You are on page 1of 5

Cardan and the Pascal Triangle

Author(s): C. B. Boyer
Source: The American Mathematical Monthly, Vol. 57, No. 6 (Jun. - Jul., 1950), pp. 387-390
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2307638 .
Accessed: 28/03/2013 23:42
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to
The American Mathematical Monthly.

http://www.jstor.org

This content downloaded from 62.204.192.85 on Thu, 28 Mar 2013 23:42:43 PM


All use subject to JSTOR Terms and Conditions

1950]

CARDAN AND THE PASCAL TRIANGLE

387

of an intersectionpoint theoremrequires that there be only a finitenumberof


intersectionpointsin the configuration.Thus it is seen that an intersectionpoint
at the maximumdistance fromany particularpoint of the plane can be so isolated. Note that the definitionof a non-trivialintersectionpoint theoremalso
requiresat least threelines on a point and threepointson a line so that we have
at least threelines of the configurationconcurrentat the isolated point and cutting the axis in ordinarypoints. With the configurationof the theoremthus established in a non-desarguesianplane, it is seen that the argumentsof the preceding section apply as beforeto show that the theoremis false in every one of
the geometriesof class U.
Thus farwe have only considerednon-trivialconstructibleintersectionpoint
theoremsof euclidean geometry.However, we can show that possible theorems
which do not hold in euclidean geometrycannot hold in our non-desarguesian
planes either.For, considerthe configurationof such a theoremconstructedin a
euclidean plane. Establish the configurationin a non-desarguesianplane by
drawingan axis isolatingit completelyin the lowerhalf-plane.Then, ifthe theorem were valid in the non-desarguesiangeometry,it would hold in euclidean
geometryalso. We may now state
THEOREM6. All non-trivialconstructible
intersection
pointtheorems
arefalse in
each ofthegeometries
ofclass U.
This yields, by elementarymeans, the result of Moufang (l.c.) to the effect
that all non-trivialconstructibleintersectionpoint theoremsare independentof
the plane axioms of projectivegeometry.It is interestingto note that a common
geometricmodel (Moulton's geometry,for example) can be exhibited to show
this independence forall non-trivialconstructibleintersectionpoint theorems,
and that each geometryof class U will serve as such a model.

CARDAN AND THE PASCAL TRIANGLE


College
C. B. BOYER, Brooklyn

It is well known that the arithmetictriangle,in one or another variant of


the form
1
2

1
1
a

4
. . . .

4
1
6
h fau . . T. .

appearedlongbeforePascal composedhisfamousTraitgdu triangle


arithm9tique

This content downloaded from 62.204.192.85 on Thu, 28 Mar 2013 23:42:43 PM


All use subject to JSTOR Terms and Conditions

388

CARDAN AND THE PASCAL TRIANGLE

[June,

(published posthumouslyin 1665). The device may have been known to Omar
Khayyam (c. 1100), it appeared in China in 1303, and it was published in
Europe by Apian in 1529. [1] Similar formsof the trianglewere given before
Pascal by numerous men, including Stifel (1544), Scheubel (1545), Peletier
(1549), Rudolph (1553), Tartaglia (1556), Stevin (1585), Girard (1629), Oughtred (1631), and Briggs (1633). [2] However, it seems still to be widely believed
[3] that "the triangulararray was investigatedby Pascal (1654) under a new
form,substantiallyas follows":
1...

4...

6.*.

4--.

1.*So significantdid Cantor, Wieleitner,and Tropfkeregardthe differencein form


that they declined to recognize any dependence of Pascal upon the earlier instances. [4] There also is a general impressionthat to Pascal is due the first
study of the relationshipsexhibited by the triangle and their application to
questions in the theoryof probability. [5] It is the purpose of this note to call
attention to the work of one whose name has not been associated with the
arithmetictriangle but who anticipated Pascal with respect both to the form
and the study of the triangle.This man, perhaps the greatest mathematician
of his day, was JeromeCardan (1501-1576).
The Ars magna, which appeared in 1545, contains no referenceto the arithmetic triangle;but in 1570 Cardan published his Opus novumde proportionibus,
and in this work the Pascal triangleappears in both formsand with varying
applications. In connectionwith the problem of the determinationof roots of
numbers, Cardan used the familiarearlier form,citing Stifel as the putative
discoverer. Here he gave the numbers in the triangle through n = 17, and he
pointed out the relationship,known to Stifel,equivalent to

{m8

(n

) (n+

m + 18

n+ 1

[6]. Later on in the Opus novumthe question of combinationsand probabilities


is taken up, and then Cardan gave the arithmetictriangle,throughn= 11, in
virtuallythe formlater made famousby Pascal. [7] Even in the case of Cardan,
however,this formwas not entirelyoriginal,forin 1556 Tartaglia had published
it in a somewhat similar square array, [8] and had used it in determiningthe
ofa binomial
in the expansionof the twelfthpower (cubo-censo-censo)
coefficients
of
the
newer
with
arrangement
In
connection
(una quantitadivisia in due parti).
and
name
his
by
known
formula
the
familiar
the triangle Cardan reiterated
enunciated by him many years before-the total numberof differentcombina-

This content downloaded from 62.204.192.85 on Thu, 28 Mar 2013 23:42:43 PM


All use subject to JSTOR Terms and Conditions

1950]

CARDAN

AND THE

PASCAL

TRIANGLE

389

objects, taking as many at a time as


tions which can be formedfromn different
one pleases, is given by 2 - 1. As an illustrationof the usefulnessof his result,
he determinedthe numberof possible combinationsof 25 differentobjects, the
result being 225_-1=16777215. Following this, Cardan gave something more
important which seems to have been overlooked by historiansand which incorrectlyhas been ascribed to Pascal. [9] This is the rule of succession, equivalent to
{n

r/

n- r+

1(nA

\r-1

The rule is not given symbolicallybut is expressedratherawkwardlyas follows:


To obtain the third element in the row correspondingto n =11, for example,
subtract one fromeleven, divide by two, and then multiply by eleven. The
result,55, is the numberdesired.To obtain the next number,subtract two from
eleven, divide by three,and multiplyby 55. Continuingin this manner,Cardan
obtained all of the elementsin the sequence of combinationsof eleven objects.
[10] Had Cardan applied his rule to the expansion of binomials,he would have
anticipated the binomialtheoremforpositiveintegralpowers.Instead he emphasized the connectionbetweenthe numbersof the arithmetictriangleand "mixed
proportions,"i.e., progressionsof higherorder,and the applicabilityof these to
musical theory.There can be little doubt but that Cardan, like Tartaglia, was
aware that the elements in the triangle are coefficientsin the expansions of
binomials. That a clear-cutstatementof Cardan's rule of succession as applied
to the binomial theoremshould have been delayed forabout anothercenturyis
one of the anomalies in the development of mathematics,and that the arithmetic triangle should be named for Pascal rather than for one of his many
anticipatorsis largelyan accident of history.
References
(2 vols.,New York, 1923-1925),vol. 2, pp. 508-511.
1. D. E. Smith,Historyofmathematics
thisMONTHLY, 56, 1949,147-157.
Cf.J. L. Coolidge,The storyofthebinomialtheorem,
de Pascal, Annalesde la
sur le trianglearithm6tique
2. HenriBosmans,Une notehistorique
de Bruxelles,31 (1906-1907),65-72. Cf. Smith,loc. cit.,and FlorianCajori,
Soci6t6Scientifique
(2nded., New York,1931),pp. 76, 183,187.
Historyofmathematics
3. Smith,op. cit.,and Coolidge,loc. cit.
uber Geschichteder Mathematik(4 vols., Leipzig, 19004. MoritzCantor,Vorlesungen
derMathematik(2 vols., Leipzig,1908-1921),
Geschichte
1908),II, 685-688;HeinrichWieleitner,
(2nded., 7 vols.,
derElementar-Mathematik
Tropfke,Geschichte
vol. II, part1, p. 93; Johannes
Berlinand Leipzig,1921-1924),VI, 34-39.
ofthemathematical
A history
theoryof
5. Cajori,Cantor,and Smith,op. cit.;I. Todhunter,
(Cambridgeand London,1865),p. 17 f.
probability
6. JeromeCardan,Operaomnia(10 vols.,Lugduni,1663),IV, 529.
7. Ibid.,IV, 557.
8. NiccoloTartaglia,Generaltrattatodi numeriet misure(part II, Vinegia,1556),book 1,
imXVI. undXVII. Jah1r
derMathematik
fol.17; bookII, fols.69-73.H. G. Zeuthen,Geschichte
Libri, Histoiredes sciences
hundert(Germaned., Leipzig,1903), pp. 102, 169,and Guillaume-

This content downloaded from 62.204.192.85 on Thu, 28 Mar 2013 23:42:43 PM


All use subject to JSTOR Terms and Conditions

390

THE FINITE FOURIER SERIES AND ELEMENTARY

GEOMETRY

[June,

en Italie(4 vols.,Paris,1838-1841),III, 158,mention


theworkofTartagliabutnot
math6matiques
thatofCardan.
9. Encyclop6diedes sciencesmath6matiques,
I (1), 1 (1904), pp. 83-84. Tropfke,loc. cit.,
ascribesCardan'sruleofsuccessionto Briggsin 1633.
10. Operaomnia,IV, 558.

THE FINITE FOURIER SERIES AND


ELEMENTARY GEOMETRY
I. J. SCHOENBERG, University
of Pennsylvania

Every freshmanknows, or soon learns, that the periodic sequence xo=l


Xl=-1,

Xn= (-

has a generaltermxS, whichmay be writtenas


X3= -1,...
X2=1,
1) . This is a very special case of representationof a periodic sequence

by the so-called finiteFourier Series. The finiteFourier series is the analogue


for periodic sequences of the ordinary Fourier series expansion of periodic
functions.Its only mentionin our textbook literatureis under the heading of
practical harmonic analysis, or trigonometricinterpolation,as found in books
on applied mathematicsor numerical methods; and yet, the range of applicationsof the finiteFourierseriesextendsbeyond thisimportantpractical problem.
It is intimatelyrelated to the Gaussian sums and has been used for number
theoreticpurposes by Eisenstein and morerecentlyby H. A. Rademacher. The
followinglines presentthe basic propertiesof'the (complex) finiteFourier series
stressingits geometricsignificanceand followedby a few applications to extremal problemsof elementarygeometry.
?1. THE FINITE FOURIER SERIES

1. The finiteFourierseries. We recallthe generalproblemof polynomial

interpolation:If we are given k distinctcomplex numberswo,w1, * * *,IWk-1, and


a second set of k arbitrarycomplex numberszo,z1, . . ., Zk-1, then there is one
+?k 1Xk-l satisfyingthe equaand only one polynomial P(x) =o+?jx+
tions P(co,) =z, (v=0, 1, * *, k-1) or
(1)
Zp)o

1lWp +

~~~~~2 ~
P2WP

k-i1L
Pk-1WP

(V =

12

. . .

k - 1). [1]

Indeed, the systemof linear equations (1) in the unknownsRphas a non-vanishing,(Vandermonde) determinantand has thereforea unique solution. We obtain
the finiteFourier series (abbreviated in the sequel to f.F.S.) if we choose the
numberscopto be the kth roots of unity; so forthe remainderof this paper we
shall assume that

(2)

w,= e2liPk

(v = 0, 1,

k
k-1).

This equation definescopforall integralvalues ofv,a factoccasionally used later.

This content downloaded from 62.204.192.85 on Thu, 28 Mar 2013 23:42:43 PM


All use subject to JSTOR Terms and Conditions