The Rise, Fall, and Possible Transfiguration of Triangle Geometry: A Mini-History

Author(s): Philip J. Davis

Source: The American Mathematical Monthly, Vol. 102, No. 3 (Mar., 1995), pp. 204-214
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2975007 .
Accessed: 07/12/2014 23:49

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 128.235.251.160 on Sun, 7 Dec 2014 23:49:45 PM

All use subject to JSTOR Terms and Conditions
 l

The Rise, FallSand Possible

Transfilgurationof Triangle Geometxy:
A Mini-history

Philip J. Davis
-

For Deborah TepperHaimo

In Friendship

"Es ist in der That bewundernswuerdig, dass eine so einfache Figur, wie das Dreieck, so
unerschoepflich an Eigenschaften ist. Wie viele noch unbekannte Eigenschaften anderer Figuren
mag es nicht geben." A.L. Crelle (1780-1855), Sammlung, v. I, 1821, p. 176.
[It is indeed wonderful that so simple a figure as the triangle is so inexhaustible in its
properties. How many as yet unknown properties of other figures may there not be?]

1. INTRODUCTION.In the great Encyklopaedieder MathematischenWissen-

schaften,put out underthe generaleditorshipof Felix Klein, there will be found a
hundred page article, completed in the Fall of 1914, on contemporarytriangle
geometry.(G. Berkhanand W. Fr. Meyer:NeuereDreiecksgeometrie, Vol. III AB
10, pp. 1173-1276.)On findingthis article,the reader'seyebrowsmay be elevated:
what on earth is triangle geometry?If the reader then goes to the index of the
Mathematical Reviews for enlightenment, he will not find the term triangle
geometry among the hundred or so subsets into which its coverage has been
partitioned.Differentialgeometry,yes; convexgeometry,yes; finite geometry,yes;
trianglegeometry,no. Yet, the Encyklopaediedevoted one of its majorarticlesto
this topic. F. Cajori,in his 1907historyof mathematics,devoteda half dozen pages
to it. So what is going on here? The subsumptionof this topic by anotherone? Or
the essential death of a topic?
What, in fact, is triangle geometry?Accordingto Encyklopaedieauthors,it is
not easy to define the subjectlogically,but it seems to boil down to this: given an
arbitrarytriangle,certainpoints (and lines and curves)are then determinedwhich
have remarkablepropertieswith respect to the triangle.Instancesof such points
are the incenter,the circumcenter,the orthocenter,and the center of gravityof the
triangle.By way of reminder,the first three are, respectively,the intersectionof
the internalangle bisectors,the intersectionof the perpendicularbisectorsof the
trianglesides, and the intersectiorlof the three altitudes.These four points were
studied in antiquity.
In 1803, a mathematicianby the name of Kluegel dubbedthese points thefour
distinguished(or remarkable)points of the triangle(merkwuerdigePunkte).In the
years that followed, a great many distinguishedpoints, lines, circles, and conics,
have been unearthed;so manyXin fact, that Berkhan and Meyer despaired of
counting them all. A point, line, circle, or conic, if sufficiently distinguished,
merited a special name, and so we have, as some further examples,the Fermat

THE POSSIBLE TRANSFIGURATION OF TRIANGLE GEOMETRY [March

204

This content downloaded from 128.235.251.160 on Sun, 7 Dec 2014 23:49:45 PM

All use subject to JSTOR Terms and Conditions
point, the Torricellipoint, the Gergonnepoint, the Brocardpoints and circle, the
Lemoine point and circle, the nine point circle, the Euler line, the symmedian
point, the Steiner point, etc. etc. In Kimberling("Central points...") will be
found a listing of more than one hundred such distinguishedobjects. Not only
special names were given, but also, quite understandably,histories of individual
distinguishedobjects were written (e.g., Mackay wrote short histories of the
symmedianpoint and of the nine-pointcircle. A recent book by Baptist presents
many facts about the developmentof trianglegeometryin the l9th century).
Accordingly,Berkhanand Meyer proposedas a definitionof trianglegeometry,
"the study of distinguishedpoints, lines, circles and conics of a triangle",leaving,
as far as I can see, the definitionof what is distinguishedor remarkableabout a
point to one's subjectivejudgment.
A somewhatmore sophisticateddefinition comes from Felix Klein himself (in
his famous ErlangerProgramm)and says that triangle geometryis the invariant
theory of five points under the projectivegroup. Perhaps this definition is less
vague, but I don't think that it catches the flavor of the subject as it has been
pursuedhistorically.
Triangle geometry as a distinguishedsubfield of mathematicsseems to have
emergedin the 1870'sin the writingsof E. Lemoine,and if one considersthe field
both forwardand backwardin time from that date, it will be found that many
distinguishedmathematicianshave contributeda little somethingto it. I leave to
the reader's subjectivejudgmentwhat constitutes a distinguishedmathematician.
Among the books that are wholly or partly devoted to this topic, one may cite
Alasia (written under the encouragementof the famous geometer Eugenio Bel-
trami, then the President of the Reale Accademia dei Lincei), and containing,
among other things, 566 metric formulasrelating to the triangle and its distin-
guished points! Some other books are: Altschiller-Court,Casey, Coolidge, Em-
merich, Johnson.In moments of euphoria,some of these authorsviewed triangle
geometry as the new and fulfilled Euclid, very much as the New Testament has
been claimed as the fulfillmentof the Old Testament.
One of the classic results of trianglegeometryis the nine point circle theorem,
which goes back, in part, to Poncelet in 1820. This theorem asserts that, given a
triangle,the followingnine points are concyclic:(i.e., lie on one circle) the three
side bisectors,the three altitude feet, and the three midpointsalong the altitudes
from the verticesto the orthocenter.This is only one of the remarkableproperties
that this circle has; for example,it is tangent to the inscribedcircle and the three
escribedcircles of the triangle.When one comes across this theorem in geometry
for the first time, there is a certainsurpriseassociatedwith it. One gets the feeling:
what wonderful coincidences! However, there is rlothingthat dissipates such a
feeling more quicklythan to see the geometrictheorem reduced to an algebraic
identityor to have it placed in a more general context.
There is a great deal more that can be said about the nine-pointcircle. Some
authoritieshave assertedthat there are no fewer than 43 distinguishedpoints lying
on the nine-point circle. The nine-point circle theorem has generated a small
mathematicalindustry.(See Gallatly). This, in itself, should now occasion no
surprise,consideringthat a few well-chosenaxiomssuch as those of group theory
can generate a majormathematicalindustry.Moreover,it will give the reader an
idea of the high regardaccordedthe nine-pointcircle to learn that some years ago,
the distinguishedanalystDame MaryCartwrighttold me that when she went up to
Cambridgeas a student(c. 1920),she was expectedto knowtwo differentproofs of
it.

1995] THE POSSIBLE TRANSFIGURATION OF TRIANGLE GEOMETRY 205

This content downloaded from 128.235.251.160 on Sun, 7 Dec 2014 23:49:45 PM

All use subject to JSTOR Terms and Conditions
 One treatise on triangle geometry(Emmerich)presents the subject from the
point of view of the Brocardpoints. I will refrainfrom givingdefinitionsbecause
the whole Brocardtheory never sent me into raptures.But I shall mention one
theorem that did; and when I was in high school, I cut my mathematicalteeth on
it.
"Napoleon's Theorem":On the sides of an arbitrarytriangle T, erect three
equilateraltrianglesoutwardly.Then:
(1) The three centersof the equilateraltriangleare thems.elvesthe verticesof an
equilateraltriangle.(Napoleon'sTriangle.)
(2) The three lines joining the vertices of the equilateral triangles to the
opposite vertices of T are collinear.They meet in a point P known as the inner
isogonic point of T. That is, P is the unique potnt in T at which the sides of T
subtendequal angles of 2pi/3.
(3) These three line segmentsare of equal length.
(4) Similarstatementswhen the equilateraltrianglesare constructedinwardly
on the initial triangle.
(5) The inner and outer Napoleon Triangleshave the same center.
(6) The areasof the inner and outer Napoleon trianglesdifferby the area of the
initial triangle.
These are merely a few of the remarkablethings that are associatedwith the
Napoleon configurationor its generalizations.(See, e.g., Court, pp. 105-107,
Sommerville,p. 165, Forder,p. 40, Hofstadter.And see Wetzel for a recent article
containingnew results and an extensivebibliography.)
I think that these examples should give the reader a good feeling for what
trianglegeometryis all about. I refer to the Encyklopaedie articleof Berkhanand
Meyerfor a numberof more complicateddevelopmentsand to older bibliographi-
cal references.For additionalrecent references,see Kimberling.
How were the theorems of triangle geometry discovered?The mathematical
literature,in general, is not often forthrightas to how its materialemerges. I can
only conjecturethat as with much of mathematics,it emergedfrom long hours of
"playingaround".Playingaroundsynthetically,in coordinatefree fashion,the way
that Euclid is writtenup; but also playingaroundwith algebra,trigonometry,and
rectangular,oblique,homogeneous,barycentric,trilinear,complex,conjugate,pro-
jective coordinates;all have been employedat one time or another. In textbooks
such as Altschiller-Court,whichare positionedas "advancedEuclid",the synthetic
approachis strong.
But another sort of playing around undoubtedlytook place. The figures of
trianglegeometrycan be drawnrelativelyeasily and fairlyaccuratelywith a ruler
and compass.I conjecturethat a numberof theoremswere discoveredvisuallyin
this way. Accurate computer graphicsare now availablefor this sort of playing
around, or to give this old and important activity its current gentrified name:
mathematicalexperimentation.

2. TRIANGLEGEOMETRYBECOMESA MUSEUMPIECE. In a certain sense,

the high regardaccordedto trianglegeometryculminatedwith the Encyklopaedie
article. One of the two authorsof the article (Berkhan)fell on the battlefield of
World War I at the age of 32, his mathematicalpotential unrealized.As though
prophetically,the subJectitself hardlysurvivedthat war.
In the USA, triangle geometrywas known as advanced geometry or college
geometry.Courseswere offered whereverthere was a facultydevotee. Textbooks
were written(e.g., N. Altshiller-Court,R. A. Johnson).In Sommerville's1924book

THE POSSIBLE TRANSFIGURATION OF TRIANGLE GEOMETRY [March

206

This content downloaded from 128.235.251.160 on Sun, 7 Dec 2014 23:49:45 PM

All use subject to JSTOR Terms and Conditions
on conics, a UK/New Zealand text, many theorems of triangle geometrywere
"downgraded"to the positionof exercisesfor the student.In more recent years, a
numberof the theorems of triangle geometryappear in Coxeter,but there is no
attempt to categorize them as such. The role of triangle geometryin European
mathematicaleducationhas been detailed by Baptist.
Here is the 1940judgmentof Eric Temple Bell on the subject:

"The geometersof the 20th Centuryhave long since piouslyremovedall these treasuresto the
museumof geometrywhere the dust of historyquicklydimmedtheir luster."(TheDeuelopment
of Mathematics,p. 323)

Joseph Malkevitch,in a recent articlethat attemptsto reviveall kindsof geometry

in the curriculum,lists fifty-eightsubfields of geometry.Subfield No. 23, called
Geometric ExtremalProblems, lists the Fermat-SteinerPoint. Other than that,
there is nothing on triangle geometryon his list. Geometric interest, even when
visual, turned elsewhere.
Yet, the subjectof trianglegeometryand its generalization,polygongeometry,
(I often use the phrasetrianglegeometryto includethis generalization),was and is
still a steady source of problemsfor the entertainmentand enjoymentof problem
buffs who read the AmericanMathematicalMonthly,the MathematicalGazette,
CruxMathematicorumand similarperiodicalsin this and other countries.Overthe
years, one man, V. Thebault,contributeda thousandproblemsin the area.
Dozens of papers on these subjects have appeared not just in the problem
solvingcontext often displayingingeniousnew approachesand new connections.
Thus, Jesse Douglas presented a complex variable approach. I. J. Schoenberg
exploited both complexvariablesand the discrete Fourier transform.Chang and
Davis looked at Napoleon'sTheoremfrom the point of view of circulantmatrices
and the Moore-Penrosegeneralized matrix inverse. (Davis, 1977, 1979, Chang,
Chang&Davis). Kimberlinghas examined triangle geometry from the point of
view of functionalequations;Baptistfrom the point of view of extremalproblems.
Other than the quiet and steadyproblemsolvingactivityand the occasionalnew
result, the subjectwas makingno waves. Mathematiciansby and large, might play
with individualitems as a relaxation;they might even derive intense satisfaction,
but they probablywould not have wanted their professional reputation to be
judged by a contributionof this sort.
"The song is ended, but the melody lingerson."

3. WHYDID TRIANGLEGEOMETRYDIE? What reasons can be given for the

short life of triangle geometryas a stronglyand coherentlydelineated corpus of
results, sanctioned by the mathematical establishment?I can suggest a few.
Though hardlyas complex a phenomenonas the decline and fall of the Roman
Empire,I will not assert that I have gotten to the heart of the matter.
(1) The perception that the subject is part of elementary, "amateur",or
recreationalmathematicsand thereforeis of low professionalstatus.The subjectis
not "deep".At the level of personalpsychology,there was a feeling that even if a
proof of a statementwas not fairly transparentor immediatelyforthcoming,one
could always"bulldoze"one's way througha proof via analyticgeometry.So why
bother?
To speak of the professionalstatus of certainproblems,one must deal with the
relationshipbetween the inner challenges of a field and the outer sociology of

1995] THE POSSIBLE TRANSFIGURATION OF TRIANGLE GEOMETRY 207

This content downloaded from 128.235.251.160 on Sun, 7 Dec 2014 23:49:45 PM

All use subject to JSTOR Terms and Conditions
mathematicians.The latter includesthe rewardstructureof mathematicalactivity.
Where one group of mathematicians may charge a second group with
"amateurism",the second may counterwith a chargeof "elitism".(See Fang and
Takayama,Wilder, 1968, 1981). There is a similarityhere with the field of music
where the profession unselfconsciouslydivides its output into "classics","light
classics", "popular","highbrow","lowbrow",and into many other status cate-
gories.
(2) The inner exhaustion of the interest and variety of its theorematic and
methodologic possibilities. The Encyklopaedie article by Berkhan and Meyer
presented no challenges or suggested directions for future development. No
outstanding,long unsolved problems emerged to capture the imaginationand
challenge mathematicalbrilliance in the way that the famous Hilbert list of
problemsdid. In a word, no reallynew ideas emergedfrom trianglegeometry.
In this connection, however, I should mention one idea that emerged briefly
from triangle geometry and is analogous to a currentlythrivingfield: computa-
tional complexity.Called geometrography, it seems to have originated in a talk
given by E. Lemoine at the Congressof the FrenchAssociationfor the Advance-
ment of Science at Orano in 1888. The reader will find writeups in Alasia, pp.
29-44, in Coolidge,Chap. III, in Lemoine, and in Mackay(1893/4).
The idea of geometographyis as follows:beginningwith a basic figure (often a
triangle),constructa distinguishedpoint or figure, often with ruler and compass,
but also with other means, and then count the numberof elementaryoperations
requiredto do so.
Alasia's elementaryconstructions(operations)are five in number(1) R: Place
a ruler'sedge througha given point. (2) R2 drawa straightline. (3) Cl: Place one
point of a compasson a given point. (4) C2: Place one point of a compasson an
indeterminatepoint of a line. (5). C3: Draw a circle. Now count up how many of
these operations are required to effectuate the required construction.Call the
total numberof operationsthe simplicity of the construction.
The simplicityof many constructionswill be found computed in the books
alluded to. Here is one result:given the side of a regularpentagon,constructthe
circle in which the pentagon can be inscribed.The count given is 8R, + 4R2 +
llC1 + 8C3,yieldinga coefficientof simplicityof 31.
Note again that the basic elementary operations are geometrical and not
arithmetic.Even so, the coefficientof simplicityis stronglyreminiscentof counting
up the total numberof floatingpoint operationsas is done in computercomplexity
theory.
As far as I am able to determine, the notion of constructivesimplicitywent
nowherein its day. It died on the vine.
(3) The increasingvisual complexityand tediousnessof the "deeper"resultsof
trianglegeometry,combinedwith
(4) A view of geometrythat had emergedby the end of the l9th Centuryand
seriouslydowngradedthe visual in favorof the algebraic/symbolic.
(5) Susceptibilityto the feeling of surprisehas its ups and downs.(My goodness,
do those three lines really intersect in one point? Who would have expected it?)
But the professionalis exposed to too many theorems and too many surprises.
Surpriseis accordinglydulled or attenuatedand thereforedevaluedpsychologically
and can easily slip into boredom.
(6) The reassignmentor the migration of some of the content of triangle
geometryto other traditionalor newly emergingfields. As examples:the famous
Desargues theorem about two triangles in perspective is seen now as part of

208 THE POSSIBLE TRANSFIGURATION OF TRIANGLE GEOMETRY [March

This content downloaded from 128.235.251.160 on Sun, 7 Dec 2014 23:49:45 PM

All use subject to JSTOR Terms and Conditions
projectivegeometry.Other theorems are viewed as part of inversivegeometryor
algebraicgeometry.
(73A deartho£ connectionsor applicationsto other fields consideredC4live'in
particularnto areas in physics etc.
However some counterexamplesexist here. Bernhard NeumannSwho has
written on triangle geometry (B. H. Neumann 1941)7 told me that his Xthera
RichardNeumann who was an electricalengineer discoveredNapoleons Theo-
rem on his own and made use of it in the theoryof three phase alternatingcurrent
circuits.(R. Neumann,1911, 1939. B. H. Neumann,,1982.)
(8) Competition arising after World War II from many other geometrical
constituencies, oRen with a strong visual component or with daims to wider
applicability.e.g., convex geometry,tilings symmetryand group theory fractalsS
graphtheo computationalgeomet etc. (See Malkevltch)
In a wordlthe problemof statusboils downto this for the reasonsJUSt outlined,
and perhapsothersSnone of the majormathematiciansof the post World War I
period considered triangle geometry to be of great importance. It would be
interestingto position its change in status within the context of the 44laws'of the
evolutionof mathematicsproposedby Croweand by Wilder (Wilder 1968 1981.)

4. ENTER:ME COMPUTER.It was clear early on that the computeroffered

the possibilityof mathematicalerimentation along visualynumerical and sym-
bolic/algebraic and logical lines; it offered the possibility of C;mechanical'or
vCautomatic'proof, and the possibility of the discoureryor generation of nevT
theorems
With the availabilityof fast computationand convenient7 high level languages7it
was inevitablethat the strategiesdirectedtowardthe abovegoals would be applied
to one of the ;easiest of the computableand decidable mathematicaltheories
wod old trianglegeometry.
The Seld of automatedreasoningis currentlyextremelyactive7boastingof an
enthusiastic corps of reseachers, international confUrences,and a number of
specializedjournals Towardthis end
tWe may pursue a numericalroad.
Let us suppose that a certain speciEc geometricalconfigurationhas been put
biward. CertainpointsSlines sa have been specifiedby their specific coordi-
nates and certainwnclusions are to be reached.Suppose that the conclusionmay
be reached by generation of a finite sequence Qf intersections of the cuses
coupled with the interpolationof new curvesto currentlyavailabledata. Carrying
out such a programnumericallywe reach our conclusionor veriifyour theoremin
the specific llumericalcase set up. We may be even in a position to do all this
visuallyusing computergraphics.
In most instances the numerical answers will be approximate in Xvorable
instances there will be single or multiple precision accuracy. By altering the
numericalparametersSa family of results can be dlsplayedrapidlySand on this
basis certaindistinguishedphenomenaSoccurrences(theorems)maybe inferredby
the investigator.This kind of thing goes on constantlyin computergraphicsSor in
computerassisted geometricindustrialdesign (CAGD).
The investigatorwhose criterion of validity demands more than approximate
numericalverificationin one specific instancemust employother trategies. If the
given initial configurationconsists solely of lines specified as passing through
points with rational coordinateslthen all the computationmay be (in principle)
performedexactlyin rationalarithmetic.

1995] THE POSSIBLE TRANSFIGURATION OF TRIANGLE GEOMETRY 209

This content downloaded from 128.235.251.160 on Sun, 7 Dec 2014 23:49:45 PM

All use subject to JSTOR Terms and Conditions
CE 1s 1I S T A

There is a strategyavailableto overcomenumericallythe lack of generalityin

one special numericalcase. It will be exhibitedin the case of the famous Pappus
theoremof projectivegeometrywhose initial configurationis Wo arbitrarystraight
lines and three arbitrarypoints on each.
If the coordinatesof the points and lines are taken as algebraicallyindependent
real numbers, then the verification of the theorem in tehatone case serves to
demonstratethe theorem generally.(Davis, 1977, Rowlandand DavisX198171981,
Jcllwaltz, llong anu lanJ.
This leaves hangingin the air the question of whatSdigitallyspeakingSa set of
algebraicallyindependentnumberscould possiblymean. Such a set would act as
our symbolswith respect to the computationsinvoWedand hence the force of the
above remark.
One might instantiatea 4pseudo-algebraicallyindependent' set of numbersby
takingthem as randomnumbers7and interpretingthe result probabilistically.One
then arrivesat the followingprinciple:(validwithin a certain limited context) if a
theoremis true for one randomlyselected set of initialconfigurations,it is true for
all configurations.This puts at risk the old caveat of mathematicsteachers(which
can have a constipatingeffect on investigations)* you must prove it in sU cas&s, not
just in one specialcase. At the same time, it offers the possibilityin some instances
of formalizingthe inductiveleaps that the mathematicalmind takes when con-
frontedwith what seems to be, logicallyspeaking,incompleteevidence.
eWe may go the road of heuristics,such as developed by George Polya in a
series of popularbooks. We may try to combinePolya hewristicswith strategiesof
ArtificialIntelligence(AI). (Newell. R. Davis and D Lenat.)
tWe may go the road of logic. Tarskihas provedthat all statementsin 44Tanki
geometry' are decidable. But it has been found that workingwith the two basic
predicatesgivenby Tarski,one £orbetweennessand one for distance,has not been
a very promising approach. (Woos, p. 206-214, Hao Wang. For a system of
4'naturaldeduction"using the programAUTOMATH,see de Bruijn).
*We may travel along the symbolic road, using computer packages such as
FORMAC (now located in the Museum of Ancient Softwaret),MAPLE, MAC
SYMA, MATHEMATICAand prove Pappus in a naive and ad hoc wayX(Daus
and Cerutti)
*We may travel a rather sophisticatedalgebraicroad7a road traveled by Wu7

Chou, Zhang7Goa and others, a road that uses such algebraic ideas as Ritt's
Principle,or GroebnerBases.
Briefly and here I follow Chou, 1988 the strategyof method is: Wu7s

Step 1: First convert the initial geometricalconfigurationinto a set o£ polyno-

mial equations ConveAthe geometricalconclusioninto a polynomialequation.
The initial configuration(hypotheses)will be specifiedby
hl(ul7 U27 + . ., Ud; Xl* * * * Xt) -°*

h2(ul7 u27. . 7 Ud; Xl7t * * 7 Xl) °-

hn(UI U27 * ' . a Ud7 Xl. * B 7 Xt) 0*

The conclusionis gwen by

210 THIEPOSSIBLE TRANSFIGURATION OF TRIANGLE GEOMETRY [March

This content downloaded from 128.235.251.160 on Sun, 7 Dec 2014 23:49:45 PM

All use subject to JSTOR Terms and Conditions
 In these equations, the u's are independent variables, while the are X5S

algebraicallydependenton the u's.

Step 2. Using pseudo division and Ritt's Principle, (or your own method),
"triangulate"the polynomials.This means replacingthe set of h's by a set which
introducesone new x at a time. Then check for irreducibility.
Step 3. Successivepseudodivisionto arriveat a final remainderR, after analysis
of certain non-degenerateconditions.Hopefully, R = O, indicatingthat the geo-
metricalimplicationis true.
To indicate the complexityof this approachin some specific examples,Chou
reports that the proof of the so-called Thebault-Taylortheorem of elementary
geometry(involvinglines, circles, intersectionsand tangencies)requiredthe ma-
nipulation of polynomialsof almost 700,000 terms. (In their naive approachto
Pappus' Theorem of projective geometry, Davis and Cerutti report that the
numberof terms in the polynomialswas almost33,000.)

5. GENERATIONOF NEW THEOREMSUSING THE COMPUTER.A number

of approacheshave been explored.
*Playingaround(i.e., mathematicalexperimentation)visually,numerically,sym-
bolically. This is quite successful. For example, I have found many theorems
(unpublished)in the area of group matrices simply by playing around with the
MATLABmatrixpackage.Most experimenterscan report similarexperiences.
In connection with visual output, I have even argued (Davis, 1974), for the
recognition of the existence of "visual theorems", i.e., stable visual patterns,
generatedby a computeralgorithm,where what the eye "sees" need not even be
verbalized, let alone formalized in traditional mathematical language. (As a
parallel, philosopher Susanne Langer, in the context of music, speaks of the
"subtlecomplexesof feeling that languagecannot even name, let alone set forth".)
*Programmedheuristics.This seems to be less promising.(See, e.e., Newell,
1981)R. Davis and D. Lenathave writtena program,AM (AutomatedMathemati-
cian) which starts from set theory, and proposes to invent new mathematical
concepts and new conjecturesrelyingon a libraryof built-inheuristics.
Whateverthe method employed,numerousnew theoremshave emerged,some
of which have attracted attention, surprise and enthusiasm. (Hofstadter,
Grunbaumand Shephard.In the last namedreference,there are some philosophi-
cal remarkson proof methodologyparallelingthose just made.)

6. THE TRANSFIGURATION OF TRIANGLEGEOMETRY.Can a subjectarise

from the dust and ashes that historyhas piled on it? Onlyif it is transformedin the
process.The focus of trianglegeometryhas now been changed.The computerhas
popped it up a metalevel, and in the process has transfiguredthe subject.
Hundredsof elementaryand not so elementarytheoremsthat were in the litera-
ture have now been provedby computer.Many new theorems have been discov-
ered, again in a varietyof ways. Trianglegeometryalwayswas a practiceground
for strategies of proof in the spirit of Euclid, and it has now become a testing
groundfor strategiesof decidability,proof, and theorem discovery.These strate-
gies have run from naive schemes to the employmentof deep and abstractresults
of modernalgebraand differentialalgebra.
But there is yet more that emergesfrom the change of focus: I believe that the
experiencegained in this change can become a prime source of raw materialfor

1995] THE POSSIBLE TRANSFIGURATION OF TRIANGLE GEOMETRY 211

This content downloaded from 128.235.251.160 on Sun, 7 Dec 2014 23:49:45 PM

All use subject to JSTOR Terms and Conditions
philosophicaldiscussionson the nature of proof, methodologiesof research,the
role and nature of intuition5educationalvalues, etc.
What are some of the implicationsof this work?While I think it is too earlyto
write with assurance,I will venture a few observations.
*Obsessed over the millennia by the vision that mathematics can provide
absolute, rock bottom "certainty",the mathematical establishment has often
expressedits displeasurewith certaintypes of "proof":visual, mechanical,experi-
mental, probabilistic.This attitude goes back as far as Archimedes(200 B.C.), if
not further.
Computerproof, theoremdiscovery,and mathematicalexperimentationare now
openly acknowledgedas legitimate methodologies and roads to mathematical
knowledge.
Thus, absolutelyrigorousmathematicalproof, as an ideal, is givingway and is
now seen as a part of a wider, more generousand more flexible notion that I like
to call "mathematicalevidence".
*Giventhat the output of the whole mathematicalworld,measuredin terms of
numbersof theorems,is of the order of one hundredthousandper annum,what
use wouldthe automaticgenerationof theoremsin a restricted,well ploughedarea
serve?As A. L. Crelle correctlyobservedin the epigraphplaced at the head of this
article, the simplestmathematicalstructurecan produce an unlimitednumberof
conclusions. What, therefore, does one "do" with mathematicalproducts that
mightbe stampedout like doughnutsin a doughnutmachine?
*In this process, the individualtheorem may stand devalued. For example, it
maybe vital in a certainapplicationto knowthat the product12563*502= 6306626,
but to the averagemathematicalmind, the theorem expressedby this multiplica-
tive identityis quite tedious.
Since this kind of theorem(of arithmetic,trianglegeometryor whatever)can now
be produced by the hundredfold, the emphasis inevitably changes from the
theorem to the means by which the theorem is produced. By and large, the
mediumbecomes the message. This is one of the lessons taughtby the subsump-
tion of geometryby algebrathat occurredas a result of the revolutionaryvision of
Descartes.
*An individualtheorem may still be judged as to its importance,practicalor
otherwise, and may turn out to have such importance.This importanceis deter-
mined by subjectiveand historicalcriteria.
The processwherebya mathematicalconcept,whetherit be a simple point in a
triangle or a whole complex theory, becomes "distinguished",is not capable of
formalization.(Woos sets up this problemas one of 33 basic researchproblemsin
automatedreasoning.)It is a historicalprocessand may involvethe whole scientific
communityor significantsubsets of that community.
*The inner complexityof some proofsby computer,often involvingpolynomials
of hundreds or thousands of terms, adds new respect and appreciationfor the
historicmethods and traditionalmannerin which the resultswere presented.
*The mysterious,omnipresent and vitally essential "mathematicalintuition"
together with its components of experience, analogy, educated guessing, and
transcendental,non-explainable"pre-knowledge"(e.g., Ramanujan),all get raised
a metaleveland now can operate in a wider arena.
*As regards mathematicaleducation, I think the message is clear. Classical
proof must move over and share the educationalstage and time with other means
of arrivingat mathematicalevidenceand knowledge.Mathematicaltextbooksmust
modifythe often deadeningrigidityof the Euclideanmodel of exposition.

THE POSSIBLE TRANSFIGURATION OF TRIANGLE GEOMETRY [March

212

This content downloaded from 128.235.251.160 on Sun, 7 Dec 2014 23:49:45 PM

All use subject to JSTOR Terms and Conditions
7. CONCLUDINGREMARKS.

"The whole culturalworld,in all it formsexists throughtradition.These formshave arisennot

merelycausally. . . they havearisenwithinour humanspacethroughhumanactivity." Edmund
Husserl,TheOriginof Geometry.

The distinguishedlogician Hao Wang, as a result of his interest in proof by

computer,was once charged with desiring to eliminate the mathematician.He
answered"No, only the inferiorones". I shouldlike to interpretthis remarkmore
charitably:that the capabilitiesof all mathematiciansare elevatedby their associa-
tion with computation.The transformationby the computerof trianglegeometry,
and of manyother areas has, paradoxically,reconfirmedand strengthenedthe vital
role of humansin the wonderfulactivityknownas doing mathematics.Put it even
more strongly:mathematicsdevelopsin such a way that the role of the mathemati-
cians is alwaysmanifest.

ACKNOWLEDGMENTS. The authorwishes to acknowledgehelpful suggestionsreceivedfrom Drs.

ChristaBinder,BrankoGrunbaum,DouglasHofstadterand from the referees.

BIBLIOGRAPHY

Alasia, Christoforo,La RecenteGeometriadel Triangolo,Cittadi Castello:S. Lapi, 1900.

Altshiller-Court,N. CollegeGeometry,Richmond:Johnson,1925.
Baptist,Peter, Die Entwicklungder NeuerenDreiecksgeometrie, Mannheim:B. I. Wissenschaftsverlag,
1992.
Bell, Eric T., TheDeuelopment of Mathematics,New York:McGrawHill, 1940.
Berkhan,G. and W. Fr. Meyer:Neuere Dreiecksgeometrie,Encyklopaediader Mathematischen Wis-
senschaften,Leipzig:Teubner,1914-1931,Vol. III, AB 10, pp. 1173-1276.
Bledsoe, W. W. and D. W. Loveland,eds. AutomatedTheoremProuing:After 25 years,Providence:
AmericanMath.Soc., 1984.
de Bruijn,N. G., A surveyof the projectAutomath.In: To H. B. Curry:Essaysin combinatorylogic,
lambdacalculusandformalism,ed. J. P. Seldinand J. R. Hindley.New York:AcademicPress 1980.
Cajori,F. A Historyof ElementaryMathematics,New York:Macmillan,1917.
Casey,John,GeometrieElementaireReocente, Paris:Gauthier-Villars,1890.
Chang,Geng-Zhe,A proof of the theoremof Douglas and Neumannby circulantmatrices,Houston
Math.J., vol 8, 1982,pp. 15-18.
Chang, Genz-Zhe, and P. J. Davis, A circulant formulationof the Napoleon-Douglas-Neumann
Theorem,LinearAlgebraand its Applications,1983,pp. 87-95.
Chou Shang-Ching,An Introductionto Wu's methodsfor mechanicaltheorem provingin geometry,
Journalof AutomatedReasoning,vol 4, 1988,pp. 237-267
Coolidge,JulianL., A Treatiseon the Circleand the Sphere,Oxford:ClarendonPress, 1917.Reprinted,
Bronx,New York:Chelsea,1971.
Coxeter,H. S. M., Introductionto Geometry,New York:Wiley, 1961.
Coxeter,H. S. M. and S. L. Greitzer,GeometryReuisited,Washington:Math.Assn. Amer, 1967.
Crowe, M. J., Ten 'laws' concerningpatternsof change within the historyof mathematics,Historia
Mathematica,v. 2, pp. 161-166. Also pp. 469-470.
Davis,P. J., VisualGeometry,ComputerGraphicsandTheoremsof PerceivedType,Proc.Symp.Appl.
Math.vol. 20, Amer. Math.Soc., 1974.
, Cyclictransformationsof polygons,and the generalizedinverse,Can.J. Math.,29, pp. 756-770,
1977.
, Proof,completeness,transcendentalsand sampling,J. Assos. Comp.Mach.,vol. 24, 1977,pp.
298-310.
, CirculantMatrices,New York:Wiley,1979,pp. 140-154.
, Are there coincidencesin mathematics?,Amer. Math.Month.,vol. 88, 1981,pp. 311-320.
Davis, P. J., and Elsie Cerutti,FORMACmeets Pappus:Some observationson elementaryanalytic
geometryby computer,Amer. Math.Monthly,vol. 76, 1969,pp. 895-905.
Davis, Randalland Douglas Lenat, eds. KnowledgeBasedSystemsin ArtificialIntelligence,New York:
McGraw-Hill,1982.

1995] THE POSSIBLE TRANSFIGURATION OF TRIANGLE GEOMETRY 213

This content downloaded from 128.235.251.160 on Sun, 7 Dec 2014 23:49:45 PM

All use subject to JSTOR Terms and Conditions
Douglas, Jesse, Geometryof Polygonsin the ComplexPlane, J. Math. and Phys.,vol. 19, 1940, pp.
93-130.
Emmerich,A., Die BrocardschenGebildeund ihre Beziehungenzu den VerwandtenMerkwuerdigen
PunktenundKreisendes Dreiecks,Berlin:GeorgReimer,1891.
Fang,J. and K. P. Takayama,Sociologyof Mathematics andMathematicians, Hauppauge,N.Y.:Paideia,
1973.
Forder,H.G., Thefoundationsof EuclideanGeometry,Cambridge:UniversityPress, 1927.
Gale, David,MathematicalEntertainments,The MathematicalIntelligencer,Spring,1992.
Gallatly,William,TheModernGeometryof the Triangle,London:1910.
, TheNinePoint Circle,Cambridge:Johnson,1907.
Grunbaum,Brankoand G. C. Shephard,FromMenelausto ComputerAssistedProofsin Geometry,to
appear.
Hofstadter,Douglas R., FromEuler to Ulam:discoveryand disectionof a geometricgem, Centerfor
Researchon Conceptsand Cognition,Bloomington,Indiana,December,1992.
Hong, Jiaweiand Xiao-nanTan, Provinginequalitiesby examples,CourantInstituteof Math.,1987.
Johnson,R. A. AduancedEuclideangeometry,New York:Dover, 1960.
Kimberling,Clark H. Central points and central lines in the plane of a triangle. To appear in
MathematicsMagazine.
Kimberling,ClarkC., Trianglecentersas functions.To appear.
, Functionalequationsassociatedwith trianglegeometry.AequationesMathematicae.
, and Peter Yff, The circumcircleand the line at infinity.To appear.
Klee, Victor and Stan Wagon, Old and new unsoluedproblemsin plane geometryand numbertheory,
Washington:Math.Assos. Amer., 1991.
Lemoine,E., Geometography, Paris:1902.
Mackay,J. S., Historyof the Nine Point Circle,Proc. Edin. Math.Soc., v. 11, 1892-3,pp. 19-57.
, Earlyhistoryof the symmedianpoint, Proc.Edin. Math.Soc., v. 11, 1892-3,pp. 92-103.
, The Geometographyof Euclid'sProblems,Proc.Edin. Math.Soc., v. 12, 1893-4,pp. 2-16.
Malkevitch,Joseph,Geometryin Utopia,Dept of Math.,York College,Jamaica,N.Y. 11451,March3,
1989.
Neumann,BernhardH., Some remarkson polygons,J. Lond. Math. Soc., vol. 16, 1941,pp. 230-245.
, Plane polygonsrevisited,pp. 113-122 of Essaysin StatisticalScience:Papersin Honourof
P.A.P. Moran,J. Gani and E. J. Hannan,eds., J. of AppliedProbability,SpecialVolume19A, 1982.
Neumann,Richard,GeometrischeUntersuchungeines Ausgleichtransfomators fuer unsymmetrische
Drehstromsysteme, Elektrotechnikund Machinenbau,vol. 39 1911,pp. 747-751.
, SymmetricalComponent Analysis,London:Pitman:1939.
Newell, Allen, The Hueristicof George Polya and its relationto artificialintelligence,Dept. Comp.
Sci., Pub. No. 81-133, Carnegie-MellonUniv. Pittsburgh,1981.
Pedoe, Daniel, Circles,London:Pergamon,1957.
Rowland,JohnH. and P. J. Davis,On the selectionof test data for recursivemathematicalsubroutines,
SIAMJ. Comput.vol. 10, 1981,pp. 59-72.
, and , On the use of transcendentalsfor programtesting,J. Assoc Comp.Mach.,vol. 28,
1981,pp. 181-190.
Schoenberg,I. J., MathematicalTimeExposures,Washington:Math.Assn. of Amer., 1982.Chaps.6-9.
Schwartz,J. T., ProbabilisticAlgorithmsfor Verificationof PolynomialIdentities,Journalof the ACM,
1980.
Sommerville,D. M. Y., AnalyticalConics,London:G. Bell, 1924.
Tarski,A., A decisionmethodfor elementaryalgebraand geometry,2nd. ed. Berkeley:1951.
Wang,Hao, Computertheoremprovingand artificialintelligence,pp. 49-72 of Bledsoe and Loveland.
Wetzel,John E., Conversesof Napoleon'sTheorem,Amer.Math.Monthly,vol. 99, 1992,pp. 339-351.
Wilder,RaymondL., Mathematicsas a CulturalSystem,Oxford:PergamonPress, 1981
, TheEuolutionof MathematicalConcepts,New York:Wiley, 1968.
Woos,Larry,AutomatedReasoning:33 BasicResearchProblems,EnglewoodCliffs:PrenticeHall, 1987.
Wu Wen-Tsuen,Basic principlesof mechanicaltheoremprovingin geometries,Journalof Automated
Reasoning,v. 2, 1986,pp. 221-252.

Division of Applied Mathematics

Brown University
Providence, RI 02912

214 THE POSSIBLE TRANSFIGURATION OF TRIANGLE GEOMETRY [March

This content downloaded from 128.235.251.160 on Sun, 7 Dec 2014 23:49:45 PM

All use subject to JSTOR Terms and Conditions