Inter-Individual Differences in Executive Functions Predict ...

Human multitasking suffers from a central attentional bottleneck preventing parallel performance of central mental operations, ... DownloadArticle DownloadPDF ReadCube EPUB XML(NLM) Supplementary Material Supplementaldata totalviews ViewArticleImpact SHAREON MarcoSteinhauser CatholicUniversityofEichstätt-Ingolstadt,Germany TiloStrobach MedicalSchoolHamburg,Germany HildeHaider UniversityofCologne,Germany Theeditorandreviewer'saffiliationsarethelatestprovidedontheirLoopresearchprofilesandmaynotreflecttheirsituationatthetimeofreview. Abstract Introduction ExperimentSeries1 ExperimentSeries2 GeneralDiscussion DataAvailabilityStatement EthicsStatement MembersoftheBrunelStudents AuthorContributions Funding ConflictofInterest Publisher’sNote Acknowledgments SupplementaryMaterial Footnotes References Opensupplementaldata Exportcitation EndNote ReferenceManager SimpleTEXTfile BibTex Checkforupdates Peoplealsolookedat ORIGINALRESEARCHarticle Front.Psychol.,11May2022Sec.Cognition https://doi.org/10.3389/fpsyg.2022.778966 Inter-IndividualDifferencesinExecutiveFunctionsPredictMultitaskingPerformance–ImplicationsfortheCentralAttentionalBottleneck AndréJ.Szameitat*andBrunelStudents CenterforCognitiveNeuroscience(CCN),DivisionofPsychology,DepartmentofLifeSciences,CollegeofHealthandLifeSciences,BrunelUniversityLondon,Uxbridge,UnitedKingdom Humanmultitaskingsuffersfromacentralattentionalbottleneckpreventingparallelperformanceofcentralmentaloperations,leadingtoprofounddefermentsintaskperformance.Whilepreviousresearchassumedthatthedefermentiscausedbyamerewaitingtime(refractoryperiod),weshowthatthebottleneckrequiresexecutivefunctions(EF;activeschedulingaccount)accountingforaprofoundpartofthedeferment.ThreeparticipantgroupswithEFimpairments(dyslexics,highlyneurotics,deprivedsmokers)showedworsemultitaskingperformancethanrespectivecontrolgroups.ThreefurthergroupswithEFimprovements(video-gamers,bilinguals,coffeeconsumers)showedimprovedmultitasking.Finally,threegroupsperformedadual-taskanddifferentmeasuresofEF(readingspan,rotationspan,symmetryspan)andshowedsignificantcorrelationsbetweenmultitaskingperformanceandworkingmemorycapacity.DemandsonEFduringmultitaskingmaycausemoreerrors,mentalfatigueandstress,withpartsofthepopulationbeingconsiderablymorepronetothis. Introduction Multitasking,i.e.,theconcurrentperformanceoftwotasks,typicallyresultsinseverecostsinformofprolongedresponsetimes(RTs)andincreasederrorrates(Pashler,1994).Onesourceofsuchcostsisacentralattentionalbottleneckpreventingtheparallelprocessingofcertainmentaloperations.Asaconsequence,thetaskscompetefortheprocessingbythebottleneck,resultingininterference.Inaddition,theprocessingofonetaskhastowaituntiltheothertaskhasbeenprocessedbythebottleneckmechanism,theso-calledrefractoryperiod(Telford,1931;Broadbent,1958).Thisbottleneckseemsvirtuallyimmutableandisaffectingalargenumberofmentalprocessescentraltohumancognition,suchasdecisionmakingormemoryretrieval(CarrierandPashler,1995;Duxetal.,2006;Tombuetal.,2011).Evenbasicprocesses,suchaspressingabuttoninresponsetoastimulus(speededchoice-responsetasks)candeferasecondresponsebyhalfasecondormore(paradigmofthePsychologicalRefractoryPeriod,PRP,Figure1;alsoseeSupplementaryFigure2).Whendrivingacarat30mph,thiswouldincreasethebrakedistancebyapprox.7m,illustratingtheseverityoftheeffect.Whileinpopularmodelsofbottleneckprocessingtherefractoryperiodisthesolesourceofmultitaskingcosts,afurthersourcehasbeensuggested1. FIGURE1 Figure1.IllustrationofthetrialdesignofthePRPdual-task(studies1and2)andthethreecomplexworkingmemoryspantasks(study2).InthePRPdual-taskparticipantsrespondedusingbothhandsonacomputerkeyboard.Assoonasthenumberofrequiredresponseswasgiven,theresponseregistrationperiodwasterminated.Insomestudies,thenumbers“1”or“2”insteadofmaleorfemalefaceswereusedasvisualstimuli.Inthespantasks(Fosteretal.,2015),responsesweregivenusingacomputermouse.Intheprocessingtask,eachscreenshowingthetask(depictedabove)wasfollowedbyanotherscreenregisteringtheanswer(notshownabove,presentingaquestionsuchas“Aretheseshapessymmetrical?”withtwoanswerboxes,“Yes”or“No,”tobeclickedwiththemouse). Thispotentialsecondsourceofmultitaskingcostsisrelatedtothewiderimplicationsthepresenceofabottleneckmighthaveontaskprocessing.Morespecifically,ithasbeensuggestedthattheinterferencebetweenthetasks,whicharisesfromthepresenceofthebottleneck,isresolvedbyexecutivefunctions(EF),suchastheinhibitionofonetaskwhiletheotherisprocessedandtheswitchingofthebottleneckbetweenthetasks(DeJong,1995;MeyerandKieras,1997;LoganandGordon,2001;Szameitatetal.,2002,2006;LuriaandMeiran,2003;Schubert,2008;WuandLiu,2008;Kochetal.,2018).DemandsonsuchEFmightprovideafurthersourceofmultitaskingcosts.Inthecontextofthecentralattentionalbottleneck,thistheoryhasbeentermedtheactive-schedulingaccount.However,otherauthorssuggestedthatEFarenotneededtocoordinatetheprocessingofthetasksatthestageofthebottleneck(Jiangetal.,2004;Lehleetal.,2009;Hartleyetal.,2012).Instead,theprocessingwouldworkpurelyonafirst-comefirst-servedbasis,atheorywhichhasbeencalledthepassivequeuingaccount. Anumberofstudieshavefoundevidencesupportingtheactive-schedulingaccount.OneofthemostcommonlyusedtaskstoassessbottleneckprocessingisthePRPtask,inwhichparticipantsarepresentedwithtwostimuli,S1andS2,whichbothrequireaspeededchoice-response,R1andR2(withtheircorrespondingresponsetimesRT1andRT2).S1andS2areeitherpresentedwithashorttemporaloffset(stimulusonsetasynchrony,SOA)orsimultaneously(SOA0ms).Participantsusuallyhavetorespondtothestimuliinacertainorder,whichimpliesthatthebottleneckhastoprocessthetasksinthisorderaswell.Evidenceforactiveschedulingwasprovided,forexample,byHirschetal.(2017)whoshowedthatthetaskpairsconstitutingaPRPdual-taskareorganizedonahierarchicallyhigherlevel,whichseemsinconsistentwithaplainpassivequeuingaccount(seealsoDeJong,1995).Supportforsuchhigher-levelcoordinationhasalsobeenprovidedbyStrobachetal.(2021b)whoshowedthatthetimeittakestoadjusttheprocessingorderofthecomponenttasksdependsonthenatureoftheprecedingtrial.Strobachetal.(2021a)alsoshowedthatparticipantscanactivelyprepareaspecificprocessingorderofthetasks(cf.alsoDeJong,1995;LuriaandMeiran,2003),againafindinginconsistentwithamerepassivefirst-comefirst-serveapproach.Somewhatindirectevidenceforactiveschedulingindual-taskssufferingfromabottleneckhasbeenprovidedbyHirschetal.(2018)whoshowedthattheamountofmultitaskingcostsinaPRPtaskcorrelateswithtaskswitchingcosts.Importantly,task-switchingcostshavebeenlinkedtocognitivecontrolprocessesmanagingthetasksets(Monsell,2003).Finally,ithasbeensuggestedthatonlydual-taskpractice,butnotsingle-taskpractice,trainsEF,suchasswitchingoperations,whicheventransfertomodifieddual-tasks(Liepeltetal.,2011;Strobach,2020). Theactiveschedulingapproachisalsosupportedbyneuroimagingdata(MaroisandIvanoff,2005).Forexample,severalstudiesemployingfunctionalmagneticresonanceimaging(fMRI)haveshownthatperformanceofaPRPdual-taskactivateslateral-prefrontalcortices(lPFC)over-additivelymorethanpredictedbythesumofthecomponenttasks(Sayliketal.,inpress;Szameitatetal.,2002;SchubertandSzameitat,2003;Stelzeletal.,2009),andthatthelPFCisstrongeractivatediftheorderinwhichtheparticipantshavetorespondtothetaskswitchesfromonetrialtothenextascomparedtowhenitstaysconstant(Szameitatetal.,2006;Stelzeletal.,2008).Furthermore,Stelzeletal.(2008)wereabletodemonstratethatthelPFCareasactivatedbydual-taskcoordinationcanbedifferentiatedfromthelPFCareaslinkedtothemaintenanceofthetask-set[seealsoStrobachetal.(2021a)forasimilarconclusionbasedonbehavioralfindings].TocomplementthefindingsfromfMRIstudies,Kübleretal.(2019)haveshownthatdisruptingthefunctionofthelPFCbymeansoftranscranialmagneticstimulation(TMS)alsodisruptsspecificallythecoordinationoftheprocessingorderofthetasks.ThefactthatthelPFCareasactivatedbyPRPdual-tasksaregenerallyassociatedwithEFsandtask-coordinationisinsupportoftheactiveschedulingaccount. However,thealternativeassumptionofthepassivequeuingaccount,whichstatesthatthetasksaresimplyprocessedonafirst-comefirst-servedbasiswithoutanyadditionalcoordination,hasalsoreceivedempiricalsupport.Onekeysupportforthefirst-comefirst-servedapproachcomesfromstudieswhichshowthatthetimesatwhichthestimuliarriveatthebottleneckdeterminetheorderinwhichthebottleneckprocessesthem(SigmanandDehaene,2006;Hendrichetal.,2012;Strobachetal.,2018).However,thefindingthatsuchcentralarrivaltimescandetermineoraffectprocessingorderdoesnotruleoutthatunderdifferenttaskconditionstheprocessingordermaybedeterminedbyhierarchicallyhighercontrolprocesses(DeJong,1995;LuriaandMeiran,2003;Leonhardetal.,2011;Strobachetal.,2021a)andthatEFsarerequiredforotheraspectsbesidestask-ordercoordination,suchasinhibitionofthesecondtasktonotinterferewiththefirsttaskorswitchingthebottleneckbetweentasks.Furtherevidenceforthepassivequeuingaccountcomesfromtheobservationthatincreasingoverlapofthetasks(i.e.,shorterSOA),whichcanbepostulatedtoincreasethedemandsonEFs,doesnotresultintheexpectedelectrodermalresponses(Hartleyetal.,2012).Finally,inanfMRIstudyJiangetal.(2004)wereunabletoobserveanybrainactivationlinkedtodual-taskprocessing,againsupportingthepassivequeuingapproach. Thecurrentstudyaimedatprovidingfurtherevidencetoresolvethecontroversybetweenactiveschedulingandpassivequeuingapproachesbytestingwhetherinter-individualdifferencesinEFcapabilitiesaffectmultitaskingperformance. ExperimentSeries1 Weaimedtoresolvethedebateaboutpassivequeuingvs.activeschedulingbyassessingthemultitaskingperformanceofpopulationsofparticipantswhoareknowntohaveeitherimpaired(dyslexics;highlyneurotics;nicotinedeprivedsmokers)orimproved(video-gamers;bilinguals;coffeeconsumers)EFcapabilitiesascomparedtorespectivecontrolgroups(non-dyslexics;lowneurotics;non-deprivedsmokers;non-gamers;monolinguals;caffeineabstinence).IfmultitaskingindeeddemandsEF(activescheduling),themultitaskingperformanceofthesegroupsshouldalsobeimpairedorimproved,respectively.Conversely,ifEFarenotrelevanttomultitasking(passivequeuing),thenallgroupsshouldshowcomparablemultitaskingperformance.Tocontrolforgenericallydifferentperformancelevelsoftheexperimentalandcontrolgroups,wecomparedthegroupswithrespecttomultitaskingcosts,i.e.,therelativeslowinginthemultitaskingascomparedtothesingle-taskperformance.Thefindingswillhelptoresolvethelong-standingdebatewhetherthepresenceofacentralattentionalbottleneckdemandsadditionalEFornot.TheywillalsoallowafirstinsightintothequestionwhethertheinvolvementofsuchadditionalEFcosttime,i.e.,contributetothedefermentofthesecondresponse.Finally,thefindingswillallowinsightsintohowpersoncharacteristics,i.e.,inter-individualdifferencesinexecutivefunctioncapabilities,affecttheperformanceinmultitaskingsituationssufferingfromacentralattentionalbottleneck. OverviewoftheGroups Wecomparedsixgroupsofparticipantswithknownmodulations,i.e.,impairmentsorimprovements,oftheirexecutivefunctioncapabilitieswithcontrols.Notably,themanipulationsinexecutivefunctioncapabilitieswerecausedbyavarietyoffactors,suchaschilddevelopment(dyslexia),personality(neuroticism),substancewithdrawal(nicotinedeprivation),training(videogamingexperience),plasticity(bilingualism),andsubstanceuse(caffeine).Whileforasinglegroupitisconceivabletoexplainpotentialeffectsonmultitaskingperformancebyalternativefactorsbesidesexecutivefunctions,wearguethatexecutivefunctionsaretheonlycommonfactoracrossallthesesixhighlydifferentgroups(butalsoseesections“ExperimentSeries2”and“GeneralDiscussion,”whichaddressthispointaswell). Inthefollowingwewilldiscussevidenceabouttheaboveproposedmodulationofexecutivefunctioncapabilitiesinmoredetail.Whilethekeydeficitsindyslexia,aneurodevelopmentaldisorderaffectingroughly5%ofthepopulation,arereadingproblems,ithasbeensuggestedthatdyslexicsalsoshowimpairmentsinEFwhichcontributetothereadingproblems(Brosnanetal.,2002).Forexample,Brosnanetal.(2002)showeddeficitsininhibitionandsequencingnotonlyinchildren,butalsoinuniversitystudents.Smith-SparkandFisk(2007)havefurthershownthatdyslexicadultsperformpoorerthannon-dyslexiccontrolsincomplexworkingmemoryspantasks,whichareagoodindicatorofEFcapabilities(cf.ExperimentalSeries2below).Inarecentmeta-analysisof26studies,Lonerganetal.(2019)foundstrongevidenceforprofounddeficitsininhibition,switchingandauditoryworkingmemoryinchildrensufferingfromdyslexia.Takentogether,thereisstrongevidencethatdyslexiaresultsinadeficitofEF.Inthepresentstudy,wecompareddyslexicparticipants,whoself-reportedhavingreceivedanofficialdiagnosis,withnon-dyslexicparticipants(seeSupplementaryInformationforamoredetailedcharacterizationofallparticipantgroups). Highlevelsofneuroticismhavebeenassociatedwithexperiencingmorestressandanxiousness(EysenckandEysenck,1986),impairingcognitiveperformance(Osorioetal.,2003).Inparticular,impairmentshavebeensuggestedinmoredifficulttasksdemandingexecutivefunctions(Studer-Luethietal.,2012).Forexample,Szameitatetal.(2016)showedthathighneuroticismisassociatedwithincreasedmultitaskingcosts,whichwasaccompaniedbyloweractivationofEF-relatedareasinthelPFC.Thisisinlinewiththefindingthatlesionstothelateral-prefrontalcorteximpairEFandatthesametimeincreasetherisktodevelopsymptomsofhighneuroticism(Forbesetal.,2014).Sayliketal.(2018)alsoshowedthathighneuroticswereimpairedonlyintasksdemandingswitchingand/orinhibition,butnotinataskdemandingthevisuo-spatialsketchpad.Therefore,weconcludethathighlevelsofneuroticismareassociatedwithlowerEFcapabilities.Inthepresentstudy,wecomparedextremegroupsofhighneurotics(meanEysenckPersonalityQuestionnairescore18,range16–24)withlowneurotics(meanscore3.89,range0–6). Nicotine,typicallyconsumedbysmokingcigarettesore-cigarettes,hasbeenshowntoimprovecognitiveperformanceinavarietyofcognitivedomains,includingworkingmemory,insmokersaswellasnon-smokers(Heishmanetal.,2010).Nicotineabstinenceinsmokers,ontheotherhand,produceswithdrawalsymptomswhichincludetheimpairmentofcognitivefunctions(Ashareetal.,2014;andnicotineconsumptionalleviatestheseimpairmentsagain:Atzorietal.,2008).Asfordyslexiaandhighneuroticism,alsotheeffectsofnicotinedeprivationappeartoparticularlyaffectEF,suchasinhibition,workingmemory,andtaskswitching(Dawkinsetal.,2007;Harrisonetal.,2009;Jansarietal.,2013),althoughtheeffectsmaynotalwaysbefullyconsistent(ButlerandLeFoll,2019).Therefore,weconcludethat,overall,thereisgoodevidencethatnicotinedeprivationimpairsEF.Inthepresentstudy,wecomparednicotinedeprivedsmokers(regularsmokerswhoabstainedforatleast50min;self-ratednicotinedeprivation8.5onascaleof1–10)withsmokerswhohadacigarettedirectlybeforethesession(meandeprivationscore1.9). AfterhavingdiscussedthreegroupsofparticipantswhoarelikelytoshowimpairedEF,wenowdiscussthreegroupswhichhavebeenproposedtoshowimprovedEF.Frequentplayingofvideogameshasbeensuggestedtoimprovecognitivefunctions[Granicetal.,2014;seeBootetal.,2011foramorecriticaldiscussion].Besideseffectsonvisualattention(GreenandBavelier,2003),alsopositiveeffectsonEFhaveconsistentlybeendemonstrated(Stanmoreetal.,2017).Forexample,inthestudybyColzatoetal.(2010)video-gamerswerebetterattask-switching[seeKarleetal.(2010),Colzatoetal.(2013)forexamplesthatnotallEFbenefitinthesameway].Strobachetal.(2012)suggestedthatvideo-gamers(andnon-gamersafterpractice)outperformnon-gamersindual-taskandtask-switchingparadigmsandsuggestedthatvideogamingresultsinimprovedexecutivecontrolskillsrelevantformultitasking.Theseobservationsareinlinewithameta-analysisshowingimprovementsininhibition,top-downattention,andmultitasking/task-switching(Bediouetal.,2018).Interestingly,improvementsofEFcanbeseeninallagegroups,includingpreschoolchildren(Yangetal.,2020)andtheoldestold(McCordetal.,2020).Inconclusion,itseemsverywellestablishedthatvideo-gamingexperienceimprovesexecutivefunctionabilities.Inthepresentstudy,wecomparedfrequentvideo-gamers(onaverage3.5h/dayactiongames,plus1.7h/dothergames)withoccasionalornon-gamers(0.6h/dactiongames,plus0.7h/dothergames). Coffeeisknowntohaveagenerallystimulatingeffect,includingimprovementofcognitivefunctions(Cappellettietal.,2015;McLellanetal.,2016).ConsumptionofcaffeinehasbeenshowntoimproveEFintaskssuchastheTowerofLondontask(Killgoreetal.,2014),grammaticalandlogicalreasoning(Kohleretal.,2006;Kamimorietal.,2015),theJansariAssessmentofExecutiveFunctions(Soaretal.,2016),andtheAttentionNetworkTest(Brunyéetal.,2010).However,theevidenceisslightlymixed,withsomestudiesfailingtofindeffectsonEF(e.g.,Gottseligetal.,2006),whichmaybecausedbytheprofoundvarietyinstudydesign(e.g.,whethersleep-deprivationwasinduced),caffeine-dosage,andparticipantcharacteristics(e.g.,whethertheyareregularcaffeineusers).Weconcludethat,overall,thereisgoodevidencethatcoffeeconsumptionhastheabilitytoimproveEFs.Inthepresentstudy,weaskedallparticipantstonotconsumeanycaffeinateddrinkforatleast4hbeforethestudyandcomparedparticipantswhichthendrankacupofcoffee(theexperimentstartedapprox.30minaftercoffeeconsumption)withparticipantswhodidnothaveanydrink. BilingualismhasbeenproposedtoimproveEF,mainlyduetotheconstantuseofEFtofocusattentiontothetargetlanguage,toswitchbetweenlanguages,andtoinhibitthecurrentlynon-usedlanguage(Bialystoketal.,2009).Andwhilethereisprofoundsupportforthissuggestion(vandenNoortetal.,2019;Grundy,2020),thererecentlyhavealsobeendoubtsabouttheconsistencyofthefindingthatbilingualismbenefitsEFs(e.g.,Lehtonenetal.,2018).Thecurrentlymostlikelysolutiontothisdebatemightbethattheconceptofbilingualismitselfandthepotentiallyaffectedcognitivedomainsrequireamorefine-graineddefinition(DeLucaetal.,2019;Bialystok,2021).Despitethisdebate,weincludedastudycomparingbilingualwithmonolingualparticipants.Becauseweusedaratherstrictdefinitionofbilingualismforthecurrentstudy(theabilitytospeakatleasttwolanguageswiththeproficiencyofanativetonguebytheageofsix)wehadtheassumptionthatthebilingualparticipantshaveimprovedEFascomparedtomonolingualparticipants(ignoringpotentialfurtherlanguageslearnedlaterinlife,e.g.,inschool). Eachgroupconstitutedanindependentstudywithaseparatecontrolgroup,runbyadifferentexperimenter.FiveofthesixreportedstudieswerefinalyearundergraduatedissertationprojectsandonestudywaspartofPh.D.work.AllstudieswereconductedatBrunelUniversityLondonbetween2013and2017.Thedataarebasedon7students’projectsandallprojectsweresupervisedbythefirstauthor,AS.Allstudentsworkedindependentlyandtestedindependentsamples.Twostudiesarebasedontwocombineddatasetseach,collectedindependentlybydifferentstudents.Thestudieswererunbythefollowingstudents:Dyslexia:CB;Smokingdeprivation:JB,MB;Neuroticism:RS(Ph.D.student);Video-gaming:NG,AO;Bilingualism:LS;Caffeine:AS. Inthefollowingthemethodsofallthedifferentstudiesaredescribedbriefly,formoredetailsseetheSupplementaryInformation. Methods Participants AllstudieswereindividuallyapprovedbytheDepartmentofLifeSciencesEthicscommittee,BrunelUniversityLondon,andwerecarriedoutinaccordancewiththeDeclarationofHelsinki.Participantsgavewritteninformedconsentbeforeparticipationandweredebriefedafterthestudy.Participantswereeitherreimbursedwithcoursecredits,£10,orvolunteeredforfree.VirtuallyallparticipantswerestudentsofBrunelUniversityLondonbetweentheageof18and25. Intotal274participantstookpart(Table1).Weexcludedparticipantsbasedonthreecriteria:(a)iftheyhadmorethan30%errorsinanytaskcondition,(b)iftheirdataweremorethan2.5standarddeviationsfromtherespectivegroupmean,and(c)iftheydidn’tfulfillthecriteriaforthegroup.Mostparticipantswereexcludedinthenicotinedeprivationstudy,becauseperceivednicotinecravingsweretoolowafteronly30–50minofdeprivation.AlldetailsonparticipantselectioncanbefoundintheSupplementaryInformation. TABLE1 Table1.Demographicinformation. Design Allstudieswerebasedonamixedtwo-factorialdesign.Groupisabetween-subjectfactorwiththelevelsexperimentalgroup(e.g.,dyslexics,highneurotics,etc.)andcontrolgroup(e.g.,non-dyslexics,lowneurotics,etc.).Taskisawithinsubjectfactorwiththelevelssingle-taskanddual-task(i.e.,multitasking).Theinteractionbetweenthetwofactorsisofparticularinterest,becauseitreflectswhetherthedecrementsindual-taskperformanceascomparedtosingle-taskperformanceweredifferentforthetwogroups.Thedependentvariableswereresponsetimes(RT)forthesingle-tasksandforeachofthetwotasksinthedual-task(RT1andRT2),aswellaserrorrates(reflectingwhetherthetrialasawholewaseithercorrect,orwhetheroneormorepotentialerrorsweremade). MaterialsandProcedure Weemployedthedual-taskparadigmofthepsychologicalrefractoryperiod(PRP),whichconsistedofthecombinationofavisualandanauditoryspeededchoiceresponsetask(Figure1,leftpanel).Atrialinthevisualsingle-taskstartedwiththepresentationofafixationcrossfor200ms,afterwhichthevisualstimuluswaspresentedfor345ms.Thevisualstimuluswaseitheramaleorfemaleface(studiesDyslexia,Neuroticism,Smokingdeprivation,Video-gaming,Bilingualism)orthenumbers“1”or“2”(Coffeestudy)presentedinthemiddleofthescreen.Participantshadtorespondusingtheirrighthandbypressingtheaccordingbuttonsonacomputerkeyboard(“n”formaleface/number1;“m”forfemale/2,respectively).Participantshadtorespondwithin2,655msafteronsetofstimuluspresentation.Aftertheresponsewasgivenor2,655mshadpassedsincestimulusonset,eitherablankscreenoranerrormessagewasdisplayedfor300ms,beforethenexttrialstarted.Atrialintheauditorytaskwasidenticalexceptthatinsteadofavisualstimulus,ablankscreenwasdisplayedandasoundwasplayedviaheadphonesfor345ms.Thissoundwaseitheradouble-syllable(/haha/or/yaya/;studiesDyslexia,Neuroticism,Smokingdeprivation,Video-gaming,Bilingualism)orabeep(300Hzor800Hz;Coffeestudy).Participantsrespondedusingtheirlefthand(“x”for/haha//300Hz;“c”for/yaya//800Hz).Atrialinthedual-taskwasidentical,exceptthatbothstimuliwerepresentedatthesametime(stimulus-onset-asynchrony,SOA,0ms)andparticipantshadtorespondtobothstimuliusingthesamekeymappingasinthesingle-tasks.Participantshadtorespondtothetasksinagivenorder(e.g.,firsttotheauditorytask)whichwasvariedacrossblocksandspecifiedbyaninstructionbeforeeachblock.Participantsreceivedanerrorfeedbackiftheypressedoneormorewrongkeys,andiftheypressedthecorrectkeys,butinthewrongorder.Insomestudies,additionalconditionswerepresented(e.g.,SOA1,000ms,randomtask-order),whicharenotrelevanttothecurrentreport. Eachcondition(auditorysingle-task,visualsingle-task,dual-task(responseorderauditory->visual),dual-task(visual->auditory)waspresentedintwoblocksof35trialseach.Fortheanalysis,thetwosingle-tasksaswellasthetwodual-taskswerecombined,resultingin140single-taskand140dual-tasktrials.Beforeeachcondition,aself-pacedinstructionabouttheupcomingtaskwaspresentedonthescreen,whichalsoservedastheopportunityforabreak.Orderofconditionsandtrialswereindividuallyrandomized.Beforethemainexperiment,participantspracticedalltaskconditions.Asessionlastedbetween40and60min.ThetaskwaspresentedusingPresentation2. Results Theaimoftheseexperimentswastoassesswhethergroupsofparticipantswithimpairedexecutivefunctionsshowimpairedmultitaskingperformance,andwhethergroupsofparticipantswithimprovedexecutivefunctionsshowimprovedmultitaskingperformance.Tocontrolforpotentialgenericdifferencesbetweenthegroups,wecalculated2×2–factorialmixedANOVAsforeachstudy,withthefactorsGroup[experimentalgroup(e.g.,dyslexics)vscontrolgroup(e.g.,non-dyslexics)]andTask(single-taskvsdual-task).Generally,thecostsofmultitaskingareevidentbypoorerperformance(increasedRTsanderrorrates)inthedual-taskascomparedtothesingle-task.Anexecutivefunctionimpairment/improvementspecifictomultitaskingshouldbeevidentbyadifferenceinthesemultitaskingcostsbetweenthegroups,whichisreflectedbytheinteractiontermoftheANOVA. First,weanalyzedtheresponsetimes.Inthedual-task,weusedtheresponsetimesofthesecondtask(RT2),becausethisismostsensitivetoanyformofdual-taskcosts(Pashler,1994;seealsoSupplementaryInformation).Inallsixstudies,participantsgenerallyshowedprofoundmultitaskingcostsof797ms(averagedacrossallstudiesandgroups)and,consequently,themaineffectofTaskwassignificantinallstudies(allp<0.001).Inmoststudies,therewasalsoanoveralldifferencebetweenthegroups,asreflectedbyasignificantmaineffectofGroup(allp<0.05),exceptforthestudiesonbilingualism(p=0.246)andcaffeine(p=0.066).Mostimportantly,inallsixstudiesweobservedasignificantinteractionbetweenGroupandTaskintheexpecteddirection(Figure2andTable2;allp-values<0.05).Forthegroupswhichareknowntohaveimpairedexecutivefunctions(dyslexics,highlyneurotics,nicotinedeprived),themultitaskingcostswereincreasedby23–36%ascomparedtotherespectivecontrolgroups(Table3).Ontheotherhand,forthegroupswhichareknowntohaveimprovedexecutivefunctions(video-gamers,bilinguals,coffeeconsumers),themultitaskingcostswerelowerby20–25%ascomparedtotherespectivecontrolgroups.Additionalanalyses(seeSupplementaryInformation)confirmedtheinteractionsarenotdrivenbyamereprolongation/shorteningoftheresponseselectionstagesintheexperimentalgroups.Thisillustratesthatindividualdifferencesinexecutivefunctioncapabilitiesaffectmultitaskingperformance.Italsoprovidesstrongsupportfortheactiveschedulingaccountofthecentralattentionalbottlenecktheory. FIGURE2 Figure2.ExperimentSeries1.Responsetimes(ms)forthethreeindependentgroupsofparticipantswhoareknowntohaveimpairedexecutivefunctions[upperpanel,(A)]andforthethreegroupswhoareknowntohaveimprovedexecutivefunctions[lowerpanel,(B)].Dual-taskcostsaretherelativeslowingofresponsetimesinthedual-task(RT2)ascomparedtothesingle-task.Effectsspecifictomultitaskingareevidentbyincreasedordecreaseddual-taskcosts,respectively,reflectedintheinteraction.Errorbarsshowstandarderrorofthemean(SEM). TABLE2 Table2.DescriptiveandinferentialstatisticsofresponsetimesinExperimentSeries1. TABLE3 Table3.Multitaskingcosts(dual-taskRT2—single-taskRT)forthecontrolandexperimentalgroupsinExperimentSeries1. ItisconceivablethatsomeEFdemandsoccurbeforeoratthestageoftheprocessingbottleneck(e.g.,inhibitionoftask2),whileotherEFdemandsoccurafterthebottleneckprocessingofthefirsttaskhasbeenfinished(e.g.,switchingthebottlenecktotask2,DeJong,1995;foramoredetailedargumentincludingagraphicalillustrationseeSupplementaryInformation).Toestimatethecostsbeforeoratthebottleneck,wecalculated(dual-taskRT1–single-taskRT)foreachparticipantindividuallyandthencomparedthegroups,i.e.,(dual-taskRT1–single-taskRT)Experimentalgroup–(dual-taskRT1–single-taskRT)Controlgroup,whichisequivalenttotheinteractiontermintheaboveanalyses,exceptthatnowRT1isusedinsteadofRT2.Toestimatethecostsintask2afterthebottleneckwecalculated(dual-taskRT2–dual-taskRT1)Experimentalgroup–(dual-taskRT2–dual-taskRT1)Controlgroup.Table4showsthegroupdifferencesinmultitaskingcostsforeachtypeofcostseparately.Itshowsthatabouttwothirdsoftheoverallgroupdifferencesinmultitaskingcostsoccurredbeforeoratthestageofthebottleneck,whileaboutonethirdoccurredintask2afterbottleneckprocessingintask1hasfinished.Inotherwords,resultsshowthatinter-individualdifferencesinEFcapabilitiesaffecttaskprocessingbefore,atandafterbottleneckprocessing,suggestingthatEFaredemandedatseveralstagesduringtheprocessingofadual-task,withthemajorityofdemandsoccurringbeforeoratthestageofthebottleneck. TABLE4 Table4.DetailedanalysisofmultitaskingcostsinExperimentSeries1. Toassesswhetherthegroupsalsodifferedintheirsingle-taskperformancealone,wecalculatedindependent-samplest-testscomparingthesingle-taskRTsoftheexperimentalgroupswiththoseofthecontrolgroups,individuallyforeachexperiment.Resultsshowedthatnostatisticallysignificantdifferencewasevidentforthestudiesondyslexia[forallcomparisonsGroup>Controls;78.141ms;t(24)=1.684;p=0.105;Cohen’sd=0.687],neuroticism[32.691ms;t(33)=1.310;p=0.196;Cohen’sd=0.459],bilingualism[29.054ms;t(23)=0.664;p=0.513;Cohen’sd=0.277],andcoffeeconsumption[−12.232ms;t(33)=0.520;p=0.606;Cohen’sd=0.181],whilethedifferencewassignificantinthestudyonvideo-gaming[−65.502ms;t(51)=2.651;p=0.011;Cohen’sd=0.742]andinthestudyonsmokingdeprivation[183ms;t(25)=3.046;p=0.005;Cohen’sd=1.218].Overall,thisshowsthattheexperimentalandcontrolgroupsperformedcomparablyinthesingle-taskconditionsinfourofthesixstudies.Notethatdifferencesinsingle-taskperformancecannotaccountforthefurtherabovereportedinteractioneffects. Fortheanalysesoferrorrates,trialswereclassifiedaseitheracorrecttrial,orasanerrortrial,irrespectiveofwhetherthereweremultipleerrorswithinasingletrial(whichwaspossibleindual-tasktrials).2×2-factorialANOVAsoftheerrorrates,comparabletothoseoftheresponsetimeanalysesabove,revealedeithernon-significantdifferences,ordifferencesofthesamenature,i.e.,slowerresponsetimeswereaccompaniedbyhighererrorrates.Theonlyexceptionwasthestudyonvideo-gaming,inwhichthevideo-gamersshowedsignificantlyhigherdual-taskcostsinerrorratesthanthenon-gamers.Fulldetailsoferror-rateanalysescanbefoundintheSupplementaryInformation. DiscussionExperimentSeries1 Resultsshowedthatthemultitaskingcostsvariedacrossgroupsaspredictedbytheirpresumedexecutivefunctioncapabilities,supportingtheactiveschedulingaccount(MaroisandIvanoff,2005;SigmanandDehaene,2005). Forsomeofthegroupswecompared,thereisanongoingdiscussionintheliteraturewhetheritisindeedtheproposedunderlyingfactorwhichcausesdifferencesintheEFcapabilities,orwhetheritcanbeexplainedotherwise.Forexample,itmightbethatfrequentvideo-gamersdifferinanumberofcharacteristicsfromnon-gamers,andthatitisnotonlyvideo-gamingexperience,butalsosomeofthesefurthercharacteristicswhichmayexplainpartofthedifferencesinEFcapabilities.WhileBediouetal.(2018)intheirmeta-analysisindeedprovidedevidenceforthis,theyatthesametimeshowedthatsuchadditionalcharacteristicscannotfullyexplainthedifferencebetweengamersandnon-gamers.However,importantly,thisisnotamajorissueforthepresentstudy,becausehereitisonlyrelevantthatthegroupsdifferintheirEFcapabilities,butnotwhyexactlytheydiffer.Thus,thepresentstudycan(tosomeextent)remainoblivioustothein-depthdiscussionsaboutthetruenatureoftheunderlyingmechanismsofEFdifferences. However,groupcomparisonsbearthenaturallimitationthattheremightbefurtherdifferencesbesidesthepostulateddifferencesinexecutivefunctioncapabilities.Whilethisislikelytobethecasewhenlookingatanindividualgroup,webelievethattheonlycommondifferenceacrossallsixgroupsisadifferenceinexecutivefunctions(seesection“GeneralDiscussion”foramorein-depthdiscussion).Nevertheless,toruleoutthepossibilitythatthecurrentfindingsareexplainedbyapotentialfurtherfactorcommontoallgroups,ExperimentSeries2wasconducted,whichwasnotbasedonbetween-groupcomparisons,butinsteadoncorrelationalwithin-subjectdesigns. ExperimentSeries2 ExperimentSeries2aimedatassessingtheEFcapabilityofeachparticipantindividuallyandcorrelatingitwiththeirmultitaskingability.ToassessEFcapabilities,weusedcomplexworkingmemoryspantasks(Redicketal.,2012).Complexworkingmemoryspantasksweredesignedtoassesstheso-calledworkingmemorycapacity(WMC).Theoreticalmodels(Kaneetal.,2007)aswellasempiricaldata(McCabeetal.,2010)suggestthatWMCiscloselyrelated,ifnotidentical,totheconceptofexecutivefunctions.Inmoredetail,Kaneetal.(2007)consideredthecentralexecutive,workingmemorycapacity,executiveattention,andsupervisoryattentionsystemassynonymous.Insupportofthisassumption,McCabeetal.(2010)measuredworkingmemorycapacity(fourdifferentcomplexworkingmemoryspantasks)andexecutivefunctions(WisconsinCardSortingTest,verbalfluency,mentalarithmetic,mentalcontrol),andfoundthatthetworesultingfactorsinafactoranalysiscorrelatedextremelyhighwitheachother(r=0.97).Theauthorsconcludethatboth,workingmemorycapacityandexecutivefunctions,shareanunderlyingattentionalability,whichtheytermexecutiveattention(EngleandKane,2004;McCabeetal.,2010).Therefore,weassumethatcomplexworkingmemoryspantasksassessexecutivefunctions. Workingmemorycapacityisassessedbycomplexworkingmemoryspantasks,suchasthereadingspantask,inwhichaprocessingtask(e.g.,determinethecorrectnessofasentence)isintermingledwithashort-termmemorytask(e.g.,memorizingletters).Tocreateareliablepicturewhichisindependentofaspecificdomain,weusedthreedifferentversionsofthistask:(a)theclassicreadingspantask,relyingonlinguisticprocessing,(b)thesymmetryspantask,relyingonvisualpattern-recognitionprocessing,and(c)therotationspantask,relyingonspatialprocessing(DanemanandCarpenter,1980;Fosteretal.,2015).Multitaskingabilitywasagainassessedinformofmultitaskingcosts,i.e.,performanceinthedual-taskminusperformanceinthesingle-tasks. Ifthepassivequeuingaccountholds,whichproposesthatthepresenceofabottleneckdoesnotdemandadditionalexecutivefunctions,thenworkingmemorycapacityshouldbeindependentofmultitaskingcosts,i.e.,nocorrelationisexpected.However,iftheactiveschedulingaccountholds,thenhigherworkingmemorycapacityshouldresultinmoreefficientprocessinginthemultitaskatthestageofthebottleneckand,consequently,thebehavioralmultitaskingcostsshouldbelower,i.e.,anegativecorrelationbetweenspanscoresandmultitaskingcostsisexpected. Methods OverviewofTasks AllthreereportedexperimentswerefinalyearundergraduatedissertationprojectsconductedatBrunelUniversityLondonbetween2013and2017,supervisedbythefirstauthor,AJS.Allstudentsworkedindependentlyandtestedindependentsamples.Thestudieswererunbythefollowingstudents:ReadingSpan:KT;SymmetrySpan:KK;RotationSpan:BS. Participants AllstudieswereindividuallyapprovedbytheDepartmentofLifeSciencesEthicscommittee,BrunelUniversityLondon,andwerecarriedoutinaccordancewiththeDeclarationofHelsinki.Participantsgavewritteninformedconsentbeforeparticipationandweredebriefedafterthestudy.Participantswereeitherreimbursedwithcoursecredits,£10,orvolunteeredforfree.VirtuallyallparticipantswerestudentsofBrunelUniversityLondonbetweentheageof18and25. Intotal85participantstookpart(Table5).WeexcludedparticipantsasinExperimentalSeries1,whichresultedintheexclusionof4participants.MoredetailsonparticipantselectioncanbefoundintheSupplementaryInformation. TABLE5 Table5.DemographicinformationforExperimentSeries2. Design AllstudiesinExperimentSeries2werebasedonacorrelationaldesigninwhicheachparticipantperformedthedual-taskaswellasonetypeofcomplexworkingmemoryspantask.Theoutcomevariablesofthemulti-taskwerethemultitaskingcostscalculatedas(a)dual-taskRT2minussingle-taskRT,and(b)error-ratedual-taskminuserror-ratesingle-task.Theoutcomevariablesinthedifferentspantasksweretherespectivepartialscores,i.e.,thenumberofitemscorrectlyrecalledinserialposition. MaterialsandProcedure ThePRPdual-taskwasidenticaltoExperimentSeries1,exceptthat,toshortentheexperiment,thedual-taskwasperformedonlyinonetaskorder(firstauditorytask,thenvisualtask).Thevisualstimuliwerealwaysthenumbers1and2,andtheauditorystimulialwaysthelow-andhigh-pitchedtones(seeMethodsExperimentSeries1).Asession,includingthedual-taskandworkingmemoryspantask,lasted60–70min. Inallcomplexworkingmemoryspantasks,aso-calledprocessingtaskwasinterwovenwithamemorytask(Figure1,rightpanel).Thenatureofthetasksvariedacrosstheexperiments,andfollowedtheprocedureasdescribedinFosteretal.(2015).Inthesymmetryspantask,participantshadtodecidewhetherapatternofblackboxesinamatrixissymmetricalarounditsverticalaxisornot(processingtask)andtomemorizethepositionofaredsquareina4×4matrix(memorytask).Intherotationspantask,participantshadtodecidewhetherarotatedletterisinitsnormalformorwhetheritismirrorreversed(processingtask)andtomemorizeanarrowwhichcouldbeeithershortorlongandpointtooneofeightdirections(memorytask).Inthereadingspantask,participantshadtoreadsentencesanddecidewhethertheymadesenseornot(processingtask)andtomemorizeletters(memorytask).Onememorytrialalwaysstartedwiththepresentationofoneprocessingtask(e.g.,onesentenceinthereadingspantask).Aftertheprocessingtaskhasbeenshown(e.g.,asentence),anotherscreenappeared(notshowninFigure1)onwhichparticipantsgavetheirresponsetotheprocessingtaskusingthecomputermouse.Forthis,aquestionwasasked(e.g.,“Doesthissentencemakesense?”)withtwoclickableanswerboxes(YesandNo).Aftertheresponsetotheprocessingtask(oratime-outerrormessage),oneitemtoberememberedwaspresented(e.g.,oneletterinthereadingspantask).Thiscyclewasrepeatedrandomlyanywherefromtwouptoseventimes(dependingonthetask).Attheendofonememorytrial,participantshadtorecallallitemstoremember(e.g.,letters)intheorderoftheirpresentation.Forthis,theyusedthecomputermouseandclickedeitherontheappropriatesquaresina4×4matrix(symmetryspantask),orontheappropriatearrowsinadisplayofall16arrows(shortandlongin8directions;rotationspantask),orontheappropriateletterinamatrixofallpotentialletters(readingspantask).Toavoidthatparticipantstradedoffperformanceintheprocessingtasktorehearsethememoryitems,theywereaskedtomaintainaccuracyintheprocessingtaskabove80%andhadanindividuallyadjustedtimewindowinwhichtheresponsehadtobegivenbeforeatime-outerrorwaspresented. Results Theaimoftheseexperimentswastoshowthatindividualdifferencesinexecutivefunctioncapabilitiesareassociatedwithmultitaskingabilitiesonasubject-by-subjectlevel.Forthis,wecorrelateddual-taskcosts(responsetimeofthesecondtaskinthedual-task(RT2)minussingle-taskRT)withdifferentmeasuresofworkingmemorycapacitywhichreflectexecutivefunctioncapabilities.Whilethepassivequeuingaccountpredictsnocorrelations,theactiveschedulingaccountpredictsanegativecorrelation,i.e.,thehighertheworkingmemoryspanthelowerthedual-taskcosts.Duetothisstronglydirectionalhypothesis(tothebestofourknowledge,thereexistsnomodelpredictingapositivecorrelation),weappliedone-sidedsignificancetesting. Results(Figure3)showedthatinallthreestudiesdual-taskcostsintermsofresponsetimesweresignificantlynegativelycorrelatedwiththerespectivespanscores(Rotationspan:N=27,Pearson’sr=−0.525,p=0.003,R2=0.276;Symmetryspan:N=20,Pearson’sr=−0.379,p=0.049,R2<0.144;Readingspan:N=34,Pearson’sr=−0.535,p<0.001,R2=0.286).Althoughwealreadycorrelatedthedual-taskcosts,whicharearelativemeasurenotdirectlydependentongenericprocessingspeedofanindividual,withthepartialspanscore,weconfirmedaboveanalysesbypartialingoutthesingle-taskperformanceasameasureofprocessingspeed.Resultsshowedthesamepattern(allp<0.05;allPearson’sr