
Clearly G(d0 ~ 0, and between each pair d~, d~+I, G has a change of sign. Choose
such that [~G(d3[ > g for all d/, and consider the polynomial of mth degree
tt(z) = F(z) + c~G(z).
Because lF(d/)l < K a n d laG(d/) i > K, H(4) has the same sign as G(4); thus H
has a change of sign between each pair di, d,.+l. Therefore all the roots of H are
real. The required decomposition is therefore
F(z) = -
Corollary of Theorems 2 and 3. If F(x,y) is a PAPPUS polynomial and degree

F = k , then there is an m ~ 2k and a v E Vm such that F = P=(v).
I now return to the interpretation of DESCARTES' assertion. If it is taken to
mean that every polynomial in two variables is a PAPPUS polynomial (and the
first lines of DESCARTES' assertion could be read thus), then the assertion is quite
easily proved wrong. A counter-example is
F(x, y) = x = + y2 + 1,
because there are no real points (x, y) satisfying F(x, y) = O; also w = F(x, y)
is not a ruled surface in R a, so by Theorem 1 and the remark preceding Theorem 3,
F(x, y) cannot occur as P~(v) for any n and v E Vn. There is no evidence that DES-
CARTESwas aware of this complication in connection with polynomials without
real roots° But, in view of the later part of the passage we are discussing, where
DESCARTES speaks about curves, we must conclude that he did not have poly-
nomials in mind but rather polynomial equations of curves, i.e. polynomials F(x,y)
such that the set {(x, y) I F(x, y) = 0) is not empty.
Even with this restriction the assertion is wrong. I shall prove this with a
dimensional argument. In particular:
Theorem 4. If n > 21, there are polynomials F(x, y) of degree n, such that F(x, y)
= 0 is a non-empty real curve and F(x, y) ---- 0 does not occur as the locus of a
problem of PAPPUS.
Proof. Let Q~ be the space of polynomials in two variables of degree ~ n. The

dimension of Q~ is (n + 1) (n + 2) Let O~ ( Qn be an open subset of Q~ consisting
2 ~*
of polynomials F(x, y) such that F(x, y) = 0 is a non-empty real curve; the di-
(n + I) (n + 2)
mension of On is also 2 . [It can be seen from the following that
÷
such a subset On exists. Let O(1,0) be the set of polynomials F with F(I, 0) > 0
and let O('-l,0~ be the set of polynomials F with F(--1, 0) < 0. Both sets are open
subsets of Q~ and their intersection O is not empty (F(x, y) = x belongs to it).
If FE 0, then F(1, 0) > 0 and F(--1, 0) < 0 so that F(x, y) = 0 must be a non-
empty real curve.] Suppose that all FE On are PAPPUSpolynomials. Then, accord-
ing to the corollary to Theorems 2 and 3, each FE On must occur as P~(v) for
some v E V~. O~ ( Qn C Q:~, so o~ must be a subset of the image of P2~ in Qz~.
Hence
dim On _--<dim image P~ ~ dim V2n,
that is
(n -+- I) (n + 2)
"__ 1 2 n + 1.
2
338 H . I . M . Bos
This condition is not satisfied if n > 21, so in that case there must be polynomials
in On which are not PAPPUS polynomials. (The estimate of dimension can easily
be sharpened e.g. to n > 13.) This completes the proof that DESCAR~S' assertion
is wrong.
Apparently NEWTON was the first to question DESCARTES' assertion that all
curves occur as loci of problems of PAPPUS and to give a proof that it is wrong
(cf. The mathematical papers of Isaac Newton (ed. D. T. WHIrr.SIOE, Cambridge
(1967-) vol. 4 (1971), pp. 340-344). The basic idea in this proof is also a dimensional
argument; NEWTON compared the number of free coefficients in the choice of
n lines and the number of coefficients in a polynomial of degree n, and found that
for n >---5 the latter is larger than the former, so, NEWTON concluded, DESCARTES
was wrong. NEWTON did not mention the occurrence of polynomials F Such that
F(x, y) = 0 is empty, nor did he consider the possibility that an equation F(x, y) --- 0
o f n th degree might occur as the locus of a problem of PAPPUS in more than 2n lines.
As we have seen, these tWO aspects make the question much more complicated.
H. G. ZEtrrnEN states, in his Geschichte der Mathematik im 16. und 17. Jahr-
hundert (1966, p. 210), that DESCARTES'assertion is wrong, but he does not provide
a proof.
Acknowledgement. A large part of this article was written during my stay in at the
Institut des Hautes l~tudes Scientifiques at Bures-sur-Yvette, France in November
1979. I wish to thank that institution for its support and hospitality.
I a m indebted to K. A2~mEN (Aarhus), W. VAN DER KALLEN(Utrecht), L VAN
MAANEN(Utrecht), A° Go MOLLAND(Aberdeen), F° OORT(utrecht), J. R. RAVETZ(Leeds)
and D. T. WHrrr.SIDE (Cambrige) for making stimulating comments on an earlier draft
of this article and to W. VAN D~R K.ALLENfor providing crucial ideas for the proofs
in the Appendix.
Mathematical Institute
Boedapestlaan 6
Utrecht, Netherlands
(Received duly 16, 1980)

On The Representation of Curves in Descartes' Géométrie

Uploaded by

Copyright:

Available Formats

On The Representation of Curves in Descartes' Géométrie

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

On The Representation of Curves in Descartes' Géométrie

Uploaded by

Copyright:

Available Formats

On the Representation of Curves in

1. Introduction. Curves and their representation in 17 th century mathematics

Finally, the curve with modern equation

i R. DESCARTES,La Ge~omEtrie, one of the essays in his Discours de la mdthode pour

2o The problem of Pappus

2.1 In his Gdomdtrie DESCARTES expounded a new programme for dealing

2.3 This classification refers to the construction of the locus. DESCARTES

I shall return to this classification in Section 3. It leads to the following classifica*

2.4 DESCARTES'solution of the problem of PAPPUS supplies a good illustration

7 See the Appendix.

3. Descartes' Programmefor Geometry

3.1 In his Gdomdtrie DESCARTESpresented a new approach to the solution

3.2 The first point of the programme--differentiating between the curves

9 See the article by 1V[OLLANDcited in Note 3.

that in fact--although DESCARTES never explicitly said so--he wanted to regard

lo DESCARTEStended to underestimate the dangers of extrapolating mathematical

4. The representation of curves in Descartes' G~omdtrie

4.1 DESCARTES dealt with the fundamental question of his programme at

understanding of its nature; significantly, DESCARTES combined the word

4.2 Since DESCARTES considered curves primarily as traced by continuous

5. Curves described by continuous motion

It is a system of linked rulers. The rulers YX and Y Z are connected in Y by a

1: DESCAgTES~Euvres (see Note 1) vol. 6, p. 19.

A ruler G L pivots at G. It is linked at L with a device N K L which can be moved

(where G A = a, K L = b, N L = c, C B = y and A B --- x) and concluded that it

"Cartesian parabola". 13 Further on in the second book (G pp. 335-337) he showed

la See G. LORIA, Spezielle algebraische u n d transzendente ebene Kurven, Leipzig

Figure for Footnote 14

t5 DESCARTESclaimed that it is easy to prove this by actual calculation. But this

The quadratrix 17 (see Figure 8) can be described by a combination of separate

x7 Often called the quadratrix of DINOSTRATU$,although the names of HiPPt~s and

plains why the first rectifications of algebraic (i.e. for DESCARTESgeometrical)

6. Pointwise construction of curves

zx DESCARTES'cryptic description of the curve is as follows: The curve is such that

as the equation for the required curve° Taking

(i.2 Dr:.SC~TES said more about the acceptability of these constructions in

to each other can be geometrically constructed. There is a similar construction for

6.3 After this passage DESCARTES repeatedly used pointwise construction

7. Curve construction using string

7.3 Further on in the Gdomdtrie DESCARTESused string constructions as an

8. Equations of curves in the G~om~trie

8.1 We have seen that DESCARTESused three different kinds of representation

9.2 It is noteworthy that DESCARTES'basic argument in rejecting the transcen-

26 A. B. I~MPE, "On a general method of describing plane curves of the n th degree

are so incommensurable that they cannot be regulated by each other in an

10. Once more: Descartes' programme

10.1 W.e have seen that in DESCARTES'programme for geometry as expounded

10.2 Even in 1619 DESCARTEShad a programme for his geometrical research.

(problems) can be solved only by curved lines originating from different

33 A.T. 10, p. 234 and p. 240.

Around this time DESCARTEShad experimented with other compasses -similar

10.3 If we compare the programme which DESCARTES outlined in his letter

10.4 Evidently, therefore, DESCARTES' programme of geometry changed be-

32 See G. MILHAUD,Descartes savant (Paris 1921), Chs. 1, 3 and 6. JOHN SCHUSTER

may have been impracticable, but it was consistent. It provided a demarcation

On the curves that occur as loci o f problems o f PAPPUS

x, "bl = FI e, (f,x + g,),

a polynomial in x of degree n - - k . But this means that

I f b l = O, Jt can be proved analogously (substitution x = -~Zt~) that P. (v) (--~-, y )