You are on page 1of 22

Mind Association

The Origin and Meaning of Geometrical Axioms

Author(s): H. Helmholtz
Reviewed work(s):
Source: Mind, Vol. 1, No. 3 (Jul., 1876), pp. 301-321
Published by: Oxford University Press on behalf of the Mind Association
Stable URL: .
Accessed: 21/03/2012 10:43

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .

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

Oxford University Press and Mind Association are collaborating with JSTOR to digitize, preserve and extend
access to Mind.
No. 3.1 [JulyI876.





MY object in this article*is to discussthe philosophical
bearinigof recent inquiriesconcerniniggeometricalaxioms and
the possibilityof working out analyticallyother systems of
geometrywithotheraxiomsthaniEtnclid's. The originialworks
on the subject, addressed to experts olnly,are particularly
abstruse,but I will tryto make it plainieven to those who are
not mathematicians. It is of course no part of my plan to
prove the new doctrinescorroctas mathematicalconclusions.
Such proofmustbe soughtin the originalworksthemselves.
A-mongthe firstelementarypropositionsof geometry,from
whichthe studentis led on by continuouischainsof reasoning
to the laws of moreand more complexfigures,are some which held not to admit of proof, though sure to be granted
by everyone who uniderstands their meaning. These are the
so-called Axioms; for example, the propositionthat if the
shortestline drawnbetweentwo pointsis called straightthere
caii be onlyone such straightline. Again, it is an axiom that
throughany threepoinltsin space, not lying in a straightline,
a plane maybe drawn,i.e., a surfacewhichwill whollyinclude
* The suibstanceof the firsthalf of the articlehas been previously
expoundedby me,in the AcadXmy of Feb, 192,1870, It is here set forth
anew as necessarycontext.
302 The Origin and, Meaning of GeometricalAxioms.
every straightline joining any two of its points. Another
axiom, about which there has been much discussion,affirms
that througha point lying withouta straightline only one
straightline can be drawnparallel to the first, two straight
lines that lie in the same plane and never meet,however far
theymay be produced,being called parallel. There are also
axiomsthat determinethe numberof dimensionsof space and
its surfaces,lines and points, showing how they are con-
tinuous; as in the propositions,that a solid is bounded by a
surface,a surface- by a line and a line by a point,thatthe point
is indivisible,that-by the movement of a point a line is
described,by that of a line a line or a surface,by that of a
surfacea surface or a solid,but by the movementof a solid a
solid and nothingelse is described.
Now what is the originof such propositions,unquestionably
trueyet incapable of proof in a sciencewhere everythingelse
is reasoned conclusion? Are they inheritedfromthe divine
source of our reason as the idealisticphilosophersthink,or is it
onlythat the ingenuityof mathematicians has hithertonot been
penetratingenough to find the proof? Every new votary,
comingwithfreshzeal to geometry,naturallystrivesto succeed
whereall beforehim have failed. And it is quite right that
each should make the trial afresh; for, as the question has
hithertostood, it is only by the fruitlessnessof one's own
effortsthat one can be convincedof the impossibility of finding
a proof. Meanwhilesolitaryinquirersare alwaysfromtime to
timeappearingwho become so deeplyentangledin complicated
trainsof reasoningthatthey can no longer discovertheir mis-
takes and believe they have solved the problem. The axiom
of parallelsespeciallyhas called fortha greatnulmber of seeming
The main difficulty in these inquiriesis and always has beeni
the readinesswithwhichresultsof everydayexperiencebecome
mixed up as apparent necessities of thoughtwith the logical
processes,so long as Euclid's methodof constructive intuitionis
exclusivelyfollowedin geometry. In particularit is extremely
difficult,on this method,to be quite sure that in the steps
prescribedforthe demonstration we ha,venot involuntarily and
unconsciouslydrawn in some most general results of experi-
ence, which the power of executing certain parts of the
operationhas alreadytaught us practically. In drawing any
subsidiaryline for the sake of his demonstration,the well-
trained geometerasks always if it is possible to draw such a
line. It is notoriousthat problems of constructionplay an
essentialpart in the systemof geometry. At firstsight,these
appear to be practical operations,introducedfor the training
The Origin antdMeaning of aeometrical Axioms. 303
of learners; but in realitythey have the force of existential
propositions. They declare that points, straight lines or
circles, such as the problem requires to be constructed,are
possible underall conditions,or theydetermineany exceptions
that there may be. The point on which the investigations
turnthat we are going to consideris essentiallyof this nature.
The foundationof all proof by Euclid's method consists in
establishingthe congruenceoflines,angles,plane figures,solids,
&c. To make the congruenceevident,the geometricalfigures
are supposed to be applied to one another,of course without
changing their form and dimensions. That this is in fact
possible we have all experienced from our earliest youth.
BuLt,when we would build necessities of thought upon this
assumption of the free translation of fixed figures with
unchanged formto everypart of space, we must see whether
the assumptiondoes not involvesome presuppositionof which
no logical proof is given. We shall see later onithat it does
contain one of most serious import. But if so, everyproof
by congruence rests upon a fact which is obtained from
I offerthese remarksat firstonlyto show what difficulties
attendthe completeanalysisof the presuppositionswe make in
employingthe commonconstructivemethod. We evade them
whenwe apply to the investigationof principlesthe analytical
methodof modern algebraical geometry. The whole process
of algebraical calculationis a purelylogical operation; it can
yield no relationbetweenthe quantitiessubmittedto it thatis
not alreadycontainedin the equationswhich give occasionfor
its being applied. The recent investigationshave accordingly
been conducted almost exclusively by means of the purely
abstractmethodsof analyticalgeometry.
However,afterdiscoveringby the abstractmethodwhat are
the points in question,we shall best get a distinctview of
them by taking a region of narrowerlimits than our own
worldof space. Let us, as we logicallymay,suppose reasoning
beings of onlytwo dimensionsto live and move on the surface
of some solid body. We will assume that theyhave not the
power of perceiving anythingoutside this surface,but that
upon it theyhave perceptionssimilarto ours. If such beings
workedout a geometry,theywould of courseassign onlytwo
dimensionsto theirspace. They would ascertainthata point
in movingdescribesa line, and thata line in movingdescribes
a surface. But they could as little representto themselves
what.furtherspatial constructionwould be generated by a
surfacemoving out of itself,as we can representwhat wbuld
be generated by a solid moving-out of the space we know.
304 The Origin and Meaning of GeometricalAciorns.
By the muchabused expression" to represent"or " to be able
to thinkhow somethinghappens" I uniderstalnd-anid I do not
see how anythingelse caln be understoodby it withoutloss of
all meaning-the powerof imaginingthe whole seriesof sen-
sible impressionsthatwould be had in such a case. Now as no
sensible impressionis known relatingto such ani unheard-of
eventas the movementto a fourthdimensioni would be to uss,
or as a moveement to our third dimensionwould be to the
inhabitantsof a surface,sucha "representation"is as impossible
as the "representation"of colourswould be to one bornblind,
thougha descriptionof thenm in general termsmightbe given
to him.
Our surface-beinigswould also be able to draw shortest
lines in their superficialspace. These would not necessarily
be straight lines in our sense, but whatare technicallycalled
geodeticlines of the surfaceon which theylive, lines such as
are described by a tense thread laid along the surfaceand
which can slide upon it freely. I will heniceforth speak of
such lines as the straightestlines of any particularsurfaceor
givenispace, so as to bring out their analogy withthe straight
line in a plane.
Now if beings of thiskind lived on an infiniteplane, their
geometrywould be exactlythe same as our planimetry. They
would affirmthat only one straightline is possible betweeni
two points,that througha third point lying withoutthis line
onlyone line can be drawn parallel to it, that the ends of a
straightline nevermeet thoughit is producedto infinity, and
so on. Their space might be inifinitely extended,but even if
there were limits to their movementand perception,they
would be able to representto themselvesa continuationbeyond
these limits, and thus their space would appear to them
infinitely extended,just as ours does to us,althoughour bodies
calnnotleave the earthand our sight only reaches as faras the
visible fixedstars.
But intelligentbeings of the kind supposed mightalso live
on the surfaceof a sphere. Their shortestor straightestline
between two pointswould then be an arc of the great circle
passing throughthemii.Every great circle passing through
two pointsis by these divided into two parts,and if theyare
unequal the shorteris certainlythe shortestline on the sphere
betweenthe two points,but also the otheror largerarc of the
same great circleis a geodetic or straightestline,'i.e., every
smallerpartofit is the shortestlinebetweenits ends. Thus the
notionof the geodeticor straightestline is not quite identical
withthat ofthe shortestline. If the two given pointsare the
ends of a diameterof the sphere,everyplane passiingthrouigh
TTheOrigin and, Afeaningof G-eom7tetrictlAxiois. 305
this diametercuts semicirclesonlthe surfaceof the sphere all
of which are shortestlines betweelnthe ends; in whichcase
thereis an infinitenumberof equal shortestlines betweenthe
given points. Accordingly,the axiom oftherebeing only one
shortestlille between two points would not hold withouta
certainexceptionforthe dwellerson a sphere.
Of parallel lines the sphere-dwellers would klnownothinig.
They would declare that any two straightestlines, sufficiently
prodtuced, must finallycut, not in, one onl.ybut in two points,
The sum of the angles of a trianglewould be always greater
thantwo rightangles, increasingas the surfaceof the triangle
grew greater. T'heycould thus have no conceptionof geome-
trical similaritybetween greater and smaller figuresof the
same kind,forwiththema greatertrianglemusthave different
angles froma smallerone. Their space would be unlimited,
but would be foundto be finiteor at least representedas such.
It is clear, theni,that such beings must set up a very
different systemof geometricalaxioms fromthatof the inhabi-
talnts of a plalne or from ours with our space of three
dimensions,though the logical powers-of all were the same;
nor are more examples necessary to show that geometrical
axioms must vary accordingto the kind of space inhabited.
But let us proceed still farther.
Let us thinkof reasoniingbeings existingon the surfaceof
an egg-shapedbody. Shortestlines could be drawnibetween
threepoinltsof such a surfaceand a triangleconstructed. But
if the attemptwere made to constructcongruenttrianglesat
differentpartsofthesurface,it wouldbe foundthattwotriangles
withthreepairsof equal sideswouldnothave theirangles equal.
The sumof the angles of a triangledrawnat the sharperpole of
the bodywould departfarther fromtwo rightangles thanif the
triangle were drawn at the bluinterpole or at the equLator.
Henlce it appears that not even such a simple figureas a
trianglecan be moved oln such a surfacewithout change of
form.It would also be foulndthatif circlesof equal radii were
colnstructed at differentparts of such a surface(the lelngthof
the radii being always measured by shortestlines along the
surface)theperipherywould be greaterat the blunterthan at
the sharperend.
We see accordinglythat,if a surface admitsof the figures
lying on it being freelynmovedwithoutchange of any of their
lines and alngles as measured along it, the propertyis a
special one and does not belong to everykind of surface. The
conditiolnuiider which a surface possesses this important
propertywas pointedout by Gauss in his celebratedtreatiseon
306 The Origi?nand Meaning of GeometricalAxioms.
the curvatureof surfaces.* The "meastureof curvature,"as
he called it, i.e., the reciprocalof the productof the greatest
and least radii of curvature,must be everywhereequal over
the whole extentof the surface.
Gauss showedat the same time that this measureof curva-
ture is not chanigedif the surfaceis bent withoutdistensionor
contractionof any part of it. Thus we can roll up a flatsheet
of paper intothe formof a cylinderor of a cone withoutany
change in the dimensionsof the figurestaken along the surface
of the sheet. Or the hemisphericalfundusof a bladder may
be rolledinto a spinidle-shape withoutaltering the dimensions
on the surface. Geometryon a plane will thereforebe the
same as on a cylindricalsurface; only in the latter case we
must imagine that any numberof layersof this suLrface, like
the layersof a rolled sheet of paper, lie one upon anotherand
that after each entire revolutionround the cylindera niew
layeris reached.
These observationsare meantto give the reader a notionof
a kind of surface the geometryof which is on the whole
similarto thatof the plane, but in whichthe axiom ofparallels
does not hold good, namely,a kind of curved surface which
geometricallyis, as it were,the counterpartof a sphere,and
whichhas thereforebeen called the pseuclospherical surfcteby
the distinguishedItalian mathematician,E. Beltrami,who has
investigatedits properties.t It is a saddle-shaped surfaceof
whichonlylimitedpieces or strips can be conilectedlyrepre-
sented in our space, but which may yet be thought of as
infinitelycontinuedin all directions,since each piece lyingat
the limitof the part constructedcaln be colnceivedas drawn
back to the middle of it and theif continued. The piece
displaced mustin the process change its flexuirebut not its
dimensions,just as happens with a sheet of paper moved
about a cone formedout of a plane rolled up. Such a sheet
fitsthe conical surfacein everypart, but must be more bent
near the vertexand cannotbe so moved over the vertex as to
be at the same time adapted to the existing cone and to its
Like the plane and the sphere,pseudosphericalsurfaceshave
theirmeasureof curvatureconstant,so thateverypiece of them
* Gauss, Wlerke, Bd. IV., p. 215, firstpublishedin Commentationes
Soc. Req. Scientt.Gottingersis
recentiores, vol. vi., 1828.
tSa,gio di Interpretazionedellat.GeometriaNon-EPclideat,Napoli,
1868.-Teoria fondamentale deyli %paziidi CtOrvatusra costante,Annwali
di Miatematica,Ser. :I., Tom. II., pp. 232-55. Both have beenitrans-
lated intoFrenchbyJ. Hoiiel,AnnalesScientifiques de 1'PEole Normale,
Tom.V., 1869.
The Origin and Meaning of aeornetricalAxiomns. 307
can be exactlyapplied to everyother piece, and thereforeall
figuresconstructedat one place on the surfacecan be trans-
ferredto any other place with perfectcongruityof form and
perfectequality of all dimensionslying in the surfaceitself.
The measure of curvatureas laid down by Gauss, which is
positiveforthe sphere and zero for the plane, would have a
constantnegative value forpseudosphericalsurfaces,becanse
the two principal curvaturesof a saddle-shaped surfacehave
A strip of a pseudosphericalsurface may,for example,be
representedby the innersurface(turnedtowardsthe axis) of a
solid anchor-ring. If the plane figure aabb (Fig. 1) is made
to revolve on its axis of symmetry AB, the two arcs ab will
Ft ~~~a

FIG, 1. FIG. 2.
describe a pseudospherical concave-convexsurface like that
of the ring. Above and below, towards aa and bb, the
surfacewill turn outwardswith ever-increasingfiexure,till it
becomes perpendicular to the axis and ends at the edge
with one curvature infinite. Or, again, half of a pseudo-
spherical surface may be rolled up into the shape of a
champagne-glass (Fig. 2) with tapering stem infinitelypro-
longed. But the surface is always necessarilybounded by
a sharp edge beyond which it cannot be directly-continued.
Onllyby supposin1geach-single piece of the edge cut loose and
drawnalong the surfaceof the ring or glass, can it be brought
to places of different flexureat which farthercontinuationof
the piece is possible.
In this way too the straightestlines of the pseudospherical
surfacemay be infinitely produced. They do not like thoseon
a spherereturnupon themselves,but, as on a plane, onlyone
shortestline is possiblebetweentwo given points. The axiom
of parallels does not howeverhold good. If a straightestline
is given on the surfaceand a pointwithoutit, a whole pencil
308 ThteOrigin and MeaninRtg
of GeometricalAaxioins.
of straightestlines mlaypass throughthepoint,no onieof which,
thoughinfinitely produced,cuts the firstline; the pencil itself
being limitedby two straightestlines, one of whichintersects
one of the enidsof the given line at an infinitedistance,the
otherthe otherend.
As it happened, a systemof geometryexcludingthe axiom
of parallels was devised on Euclid's syntheticmethod, as far
back as the year 1829, by N. J. Lobatschewsky,professorof
mathematicsat Kasan,* and it was proved that this system
could be carried out as consistentlyas Euclid's. It agrees
exactly with the geomnetry of the pseudospherical surfaces
workedout recentlyby Beltrami.
Thus we see that in the geometryof two dimensionsa sur-
face is markedout as a plane or a sphere or a pseudosplherical
surfaceby the assumptioni that any figuremaybe movedabout
in all directionswithoutchalngeof dimensions. The axiom
thatthere is only one shortest liniebetween any two points
distinguishesthe plane anldthe pseudospherical surfacefrom
the sphere,and the axiom of parallels marksoffthe plalnefrom
the pseudosphere. These three axioms are in fact necessary
and sufficientto defineas a planiethe surfaceto whichEuclid's
planimetry has reference,as distinguishedfromall othermodes
of space in two dimensions.
The difference between plane anld spherical geometryhas
been long evident,but the meaningof the axiom of parallels
couLldnot be understoodtill Gauss had developed the notionof
surfaces flexiblewithoutdilatationand consequentlythat of
the possibly infinitecontinuationof pseudosphericalsurfaces.
Inhabiting a space of three dimensionsand endowed with
organs of sense fortheirperception,we can representto our-
selves the variouscases in which beings on a surface might
have to develop theirperceptioln of space; for we have only
to limitour own perceptionsto a narrowerfield. It is easy to
thinkaway perceptionsthatwe have; but it is verydifficult to
imagineperceptionsto whichthereis nothinganalogous in our
experience. When, therefore, we pass to space of three
dimensionswe are stopped in our power of representationby
the structureof our organs and the experielncesgot through
themwhichcorrespondonllyto the space in whichwe live.
There is howeveranotherway of treatinggeometryscientifi-
cally. All knownspace-relationsare measurable,that is they
may be broughtto determination of magnitudes(lines,angles,
surfaces,volumes). Problems in geometrycan thereforebe
solved by findingmethods of calculationfor arrivingat un-
* Pr incpiet der Geometrie,
Kasan, 1829-90.
ThleOrigincand 3eaninig of GeometricalAxiowm. 309
knlownmagnitudesfromknownones. Tlhisis done 'inanalytical
geometry, where all forms of space are treated only as
quantitiesanlddetermlined by means of otherquantities. Even
the axioms themselvesmake referenceto magniitudes. The
straightline is definedas the shortestbetween two poilnts,
whichis a determination of qua.ntity. The axiom of parallels
declares thatif two straightlines in a plane do not intersect
(are parallel),the alternateangles,or the correspondingangles,
made by a thirdline intersecting them,are equal; or it may be
laid down instead that the sum of the alnglesof any triangle
is equal to two right angles. These are determinationsof
Now we may start with this view of space,,according to
which the positionof a point uiay be determinedby measure-
ments in relationlto any given figure (system of co-ordi-
nates), taken as fi-xed,and theniinquirewhat are the special
characteristics of our space as manifestedin the measurernents
that have to be made, and how it differsfromotherextended
quantitiesof like variety. This path was firstentered by onie
too early lost to science,B. Riemanniof G6ttingen.* It has
the peculiaradvantage that all its operations consist in pure
calculationof quantities,which quite obviates the dalngerof
habitualperceptionis beinlgtaken forlnecessitiesof thought.
The lnumberof measurementsniecessary to give the position
of a point is equal to the numberof dimelsionisof the space
in question. In a line the distance fromolnefixed point is
sufficient,thatis to say,one quantity; in a surfacethe distances
fromtwo fixed points must be given; in space, the distances
fromthree; or we require as oln the earthlongitude,latitude
and heightabove the sea, or,as is usual in analyticalgeometry,
the distances fromthreeco-ordinateplanes. Riemannicalls a
systemof differences in whichone thingcan be determinedby
nmeasurementsan "infoldextendedaggregate" or an "aggre-
gate of nbdimensiolns." Thus the space in whichwe live is a
three-fold, a surfaceis a twofoldand a line is a simpleextended
aggregateof poinats. Time also is an aggreglateof one dimen-
sionl. The systemofcoloursis alnaggregateofthreedinmenisions,
inasmuchas each colour,accordingto the investigationsof Tb.
Young and Clerk Maxwell,may be representedas a mixture
of three primarycolours, takelnin definitequantities. The
particularmixturescan be actuallymadlewiththe colouIr-tOp.
In the sameway we maycon-sider the systemof simpletones
as all aggregate of two dimenisionis,if we distinguishonlypitch
Ueber die Hypothesenwelche der Geometriezu Grtindeliegen,
vom 10 Juni 1854. (Abhanldl.der k6nigl.Gesedllch.
Vt Gdttingen,Bd. XIII.),
310 Axioms.
the Ortginand Meaningof Geometrbcal
and intensity and leave out of accountdifferences of timbre.
Thisgeneralisation of theidea is well-suitedto bringout the
distinctionbetweenspaceofthreedimensions andotheraggre-
gates. We can, as we knowfromdailyexperience,compare
theverticaldistanceof twopointswiththehorizontal distance
of twoothers,becausewe can applya measurefirstto theone
pairand thento the other. But we cannotcomparethe dif-
ference betweentwotonesofequalpitchand different intensity
withthatbetweentwo tonesof equal intensity and different
pitch. Riemannshowedby considerations of this kindthat
the essentialfoundation of any systemof geometry is the
expressioni thatit gives forthe distancebetweentwo points
lyingin anydirectionfromone another, beginningwiththe
intervalas infinitesimal. He took fromanalyticalgeometry
themostgeneralformforthisexpression, that,namely,which
leavesaltogether openthekindof measurements by whichthe
positionofanypointis given.* Thenhe showedthatthekind
of free mobilitywithoutchangeof formwhichbelongsto
bodies in our space can onlyexistwhen certainquantities
yielded by the calculationt-quantities that coincidewith
Gauss'smeasureof surface-curvature whentheyare expressed
forsurfaces-haveeverywhere an equalvalue. For thisreason
Riemanncallsthesequantities, whentheyhavethesamevalue
in all directions fora particularspot,themeasureof curvature
ofthespaceat thisspot. To preventmisunderstanding I will
once more observethat this so-calledmeasureof space-
curvature is a quantityobtainedby purelyanalyticalcalculation
and thatits introduction involvesno suggestion of relations
thatwouldhave a meaningonlyforsense-perception. The
name is merelytaken,as a shortexpressionfora complex
relation,fromthe one case in whichthe quantity designated
admitsof sensiblerepresentation.
Nowwhenever thevalueofthismeasureofcurvature in any
space is everywhere zero,that space everywhere conforms to
theaxiomsof Euclid; and it maybe calleda flat (homaloid)
space in contradistinction to otherspaces,analytically con-
structible,thatmaybe calledcurvedbecausetheirmeasureof
curvature has a valueotherthan zero. Analyticalgeometry
may be as completely and consistently workedout forsuch
spacesas ordinary geometry forouractuallyexistinghomaloid
* For the squareof the distanceof two infinitely
is a homogeneous
expression quadricfunctionofthedifferentials
t Theyarealgebraical
compounded from of
thevarioustermsintheexpressionforthesquareof thedistanceoftwo
contiguous theirdiflerential
Axioms. 311
The Origin and Meaning of GReometrical
If the measure of curvatureis positive we have spherical
space, in which straightestlines returnupon themselvesand
there.are nioparallels. Such a space would,like the surfaceof
a sphere,be unlimitedbut not infinitelygreat. A constant
niegativemeasureof curvatureon the otherhand gives pseudo-
sPherical space, in which straightestlines run out to infinity
and a pencil of straightestlines may be drawn in any fattest
surfacethroughany pointwhichdo not intersectanothergiven
straightestline in that surface.
Beltrami* has renderedthese last relationsimaginable by
showing that the points, lines and surfaces of a pseudo-
spherical space of three dimensions can be - so portrayed
in the interior of a sphere in Euclid's homaloid space,
that everystraightestline or flattestsurfaceof the pseudo-
spherical space is representedby a straight line or a plane,
respectively,in the sphere. The surfaceitself of the sphere
corresponds to the infinitelydistant points of the pseudo-
sphericalspace; and the differentpartsof this space, as repre-
sentedin the sphere,become smallerthe nearertheylie to the
sphericalsurface,diminishingmorerapidlyin the directionof
the radii than in that perpendicularto them. Straightlines in
the spherewhichonlyintersectbeyond its suLrface correspond
to straightestlines of the pseudosphericalspace whichnever
Thus it appeared that space, colnsideredas a region of
measurablequantities,does not at all correspondwiththe most
general conceptionof an aggregate of three dimensions,but
involves also special conditions,depending on the perfectly
free mobility of solid bodies without change of form to
all parts of it and with all possible changes of direction,
and, farther,on the special value of the measureof curvature
whichforour actual space equals, or at least is not distinguish-
able from,zero. This latter definitionis given in the axioms
of straightlines and parallels.
Whilst Riemann entered upon this new field fromthe side
of the most general and fundamentalquestions of analytical
geometry,I myself arrived at similar conclusions,t partly
from seelkingto representin space the system of colours,
inivolvingthe comparisonof one threefoldextendedaggregate
with another,and paYtlyfromilnquirieson the origin of our
ocular measure fordistancesin the field of vision. Riemann
startsby assumingthe above-mentionedalgebraicalexpression,
which represents in the most general form the distance
* Teoriafondacnentale, utsup.
t Ueber die Thatsachendie derGeometriezum Grundeliegen (Nacka
vonder konigi.Ges. d. Wiss. zu Gbttingen,
vichten Juni3, 1868).
312 Axioms.
The Origintand Mleaning of Geomiietrical

between two infinitely lnearpoints,and deduces therefrom the

conditionis of mobilityof rigid figures. I, on the other hand,
startingfrom the observed fact that the maovement of rigid
figuresis possible,in our space, with the degree of freedom
that,we know,deduce the necessityof the algebraic expression
taken by Riemann as an axiomn. The assumptionsthat I had
to make as the basis of the calculationwere the following.
First, to make algebraical treatmentpossible, it must be
assumed thatthe positiolnof any pointA can be determined,
in relatiolnto certainigiveni figures taken as fixed bases,
by measurementof some kind of magnitudes,as linles,angles
between linies,angles between surfaces and so forth. The
measurementsnecessaryfordetermininig the-positionof A are
known as its co-ordinates. In general the nnmberof co-
ordinates n-ecessaryto the complete determinationof the
positioniof a poinltmarksthe niumberof the dimenisions of the
space in question. It is furtherassumed that with the move-
menltof the poinitA the magnittudes used as co-ordinatesvary
Secondly,the definitioni of a solid body, or rigid systemof
points,mustbe made in such a way as to admitof magnitudes
being comparedby congruenice. As we mustnot at this stage
assume any special methods for the measurementof magni-
tudes, our definitioni cani,in the firstinistance,rul only as
follows: Betweenthe co-ordinatesof anlytwo poinltsbelonging
to a solid body, theremustbe an equationwhich,lhoweverthe
body is moved,expressesa constant spatial relation(proving
at last to be the distanice)betweenthe two poinlts,and which
is the same for congruentpairs of poinlts,that is to say, such
pairs as can be made successivelyto coincidein space withthe
same fixedpair of points.
Howeverindeterminate in appearance,thisdefinitioniinvolves
most importanitconsequences, because with increase in the
number of points the niumberof equations increases niuch
more quickly thanithe number of co-ordinates which they
determinie. Five points, A, B, C, D, E give ten different
pairs of points (AB, AC, AD, AE, BC, BD, BE, CD, CE, DE)
anid thereforeten equatiolns,involvilngin space of three
dimensionsfifteenvariable co-ordinates. But of these fifteen
six must remainiarbitraryif the systemof five points is to
admit of free movement anid rotation, and thus the ten
equationscan determineonlynine co-ordinatesas functionsof
the six variables. With six poinltswe obtain fifteenequations
fortwelve quanitities,with sevell poinltstwenity-one equations
forfifteen, and so oln. Now fromn indepelndent equationswe
can determinen contained quantities,and if we have more
Axio)ns. 313
The Origin antdMeaning of Geomnetr'leal
thanin equations,the superfluousones mustbe deduciblefrom
the firstn. Hence it followsthat the equationswhichsubsist
betweenthe co-ordinatesof each pair of points of a solid body
nmusthave a special character,seeinogthat,wvhenin space of
threedimensionstheyare satisfiedfornine pairs of poinitsas
formedout of any fivepoints,the equ:ationifor the tenithpair
follows by logical colnsequen-ce. Thus our assumptionlfor
the definitioni to determine
of soliditybecomes quite sufficienlt
the kind of equationsholdingbetweenthe co-ordinatesof two
Thirdly,the calculationmustfurtherbe based on the factof
a peculiarcircumstancein the movementof solid bodies, a fact
so familiarto us thatbut forthis inquiryit mightneverhave
been thoughtof as somethingthat need not be. When in our
space of threedimensiolns two points of a solid body are kept
fixed,its movementsare limitedto rotationsrounidthe straight
line colnniecting them. If we turnit completelyround once,
it again occupies exactly the position it had at first. This
factthat rotationin one directionalways brinigsa solid body
back into its original positionilneeds special menition. A
systemof geometryis possible withoutit. This is most easily
seen in the geometryof a plane. Suppose that with every
rotationof a plane figureits linear dimensionsincreased in
proportionto the angle of rotation,the figureafterone whole
rotation through 360 degrees would no longer coincidewith
itself as it was originally. But any second figurethat was
congruentwiththe firstin its originalpositionmightbe made
to coincidewithit in its second position by being also turneed
through 360 degrees. A consistent system of geometry
would be possible upon this supposition,whichdoes not come
underRiemaln's formula.
On the otherhand I have shown that the threeassumptions
taken togetherforma sufficient basis for the starting-pointof
Riemann's investigation,and thence forall his furtherresults
relating to the distinctionof differentspaces according to
theirmeasureof curvature.
It still remainedto be seen whetherthe laws of motion as
dependenton movingforces could also be consistentlytrans-
ferredto sphericalor pseudosphericalspace. This investigation
has been carried out by ProfessorLipschitz of Bonn.* It is
found that the comprehensiveexpression for all the laws of
dynamics,Hamilton'sprinciple,may be directlytransferred to
* Untersuchungenfiberdie ganzen homogenenFunctionenvon n
Diferentialen(Borchardt'sJournalfii Miathematik,Bde. 1xx. 3, 71;
eines Problemsder Variationsrechnung
lxxiiii. 3, 1); UJnterstclihung
(Ibid. Bcl lxxiv.)
814 Axcioms.
The Origii and Meaning of Geomnetrical
spaces of which the measure of curvatureis otherthan zero.
Accordingly,in this respect also the disparate systems of
geometrylead to no contradiction.
We have now to seek an explanationof the special charac-
teristicsof our own flatspace, since it appears that theyare
not implied in the general notionof an extendedquantityof
threedimensionsand of the freemobilityof bounded figures
therein. Necessitiesof thought,involvedin such a coniception,
theyare not. Let us then examinethe oppositeassumptionas
to theiroriginbeing empirical,and see if they can be inferred
fromfacts of experienceand so established,or if,when tested
by experience,they are perhaps to be rejected. If they are
of empiricaloriginwe must be able to representto ourselves
connected series of facts indicatinga different value for the
measureof curvaturefromthat of Euclid's flatspace. But if
we can imaginesuch spaces of othersorts,it cannot be main-
tained thatthe axiomsof geometryare necessaryconsequences
of an a' jpriori transcendentalform of intuition,as Kant
The distinction between spherical, pseudospherical and
Euclid's geometrydepends, as was above observed, on the
value of a certainconstant called by Riemannthe measure of
curvatureof the space in question. The value must be zero
for Euclid's axioms to hold good, If it were not zero, the
sum of the angles of a large trianglewould differfromthat of
the angles of a small one, being largerin spherical,smallerin
pseudosphericalspace. Again, the geometricalsimilarityof
large and small solids or figuresis possible only in Euclid's
space. All systemsof practical mensurationthat have been
used forthe angles of large rectilineartriangles,and especially
all systems of astronomicalmeasurementwhich make the
parallax of the immeasurablydistantfixed stars equal to zero
(in pseudosphericalspace the parallax even of infinitelydistant
points would be positive), confirmempiricallythe axiom of
parallels and show the measureof curvatureof our space thus
farto be indistinguishable fromzero. It remains,however,a
question,as Riemannobserved,whetherthe result might not
be different if we could use otherthan our limitedbase-lines,
the greatestof whichis the major axis of the earth's orbit.
Mea-nwhile, we mustnot forgetthatall geometricalmeasure-
mentsrest ultimatelyupon the principle of congruence. We
measurethe distance between points by applyingto themthe
compass,rule or chain. We measure angles by bringingthe
divided circle or theodoliteto the vertex of the angle. We
also determinestraightlines by the path of rays of lightwhich
in our experienceis rectilinear;but thatlighttravelsin shortest
Ple Origin and Meaning of Geometri,cal
Axrioms. 315
lines as long as it continuesin a mediumof constantrefraction-
would be equally true in space of a differentmeasure of
curvature. Thus all our geometricalmeasurementsdepend on
our instrumentsbeingereally,as we considerthem,invariable
in form.,or at least on theirundergoingno otherthanthe small
changes we know of as arising fromvariationof tenmperature
or fromgravityacting differently at differentplaces.
In measuringwe onlyemploythe best and surestmeans we
know of to determinewhat we otherwiseare in the habit of
makingout by sightand touch or by pacing. Here our own
body withits organsis the instrument we carryabout in space.
Now it is the hand, now the leg that serves fora compass,or
the eye turningin all directionsis ourtheodoliteformeasurinig
arcs and angles in the visual field.
Every comparativeestimate of magnitudesor measurement
of theirspatial relationsproceedsthereforeupon a supposition
as to the behaviour of certain physical things, either the
humanbody or otherinstrumentsemployed. The supposition
may be in the highestdegree probableand in closestharmony
withall otherphysicalrelationsknownto us, but yet it passes
beyondthe scope of pure space-intuition.
It is in fact possible to imagineconditionsfor bodies appa-
rentlysolid such that the measurementsin Euclid's space be-
comewhattheywould be in sphericalor pseudosphericalspace.
Let me firstremindthe readerthatif all the linear dimensions
of otherbodies and our own at the same timewere diminished
or increasedin like proportion,as forinstanceto half or double
theirsize, we should with our means of space-perceptionbe
utterlyunawareof the change. This would also be the case
if the distension or contractionwere differentin different
directions,provided that our own body changed in the same
mannerand furtherthat a body in rotatingassumed at every
moment,withoutsuffering or exertingmechanicalresistance,the
amountof dilatationin its different dimensionscorresponding
to its position at the time. Think of the image of the world
in a convex mirror. The common silvered globes set up in
gardens give the essential features,only distortedby some
opticalirregularities.A well-madeconvex mirrorof moderate
aperture representsthe objects in frontof it as apparently
solid and in fixedpositionsbehindits surface. But the images
of the distanthorizonand of the sun in the sky lie behind the
mirrorat a limiteddistance,equal to its fQcallength. Between
these and the surfaceof the mirrorare foundthe images of all
the other objects beforeit, but the images are diminishedand
flattenedin proportionto the distanceof theirobjects fromthe
mirror. The flattening, or decrease in the thirddimension,is
316 ThLeOrigin and Meaning of GeometricalAxiorns.
relativelygreaterthanthe decrease of the surface-dimensions.
Yet everystraightline or every plane in the outer world is
representedby a straightline or a plane in the image. The
image of a nmanmleasuring witha rule a straightline fromthe
mirrorwould contractmore and morethe fartherhe went,but
with his shrunkenrule the man in the image wouildcountout
exactly the same nuLmberof centimetresas the real man.
And, in general, all geometrical measurementsof lines or
angles made withregularlyvaryingimages of real instruments
would yield exactlythe samne resultsas in the outerworld,a11
congruent bodies would coincide on being applied to onie
another in the mirroras in the outerworld,all lines of sight
in the outerworld would be representedby straightlines of
sight in the mirror. In short I do not see how men in the
mirrorare to discoverthat theirbodies are not rigid solids and
theirexperiencesgood examplesof the correctnessof Euclid's
axioms. But if theycould look out uponiour worldas we can
look into theirs,withoutoversteppingthe boundary,theymust
decla-reit to be a picturein a sphericalmirror,and wouLldspeak
of us just as we speak of them; and if two inihabitanits of the
different worldscould communicatewith one another,neither,
so faras-I can see, would be able to convincetlheotherthat he
had the truLe,the other the distortedrelationis. Indeed I
cannot see that such a question would have any meaning
at all so 'long as mechaniicalconsiderationsare not mixed up
Now Beltrami'srepresentationof pseudosphericalspace in a
sphereof Euclid's space is quite similarexcept that the bac-k-
ground is not a plane as in the convex mirror,but the surface
of a sphere,and that the proportionin which the images as
theyapproach the spherical surface contract,has a different
mathematicalexpression. If we imagine then, conversely,
that in the sphere,forthe interiorofwhichEuclid's axiomshold
good, movinigbodies contractas theydepart fromthe centre
like the images in a convex mirror,and in such a way that
their representatives in pseudospherical space retain their
dimensionsunchanged,-observerswhosebodies wereregularly
subjected to the same change would obtain the same results
fromthe geometricalmeasurementstheycould make as if they
lived in pseudospheriealspace.
We can even go a step further,and iniferhow the objects in
a pseudosphericalworld,were it possible to enter one, would
appear to an observerwhose eye-measureand experiences of
space had been gained like ours in Euclid's space. Such an
observerwould continueto look upon rays of light or the lines
of vision as straightlinies,sui-chas are met withiin flat space
ThteOrigin aiut7 JIfectingof GeometricalAxiovm. 317
and as they really are in the spherical representationof
pseudospherical space. The visual image of the objects in
pseudosphericalspace would thus make the same impression
upon him as if he were at the centreof Beltrami'ssphere. He
would think he saw the mostremoteobjects round about him
at a finitedistance,*let us suppose a hundredfeetoff. But as
he approached these distant objects, theywould dilate before
him, though more in the third dimensionthan superficially,
while behind him they would: contract. He would know
that his eye judged wrongly. If he saw two straightlines
which in his estimateran parallel forthe hundredfeetto his
world's end, he would findon followingthemthat the farther
he advanced the morethey diverged,because of the dilatation
of all the objects to whichhe approached. On the otherhanid
behindhim theirdistancewould seen to diminish,so thatas he
advanced theywoutldappear always to divergemoreand more.
But two straightlines whichlfromhis firstpositionseemed to
converge to onie and the same point of the background a
h-uLndred feet distalnt,would continue,to do this however far
he went,and he wotuldneverreach their point of intersectioni.
Now we can obtain exactlysimilarimages of our real world
if we look through a large convex lenis of corresponding
negative focal length,or even througha pair of convex spec-
tacles if ground somewhatprismaticallyto resemblepieces of
one continuous larger lens. With these, like the convex
mirror,we see remoteobjects as if near to us, the mostremote
appearing no fartherdistant than the focus of the lens. In
going about with this lens before the eyes, we find that the
objects we approach dilate exactly in the manner I have
described for pseudosphericalspace. Now any one using a
len-s,were it evelnso strongas to have a focal length of only
sixtyinches,to say nothingof a hundred feet,would perhaps
observe for the firstmoment that he saw objects brought
nearer. But after going about a little the illusion would
vanish,and in spite of the false images he would judge of the
distances rightly. We have every reason to suppose that
what happens in a few hours to any one beginning to wear
spectacles would soon enough be experienced in pseudo-
spherical space. In short,pseudosphericalspace would not
seem to us very strange,comparativelyspeaking; we should
only at firstbe subject to illusions in measuringby eye the
size and distanceof the moreremoteobjects.
There would be illusionsof an opposite description,if,with
of the squareofthisdistance,
* Tlhereciprocal expressedinnegative
wouldbe the measureof curvature
quantity, of the pseudospherical
space. 22
318 Axioms.
The Origqnand Meaninngof Geoinet'ri'eai

eyes practised to measure in Euclid's space, we entered a

sphericalspace of three dimensions. We should suppose the
moredistant-objects to be moreremoteand largerthan they
are, and should find-on approachingthem that we reached
them more quicklythan we expected fromtheir appearance.
But we should also see before us objects that we can fixate
.only with diverging lines of sight, namely, all those at a
greaterdistancefromus than the quadrant of a great circle,
Such an aspect of thingswould hardlystrikeus as veryextra-
ordinary,for we can have it even as things are if we place
beforethe eye a slightlyprismaticglass with the thickerside
towards the lnose: the eyes must then become divergentto
take in distant objects. This excites a certain feeling of
unwonted strain in the eyes but does not perceptiblychange
the appearanceof the objects thus see;n. The strangestsight,
however,in the sphericalworldwould be the back of our own
head, in which all visual lines not stopped by other objects
would meetagain, and whichmustfillthe extremebackground
of the wholeperspectivepicture.
At the same time it must be noted that as a small elastic
flat disc, say of india-rubber,can onlybe fittedto a slightly
curvedspherical-surfacewithrelativecontractionof its border
and distension of its centre, so our bodies, developed in
Euclid's flat space, could not pass into curved space without
undergoingsimilardistensionsand contractionsof theirparts,
their coherencebeing of course maintainedonlyin as far as
theirelasticitypermittedtheirbendingwithoutbreaking. The
kind of distensionmiustbe the same as in passing froma small
bodyimaginedat the centreofBeltrami'ssphereto its pseudo-
spherical or spherical representation. For such passage to
appear possible, it will always have to be assumed that the
body is sufficiently elastic and small in comparisonwith the
real or imaginaryradius of curvatureof the curved space into
whichit is to pass.
These remarkswill sufficeto show the way in whichwe can
inferfrom the known laws of our sensible perceptionsthe
series of sensible impressionswhich a sphericalor pseudo-
sphericalworldwould give us, if it existed. In doing so we
nowhere meet with inconsistencyor impossibilityany more
than in the calculationof its metrical proportions. We can
representto ourselvesthe look of a pseudosphericalworld in
all directions just as we can develop the conceptionof it.
Thereforeit cannotbe allowed that the axioms of our geometry
depend on the nativeformof our perceptivefaculty,or are in
any way connectedwithit.
It is different withthe three dimensionsof space. As all
The Origin and Meaning of GeometricalAxrioms. 319
our means of sense-perceptionextend onlyto space of three
dimensions,and a fourthis not merelya modificationof what
we have but somethingperfectlynew, we find ourselves by
reason of our bodilyorganisationquite unable to representa
In conclusion I would again urge that the axioms of
geometryare not propositionspertaining only to the pure
doctrineof space. As I said before,theyare concernedwith
quantity. We can speak of quantitiesonly when we know
of some way by which we can compare, divide and measure
them. All space-measurementsand thereforein general all
ideas of quantitiesapplied to space assume the possibilityof
figures moving without change of form or size. It is
true we are accustomed in geometryto call such figures
purely geometricalsolids, surfaces,angles and lines, because
we abstract from all the other distinctionsphysical and
chemical of natural bodies; but yet one physical quality,
rigidity,is retained. Now we have no othermark of rigidity
of bodies or figuresbut congruence,whenevertheyare applied
to one alnotherat any time or place, and afterany revolution.
We cannot however decide by pure geometryand without
mechanical considerationswhether the coincidingbodies may
not both have varied in the same sense.
If it were useful for any purpose,we mightwith perfect
consistencylook upon the space in which we live as the
apparentspace behinda convex mirrorwith its shortenedand
contractedbackground; or we might consider a bounded
sphere of our space, beyond the limitsof whichwe perceive
nothingfurther,as infinitepseudosphericalspace. Only then
we should have to ascribe to the bodies whichappear as solid
and to our own body at the same time correspondingdis.
tensionsand contractions, and we must change our systemof
mechanicalprinciplesentirely; for even the propositionthat
everypointin motion,if acted upon by no force,continuesto
move withunchangedvelocityin a straightline,is not adapted
to the image of the world in the convex-mirror. The path
would indeed be straight,but the velocitywould depend upon
the place.
Thus the axioms of geometryare not concernedwith space.
relations only but als-oat the same time with the mechanical
deportmentof solidest bodies in motion,. The notionof rigid
geometricalfiguremightindeedbe conceivedas transcendental
in Kant's sense, namely, as formedindependentlyof actual
experience,whichneed not exactlycorrespondtherewith,any
morethan naturalbodies do ever in fact correspondexactlyto
the abstract notion we have obtained of them by induction.
22 *
320 lAte Oig7it actncJ3ieaningof GeometricalAxiioms.
Taking the lnotionof rigiditythus as a mere ideal, a strict
Kantian mightcertainlylook upon the geometricalaxioms as
propositionsgiven 'a prioriby transcendentalintUitionwhich
no experiencecould either confirmor refute,because it must
firstbe decided by them wlhetherany natural bodies can be
considered as rigid. B:1utthen we should have to maintain
that the axioms of geometryare not syntheticpropositiolns,
as Kanit held them: theywould merelydefinewhat qualities
and deportmenta body musthave to be recognisedas rigid.
But if to the geometrical axioms we add propositions
relating to the mechanical propertiesof natural bodies, were
it onlythe axiom of inertiaor the single propositionthat the
mnechanical and physicalpropertiesof bodies and theirmutual
reactionsare, other circumstancesremainingthe same, inde-
pendent of place, suchla system of propositionsbas a real
importwhich calnbe conifirmed or reftutedby experience,but
just forthe same reason can also be got by experience. The
mechanicalaxiomjust cited is in factof the utmostimportance
forthe wlholesys-temi of our mechanicaland physical concep-
tions. That rigid solids, as we call them, which are really
nothingelse than elastic solids of great resistance,retain the
same form in every part of space if no externalforce affects
them,is a single case fallinguniderthe generalprinciple.
For the rest, I do niot,of course, suppose that mankind
firstarrivedat space-intuitionsin agreemenit with the axioms
of Euclid by aniycarefullyexecutedsystemsof exact measure-
ment. It was rather a succession of everyday experiences,
especiallythe perceptionof the geometricalsimnilarity of great
and small bodies, only possible in flat space, that led to the
rejection,as impossible,of everygeometricalrepresentation at
variancewiththis fact. For thisInoknowledgeof the necessary
logical connectionbetween the observed fact of geometrical
similarityand the axioms w.rasneeded, but only an intuitive
apprehensionof the tvpical relations between lines, planes,
angles, &c., obtained by numerousand attentiveobservations
-all intuitionof the kind the artist possesses of the objects
he is to represent,and by means of which he decides surely
and accuratelywhethera new combinationwhichhe trieswill
correspondor not to theirnature. It is true that we have no
word but intuitionto markthis; but it is knowledgeempirically
gained by the aggregationiand reinforcement of similarrecur-
rent impressionisin memory,and not a transcendentalform
given beforeexperience. That othersuch empiricalintuitionas
of fixedtypicalrelations,whennot clearlycomprehended, have
frequentlyenioughbeen taken by metaphysicialns for 'a _rioii
principles,is a pointon whichI need not insist;
tahVe Or'q[a of Moral Ideas-. 321

To sumni up, the final outcomeof the whole inquirymay be

thus expressed:-
(1.) The axioms of geometry,taken by themselvesout of all
connectionwithmechanicalpropositions, representno relations
of real things. When thus isolated,if we regard thenm with
Kant as formsof intuitiontranscendentallygiven,theycon-
stitute a forminitowhich any empiricalcontentwhateverwill
fitand which thereforedoes not in any way limitor determiine
l)eforehandthe natureof the content. This is true,however.,
not onlyof ELuelid'saxioms,but also of the axioms of spheric1al
anidpseu-dospherical geometry.
(2.) As soon as certainprinciplesof mechanicsare conjoined
withtlhieaxiomsof geonletrywe obtaina svstemof propositions
whichhas real import,and whichcan be verifiedor overturned
by empiricalobservations,as fromexperienceit can be iniferred.
If such a systemwere to be taken as a transcendentalform of
intuitionand thought,theiremustbe asstumeda pre-established
harmolny betweenformanldreality.


CAkNthe fact that manidistilnguiliesright fromwronlgbe
explained by-the association of ideas ? This is the question
whichI mean to discuss,and as I feel compelled to aniswerit
in the negative,it is the morerequisitethat I should acknow-
ledge at the outset association to be a great and fruitful
principle,of wide range and powerfulinifluence in the melntal
ecolnomy. It is not confinedto any particularprovince of
humannature,but operatesalike among our thoughts,feelings
and volitions,bringingtheminto the mostvaried combinations.
Its laws are essential conditionsof memoryand reminiscence,
of all the powersof intellectualacquisitivenessand inventive.
ness, of imaginationand reasoniing; they are implied in the
perfectingand pervertingof every perceptive faculty and
emotional capacity; and largely determinethe growth of
characterboth in inidividualsand communities. In a word, it
is mainly through associationthat mentalenergyis accumulated
and mental change effected. It is the sovereign means of
eliciting and educating,of drawing out and developing, the
originalendowmentso the mind,and it is continuallyaltering
forthe betteror worse all temperaments,dispositions,habits,