Methodological Problems in the Balloon Analogue Risk Task ...

In the Balloon Analogue Risk Task, or BART for short, participants are presented with a computer screen showing a small balloon and a pump. They ... SkiptomaincontentHomeIssuesSubmitArticlesIssue1Meta-ResearchPublishedonOct08,2020DOI10.36850/mr1BurstBeliefs–MethodologicalProblemsintheBalloonAnalogueRiskTaskandImplicationsforItsUsebyKristelDeGrootPublishedonOct08,2020CiteSocialDownloadContentsShowdetailsBurstBeliefs–MethodologicalProblemsintheBalloonAnalogueRiskTaskandImplicationsforItsUseContents·AbstractStudiesinthefieldofpsychologyoftenemploy(computerized)behavioraltasks,aimedatmimickingreal-worldsituationsthatelicitcertainactionsinparticipants.Suchtasksareforexampleusedtostudyriskpropensity,atrait-liketendencytowardstakingoravoidingrisk.OneofthemostpopulartasksforgaugingriskpropensityistheBalloonAnalogueRiskTask(BART;Lejuezetal.,2002),whichhasbeenshowntorelatewelltoself-reportedrisk-takingandtoreal-worldriskbehaviors.However,despiteitspopularityandqualities,theBARThasseveralmethodologicalshortcomings,mostofwhichhavebeenreportedbefore,butnoneofwhicharewidelyknown.Inthepresentpaper,foursuchproblemsareexplainedandelaboratedon:alackofclarityastowhetherdecisionsarecharacterizedbyuncertaintyorrisk;censoringofobservations;confoundingofriskandexpectedvalue;andpoordecomposabilityintoadaptiveandmaladaptiveriskbehavior.Furthermore,foreveryproblem,arangeofpossiblesolutionsisdiscussed,whichoverallcanbedividedintothreecategories:usingadifferent,moreinformativeoutcomeindexthanthestandardaveragepumpscore;modifyingoneormoretaskelements;orusingadifferenttask,eitheranalternativerisk-takingtask(sequentialorotherwise),oracustom-madeinstrument.Itisimportanttomakeuseofthesesolutions,asapplyingtheBARTwithoutaccountingforitsshortcomingsmayleadtointerpretationalproblems,includingfalse-positiveandfalse-negativeresults.Dependingontheresearchaimsofagivenstudy,certainshortcomingsaremorepressingthanothers,indicatingthe(typeof)solutionsmostneeded.Bycombiningsolutionsandopenlydiscussingshortcomings,researchersmaybeabletomodifytheBARTinsuchawaythatitcanoperationalizeriskpropensitywithoutsubstantialmethodologicalproblems.Keywords:BalloonAnalogueRiskTask,risk-taking,uncertainty,expectedvalue,confounding,censoring.PurposeTheBalloonAnalogueRiskTask(BART)isoneofthemostwidelyusedbehavioraltasksinpsychologyandhasanespeciallystrongpresenceinthefieldsofdecisionresearch,addictionresearch,andneuropsychology.Butdespiteitspopularity,researchersusingtheBARTseemlargelyunawareofthetask’smethodologicalshortcomings,whichsometimesleadstoconclusionsthatarenotsupportedbythedata.Thisislikelyaresultoftheseshortcomingsnotbeingwidelyreported,as“failure”isnotconsideredapopularpublishingtheme.Therefore,thepresentpaperaimstogatherandreviewtheseshortcomings,aswellaspotentialsolutions.Take-homemessageTheBalloonAnalogueRiskTask(BART)suffersfromvariousmethodologicalshortcomings.Thepresentpaperanalyzestheseshortcomingsandofferssuggestionstomitigatetheireffects.Finally,itcallsuponresearcherstocriticallyevaluatehowtheseshortcomingsimpacttheirstudiesbeforedecidingwhetherandhowtouseBART.IntroductionToalargeextent,psychologicalsciencerestsonthepromisesofoperationalization:definingfuzzyconceptsasmeasurablevariables,orinotherwords,changingconceptualvariablesintooperationalones(Shuttleworth,2008).Thisprocessisimperativebecausemostconceptsresearchershypothesizeaboutarenotstraightforwardlyquantifiable.Bydefininghowaconceptismeasured,operationalizationallowshypothesestotakeafalsifiableformatandenablesustoreplicatefindings.Inaway,operationalizationsarearbitrary,asconceptscanbedefinedandthusmeasuredinnumerousways–noneofwhicharesurely“right”.Nonetheless,somemeasuresmaybemoresuitablethanothers.Anotableexampleofaconceptthatcanbeoperationalizedinvariouswaysisrisk-taking(Lauriola&Weller,2018),whichhasanimportantplaceinclinical,cognitive,anddevelopmentalpsychology,aswellasinthefieldsofcriminology,economics,andmanagement.Onewayrisk-takingisoperationalizedinthesefieldsisthroughself-reportmeasures,suchastheDomain-SpecificRisk-Taking(DOSPERT)scale(Blais&Weber,2006)andtheFinancialRiskToleranceassessment(Grable,1999).Anotherwayisthroughcomputerizedbehavioraltasks,liketheIowaGamblingTask(Becharaetal.,1994),theCambridgeGamblingTask(Rogersetal.,1999),theGameofDiceTask(Brandetal.,2005),theBalloonAnalogueRiskTask(Lejuezetal.,2002),andthemorerecentbutalreadywidelyusedColumbiaCardTask(Figneretal.,2009).Importantly,thequalityofastudylargelydependsonthedegreetowhichitsoperationalmeasuresreflecttheunderlyingconcept;inthiscase,one’sdispositiontowardsrisk-taking.Ifataskisapoorproxyforaconceptorissubjecttomethodologicalorinterpretationalproblems,anydataresultingfromitareoflimitedvaluetoourunderstandingoftheconcept.Inthisregard,severalstudieshavechallengedtheoperationalizationabilityofthemost-citedrisktask,theIowaGamblingTask(see,e.g.,Brandetal.,2006;Buelow&Suhr,2009;Figneretal.,2009;Maia&McClelland,2004).TheBalloonAnalogueRiskTask,whichisthesecond-mostcited,mayyetsufferfromevenmoresevereissues,hinderingitsabilitytooperationalizerisk-taking.Whilesomeindividualissueshavebeenreportedinpreviouspublications,noliteraturesofarhasdiscussedthesecollectively.Thepresentcommentaryaspirestofillthisgap.TheBalloonAnalogueRiskTaskIntheBalloonAnalogueRiskTask,orBARTforshort,participantsarepresentedwithacomputerscreenshowingasmallballoonandapump.Theyaretoldthateverytimetheyclickthepump,theballoonexpands,andafixedamountofmoney(5cents)isaddedtoatemporarybank.Everypumpalsoincreasesthechanceoftheballoonexploding(markedbya“pop”soundfromthecomputer),resultinginlosingallmoneyinthetemporarybankforthatparticularballoon(trial).Thepointatwhichaballoonexplodesvariesacrosstrials,rangingfromthefirstpumptothepointwheretheballoonfillstheentirescreen.Participantscandecidetostoppumpingtheballoonatanypointduringatrialbyclickingthe“collect”button(leftinFigure1),whichtransfersthemoneyaccumulatedintheirtemporarybanktotheirpermanentone,whileaslotmachinesoundisplayed.Onceaballoonexplodesoronceparticipantscashaballoon’sproceeds,thetrialends,andanew,uninflated,balloonappears.Figure1.Set-upoftheoriginalBalloonAnalogueRiskTaskasdescribedbyLejuezetal.(2002).Figure2.AnapproximateinteractiveillustrationoftheBARTbyLejuezetal.(2001).Note:parametersofFigure2divergefromthosedescribedbyPleskacetal.(2008)andisthereforenotmeanttobetakenasafaithfulrendition.IntheoriginalstudybyLejuezetal.(2002),participantswereinformedthattheywouldcomplete90balloons:30orange,30yellow,and30blueones.Unbeknownsttoparticipants,differentlycoloredballoonshadadifferentchanceofexploding.Theprobabilitydistributiongoverningtheirexplosionpointsconsistedofanarrayofnnnnumbersfromwhichoneverypumparandomnumberwasdrawnwithoutreplacement.Ifa1wasdrawn,theballoonexploded.Thus,theprobabilitypppoftheballoonexplodingonthefirstpumpwas1/n1/n1/n,andtheprobabilityofitexplodingonpumpiii(givennopriorexplosion)waspi= 1n − i + 1p_{i}=\\frac{1}{n\-\i\+\1}pi​= n − i + 11​.Fororangeballoons,thearrayrangedfrom1to8(hencep1=18−1+1=1/8p_{1}=\frac{1}{8-1+1}=1/8p1​=8−1+11​=1/8),foryellowballoonsfrom1to32(p1=132−1+1=1/32p_{1}=\frac{1}{32-1+1}=1/32p1​=32−1+11​=1/32),andforblueonesfrom1to128(p1=1128−1+1=1/128p_{1}=\frac{1}{128-1+1}=1/128p1​=128−1+11​=1/128).Theiraverageexplosionpointswererespectively4,16,and64,withthesame(randomlygenerated)setsofexplosionpointsbeingusedacrossallparticipantstolimitextraneousvariability.Neithertherangesnortheaverageexplosionpointswerecommunicatedtoparticipants.TheBART’sdesignisintendedtoreflectnaturalisticdecision-making,inwhichtakingmoreriskgenerallyincreasestheoddsofencounteringaloss.Thissortofdecision-makingtendstobeemotionallyengaging,instigatingasenseofincreasingtensionastheballoonincreasesinsize(Schonbergetal.,2011).InsupportoftheBART’svalidity,Lejuezetal.(2002)showedthattheaveragenumberoftimesparticipantspumpedtheblueballoonsignificantlycorrelatedwithscoresonrisk-relatedconstructs(sensationseeking,impulsivity)andwithreal-worldriskbehaviors,suchaspolydruguse,gambling,unsafesex,andstealing.Theorangeandyellowpumpswereoriginallynotexaminedwithrespecttorisk-relatedconstructs,astheirnarrowrangesofoutcomevalues(1-8and1-32)arelesssuitedforcapturingindividualdifferences.Instead,theiraveragepumpnumberswereanalyzedtogetherwiththoseoftheblueballoonstoshowthatthenumberoftimesparticipantschoosetopumpissensitivetotheprobabilityofexploding.Overall,thedatashowedtheBARTtohave“particularpromiseasabehavioralindexofrisk-taking”(Lejuezetal.,2002,p.82).Aswouldbeexpectedbasedonthisconclusion,theBART(particularlyitsblueballoon)becameapopularinstrumentforgaugingindividuals’propensityforrisk-taking,withinconsistentfindingsbeingattributedtofactorslikesamplingvariabilityandinadequatestatisticalpower(Lauriolaetal.,2014),ratherthanproblemsinherenttotheBART.However,severalauthorshavearguedthatsuchproblemsexist(DeGroot&Thurik,2018;Guetal.,2018;Schmidtetal.,2019;Schonbergetal.,2011),andthattheylimittheBART’sabilitytomeasureone’spropensityfortakingrisk.ThekeyproblemsthatcharacterizetheBARTare1)alackofclarityastowhetherdecisionsarecharacterizedbyuncertaintyorrisk,2)censoringofobservations,3)confoundingofriskandexpectedvalue,and4)poordecomposabilityintoadaptiveandmaladaptiveriskbehavior.Riskoruncertainty?Ineconomictheoriesofdecision-making,akeydistinctionisthatbetweenuncertaintyandrisk,whichisoftenaccreditedtoKnight(1921),andwasintroducedtopsychologicalthinkinginaseminalpaperbyEdwards(1954)thatliesattheoriginofbehavioraldecisiontheory.Whendecidingundertheconditionofrisk,theprobabilitiesassociatedwiththepossibleoutcomesareknown.Whendecidingunderuncertainty(whichsomeauthorscallambiguity),thisprobabilitydistributionisunknown.ForKnight(1921),thisdistinctionwasnotonlyoftheoreticalbutofpracticalimportanceaswell.Accordingtohim,uncertainty–notrisk–wasthemaindriverofentrepreneurialsuccess,asonlypeoplewhorecognizehiddenopportunitiescanseizethemandprofitfromthem.Sincethen,theempiricalrelevanceoftheuncertainty-riskdistinctionhasbeenconfirmedinvariousfieldsofresearch.Ineconomics,Ellsberg(1961)showedthatindividualspreferriskoveruncertainty,eveniftheknownprobabilitiesareunfavorableandtheuncertainoptioncouldbeaguaranteedwin.Inpsychology,studiesshowedthatuncertainandriskydecisionsinvolvedifferentmentalprocesses,asriskallowsforstatisticalthinking(tooptimize)butuncertaintyinvolvesheuristics(tosatisfice)(Volz&Gigerenzer,2012).Inlinewiththis,decision-makingunderriskisthoughttodependmoreonexecutivefunction(suchascategorizationandcognitiveflexibility)forwhichthedorsolateralprefrontalcortexisimportant,whereasdecision-makingunderuncertaintyhingesonemotionalprocesses(suchassomaticfeedback),whicharemoreassociatedwiththeventromedialprefrontalcortexandtheamygdala(Brandetal.,2006).Thismayexplainwhypatientswithexecutivedeficits,suchasthosewithParkinson’sdisease,havedifficultydecidingunderriskbuthavenotroubledecidingunderuncertainty(Euteneueretal.,2009),whereaspersonswithobsessive-compulsivedisorder,forexample,showtheoppositepattern(Starckeetal.,2009;Starckeetal.,2010).Giventhatuncertaintyandriskdifferboththeoreticallyandempirically,itisimperativeforresearcherstoknowtheconditionsunderwhichparticipantsdecide.Unfortunately,despitetheword“risk”initsname,theseconditionsarenotstraightforwardintheBART.Sinceparticipantsarenevergiven“detailedinformationabouttheprobabilityofanexplosion”(Lejuezetal.,2002,p.77),wecanassumethatatleastduringearlytrials,theydecideunderuncertainty(Bisharaetal.,2009;DeGroot&Thurik,2018;Schonbergetal.,2011).Astheymovefurtheralonginthetaskand“samplethedistribution”bypumpingballoonsandobservingtheiroutcomes,theygetabettersenseoftheprobabilities,whichgraduallymovestheirdecisionsinthedirectionofrisk.AlthoughnotstudiedintheBARTitself,suchashifthasbeenshownfortheIowaGamblingTask,whereperformanceinearlytrialsdoesnotcorrelatewiththatinlatertrialsnorwithexecutivefunction,indicatingthatpeoplefirstdecideunderuncertaintyandlaterunderrisk(Brandetal.,2006;Brandetal.,2007).WhilethiseffectmaynotbeasstrongintheBART,studiesdoshowbetterperformanceinlatercomparedtoearlytrials,suggestingthatparticipantsindeedgetabettergraspoftheprobabilitydistributionovertime(DeGroot&Strien,2019;Lejuezetal.,2002).1TheBART’stransitionfromuncertaintytowardsriskisproblematicforseveralreasons.First,itisunclearwhenexactlythisshifttranspires,makingitdifficulttodeterminewhetheradecisioninagiventrialismadeunderuncertainty,risk,orsomethinginbetween.Second,thepointwheredecisionsshiftfromuncertaintytoriskislikelytodifferbetweenindividuals,andisdependentontaskcharacteristics(Brandetal.,2006;Brandetal.,2007).Third,theshiftimpliesthattheBARTimposeslearningdemands,whichcouldinadvertentlyimpactparticipants’outcomesonthetask,withthosecapableofupdatingtheirknowledgeoftheprobabilitiesperformingbetterthanthosewhohavedifficultydoingso.Fourth,onceparticipantsmanagetoderivethetask’sprobabilities,subsequentdecisionsarenotcharacterizedbywhatisusuallyconsideredrisk.Contrarytodecisionsinwhichprobabilitiesareexplicitlydescribed(“apriori”probabilities),probabilitiesintheBARTarederivedfromexperience.Sincesuchprobabilitiesdependonfactorslikesamplingvariabilityandone’smemoryofpreviousevents,decision-makerstreatexperience-basedprobabilitydifferently,whichiscalledthedescription-experiencegap(Hauetal.,2008;Rakow&Newell,2010).Mostnotably,whendecidingbasedonexperience,peopledonotactinaccordancewithprospecttheory,butinstead,underweightrareeventsandoverweightcommonencounters.Aspeoplehavemoreandmoreencounters(e.g.,trials),theirexperienceswillapproachtheprecisionofaprioriprobabilities,thoughinpracticethisisdifficulttoattain(Knight,1921).ToaddresstheinabilityoftheBARTtodifferentiatebetweencompleteuncertainty,experience-basedrisk,anddescription-basedrisk,severalapproachesmaybeused.OneoptionistoapplyamodeltotheBART’sdatathatallowsforparticipantslearningthroughexperience.AnearlyexampleisamodelbyWallstenetal.(2005)inwhichdecision-makersupdatetheirprobabilitiesfromtrialtotrial,andcontinuallyre-evaluatetheiroptions.Alternatively,onecoulduseadifferenttask,inwhichdecisionsareeitherallcharacterizedbyuncertaintyorrisk,orwhichincludesawell-understoodshiftbetweenthetwo.Tasksthatinvolveonlyuncertaindecision-makingareratherdifficulttodesign,astheyrequireparticipantstobeignorantofprobability-relatedinformationandremainignorantofthataswell–automaticallydisqualifyingtasksthathavealearningcurve.Tasksinvolvingonlydecisionsmadeunder(apriori)riskaremuchmorecommonandincludetheCambridgeGamblingTask,theGameofDiceTask,andtheColumbiaCardTask,thelatterofwhichresemblestheBART’sdynamic,affectivenature(Schonbergetal.,2011).Finally,aknownshiftfromuncertaintyto(experience-based)riskcanbefoundintheIowaGamblingTask.Thistask’sshift,whilenotfullyunderstood,hasbeenstudiedmorethoroughlythanthatintheBART.CensoredobservationsStatisticalcensoringreferstoaconditioninwhichthevalueofanobservationisunknownbecauseitisbeyondacertainlimit.Thislimitcanexistbydesign,whichiscommoninsurvivalanalysis.Ifastudyonasurgicalinterventionfollowspatientsforupto10years,thelongevityscoresofthosewholivepastthistermarecensored,astheirlongevityisatleast10(Young&McCoy,2019).Censoringcanalsoresultfromlimitsonwhataninstrumentcanreliablymeasure.Forexample,thefullIQscoreoftheWechslerAdultIntelligenceScalerangesfrom40to160(Sattler&Ryan,2009),meaningthatIQscoresofpeopleperformingeitherextremelypoorlyorextremelywellarecutoffattheseboundariesandarethuscensored.IntheBART,censoring(bydesign)occursifaparticipantisstoppedfromtakingmoreriskinagiventrial,becausetheballoontheyarepumpingexplodes,forcingthetrialtoend.Sincesuchatrialendsprematurely,thenumberoftimestheparticipantpumpedtheballoondoesnotnecessarilyreflecttherisktheywerewillingtotake,meaningtheirriskpropensityiscensored.Thisisproblematicforvariousreasons.First,includingthesecensoredtrialsbiasestheaveragenumberofpumpsdownwards(especiallyforhigh-risktakers),underestimatingparticipants’willingnesstotakerisks(Dijkstraetal.,2020;Pleskacetal.,2008).Likewise,thebetween-subjectsvariabilityacrosstheseaveragesisreduced(Lejuezetal.,2002).Overall,the(unadjusted)averagenumberofpumpsisanill-suitedoperationalizationofriskpropensity.Ascensoringaffectsallsequentialrisk-takingtasksliketheBART(involvingmultipledecisionspertrial)andvariousotherresearchparadigms,likesurvivalanalysis,severalsolutionshavebeenproposed.InthepaperintroducingtheBART,Lejuezetal.(2002)suggestcomputinganadjustedpumpaverageusingonlytrialsinwhichparticipantsstoppedvoluntarily,thatis,inwhichtheballoondidnotburst.However,byomittingexplosiontrials,censoredobservationsareessentiallytreatedasrandomlymissing,whichisinaccurate(Pleskacetal.,2008).Themorerisksomeonetakes,themorelikelyitisthattheballoonbursts,andthatthetrialforciblyends.Theterminationoftrialsisthereforenotindependentfromparticipants’behavior.Asaresult,Lejuezetal.’sadjustedscoretendstodiscardtrialsinwhichparticipantstakealotofrisk.Thiscausestheaveragenumberofpumpstobebiaseddownwards,similartotheunadjustedscore,buttoalesserextent.Tocircumventtheproblemofcensoring,Pleskacetal.(2008)developedanautomaticresponseversionoftheBART.ContrarytothestandardBART,inwhichparticipantsinflateaballoononepumpatatime,theautomaticBARTletsthemindicatetheirintendednumberofpumpsbeforehand.Theballoontheninflatestothecorrespondingsize,oruntilitbursts.Thisprocedureallowsforanunbiasedstatisticofriskpropensity,astheintendednumberofpumpsisnowobservableinalltrials(Pleskacetal.,2008).However,itincreasesthetimebetweendecisionandoutcome,whichmaymakedecisionslessemotional(impulsive)andmorecognitive(planned)(Pleskacetal.,2008),andmayreducethesalienceoftheoutcomes.Theseeffects,inturn,canaffectparticipants’risk-taking(Young&McCoy,2019).Incontrast,however,astudyusingtheBombRiskElicitationTask(BRET;Crosetto&Filippin,2013),anotherrisktaskthatusesdelayedexplosionstocircumventcensoring,foundthatintroducingsuchdelaysdidnotimpactrisk-taking.Figure3.AnapproximateinteractiveillustrationoftheAutomatedBARTtaskbyPleskacetal.(2008).Note:parametersofFigure3divergefromthosedescribedbyPleskacetal.(2008)andisnotmeanttobetakenasafaithfulrendition.Anothersolutiontocensoringisusingariggedtask(Slovic,1966).Participantsarethentoldthatfailurecanoccuratanymoment(intheBART,atanypump),butactually,itissettooccuratthelastpossiblechoice.Hence,participantscanalwaysstopvoluntarily,andnoscoresarecensored.Toupholdcredibility,“mock”trialsareadded,inwhichfailureissettooccurearlyon.Decidingonthenumberandtimingofmocktrials,however,isachallenge.Sincebehaviorinatrialisaffectedbypreviousoutcomes,experiencing(too)fewfailurescouldincreaserisk-taking(DeGroot&Strien,2019;Dijkstraetal.,2020).Therefore,riggedtasksshouldbedesignedsuchthattheyproducefailureratessimilartonon-riggedtasksandshouldtakeintoaccountthatfailureratesdifferbetweenparticipantstoo.However,researchontheColumbiaCardTask,anothersequentialrisk-takingtask,showsthatthisisoftennotthecase(DeGroot&Strien,2019).Afinalremedy,whichaddressesthebiasbutleavestheBARTunchanged,istoapplyastatisticalmodeltotheresultingdatathatexplicitlyincorporatescensoredbehavior.Suchmodelsconsiderallobserveddata,usingthecensoredtrialsaslowerboundsindeterminingaparticipant’sactualriskpropensity.SomeofthememployBayesian(generalized)linearmixed-effectsregression(Welleretal.,2019;Young&McCoy,2019);othersusemaximumlikelihoodestimation,addingacumulativedistributionfunctiontothelikelihoodfunctiontoaccountforcensoring(Dijkstraetal.,2020;Tobin,1958).Suchmodelsperformsignificantlybetter(i.e.,havelessbiasedpredictions)thanthosethatdonotaccountforcensoring.However,asisthecaseforallstatisticalmodels,theirsoundnesshingesonthevalidityoftheirunderlyingassumptions(Schafer&Graham,2002),suchasthatofnormality,whoseviolationnotallmodelsarerobustagainst(Powell,1984).ConfoundinganddecomposabilityTheBARTwasdesignedtoresemblereal-worldrisksituations,wheretakingmodestriskisgenerallyadvantageous,buttakingexcessiveriskisincreasinglyunfavorable(Lejuezetal.,2002;Wallstenetal.,2005).Withinatrial,everysuccessfulpumpearnsparticipants5cents,whichareaddedtotheirtemporarybank.Astheamountaccumulatedinthebankgrows,therelativegainoftakingadditionalriskdecreases,whilethepotentiallossincaseofanexplosionincreases.Additionally,theprobabilityoftheballoonexplodingincreaseswitheverypump:from1/128onthefirstto1/127onthesecond,andsoon.Thiscombinationofcharacteristicsmakesthatthetask’sstructureentailsaseriousproblem.Sinceboththeballoonvalue(theamountcollectedinthetemporarybank)andtheexplosionprobabilityincreasewitheverypump,theexpectedvalueofinflatingtheballoon–theproductofthesuccesschanceandthereward,minustheproductoftheexplosionchanceandtheballoonvalue–changesacrossatrial(Schmidtetal.,2019).ThischangeisillustratedinTable1.Earlyinatrial,theexpectedvalueofthepumpispositive,sotakingadditionalriskisadvantageous.Thisprospectchangeshalfwaywhentheexpectedvalueturnsnegative,makingadditionalpumpsunfavorable(Lejuezetal.,2002).Duetotheexpectedvaluechangingwitheachdecision,itisconfoundedwithrisk(definedasthevariabilityofthepossibleoutcomes),whichvariesacrossdecisionsbydesign.Althoughsuchconfoundingcanhappeninreal-lifedecision-making,itisnotdesirableinacontrolledscientificenvironment:itmakesitdifficulttomeasureparticipants’riskpropensity,asbothriskandexpectedvaluemayinfluencetheirdecisions.Theextenttowhichindividualsare,forexample,risk-seeking,canthereforenotbedetermined,becausethiswouldrequireshowingapreferenceforhighervariancepayoffs,holdingexpectedvalueconstant(Schonbergetal.,2011).ThisconfoundingdemonstratesthattheBART’smainobservableoutcome–thenumberofpumpsparticipantspress–cannotbeinterpretedasastraightforwardindicatorofriskpropensity.Likemanybehavioraltasks,theBARTsupposedlygaugesasinglecognitiveconstruct,butitmanipulatesvariousother,potentiallyconfoundingconstructsaswell(Schonbergetal.,2011).Expectedvalueisanexampleofsuchaconstruct.Asaresult,thesinglescoreprovidedbytheBARTcannoteasilybedecomposedtoidentifythecognitiveorneuralmechanismsinvolvedinthepumpdecisions.Studyingtherisk-takingprocessinisolationusingtheBARTisthereforenotpossible.OneapproachforresolvingtheconfoundinganddecomposabilityissuesintheBARTistoapplyacomputationalmodeltoitsdatathatquantifiesthecognitivemechanismsunderlyingtheobservedbehavior(Bisharaetal.,2009).Suchmodelswerefirstproposedby(Wallstenetal.,2005),inspiredbyanexpectancy-valencemodelfordecomposingbehaviorintheIowaGamblingTask(Busemeyer&Stout,2002).Wallstenetal.(2005),explaindecisionvariabilityusingoneparameterforrisk-taking,oneforresponseconsistency,andtwoforlearning.Byapplyingthesemodels,wecanstudyrisk-taking–andotheraspectsthatdetermineBARTbehavior–inisolation,bytranslating“whatisobservedbutrelativelyuninformativetowhatisunobservedandrelativelyinformative”(vanRavenzwaaijetal.,2011,p.95).However,datafromtheBARTmaynotberichenoughtowarranttheuseofcomplicateddecompositionmodels.Forinstance,astudyonWallstenetal.’sbestperformingmodeldemonstratedthatitslearningparameterscouldnotreliablyberecovered(vanRavenzwaaijetal.,2011).Toallowformoreextensivedecomposition,onemayneedtoresorttoadifferenttask,liketheIowaGamblingTask.Alternatively,onecoulduseataskthatbydesignavoidsconfounding,suchastheColumbiaCardTask.AlthoughdynamicandaffectiveliketheBART,thistaskorthogonallyvariesrisk-relatedconstructs,sothattheycanbedecomposedintotheirunderlyingmechanisms–likesensitivitytogains,losses,andprobabilities–withouttheuseofacomputationalmodel(Dijkstraetal.,2020;Figneretal.,2009;Schonbergetal.,2011).Finally,researcherscanchoosetodesignacustomtasktoensurethattheconstructsrelevanttotheirhypothesesarenotconfounded.Forexample,arisktaskpresentedinSchmidtetal.(2013)variesthelevelofriskbutholdsexpectedvalueconstant.Solutionssuchastheseshouldbeconsideredcarefullysothatconstructscrucialtoastudy’shypothesescanbeisolatedeffectively.Table1ChangingBalloonValues,ExplosionandSuccessChances,andExpectedValuesAcrossBalloonPumpsPumpNumber(A)BalloonValueBeforePump(B)BalloonValueAfterPump(C)ChanceofExplosion(D)ChanceofSuccess(E)ExpectedValueofCurrentPump(F)ExpectedValueofAllRemainingPumps(G)1€-€0.050.007810.99219€0.04961€1.600002€0.05€0.100.007870.99213€0.04921€1.562603€0.10€0.150.007940.99206€0.04881€1.525404€0.15€0.200.008000.99200€0.04840€1.488405€0.20€0.250.008060.99194€0.04798€1.45161(…)62€3.05€3.100.014930.98507€0.00373€0.0067263€3.10€3.150.015150.98485€0.00227€0.0030364€3.15€3.200.015380.98462€0.00077€0.0007765€3.20€3.250.015630.98438€-0.00078€-0.0007866€3.25€3.300.015870.98413€-0.00238€-0.00238(…)124€6.15€6.200.200000.80000€-1.19000€-1.19000125€6.20€6.250.250000.75000€-1.51250€-1.51250126€6.25€6.300.333330.66667€-2.05000€-2.05000127€6.30€6.350.500000.50000€-3.12500€-3.12500128€6.35€6.401.000000.00000€-6.35000€-6.35000Note:Theexpectedvalueofthecurrentpump(F)iscomputedbymultiplyingthesuccesschance(E)by0.05,thensubtractingtheproductoftheexplosionchance(D)andtheballoonvaluebeforethepump(B)[F=E*0.05-D*B].Alternatively,onecanalsotakeintoaccounttheexpectedvalueofanysubsequentpumps,insofarastheyareadvantageous(G).Thisresultsinsomewhatdifferentvalues,butanidenticaltippingpointat64.ThenormativesolutionTheBARTisdesignedinsuchawaythattheballoons’averageexplosionpointliesat64,halfwaythemaximumnumberofpumps.Thisisachievedbyrandomlygeneratingcollectionsofexplosionpointsuntiloneproducesanaverageof64overalltrials,aswellaswithineachsetof10trials(Lejuezetal.,2002).Participantscanthenmaximizetheirearningsbyattemptingtopumpeveryballoon64times,whichresultsinanexplosioninabouthalfofthetrials,andanoptimaloverallexpectedvalue.GoingbacktoTable1,wecanseeexactlywhythisistheoptimal,ornormative,solutionintheBART.Uptoandincludingthe64thpump,theexpectedvalueofpumpingtheballoonispositive;after64,theexpectedvalueis(increasingly)negative.Itis,therefore,optimaltoaimfor64pumpsoneveryballoon,andthenstop.Choosingtopumpmoreorfewerthan64timeswilldecreaseexpectedearnings;andthefartheronedeviatesfromtheoptimum,thelowertheexpectedearningsbecome(Lejuezetal.,2002;Pleskacetal.,2008;Wallstenetal.,2005).Remarkably,inmosttrials,participantsstoppumpingtheballoonfarbeforetheoptimalstoppingpoint(Lejuezetal.,2002).Infact,theaverageadjustedpumpscoreistypicallybetween26and35(Pleskacetal.,2008).Real-worldrisk-avoidersandrisk-takersalikerarelypumptheballoonenoughtimestomaximizetheirexpectedearnings.ThisislessofaproblemintheautomaticBART,althoughparticipantstherestillpumpfewerthan64timesonaverage.Forexample,tworecentstudiesreportedaveragesof61.9(Bernosteretal.,2019)and58.5pumps(DeGroot&Strien,2019).Itisyetunknownexactlywhyparticipantsoftenstoppumpingbeforetheyreachtheoptimalpoint,butvariousfactorsmayplayarole.First,sincetheoriginalBARTrequiresparticipantstoinflateballoonsonepumpatatime,itisplausiblethattheygettiredofpumpingafterawhile.Second,participantsmaywanttolimittheireffortoutoflazinessoradesiretofinishearly(butseeYoung&McCoy,2019).Third,theymaybecomesatiated:duetodiminishingmarginalreturns,adding5centstoagrowingtemporarybankmaystopbeinganattractiveprospectwellbeforereachingpump64.Fourth,participantsmayneedtimetolearnwhichstrategyresultsinmaximalearnings(Lejuezetal.,2002).ThisconjectureissupportedbytheobservationthatparticipantsinboththeoriginalandtheautomaticBARTonaveragepressclosertothenormativesolutioninthefinalblockof10trialsthantheydoinpreviousblocks(Lejuezetal.,2002;DeGroot&Strien,2019).1Italsocorrespondswiththepresumedshiftfromdecidingunderuncertaintytodecidingunderrisk.IntheBART,learningtheoptimalsolutionishard,astherangeofpossibleexplosionpointsislarge(1-128),andindividualexplosionsprovidelimitedfeedback.ThisisinlinewithfindingsbyLejuezetal.(2002),whoshowthatlargerexplosionrangesresultinlargerrelativedeviationsfromtheoptimum.ThefactthatparticipantsintheBARToftenstoppumpingbeforetheoptimalstoppingpointhasseriousimplicationsforhowthedatacanbeinterpreted.Upto64pumps,therisktheytakecanbecharacterizedasadaptiveorfunctional,asitresultsinhigherearnings.Afterthatpoint,itcanbeconsideredmaladaptiveordysfunctional,asitreducesexpectedearnings.Sincepeoplegenerallypumpfewerthan64times,theBARTcannotproperlydifferentiatebetweenadaptiveandmaladaptiveriskbehavior,neitherwithinnorbetweenparticipants.Asecond,relatedproblemisthatexperimentalmanipulationsmeanttoincreaserisk-taking(suchasaddingtimepressureoradministeringacertaindrug)generallydonotleadtolowerearnings,aseventheresultinghigherpumpnumbersusuallydonotexceed64(Pleskacetal.,2008).Forexample,ifamanipulationcausesparticipantstotakemoreriskandpress50insteadof30times,theyareactually,onaverage,betteroffthanbefore,theoppositeofwhatonewouldexpectinreallife.Inshort,ifparticipantsmostlystayunder64pumps,theysimplyneverreachthepointwheretakingmoreriskbecomesdisadvantageous,whichlimitstheconclusionsonecandrawfromthedata.ThemoststraightforwardwaytomitigatetheseproblemsmaybethemodifiedBARTdevelopedbyPleskacetal.(2008),whichdiffersfromtheoriginaltaskinthreeways.First,itinvolvesanautomaticresponsemode:participantsindicatetheirintendednumberofpumpsatthestartofeachtrial,afterwhichtheballoonautomaticallyinflatestothecorrespondingsize(oruntilitbursts).Althoughmeanttomitigatecensoring,thisadjustmentmayalsopreventpeoplefromgettingtiredofpumpingandfromwantingtofinishthetasksooner.Second,theadjustedtaskprovidesexplicitfeedbackabouttheexplosionpointofeveryballoon,notmerelyofthosethatactuallyexplode.Thismayimproveparticipants’learningacrosstrials.Third,participantsare(truthfully)informedthattherangeofpumpnumbersis1-128andthatthebestoverallnumberofpumpsis64,furtherincreasingtheamountofinformationattheirdisposal.Thesethreemodificationstogethersuccessfullymovedparticipants’behaviorclosertothenormativesolutionof64,withanaveragepumpscoreof57.7forfemalesand63.7formales(Pleskacetal.,2008).Partofthiseffectcanbeattributedtotheautomaticresponsemode,astheseaveragesarehigherthanthosefromamanualBARTwithfullfeedbackandstrategyinstructionsadded.SincethismanualBARTitselfresultedinhigheraveragesthantheoriginalBART,thefeedbackandinstructionslikelyalsocontributedtotheeffect(Lejuezetal.,2002).Recentresearch,however,indicatesthatinformingparticipantsabouttheoptimalstrategyisnotnecessary,andevenill-advised.TwostudiesusinganautomaticBARTwithfullfeedback–butwithoutstrategyinstructions–foundequallyhighpumpaveragesasdidPleskacandcolleagues(Bernosteretal.,2019;DeGroot&Strien,2019).Additionally,thesestudiesfoundthatasubgroupofparticipants–oftenfromaSTEMbackground–seemtoinfertheoptimalstrategywithoutanyhelp.2Theirrepeated64-answers,therefore,reflectcognitiveabilityratherthanriskpropensityandreducetaskvariability.Informingparticipantsabouttheoptimalstrategycanincreasesuchproblematicresponses.Therefore,itseemsbesttoaddautomaticresponsesandfullfeedbacktotheBART,butnotstrategyinstructions.Thiswilllikelyelicitsufficientlyhighpumpaverages,withoutcompromisingthevalidityofthetask.DiscussionSinceitwasfirstpublishedin2002,theBARThasbecomeoneofthemostpopulartoolsinpsychologytogaugeindividuals’propensityforrisk-taking.Halfwaythrough2020,theoriginalarticledescribingtheBART(Lejuezetal.,2002)hadbeencitedover1100timesinScopus,mostofteninjournalsondecisionresearch,addiction,andneuropsychology.Thispopularityiswell-founded.TheBARTsucceedsinrecreatingthe“natural”feelingofexhilarationandtensionpeopleexperiencewhentakingrisk,andthushasexcellentecologicalvalidity.Furthermore,itcorrelateswellwithself-reportedrisk-relatedconstructs,suchasimpulsivityandsensation-seeking,andwithreal-worldriskbehaviors,likepolydruguseandunsafesex,supportingitsconvergentvalidity.Lastly,itdoesnotcorrelatewithconstructslikedepressionandanxiety,endorsingitsdiscriminantvalidity(Lejuezetal.,2002).Butdespitethesequalities,theBARTsuffersfrommethodologicalproblems,mostofwhichhavebeenacknowledgedinpreviousresearchasnegativelyimpactingitsrigor.Thepresentpaperisthefirsttogiveacomprehensiveoverviewoftheseproblems.ThefirstproblemconcernsthelackofclarityastowhetherdecisionsintheBARTaremadeunderuncertainty(whereoutcomeprobabilitiesareunknown)orrisk(wheretheyareknown).Sinceparticipantsarenotgivenanyinformationabouttheexplosionprobabilities,theyfirstdecideunderuncertainty,whichthengraduallyshiftstowardsriskastheylearnmoreabouttheprobabilitiesinthetask.Asitisunclearexactlywhenthisshifttakesplace,itisdifficulttodeterminewhetheragivendecisionismadeunderuncertainty,risk,orsomethinginbetween.Thesecondproblemconcernsstatisticalcensoring,whichoccursintrialswheretheballoonexplodes,asparticipantsarethenpreventedfromtakingadditionalrisk.Asaresult,theaveragenumberoftimesparticipantspumptheballoonunderestimatestheirriskpropensity.Third,theBARTconfoundsriskwithexpectedvalue.Sincetheseconstructschangesimultaneouslythroughoutatrial,participants’pumpbehavioragaindoesnotreflectriskpropensity,asdecisionsareinfluencedbybothriskandexpectedvalue.Thisalsomeansthatthetaskispoorlydecomposable,asitcannotdisentanglethemotivesunderlyingapumpdecision.Afinalproblemconcernsthetask’snormativesolution.Inthemajorityoftrials,participantsstoppumpingbeforethepointwhereexpectedearningsaremaximized.Therefore,participantsmostlytakeadaptiverisk,whichleadstohigherearnings.Maladaptiverisk-takinghardlyoccurs,eventhoughonewouldexpecttoseesuchbehaviorincertaincases.Despitetheseproblems,muchoftheresearchuptonowhasfocusedontheempiricalfindingsproducedbytheBART,ratherthanonthetaskitself,withthemajorityofresearchersusingthetaskwithoutcriticallyreviewingwhetheritsproblemsinterferewiththeiraims.Thiscanhaveundesirableconsequences,suchaswhenitleadstofalsepositivesorfalsenegatives.Forexample,onemayfailtoshowarelationshipwhichonlyexistsfordecisionscharacterizedbyrisk,assometrialsintheBARTarecharacterizedbyuncertaintyinstead.Conversely,ahypothesismaypertaintopeople’sresponsetochangingriskandbeunjustlysupported,asintheBART,riskandexpectedvaluesimultaneouslychangeandimpactindividuals’behavior.Findingtruepositivesandnegativeshingesonseveralfactors,animportantonebeingthevalidityofthemeasurementinstrument.Anydataresultingfrominstrumentsthatsufferfrommethodologicalorinterpretationalproblemsisoflimitedvaluetounderstandingtheconceptstheyaresupposedtooperationalize.Forthesereasons,itisimperativethatresearcherscriticallyevaluatethe“fit”betweentheirresearchandtheBARTbeforedecidingonusingit.Formanyresearchaims,onewillnowseethattheoriginalBARTdoesnotsuffice.Yetdespitethese“burstbeliefs”,therearethreetypesofapproachesonecantaketoaccountforitslimitations.First,datafromtheoriginalBARTcanbeanalyzedusingadifferent,moreinformativeindexthanLejuezetal.’saverageadjustedpumpscore.Forexample,themodelsbyWallstenetal.(2005)breakdownbehaviorintorisk-taking,responseconsistency,andlearning.Inaddition,computationalmodelscanbeusedtotakeintoaccountcensoringandtoprovideanindexofuncensoredrisk-takingintheBART(Dijkstraetal.,2020;Tobin,1958;Welleretal.,2019;Young&McCoy,2019).AsecondwayofdealingwiththeBART’slimitationsisbymodifyingthetask,forexamplebyriggingit(Figneretal.,2009;Slovic,1966),providingadditionalfeedback,orautomatingtheresponses(Pleskacetal.,2008).Third,onemayconsiderusingadifferenttask.Thiscanbeanexisting(sequential)risk-takingtask,liketheColumbiaCardTask(Figneretal.,2009),whichperformsbetterintermsofdecomposabilitythantheBART.Alternatively,researchersshouldconsidercreatingacustomtaskthatexactlysuitstheirresearch,avoidingmethodologicalflawsthatcouldendangerthesoundnessoftheirconclusions.Forinstance,ataskdevelopedbySchmidtetal.(2013)involvesdecisionsunderconditionsofexplicitriskanddoesnotconfoundriskwithexpectedvalue.Animportantgoaltokeepinmindwhendesigningsuchbespoketasksistocombinestrongecologicalvaliditywithmethodologicalrigor(Schonbergetal.,2011).Clearly,noneofthesolutionsproposedcanbeconsidereda“universal”fixthatsolvesalloftheBART’sproblems.Dependingontheaimsofanygivenstudy,certainproblemswillbemorepressingthanothers,indicatingthe(typeof)solutionsmostneeded.Bycombiningsolutions,researcherscouldworktowardsataskthatcanoperationalizeriskpropensitywithoutsubstantialmethodologicalorinterpretationalproblems.Forexample,anautomaticBARTwithfullfeedbackandexplicitinformationontheprobabilitydistributionprovidesuncensoreddecisionsmadeunderclearriskthatareattimesriskyenoughtobemaladaptive.IftheresultingdatafromthisadaptedBARTarethenanalyzedusingamodellikethatbyWallstenetal.(2005)orthatbyvanRavenzwaaijetal.(2011),allproblemsreviewedinthecurrentcommentarywouldbeaddressed.However,thisdoesnotnecessarilymeanthatthiscombinationofsolutionsconstitutesauniversalfixafterall,astheBARTmayfacemoreproblemsthantheonesdiscussedhere.Inalllikelihood,thepresentreviewisnotexhaustive.ResearchersusingtheBARTmayknowofadditionalproblems,althoughthisisunlikelytoshowintheirwork,asjournals–andbyextensionresearchers–donotconsider“failure”apopularpublishingtheme(Ferguson&Heene,2012;Songetal.,2009).Therefore,itisimportantforresearcherstonotonlycriticallyevaluatetheinstrumentstheyusebuttodisclosetheseevaluationsaswell,sothatanyandallmethodologicalshortcomingscanbeopenlydiscussedandaddressed,improvingthequalityofthemeasuresused.ConclusionThepresentpaperisthefirsttoreviewthemethodologicalshortcomingsoftheBalloonAnalogueRiskTask,ahighlypopularrisk-takingtaskinpsychology.Themainproblemsidentifiedaretheambiguitybetweenuncertaintyandrisk,censoringofobservations,confoundingofriskandexpectedvalue,andpoordecomposabilityintoadaptiveandmaladaptiverisk-taking.Inaddition,thepaperreviewssolutionsthatmitigatetheseproblems.Bypresentingthisfirst-timeinventory,thepaperhighlightsearliermentionsofproblemsintheBARTaswellasproposedsolutions.ItcallsforacriticalattitudetowardstheBARTandexperimentaltasksingeneral,astheirdesigndeservesatleastasmuchattentionasthefindingstheyproduce.Italsosetstheagendafortestingandcomparingdifferenttasksandtaskversions,toexplorewhichdesignsresultinthebestusability,reliability,andvalidity,sothatriskpropensitycanbemeasuredinthemostaccuratewaypossible.PeerreviewJOTEaimstomakethepeerreviewprocessaccessibletoitsreaders.Therefore,theinitialsubmissionwithintegratedpeerreviewcommentsisavailablehere.ReferencesBechara,A.,Damasio,A.R.,Damasio,H.,&Anderson,S.W.(1994).Insensitivitytofutureconsequencesfollowingdamagetohumanprefrontalcortex.Cognition,50(1–3),7–15.https://doi.org/10.1016/0010-0277(94)90018-3Bernoster,I.,DeGroot,K.,Wieser,M.J.,Thurik,R.,&Franken,I.H.A.(2019).Birdsofafeatherflocktogether:Evidenceofprominentcorrelationswithinbutnotbetweenself-report,behavioral,andelectrophysiologicalmeasuresofimpulsivity.BiologicalPsychology,145,112–123.https://doi.org/10.1016/j.biopsycho.2019.04.008Bishara,A.J.,Pleskac,T.J.,Fridberg,D.J.,Yechiam,E.,Lucas,J.,Busemeyer,J.R.,Finn,P.R.,&Stout,J.C.(2009).Similarprocessesdespitedivergentbehaviorintwocommonlyusedmeasuresofriskydecisionmaking.JournalofBehavioralDecisionMaking,22,435–454.https://doi.org/10.1002/bdm.641Blais,A.,&Weber,E.U.(2006).Adomain-specificrisk-taking(DOSPERT)scaleforadultpopulations.JudgmentandDecisionMaking,1,33–47.https://doi.org/10.1037/t13084-000Blais,A.-R.,&Weber,E.U.(2012).Domain-SpecificRisk-TakingScale[Dataset].AmericanPsychologicalAssociation.https://doi.org/10.1037/t13084-000Brand,M.,Recknor,E.C.,Grabenhorst,F.,&Bechara,A.(2007).Decisionsunderambiguityanddecisionsunderrisk:Correlationswithexecutivefunctionsandcomparisonsoftwodifferentgamblingtaskswithimplicitandexplicitrules.JournalofClinicalandExperimentalNeuropsychology,29,86–99.https://doi.org/10.1080/13803390500507196Brand,Matthias,Fujiwara,E.,Borsutzky,S.,Kalbe,E.,Kessler,J.,&Markowitsch,H.J.(2005).Decision-makingdeficitsofKorsakoffpatientsinanewgamblingtaskwithexplicitrules:AssociationsWithExecutiveFunctions.Neuropsychology,19(3),267–277.https://doi.org/10.1037/0894-4105.19.3.267Brand,Matthias,Labudda,K.,&Markowitsch,H.J.(2006).Neuropsychologicalcorrelatesofdecision-makinginambiguousandriskysituations.NeuralNetworks,19(8),1266–1276.https://doi.org/10.1016/j.neunet.2006.03.001Buelow,M.T.,&Suhr,J.A.(2009).ConstructvalidityoftheIowaGamblingTask.NeuropsychologyReview,19(1),102–114.https://doi.org/10.1007/s11065-009-9083-4Busemeyer,J.R.,&Stout,J.C.(2002).Acontributionofcognitivedecisionmodelstoclinicalassessment:DecomposingperformanceontheBecharagamblingtask.PsychologicalAssessment,14(3),253–262.https://doi.org/10.1037/1040-3590.14.3.253Crosetto,P.,&Filippin,A.(2013).The“bomb”riskelicitationtask.JournalofRiskandUncertainty,47(1),31–65.https://doi.org/10.1007/s11166-013-9170-zDeGroot,K.,&Strien,J.W.(2019).Event‐relatedpotentialsinresponsetofeedbackfollowingrisk‐takinginthehotversionoftheColumbiaCardTask.Psychophysiology,56(9).https://doi.org/10.1111/psyp.13390DeGroot,K.,&Thurik,R.(2018).Disentanglingriskanduncertainty:Whenrisk-takingmeasuresarenotaboutrisk.FrontiersinPsychology,9,2194.https://doi.org/10.3389/fpsyg.2018.02194Dijkstra,N.F.,Tiemeier,H.,Figner,B.C.,&Groenen,P.J.(2020).Acensoredmixturemodelformodelingrisktaking(2002.08146).arXiv.Edwards,W.(1954).Thetheoryofdecisionmaking.PsychologicalBulletin,51(4),380–417.https://doi.org/10.1037/h0053870Ellsberg,D.(1961).Risk,ambiguity,andthesavageaxioms.TheQuarterlyJournalofEconomics,75(4),643.https://doi.org/10.2307/1884324Euteneuer,F.,Schaefer,F.,Stuermer,R.,Boucsein,W.,Timmermann,L.,Barbe,M.T.,Ebersbach,G.,Otto,J.,Kessler,J.,&Kalbe,E.(2009).Dissociationofdecision-makingunderambiguityanddecision-makingunderriskinpatientswithParkinson’sdisease:Aneuropsychologicalandpsychophysiologicalstudy.Neuropsychologia,47(13),2882–2890.https://doi.org/10.1016/j.neuropsychologia.2009.06.014Ferguson,C.J.,&Heene,M.(2012).Avastgraveyardofundeadtheories:Publicationbiasandpsychologicalscience’saversiontothenull.PerspectivesonPsychologicalScience,7(6),555–561.https://doi.org/10.1177/1745691612459059Figner,B.,Mackinlay,R.J.,Wilkening,F.,&Weber,E.U.(2009).Affectiveanddeliberativeprocessesinriskychoice:AgedifferencesinrisktakingintheColumbiaCardTask.JournalofExperimentalPsychology:Learning,Memory,andCognition,35(3),709–730.https://doi.org/10.1037/a0014983Grable,J.(1999).Financialrisktolerancerevisited:Thedevelopmentofariskassessmentinstrument.FinancialServicesReview,8(3),163–181.https://doi.org/10.1016/s1057-0810(99)00041-4Gu,R.,Zhang,D.,Luo,Y.,Wang,H.,&Broster,L.S.(2018).PredictingriskdecisionsinamodifiedBalloonAnalogueRiskTask:Conventionalandsingle-trialERPanalyses.Cognitive,Affective,&BehavioralNeuroscience,18(1),99–116.https://doi.org/10.3758/s13415-017-0555-3Hau,R.,Pleskac,T.J.,Kiefer,J.,&Hertwig,R.(2008).Thedescription-experiencegapinriskychoice:Theroleofsamplesizeandexperiencedprobabilities.JournalofBehavioralDecisionMaking,21(5),493–518.https://doi.org/10.1002/bdm.598Knight,F.H.(1921).Risk,uncertaintyandprofit.SentryPress.Lauriola,M.,Panno,A.,Levin,I.P.,&Lejuez,C.W.(2014).Individualdifferencesinriskydecisionmaking:Ameta-analysisofsensationseekingandimpulsivitywiththeBalloonAnalogueRiskTask:Personalityandriskydecisionmaking.JournalofBehavioralDecisionMaking,27(1),20–36.https://doi.org/10.1002/bdm.1784Lauriola,M.,&Weller,J.(2018).PersonalityandRisk:BeyondDaredevils—RiskTakingfromaTemperamentPerspective.InM.Raue,E.Lermer,&B.Streicher(Eds.),PsychologicalPerspectivesonRiskandRiskAnalysis(pp.3–36).SpringerInternationalPublishing.http://link.springer.com/10.1007/978-3-319-92478-6_1Lejuez,C.W.,Read,J.P.,Kahler,C.W.,Richards,J.B.,Ramsey,S.E.,Stuart,G.L.,Strong,D.R.,&Brown,R.A.(2002).Evaluationofabehavioralmeasureofrisktaking:TheBalloonAnalogueRiskTask(BART).JournalofExperimentalPsychology:Applied,8(2),75–84.https://doi.org/10.1037/1076-898x.8.2.75Maia,T.V.,&McClelland,J.L.(2004).Areexaminationoftheevidenceforthesomaticmarkerhypothesis:WhatparticipantsreallyknowintheIowagamblingtask.ProceedingsoftheNationalAcademyofSciences,101(45),16075–16080.https://doi.org/10.1073/pnas.0406666101Pleskac,T.J.,Wallsten,T.S.,Wang,P.,&Lejuez,C.W.(2008).Developmentofanautomaticresponsemodetoimprovetheclinicalutilityofsequentialrisk-takingtasks.ExperimentalandClinicalPsychopharmacology,16(6),555–564.https://doi.org/10.1037/a0014245Powell,J.L.(1984).Leastabsolutedeviationsestimationforthecensoredregressionmodel.JournalofEconometrics,25(3),303–325.https://doi.org/10.1016/0304-4076(84)90004-6Rakow,T.,&Newell,B.R.(2010).Degreesofuncertainty:Anoverviewandframeworkforfutureresearchonexperience-basedchoice.JournalofBehavioralDecisionMaking,23(1),1–14.https://doi.org/10.1002/bdm.681Rogers,R.D.,Owen,A.M.,Middleton,H.C.,Williams,E.J.,Pickard,J.D.,Sahakian,B.J.,&Robbins,T.W.(1999).Choosingbetweensmall,likelyrewardsandlarge,unlikelyrewardsactivatesinferiorandorbitalprefrontalcortex.TheJournalofNeuroscience,19(20),9029–9038.https://doi.org/10.1523/jneurosci.19-20-09029.1999Sattler,J.M.,&Ryan,J.J.(2009).AssessmentwiththeWAIS-IV.JeromeM.SattlerPublisher.Schafer,J.L.,&Graham,J.W.(2002).Missingdata:Ourviewofthestateoftheart.PsychologicalMethods,7(2),147–177.https://doi.org/10.1037/1082-989x.7.2.147Schmidt,B.,Kessler,L.,Holroyd,C.B.,&Miltner,W.H.R.(2019).Wearingabikehelmetleadstolesscognitivecontrol,revealedbylowerfrontalmidlinethetapowerandriskindifference.Psychophysiology,56(12).https://doi.org/10.1111/psyp.13458Schmidt,B.,Mussel,P.,&Hewig,J.(2013).I’mtoocalm-Let’stakearisk!Ontheimpactofstateandtraitarousalonrisktaking:I’mtoocalm-Let’stakearisk!Psychophysiology,50(5),498–503.https://doi.org/10.1111/psyp.12032Schonberg,T.,Fox,C.R.,&Poldrack,R.A.(2011).Mindthegap:Bridgingeconomicandnaturalisticrisk-takingwithcognitiveneuroscience.TrendsinCognitiveSciences,15(1),11–19.https://doi.org/10.1016/j.tics.2010.10.002Shuttleworth,M.(2008).Operationalization.https://explorable.com/operationalizationSlovic,P.(1966).Risk-takinginchildren:Ageandsexdifferences.ChildDevelopment,37(1),169.https://doi.org/10.2307/1126437Song,F.,Parekh-Bhurke,S.,Hooper,L.,Loke,Y.K.,Ryder,J.J.,Sutton,A.J.,Hing,C.B.,&Harvey,I.(2009).Extentofpublicationbiasindifferentcategoriesofresearchcohorts:Ameta-analysisofempiricalstudies.BMCMedicalResearchMethodology,9(1),79.https://doi.org/10.1186/1471-2288-9-79Starcke,K.,Tuschen-Caffier,B.,Markowitsch,H.J.,&Brand,M.(2010).Dissociationofdecisionsinambiguousandriskysituationsinobsessive–compulsivedisorder.PsychiatryResearch,175(1–2),114–120.https://doi.org/10.1016/j.psychres.2008.10.022Starcke,K.,Tuschen-Caffier,B.,Markowitsch,H.-J.,&Brand,M.(2009).Skinconductanceresponsesduringdecisionsinambiguousandriskysituationsinobsessive-compulsivedisorder.CognitiveNeuropsychiatry,14(3),199–216.https://doi.org/10.1080/13546800902996831Tobin,J.(1958).Estimationofrelationshipsforlimiteddependentvariables.Econometrica,26(1),24.https://doi.org/10.2307/1907382vanRavenzwaaij,D.,Dutilh,G.,&Wagenmakers,E.-J.(2011).CognitivemodeldecompositionoftheBART:Assessmentandapplication.JournalofMathematicalPsychology,55(1),94–105.https://doi.org/10.1016/j.jmp.2010.08.010Volz,K.G.,&Gigerenzer,G.(2012).Cognitiveprocessesindecisionsunderriskarenotthesameasindecisionsunderuncertainty.FrontiersinNeuroscience,6.https://doi.org/10.3389/fnins.2012.00105Wallsten,T.S.,Pleskac,T.J.,&Lejuez,C.W.(2005).Modelingbehaviorinaclinicallydiagnosticsequentialrisk-takingtask.PsychologicalReview,112(4),862–880.https://doi.org/10.1037/0033-295x.112.4.862Weller,J.A.,King,M.L.,Figner,B.,&Denburg,N.L.(2019).Informationuseinriskydecisionmaking:Doagedifferencesdependonaffectivecontext?PsychologyandAging,34(7),1005–1020.https://doi.org/10.1037/pag0000397Young,M.E.,&McCoy,A.W.(2019).Variationsontheballoonanaloguerisktask:Acensoredregressionanalysis.BehaviorResearchMethods,51(6),2509–2521.https://doi.org/10.3758/s13428-018-1094-8DeGroot,K.(2020)-BurstBeliefs;MethodologicalProblemsintheBalloonAnalogueRiskTaskandImplicationsforItsUse.pdf459KBConnections1of2childrenandsiblingsfilterASupplementtothisPubErratum:BurstBeliefs–MethodologicalProblemsintheBalloonAnalogueRiskTaskandImplicationsforItsUsebyKristelDeGrootPublishedonNov23,2020archive.jtrialerror.comAReviewofthisPubPeerreviewof“BurstBeliefs–MethodologicalProblemsintheBalloonAnalogueRiskTaskandImplicationsforItsUse”byMichaelYoungandSihuaXuPublishedonDec16,2020archive.jtrialerror.comLicenseCreativeCommonsAttribution4.0InternationalLicense(CC-BY4.0)Comments0SortcommentsFiltercommentsCollapsethissection?LogintodiscusscommentNocommentshereWhynotstartthediscussion?JournalofTrial&ErrorAboutLegalSubmitQuestionsTheTeamLegalStatusForAuthorsWhyPubPub?ManifestoOpenAccessPolicyOpenCallsFAQRetractionsandCorrectionsPolicyISSN(Online)2667-1204RSSLegalPublishedwith