Brainactivityofoutcomeexpectation
7
ofthedoorswouldcontainaprizeonthattrial.Theyfoundthataancy(ifany)issmallwhenoutcomeisexpected.Aprediction‘‘0’’cuethatpredictedanupcominglosselicitedanERN.InerroralsooccurredwhenthepredictivecueindicatedanegativeKrigolsonandHolroyd’s(2007)study,participantsusedajoy-outcome.Thisisduetothediscrepancybetweenparticipants’sticktoperformacomputer-basedcontinuoustrackingtaskininitialexpectationoftheoutcomewhentheyselectedacardandwhichsometrackingerrorswereinevitable.Halfoftheseerrorstheirupdatedexpectationswhenthepredictivecuewaspre-wereprecededbyapredictivecue.TheyreportedthatanERN-sented.Giventhatparticipants’initialexpectationswouldnotbelikewaveformwaselicitedbytrackingerrors.Morerecently,sostrongbecausetheyhadlimitedinformation(i.e.,theirpriorBakerandHolroydreportedthatwhenparticipantswereaskedprobabilities)andthatpredictivecueisonlyprobabilistic,thesizeto ndarewardina‘‘virtualT-maze’’environment,predictiveofthepredictionerrorgeneratedbythepredictivecuewasnotascuesindicatingtheabsenceversuspresenceofarewarddiffer-bigastheonegeneratedbyunexpectedoutcomes,butlargerthanentiallymodulatedFRNamplitudes.
theonegeneratedbyexpectedoutcomes.Thatis,itwasofin-Inallofthesestudies,thecueswerede nitepredictorsofawintermediatemagnitude.ThevarianceinFRNdifference-waveorloss.Insuchstudiesthe‘‘cue’’essentiallyservesasperfectamplitudesthusseemtoberelatedtothemagnitudeofthepre-errorfeedback.Bycontrast,thecurrentstudyprovedthatFRNsdictionerrorineachcondition.
areelicitedbyprobabilisticexpectancy-generatingcuesandthatInsum,ERPstime-lockedtooutcomefeedbackreplicatetheamplitudeisgreaterwhentheoutcomeisexpectedtobeprevious ndingsthattheFRNamplitudeismodulatedbyannegative.Thus,theFRNissensitivetopositive/negativeeval-interactionofoutcomevalenceandoutcomeprediction.Inad-uationofone’sperformancerelativetoprobableoutcomes,evendition,thecurrentstudyisthe rsttodemonstratethatFRNifthoseoutcomesareuncertain.
couldbeelicitedbycuesthatprobabilisticallypredictimpendingAlso,ERPslockedtothefeedbackgivenattheendofthetrialoutcome.TheFRNelicitedbypredictivecuespeakedaroundcon rmtheinteractionbetweenoutcomevalenceandexpectancy280msafterstimulionsetandrevealedamaximumovertheinFRNamplitude(Holroyd&Coles,2002;Holroyd&Krigol-frontalmidline.ThiswassimilartoFRNelicitedbyunexpectedson,2007;Holroydetal.,2003;Nieuwenhuisetal.,2005).Differ-outcomes.ItisalsoconsistentwithpreviousreportsonFRN.encewavesbetweenpositiveandnegativetrialsshowalargerThus,theresultsareconsistentwithRL-ERNtheory.AccordingFRNdifferencefollowingunexpectedoutcomes,butarelativelytoRL-ERNtheory,FRNisarewardpredictionerrorsignal.smallFRNdifferencefollowingexpectedoutcomes.AccordingtoWhenongoingeventsareworsethanexpected,phasicdecreasesHolroydetal.(2008),negativedifferencewavesarenotduetoainthemidbraindopaminesystemareissuedtoanteriorcingulatenegativepotentialelicitedbynegativefeedback,butare,infact,cortex,suchthatalargeFRNisproduced(Holroyd&Coles,duetoapositive-goingde ectionelicitedbythepositivefeedback2002).ThetheoryholdsthatanyrewardpredictionerrorwillinthetimerangeoftheFRN.Intheauthors’view,theFRNismodulateFRNamplitude.TheresultsalsoareconsistentwithessentiallyanotherformoftheN200thatiselicitedbynegativesingle-neuronrecordingsfromdopamine-efferentcellsinnon-feedbackandismodulatedbythesuperpositionofapositive-humananimals.Single-neuronrecordingsfromDAneuronsingoingde ectiononcorrecttrials,whichtheycallthefeedback-ratsorprimateshaveshownthatwhenananimalispresentedcorrect-relatedpositivity(fCRP).Unexpectedpositivefeedbackwithaconditionedcuethatpredictsafuturereward,phasicre-elicitsalargerfCRPthanexpectedpositivefeedback.Consistentsponsesareinducedthatencodeanexpectationofthetimeofwiththisview,ourresultsdemonstratemorepositive-goingERPrewardaswellasapredictionerroriftheoutcomedoesnotoccurde ectionswhenthe naloutcomewasunexpectedandpositiveas(Roesch,Calu,&Schoenbaum,2007;Schultz,2006).Oneofthecomparedtoexpectedandpositive.Thismightcontributetothemainprojectionregionsofmidbraindopaminergicnuclei,themaineffectoflargerERPdifferencewaveswhenthe nalout-striatum,encodestheexpectedvalueofpossibleactionsaswellascomeswereunexpectedthanwhentheywereexpected.
thedifferencebetweentheexpectedandobtainedvaluesoftheThefCRP,however,doesnotseemtoaccountforthecue-resultsofactions(Samejima,Ueda,Doya,&Kimura,2005).relateddatabecausetheredoesnotseemtobeanobviousdiffer-SimilarparallelshavebeenfoundinACC,whichisthelikelyenceinpositivitybetweenthetwowaveforms.ComparedtosourceoftheFRN,andisanotherprominenttargetofDApro-FRNsthatareassociatedwith naloutcomes,thepresentresultsjections(Williams&Goldman-Rakic,1998).ACCneuronsinrevealedsmallercue-relatedFRNamplitudesthanunexpectednonhumanprimatesencodearewardpredictionthatincludesoutcome-relatedFRN,butslightlylargercue-relatedamplitudesinformationaboutrewardmagnitudeandrewardprobabilitythanexpectedoutcome-relatedFRN(marginallysigni cant).(Amiez,Joseph,&Procyk,2006;Matsumoto,Suzuki,&Tan-TheresultsareconsistentwiththepredictionofRL-ERNtheoryaka,2003).Furthermore,ACCneuronsencodeaquantitativethatFRNamplitudeispositivelyrelatedtothesizeoftheout-rewardpredictionerroroftheoutcome(Amiezetal.,2005;comepredictionerror(Holroyd&Coles,2002;Holroyd,Krigol-Matsumoto,Matsumoto,Abe,&Tanaka,2007;Seo&Lee,son,Baker,Lee,&Gibson,2009).Inthecurrentstudy,2007).ThecurrentresultssupporttheviewthatFRNisarewardunexpectedoutcomesgeneratedalargepredictionerrorwhereaspredictionerrorsignalinhumans.TheysuggestthatFRNisexpectedoutcomesgeneratedasmallpredictionerror.Thisissensitivetorewardpredictionerrorsthataresignaledbydis-becausethediscrepancybetweenparticipants’expectationandcrepantoutcomesaswellasrewardpredictionerrorsthatareoutcomeislargewhenoutcomeisunexpectedandthisdiscrep-
basedstrictlyonprobabilisticpredictivecues.
References
AmericanElectroencephalographicSociety.(1994).GuidelinesforAmiez,C.,Joseph,J.P.,&Procyk,E.(2005).Anteriorcingulateerror-standardelectrodepositionnomenclature.JournalofClinicalNeuro-relatedactivityismodulatedbypredictedreward.EuropeanJournalofphysiology,11,111–113.
Neuroscience,21,3447–3452.
热门文档