A。Ruger&ClaraE。Bussenius（1913）Author’sPrefaceIntherealmofmentalphenomena，experimentandmeasurementhavehithertobeenchieflylimitedinapplicationtosenseperceptionandtothetimerelationsofmentalprocesses。Bymeansofthefollowinginvestigationswehavetriedtogoastepfartherintotheworkingsofthemindandtosubmittoanexperimentalandquantitativetreatmentthemanifestationsofmemory。Theterm，memory，istobetakenhereinitsbroadestsense，includingLearning，Retention，AssociationandReproduction。
　　Theprincipalobjectionswhich，asamatterofcourse，riseagainstthepossibilityofsuchatreatmentarediscussedindetailinthetextandinparthavebeenmadeobjectsofinvesti-gation。Imaythereforeaskthosewhoarenotalreadyconvincedapriorioftheimpossibilityofsuchanattempttopostponetheirdecisionaboutitspracticability。
　　Theauthorwillbepardonedthepublicationofpreliminaryresultsinviewofthedifficultyofthesubjectinvestigatedandthetime-consumingcharacterofthetests。Justicedemandsthatthemanydefectsduetoincompletenessshallnotberaisedasobjectionsagainstsuchresults。Thetestswereallmadeuponmyselfandhaveprimarilyonlyindividualsignificance。Naturallytheywillnotreflectexclusivelymereidiosyncrasiesofmymentalorganisation;
　　iftheabsolutevaluesfoundarethroughoutonlyindividual，yetmanyarelationofgeneralvaliditywillbefoundintherelationofthesenumberstoeachotherorintherelationsoftherelations。Butwherethisisthecaseandwhereitisnot，IcanhopetodecideonlyafterfinishingthefurtherandcomparativeexperimentswithwhichIamoccupied。ClassicsintheHistoryofPsychology——Ebbinghaus（1885）Chapter1Memory：AContributiontoExperimentalPsychologyHermannEbbinghaus（1885）TranslatedbyHenryA。Ruger&ClaraE。Bussenius（1913）CHAPTERIOURKNOWLEDGECONCERNINGMEMORYSection1。MemoryinitsEffectsThelanguageoflifeaswellasofscienceinattributingamemorytothemindattemptstopointoutthefactsandtheirinterpretationsomewhatasfollows：
　　Mentalstatesofeverykind，——sensations，feelings，ideas，——whichwereatonetimepresentinconsciousnessandthenhavedisappearedfromit，havenotwiththeirdisappearanceabsolutelyceasedtoexist。Althoughtheinwardly-turnedlookmaynolongerbeabletofindthem，neverthelesstheyhavenotbeenutterlydestroyedandannulled，butinacertainmannertheycontinuetoexist，storedup，sotospeak，inthememory。Wecannot，ofcourse，directlyobservetheirpresentexistence，butitisrevealedbytheeffectswhichcometoourknowledgewithacertaintylikethatwithwhichweinfertheexistenceofthestarsbelowthehorizon。Theseeffectsareofdifferentkinds。
　　Inafirstgroupofcaseswecancallbackintoconsciousnessbyanexertionofthewilldirectedtothispurposetheseeminglyloststates（or，indeed，incasetheseconsistedinimmediatesense-perceptions，wecanrecalltheirtruememoryimages）thatis，wecanreproducethemvoluntarily。Duringattemptsofthissort，——thatis，attemptstorecollect——
　　allsortsofimagestowardwhichouraimwasnotdirected，accompanythedesiredimagestothelightofconsciousness。Often，indeed，thelatterentirelymissthegoal，butasageneralthingamongtherepresentationsisfoundtheonewhichwesought，anditisimmediatelyrecognisedassomethingformerlyexperienced。Itwou1dbeabsurdtosupposethatourwillhascreateditanewand，asitwere，outofnothing;itmusthavebeenpresentsomehoworsomewhere。Thewill，sotospeak，hasonlydiscovereditandbroughtittousagain。
　　Inasecondgroupofcasesthissurvivalisevenmorestriking。Often，evenafteryears，mentalstatesoncepresentinconsciousnessreturntoitwithapparentspontaneityandwithoutanyactofthewill;thatis，theyarereproducedinvoluntarily。Here，also，inthemajorityofcasesweatoncerecognisethereturnedmentalstateasonethathasalreadybeenexperienced;thatis，werememberit。Undercertainconditions，however，thisaccompanyingconsciousnessislacking，andweknowonlyindirectlythatthe"now"mustbeidenticalwiththe"then";yetwereceiveinthiswayanolessvalidproofforitsexistenceduringtheinterveningtime。
　　Asmoreexactobservationteachesus，theoccurrenceoftheseinvoluntaryreproductionsisnotanentirelyrandomandaccidentalone。Onthecontrarytheyarebroughtaboutthroughtheinstrumentalityofotherimmediatelypresentmentalimages。Moreovertheyoccurincertainregularwayswhichingeneraltermsaredescribedundertheso-called’lawsofassociation。’
　　Finallythereisathirdandlargegrouptobereckonedwithhere。Thevanishedmentalstatesgiveindubitableproofoftheircontinuingexistenceeveniftheythemselvesdonotreturntoconsciousnessatall，oratleastnotexactlyatthegiventime。Employmentofacertainrangeofthoughtfacilitatesundercertainconditionstheemploymentofasimilarrangeofthought，eveniftheformerdoesnotcomebeforetheminddirectlyeitherinitsmethodsorinitsresults。Theboundlessdomainoftheeffectofaccumulatedexperiencebelongshere。Thiseffectresultsfromthefrequentconsciousoccurrenceofanyconditionorprocess，andconsistsinfacilitatingtheoccurrenceandprogressofsimilarprocesses。Thiseffectisnotfetteredbytheconditionthatthefactorsconstitutingtheexperienceshallreturnintototoconsciousness。Thismayincidentallybethecasewithapartofthem;itmustnothappentoatoogreatextentandwithtoogreatclearness，otherwisethecourseofthepresentprocesswillimmediatelybedisturbed。Mostoftheseexperiencesremainconcealedfromconsciousnessandyetproduceaneffectwhichissignificantandwhichauthenticatestheirpreviousexistence。Section2。MemoryinitsDependenceAlongwiththisbareknowledgeoftheexistenceofmemoryanditseffects，thereisabundantknowledgeconcerningtheconditionsuponwhichdependthevitalityofthatinnersurvivalaswellasthefidelityandpromptnessofthereproduction。
　　Howdifferentlydodifferentindividualsbehaveinthisrespect！
　　Oneretainsandreproduceswell;another，poorly。Andnotonlydoesthiscomparisonholdgoodwhendifferentindividualsarecomparedwitheachother，butalsowhendifferentphasesoftheexistenceofthesameindividualarecompared：morningandevening，youthandoldage，findhimdifferentinthisrespect。
　　Differencesinthecontentofthethingtobereproducedareofgreatinfluence。Melodiesmaybecomeasourceoftormentbytheundesiredpersistencyoftheirreturn。Formsandcolorsarenotsoimportunate;andiftheydoreturn，itiswithnoticeablelossofclearnessandcertainty。
　　Themusicianwritesfortheorchestrawhathisinnervoicesingstohim;
　　thepainterrarelyrelieswithoutdisadvantagesolelyupontheimageswhichhisinnereyepresentstohim;naturegiveshimhisforms，studygovernshiscombinationsofthem。Itiswithsomethingofastrugglethatpaststatesoffeelingarerealized;whenrealized，andthisisoftenonlythroughtheinstrumentalityofthemovementswhichaccompaniedthem，theyarebutpaleshadowsofthemselves。Emotionallytruesingingisrarerthantechnicallycorrectsinging。
　　Ifthetwoforegoingpointsofviewaretakentogether——differencesinindividualsanddifferencesincontent——anendlessnumberofdifferencescometolight。Oneindividualoverflowswithpoeticalreminiscences，anotherdirectssymphoniesfrommemory，whilenumbersandformulae，whichcometoathirdwithouteffort，slipawayfromtheothertwoasfromapolishedstone。
　　Verygreatisthedependenceofretentionandreproductionupontheintensityoftheattentionandinterestwhichwereattachedtothementalstatesthefirsttimetheywerepresent。Theburntchildshunsthefire，andthedogwhichhasbeenbeatenrunsfromthewhip，afterasinglevividexperience。Peopleinwhomweareinterestedwemayseedailyandyetnotbeabletorecallthecoloroftheirhairoroftheireyes。
　　Underordinarycircumstances，indeed，frequentrepetitionsareindispensableinordertomakepossiblethereproductionofagivencontent。Vocabularies，discourses，andpoemsofanylengthcannotbelearnedbyasinglerepetitionevenwiththegreatestconcentrationofattentiononthepartofanindividualofverygreatability。Byasufficientnumberofrepetitionstheirfinalmasteryisensured，andbyadditionallaterreproductionsgaininassuranceandeaseissecured。
　　Lefttoitselfeverymentalcontentgraduallylosesitscapacityforbeingrevived，oratleastsufferslossinthisregardundertheinfluenceoftime。Factscrammedatexaminationtimesoonvanish，iftheywerenotsufficientlygroundedbyotherstudyandlatersubjectedtoasufficientreview。Butevenathingsoearlyanddeeplyfoundedasone’smothertongueisnoticeablyimpairedifnotusedforseveralyears。Section3。DeficienciesinourKnowledgeconcerningMemoryTheforegoingsketchofourknowledgeconcerningmemorymakesnoclaimtocompleteness。Toitmightbeaddedsuchaseriesofpropositionsknowntopsychologyasthefollowing："Hewholearnsquicklyalsoforgetsquickly，"
　　"Relativelylongseriesofideasareretainedbetterthanrelativelyshortones，"Oldpeopleforgetmostquicklythethingstheylearnedlast，"andthelike。Psychologyiswonttomakethepicturerichwithanecdoteandillustration。But——andthisisthemainpoint——evenifweparticulariseourknowledgebyamostextendeduseofillustrativematerial，everythingthatwecansayretainstheindefinite，general，andcomparativecharacterofthepropositionsquotedabove。Ourinformationcomesalmostexclusivelyfromtheobservationofextremeandespeciallystrikingcases。Weareabletodescribethesequitecorrectlyinageneralwayandinvagueexpressionsofmoreorless。Wesuppose，againquitecorrectly，thatthesameinfluencesexertthemselves，althoughinalessdegree，inthecaseoftheinconspicuous，butathousand-foldmorefrequent，dailyactivitiesofmemory。Butifourcuriositycarriesusfurtherandwecravemorespecificanddependencies，boththosealreadymentionedandothers，——ifweputquestions，sotospeak，concerningtheirinnerstructure——ouranswerissilence。Howdoesthedisappearanceoftheabilitytoreproduce，forgetfulness，dependuponthelengthoftimeduringwhichnorepetitionshavetakenplace？Whatproportiondoestheincreaseincertaintyofreproductionbeartothenumberofrepetitions？
　　Howdotheserelationsvarywiththegreaterorlessintensityoftheinterestinthethingtobereproduced？Theseandsimilarquestionsnoonecananswer。
　　Thisinabilitydoesnotarisefromachanceneglectofinvestigationoftheserelations。Wecannotsaythattomorrow，orwheneverwewishtotaketime，wecaninvestigatetheseproblems。Onthecontrarythisinabilityisinherentinthenatureofthequestionsthemselves。Althoughtheconceptionsinquestion——namely，degreesofforgetfulness，ofcertaintyandinterest——arequitecorrect，wehavenomeansforestablishingsuchdegreesinourexperienceexceptattheextremes，andeventhenwecannotaccuratelylimitthoseextremes。Wefeelthereforethatwearenotatallinaconditiontoundertaketheinvestigation。Weformcertainconceptionsduringstrikingexperiences，butwecannotfindanyrealisationoftheminthesimilarbutlessstrikingexperiencesofeverydaylife。Viceversathereareprobablymanyconceptionswhichwehavenotasyetformedwhichwouldbeserviceableandindispensableforaclearunderstandingofthefacts，andtheirtheoreticalmastery。
　　Theamountofdetailedinformationwhichanindividualhasathiscommandandhistheoreticalelaborationsofthesamearemutuallydependent;theygrowinandthrougheachother。Itisbecauseoftheindefiniteandlittlespecialisedcharacterofourknowledgethatthetheoriesconcerningtheprocessesofmemory，reproduction，andassociationhavebeenuptothepresenttimeofsolittlevalueforapropercomprehensionofthoseprocesses。
　　Forexample，toexpressourideasconcerningtheirphysicalbasisweusedifferentmetaphors-storedupideas，engravedimages，well-beatenpaths。
　　Thereisonlyonethingcertainaboutthesefiguresofspeechandthatisthattheyarenotsuitable。
　　Ofcoursetheexistenceofallthesedeficiencieshasitsperfectlysufficientbasisintheextraordinarydifficultyandcomplexityofthematter。Itremainstobeprovedwhether，inspiteoftheclearestinsightintotheinadequacyofourknowledge，weshallevermakeanyactualprogress。
　　Perhapsweshallalwayshavetoberesignedtothis。Butasomewhatgreateraccessibilitythanhassofarbeenrealisedinthisfieldcannotbedeniedtoit，asIhopetoprovepresently。Ifbyanychanceawaytoadeeperpenetrationintothismattershouldpresentitself，surely，consideringthesignificanceofmemoryforallmentalphenomena，itshouldbeourwishtoenterthatpathatonce。Forattheveryworstweshouldprefertoseeresignationarisefromthefailureofearnestinvestigationsratherthanfrompersistent，helplessastonishmentinthefaceoftheirdifficulties。
　　ClassicsintheHistoryofPsychology——Ebbinghaus（1885）Chapter2Memory：AContributiontoExperimentalPsychologyHermannEbbinghaus（1885）TranslatedbyHenryA。Ruger&ClaraE。Bussenius（1913）CHAPTERIITHEPOSSIBILITYOFENLARGINGOURKNOWLEDGEOFMEMORYSection4。TheMethodofNaturalScienceThemethodofobtainingexactmeasurements——i。e。，numericallyexactones——oftheinnerstructureofcausalrelationsis，byvirtueofitsnature，ofgeneralvalidity。Thismethod，indeed，hasbeensoexclusivelyusedandsofullyworkedoutbythenaturalsciencesthat，asarule，itisdefinedassomethingpeculiartothem，asthemethodofnaturalscience。Torepeat，however，its1ogicalnaturemakesitgenerallyapplicabletoallspheresofexistenceandphenomena。Moreover，thepossibilityofdefiningaccuratelyandexactlytheactualbehaviorofanyprocesswhatever，andtherebyofgivingareliablebasisforthedirectcomprehensionofitsconnectionsdependsabovealluponthepossibilityofapplyingthismethod。
　　Weallknowofwhatthismethodconsists：anattemptismadetokeepconstantthemassofconditionswhichhaveproventhemselvescausallyconnectedwithacertainresult;oneoftheseconditionsisisolatedfromtherestandvariedinawaythatcanbenumericallydescribed;thentheaccompanyingchangeonthesideoftheeffectisascertainedbymeasurementorcomputation。
　　Twofundamentalandinsurmountabledifficulties，seem，however，toopposeatransferofthismethodtotheinvestigationofthecausalrelationsofmentaleventsingeneralandofthoseofmemoryinparticular。Inthefirstplace，howarewetokeepevenapproximatelyconstantthebewilderingmassofcausalconditionswhich，insofarastheyareofmentalnature，almostcompletelyeludeourcontrol，andwhich，moreover，aresubjecttoendlessandincessantchange？Inthesecondplace，bywhatpossiblemeansarewetomeasurenumericallythementalprocesseswhichfiltbysoquicklyandwhichonintrospectionaresohardtoanalyse？Ishallfirstdiscusstheseconddifficultyinconnection，ofcourse，withmemory，sincethatisourpresentconcern。Section5。IntroductionofNumericalMeasurementsforMemoryContentsIfweconsideroncemoretheconditionsofretentionandreproductionmentionedabove（sec。2），butnowwithregardtothepossibilityofcomputation，weshallseethatwithtwoofthem，atleast，anumericaldeterminationandanumericalvariationarepossible。Thedifferenttimeswhichelapsebetweenthefirstproductionandthereproductionofaseriesofideascanbemeasuredandtherepetitionsnecessarytomaketheseseriesreproduciblecanbecounted。Atfirstsight，however，thereseemstobenothingsimilartothisonthesideoftheeffects。Herethereisonlyonealternative，areproductioniseitherpossibleoritisnotpossible。Ittakesplaceoritdoesnottakeplace。Ofcoursewetakeforgrantedthatitmayapproach，underdifferentconditions，moreorlessneartoactualoccurrencesothatinitssubliminalexistencetheseriespossessesgradeddifferences。Butaslongaswelimitourobservationstothatwhich，eitherbychanceoratthecallofourwill，comesoutfromthisinnerrealm，allthesedifferencesareforusequallynon-existent。
　　Bysomewhatlessdependenceuponintrospectionwecan，however，byindirectmeansforcethesedifferenceintotheopen。Apoemislearnedbyheartandthennotagainrepeated。Wewillsupposethatafterahalfyearithasbeenforgotten：noeffortofrecollectionisabletocallitbackagainintoconsciousness。Atbestonlyisolatedfragmentsreturn。Supposethatthepoemisagainlearnedbyheart。Itthenbecomesevidentthat，althoughtoallappearancestotallyforgotten，itstillinacertainsenseexistsandinawaytobeeffective。Thesecondlearningrequiresnoticeablylesstimeoranoticeablysmallernumberofrepetitionsthanthefirst。Italsorequireslesstimeorrepetitionsthanwouldnowbenecessarytolearnasimilarpoemofthesamelength。Inthisdifferenceintimeandnumberofrepetitionswehaveevidentlyobtainedacertainmeasureforthatinnerenergywhichahalfyearafterthefirstlearningstilldwellsinthatorderlycomplexofideaswhichmakeupthepoem。Afterashortertimeweshouldexpecttofindthedifferencegreater;afteralongertimeweshouldexpecttofinditless。Ifthefirstcommittingtomemoryisaverycarefulandlongcontinuedone，thedifferencewillbegreaterthanifitisdesultoryandsoonabandoned。
　　Inshort，wehavewithoutdoubtinthesedifferencesnumericalexpressionsforthedifferencebetweenthesesubliminallypersistentseriesofideas，differenceswhichotherwisewewouldhavetotakeforgrantedandwouldnotbeabletodemonstratebydirectobservation。Therewithwehavegainedpossessionofsomethingthatisatleastlikethatwhichweareseekinginourattempttogetafootholdfortheapplicationofthemethodofthenaturalsciences：namely，phenomenaonthesideoftheeffectswhichareclearlyascertainable，whichvaryinaccordancewiththevariationofconditions，andwhicharecapableofnumericaldetermination。Whetherwepossessinthemcorrectmeasuresfortheseinnerdifferences，andwhetherwecanachievethroughthemcorrectconceptionsastothecausalrelationsintowhichthishiddenmentallifeenters——thesequestionscannotbeansweredapriori。Chemistryisjustaslittleabletodetermineaprioriwhetheritistheelectricalphenomena，orthethermal，orsomeotheraccompanimentoftheprocessofchemicalunion，whichgivesititscorrectmeasureoftheeffectiveforcesofchemicalaffinity。Thereisonlyonewaytodothis，andthatistoseewhetheritispossibletoobtain，onthepresuppositionofthecorrectnessofsuchanhypothesis，wellclassified，uncontradictoryresults，andcorrectanticipationsofthefuture。
　　Insteadofthesimplephenomenon——occurrenceornon-occurrenceofareproduction——whichadmitsofnonumericaldistinction，Iintendthereforetoconsiderfromtheexperimentalstandpointamorecomplicatedprocessastheeffect，andIshallobserveandmeasureitschangesastheconditionsarevaried。BythisImeantheartificialbringingaboutbyanappropriatenumberofrepetitionsofareproductionwhichwouldnotoccurofitsownaccord。
　　Butinordertorealisethisexperimentally，twoconditionsatleastmustbefulfilled。
　　Inthefirstplace，itmustbepossibletodefinewithsomecertaintythemomentwhenthegoalisreached——i。e。，whentheprocessoflearningbyheartiscompleted。Foriftheprocessoflearningbyheartissometimescarriedpastthatmomentandsometimesbrokenoffbeforeit，thenpartofthedifferencesfoundunderthevaryingcircumstanceswouldbeduetothisinequality，anditwouldbeincorrecttoattributeitsolelytoinnerdifferencesintheseriesofideas。Consequentlyamongthedifferentreproductionsof，say，apoem，occurringduringtheprocessofitsmemorisation，theexperimentermustsingleoutoneasespeciallycharacteristic，andbeabletofinditagainwithpracticalaccuracy。
　　Inthesecondplacethepresuppositionmustbeallowedthatthenumberofrepetitionsbymeansofwhich，theotherconditionsbeingunchanged，thischaracteristicreproductionisbroughtaboutwouldbeeverytimethesame。Forifthisnumber，underconditionsotherwiseequivalent，isnowthisandnowthat，thedifferencesarisingfromvariedconditionslose，ofcourse，allsignificanceforthecriticalevaluationofthosevaryingconditions。
　　Now，asfarasthefirstconditionisconcerned，itiseasilyfulfilledwhereveryouhavewhatmayproperlybecalledlearningbyheart，asinthecaseofpoems，seriesofwords，tone-sequences，andthelike。Here，ingeneral，asthenumberofrepetitionsincreases，reproductionisatfirstfragmentaryandhalting;thenitgainsincertainty;andfinallytakesplacesmoothlyandwithouterror。Thefirstreproductioninwhichthislastresultappearscannotonlybesingledoutasespeciallycharacteristic，butcanalsobepracticallyrecognised。ForconvenienceIwilldesignatethisbrieflyasthefirstpossiblereproduction。
　　Thequestionnowis：——Doesthisfulfillthesecondconditionmentionedabove？Isthenumberofrepetitionsnecessarytobringaboutthisreproductionalwaysthesame，theotherconditionsbeingequivalent？
　　However，inthisform，thequestionwillbejustlyrejectedbecauseitforcesuponus，asifitwereanevidentsupposition，therealpointinquestion，theveryheartofthematter，andadmitsofnonebutamisleadinganswer。Anyonewillbereadytoadmitwithouthesitationthatthisrelationofdependencewillbethesameifperfectequalityofexperimentalconditionismaintained。Themuchinvokedfreedomofthewill，atleast，hashardlyeverbeenmisunderstoodbyanybodysofarastocomeinhere。Butthistheoreticalconstancyisoflittlevalue：HowshallIfinditwhenthecircumstancesunderwhichIamactuallyforcedtomakemyobservationsareneverthesame？SoImustratherask：——CanIbringundermycontroltheinevitablyandeverfluctuatingcircumstancesandequalisethemtosuchanextentthattheconstancypresumablyexistentinthecausalrelationsinquestionbecomesvisibleandpalpabletome？
　　Thusthediscussionoftheonedifficultywhichopposesanexactexaminationofthecausalrelationsinthementalspherehasledusofitselftotheother（sec。4）。Anumericaldeterminationoftheinterdependentchangesofcauseandeffectappearsindeedpossibleifonlywecanrealisethenecessaryuniformityofthesignificantconditionsintherepetitionofourexperiments。Section6。ThePossibilityofMaintainingtheConstancyofConditionsRequisiteforResearchHewhoconsidersthecomplicatedprocessesofthehighermentallifeorwhoisoccupiedwiththestillmorecomplicatedphenomenaofthestateandofsocietywillingeneralbeinclinedtodenythepossibilityofkeepingconstanttheconditionsforpsychologicalexperimentation。Nothingismorefamiliartousthanthecapriciousnessofmentallifewhichbringstonoughtallforesightandcalculation。Factorswhicharetothehighestdegreedeterminativeandtothesameextentchangeable，suchasmentalvigor，interestinthesubject，concentrationofattention，changesinthecourseofthoughtwhichhavebeenbroughtaboutbysuddenfanciesandresolves——alltheseareeithernotatallunderourcontroloraresoonlytoanunsatisfactoryextent。
　　However，caremustbetakennottoascribetoomuchweighttotheseviews，correctinthemselves，whendealingwithfieldsotherthanthoseoftheprocessesbytheobservationofwhichtheseviewswereobtained。
　　Allsuchunrulyfactorsareofthegreatestimportanceforhighermentalprocesseswhichoccuronlybyanespeciallyfavorableconcurrenceofcircumstances。
　　Themorelowly，commonplace，andconstantlyoccurringprocessesarenotintheleastwithdrawnfromtheirinfluence，butwehaveitforthemostpartinourpower，whenitisamatterofconsequence，tomakethisinfluenceonlyslightlydisturbing。Sensorialperception，forexample，certainlyoccurswithgreaterorlessaccuracyaccordingtothedegreeofinterest;
　　itisconstantlygivenotherdirectionsbythechangeofexternalstimuliandbyideas。But，inspiteofthat，weareonthewholesufficientlyabletoseeahousejustwhenwewanttoseeitandtoreceivepracticallythesamepictureofittentimesinsuccessionincasenoobjectivechangehasoccurred。
　　Thereisnothingaprioriabsurdintheassumptionthatordinaryretentionandreproduction，which，accordingtogeneralagreement，isrankednexttosensorialperception，shouldalsobehavelikeitinthisrespect。Whetherthisisactuallythecaseornot，however，IsaynowasIsaidbefore，cannotbedecidedinadvance。Ourpresentknowledgeismuchtoofragmentary，toogeneral，toolargelyobtainedfromtheextraordinarytoenableustoreachadecisiononthispointbyitsaid;thatmustbereservedforexperimentsespeciallyadaptedtothatpurpose。Wemusttryinexperimentalfashiontokeepasconstantaspossiblethosecircumstanceswhoseinfluenceonretentionandreproductionisknownorsuspected，andthenascertainwhetherthatissufficient。Thematerialmustbesochosenthatdecideddifferencesofinterestare，atleasttoallappearances，excluded;equalityofattentionmaybepromotedbypreventingexternaldisturbances;suddenfanciesarenotsubjecttocontrol，but，onthewhole，theirdisturbingeffectislimitedtothemoment，andwillbeofcomparativelylittleaccountifthetimeoftheexperimentisextended，etc。
　　When，however，wehaveactuallyobtainedinsuchmannerthegreatestpossibleconstancyofconditionsattainablebyus，howarewetoknowwhetherthisissufficientforourpurpose？Whenarethecircumstances，whichwillcertainlyofferdifferencesenoughtokeenobservation，sufficient1yconstant？
　　Theanswermaybemade：——Whenuponrepetitionoftheexperimenttheresultsremainconstant。Thelatterstatementseemssimpleenoughtobeself-evident，butoncloserapproachtothematterstillanotherdifficultyisencountered。Section7。ConstantAveragesWhenshalltheresultsobtainedfromrepeatedexperimentsundercircumstancesasmuchalikeaspossiblepassforconstantorsufficientlyconstant？Isitwhenoneresulthasthesamevalueastheotheroratleastdeviatessolittlefromitthatthedifferenceinproportiontoitsownquantityandforourpurposesisofnoaccount？
　　Evidentlynot。Thatwouldbeaskingtoomuch，andisnotnecessarilyobtainedevenbythenaturalsciences。Then，perhapsitiswhentheaveragesfromlargergroupsofexperimentsexhibitthecharacteristicsmentionedabove？
　　Againevidentlynot。Thatwouldbeaskingtoolittle。For，ifobservationofprocessesthatresembleeachotherfromanypointofviewarethrowntogetherinsufficientlylargenumbers，fairlyconstantmeanvaluesarealmosteverywhereobtainedwhich，nevertheless，possesslittleornoimportanceforthepurposeswhichwehavehere。Theexactdistanceoftwosignalpoles，thepositionofastaratacertainhour，theexpansionofametalforacertainincreaseoftemperature，allthenumerouscoefficientsandotherconstantsofphysicsandchemistryaregivenusasaveragevalueswhichonlyapproximatetoahighdegreeofconstancy。Ontheotherhandthenumberofsuicidesinacertainmonth，theaveragelengthoflifeinagivenplace，thenumberofteamsandpedestriansperdayatacertainstreetcorner，andthelike，arealsonoticeablyconstant，eachbeinganaveragefromlargegroupsofobservations。Butbothkindsofnumbers，whichIshalltemporarilydenoteasconstantsofnaturalscienceandstatisticalconstants，are，aseverybodyknows，constantfromdifferentcausesandwithentirelydifferentsignificancefortheknowledgeofcausalrelations。
　　Thesedifferencescanbeformulatedasfollows：——
　　Inthecaseoftheconstantsofthenaturalscienceseachindividualeffectisproducedbyacombinationofcausesexactlyalike。Theindividualvaluescomeoutsomewhatdifferentlybecauseacertainnumberofthosecausesdonotalwaysjointhecombinationwithexactlythesamevalues（e。g。，therearelittleerrorsintheadjustmentandreadingoftheinstruments，irregularitiesinthetextureorcompositionofthematerialexaminedoremployed，etc。）。However，experienceteachesusthatthisfluctuationofseparatecausesdoesnotoccurabsolutelyirregularlybutthatasaruleitrunsthroughor，rather，triesoutlimitedandcomparativelysmallcirclesofvaluessymmetricallydistributedaroundacentralvalue。
　　Ifseveralcasesarebroughttogethertheeffectsoftheseparatedeviationsmustmoreandmorecompensateeachotherandtherebybeswallowedupinthecentralvaluearoundwhichtheyoccur。Andthefinalresultofcombiningthevalueswillbeapproximatelythesameasiftheactuallychangeablecauseshadremainedthesamenotonlyconceptuallybutalsonumerically。Thus，theaveragevalueisinthesecasestheadequatenumericalrepresentativeofaconceptuallydefiniteandwelllimitedsystemofcausalconnections;
　　ifonepartofthesystemisvaried，theaccompanyingchangesoftheaveragevalueagaingivethecorrectmeasurefortheeffectofthosedeviationsonthetotalcomplex。
　　Ontheotherhand，nomatterfromwhatpointofviewstatisticalconstantsmaybeconsidereditcannotbesaidofthemthateachseparatevaluehasresultedfromthecombinationofcauseswhichbythemselveshadfluctuatedwithintolerablynarrowlimitsandinsymmetricalfashion。Theseparateeffectsarise，rather，fromanoftimesinextricablemultiplicityofcausalcombinationsofverydifferentsorts，which，tobesure，maysharenumerousfactorswitheachother，butwhich，takenasawhole，havenoconceivablecommunityandactuallycorrespondonlyinsomeonecharacteristicoftheeffects。Thatthevalueoftheseparatefactorsmustbeverydifferentis，sotosay，selfevident。That，nevertheless，approximatelyconstantvaluesappearevenherebythecombiningoflargegroups——thisfactwemaymakeintelligiblebysayingthatinequalandtolerablylargeintervalsoftimeorextentsofspacetheseparatecausalcombinationswillberealisedwithapproximatelyequalfrequency;wedothiswithoutdoingmorethantoacknowledgeasextantapeculiarandmarvellousarrangementofnature。
　　Accordinglytheseconstantmeanvaluesrepresentnodefiniteandseparatecausalsystemsbutcombinationsofsuchwhicharebynomeansofthemselvestransparent。Thereforetheirchangesuponvariationofconditionsaffordnogenuinemeasureoftheeffectsofthesevariationsbutonlyindicationsofthem。Theyareofnodirectvalueforthesettingupofnumericallyexactrelationsofdependencebuttheyarepreparatorytothis。
　　Letusnowturnbacktothequestionraisedatthebeginningofthissection。Whenmayweconsiderthatthisequalityofconditionswhichwehavestriventorealiseexperimentallyhasbeenattained？Theanswerrunsasfollows：Whentheaveragevaluesofseveralobservationsareapproximatelyconstantandwhenatthesametimewemayassumethattheseparatecasesbelongtothesamecausalsystem，whoseelements，however，arenotlimitedtoexclusivelyconstantvalues，butmayrunthroughsmallcirclesofnumericalvaluessymmetricalaroundamiddlevalue。Section8。TheLawofErrorsOurquestion，however，isnotansweredconclusivelybythestatementjustmade。Supposewehadinsomewayfoundsatisfactorilyconstantmeanvaluesforsomepsychicalprocess，howwouldwegoaboutittolearnwhetherwemightormightnotassumeahomogeneouscausalcondition，necessaryfortheirfurtherutilisation？Thephysicalscientistgenerallyknowsbeforehandthathewillhavetodealwithasinglecausalcombination，thestatisticianknowsthathehastodealwithamassofthem，everinextricabledespiteallanalysis。Bothknowthisfromtheelementaryknowledgetheyalreadypossessofthenatureoftheprocessesbeforetheyproceedwiththemoredetailedinvestigations。Justas，amomentago，thepresentknowledgeofpsychologyappearedtoustoovagueandunreliabletobedependeduponfordecisionaboutthepossibilityofconstantexperimentalconditions;
　　sonowitmayproveinsufficienttodeterminesatisfactorilywhetherinagivencasewehavetodealwithahomogeneouscausalcombinationoramanifoldofthemwhichchancetooperatetogether。Thequestionis，therefore，whetherwemaythrowlightonthenatureofthecausationoftheresultsweobtainunderconditionsasuniformaspossiblebythehelpofsomeothercriterion。
　　Theanswermustbe：Thiscannotbedonewithabsolutecertainty，butcan，nevertheless，bedonewithgreatprobability。Thus，astarthasbeenmadefrompresuppositionsassimilaraspossibletothosebywhichphysicalconstantshavebeenobtainedandtheconsequenceswhichflowfromthemhavebeeninvestigated。Thishasbeendoneforthedistributionofthesinglevaluesabouttheresultingcentralvalueandquiteindependentlyoftheactualconcretecharacteristicsofthecauses。Repeatedcomparisonsofthesecalculatedvalueswithactualobservationshaveshownthatthesimilarityofthesuppositionsisindeedgreatenoughto1eadtoanagreementoftheresult。Theoutcomeofthesespeculationscloselyapproximatestoreality。Itconsistsinthis，——thatthegroupingofalargenumberofseparatevaluesthathavearisenfromcausesofthesamekindandwiththemodificationsrepeatedlymentioned，maybecorrectlyrepresentedbyamathematicalformula，theso-calledLawofErrors。Thisisespeciallycharacterisedbythefactthatitcontainsbutoneunknownquantity。Thisunknownquantitymeasurestherelativecompactnessofthedistributionoftheseparatevaluesaroundtheircentraltendency。Itthereforechangesaccordingtothekindofobservationandisdeterminedbycalculationfromtheseparatevalues。
　　NOTE。Forfurtherinformationconcerningthisformula，whichisnothereourconcern，Imustrefertothetext-booksonthecalculationofprobabilitiesandonthetheoryoferrors。Forreadersunfamiliarwiththelatteragraphicexplanationwillbemorecomprehensiblethanastatementanddiscussionoftheformula。Imagineacertainobservationtoberepeated1，000times。Eachobservationassuchisrepresentedbyaspaceofonesquaremillimeter，anditsnumericalvalue，orratheritsdeviationfromthecentralvalueofthewhole1，000observations，byitspositiononthehorizontallinepqoftheadjoiningFigure1。
　　Foreveryobservationwhichexactlycorrespondswiththecentralvalueonesquaremillimeterislaidoffontheverticallinemn。Foreachobservedvaluewhichdeviatesbyoneunitfromthecentralvalueupwardonesquaremillimeterislaidoffonaverticallinetorightofmnanddistantonemillimeterfromit，etc。Foreveryobservedvaluewhichdeviatesbyxunitsabove（orbelow）thecentralvalue，onesq。
　　mm。isplacedonaverticallinedistantfrommnbyxmms。，totheright（orleft，forvaluesbelowthecentralvalue）。Whenalltheobservationsarearrangedinthiswaytheoutercontourofthefiguremaybesocompactedthattheprojectingcornersoftheseparatesquaresaretransformedintoasymmetricalcurve。Ifnowtheseparatemeasuresareofsuchasortthattheircentralvaluemaybeconsideredasaconstantasconceivedbyphysicalscience，theformoftheresultingcurveisofthekindmarkedaandbinFig。1。Ifthemiddlevalueisastatisticalconstant，thecurvemayhaveanysortofaform。（Thecurvesaandbwiththelinespqincludeineachcaseanareaof1，000sq。mms。Thisisstrictlythecaseonlywithindefiniteprolongationofthecurvesandthelinespq，buttheselinesandcurvesfinallyapproacheachothersocloselythatwherethedrawingbreaksoffonlytwoorthreesq。mms。ateachendofthecurvearemissingfromthefullnumber。Whether，foracertaingroupofobservations，thecurvehasamoresteepormoreflatformdependsonthenatureofthoseobservations。
　　Themoreexacttheyare，themorewilltheypileuparoundthecentralvalue;andthemoreinfrequentthelargedeviations，thesteeperwillthecurvebeandviceversa。Fortherestthelawofformationofthecurveisalwaysthesame。Therefore，ifaperson，inthecaseofanyspecificcombinationofobservations，obtainsanymeasureofthecompactnessofdistributionoftheobservations，hecansurveythegroupingofthewholemass。Hecouldstate，forinstance，howoftenadeviationofacertainvalueoccursandhowmanydeviationsfallbetweencertainlimits。Or——
　　asIshallshowinwhatfollows——hemaystatewhatamountofvariationincludesbetweenitselfandthecentralvalueacertainpercentofalltheobservedvalues。Thelineswand-wofourfigure，forinstance，cutoutexactlythecentralhalfofthetotalspacerepresentingtheobservations。Butinthecaseofthemoreexactobservationsof1btheyareonlyonehalfasfarfrommnasin1a。Sothestatementoftheirrelativedistancesgivesalsoameasureoftheaccuracyoftheobservations。
　　Therefore，itmaybesaid：whereveragroupofeffectsmaybeconsideredashavingoriginatedeachtimefromthesamecausalcombination，whichwassubjecteachtimeonlytoso-calledaccidentaldisturbances，thenthesevaluesarrangethemselvesinaccordancewiththe"lawoferrors。"
　　However，thereverseofthispropositionisnotnecessarilytrue，namely，thatwhereveradistributionofvaluesoccursaccordingtothelawoferrorstheinferencemaybedrawnthatthiskindofcausationhasbeenatwork。
　　Whyshouldnaturenotoccasionallybeabletoproduceananalogousgroupinginamorecomplicatedway？Inrealitythisseemsonlyanextremelyrareoccurrence。Foramongallthegroupsofnumberswhichinstatisticsareusuallycondensedintomeanvaluesnotonehasasyetbeenfoundwhichoriginatedwithoutquestionfromanumberofcausalsystemsandalsoexhibitedthearrangementsummarisedbythe"lawoferrors。"[1]
　　Accordingly，thislawmaybeusedasacriterion，notanabsolutelysafeonetobesure，butstillahighlyprobableone，bymeansofwhichtojudgewhethertheapproximatelyconstantmeanvaluesthatmaybeobtainedbyanyproceedingmaybeemployedexperimentallyasgenuineconstantsofscience。TheLawofErrorsdoesnotfurnishsufficientconditionsforsuchausebutitdoesfurnishoneofthenecessaryones。Thefinalexplanationmustdependupontheoutcomeofinvestigationstotheveryfoundationsofwhichitfurnishesacertainsecurity。ThatiswhyIappliedthemeasureofferedbyittoanswerourstillunansweredquestion：Iftheconditionsarekeptasmuchalikeasispossible，istheaveragenumberofrepetitions，whichisnecessaryforlearningsimilarseriestothepointoffirstpossiblereproduction，aconstantmeanvalueinthenaturalsciencesense？AndI
　　mayanticipatebysayingthatinthecaseinvestigatedtheanswerhascomeoutintheaffirmative。Section9。Resumé
　　Twofundamentaldifficultiesariseinthewayoftheapplicationoftheso-calledNaturalScienceMethodtotheexaminationofpsychicalprocesses：
　　（1）Theconstantfluxandcapriceofmentaleventsdonotadmitoftheestablishmentofstableexperimentalconditions。
　　（2）Psychicalprocessesoffernomeansformeasurementorenumeration。
　　Inthecaseofthespecialfieldofmemory（learning，retention，reproduction）
　　theseconddifficultymaybeovercometoacertainextent。Amongtheexternalconditionsoftheseprocessessomearedirectlyaccessibletomeasurement（thetime，thenumberofrepetitions）。Theymaybeemployedingettingnumericalvaluesindirectlywherethatwouldnothavebeenpossibledirectly。
　　Wemustnotwaituntiltheseriesofideascommittedtomemoryreturntoconsciousnessofthemselves，butwemustmeetthemhalfwayandrenewthemtosuchanextentthattheymayjustbereproducedwithouterror。TheworkrequisiteforthisundercertainconditionsItakeexperimentallyasameasureoftheinfluenceoftheseconditions;thedifferencesintheworkwhichappearwithachangeofconditionsIinterpretasameasureoftheinfluenceofthatchange。
　　Whetherthefirstdifficulty，theestablishmentofstableexperimentalconditions，mayalsobeovercomesatisfactorilycannotbedecidedapriori。Experimentsmustbemadeunderconditionsasfaraspossiblethesame，toseewhethertheresults，whichwillprobablydeviatefromoneanotherwhentakenseparately，willfurnishconstantmeanvalueswhencollectedtoformlargergroups。However，takenbyitself，thisisnotsufficienttoenableustoutilisesuchnumericalresultsfortheestablishmentofnumericalrelationsofdependenceinthenaturalsciencesense。Statisticsisconcernedwithagreatmassofconstantmeanvaluesthatdonotatallarisefromthefrequentrepetitionofanideallyfrequentoccurrenceandthereforecannotfavorfurtherinsightintoit。Suchisthegreatcomplexityofourmentallifethatitisnotpossibletodenythatconstantmeanvalues，whenobtained，areofthenatureofsuchstatisticalconstants。Totestthat，Iexaminethedistributionoftheseparatenumbersrepresentedinanaveragevalue。Ifitcorrespondstothedistributionfoundeverywhereinnaturalscience，whererepeatedobservationofthesameoccurrencefurnishesdifferentseparatevalues，Isuppose——tentativelyagain——thattherepeatedlyexaminedpsychicalprocessinquestionoccurredeachtimeunderconditionssufficientlysimilarforourpurposes。Thissuppositionisnotcompulsory，butisveryprobable。Ifitiswrong，thecontinuationofexperimentationwillpresumablyteachthisbyitself：thequestionsputfromdifferentpointsofviewwillleadtocontradictoryresults。Section10。TheProbableErrorThequantitywhichmeasuresthecompactnessoftheobservedvaluesobtainedinanygivencaseandwhichmakestheformulawhichrepresentstheirdistributionadefiniteonemay，ashasalreadybeenstated，bechosendifferently。
　　Iusetheso-called"probableerror"（P。E。）——i。e。，thatdeviationaboveandbelowthemeanvaluewhichisjustasoftenexceededbytheseparatevaluesasnotreachedbythem，andwhich，therefore，betweenitspositiveandnegativelimits，includesjusthalfofalltheobservationalresultssymmetricallyarrangedaroundthemeanvalue。Asisevidentfromthedefinitionthesevaluescanbeobtainedfromtheresultsbysimpleenumeration;itisdonemoreaccuratelybyatheoreticallybasedcalculation。
　　Ifnowthiscalculationistriedouttentativelyforanygroupofobservations，agroupingofthesevaluesaccordingtothe"lawoferrors"isrecognisedbythefactthatbetweenthesub-multiplesandthemultiplesoftheempiricallycalculatedprobableerrorthereareobtainedasmanyseparatemeasuressymmetricallyarrangedaboutacentralvalueasthetheoryrequires。Accordingtothisoutof1，000observationsthereshouldbe：
　　WithinthelimitsNumberofseparatemeasures±1/10P。E。54
　　±1/6P。E。89。5
　　±1/4P。E。134
　　±1/2P。E。264
　　±[sic]P。E。500
　　±11/2P。E。688
　　±2P。E。823
　　±21/2P。E。908
　　±3P。E。957
　　±4P。E。993Ifthisconformityexistsinasufficientdegree，thenthemerestatementoftheprobableerrorsufficestocharacterisethearrangementofalltheobservedvalues，andatthesametimeitsquantitygivesaserviceablemeasureforthecompactnessofthedistributionaroundthecentralvalue-i。e。，foritsexactnessandtrustworthiness。
　　Aswehavespokenoftheprobableerroroftheseparateobservations，（P。E。o），socanwealsospeakoftheprobableerrorofthemeasuresofthecentraltendency，ormeanvalues，（P。E。m）。Thisdescribesinsimilarfashionthegroupingwhichwouldarisefortheseparatemeanvaluesiftheobservationofthesamephenomenonwererepeatedverymanytimesandeachtimeanequallygreatnumberofobservationswerecombinedintoacentralvalue。Itfurnishesabriefbutsufficientcharacterisationofthefluctuationsofthemeanvaluesresultingfromrepeatedobservations，andalongwithitameasureofthesecurityandthetrustworthinessoftheresultsalreadyfound。
　　TheP。E。misaccordinglyingeneralincludedinwhatfollows。
　　Howitisfoundbycalculation，again，cannotbeexplainedhere;sufficeitthatwhatitmeansbeclear。Ittellsus，then，that，onthebasisofthecharacterofthetotalobservationsfromwhichameanvaluehasjustbeenobtained，itmaybeexpectedwithaprobabilityof1to1[sic]thatthelattervaluedepartsfromthepresumablycorrectaveragebynotmoreatthemostthantheamountofitsprobableerror。Bythepresumablycorrectaveragewemeanthatonewhichwouldhavebeenobtainediftheobservationshadbeenindefinitelyrepeated。Alargerdeviationthanthisbecomesimprobableinthemathematicalsense——i。e。，thereisagreaterprobabilityagainstitthanforit。And，asaglanceattheaccompanyingtableshowsus，theimprobabilityoflargerdeviationsincreaseswithextremerapidityastheirsizeincreases。Theprobabilitythattheobtainedaverageshoulddeviatefromthetrueonebymorethan21/2timestheprobableerrorisonly92to908，thereforeabout1/10;theprobabilityforitsexceedingfourtimestheprobableerrorisveryslight，7to993（1to142）。
　　Footnotes[1]Thenumbersrepresentingthebirthsofboysandgirlsrespectively，asderivedfromthetotalnumberofbirths，aresaidtogroupthemselvesinveryclosecorrespondencewiththelawoferrors。
　　Butinthiscaseitisforthisveryreasonprobablethattheyarisefromahomogeneouscombinationofphysiologicalcausesaimingsotospeakatthecreationofawelldeterminedrelation。（SeeLexis，ZurTheoriederMassenerscheinungenindermenschlichenGesellschaft，p。64andelsewhere。）
　　ClassicsintheHistoryofPsychology——Ebbinghaus（1885）Chapter3Memory：AContributiontoExperimentalPsychologyHermannEbbinghaus（1885）TranslatedbyHenryA。Ruger&ClaraE。Bussenius（1913）CHAPTERIIITHEMETHODOFINVESTIGATIONSection11。SeriesofNonsenseSyllablesInordertotestpractically，althoughonlyforalimitedfield，awayofpenetratingmoredeeplyintomemoryprocesses——anditistothesethattheprecedingconsiderationshavebeendirected——Ihavehituponthefollowingmethod。
　　Outofthesimpleconsonantsofthealphabetandourelevenvowelsanddiphthongsallpossiblesyllablesofacertainsortwereconstructed，avowelsoundbeingplacedbetweentwoconsonants。[1]
　　Thesesyllables，about2，300innumber，weremixedtogetherandthendrawnoutbychanceandusedtoconstructseriesofdifferentlengths，severalofwhicheachtimeformedthematerialforatest。[2]
　　Atthebeginningafewruleswereobservedtoprevent，intheconstructionofthesyllables，tooimmediaterepetitionofsimilarsounds，butthesewerenotstrictlyadheredto。Latertheywereabandonedandthematterlefttochance。Thesyllablesusedeachtimewerecarefullylaidasidetillthewholenumberhadbeenused，thentheyweremixedtogetherandusedagain。
　　Theaimofthetestscarriedonwiththesesyllableserieswas，bymeansofrepeatedaudibleperusaloftheseparateseries，tosoimpressthemthatimmediatelyafterwardtheycouldvoluntarilybereproduced。Thisaimwasconsideredattainedwhen，theinitialsyllablebeinggiven，aseriescouldberecitedatthefirstattempt，withouthesitation，atacertainrate，andwiththeconsciousnessofbeingcorrect。Section12。AdvantagesoftheMaterialThenonsensematerial，justdescribed，offersmanyadvantages，inpartbecauseofthisverylackofmeaning。Firstofall，itisrelativelysimpleandrelativelyhomogeneous。Inthecaseofthematerialnearestathand，namelypoetryorprose，thecontentisnownarrativeinstyle，nowdescriptive，ornowreflective;itcontainsnowaphrasethatispathetic，nowonethatishumorous;itsmetaphorsaresometimesbeautiful，sometimesharsh;itsrhythmissometimessmoothandsometimesrough。Thereisthusbroughtintoplayamultiplicityofinfluenceswhichchangewithoutregularityandarethereforedisturbing。Suchareassociationswhichdarthereandthere，differentdegreesofinterest，linesofverserecalledbecauseoftheirstrikingqualityortheirbeauty，andthelike。Allthisisavoidedwithoursyllables。Amongmanythousandcombinationsthereoccurscarcelyafewdozenthathaveameaningandamongthesethereareagainonlyafewwhosemeaningwasrealisedwhiletheywerebeingmemorised。
　　However，thesimplicityandhomogeneityofthematerialmustnotbeoverestimated。Itisstillfarfromideal。Thelearningofthesyllablescallsintoplaythethreesensoryfields，sight，hearingandthemusclesenseoftheorgansofspeech。Andalthoughthepartthateachofthesesensesplaysiswelllimitedandalwayssimilarinkind，acertaincomplicationoftheresultsmuststillbeanticipatedbecauseoftheircombinedaction。
　　Again，toparticularise，thehomogeneityoftheseriesofsyllablesfallsconsiderablyshortofwhatmightbeexpectedofit。Theseseriesexhibitveryimportantandalmostincomprehensiblevariationsastotheeaseordifficultywithwhichtheyarelearned。Itevenappearsfromthispointofviewasifthedifferencesbetweensenseandnonsensematerialwerenotnearlysogreatasonewouldbeinclinedaprioritoimagine。AtleastIfoundinthecaseoflearningbyheartafewcantosfromByron’s"DonJuan"nogreaterrangeofdistributionoftheseparatenumericalmeasuresthaninthecaseofaseriesofnonsensesyllablesinthelearningofwhichanapproximatelyequaltimehadbeenspent。Intheformercasetheinnumerabledisturbinginfluencesmentionedaboveseemtohavecompensatedeachotherinproducingacertainintermediateeffect;
　　whereasinthelattercasethepredisposition，duetotheinfluenceofthemothertongue，forcertaincombinationsoflettersandsyllablesmustbeaveryheterogeneousone。
　　Moreindubitablearetheadvantagesofourmaterialintwootherrespects。
　　Inthefirstplaceitpermitsaninexhaustibleamountofnewcombinationsofquitehomogeneouscharacter，whiledifferentpoems，differentprosepiecesalwayshavesomethingincomparable。Italsomakespossibleaquantitativevariationwhichisadequateandcertain;whereastobreakoffbeforetheendortobegininthemiddleoftheverseorthesentenceleadstonewcomplicationsbecauseofvariousandunavoidabledisturbancesofthemeaning。
　　Seriesofnumbers，whichIalsotried，appearedimpracticableforthemorethoroughtests。Theirfundamentalelementsweretoosmallinnumberandthereforetooeasilyexhausted。Section13。EstablishmentoftheMostConstantExperimentalConditionsPossibleThefollowingrulesweremadefortheprocessofmemorising。
　　1。Theseparateserieswerealwaysreadthroughcompletelyfrombeginningtoend;theywerenotlearnedinseparatepartswhichwerethenjoinedtogether;neitherwereespeciallydifficultpartsdetachedandrepeatedmorefrequently。Therewasaperfectlyfreeinterchangebetweenthereadingandtheoccasionallynecessarytestsofthecapacitytoreproducebyheart。
　　Forthelattertherewasanimportantruletotheeffectthatuponhesitationtherestoftheserieswastobereadthroughtotheendbeforebeginningitagain。
　　2。Thereadingandtherecitationoftheseriestookplaceataconstantrate，thatof150strokesperminute。Aclockworkmetronomeplacedatsomedistancewasatfirstusedtoregulatetherate;butverysoonthetickingofawatchwassubstituted，thatbeingmuchsimplerandlessdisturbingtotheattention。Themechanismofescapementofmostwatchesswings300
　　timesperminute。
　　3。Sinceitispracticallyimpossibletospeakcontinuouslywithoutvariationofaccent，thefollowingmethodwasadoptedtoavoidirregularvariations：eitherthreeorfoursyllableswereunitedintoameasure，andthuseitherthe1st，4th，7th，orthe1st，5th，9th……syllableswerepronouncedwithaslightaccent。Stressingofthevoicewasotherwise，asfaraspossible，avoided。
　　4。Afterthelearningofeachseparateseriesapauseof15secondswasmade，andusedforthetabulationofresults。Thenthefollowingseriesofthesametestwasimmediatelytakenup。
　　5。Duringtheprocessoflearning，thepurposeofreachingthedesiredgoalassoonaspossiblewaskeptinmindasmuchaswasfeasible。Thus，tothelimiteddegreetowhichconsciousresolveisofinfluencehere，theattemptwasmadetokeeptheattentionconcentratedonthetiresometaskanditspurpose。Itgoeswithoutsayingthatcarewastakentokeepawayallouterdisturbancesinordertomakepossibletheattainmentofthisaim。Thesmallerdistractionscausedbycarryingonthetestinvarioussurroundingswerealsoavoidedasfarasthatcouldbedone。
　　6。Therewasnoattempttoconnectthenonsensesyllablesbytheinventionofspecialassociationsofthemnemotechnik[sic]type;learningwascarriedonsolelybytheinfluenceofthemererepetitionsuponthenaturalmemory。
　　AsIdonotpossesstheleastpracticalknowledgeofthemnemotechnicaldevices，thefulfillmentofthisconditionofferednodifficultytome。
　　7。Finallyandchiefly，carewastakenthattheobjectiveconditionsoflifeduringtheperiodofthetestsweresocontrolledastoeliminatetoogreatchangesorirregularities。Ofcourse，sincethetestsextendedovermanymonths，thiswaspossibleonlytoalimitedextent。But，evenso，theattemptwasmadetoconduct，underassimilarconditionsoflifeaspossible，thoseteststheresultsofwhichweretobedirectlycompared。