4
Holroyd&Krigolson,2007).Speci cally,foreachparticipantandchannel,wecreatedthreeFRNdifferencewavesby(1)sub-tractingthepositiveexpectancyERPfromthenegativeexpec-tancy,creatinga‘‘expectancy’’differencewave;(2)subtractingthepositiveoutcomerelatedERPintheexpectedconditionfromthenegativeoutcomerelatedERPinexpectedcondition,creat-ingan‘‘expected-outcome’’differencewave;and(3)subtractingthepositiveoutcomerelatedERPintheunexpectedconditionfromthenegativeoutcomerelatedERPintheunexpectedcon-dition,creatingan‘‘unexpected-outcome’’differencewave.We rstdetectedthepeaksofFRNsasthemostnegativede ectionwithin150to500msfollowingthewinningshapeorthefeedbackstimulusonsetforeachindividual.MaximumFRNamplitudeswereaggregatedwithneighboringamplitudevalues(peakmax-imumÆ10ms)andanalyzedformidlinechannelsFz,Cz,andPz.TheFRNwasstatisticallyevaluatedusingSPSS(version15.0)software.Forallanalysesofvariance(ANOVAs),thede-greesoffreedomwereGreenhouse–Geissercorrectedwhenap-propriate.Currentsourcedensitywascomputedusingsphericalsplineinterpolation(Perrin,Pernier,Bertrand,&Echallier,1989,1990).Thesplinenumberwassettobe4andthemaximaldegreeofLegendrepolynomialswas10.Thedefaultlambdawassetto1E-5.
Results
BehavioralResults
Toensurethevalidityofthedata,weanalyzedwhetherpartic-ipantswere,infact,tryingto gureoutthewinningrule.Thiswouldindicatethattheyrememberedtheinstructionsandweremotivatedtosucceed.Toruleoutanysimplisticstrategysuchasasidebias,weexaminedthedatafornonrandompatternsofan-swersthatwerenottiedtothecuesorrewardcontingencies.Tothisend,wecalculatedthepercentageoftrialsonwhicheachparticipantselectedtheleftcard(P[l])andtherightcard(P[r]).Wethenexaminedhowoftenparticipantsselectedtheleftcardjustafterselectingtheleftcard(P[l/l])orrightcard(P[l/r]),andhowoftentheyselectedtherightcardjustafterchoosingtheleftcard(P[r/l])ortherightcard(P[r/r]).Weassumedthatifpar-ticipantsfollowedthetaskdemandto gureouttherulethatdeterminedthewinningshape,theywouldnotrandomlyselectthecardsbutwouldattempttoinfersomekindofrule,whichwouldbeevidentinnonrandomsequencesofresponses.Thus,ifparticipantshaveasidebias,theiroverallchoiceofleftversusrightcardsshouldbenonrandom;bycontrast,iftheyarechoos-ingstrategically,theircurrentresponseshouldnotbeindepen-dentofthepreviouschoice.
Overall,participantsselectedleftcardsasoftenasrightcards.NodifferencewasfoundbetweenP(l)(mean5.52;SD5.05)andP(r)(mean5.48;SD5.05).Thus,participantshadnosidebias.However,P(l/l)wassigni cantlyhigherthanP(r/l),t(14)54.39,po.01,andP(r/r)wassigni cantlyhigherthanP(l/r),t(14)52.51,po.05,indicatingthatcardselectionwaspartlydependentontheprevioustrial.Theresultsuggeststhatparticipantsbelievedthattheycouldlearntheruleandthereforecontinuouslytriedtoupdatetheirexpectancies.Participants’exitinterviewsalsocon rmedthis:Allparticipantsreportedthattheytriedto gureouttherulesthatdictatedthewinningshape.Thus,thebehavioraldatashowedthatparticipantsweremotivatedandfollowedthetaskinstructions.
Y.Liaoetal.
ERPsTime-LockedtothePredictiveCue
Negative-expectancytrialswereassociatedwithafrontalnega-tive-goingde ectionthatpeakedapproximately280msafteronsetofthepredictivecue(thewinningshape;Figure3).Figure3alsopresentsthedifferencewavesforthethreemidlineelectrodes(Fz,Cz,andPz)obtainedbysubtractingpositiveexpectancyfromnegativeexpectancy.Table1presentstheaverageFRNamplitudesateachrecordingsite.TheFRNamplitudewasmaximalatchannelFzanddeclinedtowardmoreposteriorsites.ArepeatedmeasuresANOVArevealedsigni cantdifferencesforchannellocation(frontal,centralandparietal),F(2,28)59.44,po.01,Z25.403.AposthocLSDtestshowedFRNamplitudetobesigni cantlylargeratFzascomparedtoCz(po.05or,atPz,po.01),andFRNamplitudetobelargeratCzascomparedtoPz(po.05).Figure3alsoshowsthescalpcurrentdensityofthedifferencewaveat288msaswellasthetopographicdistri-butionofthecomponent.Bothofthesefeaturesrevealedaprominentfrontalmidlinedistribution.
FRNTime-LockedtotheFinalOutcome
WeobtainedaverageERPsforpositiveandnegativefeedbackforexpectedandunexpectedoutcomesatFz,Cz,andPz(seeFigure4).Consistentwithpreviousstudies,negativeoutcomewasassociatedwithafrontallymaximalnegativede ectionthatpeakedapproximately290msfollowingthe naloutcome.Fig-ure5presentsthedifferencewaveobtainedbysubtractingpos-itivefromnegativefeedbackforbothlevelsofexpectancy(expectednegativeminusexpectedpositiveandunexpectedneg-ativeminusunexpectedpositive)aswellastheassociatedsourcedensities.Table1presentstheaverageFRNamplitudesateachrecordingsite.A2Â3repeatedmeasuresANOVAondifferencewaveamplitudeswiththefactorsexpectancy(expectedvs.un-expected)andlocation(Fz,Cz,andPz)revealedsigni cantlylargerFRNamplitudesforunexpectedthanexpectedoutcomes,F(1,14)521.16,po.001;Z25.602.Thatis,whenthefeedbackwasunexpected,ithadagreaterimpactonFRNamplitudes.Inaddition,therewasamaineffectoflocation,F(2,28)56.98,po.01;Z25.332.Consistentwiththefronto-centralmaximumreportedinpreviousstudies,posthoctestsindicatedthattheFRNwaslargeratFzandCzascomparedtoPz(po.01andpo.01,respectively)butthatFRNamplitudewasnotsigni -cantlydifferentatFzandCz(p5.903).Theinteractionofexpectancyandlocationdidnotreachsigni cance.
ComparisonbetweenCue-RelatedandOutcome-RelatedFRNA3Â3repeatedmeasuresANOVAfordifferencewaveampli-tudeswiththefactorslocation(Fz,Cz,andPz)andFRNclass(cue-relatedFRN,expectedoutcome-relatedFRNvs.unex-pectedoutcome-relatedFRN)revealedsigni cantdifferencesinFRNamplitudesatdifferentelectrodelocations,F(2,28)511.29,po.01,Z25.446.PosthoctestsindicatedthatFRNamplitudeswerelargeratFzandCzascomparedtoPz(allpso.01)butshowednodifferencebetweenFzandCz(p5.618).Inaddition,themaineffectofFRNclassreachedsigni cance,F(2,28)513.91,po.01,Z25.50.PosthoctestsrevealedthatpredictivecueselicitedsmallerFRNamplitudesascomparedtounexpectedoutcomes(po.01),butpredictivecueselicitedmarginallysigni cantlargerFRNamplitudesthanexpected
热门文档