Page 1
April7,2011
1
IntuitionsandExperiments
JenniferNagel,UniversityofToronto
Contemporaryepistemologistsemployvariousmethodsinthecourseofarticulatinganddefending
theirtheories.Amethodthathasattractedparticularscrutinyinrecentyearsinvolvesthe
productionofintuitiveresponsestoparticularcases:epistemologistsdescribeapersonmaking
somejudgmentandtheninvitetheiraudiencestocheckthisjudgment’sepistemicstatusfor
themselves.“DoesSknowthatP?”—Awell‐constructedcasecanelicitapowerfulintuitiveverdict.
Butthepowerofintuitiveresponsesissomewhatmysterious,forreasonstobediscussedinsome
detailinwhatfollows.Onecrucialcharacteristicofintuitionisthatitsworkingsarenotexposedto
usatthemomentofjudgment.Whenweareaskedwhetherasubjectinsomescenariohas
knowledge,ananswermaycometomindmoreorlessforcefully,withoutitsbeingimmediately
transparenttousexactlywhytheanswerhasthevalenceortheforcethatitdoes.
Thisfailureoftransparencydoubtlessmakesiteasierforus,asself‐conscious
epistemologists,towonderabouttheepistemiclegitimacyofthemethodofcases.Experimental
philosophershavesuggestedthatcloseattentiontothemechanicsofintuitivejudgmentshould
makeusuneasyaboutthemethod:intuitionsmayexhibitworrisomeinstabilityeitherwithinan
individual(Swain,Alexander,&Weinberg,2008,335),orbetweengroupsofindividuals,perhaps
alongsuchepistemicallyscaryfaultlinesasethnicity(Weinberg,Nichols,&Stich,2001)orgender
(Buckwalter&Stich,2011).IfacertainintuitiveresponsetosomeGettiercaseworksforthe
majorityofepistemologists,wehavenoguaranteethatitwillworkforothersoutsideour
professionalclub,andnogoodreason—theexperimentalistsargue—tosupposethatitsupplies
genuineevidenceaboutthenatureofknowledge.Ontheexperimentalistwayofthinking,
philosophershavetakenfalsecomfortinachievingconsensus(ornear‐consensus)amongst
themselvesonparticularcases,andneedtoberemindedofthedifferencebetweenagreeingwith
one’speersaboutsomethingandbeingrightaboutit.Accordingtoonegroupofexperimentalists,
‘‘experimentalevidenceseemstopointtotheunsuitabilityofintuitionstoserveasevidenceatall”
(Alexander&Weinberg,2007,63).Whateverexactlyisgoingonwhenphilosopherssolicit
intuitionsandusethemaspremisesintheirarguments,it’sthoughttobeepistemically
questionable,andratherdifferentinkindfromwhatgoesonwhenpeopleengageinalegitimate
cognitiveenterpriselikeempiricalscience.
Page 2
April7,2011
2
Iagreewiththeexperimentaliststhatthequestionoftheepistemicstatusofepistemic
intuitionsisanexcellentquestion,andthatwecantakesomestepstowardsansweringitbytaking
acloselookattheempiricalfactsaboutintuition.Ialsoagreethatphilosophersriskmistaking
consensusforcorrectnesswhenweighingtheevidentialvalueoftheirintuitions.Infact,according
tothebestavailablemodeloftherelationshipbetweensubjectiveconfidenceandaccuracyof
intuition,thestrengthofanintuitiongenerallycorrelateswithitsconsensualityratherthanits
correctness.However,thismodelappliesequallytodomainswhoseepistemiclegitimacythe
experimentalistswouldnotwanttodispute;inparticular,itappliestoperceptualjudgments.
Ratherthanunderminingthecasemethodinepistemology,closeattentiontothemechanicsof
intuitivejudgmentrevealssomedeepsimilaritiesbetweenepistemicintuitionandperception.1
Section1beginswithageneraldiscussionofintuitivejudgmentandthenisolatesthetype
ofintuitionthatservesasignificantdialecticalroleincontemporaryepistemology.Tobe
dialecticallyeffective,anintuitionaboutknowledgedoesnotneedtobecorrect:itjustneedstobe
sharedbyone’saudience.Itisaninterestingquestionhowskilledpractitionersofthemethodof
casescanknowinadvancewhichcaseswillresonatewiththeiraudiences:Section2showshow
thatquestionwouldbeansweredwithintheleadingmodelofconfidenceinintuitivejudgment,
Koriat’sSelf‐ConsistencyModel(SCM).Onthismodel,thestrongeranintuitionfeelstoan
individual,themorelikelyitistobesharedwithothers.IftheSCMappliestoepistemicintuitions,
variationsinthestrengthofepistemicintuitioncannotbereadatfacevalue:someintuitionsmay
bestrongandwidelysharedbutinaccurate.However,thefactthatwecandissociatethestrength
andaccuracyofepistemicintuitionsdoesnotentailthattheseintuitionsgenerallylackevidential
value.Similardissociationsarisewithinperception:wecallthemperceptualillusions.As
DescartesobservesintheFirstMeditation,thefactthatournaturalperceptualcapacitieshave
somevulnerabilitytoillusiondoesnotobligeustobegenerallyskepticalaboutthosecapacities.
Andifourepistemicintuitionsarisefromasimilarlyreliablenaturalcapacity,theycouldbe
similarlytrustworthy.
Section3arguesthatpre‐theoreticalepistemicintuitionsdoarisefromagenerallyreliable
naturalcapacity.Knowntopsychologistsas‘folkpsychology’or‘mindreading’,thisisourordinary
1Theobservationthatintuitionandperceptionhavesomeepistemicallyimportantfeaturesincommonhasbeenmadebyvariousdefendersofarmchairepistemology(e.g.Chudnoff,2010;Sosa,2007;Williamson,2007).ThesimilaritiesIexamineinthispaperhavenotpreviouslyattractedmuchattention,however.
Page 3
April7,2011
3
resourceforascribingstatesofknowledge,beliefanddesire.Experimentalistswhochallenge
epistemiccaseintuitionsdonotrejectallintuitivecapacities—andcannotdoso,onpainof
collapsingintoageneralskepticism.Atleastoneprominentexperimentalisthasexplicitly
identifiedmindreadingasthekindofintuitivecapacitythatcanbetrusted,andnotwithoutreason.
Itisentirelyplausiblethatthiscapacityislargelyreliableinitsdeliverances,notleastbecauseour
intuitivemindreadinggeneratespredictionsaboutwhatotherswilldoandsay,andthese
predictions—includingpredictionsaboutthedifferencesbetweenthinkingandknowing2—are
subjecttofeedbackandcorrectionovertime.Ourmindreadingcapacitiesareuniversal;thissection
alsopresentsevidencethatneitherethnicitynorgenderhasasignificantimpactonknowledge
ascriptioningeneral,noronepistemologicallyinterestingcasesinparticular.Inaddition,the
majorityofuntrainedsubjectsproducethestandardphilosophicalresponses,furtherevidencethat
wearedrawingonacommoncapacityinthecasemethod.
Althoughperceptionandmentalstateattributionarebothlargelyaccuratenatural
capacities,botharesubjecttocertainnaturalillusions.Wherestrengthandaccuracyofintuition
comeapartwehavevariousresourcesavailableforcorrectingourselves;Section3concludesby
notingsomesimilaritiesbetweentheresourcesavailableforself‐correctioninthedomainsof
epistemicintuitionandperceptualjudgment.ForthepurposesofthispaperIassumethatitis
epistemicallylegitimatetotakesensoryobservationsasyieldingevidenceaboutthephysicalworld;
theaimistoshowthatwehaveroughlysimilarreasonstotakeepistemicintuitionsasyielding
evidenceaboutknowledge.
1.Intuitivejudgmentingeneral,anddialecticallyusefulepistemicintuitionsinparticular
Therearemanycharacterizationsofthesplitbetweenintuitiveandnon‐intuitivejudgments,both
withinphilosophyandwithinpsychology(forreviews,seeEvans,2007;Nagel,2007;Sloman,1996;
Stanovich&West,2000).Fortunately,thereisveryconsiderablecommongroundbetweenthe
majortheoriesofintuition,andforpresentpurposesitmattersonlythatepistemicintuitionscount
asintuitiveinsomerelativelyuncontroversialsense.Itshouldbefairlyunsurprisingthatepistemic
intuitionsareintuitiveinsomemainstreampsychologicalsenseof‘intuitive’,butabriefreviewofa
2Theidentificationofknowledgeasamentalstateisuncontroversialwithinthepsychologyofmentalstateascription;thisissuewillbediscussedinmoredetailbelow.
Page 4
April7,2011
4
psychologicalaccountmayhelptomakethisevident,andhelptojustifytheapplicationof
psychologicaltheoriesofintuitiveconfidencetoepistemicintuitionsinparticular.
Bothinphilosophyandinpsychology,intuitivejudgmentsareseenincontrasttothe
judgmentsweproducethroughdeliberatereasoning.Becausethispointisemphasizedina
particularlyclearfashionbyHugoMercierandDanSperber,Ifollowtheoutlinesoftheirviewin
whatfollows,andusetheirterms‘intuitive’and‘reflective’forthetwocontrastingkindsof
judgment.MercierandSperberdescribeintuitivejudgmentsasgeneratedby‘processesthattake
placeinsideindividualswithoutbeingcontrolledbythem’(Mercier&Sperber,2009,153).The
spontaneousinferencesproducedbytheseprocessesmodifyorupdatewhatwebelieve‘without
theindividual’sattendingtowhatjustifiesthismodification’(ibid.).
Somecareisneededinhandlingtheclaimthatwedonotattendtothefactorsjustifyingour
intuitivelyproducedbeliefs.Failuretoattendisnotakindofblindnessorinsensitivity.Whenwe
readtheemotionsofothersintheirfacialexpressions—totakeanexampleofanuncontroversial
caseofintuitivejudgment—neurotypicaladultsareremarkablyaccurateatdetectinganddecoding
theminuteshiftsinbrowpositionandnostrilcontourthatdistinguishemotionssuchassurprise
andfear(Ekman&Friesen,1975).Butjudgmentsreflectthesecueswithoutourattendingtothe
cues:thecross‐culturallyrobustabilitytorecognizebasicemotionsdoesnotdependonany
personal‐levelattentiontothefacialconfigurationsandmovementsthatjustifytheseswiftintuitive
classifications(Ekman,1989;Ekman&Friesen,1975).Whatweareexplicitlyawareofatthe
personallevelistheemotionweseeexpressedinthefaceofanother;wearenottypicallyawareof
relevantsetofsubtlefacialcuesassuch,letalonethesubpersonalprocessingthatconnectsthem
withtheconceptualtemplateofthecorrespondingemotion.Sowedonotneedtohaveanexplicit
theoryofthefacialdifferencesbetweensurpriseandfearinordertodiscriminatetheseconditions
successfullyinthefacesofothers,andindeedwemayhaveconsiderabledifficultyinformulatingan
explicittheorythatreflectsouractualcompetence.Inintuitivejudgmentweare‘consciousonlyof
theresultofthecomputation,nottheprocess.’(Sloman,1996).
Inreflectivejudgment,bycontrast,weengageinexplicitreasoninganddevotepersonal‐
levelattentiontothegroundsoftheconclusionswereach.Becauseofthestrictcapacitylimitations
onconsciousattention,reflectivethinkingissequentialincharacter;whereintuitivejudgmentcan
integratelargeamountsofinformationveryrapidlyinassociativeparallelprocessing,reflective
judgmentisrestrictedbythebottleneckoflimitedworkingmemoryspace(fordetaileddiscussion,
seeEvans,2007).However,whatitlacksinspeed,reflectivejudgmentmakesupinflexibility(a
Page 5
April7,2011
5
pointparticularlyemphasizedinStanovich,2005).Intuitivejudgmentsuppliesroutineanswers,
say,automaticallyandeffortlesslyrecognizingthefaceofafriend,orproducingananswerwhen
oneisaskedtosum2+2;reflectivejudgmentcantacklenovelproblems,forexampleapplyinga
controlledsequenceofoperationstosumasetoflargenumberswehavenotpreviously
encountered.Thecontrolledsequentialcharacterofreflectiveprocesseslikecomplexarithmetical
calculationandconsciousdeductivereasoningkeepsthemopentoview:havingengagedinan
explicitprocesssuchassyllogisticreasoningorlongdivisionitisnotparticularlymysterious,from
thefirst‐personperspective,whywehavearrivedatagivenanswer.
Thecontrastbetweentheintuitiveandthereflectivedoesnotmeanthatthesetwotypesof
processingareisolatedfromoneanother.Theyareintimatelyconnectedinvariousways.Tobegin,
reflectiveprocessestaketheproductsofintuitiveprocessesasinput:inthereflectivecalculationof
anovellongdivisionproblem,forexample,theindividualstepsconsistintheapplicationof
intuitivesingle‐digitcalculation.Theintegrationbetweenthesekindsofprocessingcanalsogothe
otherway.Wecanalsohaveintuitiveresponsestotheresultsofreflectivecalculations;for
example,wemayhaveintuitivelygeneratedfeelingsofsurpriseorreliefwhenexplicitdeduction
producesanunanticipatedordesiredresult.
Complexrelationshipsbetweencontrolledandautomaticprocessingmayalsomakeitless
thanobviouswhichkindofprocessingisengagedinsometask.Becauseasenseofeffort
accompaniesthecontrolledallocationofattentioninreflectivejudgment,doingsomethinglikelong
divisionfeelshardwhilemerelyadding2+2feelseasy.Butenergycanbeapplieddifferentlyto
differentpartsofatask.AsMercierandSperberpointout,asenseofeffortcanalsoaccompanya
personal‐leveldecisiontokeepattending,overaperiodoftime,totheoutputofsomeparticular
intuitivemodule,liketheintuitivecapacityforfacerecognition(Mercier&Sperber,2009).Ittakes
efforttosustainthetaskofscanningfacesinacrowd,searchingforafriend,althoughineach
instancetheprocessingthatdecideswhetherornotagivenindividualisrecognizedisitself
subpersonalandeffortless.Itmayalsotakeefforttoscananepistemologicalscenarioforthe
presenceorabsenceofknowledge,eveniftheresultantrecognitionofknowledgeormerebeliefis
itselfprocessedintuitively.
Page 6
April7,2011
6
Thesenseinwhichepistemicintuitionsareintuitiveisadelicatematter.3Whenwereada
scenario,weareoftenawareofaseriesofconsiderationsthatcouldbearonwhetherthesubjectof
thecasehasknowledge.Wemaybeparticularlyconsciousofthiswhenreadingascenariothat
invitesustorepresentorretracethestepsofasubjectwhoisthinkingreflectivelyherself.Ihave
arguedelsewherethatourepistemicevaluationsaresensitivetowhetherthejudgmentsbeing
madebytheprotagonistofthecasewouldnaturallybemadeintuitivelyorreflectively(Nagel,
2011).Bydefault,andinroutinecircumstances,wethink(andexpectotherstothink)intuitively.
Theextraeffortofreflectivethinkingistriggered(andanticipatedtobetriggeredinothers)bysuch
factorsashighstakes,novelorunusualcasesandtheneedtonegatehypotheticalpossibilities.
Consequentlywetendtohavesomewhatdifferentexpectationsaboutaroutinejudgment,likethe
recognitionthatacertainzooanimalisazebra,ascontrastedwithamoreeffortfuljudgmentlike
thejudgmentthatthesamecreatureisnotacleverlydisguisedmule.Thereaderofascenarioneeds
toengageinreflectivecognitiontorepresentthesecondtypeofjudgment,ortograsphigh‐stakes
judgments,likethejudgmentsofapersonwhoisexceptionallyanxiousaboutwhetherthebankwill
beopentomorrow,andactivelycontemplatingpossibilitiesinvolvingchangesinthebank’shours.
However,thisisnottosaythatclassifyingtheresultsofsuchreflectivejudgmentsasknowledge(or
merebelief)woulditselfneedtobeareflectivematter.Thefactthattherearevariousdifferences
betweenintuitiveandreflectivejudgmentsdoesnotentailthatthereareanydifferencesinthe
qualitiesthatmaketheproductsofeithertypeofjudgmentcountasknowledge.
Asanexample,ifthereliabilityofanobservedsubject’s(reflectiveorintuitive)belief‐
formingprocessisaqualitythathelpstodeterminewhetherthesubjectknows,thenintuitive
knowledgeascriptioncouldbesensitivetodifferencesinperceivedreliabilitywithoutanyexplicit
personalattentiononourparttotheissueofreliabilityassuch.Evenifweneedsomemixtureof
intuitiveandreflectivecognitiontofollowthestory,itispossiblethatweuseintuitiveprocessing
acrosstheboardindeterminingwhetherthekeymentalstateinthestoryisaninstanceof
knowledgeormerebelief.3Thisquestionisfurthercomplicatedbythediversityofopiniononhowtocharacterizeintuition:therearecertainlyunderstandingsof‘intuitive’onwhichnotallepistemologicalcaseresponseswouldcountasintuitive.Forexample,thelinebetweenintuitiveandnon‐intuitivejudgmentissometimesdrawnstrictlyintermsofexecutivefunction,andthereisevidencethatmentalstateattributionmayundersomeconditionsdrawuponexecutiveresources(Apperly,Back,Samson,&France,2008;Apperly,Riggs,Simpson,Chiavarino,&Samson,2006).Theinterpretationoftheseresultsissomewhatcontroversial,however(Cohen&German,2009),asisthemoregeneralquestionofhowexactlytheintuitive/non‐intuitivedistinctioncouldbestbecharacterized(e.g.Evans,2007).Theproblemshereliebeyondthescopeofthepresentpaper;forpresentpurposes,itisenoughtoobservethatspontaneousverdictsonepistemicscenariosarenotgeneratedbythekindoffullytransparentprocessofreasoningweseeinlongdivisionortheapplicationofanexplicittheory.
Page 7
April7,2011
7
Wehaveavarietyofevidencethatouridentificationofknowledgewouldgenerallybe
intuitive.Mostnotably,wedonotneedtopossessorapplyanyexplicittheoryofknowledgein
ordertogainthesensethattheprotagonistofsomescenariohasorlacksknowledge.Wearenot
fullyconsciousofthegroundsofourjudgmentinthewaythatwewouldbeifweweremakinga
reflectivecategorizationonthebasisofanexplicittheory.Inmakingupourmindsaboutsome
particularcase,wearenottypicallyconsciousofmatchingfeaturesofthesubject’sjudgmentto
featuresofsometheoryofknowledge.Itisnotclearthatsuchatheoryisavailabletousinany
event.Eventhetheoristwhodoesadoptanexplicitworkingtheoryofknowledge—say,someone
attractedtoGoldman’searlycausaltheory(Goldman,1967)—canreadilyfindhimselfmovedto
classifynovelcasesinwaysthatdirectlyconflictwiththisworkingtheory(asinGoldman,1976).
TheresilienceoftheGettierproblemsuggeststhatitisdifficult(ifnotimpossible)todevelopany
explicitreductivetheoryofknowledgethatfullycapturesouractualpatternsofresponseto
particularexamples(Williamson,2000;Zagzebski,1994).
Furtherevidencefortheintuitivecharacterofknowledgerecognitioncouldbefoundinthe
speedandfrequencyofourreal‐timedecisionsbetweenverbsofthinkingandknowing—‘knows’
and‘thinks’arebothheavilyused,rankingat#8and#12oftheOxfordEnglishCorpuslistofour
mostcommonverbs.Judgmentsnaturallybecomeintuitivewhenmadeveryfrequently.While
criticsofphilosophicalmethodshavefocusedondeliberateassessmentsofhypotheticalscenarios,
wehavereasontobelievethattheabilitiesbroughttobearonthosecaseswouldbeequally
operativeinourveryfrequentandspontaneousassessmentsofreal‐lifesituations(Saxe,2006;
Williamson,2007).Forexample,similarbrainregionsareactivatedwhenexperimentalsubjects
readnarrativesabouttheknowledgeandbeliefsofothers(butnotabouttheirappearanceor
subjectivestatessuchashungerorthirst),andwhentheyareengagedinliveinteractionwith
others,playingagamethatrequiresthemtoattributeknowledgeorbelieftoapartner(Redcayet
al.,2010;Saxe&Wexler,2005).Justastheintuitivemechanismsenablingrecognitionofemotion
arestimulatedinsimilarwaysbylivefaces,videotapesofactualfacesoranimatedcartoons,our
resourcesforrapidlydetectingknowledgeandbeliefarethoughttobestimulatedinsimilarways
byliveinteractionsandappropriatenarrativerepresentations.4Whereattributionsofemotionare
4Presentationformatmayhoweveraffectthedegreetowhichthekeyresourcesareactivated.InarecentfMRIstudyofresponsestocloselymatchedanimatedandlive‐actionmoviesequences,RaymondMarandcollaboratorsfoundjustthesamementalstateattributionareasactivated,buttoagreaterdegreebythelive‐actionsequencesthanbytheircartooncounterparts(Mar,Kelley,Heatherton,&Macrae,2007).Activationoftheseregionswasspontaneous;participantswereinstructedtowatchthemovieclipscloselybutweregivennoinstructionstomakementalstateinferencesconcerningthecharacters(cf.Cohen&German,2009).
Page 8
April7,2011
8
triggeredbycertainpatternsoffacialfeatures,attributionsofmentalstatesaretriggeredbythe
recognitionofpatternsof‘input’toanother’sperceptualandinferentialcapacities—forexample,by
ourautomaticcalculationsofwhatanobservedagentcansee(Samson,Apperly,Braithwaite,
Andrews,&Scott,2010)—andbyrecognitionofanagent’s‘output’ofintentionalactionandspeech
(forreviews,seeApperly,2011;Goldman,2006).
Tosuggestthatourclassificationsofknowledgewouldordinarilybeintuitiveisnottosay
thatitisimpossibletodeviseasituationinwhichwewouldthinkreflectivelyinidentifyinga
judgmentasaninstanceofknowing.Forexample,reflectiveclassificationscanbemadewhenwe
areconsideringcasesatacertainlevelofabstraction.Supposewestipulatethatthereissomeset
ofsomepropertiesΓsuchthatanyjudgmentwillinstantiateknowledgeifandonlyifithasallthe
propertiesinΓ,andthenstipulatethatjudgmentJhasallthepropertiesinΓ;thesubsequent
conclusionthatjudgmentJisaninstanceofknowledgecanbemadereflectively,withfull
consciousnessofthegroundsofourcategorization.Suchformallyacceptablebutmaterially
uninformativeclassificationsarenotwhatfigureinthericherandmoreunpredictablecasemethod
thatexperimentalphilosophershavebeenattacking.
Itshouldalsobeacknowledgedthatnotallintuitiveidentificationsofknowledgeare
equallyrichandunpredictable.Itshouldbepossibletohavetheory‐drivenepistemicintuitions,for
example,afterbecomingverywell‐rehearsedinapplyingtheverdictsofsomeparticularanalysisof
knowledge.Suchclassificationsmayfailtoresonatewiththosewholackacommitmenttothat
analysis,andwouldsupplynoindependentevidenceforthetheorythatproducesthem.Asrote
exercisesintheapplicationofanexistinganalysis,theywouldlackintuition’sordinarypowerto
convinceothers,ortosurpriseusandreshapeourtheories.Someepistemicintuitionsmayindeed
betheory‐driven,butnotallintuitionsincurrentepistemologycouldhavethisstatus.TheGettier
resultthatjustifiedtruebeliefwasinsufficientforknowledgedidcomeasasurprise,andresonated
withanaudienceofphilosopherswhohadlargelybeencommittedtooneoranotherformofJTB
theory.Atleastsomedialecticallypowerfulepistemicintuitionsarepre‐theoretical.
Itisaninterestingquestionhowthepersonwhoexperiencesapre‐theoreticalintuition
couldseeitsvalue,especiallywhenitrunsagainstestablishedtheory.WhenEdmundGettierfound
himselfinclinedtojudgethatSmithdoesnotknowthatthemanwhowillgetthejobhastencoinsin
hispocket,hecouldhavedismissedthisinclinationasamistakeonhispart.Giventhebroad
acceptanceoftheJTBtheory,itmightseemGettiercouldreasonablyhaveconcludedthathewas
subjecttosomemomentaryfailureofinsightorhadsomeindividualpeculiaritydrivinghimto
Page 9
April7,2011
9
misclassifythecase.Heclaimedinsteadthathisverdictwasclearlyright,asifanticipatingthathe
himselfwouldmakethesamejudgmentaboutthiscaseonotheroccasions,andexpectinghis
audiencetodolikewise,notwithstandingtheirknownpriorinclinationstoaccepttheJTBtheory
(Gettier,1963).GettierdoesnotpresentanexplicitargumentshowingexactlywhySmith’s
judgmentdoesnotamounttoknowledge:hedoesnotofferanypositiveanalysisofknowledgeof
hisown,nordoeshespecifyanynecessaryconditionsonknowledgewhicharelackinginthiscase.
Butdespiteneitherexplainingnorperhapsevenknowingexactlywhyhefeelsthewayhedoes
aboutthiscase,Gettierseemsconvincedthathisintuitionaboutitwillbefeltbyothersandby
himselfonotheroccasions.Whatisitaboutthecharacterofintuitivejudgmentthatcouldhave
madehimfeelthatway?Thenextsectionexaminesthisquestion.
2.Confidenceandconsensualityinintuitivejudgment:Koriat’sSCM
Althoughwearenotconsciousoftheinnerworkingsofintuitionatthemomentofjudgment,weare
consciousofcertaindifferencesinitsdeliverances:someintuitionsfeelclearandstrong;othersare
weakerandmoreobscure.Thesedifferencesinconfidencehavemorethanmomentaryandprivate
significance.Whenanintuitionisstrong,itislikelytobestable,andlikelytobefeltbyothers.Toa
firstapproximation,thisisbecauseconfidenceinintuitivejudgmentisdeterminedbytheeasewith
whichonemakesthejudgment(Alter&Oppenheimer,2009;Kelley,1993;Reber&Schwarz,1999).
Ingeneral,whatisintuitivelyeasyononeoccasionislikelytobeeasyonsubsequentoccasions,and
similarlyeasyforotherpeople.Anelegantmodelexplainingthisphenomenonhasbeendeveloped
inrecentyearsbyAsherKoriat.Usingintuitiveresponsestotwo‐answerforced‐choicequestions
fromavarietyofdomains,Koriatestablishedthataperson’slevelofconfidenceinanintuitive
judgmentpredictsthedegreetowhichthatindividualwillmakethesamejudgmentagainwhen
presentedwiththesameproblemagain,and,furthermore,predictsthejudgment’sconsensuality—
theextenttowhichotherswillmakethesamejudgment(Koriat,2008;Koriat,2011).
Koriat’smodelofintuitiveconfidence,theSelf‐ConsistencyModel(SCM),tracesitsorigins
tosomecuriousfindingsofhisfromthemid‐1970s.Inthecourseofinvestigatingourabilityto
monitorourownmemoryperformance,hedecidedtoexamineanobscurememorydomain,where
subjectswouldpresumablynothavehadpreviousexperienceinmonitoringtheirtrackrecord.
Accordingtothe(nowderelict)theoryofuniversalphoneticsymbolism,therearecertainsounds
whichhaverobustcross‐culturalsignificance.Certainsound‐meaningrelationshipswere
Page 10
April7,2011
10
hypothesizedtobeencodedinthememoryofallhumanbeings.Someempiricalevidencehadbeen
offeredinsupportofthetheory:forexample,peopleperformatsignificantlybetterthanchance
whenaskedtomatchantonymsofvalue‐laden,magnitudeorsensationterms(beautiful,ugly;large,
small;hard,soft)fromtheirownlanguagewiththoseofanon‐cognatelanguagetowhichtheyhave
hadnopreviousexposure(Slobin,1968).
InKoriat’sfirststudyinthisarea(Koriat,1975),heaskedEnglish‐speakingparticipantsto
match56antonympairsinThai,KannadaandYorubatotheirEnglishtranslations,presentedin
randomorder.Aftertakingtheirbestguessasto,say,whichofthepair[tuun,luk]meant“deep”
andwhichmeant“shallow”,participantswereaskedtoratetheirconfidence—orwhatKoriat
describedinhisinstructionstothemas“theunexplainablefeelingthatyoumaybecorrect”—inthe
choicetheyhadmade.Ratingsweremadeona1‐4scale,where1wastobeusedfor“atotallywild
guess”,and4fortheanswersaboutwhichmostconfidencewasfelt.Onthematchingtask,Koriat
replicatedthepatternfoundbySlobin,with21wordpairsyieldingasignificantlybetterthan
chancematching,andonly7pairsyieldingasignificantlyworsethanchanceperformance.49out
ofhis55subjectsmatchedmorethanhalfthepairscorrectly,theremaining6closeto50‐50.The
resultsoftheconfidencetaskseemedtosuggestthatparticipantshadsomeaccesstothevalidityof
theiranswers:only53.35%ofthepairstowhichtheyhadassignedconfidencelevel1were
correctlymatched,versus66.10%ofthelevel‐4ratedpairs.Accuracyrosemonotonicallywith
confidenceratingacrosstheboard.
Inasubsequentstudy,Koriattookacloserlookattherelationshipbetweenconfidence,
accuracy,andconsensuality,ortheextenttowhichanintuitionwassharedacrossparticipants
(Koriat,1976).Drawingonearlierwork,Koriatcompiledalistof85antonympairsinsixnon‐
cognatelanguagestowhichhisparticipantsclaimednopreviousexposure;eachofthesepairswas
ultimatelyclassifiedaseitherconsensuallycorrect(CC),consensuallywrong(CW)ornonconsen‐
sual(NC).AnantonympaircountedasCCifastatisticallysignificantpercentageofsubjects(inthis
caseatleast60%)selectedthecorrecttranslation,asCWifasimilarpercentageagreedonthe
incorrectresponse,andasNCifneitherresponsewassignificantlypreferred.The1975studyhad
shownapositiveoverallcorrelationbetweenconfidenceandaccuracy,butinthatstudyCCpairs
outnumberedCWpairsthreetoone,leavingitunclearwhetherconfidencewasreallytracking
correctnessorconsensuality.Thenewstudyanalyzedtheconsensualityandreportedconfidence
foreachcategoryofitem.ConfidencestillcorrelatedpositivelywithaccuracyforCCpairs,suchas
theChinese[ching,chung],forwhich94%ofparticipantscorrectlychose[light,heavy]ratherthan
Page 11
April7,2011
11
[heavy,light].AcrossresponsestoitemsintheCCcategory,lowerconfidenceratings(1‐2)were
69.63%likelytobeaccurateandhigherconfidenceratings(3‐4)were79.55%likelytobeaccurate.
Fornonconsensualpairs,therewasnosignificantrelationshipbetweenconfidenceandaccuracy.
Meanwhile,forCWpairs,suchastheHindi[ranjida,khush],forwhichonly18%ofparticipants
correctlychose[sad,happy],confidenceandaccuracywerenegativelycorrelated:34.45%ofthe
lowconfidenceanswerswerecorrect,comparedwith24.22%ofthehighconfidenceanswers.
Strikingly,thosewhohadselectedaminorityresponse(eitherthewrongtranslationforaCCitem
orthecorrectoneforaCWitem)tendedtofeellessconfidentabouttheirchoicethanthosewho
hadchosenasthemajoritydid.Koriatconcludedthateachindividual’sconfidencewassomehow
“attunedtotheconsensualityoftheresponse,regardlessofitsaccuracy”(1976,247).
Onemightwonderhowisolatedparticipantscompletingthetaskindividuallywereableto
respondasiftrackinghowotherswereresponding;onemightalsowonderwhetherthesefindings
areparticulartotheintuitivedomainKoriatwasprobing,orwhethertheyariseasaresultofsome
generalorstructuralfeaturesofintuitivejudgment.Someprogressonthesequestionshasbeen
madeinrecentyears.Thereisevidencethatunderlyingstructuralfeaturesofintuitivejudgment
areatworkhere.Resultslikethosefoundinthephoneticsymbolismtaskhavebeenfoundfora
widevarietyofothertwo‐alternativeforced‐choicetasks,includinglong‐termmemoryfortrivia
(Koriat,1995),short‐termmemoryforsentenceswithandwithoutinterferencefromschema‐based
inference(Brewer&Sampaio,2006),socialattitudes(Koriat&Adiv,inpress),andperceptual
judgments(Koriat,2011).5Inthecourseofthiswork,twofurthercorrelationswereuncovered:a
participant’sreportedconfidencelevelinanygivenjudgmentpredictsnotonlytheextenttowhich
otherswillrespondsimilarly,butalsothelikelihoodthatthisparticularparticipantwillrepeatthat
responsewhenpresentedwiththesametwo‐alternative,forced‐choiceproblemagainlateron
(Koriat,2008;Koriat,2011).Furthermore,inastudywhichaskedsubjectstoanswerthesame50
intuitivequestionsseventimesoverthecourseofseveraldays,asubjectwhofluctuatedbetween
differentanswerstoanintuitiveprobewasmoreconfidentwhenhegavehismajorityanswer(the
answerthathegavemoreoften)thanwhenhegavehisminorityanswer(Koriat&Adiv,inpress).
TheselastcorrelationsprovidedtheguidingideaoftheSelf‐ConsistencyModel(Koriat,
2011;Koriat&Adiv,inpress).AccordingtotheSCM,subjectiveconfidenceisabyproductofthe5Suchresultsarenotrestrictedtoself‐reportofconfidence,buthavealsobeenfoundfornon‐verbalbehavior.Forexample,independentoftheiraccuracy,moreconsensualanswersattracthigherwagersinbettingtasks,evenforsubjectswhohavenotbeenaskedtogiveanyexplicitreportoftheirlevelofconfidence(Koriat,2011;Simmons&Nelson,2006).
Page 12
April7,2011
12
processofintuitivejudgment.Theparallelprocessingthatunderpinsintuitiveandperceptual
judgmentdrawsonavastrangeofinformationpotentiallyrelevanttotheproblemathand.Wedo
notconsulttheentirerangeonanygivenoccasion.Ingeneratingananswertoatwo‐alternative
forced‐choicequestion,asampleofrepresentationsisdrawnfromthepoolinorderproducea
response,whereeachrepresentationissomeconsiderationinfavorofoneortheotheranswer.
Samplingcontinuesuntileitheracriticalnumberofrepresentationsfavoringonesidehasbeen
amassed,oruntilapresetnumberofsampleshasbeendrawn.6One’sresponseisdeterminedby
thedirectionofthesample’smajority,andone’sconfidenceinthatresponseisdeterminedbythe
sizeofthesample’smajority,theinternalconsistencyofthesampleinonedirectionortheother.
Whereanalmostequalnumberofrepresentationsinthesamplespeakinfavorofeachside,one’s
confidencewillbemuchlowerthanitiswhenthesampleisunivocal.Onthismodel,neitherone’s
answernorone’sconfidencelevelwilldirectlymeasuretheextenttowhichtheunderlyingpoolof
informationsupportsananswertothequestion:itispossibletodrawanunrepresentativesample
fromapoolwhichoverallstronglyfavorsP,andrespondthatnot‐P.However,whenwehavedone
suchathing,theoddsareoverwhelmingthatoursamplewillspeakonlyweaklyinfavorofthe
unrepresentativechoice.7IfthepoolstronglyfavorsP,wearelikelytodrawaveryconsistent
samplefromit,andrespondwithhigherconfidenceforP.Ifconfidenceisdeterminedbythe
consistencyofone’ssample,thenconfidencewillpredictthelikelihoodofone’smakingthesame
choiceonsubsequentoccasions.Furthermore,ifothersaredrawingfromasimilarpoolof
representations,individualconfidencewillpredictconsensualityaswell:achoiceveryheavily
favoredbyone’sownsampleislikelytobefavoredbyothersalso.
TheSCMisageneralmodelofintuitivejudgmentfortwo‐alternativeforced‐choice
questions.Itisconceivablethatintuitivejudgmentsinepistemologyaremanagedinsomequite
differentway,althoughwehavenopositivereasontothinkso.Thequestionaskedattheendofthe
typicalcase—“DoesSknowthatp?”isatwo‐alternativeforced‐choicequestion.Theonlypublished
workbearingdirectlyontherelationshipbetweenintuitiveconfidenceandconsensualityin
6Whilethefocusofthepresentpaperisintuitivejudgment,itshouldbenotedthatsimilarsamplingmodelshavebeenproposedforreflectivejudgment.BenjaminNewellandMichaelLee,forexample,havearguedthatwealwayssampleevidencesequentiallyuntilwereachapresetthreshold(Newell&Lee,2010).Suchmodelscouldextendtheconclusionsofthissectiontothosewhoareinclinedtothinkthatepistemologicalcaseresponsesmaytosomeextentbegeneratedbysequentialorreflectivethinking.7Koriatsuppliesthefollowingexample(usingtheabbreviationpmajfortheproportionofrepresentationssupportingthemajorityanswer):‘withpmaj=.75,asampleofsevenrepresentationshasa.445likelihoodofyieldingsixorsevenrepresentationsthatfavorthemajoritychoice.Incontrast,thelikelihoodthatitwillyieldsixorsevenrepresentationsthatfavortheminoritychoiceisonly.001’(Koriat,2011,121).
Page 13
April7,2011
13
responsestoepistemicscenarioshasgeneratedresultsdirectlyinlinewiththepredictionsofthe
SCM(Wright,2010).
IftheSCMappliestointuitivejudgmentsinepistemologyasitappliestoothertypesof
intuitivejudgments,thestrengthofanepistemicintuitionisanindexofitsreproducibility.Justasa
goodexperimentisonewhichcanbereproducedbyothersorbyoneselfonotheroccasions,soalso
agoodintuition.Aslongastheunderlyingpoolofrepresentationsissimilarforoneselfandothers,
one’sconfidenceinanintuitiveresponseisamarkerofitsdialecticalvalue:astrongintuitionwill
typicallybestable,andotherswillfeelit,too.IfGettierfeltstronglythathisintuitionwouldbe
sharedbyhisaudience,hewasnotwrongaboutthat:whateverintuitivetechniquesheappliedin
categorizingthecase—orwhateversetofrepresentationshepulledupfromthepooltoidentifyhis
Smithasfailingtoknow—contemporaryepistemologistshavegenerallysharedhisexperience.
EvenepistemologistswhoexpressreluctancetorejecttheJTBtheoryonthebasisofGettier
intuitionsdonotdenytheforceofthoseintuitions(e.g.Weatherson,2003).Theapplicabilityofthe
SCMtoepistemicintuitionswouldhelptoexplainthedialecticalsuccessofappealtointuitions.If
theSCMapplies,however,thereisadistinctionbetweendialecticalvalueandaccuracy.Intuitive
confidenceisnotdirectlycorrelatedwiththetruthofone’sresponse:therearesomeintuitive
problemsforwhichone’sunderlyingpoolofrepresentationswillhavearobusttendencyto
generatethewronganswer.
Onthesideofperception,weclassifythese‘robustlyandconsensuallywrong’casesas
illusions.OurintuitiveresponsestoclassicillusionssuchastheMüller‐Lyercanbecloudedbyour
explicitknowledgeoftheirdeceptiveness.Itisusefultoconsidersomefreshcases,suchasthe
following,takenfrom(Koriat,2011):
Whichoftheselinesegmentsislonger?
Whichofthesefiguresislarger?
and the next trial began. The order of the 40 experimental pairswas determined randomly for each participant and for each block.There were short breaks between the blocks. The experiment lastedabout 45 min.
Results and Discussion
By and large, participants tended to give the same response toeach pair across the five blocks. Thus, the probability of makingthe Block-1 response again over the next four blocks averaged .76across participants.
The results were organized around four topics: (a) reproducibil-ity, (b) response consistency, (c) response consensus, and (d) theconsensuality principle. Within each topic, the results for confi-dence judgments are presented first, followed by those for choicelatency. In the final section, several analyses that connect some ofthe previously mentioned topics are presented.
Reproducibility. The assumption that confidence acts as amonitor of reliability implies that confidence in a choice predictsthe likelihood that an individual will make the same choice in asubsequent presentation of the item. To examine this possibility, Igrouped the confidence judgments in Block 1 into six categories,and calculated repetition proportion—the likelihood of making theBlock-1 response across the subsequent four blocks—across allparticipants and items. The results are presented in Figure 3A. Thefunction is monotonic; the Spearman rank-order correlation overthe six values was .94, p ! .005.2
Choice speed also predicted reproducibility. In all of the anal-yses of choice latency reported in this article, latencies that werebelow or above 2.5 SDs from each participant’s mean latency foreach block were eliminated (3.2% across all blocks). The choice
2 Other binning procedures led to similar results.
Figure 2. Examples of the stimuli used in Experiments 1 and 2, divided into those for which the consensualanswer was the correct answer (consensually correct) and those for which the consensual answer was the wronganswer (consensually wrong).
123CONFIDENCE IN PERCEPTUAL JUDGMENTS
and the next trial began. The order of the 40 experimental pairswas determined randomly for each participant and for each block.There were short breaks between the blocks. The experiment lastedabout 45 min.
Results and Discussion
By and large, participants tended to give the same response toeach pair across the five blocks. Thus, the probability of makingthe Block-1 response again over the next four blocks averaged .76across participants.
The results were organized around four topics: (a) reproducibil-ity, (b) response consistency, (c) response consensus, and (d) theconsensuality principle. Within each topic, the results for confi-dence judgments are presented first, followed by those for choicelatency. In the final section, several analyses that connect some ofthe previously mentioned topics are presented.
Reproducibility. The assumption that confidence acts as amonitor of reliability implies that confidence in a choice predictsthe likelihood that an individual will make the same choice in asubsequent presentation of the item. To examine this possibility, Igrouped the confidence judgments in Block 1 into six categories,and calculated repetition proportion—the likelihood of making theBlock-1 response across the subsequent four blocks—across allparticipants and items. The results are presented in Figure 3A. Thefunction is monotonic; the Spearman rank-order correlation overthe six values was .94, p ! .005.2
Choice speed also predicted reproducibility. In all of the anal-yses of choice latency reported in this article, latencies that werebelow or above 2.5 SDs from each participant’s mean latency foreach block were eliminated (3.2% across all blocks). The choice
2 Other binning procedures led to similar results.
Figure 2. Examples of the stimuli used in Experiments 1 and 2, divided into those for which the consensualanswer was the correct answer (consensually correct) and those for which the consensual answer was the wronganswer (consensually wrong).
123CONFIDENCE IN PERCEPTUAL JUDGMENTS
Page 14
April7,2011
14
Inbothcases,thecorrectansweris:theoneontheright.However,theconsensualanswerforboth
casesis:theoneontheleft(endorsedby84.62%ofKoriat’sparticipantsinthefirstcaseand
82.93%inthesecond).Thesepairsofstimulicouldbecomparedwiththefollowing,alsofrom
(Koriat,2011):
Whichoftheselinesegmentsislonger?
Whichofthesefiguresislarger?
Theselasttwopairsfallintotheconsensuallycorrect(CC)category:astrongmajorityofsubjects
correctlyidentifytheright‐handfigureaslargerandlonger(83.59%and89.75%respectively).
Thesubjectiveexperienceofcomparingthemagnitudesissimilarfordeceptiveandnon‐deceptive
itemsofequalconsensuality:wescanthefiguresuntiloneanswerortheotherseemsright.Itisnot
transparenttousjustwhattechniquesweareapplyinginthecourseofmakingourevaluation.But
whateverourtechniquesmightbe,theyaresufficientlysimilaracrosspersonsthattheyshowclear
trendsofworkingmoreswiftlyandeasilyinsomecasesthaninothers.Whetherornottheywere
beingjudgedcorrectly,pairsonwhichtherewasmoreconsensuswerejudgedsignificantlymore
quicklyandwithsignificantlygreaterconfidencethannon‐consensualpairs(Koriat,2011).
Susceptibilitytoacommonsetofillusionsisstrongevidenceofsystematicsimilaritiesin
ourwaysofjudgingmagnitude.Ordinaryandillusorycasesarenotdistinguishedinourimmediate
subjectiveexperience,andcommonmechanismsaretakentounderpinboth;indeed,asHermann
vonHelmholtzobservedmorethanacenturyago,perceptualillusionsare“particularlyinstructive
fordiscoveringthelawsoftheprocessesbywhichnormalperceptionoriginates”(vonHelmholtz,
1893,75).Thefactthatwearepronetocertainillusionsofthetypeillustratedaboveisofcourse
notareasonforgeneralskepticismaboutourcapacitytodistinguishmagnitudesvisually;rather,
ourgeneraltrustinthiscapacitycontinuestobewarrantedaslongasthecapacityisgenerally
and the next trial began. The order of the 40 experimental pairswas determined randomly for each participant and for each block.There were short breaks between the blocks. The experiment lastedabout 45 min.
Results and Discussion
By and large, participants tended to give the same response toeach pair across the five blocks. Thus, the probability of makingthe Block-1 response again over the next four blocks averaged .76across participants.
The results were organized around four topics: (a) reproducibil-ity, (b) response consistency, (c) response consensus, and (d) theconsensuality principle. Within each topic, the results for confi-dence judgments are presented first, followed by those for choicelatency. In the final section, several analyses that connect some ofthe previously mentioned topics are presented.
Reproducibility. The assumption that confidence acts as amonitor of reliability implies that confidence in a choice predictsthe likelihood that an individual will make the same choice in asubsequent presentation of the item. To examine this possibility, Igrouped the confidence judgments in Block 1 into six categories,and calculated repetition proportion—the likelihood of making theBlock-1 response across the subsequent four blocks—across allparticipants and items. The results are presented in Figure 3A. Thefunction is monotonic; the Spearman rank-order correlation overthe six values was .94, p ! .005.2
Choice speed also predicted reproducibility. In all of the anal-yses of choice latency reported in this article, latencies that werebelow or above 2.5 SDs from each participant’s mean latency foreach block were eliminated (3.2% across all blocks). The choice
2 Other binning procedures led to similar results.
Figure 2. Examples of the stimuli used in Experiments 1 and 2, divided into those for which the consensualanswer was the correct answer (consensually correct) and those for which the consensual answer was the wronganswer (consensually wrong).
123CONFIDENCE IN PERCEPTUAL JUDGMENTS
and the next trial began. The order of the 40 experimental pairswas determined randomly for each participant and for each block.There were short breaks between the blocks. The experiment lastedabout 45 min.
Results and Discussion
By and large, participants tended to give the same response toeach pair across the five blocks. Thus, the probability of makingthe Block-1 response again over the next four blocks averaged .76across participants.
The results were organized around four topics: (a) reproducibil-ity, (b) response consistency, (c) response consensus, and (d) theconsensuality principle. Within each topic, the results for confi-dence judgments are presented first, followed by those for choicelatency. In the final section, several analyses that connect some ofthe previously mentioned topics are presented.
Reproducibility. The assumption that confidence acts as amonitor of reliability implies that confidence in a choice predictsthe likelihood that an individual will make the same choice in asubsequent presentation of the item. To examine this possibility, Igrouped the confidence judgments in Block 1 into six categories,and calculated repetition proportion—the likelihood of making theBlock-1 response across the subsequent four blocks—across allparticipants and items. The results are presented in Figure 3A. Thefunction is monotonic; the Spearman rank-order correlation overthe six values was .94, p ! .005.2
Choice speed also predicted reproducibility. In all of the anal-yses of choice latency reported in this article, latencies that werebelow or above 2.5 SDs from each participant’s mean latency foreach block were eliminated (3.2% across all blocks). The choice
2 Other binning procedures led to similar results.
Figure 2. Examples of the stimuli used in Experiments 1 and 2, divided into those for which the consensualanswer was the correct answer (consensually correct) and those for which the consensual answer was the wronganswer (consensually wrong).
123CONFIDENCE IN PERCEPTUAL JUDGMENTS
Page 15
April7,2011
15
accurate.Illusionsarerareenoughthatthereisoverallastrongpositivecorrelationbetweenour
confidenceinourperceptualjudgmentsandtheaccuracyofthosejudgments.Evensettingaside
illusions,wearenotperfectlycalibrated—weseem8tobesomewhatunderconfidentaboutvery
easyjudgmentsandoverconfidentaboutdifficultones(e.g.Baranski&Petrusic,1994)—butin
general,theperceptualjudgmentsaboutwhichwefeelmostconfidentaresubstantiallymorelikely
toberightthanthejudgmentsaboutwhichwefeelleastconfident.Similarpatternsapplytoother
domains,suchaslong‐termmemory(Baranski&Petrusic,1995).Whereintuitivejudgmentiseasy
andswiftforsomeintuitivecapacity,wefeelmostconfident;theslowanddifficultjudgmentsthat
ariseatthebordersofanintuitivecapacitynaturallygeneratelessconfidence(Alter&
Oppenheimer,2009).Itisunsurprisingthatvariationsinsubjectiveconfidencewouldgenerallybe
adaptiveinthismanner:weakintuitiveresponsescansignalustobecautiousortosupplementour
intuitiveresponseswithreflectingthinkingintheinterestsofincreasingaccuracy(Alter,
Oppenheimer,Epley,&Eyre,2007).Naturalintuitivecapacitiesproducevariationsinsubjective
confidencethataremeaningfulandlargelyhelpful.
Incriticizingthecasemethod,JonathanWeinberghassuggestedthatintuitionis“basicallya
1‐bitsignal:isppossible,yesorno?Or:Doesthehypotheticalsituationfallundertheconceptor
not?”(Weinberg,2007,335)Heacknowledgesthatwesometimesexperiencevariationsinthe
apparentforceofintuition,butsaysthatitis“completelyunknowntowhatextentwehaveany
inter‐orevenintra‐subjectiveagreementaboutit.”(ibid.)Ihaveaimedtoshowthatifepistemic
intuitionsworklikeothertypesofintuitivejudgment,significantinter‐andintra‐subjective
agreementistobeexpected.Intuitionisasignalwhichcarriesinformationnotonlyabouttheyes‐
or‐noquestion,butalsoabouttheextenttowhichone’sownresponsewillbestableandotherswill
respondsimilarly.Thewayvariationsinsubjectiveconfidencearesharedhelpsexplainthe
dialecticalsuccessofthecasemethod.9Thefactthatphilosopherslargelyagreewitheachother
8Failuresofcalibrationarecontroversial;Gigerenzerandcollaboratorshavearguedthatanecologicallyrepresentativesamplingoftaskswouldproducesomethingclosertoperfectcalibration(Gigerenzer,Hoffrage,&Kleinbolting,1991).Cesariniandcollaboratorsarguethatthe‘hard‐easy’effectisatleastinpartastatisticalartifact,becauseerrornearthepolesoftheconfidencescalecanpushusonlytowardsthemiddle(Cesarini,Sandewall,&Johannesson,2006).9TheexperimentalistchallengetowhichIamrespondinghereadmitsthedialecticaleffectivenessofthemethodofcases.Anotherpossiblechallengetothecasemethodwoulddisputeeventhat,perhapssuggestingthatmostphilosophersdonotactuallyfeeltheGettierintuitionbutonlypretendto.Whileitistruethattherearesomecaseswhichprovokeadividedresponse—andmorewillbesaidaboutsuchcasesinSection3—thesuggestionthatcasesgenerallylackdialecticalvalueisextremelyhardtosquarewithempiricalfactsaboutthepracticeofcontemporaryepistemology.Inthedebatesbetweencontextualists,relativists,interest‐relativeandstrictinvariantists,forexample,thereisalmostuniformagreementontheintuitivepowerofthe
Page 16
April7,2011
16
aboutGettier’scasesisevidencethatwearedrawingfromsimilarpoolsofrepresentationsin
respondingtothem.However,theexperimentalistcanstillwonderwhetherepistemologists’
systematicallysharedintuitionsaregenerallyindicativeofthenatureofknowledge.Because
intuitiveconfidencecorrelateswithconsensusratherthanaccuracy,thecasemethodcouldbe
pursuedwithgreatdialecticaleffectivenesseveniftheintuitionsonwhichprofessionalsstrongly
tendedtoagreeweretypicallysharedillusionsratherthansharedinsights.Itisnaturaltowonder
whetherepistemiccaseresponsesarisefromthesortofnaturalcompetenceforwhichstronglyfelt
intuitionsaretypicallymorelikelytoberightandillusionsrelativelyrare,orwhetherthecommon
poolfromwhichepistemologistsaredrawingtheirsharedintuitivejudgmentsisanunfortunate
productoftrainingorselectioneffects.Thenextsectiontacklesthatissue.
3.Intuitivemindreading,amateurandprofessional
Howdoweattributeknowledgeormerebelieftotheprotagonistofanepistemologicalscenario?
Thedefaultanswertothisquestionis:thesamewaywegenerallyattributeknowledgeorbeliefto
anyoneelse.Unlessthereisaspecialreasontothinkthatknowledgeattributionsworkquite
differentlywhenwearereadingphilosophypapers—andI’llshortlysurveysomeevidenceagainst
thatsortofexceptionalism—weshouldexpecttofindthatepistemiccaseintuitionsaregenerated
bythenaturalcapacityresponsibleforoureverydayattributionsofstatesofknowledge,beliefand
desire.Thiscapacityhasbeengivenvariouslabels,including‘folkpsychology’,‘mindreading’,and
‘theoryofmind’.IwillfollowIanApperlyandothersincallingthiscapacity‘mindreading’,but
intendtoremainasneutralaspossibleaboutthecurrentdebateswithinpsychologyabouttheexact
natureofthiscapacity.
Althoughagreatdealofworkinmindreadinghasfocusedonnaturalillusionsofmental
stateattribution,particularlyinchildrenwhosecapacitiesarestillimmature,itisgenerallyagreed
thatadultcapacitiesformentalstateascriptionprovidefairlyreliabletrackingofwhatothersthink
andknow.Interestinglyenough,inhiscriticismofthecasemethod,Weinbergexplicitlyexempts
mindreadingfromthetroublezoneofsuspectintuitions:heiskeentoestablishthatsomeintuitions
areacceptable,inordertoshowthattheexperimentalistattackonphilosophicalintuitionsdoesnot
centralcases,evenwherethesecasesareawkwardordifficultforadvocatesofoneortheothertheoryanditmightbeverytemptingtoclaimonedidnotfeeltherelevantintuition(seee.g.DeRose,2009;Stanley,2005;Williamson,2005).
Page 17
April7,2011
17
havetocollapseintoageneralskepticism.Weinbergveryreasonablyobservesthatmindreading
intuitions“makeallsortsofpredictionsabouttheworld,”andfurthernotesthat“weseematleast
somewhatcapableoflearningfromtheoccasionalfailureofthosepredictions”(Weinberg,2007,
339).Manyresearchersinmindreadingwouldmakestrongerclaimsaboutthewaysinwhichour
capacitytoascribestatesofdesire,knowledgeandbeliefgetshapedbyfeedbackfromsuccessful
andunsuccessfulpredictions,butWeinberg’sminimalconcessionsareenoughforpresent
purposes.IfGettier’sintuitionsaboutwhatSmithdoesordoesn’tknowcomefromoureveryday
mindreadingcapacityforascribingstatesofknowledgeandbelief,andifthiscapacityisgenerally
reliable,thenourepistemiccaseintuitionshavesomepositiveclaimtoepistemiclegitimacy.
BecauseWeinberghimselfexplicitlylocatesGettierintuitionsinthetroublezoneofsuspect
intuitions(2007,335),hecan’tbeunderstandingthemasordinarymindreadingintuitions.Thisis
probablynotacapriciousmoveonhispart:hisownexperimentalresultscouldseemtomakeit
veryreasonabletotakeGettierintuitionstoarisefromsomecapacityotherthanmindreading.
Mindreadingcapacitiesaregenerallythoughttobecross‐culturallyuniversal,notleastbecauseof
developmentalsimilaritiesinmindreadinginradicallydifferentcultures(Wellman,Cross,&
Watson,2001).Naturalillusionswithinthemindreadingcapacityarealsocross‐culturally
shared.10Furthermore,mindreadingcapacitiesarethoughttoexhibitnogenderdifferencesin
typicaladults;extensiveresearchonmentalstateascriptionhasshownonlyamildadvantagefor
girlsbeforetheageoffour,perhapssimplyasafunctionofgirls’earlierlinguisticdevelopment,and
nodiscernablegenderdifferencesinthenon‐clinicalpopulationbeyondthatpoint(Charman,
Ruffman,&Clements,2002).Forthosewhosuspectthatthereiscross‐culturalvariationinGettier
caserecognition(followingWeinbergetal.,2001),andforthosewhosuspectthatGettierintuitions
maynotbefeltequallybymenandwomen(followingBuckwalter&Stich,2011),itmayseem
unlikelythatepistemologists’responsestoGettiercaseswouldbeproducedbyourcommon
mindreadingcapacity.
AnotherbarriertoallocatingGettiercaseresponsestoourcommonmindreadingcapacity
couldbethewidespreadphilosophicalopinionthatbeliefisamentalstatewhileknowledgeisnot.
10Inparticular,weshareabiasknownashindsightormorebroadlyepistemicegocentrism,abiaswhichdistortsourmindreadingofthosewhooccupyamorenaïveperspective,includingourownmorenaïvepastselves.Therehadbeensomesuggestionsthatthisbiasdiffersacrosscultures:forexample,onestudyfoundthatthebiasisworseundercertaininstructionsforWestern(Canadian)subjectsthanforEastern(Japanese)subjects(Heine&Lehman,1996),whileanotherfoundgreaterhindsightontheEastern(Korean)thanontheWestern(American)side(Choi&Nisbett,2000).Morecomprehensiveworkhasunderminedclaimsofcross‐culturaldifferencesineitherdirection(Pohl,Bender,&Lachmann,2002).
Page 18
April7,2011
18
Experimentalistswhoholdsuchaviewcouldthinkthatevenifphilosopherswouldsharewith
othersthenaturalandepistemicallyunproblematiccapacitytoattributebeliefs,perhapsthe
attributionofknowledgeisafurtheroperationdrawingonsomepoolofmentalresourcesof
dubiousvalue.Hereitisusefultoobservethatpsychologistsworkingonmentalstateascription
explicitlyclassifybothknowledgeandbeliefasmentalstates(e.g.Apperly,2011;O'Neill,Astington,
&Flavell,1992;Shatz,Wellman,&Silber,1983).Theattributionofknowledgeisnotseenasan
additionalsteptobetakenafterbeliefhasbeenascribed,andwiththehelpofauxiliaryfaculties;if
anythingitisbeliefthatisregardedasthemoredifficultandproblematicstatetoattribute(onthis
point,seeespeciallyApperly,2011).Somephilosopherssharetheviewthatknowledgeisamental
stateinitsownright,andnotacompositeofbeliefandotherfactors(e.g.McDowell,1995;
Williamson,2000).OnWilliamson’sview,forexample,theconceptofbeliefisderivativefromthe
conceptofknowledge:ratherthanourhavingtoaddextraingredientstoanascriptionofbeliefto
getanascriptionofknowledge,weinsomesensesubtractfromknowledgetoascribebelief—
indeed,‘merebeliefisakindofbotchedknowing’(Williamson,2000,47).Thepsychological
literatureonmindreadingalignsmorecloselywiththisviewthanwiththemorewidespread
philosophicalopinionaccordingtowhichknowledgeisnotconsideredamentalstate.Itmightbe
objectedthatWilliamsonhaswhatisfromthemainstreamphilosophicalperspectiveanon‐
standardunderstandingof‘mentalstate’,butevenifthatpointwereconceded,thelargerissue
wouldbeuntouched:itisuncontroversialinpsychologytoseeintuitiveattributionsofknowledge
asfallingunderthescopeofourmindreadingcapacity,soanyonewhogenerallytruststhatlatter
capacityhasreasontotrustintuitiveknowledgeascriptions.
Ifintuitiveknowledgeascriptionsforepistemologicalcasesdoarisefromourordinary
mindreadingcapacity,weneedtoexplainreportedpatternsofvariationinresponsestoGettier
casesandthelike.Areviewoftheevidenceisinorder.
Intheirinfluential2001paperonthecasemethod,Weinberg,NicholsandStichdiscussed
theresultsofpollingundergraduatesonthefollowingscenario:
Bobhasafriend,Jill,whohasdrivenaBuickformanyyears.BobthereforethinksthatJilldrives
anAmericancar.Heisnotaware,however,thatherBuickhasrecentlybeenstolen,andheis
alsonotawarethatJillhasreplaceditwithaPontiac,whichisadifferentkindofAmericancar.
DoesBobreallyknowthatJilldrivesanAmericancar,ordoesheonlybelieveit?
REALLYKNOWS ONLYBELIEVES
Page 19
April7,2011
19
Forthiscase(theonlyGettiercasetestedbyWeinberg,NicholsandStich),74%ofparticipantswho
self‐identifiedas“Western”hadthestandardknowledge‐denyingresponse;however,amongthe
EastAsiansubjectsonly43%hadthisresponse,andamongtheSouthAsiansubjectsonly39%did.
Ontwofurtherepistemologicalcases(both‘skepticalpressure’casesinwhichpossibilitiesoferror
werementionedbutstipulatedtobenon‐actual)theresponsesofWesternandEastAsiansubjects
werenotsignificantlydifferent,butSouthAsiansubjectswerestilloutofline,respondingatlevels
of70and50%whereWesternsubjectsregistered89and69%(Weinbergetal.,2001).Theseare
statisticallysignificantdifferencesinperformance.Butseveralfactorsraisequestionsabout
whethertheseresultsarebestinterpretedaspointingtounderlyingdifferencesinintuitive
competence.Onemightnoticethatifwetakethestandard“Western”responsesasnormative,the
non‐normativeresponseslieclosertothe50‐50splitthatoneseeswhensubjectsarenotinterested
inaproblemandarejustansweringrandomly.Infact,thecasesthatshownodifferencebetween
WesternandEastAsiansubjectsbothproducesimilardropstowardsrandomnessintheSouth
Asianparticipants.Itisnotclearwhy.Onepossibleexplanationherewouldinvolvedifferencesin
motivation:itispossiblethatthe24SouthAsianparticipantsinthestudywereonaveragesimply
lessengagedwiththeseparticularproblems.IfforexampleSouthAsianstudentsonthecampus
wherethisresearchwasconductedweresignificantlylesslikelytobehumanitiesmajors,11andif
participantswerethinkingoftheseriesofpuzzlesasanexerciseinapplyingphilosophicalmethods,
thentheprospectofponderingthesecasesmayhavebeenlessattractivetothem.Onemightalso
wonderwhetherthecontentoftheBobandJillscenarioresonatedmorewithmembersofsome
ethnicgroupsthanothers,orwhethersomegroupsfounditeasiertoread.12
ForamorereliablemeasureofvariationinresponsestoGettiercases,itisusefultoexamine
responsestomultiplecases.Inanewstudyinvolving222undergraduateparticipantsfroma
varietyofethnicbackgrounds(71participantsself‐indentifiedasWhite;58asSouthAsian,28as
EastAsian13,16LatinAmerican,15Black,6WestAsian,4Araband24other),eightdifferentGettier
11IdonothavedataforRutgersUniversityinparticular,butintheUnitedStates,therearestatisticallysignificantdifferencesinthecollegemajorsofdifferentethnicgroups(e.g.comparedtoWhitestudents,AsianstudentswereatthetimeofWeinberg,NicholsandStich’sstudymorethantwiceaslikelytomajorinbiologyorengineering).Source:NationalCenterforEducationStatistics,StatusandTrendsintheEducationofRacialandEthnicMinorities,table25.2,Percentageofdegreesconferredbydegree‐grantinginstitutionsinmostpopularfieldsofstudy,byrace/ethnicityandlevelofstudy:2003‐04.12Forexample,thestorywillbemoreeasilyunderstoodbyreadersforwhomitisanautomaticinferencethataBuickisanAmericancar.ThankstoMeredythDanemanforthisobservation.13ParticipantsinourEastAsiangroupactuallyidentifiedthemselvesasChinese,Japanese,Korean,FilipinoorSoutheastAsian(Cambodian,Vietnamese,etc.)followingthestandardCanadiancensuscategories.
Page 20
April7,2011
20
casesweretested,togetherwitheightordinaryknowledgecases,eightjustifiedfalsebelief(JFB)
cases,eightskepticalpressurecasesandeightfillerquestionsinvolvingvarioustypesofjustified
andunjustifiedbelief(Nagel,SanJuan,&Mar,inprep).The32experimentalitemswere‘matched’:
foreachGettiercase,therewasacorrespondingknowledgecase,skepticalpressurecaseandJFB
casewithanequalwordcount.Theseitemsweredistributedinabetween‐subjectsdesign,sothat
eachparticipantsawjustonememberofeachsetofmatcheditems(forexample,oneofA‐D
below);eachparticipantjudgedtwoexperimentalitemsofeachkind.
(A). [Ordinaryknowledge:]Wandaisoutforaweekendafternoonwalk.Shelivesinalargenewcondominiumtowerdowntown,andhersuiteisfairlysmallanddoesnothaveanywindowsthatopen,soshereallylikestogetoutforsomefreshair.Passingnearthetrainstation,Wandawonderswhattimeitis.Sheglancesupattheclockonthetrainstationwallandseesthatitsays4:15pm.Itisinfact4:15pmatthatmoment.
(B). [Skepticalpressurevariant:]Wandaisoutforaweekendafternoonwalknearthetrainstationandwonderswhattimeitis.Sheglancesupattheclockonthetrainstationwallandseesthatitsays4:15pm.Itisinfact4:15pmatthatmoment.Thestationclockisinfactworking,butithasnosecondhand,andWandaonlylooksatitforamoment,soshewouldnotbeabletotelliftheclockwerestopped.
(C). [Gettiercasevariant:]Wandaisoutforaweekendafternoonwalk.Asshepassesnearthetrainstation,shewonderswhattimeitis.Sheglancesupattheclockonthetrainstationwallandseesthatitsays4:15pm.Whatshedoesn’trealizeisthatthisclockisbrokenandhasbeenshowing4:15pmforthelasttwodays.Butbysheercoincidence,itisinfact4:15pmjustatthemomentwhensheglancesattheclock.
(D). [JFBvariant:]Wandaisoutforaweekendafternoonwalk.Shelivesinasmallcondodowntown,butenjoystheoutdoors.Asshepassesnearthetrainstation,shewonderswhattimeitis.Sheglancesupattheclockonthetrainstationwallandseesthatitsays4:15pm.Itisinfact4:53pmatthatmoment.Whatshedoesn’trealizeisthatthisclockisbrokenandhasbeenshowing4:15pmforthelasttwodays.
Aseachscenariowaspresentedonscreen,participantswereaskedtoverifythekeypropositionin
thestory(e.g.“Accordingtothestory,whattimeisitwhenWandalooksattheclock?”).
ParticipantswhoansweredthisquestioncorrectlywerethenaskedaboutWanda’sstateofmind
(thescenarioremainedonscreenthroughout).Unsurprisingly,participantsweremostlikelyto
ascribeknowledgeinthetype‐Abaselinestories(overall,72.0%agreedthatthesubjectknewthe
keypropositioninsuchcases),andleastlikelytoascribeknowledgeinthetype‐Dfalsebelief
stories(15.8%).Gettiercasesandskepticalpressurecaseswereseenbymostparticipantsas
involvingafailuretoknow,attractingknowledgeascriptionratesof32.9and39.8%respectively.
NoteveryoneproducedthestandardresponsetoeveryGettiercase,butdifferencesbetween
standardandnon‐standardresponseswerenotafunctionofethnicityorgender.Wedidnotfind
Page 21
April7,2011
21
statisticallysignificantcorrelationsbetweenethnicityorgenderandknowledgeascriptionforany
ofthefourkeytypesofscenario.14
Ifundergraduateswithlittleornophilosophicaltraining15generallyclassifyepistemology
casesthewayepistemologistswould,andifthereisnoappreciablegenderorethnicvariationin
caseresponses,thenthereisnogoodreasontodoubtthedefaulthypothesisthatwearerelyingon
acommonmindreadingcapacityinrespondingtothesecases.IfIamrightthatourknowledge‐
denyingintuitiveresponsestoskepticalpressurecasesarenaturalillusionsofthemindreading
capacity(Nagel,2010),thentheevidencethatwearerespondingtoepistemiccasesonasimilar
basisisstrongerstill:theuntrainedsharenotonlyourordinaryresponsesbutalsoshareatleast
oneimportantillusion.
Criticsofthecasemethodmayhoweverworrythatthesignalfromthelayresponsesisnot
asstrongasitshouldbe.IfepistemologistshaveclearintuitionsthatGettiercasesinvolveafailure
toknow,thenonemighthaveexpectedahigherlevelofconsensualityinamateurjudgmentsof
thosecases.Ouramateursgenerallyagreedwiththestandardprofessionallineontheeightcore
Gettiercaseswetested,butataperhapsdishearteningaveragerateof67.1%(tosaynothingofthe
15.8%ofparticipantswillingtoascribeknowledgeinJFBcases).
Therearesomemundanereasonswhyonemightfindaworsesignal‐to‐noiseratioinlay
judgmentsofepistemologicalscenarios.Whileintuitivejudgmentitselfiseffortless,effortis
requiredtoreadthestoriescloselyenoughtoregistertherelevantdetailsabouttheinputstothe
subject’sjudgment(cf.Williamson,2011).Participantswhoarenotinterestedinaparticularstory
maybemoreinclinedtorespondtoitrandomly.16Philosophersandothersmayhavethesame
14ForGettiercasesinparticular,thecorrelationbetweenethnicityandknowledgeascriptionwas0.018,p=.791(two‐tailed),n=222;thecorrelationbetweengenderandknowledgeascriptionwas0.019,p=.774(two‐tailed),n=222.Thep‐valueorsignificancelevelindicatestheprobabilityoftheobservedpatternsontheassumptionthatthenullhypothesisistrue(i.e.thatethnicity/genderdoesnotinfluenceknowledgeascription).Correlationsarenotregardedassignificantinpsychologyifthesignificancelevelisgreaterthan.05.Wrightalsofoundnogenderdifferences;shedoesnotreportonethnicity(Wright,2010).1558outofour222participants(26%)reportedhavingtakenatleastonephilosophycourse;ofthese39reportedhavingtakenonlyonesuchcourse.Knowledgeascriptionrateswerenotcorrelatedwiththenumberofphilosophycoursestaken,aresultconsistentwithJenniferColeWright’sstudiesofresponsestocaseslikeLehrer’sTruetempandtheFakeBarnGettiercase(Wright,2010).16Oneofthefillerquestionscompletedbyall222participantswasaverbatimcopyoftheJill/BobPontiacandBuickcasefrom(Weinbergetal.,2001).WedidnotreproducetheoriginalfindingthatSouthAsianstendedtogivethenon‐standardresponse:only40%ofourSouthAsianparticipantsascribedknowledgeinthiscase(vs.61%ofSouthAsiansinWeinberg,NicholsandStich’soriginalstudy).WedidhoweverseeaweakerresponsetothisparticularcaseamongSouthAsiansthanamongself‐identifiedWhites:only14%ofourWhiteparticipantsascribedknowledgeinthiscase(vs.26%of‘Western’participantsintheoriginal
Page 22
April7,2011
22
basicintuitivecapacitytoregisterthepresenceorabsenceofknowledge,butphilosophersmaybe
moremotivatedtoreadepistemicvignetteswithaneyetoexercisingthiscapacity.Ifourcapacity
torecognizeknowledgeisintuitive,wedonotnecessarilyhavetodevotepersonal‐levelattention
tothefactorsjustifyingourjudgments:forexample,evenifreliabilityisanecessaryconditionof
knowing,wewouldnotneedtomakeexplicitpersonal‐leveljudgmentsofreliabilityinordertosee
someoneasknowing.However,someinterestintherelevantpartsofthestorymaybeneededin
ordertoactivatetherelevantspontaneousprocessing:participantswhoskimorreadinattentively
mayretainonlyaroughsenseofthegistofthestorywhentheyreachthequestionconcerning
mentalstate.Gettiercasescanbetaxingtofollow:amateurparticipantswhodonothavethe
motivationoftestinganepistemologicaltheorymaybelessinclinedtoreadclosely.17
Motivationmayalsoplaysomeroleintheothertypeofvariationthathasbeenconsidered
problematicforthecasemethod.Experimentalistshaveshownthatresponsestophilosophical
scenariosexhibitorderorcontrasteffects:forexample,naïvesubjectsaresignificantlymorelikely
toattributeknowledgetothesubjectofaversionofKeithLehrer’sTruetempcasewhenitis
presentedfollowingaveryclearcaseofignorancethanwhenitfollowsaveryclearcaseof
knowledge(Swainetal.,2008;Wright,2010).Intendedasacounterexampletoreliabilism,this
caseinvolvesapersonwhosebrainhasbeensurreptitiouslychangedtoproduceaccuratebeliefs
aboutthetemperaturewithouthisbeingawareofthechange;manyepistemologistshavethe
intuitionthatthesereliablytruebeliefsdonotconstituteknowledge.Observingthatthecasemay
ormaynotappeartobeacaseofknowledgedependingonitscontext,Swainandcollaborators
study).BecausewedidnotseeanysignificantcorrelationbetweenethnicityandknowledgeascriptionacrosstheeightotherGettiercasestested,itseemsmorereasonabletoattributethedifferenceinresponseratesheretosomethinglikedifferinglevelsofinterestorengagementwiththisparticularstoryasopposedtodifferencesinknowledgeascriptionforGettiercasesperse.Idonotknowenoughaboutthetestingconditionsoftheoriginalstudytospeculateaboutwhywemighthaveseenaclearersignalfrombothgroupsinourstudy.17Contingentdifferencesinmotivation,asopposedtounderlyingcognitivedifferences,havebeenpositedasexplainingmanyofthefindingsoncognitive‘cross‐culturaldifferences’thatservedasthetheoreticalbasisofWeinberg,NicholsandStich’sclaimofdiversityinepistemicintuitions.RichardNisbettandcollaboratorsmaintainthattherearefundamentalqualitativedifferencesinthereasoningofEasternandWesternpeoples,ascribingan“intuitive”,“holistic”and“experience‐based”wayofthinkingtothoseontheEasternside,andan“analytic”,“rule‐based”and“decontextualized”modeofthoughttotheWesterners(Nisbett,Peng,Choi,&Norenzayan,2001).HugoMercierobservesthatthesegeneralizationsdonotsquarewellwiththehistoricalrecordofChineseworkonlogic.HearguesthatNisbett’sEasternsubjectshavegivenlessanalytic—andmoreshallow—responsesbecausethesesubjectpoolshappenedtobelessinterestedinNisbett’stasks(Mercier,forthcoming).Thereisevidencethatdifferencesarenotfoundwhencareistakentoensurethatmotivationisuniform:forexample,evenjustimaginingthatonehasastakeinsomeissueundercontentioncansufficetoeraseculturaldifferencesinreasoningbetweenEastern(Japanese)andWestern(French)subjects(VanderHenst,Mercier,Yama,Kawasaki,&Adachi,2006).
Page 23
April7,2011
23
contendthatvariationofthiskindunderminesthecasemethod:theyareconcernedthatsensitivity
towhatonehasrecentlybeenthinkingrendersintuitionsdisturbinglyopentomanipulationand
onlyquestionablyconnectedtothesubjectmatterofinterest.
Thepossibilityofcontextualinfluencesonsomecapacitydoesnotonitsownshowthatthis
capacitycannotbeasourceofevidence.Forexample,onefindscontrasteffectsinvisualjudgments
ofcolor:
Swainetal.explicitlyconsidertheobjectionthatthetypesofvariabilitytheyhaveidentifiedare
alsofoundinperception.Theycontendthereisarelevantdisanalogy:“Weareawareofthegreat
majorityofthecircumstancesunderwhichperceptualjudgmentsarelikelytobeunreliable.For
instance,weknowthatvisualperceptionrequiresacertainamountofillumination,andvisual
perceptionitselfprovidesuswithknowledgeofwhetherenoughilluminationispresent”(Swainet
al.,2008,148).Swainetal.observethatwenaturallyhavelowerconfidenceinperceptual
judgmentsinconditionsoflowillumination;thedefenderofthecasemethodmightobservethatwe
naturallyhavelowerconfidenceinborderlineintuitivejudgmentsaswell,includingtheversionof
theTruetempcasethatSwainetal.aretesting(Wright,2010).Swainetal.alsosuggestthat
correctingforunreliabilitywillbeeasierintheperceptualcase:whilewecancounterdim
illuminationinobviousways,“wedon’tknowwhatistheparallelforintuitionofmakingsurethat
thelightison;thatis,wedonotknowwhicharethecircumstancesthatrenderintuitionreliableor
unreliable”(ibid.).
Itisdoubtlesseasyenoughtokeepthelighton,butitisnotclearthatlowilluminationisa
goodparallelfortheordereffectsidentifiedfortheTruetempcase.Acloseranalogyfororder
effectsfromthedomainofvisualperceptionwouldbeordereffectsincolorjudgments:
argument that anticipated the psychophysics of Weber andFechner by more than a century, Bernoulli concluded thatthe utility function of wealth is logarithmic. Economistsdiscarded the logarithmic function long ago, but the ideathat decision makers evaluate outcomes by the utility ofwealth positions has been retained in economic analyses foralmost 300 years. This is rather remarkable because theidea is easily shown to be wrong; I call it Bernoulli’s error.
Bernoulli’s (1738/1954) model of utility is flawedbecause it is reference independent: It assumes that theutility that is assigned to a given state of wealth does notvary with the decision maker’s initial state of wealth. Thisassumption flies against a basic principle of perception,where the effective stimulus is not the new level of stim-ulation but the difference between it and the existing ad-aptation level. The analogy to perception suggests that thecarriers of utility are likely to be gains and losses ratherthan states of wealth, and this suggestion is amply sup-ported by the evidence of both experimental and observa-tional studies of choice (see Kahneman & Tversky, 2000).The present discussion relies on two thought experimentsof the kind that Tversky and I devised in the process ofdeveloping the model of risky choice that we called pros-pect theory (Kahneman & Tversky, 1979).
Problem 2
Would you accept this gamble?
50% chance to win $150
50% chance to lose $100
Would your choice change if your overall wealth were lower by$100?
There will be few takers of the gamble in Problem 2. Theexperimental evidence shows that most people reject agamble with even chances to win and lose unless thepossible win is at least twice the size of the possible loss
(see, e.g., Tversky & Kahneman, 1992). The answer to thesecond question is, of course, negative.
Next, consider Problem 3.
Problem 3
Which would you choose?
Lose $100 with certainty
or
50% chance to win $50
50% chance to lose $200
Would your choice change if your overall wealth were higher by$100?
In Problem 3, the gamble appears much more attractivethan the sure loss. Experimental results indicate that risk-seeking preferences are held by a large majority of respon-dents in choices of this kind (Kahneman & Tversky, 1979).Here again, the idea that a change of $100 in total wealthwould affect preferences cannot be taken seriously.
Problems 2 and 3 evoke sharply different preferences,but from a Bernoullian perspective, the difference is aframing effect: When stated in terms of final wealth, theproblems only differ in that all values are lower by $100 inProblem 3—surely, an inconsequential variation. Tverskyand I examined many choice pairs of this type early in ourexplorations of risky choice and concluded that the abrupttransition from risk aversion to risk seeking could notplausibly be explained by a utility function for wealth.Preferences appeared to be determined by attitudes to gainsand losses, defined relative to a reference point, but Ber-noulli’s (1738/1954) theory and its successors did not in-corporate a reference point. We therefore proposed analternative theory of risk in which the carriers of utility aregains and losses—changes of wealth rather than states ofwealth. Prospect theory (Kahneman & Tversky, 1979) em-
Figure 5Simultaneous Contrast and Reference Dependence
704 September 2003 ! American Psychologist
Page 24
April7,2011
24
identificationofcolorsissensitivetowhatothercolorswehavebeenseeing,sothatwewill
describethesamecolorpatchasgreenifweapproachitfromoneendofthespectrumandasblueif
weapproachitfromtheother(Kalmus,1979).18Thesekindsofcontextualeffectsdonotannounce
themselvesasobviouslyaslowilluminationdoes,anditisnotatrivialmattertofigureouthowwe
shouldassessthem,whethersucheffectsshouldbeseenasoccasioningerrorsinourapplicationof
colorconcepts,orasshowingthattheboundariesofourconceptsaremorevariablethanwemight
havesupposed(Raffman,2005).Naïveperceptiononitsowndoesnottelluswhattodohere.
Meanwhile,correctingforordereffectsinthecasemethodmightnotbeinsuperably
difficult.SimonCullenhasarguedthattheordereffectsfoundbySwainetal.arisefromadherence
toconversationalnormsinthesurveycontext:participantsnoticedtheobviouscontrastbetween
theTruetempandextremecasesofknowledgeandignorance,andunderstoodthequestionabout
Truetemp’sstateofmindasaninvitationtocomparehimwiththesalientcontrastingcase.Onthis
view,participantsgivenadifferentorderofcaseswouldhaveadifferentconceptionoftheirtask.
CullentestedthepairsofcasesthatgeneratedthestrongestordereffectsforSwainetal.,butthis
timewithinstructionsurgingparticipantsto“considereachindependently.”Clarifyingthe
pragmaticsinthismannerwasenoughtocanceltheordereffects:withthenewinstructionsno
significantvariationarosefrompresentingtheTruetempcaseafteraclearcaseofknowledgeas
opposedtoaclearcaseofignorance(Cullen,2010).Ifepistemologistsaregenerallymakingan
efforttoconsidertheircasesindependently,theSwainresultsshouldnotbecauseforalarm.
TheTruetempstudiesmayraiseworriesofanotherkind,however.AsSwainetal.observe,
mostepistemologiststendtohavetheintuitionthatTruetempdoesnotknowthetemperature:
evenreliabilistslikeAlvinGoldmanseethecaseasprimafacieintuitiveevidenceagainstreliabilism
(Goldman,1994).BothCullenandSwainfoundconsiderableambivalenceaboutthiscaseamong
amateurs,however,withmeanresponsesofabout2.5to3.3ona1‐5scalewhere3wasneutraland
4or5wouldindicateadenialorstrongdenialofknowledge;Wright’sresultsweremorenegative
butalsofailedtoshowstrongconsensusonthecase.Onemightworrythatdifferencesbetween
amateursandprofessionalsherepointtotheinvolvementofsomecapacityotherthantheuniversal
mindreadingcapacityindecidingwhetherTruetemphasknowledge.
Itispossiblethatgenericmotivationproblemsplayedarolehere:perhapsthecaseis
confusingenoughthatamateurparticipantsgiveuponfollowingwhatishappeningandrespond
nearthemidpointsimplytoexpressuncertaintyaboutthetask(cf.DeBruin,Fischhoff,Millstein,&
18ThankstoDianaRaffmanfortheexampleanddiscussionofthepoint.
Page 25
April7,2011
25
Halpern‐Felsher,2000).Anotherpossibilityisthatthecaseissomewhatunder‐described,and
amateurparticipantsarenotallfleshingoutthedetailsthesameway(aseachother,oras
professionalswould).HereistheversionofthecaseusedbySwain,CullenandWright:
OnedayCharleswasknockedoutbyafallingrock;asaresulthisbrainwas‘‘rewired’’sothatheisalwaysrightwheneverheestimatesthetemperaturewhereheis.Charlesisunawarethathisbrainhasbeenalteredinthisway.Afewweekslater,thisbrainrewiringleadshimtobelievethatitis71degreesinhisroom.Apartfromhisestimation,hehasnootherreasonstothinkthatitis71degrees.Infact,itis71degrees.
Justgiventhistext,therearevariouswaystoconstrueCharles’spredicament.Onemightwonder,
inparticular,whetherheisinanywayself‐consciousaboutthepeculiarityofhisobviouslypeculiar
newtendencytoformprecisebeliefsaboutthetemperature.KeithLehrer’soriginalpresentation
oftheexample—involvinga‘tempucomp’implantedbyabrainsurgeon—providedmoredetailon
thisscore.InLehrer’sversion,Truetempis“slightlypuzzledaboutwhyhethinkssoobsessively
aboutthetemperature”.LehrergoesontostipulatethatTruetemp“neverchecksathermometerto
determinewhetherthesethoughtsaboutthetemperaturearecorrect.Heacceptsthem
unreflectively,anothereffectofthetempucomp”(Lehrer,1990,163).Oneoftheinteresting
featuresofLehrer’sversionisthatthetempucompisdescribedashavingtwodistincteffects:first,
theproductionofaccuratetemperaturethoughtsinTruetemp’smind,andsecond,theproduction
ofTruetemp’spersonal‐levelbutsomehowunreflectiveacceptanceofthesethoughts.19Whenwe
arereconstructingTruetemp’smentalstate,wegothroughatwo‐stepprocess,wherethesecond
step(inwhichthetemperaturethoughtsareacceptedwithoutreason)isproblematic,perhapseven
fromareliabilistperspective.Ifwenaturallyseethetempucompasmechanicallydisabling
Truetemp’sordinarycapacitytoweighevidenceashedecideswhattoaccept,andifweseethat
ordinarycapacityasgenerallyconducivetotheproductionoftruebelief,thenthetempucomp
interventioncouldwellcomeacrossasthetypeofthingthatwouldgenerallycompromise
reliability.20
19Onthethoughtthatacceptanceisapersonal‐levelphenomenon,see(Frankish,2009).Asimilartwo‐stagestructureisusedinanothermuch‐discussedinternalistcasedevelopedbyLaurenceBonJour:hisunwittingclairvoyantnotonlyformsvariousaccuratebeliefsthroughherpsychicfaculty,butalso,inadistinctstep,“acceptsthebeliefsinquestion”(BonJour&Sosa,2003,28).20Onthisreading,thecaseisnotintheendaneffectivecounterexampletoreliabilism.Forpresentpurposesitisenoughthatthecaseisambiguous,butmorenegativeinitsoriginalTruetempform;acleardecisionaboutwhattypeofprocessisintuitivelyseenasinstantiatedwouldrequiregrapplingwiththenotoriouslydifficultGeneralityProblem(Conee&Feldman,1998),notataskIwillundertakehere.
Page 26
April7,2011
26
ThesimplerCharlesexampledoesnotdwellontheseparationbetweenthegenerationand
theacceptanceofpropositionsabouttemperature:rather,Charles’sbrainre‐wiringleadsdirectly
toabelief.WheretheoriginalTruetempseemedtobemakingdubiousreflectivejudgments,
perhapsCharlescomesacrossasmakingbetterintuitivejudgments.Butthenewcasedoesnot
strictlyprecludeamorenegativereading:whenwereadthatCharleshas‘apartfromhis
estimation…nootherreasonstobelievethatitis71degrees’,itisopentoustorepresenthimas
activelythinkingtohimselfthathehasnoreasonsforhisbelief,andifweseehimasengagedin
thattypeofreflectiononthewaytomaintaininghisbelief,thenhisjudgmentonceagainappears
unreliable.ThismorereflectiveconstrualoftheCharlescasemaybemoreavailableto
professionalsfamiliarwiththesimilarlynegativeoriginalcase.
Wheretherearemultiplewaysofunderstandinganunder‐describedstory,better
consensusamongprofessionalsmaybearesultofournaturallyfavoringtheconstrualthatmakes
moredialecticalsenseinitsoriginalargumentativecontext.21Someonewhosewayoffleshingouta
caseisnon‐standardcouldhavestrongconfidenceinwhatseemstobeaminorityintuition;
divergentintuitionsonagivencaseinepistemologycouldinsomecircumstancesarisefrom
differentwaysofconstruingvarioussubtlefeaturesofthesubject’swayofthinking,ratherthan
differencesintheassessmentofknowledgeperse.22Itisdoubtlessariskofthecasemethodthat
wecanfillinunder‐describedcaseswithphilosophicallyimportantcontent.23Theproblemisnot
anintractableone,however:disagreementaboutcasescanhelpustoidentifythiscontent.The
problemisalsonotuniquetothecasemethod:perceptioncanalsopresentuswithstimulithatare
ambiguousbetweentwoormoreconstruals,stimulisuchastheNeckercube.
Sofar,theaimhasbeentoshowthatepistemicintuitionsdonotshowpeculiaror
particularlyproblematicformsofvariation.Thekindsofinstabilityfoundinepistemicintuitions
arealsofoundinperceptualjudgments.Amateurjudgmentsofordinaryknowledge,Gettier,
21Thispossibilityisparticularlyrelevanttoourunderstandingofthepairedcasesmotivatingcontextualismandinterest‐relativeinvariantism,whereitisclearthatthepointisthatthereissomewayofreadingthepairsothattheywilldelivercontraryverdicts.Cooperativereaderscanthenworktoconstruethecasesinsuchaway.TomotivatecontextualismorIRI,thesewaysofreadingthecasesdonotnecessarilyhavetobethewaysthatonewouldmostnaturallyreadthesecasesifencounteringtheminisolation.22ItispossiblethatthethinkingofthesubjectintheFakeBarncaseisalsoopentobeingunderstoodinvarioussignificantlydifferentwaysaswell,andthatthisunder‐descriptionalsoexplainslukewarmamateurresponsesandsomedivisioninprofessionalresponsestothecase.23TheimportanceoftacitcommitmentsinourunderstandingofcaseshasbeenexploredmostsystematicallybyTamarGendler:seetheessayscollectedin(Gendler,2011).
Page 27
April7,2011
27
skepticalpressureandfalsebeliefcasesalignroughlywiththejudgmentsofprofessionals,anddo
notseemtovarywithethnicityorgender.Therearenoobviousbarrierstoseeingepistemiccase
intuitionsastheproductsofourordinarymindreadingcapacities.
Alastproblemconcernsourhandlingofillusions.Mindreading,likeperceptionandlike
otherintuitivecapacities,issusceptibletocertainnaturalillusions,mostnotablyepistemic
egocentrism,orthetendencytomisrepresentthecognitionofthosewhooccupyamorenaïve
positionthantheobserver(e.g.Birch&Bloom,2004;2007;Nickerson,1999).Thelegitimacyofthe
casemethoddoesnotrequirethatphilosophersbeimmunetotheseillusions,orabletodistinguish
illusoryfromnon‐illusorycasesimmediatelyastheyareexperienced.Evenadvancedreflection
andtrainingdoesnotinsulateonefromillusion:forexample,physicsgraduatestudentsand
postdoctoralresearchersstillexperiencethecharacteristiccognitive‐perceptualillusionsofnaïve
‘impetustheory’physics,illusionsthatareexplicitlyatoddswiththeirtheoreticalknowledgeand
easilydiscountedonreflection(Kozhevnikov&Hegarty,2001).
Fortunately,inepistemologyasinphysics,intuitionisnottheonlytoolatourdisposal:
considerationsoftheoreticalunificationcanalsosupplysomeguidance.24Wesuspectsensory
illusionwherethedeliverancesofthesensesappeartoconflictwithoneanother,asintheMüller‐
Lyerillusion;wemayhavesimilarsuspicionswherethereisapparentconflictamongourepistemic
intuitions,forexample,conflictofthesortfoundinthecasesmotivatingcontextualism.Itisnot
transparentthattheseapparentlyconflictingintuitionsareillusions;theoristsofvarious
inclinationshavedevelopedinnovativeandsometimesstrangetheoriesofknowledgeand
knowledgeascriptiononwhichtheapparentconflictisnomorethanapparent.Tojudgewhether
thesetheoriesaretrueorfalse,wecandrawonagreatarrayofconsiderationsfromlogic,
linguistics,psychologyandphilosophy;wecanalsodevisenewcasestoofferpositivesupportto
ourtheoriesortoserveascounterexamples.Inepistemology,asinempiricalscience,itisnot
alwaysatrivialmattertodeterminewhetherwearesubjecttoanillusion,orwhetherthe
phenomenonweareinvestigatingisstrangerthanwehadthought.Butthefactthataformof
inquiryisdifficultdoesnotentailthatthereisanythingfundamentallywrongwithitsmethods.25
24Foramoredetaileddiscussedofthispoint,see(Ichikawa,forthcoming).25ForcommentsonanearlierdraftIamgratefultoJaneFriedman,DianaRaffman,SergioTenenbaum,JonathanWeinbergandJonathanWeisberg.
Page 28
April7,2011
28
References
Alexander,J.,&Weinberg,J.M.(2007).Analyticepistemologyandexperimentalphilosophy.PhilosophyCompass,2(1),56‐80.
Alter,A.,&Oppenheimer,D.(2009).Unitingthetribesoffluencytoformametacognitivenation.PersonalityandSocialPsychologyReview,13(3),219.
Alter,A.,Oppenheimer,D.,Epley,N.,&Eyre,R.(2007).Overcomingintuition:Metacognitivedifficultyactivatesanalyticreasoning.JournalofExperimentalPsychology:General,136(4),569‐576.
Apperly,I.(2011).Mindreaders:TheCognitiveBasisof"TheoryofMind".HoveandNewYork:PsychologyPress.
Apperly,I.,Back,E.,Samson,D.,&France,L.(2008).Thecostofthinkingaboutfalsebeliefs:Evidencefromadults'performanceonanon‐inferentialtheoryofmindtask.Cognition,106(3),1093‐1108.
Apperly,I.,Riggs,K.,Simpson,A.,Chiavarino,C.,&Samson,D.(2006).Isbeliefreasoningautomatic?PsychologicalScience,17(10),841.
Baranski,J.V.,&Petrusic,W.M.(1994).Thecalibrationandresolutionofconfidenceinperceptualjudgments.Attention,Perception,&Psychophysics,55(4),412‐428.
Baranski,J.V.,&Petrusic,W.M.(1995).Onthecalibrationofknowledgeandperception.CanadianJournalofExperimentalPsychology,49(3),397‐407.
Birch,S.,&Bloom,P.(2004).Understandingchildren'sandadults'limitationsinmentalstatereasoning.Trendsincognitivesciences,8(6),255‐260.
Birch,S.,&Bloom,P.(2007).Thecurseofknowledgeinreasoningaboutfalsebeliefs.PsychologicalScience,18(5),382‐386.
BonJour,L.,&Sosa,E.(2003).EpistemicJustification:Internalismvs.Externalism,Foundationsvs.Virtues.Malden,MA:Blackwell.
Brewer,W.F.,&Sampaio,C.(2006).Processesleadingtoconfidenceandaccuracyinsentencerecognition:Ametamemoryapproach.Memory,14(5),540‐552.
Buckwalter,W.,&Stich,S.(2011).GenderandthePhilosophyClub.ThePhilosophers'Magazine(52),60‐65.Cesarini,D.,Sandewall,&Johannesson,M.(2006).Confidenceintervalestimationtasksandtheeconomicsof
overconfidence.JournalofEconomicBehavior&Organization,61(3),453‐470.Charman,T.,Ruffman,T.,&Clements,W.(2002).Isthereagenderdifferenceinfalsebeliefdevelopment?
SocialDevelopment,11(1),1‐10.Choi,I.,&Nisbett,R.E.(2000).Culturalpsychologyofsurprise:Holistictheoriesandrecognitionof
contradiction.Journalofpersonalityandsocialpsychology,79(6),890‐905.Chudnoff,E.(2010).Thenatureofintuitivejustification.PhilosophicalStudies,153(2),313‐333.Cohen,A.,&German,T.(2009).Encodingofothers'beliefswithoutovertinstruction.Cognition,111(3),356‐
363.Conee,E.,&Feldman,R.(1998).Thegeneralityproblemforreliabilism.PhilosophicalStudies,89(1),1‐29.Cullen,S.(2010).Survey‐drivenromanticism.ReviewofPhilosophyandPsychology,1,275‐296.DeBruin,W.B.,Fischhoff,B.,Millstein,S.G.,&Halpern‐Felsher,B.L.(2000).VerbalandNumerical
ExpressionsofProbability.OrganizationalBehaviorandHumanDecisionProcesses,81(1),115‐131.DeRose,K.(2009).TheCaseforContextualism:Knowledge,Skepticism,andContext,Volume1.NewYork:
OxfordUniversityPress.Ekman,P.(1989).Theargumentandevidenceaboutuniversalsinfacialexpressionsofemotion.InH.Wagner
&A.Manstead(Eds.),Handbookofsocialpsychophysiology(pp.143‐164).Oxford:JohnWileyandSons.
Ekman,P.,&Friesen,W.V.(1975).Unmaskingtheface:Aguidetorecognizingemotionsfromfacialclues:Prentice‐HallEnglewoodCliffs,NJ.
Evans,J.(2007).Dual‐processingaccountsofreasoning,judgment,andsocialcognition.AnnualReviewofPsychology,59,255‐278.
Frankish,K.(2009).Systemsandlevels:dual‐systemtheoriesandthepersonalsubpersonaldistinction.InTwoMinds:DualProcessesandBeyond,89‐107.
Gendler,T.S.(2011).Intuition,Imagination,andPhilosophicalMethodology.NewYork:OxfordUnivPr.Gettier,E.L.(1963).IsJustifiedTrueBeliefKnowledge?Analysis,23,121‐123.
Page 29
April7,2011
29
Gigerenzer,G.,Hoffrage,U.,&Kleinbolting,H.(1991).Probabilisticmentalmodels:ABrunswikiantheoryofconfidence.Psychologicalreview,98(4),506‐528.
Goldman,A.(1967).ACausalTheoryofKnowing.TheJournalofPhilosophy,64(12),357‐372.Goldman,A.(1976).DiscriminationandPerceptualKnowledge.TheJournalofPhilosophy,73(20),771‐791.Goldman,A.(2006).Simulatingminds:Thephilosophy,psychology,andneuroscienceofmindreading.New
York:OxfordUniversityPress.Goldman,A.I.(1994).Naturalisticepistemologyandreliabilism.MidwestStudiesinPhilosophy,19(1),301‐
320.Heine,S.J.,&Lehman,D.R.(1996).Hindsightbias:Across‐culturalanalysis.JapaneseJournalofExperimental
SocialPsychology,35,317‐323.Ichikawa,J.(forthcoming).Experimentalistpressureagainsttraditionalphilosophicalmethodology.
PhilosophicalPsychology.Kalmus,H.(1979).Dependenceofcolournamingandmonochromatorsettingonthedirectionofpreceding
changesinwavelength.TheBritishjournalofphysiologicaloptics,33(2),1.Kelley,C.(1993).Rememberingmistakenforknowing:Easeofretrievalasabasisforconfidenceinanswers
togeneralknowledgequestions,JournalofMemoryandLanguage(Vol.32,pp.1).Koriat,A.(1975).Phoneticsymbolismandfeelingofknowing.Memory&Cognition,3(5),545‐548.Koriat,A.(1976).Anotherlookattherelationshipbetweenphoneticsymbolismandthefeelingofknowing.
Memory&Cognition,4(3),244‐248.Koriat,A.(1995).Dissociatingknowingandthefeelingofknowing:Furtherevidencefortheaccessibility
model.JournalofExperimentalPsychology:General,124(3),311‐333.Koriat,A.(2008).Subjectiveconfidenceinone'sanswers:Theconsensualityprinciple.JournalofExperimental
Psychology:Learning,Memory,andCognition,34(4),945‐959.Koriat,A.(2011).SubjectiveConfidenceinPerceptualJudgments:ATestoftheSelf‐ConsistencyModel.
JournalofExperimentalPsychology:General,140(1),117‐139.Koriat,A.,&Adiv,S.(inpress).Theconstructionofattitudinaljudgments:Evidencefromattitudecertainty
andresponselatency.SocialCognition.Kozhevnikov,M.,&Hegarty,M.(2001).Impetusbeliefsasdefaultheuristics:Dissociationbetweenexplicit
andimplicitknowledgeaboutmotion.PsychonomicBulletin&Review,8(3),439.Lehrer,K.(1990).TheoryofKnowledge.Boulder:WestviewPress.Lindsey,D.,Raffman,D.,&Brown,A.(MS).Psychologicalhysteresisandthenontransitivityofinsignificant
differences.Mar,R.A.,Kelley,W.M.,Heatherton,T.F.,&Macrae,C.N.(2007).Detectingagencyfromthebiologicalmotion
ofveridicalvsanimatedagents.Socialcognitiveandaffectiveneuroscience,2(3),199.McDowell,J.(1995).KnowledgeandtheInternal.PhilosophyandPhenomenologicalResearch,55(4),877‐893.Mercier,H.(forthcoming).Ontheuniversalityofargumentativereasoning.JournalofCognitionandCulture.Mercier,H.,&Sperber,D.(2009).Intuitiveandreflectiveinferences.J.St.BTEvans&K.Frankish(Eds.),Intwo
minds:Dualprocessesandbeyond.Oxford,UK:OxfordUniversityPress.Nagel,J.(2007).Epistemicintuitions.PhilosophyCompass,2(6),792‐819.Nagel,J.(2010).Knowledgeascriptionsandthepsychologicalconsequencesofthinkingabouterror.
PhilosophicalQuarterly,60(239),286‐306.Nagel,J.(2011).ThepsychologicalbasisoftheHarman‐Vogelparadox.Philosophers'Imprint,11(5),1‐28.Nagel,J.,SanJuan,V.,&Mar,R.(inprep).GettierCaseRecognition.Newell,B.,&Lee,M.(2010).TheRightToolfortheJob?EvidenceAccumulationinDecisionMaking.
Psychologicalreview.Nickerson,R.S.(1999).Howweknow‐‐andsometimesmisjudge‐‐whatothersknow:Imputingone'sown
knowledgetoothers.PsychologicalBulletin,125(6),737‐759.Nisbett,R.E.,Peng,K.,Choi,I.,&Norenzayan,A.(2001).Cultureandsystemsofthought:Holisticversus
analyticcognition.PsychologicalReview,108(2),291‐310.O'Neill,D.,Astington,J.,&Flavell,J.(1992).Youngchildren'sunderstandingoftherolethatsensory
experiencesplayinknowledgeacquisition.ChildDevelopment,63(2),474‐490.Pohl,R.F.,Bender,M.,&Lachmann,G.(2002).Hindsightbiasaroundtheworld.ExperimentalPsychology,
49(4),270‐282.Raffman,D.(2005).Howtounderstandcontextualismaboutvagueness:replytoStanley.Analysis,65(287),
244‐248.
Page 30
April7,2011
30
Reber,R.,&Schwarz,N.(1999).Effectsofperceptualfluencyonjudgmentsoftruth,Consciousnessandcognition(Vol.8,pp.338‐342).Orlando,Fla.:AcademicPress.
Redcay,E.,Dodell‐Feder,D.,Pearrow,M.J.,Mavros,P.L.,Kleiner,M.,Gabrieli,J.D.E.,&Saxe,R.(2010).Liveface‐to‐faceinteractionduringfMRI:Anewtoolforsocialcognitiveneuroscience.NeuroImage,50(4),1639‐1647.
Samson,D.,Apperly,I.,Braithwaite,J.,Andrews,B.,&Scott,S.(2010).Seeingittheirway.JournalofExperimentalPsychologyHumanPerceptionandPerformance,36(5),1255‐1266.
Saxe,R.(2006).WhyandhowtostudyTheoryofMindwithfMRI.BrainResearch,1079(1),57‐65.Saxe,R.,&Wexler,A.(2005).Makingsenseofanothermind:Theroleoftherighttemporo‐parietaljunction.
Neuropsychologia,43(10),1391‐1399.Shatz,M.,Wellman,H.M.,&Silber,S.(1983).Theacquisitionofmentalverbs:Asystematicinvestigationof
thefirstreferencetomentalstate.Cognition,14(3),301‐321.Simmons,J.P.,&Nelson,L.D.(2006).Intuitiveconfidence:Choosingbetweenintuitiveandnonintuitive
alternatives.JournalofExperimentalPsychologyGeneral,135(3),409‐427.Slobin,D.I.(1968).Antonymicphoneticsymbolisminthreenaturallanguages.Journalofpersonalityand
socialpsychology,10(3),301‐305.Sloman,S.(1996).Theempiricalcasefortwosystemsofreasoning.PsychologicalBulletin,119(1),3‐22.Sosa,E.(2007).Experimentalphilosophyandphilosophicalintuition.PhilosophicalStudies,132(1),99‐107.Stanley,J.(2005).KnowledgeandPracticalInterests.NewYork:OxfordUniversityPress.Stanovich,K.,&West,R.(2000).Individualdifferencesinreasoning:implicationsfortherationalitydebate?
TheBehavioralandbrainsciences,23(5),645.Stanovich,K.E.(2005).Therobot'srebellion:FindingmeaningintheageofDarwin:UniversityofChicago
Press.Swain,S.,Alexander,J.,&Weinberg,J.M.(2008).Theinstabilityofphilosophicalintuitions:Runninghotand
coldonTruetemp.PhilosophyandPhenomenologicalResearch,76(1),138‐155.VanderHenst,J.B.,Mercier,H.,Yama,H.,Kawasaki,Y.,&Adachi,K.(2006).Dealingwithcontradictionina
communicativecontext:Across‐culturalstudy.InterculturalPragmatics,3(4),487‐502.vonHelmholtz,H.(1893).Popularlecturesonscientificsubjects.London:Longmans,Greenandco.Weatherson.(2003).Whatgoodarecounterexamples?Philosophicalstudies,115,1‐31.Weinberg,J.M.(2007).Howtochallengeintuitionsempiricallywithoutriskingskepticism.MidwestStudiesin
Philosophy,31,318‐343.Weinberg,J.S.,Nichols,S.,&Stich,S.(2001).NormativityandEpistemicIntuitions.PhilosophicalTopics,29(1),
429‐460.Wellman,H.,Cross,D.,&Watson,J.(2001).Metaanalysisoftheoryofminddevelopment:Thetruthabout
falsebelief.ChildDevelopment,72(3),655‐684.Williamson,T.(2000).KnowledgeanditsLimits.NewYork:OxfordUniversityPress.Williamson,T.(2005).Contextualism,Subject‐SensitiveInvariantismandKnowledgeofKnowledge.The
PhilosophicalQuarterly,55,213‐235.Williamson,T.(2007).ThePhilosophyofPhilosophy.NewYork:Wiley‐Blackwell.Williamson,T.(2011).PhilosophicalExpertiseandtheBurdenofProof.Metaphilosophy,42(3),215‐229.Wright,J.C.(2010).Onintuitionalstability:Theclear,thestrong,andtheparadigmatic.Cognition,115(3),
491‐503.Zagzebski,L.(1994).TheinescapabilityofGettierproblems.ThePhilosophicalQuarterly,44(174),65‐73.