47
1 “The Geometry Behind Poincaré’s Conventionalism” Jeremy Heis University of California, Irvine Under Review July 2019 How does Poincaré argue for his conventionalism about geometry? In particular, what features of geometry does his argument rely on? According to some common interpretations, Poincaré’s argument depends on features that are not unique to geometry. For instance, on one recent reading (Gimbel 2004), Poincare’s conventionalism amounts to nothing more than trivial semantic conventionalism: the truth of a geometrical sentence such as “the distance from A to B > the distance from C to D” depends inter alia on the meaning of the word “distance”; but which meaning we assign to the word “distance” is simply a conventional fact about our language; so the truth of the sentence is a matter of convention. Since it is clear that this argument could be run for any sentence (this is why it is “trivial”), this interpretation of Poincaré’s conventionalism turns only on the mundane fact that sentences of geometry are composed of words. Interpretations that assimilate Poincaré’s argument to Duhemian arguments 1 based on underdetermination also turn on a feature that is not unique to geometry: 1.The metric of space is underdetermined by our a priori and empirical evidence. 1 For examples of interpreters who attribute this argument to Poincaré, see Stump 1989, note 52; Ben Menahem 2006 presents a more sophisticated version of this reading.

“The Geometry Behind Poincaré’s Conventionalism” Jeremy ......Poincare’s famous sphere argument in “Space and Geometry” (1895/1913)3 might be cited in support of premise

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

  • 1

    “TheGeometryBehindPoincaré’sConventionalism”

    JeremyHeisUniversityofCalifornia,Irvine

    UnderReviewJuly2019

    HowdoesPoincaréargueforhisconventionalismaboutgeometry?Inparticular,

    whatfeaturesofgeometrydoeshisargumentrelyon?Accordingtosomecommon

    interpretations,Poincaré’sargumentdependsonfeaturesthatarenotuniqueto

    geometry.Forinstance,ononerecentreading(Gimbel2004),Poincare’s

    conventionalismamountstonothingmorethantrivialsemanticconventionalism:

    thetruthofageometricalsentencesuchas“thedistancefromAtoB>thedistance

    fromCtoD”dependsinteraliaonthemeaningoftheword“distance”;butwhich

    meaningweassigntotheword“distance”issimplyaconventionalfactaboutour

    language;sothetruthofthesentenceisamatterofconvention.Sinceitisclearthat

    thisargumentcouldberunforanysentence(thisiswhyitis“trivial”),this

    interpretationofPoincaré’sconventionalismturnsonlyonthemundanefactthat

    sentencesofgeometryarecomposedofwords.

    InterpretationsthatassimilatePoincaré’sargumenttoDuhemianarguments1

    basedonunderdeterminationalsoturnonafeaturethatisnotuniquetogeometry:

    1.Themetricofspaceisunderdeterminedbyouraprioriandempirical

    evidence.

    1ForexamplesofinterpreterswhoattributethisargumenttoPoincaré,seeStump1989,note52;BenMenahem2006presentsamoresophisticatedversionofthisreading.

  • 2

    2.Givensuchunderdetermination,onlyconventioncanbeusedtodecideona

    metric.

    3.Mattersofconventionarenotmattersoffact.2

    So,themetricofspaceisnotamatteroffact,butofconvention.

    Poincare’sfamoussphereargumentin“SpaceandGeometry”(1895/1913)3might

    becitedinsupportofpremise1.Poincareimaginesbeingslivingintheinteriorofa

    spherewithatemperaturefieldthatdecreasestoabsolutezeroattheboundaryof

    thesphereaccordingtotheformulaR2-r2,whereRistheradiusofthesphereandr

    isthedistanceofapointfromthecenter.Ifthesizesofbodiescontract

    proportionatelytotheirtemperature,nobodycouldtravelinfinitetimeandatfinite

    speedfromthecenterofthespheretoitsboundary,sincethebodywillgetsmaller

    andsmallerasittravelstowardtheboundary.Itwouldbeconsistentwithallofour

    aprioriandempiricalevidencetosayofsuchbeingsthattheyliveinaninfinite

    worldwherethemetricishyperbolic.Alternately,itwouldbeconsistentwiththe

    empiricalandapriorievidencetoclaimthattheyliveinafiniteregionwithina

    spacewherethemetricisEuclidean,aslongaswemodifyourusualphysicallawsto

    includethetemperaturefieldIdescribedabove.Furthermore,Poincarésupposes

    thatlightwithinthediscrefractswithanindexofrefractioninverselyproportional

    2AsPoincaréusestheterm“convention,”a“convention”isnotatruthandsoalsonotamatteroffact.See,e.g.,(1891/1913,39).3TheoriginalFrenchversionofthisessayappearedin1895.Itwaslightlyrevised–i.e.,oneparagraphwasremoved–andrepublishedin1902intheFrenchversionofScienceandHypothesis,whichwasthentranslatedintoEnglishin1905,andthenre-translatedintoEnglish(byGeorgeHalsted)inafarsuperiortranslationfrom1913.Icitepagenumbersfromthe1913Englishre-translation.ButIalsoincludethepublicationdateoftheoriginalessaysaswell,sincetheoriginalpublicationdateswillbeimportantinmydiscussionofRussellandPoincarébelow.

  • 3

    toR2-r2.Insuchaworld,beamsoflightwilltravelincirculararcswithinthesphere

    (seefigure1).Again,wecandescribethepathsofthelightbeamsasstraightlines

    thatobeyhyperbolicgeometry,orwecandescribethemascirculararcsthatobey

    Euclideangeometry–aslongaswemodifyourusualphysicallawstoincludethis

    lawofrefraction.Sinceourtotalphysicaltheory–includingourphysicallawsand

    ourgeometry--facesexperienceonlyasunit,wegettheusualDuhemianconclusion

    thatbothalternativesareconsistentwithallofourevidence,andsoitisonly

    conventionandnotmattersoffactthatdeterminewhichgeometrytochoosefor

    suchaworld.

    OntheDuhemianreading,therelevantfactaboutgeometrythatleadsto

    conventionalismisjustthefactthatgeometricalsentencesarepartofourtotal

    physicaltheory.ThismakestheDuhemianreadingveryunattractive.4Itwouldseem

    thenthateverysentenceineveryphysicaltheorywouldturnoutconventional,

    whichwouldjustcollapsethedistinctionbetweenmatteroffactandconventionand

    therebyleaveconventionalismwithoutmuchinterest.Moreover,therearepartsof

    ourtotalsciencethatPoincaréclearlythinksarenotconventional.Inparticular,

    Poincareseemstobelievethatitisnotaconventionthatspaceallowsfor

    displacementsthatformagroup,5andhealsobelievesthatallofarithmeticconsists

    ofsyntheticaprioritruths,notconventions.

    Becauseofthefailuresofthesereadings,BenMenahem2006hasarguedthat

    thisDuhemianargumentneedstobesupplementedbyspecificfactsabout

    geometry.Onherreading,whatisimportantaboutthespheremodelofhyperbolic 4SeeStump1989andFriedman1999.51895/1913,53.

  • 4

    geometryisthatitprovidesarecipeforsystematicallyredescribingeveryfactabout

    thegeometryofaEuclideanworldinhyperbolictermsandviceversa.Itthusgivesa

    convincingcasefortheunderdeterminationofthegeometryofphysicalspacebyall

    possibleevidence.Poincare’sargument,onherreading,doesnotemployorrequire

    ageneralDuhemianargument(based,say,onglobalconfirmationholismaboutour

    totalphysicaltheory),andsodoesnotgeneralizebeyondtheveryspecialcaseofthe

    metricofspace.

    Onedifficultywiththiskindofreading–adifficultyI’llreturntobelow–is

    thatthisargumentforconventionalismdependsonlyontheexistenceofEuclidean

    modelsofnon-Euclideanspaces,andthesemodelspredatedPoincare’swritingsin

    thephilosophyofgeometry.ThisreadingthusleavesitunexplainedwhyPoincaré

    wasledtoconventionalismwhenothergeometers(whoalsounderstoodthese

    modelsperfectlywell)werenot.

    Areadingthat,ifsuccessful,wouldexplainwhyPoincaréwasledto

    conventionalismwhenothergeometerswerenotisprovidedbyMichaelFriedman.

    AccordingtoFriedman1999,Poincaréargues(onphilosophicalgrounds)thatwe

    canknowapriorithesyntheticclaimthatspacehasagrouptheoreticstructurethat

    allowsforfreemobility.BytheHelmholtz-Lietheorem,6thisrequirementrestricts

    thepossiblegeometriestothosewithconstantnegative,positive,orzerocurvature,

    butdoesnotprivilegeoneoveranother.Moreover,FriedmanarguesthatPoincaré

    wascommittedtoahierarchyofthesciences,wherethesciencesareorderedby

    leveloffundamentalityinsuchawaythatnofactinamorefundamentalsciencecan

    6Stein1977.

  • 5

    bedeterminedbyafactofalessfundamentalscience.Thishierarchicalpicturethus

    rulesoutappealingtoempiricalfactstodecideonthecorrectgeometry.This

    readingthusgivesanargumentbyeliminationforconventionalism:

    1. Euclid'spostulateiseitherananalytictruth,asyntheticaprioritruth,an

    empiricaltruth,oraconvention.

    2. Theexistenceofmodelsofnon-EuclideangeometryshowthatEuclid's

    postulateisnotanalytic.

    3. Euclid’spostulateisnotsyntheticapriori,sincewecanknowapriori

    onlythatspacehasagrouptheoreticstructurethatallowsforfree

    mobility,which(byHelmholtz-Lie)doesnotdecidethetruthofEuclid’s

    postulate.

    4. Euclid'spostulatecannotbeanempiricaltruth,sincethesciencesare

    arrangedhierarchically.

    5. So,itisaconvention.

    Therearemanyfeaturesofthisinterpretationthatareattractive.Itexplains

    whyPoincaréwasledtohisconventionalismwhenothergeometerswerenot,and

    whyPoincarédidnotconsiderallofgeometrytobeconventional,letaloneall

    sciences.Moreover,thisreconstructionoftheargumentnicelycapturesthe

    argumentativestructureof“Non-EuclideanGeometries”(1891/1913)–Poincaré’s

    firstsustaineddefenseofconventionalism–whichclearlyarguesinaneliminative

    way.7Onemighttakeissuewithsomefeaturesofthisreconstruction,8butitisnot

    7AsimilareliminativeargumentappearsinPoincaré1887,214-6.8Inparticular,Dunlop2016arguesthatthereislittlegroundforattributingahierarchicalpictureofthesciencestoPoincaré.

  • 6

    mygoalinthispapertodecideonitsfidelity.(Indeed,asI’llclaimbelow,Idonot

    believethatthereisauniqueargumentthatPoincaréputsforwardfor

    conventionalism.Rather,hesupplementedhisargumentsandaddednewonesover

    thenearlytwentyyearsofhiswritingsonthephilosophyofgeometry.)Whatisnot

    wellknown,isthattherewasaquitedifferentinterpretationofPoincaréthatshares

    thesameinterpretivevirtuesandwasinfactofferedupinPoincaré’slifetimebyone

    ofhisforemostcritics.BertrandRussell,inhisearlybookEssayontheFoundations

    ofGeometry,arguedthatPoincaréwasledtohisconventionalismbyadistinctive,

    mathematicalinterpretationofthemodelsofnon-Euclideangeometry.This

    mathematicalinterpretationwascommontoPoincare,Klein,andCayley,but

    differedfromtheinterpretationthat,say,Beltramigaveofhismodel.

    SincethesesystemsareallobtainedfromaEuclideanplane,byamere

    alterationinthedefinitionofdistance,CayleyandKlein[thoughnot

    Beltrami]tendtoregardthewholequestionasone,notofthenatureof

    space,butofthedefinitionofdistance.Sincethisdefinition,ontheirview,is

    perfectlyarbitrary,thephilosophicalproblemvanishes…,andtheonly

    problemthatremainsisoneofconventionandmathematicalconvenience.

    ThisviewhasbeenforcefullyexpressedbyPoincaré*:"Whatoughtoneto

    think,"hesays,"ofthisquestion:IstheEuclideangeometrytrue?The

    questionisnonsense."Geometricalaxioms,accordingtohim,aremere

    conventions:theyare"definitionsindisguise."9

    9Russell1897,§33.RussellisquotingPoincaré1891/1913,39.

  • 7

    RussellgoesontoarguethatwhatmatteredforPoincaréwasthatdistancewasnot

    aprimitivenotionasitwasforBeltrami,butwasmathematicallydefinedusing

    intrinsicallynon-metricnotionsdrawnfromadifferentareaofmathematics

    (namely,projectivegeometry).AsI’llexplainbelow,RussellclaimsthatPoincaré’s

    argumentdependsondefiningthedistancebetweentwopointsintermsofthe

    crossratioofthosetwopointsandtwootherarbitrarilychosenpoints.But,Russell

    claimed,distanceisinfactaprimitivenotionandsotheargumentfor

    conventionalismcollapses.

    ThereareafewfeaturesofthisreadingthatIwouldliketohighlight.First,it

    assimilatesPoincaré’smathematicalworktoearlierworkbyCayleyandKlein.

    Second,itclaimsthatPoincaré’sargumentforconventionalismdependsonvery

    specificfeaturesofhismathematicalwork.10Third,thereconstructiondependsin

    nowayonDuhemianunderdeterminationarguments,andinfactdoesnotturnon

    theapplicationofgeometrytophysics.Rather,theargumentturnsonthe

    applicationofonemathematicaltheorytoanother.Insteadofarguingthatthereisa

    10RussellisnottheonlyreadertoseePoincaré’sconventionalismasdependentonhisparticularmathematicalwork.Zahar(1997,185)arguesthat"strictlyinternalfactorsconnectedwithhisworkonFuchsianfunctionsgaverisetohisso-called'conventionalism.'"Zahar'sprincipalconclusionisthatPoincarédidnotemployRiemanniangeometryinhisinvestigationsoftheinvariantsinFuchsianfunctions.ThisexplainswhyPoincaréwasnottemptedtoconsiderRiemanniangeometriesofvariablecurvatureaslegitimategeometries(1891/1913,37)–animportantmoveinhisdefenseofconventionalism.IagreewithZahar'sreadingofPoincaré'smathematicalworlkonFuchsianfunctions,astherestofthispaperwillshow.However,IbelievethattheconnectionbetweenPoincaré'smathematicalworkandhiscommitmenttoconventionalismrunsdeeper.Afterall,therewerecontemporariesofPoincaré's(e.g.,HelmholtzandRussell)whoalsodeniedthatRiemanniangeometriesofvariablecurvaturewerelegitimategeometries,andyetdidnottakethefurtherstepandembraceconventionalism.ThereneedstobesomeexplanationforwhyPoincaréinparticulartookthisfurtherstep.

  • 8

    loosenessoffitintheapplicationofpuremathematicstothephysicalworld–thus

    leavingopenadegreeoffreedomthathastoberestrictedbyconvention–itturns

    ontheloosenessoffitindefiningdistance(apurelymathematicalnotioninmetric

    geometry)usingtermsdrawnfromanotherareaofpuremathematics.

    Inthispaper,Iwanttoexplainandcriticallyevaluatethisreading.InSection

    I,I’llexplainRussell’sobjectiontoPoincaré,itsphilosophicalmotivations,andwhy

    itultimatelyfailsasareadingofPoincaré.InsectionII,I’llarguethatthereisa

    successfulmodifiedRussellianreading.Thatis,thereispresentinPoincaréan

    argumentforconventionalismfromthepossibilityofalternativedefinitionsof

    distancewithinpuremathematics.Thisargumentdoesnotderivefromthe

    underdeterminationofphysicalgeometrybyexperience,butbytheapplicationof

    onemathematicaltheorytoanother.

    SectionI:EarlyRussellversusPoincaré

    Russell’scriticismofPoincaré’sconventionalismappearedinhis1897book,Essay

    ontheFoundationsofGeometry[EFG].Thisbookappearedfiveyearsbefore

    Poincaré’sclassictreatmentofthespaceprobleminPartTwoofScienceand

    Hypothesis[SH],whichcollectstogetherandre-arrangespapersthatwerepublished

    between1891and1900.InEFG,Russellcitestwopapersthateventually

    reappearedinSH:“Non-EuclideanGeometry”((1891/1913),whichwasreprinted

    withminor–thoughsignificant–deletionsaschapter3),and“SpaceandGeometry”

    ((1895/1913),whichwasreprintedaschapter4ofSH,withoneparagraph

  • 9

    deleted).11ThefirstchapterofEFGistitled“AShorthistoryofMetageometry”and

    arguesthatthehistoryofmathematicalworkonnon-Euclideangeometryshouldbe

    dividedintothreeperiods,withthesecondperiodcharacterizedbytheuseof

    differentialgeometrybyGauss,Riemann,andBeltrami,andthethirdperiod

    characterizedbytheuseofprojectivegeometrybyCayleyandKlein.Itisinthis

    contextthatRussellarguesthatconventionalistinterpretationsofmodelsofnon-

    Euclideangeometrymakesenseonlywhenthesemodelsareconstructedusing

    techniquescharacteristicofthethirdperiod,notthesecond.

    ToseewhatRussellisgettingat,weneedtoexaminethedifferentroutesthat

    BeltramiandKleintooktoconstructingtheirmodelsofhyperbolicgeometry.12

    Beltramiusedmethodsfromdifferentialgeometry,and(inhis1868Saggio)

    constructedamodelof2-dhyperbolicgeometryinEuclidean3-space.Beltrami

    beganwiththestandardwayofprojectingthedistancefunctiondefinedonpoints

    onthesouthernhemisphereofasphereontoaplane(figure2).Theinversefunction

    11Russellquotes1891/1913at§33(quotedabove).Hementions1895/1913inalongfootnoteto§100,whichgivesalonglistofthe“mostimportantrecentFrenchphilosophicalwritingsonGeometry.” ThatfootnotealsolistsPoincaré1897,whichisPoincaré’sreplytocriticismsofhisphilosophyofgeometryleveledbyLechalasandCouturat.Thispaperisdevotedlargelytoline-by-lineresponsestospecificclaimsofLechalasandCouturat,andcontainsnonewclaimsmateriallydifferentfromthosein1891/1913and1895/1913.Moreover,thereisnoevidencethatRussellengagedwiththispaperatallinEFG,beyondsimplylistingit.SoIwillfollowRussell’sleadinignoringitinthispaper. Twoparagraphsaredeletedfrom1891/1913inch.3ofSH.Thelastparagraphissimplymovedonepageovertobecometheopeningofch.4ofSH.Thepenultimateparagraph–apassagedevotedtothepossibilityofdeterminingthecorrectgeometryempiricallybymeasuringstellarparallaxes–ismovedtoch.5ofSH.Seenote16andpage37below.12MypresentationofBeltramiisderivedfromStillwell1996.

  • 10

    thatmapsthecoordinateofthepointontheplane(x,-R,z)backontothesphereis

    then

    𝑥!,𝑦!, 𝑧! =𝑅

    (𝑅! + 𝑢! + 𝑣! (𝑥,−𝑅, 𝑧)

    Nextwedefineaninduceddistancefunctionds2onpointsontheplanesothatthe

    distancebetween(x1,-R,z1)and(x2,-R,z2)isjustthestandardlengthofthegeodesic

    onthesurfaceofthesphereconnecting(x1’,y1’,z1’)and(x2’,y2’,z2’).(Thisisthekind

    ofinducedfunctionthatmapmakersareinterestedinwhentheyrepresent

    distancesonaglobeusingflatmapsinanatlas.)Beltramithennoticedthatthis

    induceddistancefunctiondependedonlyonx1,z1,x2,z2,andR2,andsowouldbe

    meaningfulifRwasreplacedwithR√-1.BeltramithenshowedthatreplacingRwith

    R√-1inducesonaplaneanon-Euclideanmetricthathasconstantnegative

    curvature,andthepointsprojectedfromContotheplanearenowallinsideadisk

    (figure3).

    Klein("Ontheso-calledNon-EuclideanGeometry":Klein1871])arrivedat

    thesamemodelutilizingtechniquesforbuildingadistancefunctiononthecomplex

    projectiveplaneusingonlyresourcesdrawnfromprojectivegeometry.Kleinstarted

    withanideafromVonStaudt.VonStaudtshowedhowtoinduceasetofrational

    coordinatesonthepointsonaprojectivelinebyrepeatedapplicationofthe

    quadrilateralconstruction,apurelyprojectiveconstructionthatrequiresonlypencil

    andstraightedge(figure4).Itwasawell-knowntheoremofprojectivegeometry

    thatthefourpointsonalinepickedoutbythequadrilateralconstructionhavethe

    samecrossratio:

  • 11

    𝐶𝑅 𝑃,𝐴,𝐵,𝑄 =𝑃𝐴𝑃𝐵 ×

    𝑄𝐵𝑄𝐴

    anditwaswellknownthatthecrossratiooffourpointswasinvariantunder

    projection.VonStaudtthenshowedthatiftwoofthesepointsarearbitrarilylabeled

    nand∞,thenrepeatedapplicationsofthisconstructionwillassignthestandard

    Euclideanmetrictothepointsontheline.Ingeneral,ifwetakeaprojectiveline

    withtwopointsfixed(suchasPQinfigure5),wecandefineafunctionthatassignsa

    distancebetweentwopoints:

    𝑑 𝐴,𝐵 = 𝑐[log𝐶𝑅(𝑃,𝐴,𝐵,𝑄)]

    Employingthelogofthecrossratioensuresthatthedistancefunctionhastheright

    additiveproperty

    𝑑 𝐴,𝐵 + 𝑑 𝐵,𝐶 = 𝑐[log𝐶𝑅(𝑃,𝐴,𝐵,𝑄)]+ 𝑐[log𝐶𝑅(𝑃,𝐵,𝐶,𝑄)]

    = 𝑐[log𝐶𝑅(𝑃,𝐴,𝐶,𝑄)] = 𝑑(𝐴,𝐶)

    sincethecrossratioofalinesegmentistheproduct,notthesum,ofcrossratiosof

    itsparts:

    𝐶𝑅 𝑃,𝐴,𝐵,𝑄 ×𝐶𝑅 𝑃,𝐵,𝐶,𝑄 = 𝐶𝑅(𝑃,𝐴,𝐶,𝑄)

    Adistancefunctionforthewholeplanecouldthusbedefinedpurely

    projectively,usingnoundefinednotionsfrommetricgeometry–ifonlywehada

    systematicwaytopickouttwoarbitrarypoints,nand∞,onanylineintheplane.

    Buteverylineintersectsaconicinthecomplexprojectiveplaneintwopoints,and

    sowecandefinedistanceprojectivelyifwejustpickanarbitraryconicinthe

  • 12

    plane.13Infact,Cayley(“SixthMemoironQuantics”:Cayley1859)hadshownhowto

    constructthestandardEuclideandistancefunctioninthisway,inthespecialcase

    wheretheconicpickedoutisimaginaryanddegeneratesintoapointpair.Klein's

    ideawasthatdifferentkindsofmetricgeometrieswouldariseiftheconic(whichhe

    calledthe“fundamentalconic”)werechosendifferently.14Inparticular,heshowed

    thatifthefundamentalconicisimaginaryandnon-degenerate,thenwehave

    sphericalgeometry(spacesofconstantpositivecurvature),andifthefundamental

    conicisrealandnon-degenerate,thenwehavehyperbolicgeometryinthespace

    enclosedwithintheconic(spacesofconstantnegativecurvature).Thelattercaseis

    depictedinfigure5.ItiseasytoseethatBeltrami’smodelisamodelofthislatter

    Kleiniankind,wherethediskaroundtheoriginisaspecialcaseofareal,non-

    degenerateconic.

    KleinandBeltramithusproducedtheirmodelintwoconceptuallyquite

    distinctways.Beltramiusedtechniquesfromdifferentialgeometry,whose

    fundamentalobjectsarevariousspaceswithdistancefunctionsonthem;Kleinused

    techniquesfromprojectivegeometry,whosefundamentalobjectiscomplex

    projectiven-spacewithprojectiverelationsoutofwhichcanbeconstructedvarious

    distancefunctions.Russellwasabsolutelycorrect,then,toarguethatoneandthe 13Again,pickingaconicisprojectivelyacceptableaswell,sinceSteinerhadshownthatconicscanbeconstructedpoint-by-pointbytakingtheintersectionpointsoflinesthatstandtooneanotherinconstantcrossratios(Steiner1881,§41.I).14Klein’sadvanceoverCayleywastwofold.First,hegeneralizedCayley’sproceduretodistancefunctionsotherthanEuclideanones.Second,heusedthetechnicalresourcesfromVonStaudttoshowthatthedistancefunctionscouldbedefinedinapurelyprojectiveway.Thisremovedtheworryofcircularityindefiningdistanceintermsofcrossratio,whichwas–afterall–initiallydefinedastheproductofquotientsofstandardEuclideandistances.Athirdadvance,discussedbelow,isKlein’sinterpretationofhisclassificationofgeometriesgroup-theoretically.

  • 13

    samemodelcanbeunderstoodinmathematicallyquitedistinctways.Russell

    furtherassimilatedPoincaré’sapproachtoKlein’s.TounderstandRussell’sclaim,

    weneedtounderstandtherelationbetweenPoincaré’sspheremodel(a2-dversion

    ofwhichisfigure1)andtheKlein-Beltramimodel.

    AsI’llshowinsection2,Poincaréarrivedathismodel(inPoincaré1880a)

    whiletryingtorepresentgeometricallytheinversesofquotientsofsolutionsto

    secondorderlineardifferentialequations.Butthoughheproducedthismodelusing

    techniquesdistinctfrombothKlein’sprojectiveandBeltrami’sdifferentialmethods,

    heshowedgeometricallyhowitcouldbeconstructedfromtheKlein-Beltrami

    model.BeginbytakingaspherewhoseequatorialdiskisaKlein-Beltramimodel

    (figure6).Projecttheequatorialplaneontothesouthernhemispherebyparallel

    projectionsorthogonaltotheequatorialplane(figure7).Thestraightlinesinthe

    Klein-Beltramimodelarenowprojectedontocirculararcsthatmeettheequatorat

    rightangles.Thenprojectthebottomhemisphereontotheplanetangenttothe

    southpoleofthesphere,usingasthecenterofprojectionthenorthpoleofthe

    sphere(figure8).Thecirculararcsonthesouthernhemispherewillbeprojected

    ontocirculararcsofthedisk,andtherightanglesatwhichthecirculararcsonthe

    southernhemispheremeetthediskwillbeprojectedontorightangleswhere

    circulararcswithinthediskmeetthecircumferenceofthedisk.Theregionofthe

    planeboundedbytheprojectionoftheequatorofthespherewillbetheboundaryof

    Poincaré’smodel(figure1).

    RussellinterpretedPoincaré’sargumentforconventionalisminsuchaway

    thatitdependedonKlein’swayofconstructinghismodel.Therearethree

  • 14

    philosophicallysignificantfeaturesofKlein’sprocedurethatdistinguishitfrom

    Beltrami’s.First,inKlein’sconstructiontheEuclidean,hyperbolic,andspherical

    metricsarealldefinedononeandthesameunderlyingcomplexprojectiveplane.It

    thereforeseemsmorenaturaltodescribethesemetricgeometriesasthreedifferent

    waysoftalkingaboutthesameunderlyingreality.Second,inKlein’sconstruction

    thedistancefunctionisnotprimitive,butdefined.Itthereforeseemsnaturalto

    characterizethechoiceofametricasachoiceofadefinition,and–inasmuchas

    definitionsofwordsarearbitrary–asnotamatteroffact.15Third:

    Theprojectivegeometer…whenheintroducesthenotionofdistance,he

    definesit,intheonlywayprojectiveprinciplesallowhimtodefineit,asa

    relationbetweenfourpoints.(EFG,§37)

    RussellthereforereconstructsPoincaré’sargumentinthefollowingway.The

    Cayley-Kleinmetricdefinesd(A,B)intermsofthecrossratioofAandB,andtwo

    otherarbitrarilychosenpointsPandQ.Moreover,theproceduresforpicking

    arbitrarypointsP,Qonalinecorrespondtothethreegeometriesofconstant

    curvature.So,thechoiceamongthethreegeometriesofconstantcurvatureis

    arbitrary.

    OnRussell’sreading,Poincaré’sargumentforconventionalismdoesnot

    dependonanyfactsaboutphysicalgeometry,norontheunderdeterminationoftotal

    theoriesbyphysicalevidence.Thisisnotsurprising,sincePoincaré’sextended

    treatmentinScienceandHypothesisofthepurportedempiricalbasisofgeometryis

    15Onthesefirsttwopoints,seeEFG,§33,quotedabove.

  • 15

    ch.5(“ExperienceandGeometry”),whichappearedafterRussell’sbook.16Indeed,as

    DavidStump(1989,348)haspointedout,thereislittleindicationinthetextof

    1895/1913thatPoincaréintendedthesphereargumenttobemakingapointabout

    thewayinwhichphysicalgeometryisunderdeterminedbyourempiricalevidence.

    Instead,thesphereargumentissupposedtoshowthatwecanimaginespaces

    whereEuclid’saxiomfails.(Andinanycase,Russelldoesnotmentionthesphere

    argumentinEFG,despiteciting1895/1905.)Infact,Poincaréintroducesthesphere

    thoughtexperimentinthisway:

    Ifgeometricspacewereaframeimposedoneachofourrepresentations,consideredindividually,itwouldbeimpossibletorepresenttoourselvesanimagestrippedofthisframe,andwecouldchangenothingofourgeometry.Butthisisnotthecase;geometryisonlytheresumeofthelawsaccordingtowhichtheseimagessucceedeachother.Nothingthenpreventsusfromimaginingaseriesofrepresentations,similarinallpointstoourordinaryrepresentationsbutsucceedingoneanotheraccordingtolawsdifferentfromthosetowhichweareaccustomed.(1895/1913,49)

    Sothepointofthespheremodelisthatthestructureofspaceisnotsomethingthat

    isintrinsicinanygivenrepresentation,butemergesonlyfromthelawsof

    connectionamongrepresentations.AsPoincaréputsthepointafewpageslater

    (1895/1913,59),theobjectofgeometryisnotaformofsensibility,butaformof

    ourintellect.Andsowecanimagineadifferentgeometrybyimaginingaworld

    whereoursensationsrelatetooneanotheraccordingtodifferentlaws.Usingthe

    eliminativeargumentforconventionalismPoincaréintroducesin1887and 16Indeed,thebulk–i.e.sections4-7–ofPoincaré1902/1913(titled,“ExperienceandGeometry”)consistsofpassagesfromPoincaré1899and1900,wherePoincarédevelopedamoresustainedargumentagainsttheclaimthatEuclid’spostulateisanempiricaltruth,inresponsetoRussell’sclaiminEFGthatitstruthcouldbedecidedbyexperiment(EFG§140).Theexceptionis§3,aparagraphthatoriginallyappearedattheendofPoincaré1891/1913,butwasremovedwhenthechapterwasrepublishedinSH.Seealsop.37below.

  • 16

    1891/1913,thespheremodelisthereforeintendedtoshowthatEuclid’spostulate

    isnotasyntheticaprioritruth.

    RussellrejectsPoincaré’sconventionalismbecauseherejectsthesecondand

    thirdfeaturesofKlein’sprocedure.Inparticular,withrespecttothethirdfeature,

    Russellargues:

    Distance,intheordinarysense,remainsarelationbetweentwopoints,not

    betweenfour;anditisthefailuretoperceivethattheprojectivesensediffers

    from,andcannotsupersede,theordinarysense,whichhasgivenrisetothe

    viewsofKleinandPoincaré.(EFG,§37)

    Russellmaintainsthatdistanceisarelationbetweentwopointsbecausehewants

    geometrytoplayatranscendentalrole:toprovidetheformofexternality.Geometry

    should“permitknowledge,inbeingswithourlawsofthought,ofaworldofdiverse

    butinterrelatedthings”(§58).Distance,asaprimitivetwo-placerelation,allowsus

    todistinguishbetweentwopointswhilealsointerrelatingthem.Theabilityto

    perceivetwothingsatdistancefromoneanotheristhustheabilitytoperceive

    identityindifference–themostprimitiveabilitywithoutwhichwecouldnot

    cognizeobjectsofperception.Ifdistancewerearelationamongfourpoints,thenit

    wouldpresupposesomeperceptualwayofdistinguishingthosefourpointsfrom

    oneanother,andwouldthuspresupposetheformofexternalityinsteadof

    constitutingit.Andso:"BeltramiremainsjustifiedasagainstKlein"(§33).

    Russell’sinterpretationisnotcorrect.InhisreviewofEFG,Poincarépointed

    outthatRussellhadmisinterpretedhisargumentbymakingitdependentonthe

  • 17

    metricofCayleyandKlein(Poincaré1899,273).17Infact,inanotherpaper,

    publishedafterRussell’sbookbutbeforePoincaréknewofit,Poincareexplicitly

    deniedthattheproperwaytodefinedistancefromAtoBisintermsofthecross

    ratioofA,B,andtwootherpoints:

    [VonStaudt]obtain[s]themetricalproperties[by]definingaharmonic

    penciloffourstraightlines,takingasdefinitionthewell-knowndescriptive

    property.Thentheanharmonic[i.e.cross]ratiooffourpointsisdefined,and

    finally,supposingthatoneofthesefourpointshasbeenrelegatedtoinfinity

    theratiooftwolengthsisdefined.Thislastistheweakpointoftheforegoing

    theory,attractivethoughitbe.Toarriveatthenotionoflengthbyregarding

    itmerelyasaparticularcaseoftheanharmonicratioisanartificialand

    repugnantdetour.Thisevidentlyisnotthemannerinwhichourgeometric

    notionswereformed.(Poincaré1899,§XVII)

    DespitethefactthatRussell’sreadingmissesthemark,isthereamore

    successfulreadingofPoincaréthatsharesmanyofthefeaturesandvirtuesof

    Russell’stheory?SuchareadingwouldmakePoincaré’sargumentdependon

    specificfeaturesofhismathematicalworkingeometryandwouldexplainwhy

    conventionalismseemednaturaltohimwhenitdidnotforothergeometerswho

    understoodthemodelsofnon-Euclideangeometryequallywell.Itwoulddependon

    thespecificwayinwhichdistanceisdefinedusingresourcesfromadifferentareaof

    mathematics,andnotonthewaymathematicsisappliedinphysicalscience.Inthe

    nextsection,I’llarguethatthereissuchanargument.Inordertounderstand 17PoincaréwroteareviewofRussell’sbook(Poincaré1899)andafurtherpiece(Poincaré1900)inresponsetoRussell’sreply(Russell1900).

  • 18

    Russell’sreadingofPoincaré,wehadtolookatthedifferentmathematicalroutes

    thatBeltramiandKleintooktoarriveattheirmodels.Inthenextsection,we’lldo

    thesameforPoincaré.

    SectionII:FuchsianFunctions

    Poincaré’searliestargumentsforconventionalism,including1891/1913(which

    wasthefirstsustainedphilosophicaltreatmentofgeometry,andtheworkthat

    Russellcitesanddiscusses),explicitlydrawonhisearlierworkinpure

    mathematics.

    Ifgeometryisnothingbutthestudyofagroup,onemaysaythatthetruthof

    thegeometryofEuclidisnotincompatiblewiththetruthofthegeometryof

    Lobachevsky,fortheexistenceofagroupisnotincompatiblewiththatof

    anothergroup.(1887,215)

    Nothingremainsthenoftheobjection[thattheremaybeahidden

    contradictioninLobachevskiangeometry]aboveformulated.Thisisnotall.

    Lobachevski'sgeometry,susceptibleofaconcreteinterpretation,ceasesto

    beavainlogicalexerciseandiscapableofapplications;Ihavenotthetimeto

    speakhereoftheseapplications,noroftheaidthatKleinandIhavegotten

    fromthemfortheintegrationoflineardifferentialequations.(1891/1913,

    34).

  • 19

    Thefirstquotationisfromtheconcludingpagesof“Surleshypothèses

    fondamentalesdelagéométrie,”whichisamathematicalpaperthatendswitha

    pageandahalfofprogrammaticphilosophicalremarks.HerePoincaréclaimsthat

    thechoicetoacceptorrejectEuclid’spostulateisnotamatteroftruth,andhis

    argumentclearlydependsonthinkingofthevariousmetricgeometriesasarising

    fromselectingamongallthepossiblegroupsofcoordinatetransformationsonone

    andthesameunderlyingspace.Thesephilosophicalremarksarethefirststatement

    ofPoincaré’sconventionalism,andareexpandeduponinhisfirstpurely

    philosophicalessay,1891/1905.Inthelatteressay,Poincaréarguesthatwhat

    movesnon-Euclideangeometryfroma“merelogicalcuriosity”tosomethingmore,

    isnotitsphysicalapplications,butitsapplicationingrouptheoryandinthetheory

    ofdifferentialequations.

    TounderstandwhatPoincaréisgettingat,weneedtolookatPoincaré’s

    mathematicalworkfromtheearly1880s.Poincaré’sfirstapplicationofhyperbolic

    geometrywasinthetheoryofFuchsianfunctions.Fuchsianfunctionsarea

    generalizationofellipticfunctions,whichwereoneofthemostwidelystudied

    topicsinnineteenthcenturymathematics.18Inordertokeeptrackoftheirorigins

    andapplications,weneedtointroduceellipticfunctionsinthecontextofelliptic

    integralsandellipticcurves.Anellipticintegralisanintegraloftheform

    𝑅 𝑡, 𝑝(𝑡) 𝑑𝑡

    18SeeBottazziniandGray2013,Ch.1.Foradiscussionofthephilosophicalsignificanceofellipticfunctions,seeTappenden2006.MypresentationhereisindebtedtoStillwell2010,chapter12.

  • 20

    whereRisarationalfunction19andp(t)isapolynomialofdegree3or4.An

    especiallysimpleexampleofsuchanintegralisthelemniscaticintegral

    𝑑𝑡1− 𝑡!

    !

    !

    whichgivesthearclengthofthelemniscateofBernoulli.TheCartesianequationof

    Bernoulli’slemniscateis

    𝑥2 + 𝑦! ! = 𝑥2 − 𝑦!

    anditsgraphisfigure9.

    TheseintegralshavebeenstudiedsinceLeibniz.Theyappearinvery

    elementarysettings:e.g.,theintegralthatgivesthearclengthofanellipseisan

    ellipticintegral–hencethename.Earlyon,itwasdiscoveredthattheycannotbe

    expressedintermsofelementaryfunctions,whichfrustratedtheintuitivenatural

    idea(beginningwithLeibnizhimself)thatthesolutionofeveryintegrationproblem

    shouldbeexpressedintermsofelementaryfunctions.20Abreakthroughcame

    around1800whenGauss(and,later,AbelandJacobi)studiedtheirinversesand

    consideredtheirbehaviorinthecomplexplane.Theinverseofanellipticintegralis

    calledanellipticfunction.Forexample,theinverseoftheleminscaticintegralGauss

    calleda"lemniscaticsinefunction,"standardlyabbreviatedsl(x),onanalogywith

    thesinefunction.Recallthatthesinefunctionistheinverseof

    sin!! 𝑥 =𝑑𝑡1− 𝑡!

    !

    !

    19Arationalfunctionisaquotientofpolynomials.20Afunctioniselementaryifitcanbedefinedbyarithmeticaloperationsonafinitenumberofexponentials,logarithms,constants,andnthroots.

  • 21

    whichistheequationforthearclengthofacircle.21Thesinefunction,togetherwith

    itsfirstderivative–thecosinefunction–,canbeusedtoparameterizetheequation

    ofacircle

    𝑥2 + 𝑦! = 𝑟!

    withx=cost=sin'tandy=sint.Justassintcanbeusedtoparameterizethe

    equationofacircle,sotoocanellipticfunctionsbeusedtoparameterizethe

    equationofcertaincurves.Thatis,iff(x)isanellipticintegral,thentherearecurves

    thatcanbeparameterizedsothat

    𝑥 = 𝑓!!(𝑢)

    and

    𝑦 = 𝑓!!′(𝑢)

    wheref-1(x)isanellipticfunction.Thosecurvesthatcanbeparameterizedby

    ellipticfunctionscametobecalledellipticcurves.22

    MathematiciansbeforeRiemannstandardlydefinedellipticfunctionsin

    termsoftheirpowerseriesexpansions.AbelandJacobiinthe1820sdiscoveredthat

    ellipticfunctionsaredoublyperiodic,andafterRiemannmathematiciansbeganto

    definethemsoastohighlighttheirdoubleperiodicity.Consideragainthe

    elementarysinefunction.Thesinefunctionissinglyperiodic(seefigure10):itis

    invariantundersubstitutions

    𝑥 ↦ 𝑥 + 2𝑛𝜋

    21Thisiswhysin-1iscalled“arcsin.”22Infact,anycubiccurvecanbeparameterizedbyanellipticfunction–aresultannouncedbySteinerbutfirstprovenbyClebschin1864.

  • 22

    Interpretedgeometrically,thismeansthatwecanslidetheentireplane2πunits

    alongthexaxisntimeswithoutchangingwhichpointslieonthefunctionsinx.

    Gaussnoticedthatthelemniscaticsinefunctionisdoublyperiodicinthecomplex

    plane:

    𝑓 𝑥 = 𝑓(𝑤 +𝑚𝜔! + 𝑛𝜔!)

    withm,nintegersandω1andω2complexnumbers.Thiscanberepresented

    geometricallybyaplanetiledbyparallelograms,asinfigure11.Themappingof*-

    pointstopointsonthecurvewillbeunaffectedby

    𝑥 ↦ 𝑥 +𝑚𝜔! + 𝑛𝜔!

    Interpretedgeometrically,thismeansthatwecanslidetheentirecomplexplaneω1

    unitsalongoneaxisntimesandω2unitsalongtheotheraxismtimeswithout

    changingthevaluesoftheellipticfunction.Thesemappingsarejustrigid

    (Euclidean)translationsoftheplanethatkeepthetilingintact.Therigid

    translationsthusformagroup,witheachelementofthegroupcorrespondingtoa

    tile.Forexample,theparallelogramwhosebottomleftcorneristheorigin

    correspondstothegroupidentity,thatis,theoperationthatleavesallpointsonthe

    planeunaffected.Theparallelogramwhosebottomleftcorneristhepoint(ω1,0)

    correspondstotherigidtranslationthatmoveseverypointtotherightbyω1.And

    soon.

    Thetheoryofellipticfunctionsprovidedagreatsimplificationand

    unificationinthetheoryoffunctions.Forinstance,inlecturesdeliveredin1874-5,

    Weiserstrass23showedthatanyanalyticfunctionofasinglecomplexvariablethat

    23SeeBottazinniandGray2013,424-9.

  • 23

    admittedanalgebraicadditionformulawasexpressibleasarationalfunctionofthe

    simplestellipticfunction,whichWeierstrasscalled℘

    𝓅 𝑧 =1

    𝑧 +𝑚𝜔! + 𝑛𝜔! !

    !

    !,!!!!

    whichparameterizestheellipticcurve

    𝑦! = 4𝑥! − 𝑔!𝑥 − 𝑔!.

    Butcouldthistheorybefurthergeneralized?Aretherefruitfulgeneralizationsof

    ellipticfunctions–generalizationsthatwouldparameterizealargerclassof

    algebraiccurvesandfurthersimplifythetheoryofcomplexfunctions?Inpapers

    from1880LazarusFuchstriedtodopreciselythis.Fuchsbeganwithsecondorder

    lineardifferentialequationswithrationalcoefficients:

    𝑦!! + 𝑃 𝑧 𝑦! + 𝑄 𝑧 𝑦 = 0

    wherePandQarerationalfunctions.Sincethefunctionissecondorderandlinear,

    allofitssolutionscanbeexpressedaslinearcombinationsoftwosolutions,f(z)and

    ϕ(z).Fuchsthenclaimed(Fuchs1880)thatthequotientofthesetwosolutions

    𝜁(𝑧) =𝑓(𝑧)𝜑 𝑧

    couldundercertainconditionsbeinvertedtoformawell-defined,single-valued,

    meromorphic24function:

    𝐹 = 𝜁!!.

    24Ameromorphicfunctionisafunctionthatisholomorphicexceptatisolatedpoints.Aholomorphicfunctionisafunctionthatiscomplexdifferentiableintheneighborhoodofeverypointinthecomplexplane.(AsPoincaréshowed,theisolatedpointsatwhichaFuchsianfunctionisnotholomorphicarethepointsontheboundaryofthedisk.)

  • 24

    ItisthisfunctionFthatisanalogoustoanellipticfunction,with𝜁akintoanelliptic

    integral.

    Inhisprizeessaysubmittedon28May1880,25Poincaré–thenayoungand

    unknownmathematician–showedthattheconditionsFuchsidentifiedwereneither

    necessarynorsufficientforFtobesingle-valued,well-defined,andmeromorphic

    (1880a,331).Underwhatconditions,then,doesFexistwiththespecified

    properties?PoincarébeganwiththefactthatanyFwouldbeinvariantunderlinear

    fractionaltransformations

    𝐹 𝑧 = 𝐹 !"!!!"!!

    .26

    So,insteadofbeingdoublyperiodic,Fwouldbeinvariantunderlinearfractional

    transformations:

    𝑧⟼ !"!!!"!!

    .

    Hefurthershowed(Poincaré1880a,346ff.)thatthepathofFwouldbeconfined

    insideacurvilinearpolygon oαγα’untilitspathcrossesaboundary(say,oα),at

    whichpointFwouldtraceoutanidenticalpathwithinthenewcurvilinearpolygon

    oαγ1α’1.WhenFcrossesaboundaryofthisnewpolygon(say,oα’1),itagainrepeats

    itspathwithinyetanewcurvilinearpolygonoα1γ2α’1–andsoon(seefigure12).

    Eachofthesepolygonscanbegeneratedfromtheoriginalpolygonbyrepeated

    applicationsoflinearfractionaltransformations,andareboundedbyarcsofcircles

    25MyunderstandingofPoincaré’sworkonFuchsianfunctionsdrawsonaseriesofworksbyJohnStillwellandJeremyGray:(Stillwell1985),(Gray1986,ch.6),(Gray1999),(BottazziniandGray2013),(GrayandWalter1997),(Gray2013).26(1880a,318).ThisfactfollowsmoreorlessimmediatelyfromthefactthatFistheinverseofthequotientofsolutionstoadifferentialequation,whereeverysolutionisalinearcombinationoftwosolutions,f(z)andϕ(z).

  • 25

    thatmeetatrightanglesacircleHH’centeredontheorigin.Therewillfurthermore

    beaninfinitenumberofthesepolygons,whichwillcovertheinteriorofthecircle

    HH’(figure1).ItfollowsthatFonlyexistswithinthediskHH’.(1880a,352).

    ButisFsingle-valued,asFuchsclaimed?Thisalldependsonwhetherthe

    curvilinearpolygonsgeneratedwithinthediskasFtracesitspatheveroverlap:if

    thepolygonsdooverlap,thenFwillbemulti-valuedintheoverlappingregion

    (1880a,351).Toshowthatthereisnooverlap,Poincarétakesthediskcoveredin

    curvilinearpolygonsandprojectsitstereographicallyandthenorthogonallyonto

    theequatorialplaneofthesphereinthewaydescribedinsection1(seefigures6,7,

    and8).Thecurvilinearpolygonsareprojectedontorectilinearpolygons,andsince

    simpleelementarygeometricalreasoningshowsthattheserectilinearpolygonsdo

    notoverlap,therectilinearpolygonsdon’teither.Fisthuswell-defined,single

    valued,andmeromorphic.

    Poincarécalledthesefunctions“Fuchsianfunctions.”Moreformally

    (Poincaré,1881a,47-8):aFuchsianfunctionisanymeromorphicfunctionthatis

    invariantundera“Fuchsiangroup,”whereaFuchsiangroupisadiscontinuous27

    groupoflinearfractionaltransformationsonthecomplexplanethatleaveinvariant

    acirclearoundtheorigin.Thatis,aFuchsiangroupisagroupofoperationsonthe

    complexplanethatleavethetessellationofthediskHH’intocurvilinearpolygons

    intact.EachFuchsiangroupcorrespondstoawayoftessellatingthediskHH’with

    curvilinearpolygons,andeachelementofagroupcorrespondstoacurvilineartile.

    27Agroupisdiscontinuousifitdoesnotcontainaninfinitesimaloperation.Fuchsiangroupshavetobediscontinuous,becausetheycorrespondtowaysofmovingthepointswithinthediskthatleavethetilingintact,andnotileisinfinitesimal.

  • 26

    Inhisprizeessay,PoincaréhadfoundanexampleofaFuchsianfunction–a

    functionwhoseFuchsiangroupcorrespondstothetessellationofthediskwitha

    certainkindofquadrilateral.ButthiswasjustoneexampleofaFuchsianfunction,

    andjustonewayoftessellatingHH’.Inthecaseofellipticfunctions,the

    correspondingtessellationsoftheplanearesimpletounderstand,sincetheyareall

    akintoparallelogramtessellations.28ThetessellationscorrespondingtoFuchsian

    groups,ontheotherhand,areinfinitelyvarious,andtheprojectofprovinggeneral

    propertiesofFuchsianfunctionscouldnotproceedunlessthereweresomegeneral

    waytosurveyallofthewaysthatHH’couldbetessellatedintocurvilinearpolygons

    whosesidesarecirclesmeetingHH’inrightangles.29AsPoincaréputitinapaper

    from1881:“ItisnecessaryfirsttoconstructallFuchsiangroups;thisIhavedone

    withtheaidofnon-Euclideangeometry”(1881a,48).

    Inthemonthaftersubmittinghisprizeessay,Poincarérealizedthatthe

    rectilinearpolygonsontowhichhehadprojectedthecurvilinearpolygonsofHH’in

    factweretheKlein-Beltramimodelofhyperbolicgeometry,andsothetessellations

    28Stillwell1985,19.29TherelationbetweenFuchsiangroupsandFuchsianfunctionsisrathersubtle.Fuchsianfunctionscannotbepairedup1-1withFuchsiangroups,anditneedstobeshownthatforeveryFuchsiangroupthereexistFuchsianfunctions.Onthefirstpoint,PoincarédiscoveredthatanytwoFuchsianfunctionsthatcorrespondtothesameFuchsiangrouparerelatedalgebraically(1881b)–afactthateventuallyledhimtohisfamousuniformizationtheorem.Onthesecondpoint,PoincaréprovedthateveryFuchsianfunctioncouldbeconstructedasthequotientoftwo“theta-Fuchsian”functionsthatcorrespondtothesameFuchsiangroup,whereatheta-FuchsianfunctionΘisameromorphicfunctionsuchthat

    Θ𝑎𝑧 + 𝑏𝑐𝑧 + 𝑑 = Θ(𝑧)(𝑐𝑧 + 𝑑)

    !!withmaninteger.Theexistenceoftheta-Fuchsianfunctionscouldthenbeprovedbytheconvergenceofaninfiniteseries.(Thisresultisannouncedin1881a,andprovedsystematicallyin1882b).

  • 27

    ofHH’inducedbyFuchsiangroupswerealsomodelsofhyperbolicgeometry.30Ina

    supplementtotheprizeessaywrittenonJune281880(Poincaré1880b),Poincaré

    describedthesituationasfollows:

    Thereisadirectconnectionbetweentheprecedingconsiderationsandthe

    non-EuclideangeometryofLobachevskii.Whatindeedisageometry?Itis

    thestudyofagroupofoperationsformedbythedisplacementsonecan

    applytoafigurewithoutdeformingit.InEuclideangeometrythisgroup

    reducestorotationsandtranslations.Inthepseudo-geometryof

    Lobachevskiiitismorecomplicated…TostudythegroupofoperationsM

    andN[viz,theoperationsthatmoveonepolygoninHH’ontoanother]is

    thereforetohavetodothegeometryofLobachevskii.Thepseudogeometry,

    asaconsequence,isgoingtofurnishuswithaconvenientlanguagefor

    expressingwhatwewillhavetosayaboutthisgroup.31

    TherelationbetweenFuchsiangroupsandnon-Euclideanisometrieswaslaidout

    systematicallyinsections1and2of1882a.There,Poincarédefinedtwofigures

    withinHH’ascongruentiftheycanbetransformedintooneanotherbyalinear

    fractionaltransformationwherea,b,c,anddarerealnumbers.32Heusedthis

    30Thisrealization–which,Poincaréclaimed,hithimoutoftheblueashewasboardingabusonaminingexpedition--wasfamouslydescribedinPoincaré1909.JeremyGrayhasshownthatthisrealizationmusthavetakenplacebetweenMay29andJune281880(1986,266-8).31QuotedandtranslatedinGray1986,258-9.Graydiscoveredthesesupplements,whichwerepreviouslyunpublished,anddescribedtheircontentsinGray1986.Theyhavesincebeenpublished(inFrench)asPoincaré1997.32Therequirementthata,b,c,anddberealnumbersforcesthelinearfractionaltransformationtokeeptherealaxisinvariant,andthusmodelsnon-Euclideangeometryinthehalfofthecomplexplanelyingabovetherealaxis.Infact,in1882a,PoincarérepresentsFuchsiangroupsusingtheupper-halfplanemodelinsteadof

  • 28

    notiontodefinestraightline,length,area,anddistance.Thepreviouslyintractable

    problemofidentifyingFuchsiangroupshadthusbeenreducedtothetractable

    problemofidentifyingnon-Euclideanisometries.

    Withthisapplicationofnon-Euclideangeometry,Poincaréwasableto

    achieveextraordinaryresultsthatgeneralizeinpowerfulwaystheresultsobtained

    usingellipticfunctions.Thisisbecause,asPoincaréputitinthesupplementfrom

    June281880,“theFuchsianfunctionsaretothegeometryofLobachevskiiwhatthe

    doublyperiodicfunctionsaretothatofEuclid”(1880b,translatedinGray1986,

    269).Morespecifically,aFuchsiangroupwherethelinearfractionaltransformation

    issuchthat(a+d)2=4isagroupofdiscontinuousEuclideanisometries(Poincaré

    1882a,58)–namely,thegroupoftranslationsthatkeepsthetilingoftheplaneinto

    parallelogramsintact(figure11).InthiswayFuchsianfunctionscomprise“avery

    extensiveclassoffunctionsofwhichtheellipticfunctionsareaspecialcase”(1881b,

    54).Justasellipticfunctionshadbeenusedtointegratealgebraicdifferentials,

    Poincaréshowedthatamuchwiderclassofequations,lineardifferentialequations

    withalgebraiccoefficients,couldbesolvedusingFuchsianfunctions(1882a,55).I

    notedabovethatellipticfunctionscanparameterizecertainkindsofalgebraic

    curves,so-calledellipticcurves–aclassthatincludesallcubicsandsomeother thePoincarédiskmodel.(TheupperhalfplanemodelresultsfromthePoincarédiskbyprojectingitstereographicallybackontothesouthernhemisphere(figure8),switchingthesouthernandnorthernhemispheresonthesphereandthenprojectingthenorthernhemispherefromapointontheequatorontoaplanetangenttothesphereatthepointontheequatoroppositethepointofprojection.)Heswitchedfromthedisktothehalf-planemodelinresponsetoanobjectionfromKlein,whodoubtedthateveryFuchsiangroupcouldberepresentedasanon-Euclideantessellationofthediskmodel(Gray1986,280,285-7).Poincaréusesthehalf-planemodelto“translate”hyperbolicgeometryintoEuclideangeometryin1891/1913,33-34.

  • 29

    specialcases.In1881,Poincaréannouncedthediscoveryofhisuniformization

    theorem,thatanyalgebraiccurvewhatsoevercanbeparameterizedbyFuchsian

    functions.Thatis,ifA(x,y)=0istheequationofanalgebraiccurve,itcanbe

    rewrittenasA(f(t),φ(t))=0,withfandφFuchsianfunctions(Gray1999,81).

    SectionIII:FromFuchsianFunctionstoConventionalism

    WenowunderstandwhatPoincarémeantwhenhespokeofthe“applications”of

    hyperbolicgeometry,and“theaidthatKleinandIhavegottenfromthemforthe

    integrationoflineardifferentialequations.”Poincarédidnotsimplyfindamodelof

    non-Euclideangeometryinasurprisingplace.Rather,thisapplicationofhyperbolic

    geometryincomplexanalysisprovideddeepandpowerfultheoremsinthetheory

    ofalgebraiccurvesandthetheoryoflineardifferentialequations–resultsthat

    couldnothavebeenobtainedotherwise.Thisworkwasthehighpointofthe

    extremelyactiveandfruitfulnineteenthcenturyworkincomplexanalysisthat

    beganwithGauss.Itbroughttogetheranalysis,geometry,andalgebraina

    surprisingandmutuallyilluminatingway.

    ItisfurtherclearthatthismathematicalworkonFuchsianfunctions

    providedPoincaréthemotivationforhisconventionalism.Toseethepossibilityof

    applyingnon-Euclideangeometry,Poincaréhadtomaketwonovelconceptual

    moves.First,hehadtothinkofthegroupofoperationsthatleavethevaluesofa

    Fuchsianfunctioninvariantaswaysofmovingaspatialobject–thegraphofthe

    function–aroundinspacewithoutchangingitsshapeorsize.Second,hehadto

  • 30

    conceiveofgeometryasfundamentallythestudyofthegroupofrigidmotionsof

    bodiesinspace.Asheputthissecondpointinhisfirstsupplementtohisprizeessay,

    whichhewrotejustoneortwoweeksafterseeingthatFuchsiangroupsarenon-

    Euclideanisometries:

    Whatindeedisageometry?Itisthestudyofagroupofoperationsformedby

    thedisplacementsonecanapplytoafigurewithoutdeformingit.(1880b,

    translatedinGray1986,258)

    Butoncegeometryisconceivedofinthisway,itisashortsteptoconcludingthat

    Euclideangeometryisnomoretruethannon-Euclideangeometry.Afterall,many

    differentkindsofgroupsofrigidbodycoordinatetransformationscanbedefinedon

    oneandthesamecomplexplane.Andthisisapointthat,again,Poincaréasserts

    explicitlyinPoincaré1887–amathematicalpaperthatappearedfouryearsbefore

    hisfirstphilosophicalpaper:

    Ifgeometryisnothingbutthestudyofagroup,onemaysaythatthetruthof

    thegeometryofEuclidisnotincompatiblewiththetruthofthegeometryof

    Lobachevsky,fortheexistenceofagroupisnotincompatiblewiththatof

    anothergroup.(1887,215,quotedabove)

    Afterall,thechoiceofagroupofdisplacementsinthecomplexplaneamountsto

    choosingwhichwaytotessellatetheplane.Butoneandthesamecomplexplanecan

    betessellatedinmanyways.Itmakesnomoresensetoask“Whichistherightway

  • 31

    totessellatetheplane,figure1orfigure11?”thanitdoestoask“Whicharethetrue

    functionsinthecomplexplane,ellipticorFuchsianfunctions?”33

    Thiswayofconceivingofgeometry–asthestudyofthegroupofrigid

    displacementsofbodies–iscontroversial.Itisafarcryfromtheolderviewthat

    geometryisthestudyofextensivemagnitude,anditisalsoopposedtothenewer

    viewthatgeometryisfundamentallythestudyofspace.34Moreover,itisalso

    opposedtoRussell’sviewthatgeometryisfundamentallythestudyofdistance,

    wheredistanceisaprimitivenotionnotdefinableintermsofothernotions.In 33TherearetwowaysinwhichPoincaré’spresentationofhisconventionalisminhislater,philosophicalpapersrefinestheargumentpresentedintheoffhandphilosophicalcommentshemakesin1880band1887.First,in1887PoincarésaysthatEuclideangeometryis“nomoretrue”thanitsnon-Euclideanrivals.Itakethistobealessperspicuouswayofsayingwhathecametosaylater,thatthechoiceofametricforspaceisnotamatteroftruthorfalsehoodatall,butinsteadamatterofconvention(wheremattersofconventionareopposedtomattersoffact).Second,thegroupofrigiddisplacementsinspaceisacontinuousgroup,notadiscontinuousgroup.Poincaréwaswellawareofthisfactfromhisearliestdiscussionsin1881aand1882a,wherehewouldfirstdefinethehyperbolicgroupof(continuous)displacementsandthendefineaFuchsiangroupbytakingadiscontinuoussubgroupthatleavesthefundamentalcircleHH’invariant.Inhislater,philosophicalpapers,suchas1898,hewouldmakeclearthatgeometryisthestudyofthecontinuousgroupofdisplacements–whichisaLiegroup,notaFuchsiangroup.Itakethesetwochangestoberefinementsoftheargumentfirstadumbratedintheseoffhandphilosophicalcommentsinhismathematicalpapers,thoughnotrefinementsthatinanywayalterthespiritoftheargument.34ForPoincare,spaceisnottheprimitivenotionofgeometry–thenotionofagroupofdisplacementsis.Hemakesthispointexplicitlyin1895/1913.There,afterdescribing“aparticularclassofphenomenawhichwecalldisplacements,”heassertsthatthe“lawsofthesephenomenaconstitutetheobjectofgeometry”(48);“itisfromthepropertiesofthisgroupwehavederivedthenotionofgeometricspace”(52). JeremyGrayhasnicelyemphasizedthedistinctivenessofPoincaré’sconceptionofgeometry.CommentingonPoincaré1898,hewrites:“Poincaréinsistedonananalysisofdistancethatwastheopposite,headmitted,oftheoneheldbyHelmholtz,Lie,andalmosteveryoneelse.Thesemathematicianssaidthatthematterofthegroupexistedbeforeitsform,thematterbeingthree-dimensionalmanifoldofspace.Whereasforhimself,saidPoincaré,theisomorphismclassofthegroupweusetoconstructspacecomesfirst”(Gray2013,56).

  • 32

    1882a,Poincarébeginswiththeideaofalinearfractionaltransformationthat

    leavesthefundamentalcircleinvariant.Hethendefinestwofiguresascongruentif

    theycanbetransformedintooneanotherbysomechosenlinearfractional

    transformationthatleavesthecircleinvariant.Last,thedistancefromAtoBequals

    thedistancefromCtoDifthecirculararcthatconnectsAandBiscongruenttothe

    circulararcthatconnectsCtoD.ForPoincaré,distanceisnotprimitive,butis

    defined.

    Inthisway,RussellgetssomeofPoincaré’sargumentright,andsomeofit

    wrong.AsIexplainedinsectionI,hepurportedtoidentifythreeimportantfeatures

    ofPoincaré’sargument:first,thedifferentgeometriesarealldefinedinthesame

    underlyingcomplexplane;second,distanceisadefined,notprimitivenotion;third,

    distanceisafour-placerelation,notatwoplacerelation.Russellbelievedthatthe

    argumentforconventionalismreliedessentiallyonthethirdpurportedfeature,

    whichherejectedasartificial.WecannowseethatRussellgotPoincarébadly

    wrong.MissingfromRussell’sinterpretationistheallimportantnotionofagroupof

    rigiddisplacements;instead,heassimilatesPoincaré’smathematicalworktoKlein’s

    approach,whichremainedweddedtoprojectivewaysofthinkingthatPoincaré

    rejected.Still,thoughRussell’sthirdclaimmissesthemark,heiscorrecttoseethat

    Poincaré’sargumentdoesrelyonthefirsttwopoints:bothEuclideanandnon-

    EuclideangeometryareconstructedinPoincaré’swaybytakingthesame

    underlyingcomplexplaneanddefininganotionofdistanceintermsofsomemore

    fundamentalmathematicalnotionthatisimportedfromanotherareaof

    mathematics.

  • 33

    InsectionI,InotedthatPoincaré’searliestargumentsforconventionalism–

    theargumentsthatRussellknewwhenwritingEFG–arepresentedasargumentsby

    elimination:somefeatureofgeometryiseitherananalytictruth,asyntheticapriori

    truth,empiricaltruth,oraconvention;butitisnotanyofthefirstfourdisjuncts;so,

    itisaconvention.Theinterpretivequestion,again,iswhichdisjunctPoincaréwas

    hopingtoeliminatebyinvokinghismodelsofhyperbolicspaceinEuclideanspace.It

    isuncontroversial,ofcourse,thatthesemodelsshowthatthereisnocontradiction

    innegatingEuclid’saxiomofparallels,andsoEuclid’saxiomscannotallbeanalytic.

    Butarethesemodelsintendedtodomorethanthis?OnDuhemianreadings,these

    modelsdemonstratefurthertheempiricalunderdeterminationofgeometry,andso

    showthatgeometryisnotempirical.IarguedinsectionI,ontheotherhand,that

    Poincaré’sdiscussionofthemodelin1895/1913makesclearthatheintendeditto

    underminetheclaimthatEuclid’saxiomofparallelsisasyntheticaprioritruth.

    Furtherconfirmationofthisreadingisprovidedbythepassage–quotedalreadyin

    thefirstparagraphofsectionII–wherePoincarépointstothe“applications”ofnon-

    Euclideangeometryinthetheoryoflineardifferentialequations.There,after

    concludingthatEuclid’saxiomisnotananalytictruthhesays,“Thisisnotall.

    Lobachevski'sgeometry,susceptibleofaconcreteinterpretation,ceasestobea

    logicalexerciseandiscapableofapplications”(emphasisadded).The

    “interpretation”ofhyperbolicgeometryisofcoursetheuseofmodelsofnon-

    Euclideangeometryinthecomplexplane.Inotherwords,theexistenceofthese

    modelsshowsthatnon-Euclideangeometrydoesnotcontraveneanyanalytictruth;

    theirapplicationshowsthatnon-Euclideangeometryisnotjustlogicallypossible–

  • 34

    itdoesnotviolateanysyntheticaprioritrutheither.Amathematiciansuchas

    BeltramiwhocouldproduceamodelofhyperbolicgeometryinEuclideangeometry

    wouldbeabletoseethatitislogicallypossible.Amathematicianwhounderstood

    theapplicationsofhyperbolicgeometrywouldbeabletoseethatitisreallypossible

    aswell.

    WecanfruitfullycomparewhatPoincarésaysaboutLobachevskii’sgeometry

    withthe“geometries”ofvariablecurvatureproposedbyRiemann.Afewpageslater,

    Poincaréwrites“thesegeometriesofRiemann,inmanywayssointeresting,could

    neverthereforebeotherthanpurelyanalytic”(1891/1913,37)However,though

    mostmathematiciansthinkofLobachevski'sgeometryalso“onlyasamerelogical

    curiosity,”Poincarédisagrees,arguingthatitsimpossibilityisnotshownby

    “syntheticapriorijudgments,asKantsaid”(37).Anotherwayofputtingthisclaimis

    thatRiemann’sgeometries,butnotLobachevskii’s,areinconsistentwithsome

    necessaryfeatureofgeometry.Andthisnecessaryfeatureofgeometryisprecisely

    whatPoincaréidentifiedinJune1880,whenhefirstannounceshisdiscoverythat

    hisgraphicalrepresentationsofFuchsiangroupsaremodelsofLobachevskiian

    geometry:thatgeometryisthestudyofagroupofoperationsformedbythe

    displacementsonecanapplytoafigurewithoutdeformingit.Riemann’sgeometries

    ofvariablecurvature,sincetheydonotallowdisplacementwithoutdeformation,

    arethusnotreallygeometries.Theyarejustlogic,notgeometry.

    Butwhyconceiveofgeometryasfundamentallythestudyofgroupsofrigid

  • 35

    displacements?35NoargumentisgiveninthemathematicalpapersPoincaré1880b

    and1887,norinhisfirstphilosophicalpaper,1891/1913.However,onevery

    powerfulmotivationforthisconceptionisprovidedbytheapplicationsthat

    PoincarémadeofhyperbolicgeometryinthetheoryofFuchsianfunctions.These

    applicationsrequiredthinkingofFuchsiangroupsaslikegroupsofdisplacementsof

    rigidbodies,andgeometryasthestudyofsuchgroups.Thinkingofgeometryinthis

    wayallowsustotransfergeometricreasoningaboutcomplexellipticfunctionsinto

    geometricreasoningaboutFuchsianfunctions–itallowsustomaketheall-

    importantanalogybetweenellipticfunctionsandFuchsianfunctions.The

    extraordinarypoweroftheresultsobtainedmakesthisreconceptualizationvery

    attractive.Furthermore,thiswayofthinkingofgeometryalsoallowedforPoincaré

    tounifyalgebra,geometry,andthetheoryoffunctionsinahighlyilluminatingway–

    thekindofwaythatwouldstronglysuggesttoaworkingmathematicianthathehad

    hitontherightwayofthinkingofgeometry.

    Thispaperbeganwithaquestion,"HowdoesPoincaréargueforhis

    conventionalismaboutgeometry?"Thispaperhasidentifiedoneway–infact,the

    35Thehistoricalquestion–Fromwhomdidhegetthisidea?–isnoteasytoanswer.OnemightbetemptedtoconcludethathegotitfromKlein’sErlangerProgramm,ifonlytherewereanyhistoricalevidencethatPoincaréknewofKlein’sworkinJune1880.WhentheKlein-Poincarécorrespondencebeginsoneyearlaterin1881,itisclearthatPoincarédidnotknowKlein’swork,norindeedvirtuallyanyGerman-languagework.AmoreplausiblehypothesisisthatPoincaréwasinspiredtodevelopthisideafromreadingHelmholtz’sessays,whichweretranslatedintoFrenchinthe1870s(Gray2013,40).Intheseessays,Helmholtz–thoughhecertainlydoesnotdeveloptheviewthatgeometryisthestudyofthegroupofdisplacements–arguesthatwecometoknowEuclid’saxiomfrommovingarigidmeasuringrodthroughspace.

  • 36

    earliestway,whichwascontainedalreadyinhismathematicalpapersfromthe

    1880s--thatPoincaréarguedforconventionalism.Thislineofargumentis

    distinctiveinasmuchasitinnowayreliesonDuhemianunderdetermination,indeed

    doesnotdependonfactsabouttheapplicationofgeometrytophysicalscienceatall.

    Instead,thisargumentarisesnaturallyfromreflectingontheverypowerfulresults

    Poincaréobtainedbyapplyingonemathematicaltheory(metricgeometry)to

    another(thetheoryofFuchsianfunctions).Moreover,Ibelievethatthe

    mathematicalresultsdescribedinthispaperexplainclearlywhyPoincaré,having

    donethekindofmathematicalworkthathehaddone,foundconventionalismso

    natural,whereasothermathematicianswhowerealsoawareofEuclideanmodelsof

    hyperbolicgeometryweredrawntootherphilosophicalviews.Whatinitiallyand

    powerfullydrewPoincarétoconventionalism,then,werenotjustphilosophical

    reflectionsonspace,geometry,andphysics,butveryspecificfeaturesofhisworkin

    puremathematics.

    Ofcourse,Poincaré'sargumentsforconventionalismevolvedandwere

    expandedovertime,partiallyasaresponsetocriticisms,andpartiallyasaresultof

    Poincaré–whosewide-ranginggeniusventuredintonearlyallofthesciences–

    expandinghisreflectionsonspaceandgeometryintophysicsandpsychology.For

    example,aswe'veseen,Poincaré'sargumentfortheconventionalityofaspatial

    metricwasoftenposedasaquadrilemma.Theexistenceofmodelsofnon-Euclidean

    geometryshowthatthetruthofaparticularmetriccannotbeanalytic;thepowerful

    applicationsofbothEuclideanandhyperbolicmetricsincomplexanalysisshowthat

    itcannotbesyntheticapriori;andsince(Poincaréarguedinitially)itcouldnot

  • 37

    possiblybeempirical,itmustbeconventional.However,inPoincaré'searliest

    philosophicalpapers(1891/1913,38and1895/1913,53),theargumentagainst

    empiricismisweakandundeveloped.Heargues,briefly,thatgeometryisanexact

    andunrevisablescience,whereasallempiricalsciencesareonlyapproximatelytrue

    andaresubjecttoconstantrevision.Partiallyasaresultofreadingandreviewing

    Russell1897(whicharguedthatmetricgeometryisempirical),Poincarébeganto

    refineandexpandhisargumentsagainstempiricism.Theseargumentsarecollected

    inCh.5ofScienceandHypothesis(Poincaré1902/1913),whichconsistslargelyof

    reprintedsectionsofhisexchangewithRussell(Poincaré1899andPoincaré

    1900).36Furthermore,geometryisintimatelyconnectedwithperceptual

    psychology,sinceweperceiveobjectsinspaceandgeometricalnotions(suchas

    point,line,anddistance)appearinperceptualpsychologyaswell.Poincarébeganin

    1895/1913(andcontinuinginchapters3and4ofPoincaré1905)toputtogethera

    richtheoryoftheoriginsofspatialnotionsinperception–atheorythatreinforces 36Asmentionedearlier(notes11and16)thepenultimateparagraphofPoincaré1891/1905–apassagedevotedtothepossibilityofdeterminingthecorrectgeometryempiricallybymeasuringstellarparallaxes–wasmovedtoch.5ofSH,whereitappearsontheopeningpage(page81)ofthepaper,followedinthesubsequentpagesbyotherparagraphsthathadappearedinPoincaré1899and1900.Inthispassage,Poincarépointsoutthat,ifLobachevskiangeometryweretrue,theparallaxofdistantstarswouldhaveapositivelimit,providedthatlightraystravelalwaysinstraightlines.Butshouldexperimentsshowthatindeedstellarparallaxesdohaveapositivelimit,weareleftwithtwochoices:rejectEuclideangeometryorrejecttheclaimthatlightraystravelinstraightlines.Thelatterwouldbemoreadvantageous,andso"Euclideangeometryhasnothingtofearfromnewexperiments." Poincarédoesnotconnectthislineofreasoningtothespheremodelin1895/1913,which(asremarkedabove)isusedtoshowthatwecanimagineanon-Euclideanworld.Thereasoninginthispassageisverybrief,thoughclearlysuggestive.Thereisgoodreason,then,whyPoincaréneededtofurtherexpandanddeepenhisarguments(inthepaperspublishedin1899andlater)againsttheclaimthatthemetricofspaceisempirical.

  • 38

    theargumentforconventionalismbycapturingtheobjectivityofspatialperceptions

    ingroup-theoreticterms.

    Theselaterdevelopmentsarepowerfulandsuggestive.Itisnotsurprising,

    then,thatdebateovertheconventionalismofgeometrybegantofocusmoreon

    Poincaré'sreflectionsontherelationsbetweengeometry,physics,andpsychology,

    asopposedtohisoriginalargumentsbasedontherelationsbetweengeometryand

    otherbranchesofpuremathematics.Inconcertwithotherhistoricaldevelopments

    –hereweshouldcertainlyincludethereceptionandexpansionofPoincaré's

    argumentsbylogicalempiricistssuchasSchlickandReichenbach–theconnections

    betweenPoincaré'sconventionalismandhismathematicalworkinFuchsian

    functionswasde-emphasized,leadingtothesortofreadingsrecountedinthe

    openingpagesofthispaper.Buttheseconnectionsareworthremembering,

    becausetheymotivateadistinctiveargumentforconventionalismbasedonthe

    loosenessoffitbetweentwomathematicaltheories.Andtheyshowclearlythatit

    wasnoaccidentalfactthatamathematicianwhoobtainedthekindofresultsusing

    thekindofmethodsPoincaréemployedwasledtoconventionalism.

  • 39

    WORKSCITED

    Beltrami,E.1868."Saggiodiinterpretazionedellegeometrianon-Eucliden,"GiornalediMath.,VI:284-312.Ben-Menahem,Yemima.2006.Conventionalism.CambridgeandNewYork:CambridgeUniversityPress.Bottazzini,UmbertoandJeremyGray.2013.Hiddenharmony--geometricfantasies:theriseofcomplexfunctiontheory.NewYork,NY:Springer.Cayley,Arthur.1859.“SixthMemoironQuantics.”PhilosophicalTransactionsoftheRoyalSocietyofLondon149:61-90.Dunlop,Katherine.2016."PoincaréOnTheFoundationsOfArithmeticAndGeometry.Part1:Against“Dependence-Hierarchy”Interpretations."HOPOS:TheJournaloftheInternationalSocietyfortheHistoryofPhilosophyofScience6:274-308.Friedman,Michael.1999."Poincaré'sConventionalismandtheLogicalPositivists."InReconsideringLogicalPositivism(CUP),71-86.Fuchs,Leonard.1880."UbereineKlassevonFunktionenmehrererVariablenwelchedurchUmkehrungderlntegralevonLösungenderlinearenDifferentialgleichungenmitrationalenCoeffizientenentstehen."JahrschriftfürreineundangewandteMathematik89:150-169.Gimbel,Steven.2004."Un-conventionalwisdom:theory-specificityinReichenbach’sgeometricconventionalism."StudiesinHistoryandPhilosophyofModernPhysics35:457-481.Gray,Jeremy.1985.LineardifferentialequationsandgrouptheoryfromRiemanntoPoincare.Boston:Birkhaüser.Gray,Jeremy.1999.Non-Euclideangeometryinthetheoryofautomorphicfunctions.editedbyJeremyGrayandAbeShenitzer.IntroductionbyJeremyGray.Providence,R.I.:AmericanMathematicalSociety.Gray,Jeremy.2013.HenriPoincaré:ascientificbiography.Princeton,NJ:PrincetonUniversityPressKlein,Felix.1871."Ontheso-calledNon-EuclideanGeometry."Tr.ByJohnStilwellinStillwell1996,69-110.Poincaré,Henri.1880a."ExtractfromanuneditedmemoirofHenriPoincaré."Ed.byN.E.Nörlund.TransbyJohnStillwell.InStillwell1985,305-356.

  • 40

    Poincaré,Henri.1880b."PremierSupplément."InPoincare1997,31-74.Poincaré,Henri.1881a."Surlesfonctionsfuchsiennes."Comptesrendusdel'AcademedesSciences92(14Feb1881):333-335.TranslatedbyStillwellas"OnFuchsianFunctions,"inPoincaré1985,47-50.Poincaré,Henri.1881b."Surlesfonctionsfuchsiennes."Comptesrendusdel'AcademedesSciences92:(21Feb1881):395-398.TranslatedbyStillwellas"OnFuchsianFunctions,"inPoincaré1985,51-54.Poincaré,Henri.1882a."TheoriedesgroupesFuchsiens."ActaMathematicaI:1-62.TranslatedbyJohnStillwellas"TheoryofFuchsianGroups,"inStillwell1985,55-127.Poincaré,Henri.1882b."MemoiresurlesfonctionsFuchsiennes."ActaMathematicaI:193-294.TranslatedbyJohnStillwellas"OnFuchsianFunctions,"inStillwell1985,128-254.Poincaré,Henri.1887.“Surleshypothèsesfondamentalesdelagéométrie.”BulletindelaSociétémathématiquedeFrance15:203-216.Poincaré,Henri.1891/1913."Lesgéométriesnoneuclidiennes."Revuegénéraledessciencespuresetappliquées2(1891):769-774.Reprinted,withminoralterations,inPoincaré1902.TranslatedbyHalstedas“TheNon-EuclideanGeometries,”inPoincaré1913,29-39.Poincaré,Henri.1895/1913."L'espaceetlagéométrie."Revuedemétaphysiqueetdemorale3(1895):631-646.Reprinted,withminoralterations,inPoincaré1902.TranslatedbyHalstedas“SpaceandGeometry,”inPoincaré1913,40-54.Poincaré,Henri.1897."Réponseàquelquescritiques."Revuedemétaphysiqueetdemorale5:59-70.Poincaré,Henri.1898."OntheFoundationsofGeometry."Monist9(1898):1-43.ReprintedinWilliamEwald,ed.,FromKanttoHilbert,vol.2(1996,OUP),982-1011.Poincaré,Henri.1899."Desfondementsdelagéométrie;àproposd'unlivredeM.~Russell."Revuedemétaphysiqueetdemorale7:251-279.Poincaré,Henri.1900."Surlesprincipesdelagéométrie;réponseàM.~Russell."Revuedemétaphysiqueetdemorale8:73-86.Poincaré,Henri.1902.LaScienceetl'hypothèse.Flammarion.

  • 41

    Poincaré,Henri.1902/1913.“ExperienceandGeometry.”InPoincaré1902.TranslatedbyHalstedinPoincaré1913,55-65.Poincaré,Henri.1905.LaValeurdelascience.Flammarion.Poincaré,Henri.1909."L'inventionmathématique."Annéepsychologique15:445-459.Poincaré,Henri.1913.TheFoundationsofScience.Trans.byGeorgeBruceHalsted.NewYork:theSciencePress.Poincaré,Henri.1985.PapersonFuchsianFunctions.Trans,withanintroduction,byJohnStilwell.SpringerVerlag.Poincaré,Henri.1997.Troissupplémentssurladécouvertedesfonctionsfuchsiennes.J.GrayandS.Walter,eds,Berlin:AkademieVerlag.Russell,Bertrand.1897.EssayontheFoundationsofGeometry.CUP.Russell,Bertrand.1900."Surlesaxiomsdelagéométrie."Revuedemétaphysiqueetdemorale7:684-707.Stein,Howard.1977.“SomePhilosophicalPre-HistoryofGeneralRelativity."InFoundationsofSpace-TimeTheories,MinnesotaStudiesinPhilosophyofScience,Vol.8,J.Earman,C.Glymore,J.Stachel(eds.).UniversityofMinnesotaPress.Steiner,Jakob.1881.SystematischeEntwicklungderAbhängigkeitgeometrischerGestaltenvoneinander.InGesammelteWerke.Vol.1.,editedbyK.Weierstrass(Berlin:G.Reimer).Stillwell,John.1985."Translator'sIntroduction."InHenriPoincaré,PapersonFuchsianFunctions.Trans,withanintroduction,byJohnStilwell.SpringerVerlag.Stillwell,John.1996.SourcesofHyperbolicGeometry.AmericanMathematicalSociety.Stillwell,John.2010.MathematicsanditsHistory.Springer.Stump,David.1989,"HenriPoincaré'sPhilosophyofScience."StudiesinHistoryandPhilosophyofModernPhysics20:335–363.Tappenden,Jamie.2006."TheRiemannianBackgroundtoFrege'sPhilosophy."InTheArchitectureofModernMathematics:EssaysinHistoryandPhilosophy.Ed.byJ.FerreirósandJ.J.Gray.OUP.

  • 42

    Zahar,Elie.1997.“Poincaré’sPhilosophyofGeometry,orDoesGeometricConventionalismDeserveItsName?”StudiesinHistoryandPhilosophyofModernPhysics28:183–218.

  • Figures

    Figure1:thePoincaréDiskmodelofhyperbolicgeometry

    Figure2:Projectionontotheplaneofthedistancefunctiononasphere

  • Figure3:BeltramiModelofHyperbolicGeometry

    Figure4:Thequadrilateralconstruction

  • Figure5:TheKleinmodel

    Figure6:TheBeltrami-Kleinmodelontheequatorialplane

  • Figure7:TheBeltrami-Kleinmodelprojectedontothesouthernhemisphere

    Figure8:TheBeltrami-KleinmodelprojectedontothePoincarémodel

    Figure9:TheLemniscateofBernoulli

  • Figure10:Thesinefunction

    Figure11:Tessellationofthecomplexplanebyellipticfunctions

    Figure12:TessellationofthediskinthecomplexplanebyaFuchsianfunction