Onthebasisofthishypothesis，theexplanationofthefactthatthenumbersofthefirstgroupcomeoutsomewhattoolargeisthat，inthecourseoftheexperiment，theexistenceofasavinginworkinthecaseofthederivedserieswasanticipated，andforthisreasonthelearningoftheseriestookplaceinvoluntarilywithasomewhatgreaterconcentrationofattention。
Ontheotherhand，itmaybethat，inconsequenceoftheexcludedknowledge，therehasbeenatworkinthecaseofthenumbersofthesecondgroupadisturbingelementwhichhasmadethemsmaller。Here，tobesure，duringthelearningofthederivedseriesaverylivelycuriositydevelopedconcerningthecategoryoftransformationtowhichtheserieswhichhadjustbeenlearnedbelonged。Thatthismusthavehadadistracting，andthereforeretarding，influenceisprobablenotonlyinitselfbutalsothroughtheresultobtainedfromtheseriesderivedbypermutationofsyllables。Itwastobeexpectedthattheidentityofthesyllables，aswellasoftheinitialandendterms，wouldmakeitselffeltinthiscasebyasavingofwork，howeversmallthatsavingmightbe。Thelattereffectappears，itistrue，intheexperimentsofthefirstgroup。Withthoseofthesecondgroup，however，thereisnoticeable，insteadofthissavingofwork，aslightadditionalexpenditureoftime。This，ifitisnotmerelyaccidental，canscarcelybeexplainedotherwisethanthroughthedistractingcuriositymentioned。
Itispossiblethatbothinfluenceswereatworksimultaneouslysothatthefirstexperimentsgaveresultswhichweresomewhattoohigh;andthesecond，resultsthatweresomewhattoolow。Itisallowable，underthishypothesis，toputthetwosetsoffigurestogethersothatthecontrastingerrorsmaycompensateeachother。Inthiswaytherewasfinallyobtainedoutofthe85doubleteststhefollowingtable。Section39。DiscussionofResultsIntheforegoingtableanespecialinterest，itseemstome，isconnectedwiththelast，andalsowiththenexttothelast，rowofnumbers。Whentherewascompleteidentityofallthesyllablesandtheinitialandendtermswereleftintheirplaces，theaveragesavingoftimefor17testsdealingwiththelearningofthederivedserieswassoslightthatitwashardlytobedetermined。Itfellwithinhalfofitsprobableerror。Thesyllableswere，therefore，inthemselves，outsideoftheirconnection，sofamiliartomethattheydidnotbecomenoticeablymorefamiliarafterbeingrepeated32times。Onthecontrarywhenarelatedserieswasrepeatedthesamenumberoftimes，eachsyllablebecamesofirmlyboundtothesyllablewhichfollowed8placesbeyondthat24hourslatertheinfluenceofthisconnectioncouldbedeterminedinnodoubtfulfashion。Itattainsavalue6timestheprobableerror。Itsexistence，therefore，mustbeconsideredtobefullyprovedalthoughnaturallywecannotbesosurethatitssizeisexactlywhatitwasfoundtobeintheexperiments。Althoughitsabsolutevalueissmall，yetitsinfluenceamountstoonetenthofthatoftheconnectionwhichbindseverymembertoitsimmediatesuccessor。Itissosignificant，andatthesametimethedecreaseintheafter-effectofconnectionswhichwereformedover2，3，7interveningmembersissogradualaone，thattheassertioncanbemade，amthesegroundsalone，thateventhetermswhichstandstillfurtherfromoneanothermayhavebeenboundtoeachothersubconsciouslybythreadsofnoticeablestrengthatthetimeofthelearningoftheseries。
Iwillsummarisetheresultssofargiveninatheoreticalgeneralisation。
Asaresultoftherepetitionofthesyllable-seriescertainconnectionsareestablishedbetweeneachmemberandallthosethatfollowit。Theseconnectionsarerevealedbythefactthatthesyllable-pairssoboundtogetherarerecalledtomindmoreeasilyandwiththeovercomingoflessfrictionthansimilarpairswhichhavenotbeenpreviouslyunited。Thestrengthoftheconnection，andthereforetheamountofworkwhichiseventuallysaved，isadecreasingfunctionofthetimeorofthenumberoftheinterveningmemberswhichseparatedthesyllablesinquestionfromoneanotherintheoriginalseries。Itisamaximumforimmediatelysuccessivemembers。Theprecisecharacterofthefunctionisunknownexceptthatitdecreasesatfirstquicklyandthengraduallyveryslowlywiththeincreasingdistanceoftheterms。