Ihaveseenamillionblackcrows。CanIsaythatthemillionandonethcrowwillbeblack？Toanswerthiswemustaskwhetherblacknessisapropertyof’kind。’79*Iftheblacknessbe，’asitwere，anaccident，’ornotapropertyofkind，itmust，hesays，beacaseofcausation。Ifnotacaseofcausation，itmustbeapropertyofkind。Hencewehavethesingularresult，thatifthecoexistencebecasual，itiscaused，andifinvariable，notcaused。As’causation’meansaccordingtohimsimplyunconditionalconnection，thestatementseemstobeespeciallyparadoxical。Itis，however，explicable。
Theblacknessofthecrowmayberegardedas’accidental，’ifitisduetotheexternalcause。Thecrow，perhaps，hasfallenintoapaint-pot。Theblacknessis’caused，’then，bythepropertiesofpaintandbythe’accidental’collocation。Itisan’accident’inthecrow，thoughcausedinrespectofthegeneralarrangementsoftheuniverse。Butwhy，ifapropertyof’kind，’
shoulditbecalled’notcaused’？HerewehaveacuriousresultofMill’sviewofcausation。Ournaturalreplywouldbethatthecolourisstillcausedaseverythingelseiscaused。80*Weassume，thatis，that’crow’impliessuchaconstitutionthatunderagivenenvironmentcrowswillbeblack。Changesomethingoutsidethecrowandhemayturnwhite。Orfindawhitecrowinthesame’environment，’andweinfersomedifferenceinhisconstitution。Thereisarelation，weassume，thoughwecannotspecifyitsnature，whichdeterminesthecolour，andasinallcaseswehaveatoncecollocationandcausation。HereisMill’speculiardifficulty。Causation，asheisprofoundlyconvinced，alwaysmeansabeginning。Itisonly，aswehaveseen，concernedwithchanges，notwithpersistence。Therefore，iftwothings，likeblacknessandcrowness，existsidebyside，itisacaseofcollocation，andconsequently，ashesupposes，notacaseofcausation。Hecannotrecogniseareciprocalrelation，althoughitisclearthatifonethingisfoundalwaystoaccompanyanother，theargumentisthesameasthoughonealwaysfollowedanother。
Indeed，hiswholetheoryofinductionimpliesthepossibilityofreasoningfromonepropertyorattributetoanother。Makeachangeinoneandtheothermustbechanged。Heseesthisclearlyinthecaseoforganisedbodies。81*Inthatcase，hesays，thereisa’presumption’thatthepropertiesare’derivable’andtherefore’caused，’becausetherewehavesequencesoroneprocessfollowinganother。Hethusseemstolimithis’naturalkinds’chieflytochemicalcompounds。Therewehavepropertieslyingsidebysideandnot’derivable，’thatis，nottobeinferredbyusfromthepropertiesoftheelementaryconstituents。Theveryattempttoderivethemisidle。Asanyeventmaycauseanyother，howeverunlike，soanysetofpropertiesmaybesimplystucktogether。Baconisagainreproved82*forassumingthat’everyobjecthasaninvariablecoexistent。’Theultimateproperties，sofaraswecanconjecture，are’inherentlypropertiesofmanydifferentkindsofthings，notalliedinanyotherrespect。’Theysimplyliesidebyside，withoutreferencetoeachother。ThusMillpusheshisempiricismtoassumingnotonlythatourknowledgeofpropertiesmustrestupondirectobservation，butthatthereisabsolutelynoconnectionor’cause’tobeknown。The’kind’afterall，whichwasmeanttobeanessentialbond，turnsouttobeitselfapurelyarbitrarycollectionofattributes，andwehavetoaskwhetheritdoesnotloseallthesignificancewhichheattachedtoit。The’collocation’meansthattheattributessimplyliesidebyside，andyetarealwaysconjoined。Thetiewhichcombinesthemisundiscoverable，andthereforeforusnon-existent。Itis，asherightlyinsists，importantthatourclassificationshouldcorrespondtonaturalkinds。’Kinds，’hesays，areclasses’betweenwhichthereisanimpassablebarrier’；
thelogicalclassisarbitrary，buttherealclassisanessentialfact，Hisillustrationisremarkable。Heholdsthatthe’speciesofplantsarenotonlyrealkinds，butareprobablyallofthemreallowestkinds，infimaespecies，’andthatfurthersubdivisionwouldleadtonovaluableresults。83*Thedoctrinethatthespeciesofbotanymustcorrespondto’realkinds’iscuriousinawriterwhowashimselfabotanistandfamiliarwiththedifficultyofmakingabsolutedivisionsbetweenkinds。TheconflictwiththeconceptionsimpliedinDarwinismisofthehighestimportance。84*Thedistinctionbetween’kinds，’
accordingtoDarwin，isnotabsolute，foritistheproductofgradualdivergencefromasingleform。But，ontheotherhand，thekindsexistingatagiventimearediscrete。Therearegapsbetweenthem，asMillremarks；though，insofarastheyhaveacommonorigin，notabsolutelyinsuperablegaps。Thisimpliesthattheorganismdoesnotcorrespondtoamereaggregateofdisconnectedattributes，sothatthedifferenceofkindswouldbesimplyadifferenceofmoreorless，andeachtypepassintotheotherbyimperceptiblegradations。Weareobligedtosupposeasystemofreciprocalrelations，sothatanychangeinoneorganinvolvescorrelatedchangesinothers；andthusspeciesdivergealongdifferentlinesinsteadofremainingconstantorsimplyaddingonnewproperties。Mill，itseems，hastoadmitofkindsinordertoaccountforthepossibilityofinference；butthen，ashewishestoavoid’mysticalbonds，’andinferencesfrom’definitions，’andthescholasticbeggingsofthequestion，hedeclarestherelationbetweentheattributestobe’accidental’
or’uncaused。’Hence，thoughheseesthedifficultyandrecognisestheprobabilityof’causation’inorganisedbodies，hereallyreducesthe’kind’tobeamereaggregate，anddestroystheveryorganicbondofwhichheisinneed。
VII。UNIFORMITYOFNATURE
Theeffectofthuscontra-distinguishing’collocation’fromcausation，andadmittingthat’uncausedcoexistences’coveralargepartofallobservablephenomena，istomaketheuniformityofnatureexceedinglyprecarious。Indeed，Milldeniesittobeconclusivelyproved。Thechapterinwhichhesumsup’theevidenceofthelawofuniversalcausation’leadstoremarkableresults。Noone，hethinks，withaproperlytrainedimaginationwillfindanydifficultyinconceivingthatinremotepartsoftheuniverse’eventsmaysucceedoneanotheratrandomwithoutanyfixedlaw’。Heconcludesbyassertingthatitwouldbe’follytoaffirmconfidently’thatthelawdoesprevailin’distantpartsofthestellarregions。’85*Atruthwhichdependsuponlocalitymight，foranythingonesees，breakdowninAustraliaandevenatParisaseasilyasatSirius。Mill，accordingly，reachesathoroughlyscepticalconclusion，86*andreducestheevidenceforuniversalcausationtoaninductionperenumerationemsimplicem。Thewiderthegeneralisation，thegreatertheefficacyofsuchinduction，uponwhichdependsnotonlythelawofcausationbuttheprinciplesofnumberandgeometry。Ifthe’subject-matterofanygeneralisation’besowidelydiffusedthatitcanbetestedateverytimeandplace，andifit’beneverfoundotherwisethantrue’itstruthcannotdependonanycollocations，unlesssuchasexistatalltimesandplaces，norcanitbefrustratedbyanycounteractingagenciesexceptsuchasneveractuallyoccur。’87*Nownoexceptiontothe’lawofcausation’haseverbeenfound，andapparentexceptionshaveonlyconfirmedit。Itisnodoubttruethatifalawbeuniversal，itwillbeconfirmedbyallourexperiments；
butithardlyfollowsthat，becauseallourexperimentshavefailedtodetectanexception，itistrueuniversally。Allourexperimentshavecoveredbutasmallfragmentofnature，andtheydonotjustifyus，asheexpresslyasserts，inreasoningaboutthestellarregions。Itisdifficult，moreover，toseehowan’exception’couldeverbeproved，since，whereverwedonotseea’cause，’wecanalwayssuppose，anddoinfactsuppose，aninvisiblecause。Finally，thetheoryof’naturalkinds，’asithasnowbeeninterpreted，seemstofailusinourneed。Hetakesittoindicate，indeed，thatthereareconnectionsinnature，which，ifknown，wouldjustifycertaingeneralinferences；butitdoesnotappearthatwecanknowwhataretheseconnections，andas，moreover，wehavebeencarefullytoldthattheyareultimateornot’derivative，’wehavenorighttobecertainthattheywillrecur。Wedonotknow，forexample，whetherblacknessbeapropertyofkind。Ifwefoundablackcrowamongwhiteones，thepropertywouldbecasual，andtherefore’caused。’IfwefoundaraceofwhitecrowsinAustralia，weshouldsimplysaythattherewasakindhithertooverlooked。88*Suchadiscovery，hesays，isnotatallincredible。Itmightbeprovedbytheevidenceofasinglecrediblewitness。Itmerelysupposesthatthereisakindwithadifferentsetofattributes，andastheattributesareinnoway’derivative，’thereisnoimprobabilityinthis。Themoregeneraltherule，however，thegreatertheprobabilityofitsholding，becausethegreatertheimprobabilityoftheexceptionbeingoverlooked。Weshouldeasilybelieveinwhitecrows，butnotsoeasilyincrowswithaproperty’atvariancewithanygenerallyrecogniseduniversalpropertyofbirds，’andstillless，ifitwere’atvariancewithsuchapropertyofanimals。’89*Wecouldhardly，thatis，believeincrowswiththestomachsofwolves，orincrowswithoutstomachsatall。Butthedifficultyappearstodependuponnothingelsethantheimprobabilitythatsuchanimals，hadtheyexisted，wouldhavebeenunnoticed。
Withouttryingfullytounravelthislogic，wemaynoticeonecharacteristic。Mill，tryingtorefereverythingto’experience，’
hasgonefartomakeexperienceimpossible。Whathasdroppedoutofthistheoryofknowledgeistheconstructivepart。Hesubstitutesfororganisationcombinationsofdisparate’things。’
Hewilladmitofnologic，exceptthat。ofanexternalconnectionofradicallydifferentobjects。Attributesmustbestucktogetherwithoutanyreciprocalrelations。Allcausationbecomesinhisphrase’collocation，’thoughhedeclarescausationandcollocationtobenotonlydifferentbutmutuallyexclusive。Hisonelogicalformulaisthenotanotaeestnotareiipsius。90*
Thingsaremarksofeachother，notimpliedbyeachother。Heforcesthislanguageevenuponmathematics。Evenageometrical’kind，’ifwemayusetheword，anellipse，oracurveofthesecondorder，istreatedbytheformulaapplicabletopurelyempiricalconjunctions。Theequalityoftwostraightlines，itseems，issimplya’mark’thatifappliedtoeachothertheywouldcoincide；thefactthattwothingsaresumsofequalsisa’mark’thattheyareequal，andsoforth。91*Hewouldapparentlybeinclinedtosaythatathing’sexistenceisamarkofitsnotbeingnothing。Thus，eventhe’naturalkind’becomesmerelyapermanentcombination。Whenthepropertiesofacurvearemerelyconnectedby’marks，’itisnowonderthatthepropertiesofcrowsshouldbemerebundles。Ifitisonlyonsuchtermsthatwecanthoroughlygetridof’intuitions，’theadvantageisdoubtful。