2
elicitednotonlybyadiscreteoutcome,butalsobycuesthatupdateormodifyaprediction.DunningandHajcak(2007)reportedthattheerror-relatednegativity(ERN)1waselicitedbycuesthatpredictedwithcertaintyanimpendingloss,butnotbycuesthatpredictedwithcertaintyanimpendingwin.KrigolsonandHolroyd(2007)alsodemonstratedthatexternalpredictivevisualinformation–predictiveerrorcomparisonbetweenaninternalmotorcommandandrelatedbehaviorelicitsanERN-likewaveform.BakerandHolroyd(2009)reportedthattheFRNismodulatedbypredictivecuesindicatingtheabsenceversusthepresenceofarewardina‘‘virtualT-maze’’environ-ment.TheseresultssuggestthatFRN/ERNcanbeelicitednotonlybydiscrepantoutcomes,butalsobycuesthatsignifyafutureoutcome.
However,predictioncuesintheseearlierstudiesperfectlypredictedalloftheoutcomes,anditremainstobedeterminedwhetherexternalcuesthatprobabilisticallypredictanoutcomewillalsoelicitaFRN.Cuestoafutureoutcomecanleadustomodifyourpredictions,butthesecuesarealmostneverperfectlypredictive.Inmanysituationswepredictoutcomesinaprob-abilisticandincrementalfashion,basedonanaccrualofimper-fectcuesandinformation.Evenwhenwearecon dentaboutaprediction,itisstillpossiblethatanoutcomewillviolatetheprediction;thatis,outcomecuesseldomareabsolutelyperfect.Thus,itisimportanttotestwhetherRL-ERNtheorycanpredictpeople’sresponsestoprobabilisticoutcomecues.Thiswillindi-catetowhatextentRL-ERNtheorycanaccountforbrainactivityassociatedwitheverydayprediction.
ToinvestigatewhetherFRNisassociatedwithcuesthatindicateprobabilisticpendingoutcome,thecurrentstudyemployedamodi edgamblingtaskinwhichpredictivecueswithprobabilisticoutcomeswerepresentedafterparticipantsmadeselectionsbutpriortothe naloutcomes.ERPstime-lockedtocuestimuliwereexamined,aswellasERPstime-lockedto naloutcomes.AccordingtoRL-ERNtheory,FRNshouldbeelicitedbypredictivecues,andlargerFRNamplitudeshouldbeelicitedbycuesindicatingnegativeoutcomesthancuesindi-catingpositiveoutcomes.
Method
Participants
Nineteenundergraduatestudentsvolunteeredtoparticipate.Foursubjectshadtoberemovedfromthestudybecauseofex-cessiveartifactsintheelectroencephalogram(EEG)recordings.Theremainingsubjects(8women,7males),aged20to25years(mean522.1,SD51.2)hadnormalorcorrected-to-normalvi-sion,wereright-handed,andhadnoself-reporteddiagnosedneurologicalorpsychologicaldisorders.TheUniversity’sEthicsCommitteeapprovedthestudy.ParticipantsreceivedRMB---Y23–---Y26(about$3.80–$4.30)fortheirparticipation,withexactpaymentdependentonperformance(seebelow).
1.ERNisanegative-going,midfrontallydistributedpotentialthatiselicitedbyerrorsonspeededresponsetasks.IthasalsobeenusedforanentireclassoferrorrelatednegativityERPcomponents(i.e.,responseERN,FRN,etc.).TheERNhereindictsanERPcomponentthatsharedasimilarlatencyandmorphologywiththetraditionalFRN.
Y.Liaoetal.
ExperimentalDesignandProcedure
Participantswereseatedcomfortably$1minfrontofacom-puterscreeninanelectromagneticallyshieldedroom.Beforetheinitialtrainingphase,participantsweretoldthattheywouldplayagamblinggameandthattherewasatriangleorasquareonthebacksideoftwo‘‘cards’’(i.e.,squaresonacomputermonitor).Participantsreceived20trainingtrialstomakesurethattheyunderstoodthetask.Oneachtrialoneoftheshapeswasdes-ignatedthewinner.Participantswereaskedtoguesswhichofthetwocardshadthewinningshapeontheotherside.Participantsweretoldtheywouldreceive100pointsforeverywinningtrial,andtheir nalpaymentwouldbebasedontheirtotalpointsattheendoftheexperiment.Theywereencouragedtomaximizetheirwinnings.
Additionally,participantsweretoldthatineachblocksomecomplicatedrulewoulddictatewhichshapewouldwin.Theyweretoldthatinmosttrials(80%,althoughparticipantswerenottoldtheexactpercentage)thecardwiththetrianglewasontheleftandthesquarewasontheright;however,thetwocardswouldoccasionallyswitchpositions.Participantswereadvisedtoassumethattheleftcardwouldhavethetriangleandtherightcardwouldhavethesquareinordertomaximizetheirchancestowin.Whenparticipants nishedthetask,theywereinterviewedaboutwhether,andhow,theyhadtriedto gureouttherulethroughoutthetask.
AsshowninFigure1,eachtrialhadtwostages.Participants’outcomepredictionswereinducedinthe rststage:expectancygeneration.Inthisstage,twocardsweredisplayedfacedown,andparticipantscouldchooseacardbypressing‘‘1’’fortheleftcardand‘‘2’’fortheright,usingtheindexandmiddle ngersoftheirrighthand.Afterparticipantsselectedacard,apredictioncueFthewinningshapeFwaspresentedinthecenterofthemonitor.Thisindicatedwhichshapewasthewinnerinagiventrial.Thewinningshapeoneverytrialwasselectedrandomly(withacumulativeproportionof.50pershape).Becausepar-ticipantsknewwhichcard(i.e.,leftorright)wasmostlikelytohaveeachshape(i.e.,triangleorsquare),theycouldpredictwhethertheirchoicewouldberightorwrong,thatis,whethertheoutcomewouldbepositiveornegative.Thispredictionwasthegeneratedexpectancy.Theirexpectationwaspositiveifthewin-ningshapewasusuallyatthechosenlocationandnegativeifitwasusuallyatthenonchosenlocation.Tocheckthatexpectancygenerationworkedasintended,duringthe rst vepracticetri-als,afterthewinningshapewasshown,participantswereasked,‘‘Doyouthinkyouwillwinorloseinthistrial?’’Alloftheiranswersindicatedthattheyhadlearnedtheprobabilisticrulethatwouldgenerateexpectancies.
Inthesecondstageofeachtrial,participantsreceiveda naloutcome.In80%oftrialsthiswascongruentwithparticipants’expectancies;in20%itwasnot.Inhalfofthecongruenttrials(40%ofalltrials),participantsexpectedtobecorrect,andindeedpositivefeedbackwasdisplayed(‘‘---Y---Y---Y---Y’’).Intheotherhalf
(40%),participantsexpectedtobeincorrectandindeedsawnegativefeedback(‘‘0000’’).Inhalfoftheincongruent(unex-pected)trials(10%ofalltrials)negativefeedback(‘‘0000’’)wasdisplayedafterapositiveexpectancy.Intheotherhalfofincon-gruenttrials(10%)positivefeedback(‘‘---Y---Y---Y---Y’’)wasdisplayed
afternegativeexpectancy.Unexpectedtrialswereseparatedbyatleastthreeexpectedtrials.
Thisdesignproducedfourconditionswhenthe naloutcomewaspresented:(i)expectedpositiveoutcomes,(ii)expected
热门文档