sequencesHazeltine2002HomannSebaldStoecker2001Ziessler1998ZiesslerNattkemper20012002ElsnerandHommel2001seealsoHommel19971998haveproposedatwostagemodelofgoaldirectedactionthatisth ID: 247754
Download Pdf The PPT/PDF document "ORIGINALARTICLEBirgitElsnerBernhardHomme..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
ORIGINALARTICLEBirgitElsnerBernhardHommelContiguityandcontingencyinaction-effectlearningAccepted:11August2003/Publishedonline:18December2003Springer-Verlag2003AbstractAccordingtothetwo-stagemodelofvoluntaryaction,theabilitytoperformvoluntaryactionisac-quiredintwosequentialsteps.Firstly,associationsare sequences(Hazeltine,2002;Homann,Sebald&Stoecker,2001;Ziessler,1998;Ziessler&Nattkemper,2001,2002).ElsnerandHommel(2001;seealsoHommel,1997,1998)haveproposedatwo-stagemodelofgoal-directedactionthatisthoughttocapturethetransitionfromtheacquisitionofknowledgeaboutaction-eectdependen-ciestotheemergenceofintentionalactioncontrol.Stage1ofthemodelanditisthisstagethepresentstudyfocusedonisconcernedwiththeacquisitionofregu-laritiesbetweenmovementsandeects.Firstly,itisas-sumedthat,whenamovementiscarriedout,allsensoryeventsaccompanyingitareregisteredandcodedwithinthecognitivesystem.Secondly,ifaparticularmovementandaparticularsensoryeventco-occurrepeatedlysothattheircognitivecodesarefrequentlyco-acti-vatedthisleadstotheautomaticintegrationofthesecodes,i.e.,toabi-directionalassociationofmovement-relatedandeect-relatedcodes.Accordingly,activatingoneoftheintegratedcodesonlateroccasionswilltendtoactivatetheotherassociatedcodestoo.Thisbi-directionalactivationisassumedtobeimportantforStage2ofthemodel,whichdealswithhowactionsarerecruitedtoachievedesiredgoalsbyanticipatingthemovementseects.Thinkingofagoalisclaimedtoactivatethecodesofitsperceptualcharacteristics.Iftheseperceptualcodeshavealreadybecomeintegratedwiththecodesofamovement,activatingthecodesofthegoalwillspreadactivationtotherespectivemotorcodesandtherebyprimethemovementthathaspro-ducedthedesiredeectsinthepast.Thisallowstheselectionofactionsbyanticipating(i.e.,activatingthecodesof)desiredactioneects.Firstevidencesupportingthismodelwasprovidedbylearningstudies(Elsner&Hommel,2001;seealsoHommel,1993,1996)inwhichparticipantsrstworkedthroughalearningphasebyperformingkeypressesthatwerefollowedbycertaintones.Forinstance,oneactionproducedalow-pitchedtone(A1low)whileanotheractionproducedahigh-pitchedtone(A2high).Inasubsequenttestphase,participantsperformedthesameactions,butnowinresponsetothesametones,whichwerepresentedasimperativestimuli.Itturnedoutthatanactionwasperformedfasterinresponsetoatonethatithadpreviouslyproduced(lowA1,highthantoatonethathadbeenproducedbythealternativeaction(lowA2,highA1).Hence,anacquisition-consistentmappingoftoneandactioninthetestphaseallowedforbetterperformancethananacquisition-inconsistentmapping.Thisobservationhastwoimpor-tantimplications.Firstly,asthetonesinElsnerandHommels(2001)learningphasewereirrelevanttothetask,producinganoveleectbyperforminganactionautomaticallycreatesanassociationbetweenthecodesoftheactionandthecodesoftheeect.And,secondly,thisassociationmustbebi-directional,aseecttonesprimedthecorrespondingactionseventhoughthesequenceofeventsinthetestphasereversedthesequenceofactionandtoneexperiencedinthelearningphase.Accordingly,perceivingalearnedactioneectleadstoabackwardactivationoftheassociatedmovement.Thepresentstudywascarriedouttoshedsomemorelightontheconditionsthatsupporttheacquisitionofaction-eectknowledge.AccordingtoElsnerandHommels(2001)model,suchknowledgeisaccumulatedbyassociativelearningmechanisms.Ifso,action-eectlearningwouldbeexpectedtobeinuencedbyfactorsthatareknowntoaectassociativeinstrumentallearn-ing.SomeofthesefactorswerealreadydiscussedbyHume(1739/1964),whosoughttogureoutwhichas-pectsofarelationbetweentwogiveneventsleadustoassumethatoneeventwascausingtheother.Humefavoritecuestocausalitywere,amongothers,thecontiguity(i.e.,temporalandspatialproximity)andthecontingency(i.e.,constantco-variation)ofevents.In-deed,laterempiricalstudieshavegatheredconsiderableevidencethatthesefactorsarenotonlycrucialforassociativelearninginanimalsandhumans,buttheyalsoinuencehumanperceptionofcausalitybetweenmovementsandtheirconsequences(seeoverviewsbyShanks&Dickinson,1987;Wasserman,1990;Young,1995).Ifso,andifaction-eectacquisitionreallyisaresultofassociativelearningmechanisms,theeaseofacquiringaction-eectassociationswouldbeexpectedtodependonthecontiguityandthecontingencybetweenactionsandeectsduringthelearningexperience.Wetestedthisexpectationintwoexperimentsmod-eledafterElsnerandHommels(2001)experiments.InthelearningphaseofExperiment1,thetemporaldelaybetweenkeypressandtone(i.e.,temporalcontiguitywasvaried,whileinthelearningphaseofExperiment2,thecovariationofkeypressandtone(i.e.,contingencywasmanipulated.Thetestphasewasalwaysthesame.Theformeraction-eecttoneswerepresentedasimperativestimuli,which,inseparateblocks,weremappedontokeypressresponsesinanacquisition-consistentoracquisition-inconsistentfashion.Ifaction-eectlearningdependsoncontiguityandcontingency,theexpectedperformancedierencebetweenacquisi-tion-consistentandacquisition-inconsistenttestblocksshouldincreasewithincreasingtemporalproximityandincreasingco-variationofactionandeect,i.e.,wetakeperformancedierencesbetweenacquisition-consistentand-inconsistenttestresponses(i.e.,theacquisition-consistencyeect)toindicatethestrengthofthelearnedaction-eectassociation.Experiment1:ContiguitySincethebeginningofthesystematicstudyofanimallearning,ithasgenerallybeenacceptedthatthetem-poralcontiguitybetweentheconditionedandtheunconditionedstimulusisacriticalfactorinclassicalconditioning(Pavlov,1927).Later,Grices(1948)studyestablishedthatforinstrumentallearning,thetemporalcontiguitybetweenresponseandreinforcementisalsocritical.Typically,learningperformancedecreaseswith anincreasingintervalbetweenresponseandreinforce-ment,i.e.,withdecreasingcontiguity.Theperformancedeclineiscommonlyexplainedbytheprocessoftraceconditioning(Pavlov,1927),whichimpliesthatthementaltraceoftherstevent(i.e.,theresponseininstrumentallearning)extendsbeyondthateventsac-tualpresenceintheenvironment,butissubjecttocon-tinuousdecay.Ifthesecondevent(i.e.,thereinforceroreect)occursatatimewhenthersttraceisalreadyweak,thelearnedassociationwillalsobeweak.Ifthedelaybecomestoolong,theactivationsofthetwoeventswillnolongeroverlapandnoassociationswillbeformedatall(Grice,1948;Pavlov,1927).Recently,severalresearchershavedemonstratedthatcontiguityaectsconditionedrespondinginanimalsandratingsofcausaleectivenessinhumansinacompara-blefashion,andthus,bothphenomenamayrelyonthesamemechanisms(e.g.,Allan,1993;Shanks&Dick-inson,1987;Wasserman,1990).Inanimals,delayingreinforcementreduceslevelsofconditionedresponding(e.g.,Reed&Reilly,1990),andinhumans,delayingtheoutcomeofamovementreducestheperceivedcausaleectivenessofthatmovement.Forinstance,ifpartici-pantsareaskedtoestimatetheextenttowhichpressingakeycausestheashingofatriangle,causalityjudg-mentsmatchtheactualdependencyquiteaccuratelywhenmovementandeectareseparatedbydelaysof0 2s,butjudgmentsbecomeincreasinglyinaccuratewithlongerdelays(Shanks,Pearson,&Dickinson,1989).Thus,theeectiveaction-eectintegrationwindowforsimplekeypressesseemstoberathershort.Underconditionswithlongmovement-eectdelays,onemethodofimprovinglearningisllingthetemporalgapwithirrelevanteventsthatarecontiguouswithboththemovementandtheoutcome(seeEinhorn&Hogarth,1986;Gruber,Fink,&Damm,1957).Inani-malstudies,ReedandReilly(1990)demonstratedthatastimuluspresentedduringaresponse-reinforcerdelayamelioratesthedecitinconditionedrespondingthatthedelaywouldotherwiseproduce.Likewise,studiesofhumanjudgmentshaveshownthatasignalpresentedbetweenmovementandoutcomewillrelievethedelay-induceddecitinratingsofcausaleectiveness(Reed,1992,1999;Shanks,1989).KaplanandHearst(1982)andRescorla(1982)demonstratedthataninterveningstimulus(IS)ismosteectivewhenitllstheintervalbetweenthetwoeventscompletely.TheinuenceoftheISistypicallyexplainedbyconditionedreinforcement(Grice,1948;Reed,1999).ItisassumedthatduetothetemporalcontiguityofISandeect(E),theparticipantslearnanIS-Eassociation.BecausetheISisalsocon-tiguoustotheresponse(R),anadditionalR-ISassoci-ationmaybeformed.Ifthisoccurs,theresponseacquirescausalecacywithrespecttotheoutcomeviatheR-IS-Eassociationchain.AnalternativeexplanationfortheinuenceoftheISisthatithelpstodiscriminatethemovement-eectintervalfromthesubsequentin-tertrialinterval(ITI).Thisassumptionissupportedbythefactthattheacquisitionofresponse-outcomeasso-ciationsisnotonlyfacilitatedbyanISpresentedintheresponse-eectinterval,butalsobyanISpresentedintheITI(Kaplan&Hearst,1982).Ifweassumethattheacquisitionofnovelactionef-fectsasinvestigatedbyElsnerandHommel(2001)reliesonassociativelearningmechanisms,wewouldexpecttheeciencyorlikelihoodofacquiringthoseeectstobesensitivetothesamemanipulationsthatweredemon-stratedtohaveanimpactonmorestandardmeasuresofresponse-outcomelearning.Experiment1testedwhetherthisisthecaseforthetemporalcontiguitybetweenmovementandeect.Inthelearningphase,participantswereaskedtorespondtoanimperativestimuluswithoneoffourkeypresses(Fig.1).Aftereachkeypress,theparticipantsheardacertaintone,sotheywereexpectedtoacquirefouraction-eectassociations.Inveexperi-mentalgroups,temporalcontiguitywasvariedbymanipulatingthetemporaldelaybetweenkeypressandtone,whichshouldaectaction-eectlearning.Theimpactofthecontiguityvariationwasinvesti-gatedinasubsequenttestphase,whichwassimilarforallparticipants.Here,theformereecttoneswerepre-sentedasimperativestimuliforatwo-choiceresponsetask.Theparticipantshadtorespondtoeachtonewithacertainkeypress,andthekeypressesnolongerpro-ducedauditoryeects.Thetestphaseconsistedoftwoblocks.Participantsrespondedinaconsistentfashioninoneofthetestblocks,andinaninconsistentfashionintheothertestblock(seeFig.1).Ifparticipantshadac-quiredaction-eectassociationsinthelearningphase,perceivingaformereecttoneshouldprimetheasso-ciatedmovement,andtheparticipantsshouldperformbetterintheacquisition-consistenttestblockthanintheacquisition-inconsistenttestblock.Inthelearningphaseofthreecontiguitygroups(Group50,Group1000,andGroup2000),thetone Atrstsight,causalityratingsmayseemtobemorerelatedtoclassicalconditioningthantoinstrumentallearning.Indeed,justlikecausaljudgments,classicalconditioningisabouttherelation-shipof,andtheresultingassociationbetweentwoevents,suchasaperceivedcauseandaperceivedeect.Thatsuchanassociationmaybethebasisforourperceptionofcausalitywasthecoreofs(1739/1964)approach.Incontrast,originalinterpretationsofinstrumentallearningdidnotassumethattheeectofanactionisactuallylearned;instead,aneectwasclaimedtoonlysignalwhethertheresponseproducingitistobelearnedorstrengthened(e.g.,Thorndike,1927).Thus,theoriginalconceptofinstrumentallearningtreatsactioneectsasmerelyprovidingtheglueneededtoassociateresponsesandthestimuliprecedingthem(Walker,1969)butnotaselementsoftheresultingmemorytrace(Hommel,1998).However,morerecentstudieshaveshownthatevenanimalsacquireknowledgeaboutthecontentofaction-eectrelations(Brogden,1962;Meck,1985;Rescorla,1992;Trapold,1970;Ur-cuioliDeMarse,1996;foranoverview,seeElsnerHommel,2001),whichlendscredittoTolman,Hall,andBretnalls(1932)viewofactioneectsasinformationtobeintegrated,i.e.,knowledgeaboutactioneectsmaywellbecomeapartofanactionsrepresentation(Hommel,1998),whichagainmayunderlieourabilitytoperceivethatactionascausingtheeect(Haggard,Aschersleben,Gehrke,Prinz,2002).However,eventhoughwetendtorelateaction-eectlearningmoretoinstrumentallearningthantoclassicalcon-ditioning,nothinginourconclusionsdependsonthispreference. appeared50ms,1,000ms,or2,000msafterkeypressonset.Ifaction-eectlearningisaectedbytemporalcontiguity,weexpecttheacquisition-consistencyeectinthetestphasetodecreasewithincreasingmovement-eectdelayinthelearningphase.However,inGroup2000themovement-eectintervalandITIbothlasted2,000ms,andthismayhamperthedistinctionbetweentheintervalsandthusaectaction-eectlearning.Therefore,weincludedafourthcontiguitygroup(Group1000-ITI),inwhichbothintervalslasted1,000ms.Ifthediscriminabilityoftheintervalshasanimpactonaction-eectlearning(Kaplan&Hearst,1982),thetestphaseresultsinGroup1000-ITIshouldcorrespondtothoseofGroup2000.Inthelearningphaseofafthgroup(Group2000-IS),wepresentedanirrelevantstimulusthatalmostcompletelylledthe2,000-msmovement-eectdelay.IftheISamelioratesthelearningdecitthatmaybeproducedbythelongdelay(Reed,1999),theconsistencyeectshouldbegreaterinGroup2000-ISthaninGroup2000.MethodFortyadults(27female,13male;1left-handed,39right-handed)werepaidtoparticipateintheexperiment.Theiraverageagewas24years.Theparticipantsreportedhavingnormalorcorrected-to-normalvisionandaudition,andtheywereblindtothepurposeoftheexperiment.Forthecontiguityvariation,theparticipantswererandomlyassignedtovegroupsofeightparticipantseach.StimuliandapparatusThedisplayandtimingwascontrolledbyaHewlettPackardVectraQS/20computer,interfacedtoanEizoFlexscanmonitor.Thevi-sualstimuliweredisplayedonablackbackground.Acentralwhiteasterisk(*)servedasxationpoint,andablackarrowwithwhiteoutlinewaspresented3.5totheleftorrightofthexationpointtosignaltheresponse.Fromaviewingdistanceofabout60cm,thearrowsubtendedavisualangleof3inwidthand1.5inheight.Thearrowheadpointedeithertotheleftortotheright.Thepar-ticipantsrespondedwiththeirleftandrightindexngeronakeyboardplacedcentrallyinfrontofthemonitor,whichconsistedofsix2--2-cmkeysarrangedinahorizontalrow.Thethreeleft-handkeysandthethreeright-handkeyswere1cmapart,andthetwogroupsofthreekeyswereseparatedby22cm.AuditorystimuliweresinusoidaltonesorMIDItones(i.e.,belltones;instrumentMarimba)of400Hz(lowpitch)or800Hz(highpitch),presentedsimultaneouslythroughtheleftandrightspeakerofaheadphone.TheIS,whichwasonlypresentedinGroup2000-IS,wasaMIDItone(instrumentFlute)of600Hz.Theexperimentconsistedofasinglesessionofabout40 60min,dependingonthecontiguitycondition.Thesessionwasdividedintoalearningphaseandatestphase.LearningphaseThesubjectsleftandrightindexngerrestedonthesecondandthefthkeysrespectively,countingfromlefttoright(Fig.1).Onappearanceofthearrow,oneindexngerwastobemovedfromthehomekeytotheadjacentleftorrightkeyasquicklyaspossible.Thelocationofthearrowindicatedthehandtobemoved,i.e.,anarrowleftofthexationpointsignaledthelefthand,andaarrowontherightsidetherighthand.Thedirectionofthearrowheadindicatedthedirectionofthemovement,thus,aleft-pointingarrowsignaledaleftwardmovement(touchingkey1withtheleft,orkey4withtherightindexnger),andaright-pointingarrowarightwardmovement(touchingkey3withtheleft,orkey6withtherightindexnger).Eachmovementendedbymovingthengerbacktothehomekey.Eachkeypresstriggeredaparticulartone.Forallparticipants,pressesoftheleftkey(i.e.,key1orkey4)triggeredasinusoidaltone,andpressesoftherightkey(i.e.,key3orkey6)triggeredabell(i.e.,MIDI)tone.Thepitchofthetonewasmappedonthehandandwasbalancedacrossparticipants.Foronehalfoftheparticipants,theleftindexngertriggeredthelowtonesandtherightindexngertriggeredthehightones(action-eectmappingA),whiletheotherhalfoftheparticipantsreceivedtheoppositeaction-eectmappingB.Participantswerenotinformedabouttheaction-eectmappingandweretoldthatthetoneswerecompletelyirrelevanttothetaskandshouldthereforebeignored.Eachcontiguitygroupconsistedoffourparticipantswithac-tion-eectmappingAandfourwithaction-eectmappingB.Theprocedureinthreeofthevegroupsdieredonlyaccordingtothetemporaldelaybetweenresponseandeect(R-Einterval).Withineachgroup,thetemporalcontiguitywasthesameforallfourac-tion-eectpairs.AsshowninTable1,threegroupsobtainedaconstantITIof2,000ms,andtheeecttoneappeared50ms(Group50),1,000ms(Group1000),or2,000ms(Group2000)afterthekeypress.ThedesignofGroup1000-ITIwasidenticaltothatofGroup1000,buttheITIwas1,000ms.ThedesignofGroup2000-ISwasidenticaltothatofGroup2000,but50msafterthekeypresstheISwaspresentedfor1,950ms,anddirectlyafterthat,theeecttoneappearedfor200ms.Thus,theR-EintervalinGroup2000-ISwasalmostcompletelylledbytheIS.FollowingtheITI,eachacquisitiontrialstartedwitha500-msdisplayofthexationpoint.Afterthat,anarrowappearedfor200mstotheleftorrightofthexationpoint.Fromarrowonset,theprogramwaitedupto1,000msforakeypress.Ifthecorrect Fig.1BasicdesignofExperiments1and2.Examplesofthevisualstimuli,responsescurvedarrow:ngerismoved,crossinbox:ngerisnotmoved),andeectsinthelearningphaseandtestphaseforaparticipantunderresponseeectmappingA responsekeywaspressed,thecorrespondingeecttonewaspre-sentedfor200ms,starting50,1,000,or2,000msaftertheonsetofthekeypress,dependingonthecontiguitycondition.Trialswithresponselatenciesexceeding1,000mswerecountedasmissing,andresponsesfasterthan100mswerecountedasanticipationerrors.Incorrectkeypresses,responseomissions,andanticipationswererecorded,fedbacktotheparticipantsbya1,000-mswarningmessage,andrepeatedinarandompositionduringtheremainderoftheblock.Participantsworkedthrougheightpracticetrialsand400validacquisitiontrials(i.e.,fouraction-eectpairswith100repetitionseach).xTestphaseAftercompletingtheacquisitiontrials,participantswereverballyinstructedforthetestphase,whichwasthesamefortheparticipantsofallgroups.Ineachtesttrial,oneofthefoureecttoneswaspresentedasanimperativestimulus,andthepar-ticipantswereaskedtorespondtothisstimulusasquicklyandascorrectlyaspossibleaccordingtoaxedstimulus-response(S-R)mapping.Thetestphasewasdividedintotwoblocks,inwhichonlythetonesofonesoundcategorywerepresentedasimperativestimuli.Becausethesoundcategorieswerepreviouslymappedontothekeys,theparticipantsrespondedwithbothhandsineachblock,buttheyusedonlytheleftkeys(key1andkey4)inoneblock,andtherightkeys(key3andkey6)intheother.Intheacquisition-consistenttestblock,theparticipantshadtopressthekeythatprecededthetoneintheacquisitionphase.Forinstance,partici-pantswhohadexperiencedaction-eectmappingA(lefthandlowtone,righthandhightone)werenowtorespondtothelowtonewiththeleftindexngerandtothehightonewiththerightindexnger.Intheacquisition-inconsistenttestblock,thesameparticipantsweretorespondtothelowtonewiththerightindexnger,andtothehightonewiththeleftindexnger.Halfoftheparticipantsworkedthroughtheacquisition-consistenttestblockrst,theotherhalfworkedthroughtheacquisition-inconsistentblockrst.Additionally,halfoftheparticipantsreceivedtheacquisition-consistentinstructionforthesinusoidaltones,andtheotherhalfforthebelltones.Eachtesttrialstartedaftera1,500-msITIwitha200-mspre-sentationofaloworhightone.Thentheprogramwaitedupto1,000msforakeypress.Afterthekeypress,notoneappeared,butthenextITIstartedimmediately.Incorrectkeypresses,responseomissions,andanticipationsweretreatedasintheacquisitionphase.Participantsworkedthrough10practicetrialsand100validtesttrialsintherstblock.Afterthat,ascreenmessageinformedtheparticipantsthatonlythetonesoftheothersoundcategorywouldbepresentedinthenextblock,andinstructedthemforthenewS-Rmapping.Thenumberoftrialswasidenticalinthetwotestblocks,sothattheparticipantsworkedthroughatotalof20practicetrialsand200validtesttrials(twoblockswithtwoeect-responsepairsand50repetitionseach).ResultsThesignicancecriterionwassetto.05forallanalyses(one-tailedforsinglecontrastsofpredictedef-fects).LearningphaseAfterexcludingresponseomissions(1.0%)andantici-pations(0.1%),individualmeanRTswerecalculated.Aone-wayANOVAdidnotrevealaneectofcontiguitygroup,indicatingthatRTsinthelearningphasewerestatisticallycomparableacrossgroups(meanRTandstandarderror:Group50:450.3[13.9],Group1000:487.8[16.4],Group2000:485.9[15.5],Group1000-ITI:483.9[25.3],Group2000-IS:479.1[22.1]).TestphaseTrialswithresponseomissions(0.7%)wereexcluded,andanticipationsdidnotoccur.MeanRTsandper-centagesoferrorwerecalculatedandanalyzedasafunctionofgroupandacquisition-consistencyblock(S-Rmappingacquisition-consistentvs.-inconsistent).AsshowninFig.2,theparticipantsofGroup50,Group1000,andGroup1000-ITIrespondedfasterintheacquisition-consistentblockthanintheacquisition-inconsistentblock,andseparate-testsrevealedthattheacquisition-consistencyeectwassignicantineachofthesegroups,s(7)=2.58(=.02),2.04(=.04),and2.08(=.04)respectively.InGroup2000andGroup2000-IS,theconsistencyeectwasnotsigni-cant(Group2000:(7)=-1.01,=.16;Group2000-(7)==.12).Thus,areliableeectofacquisitionconsistencywasobtainedinallgroupswithaction-eectlearningdelaysofupto1,000ms.A5(contiguitygroup)2(consistencyblock)analysisofvariance(ANOVA)onlyrevealedasignicantinter-action((4,35)=3.84,=.01),showingthatthecon-sistencyeectdiersbetweenthecontiguitygroups.Separateanalyseswereconductedtoinvestigatethegroupdierences.ForGroup50,Group1000andGroup2000,aseparate3(contiguitygroup)2(consis-tencyblock)ANOVAalsoyieldedonlyasignicantinteraction((2,21)=4.5,.02).Asdeterminedby Table1Experiment1:Lengthoftheintertrialinterval()andofthedelaybetweenresponseandeect(R-Einterval)inthelearningphaseofthevecontiguitygroupsGroupITI(ms)R-Einterval(ms)FillertoneGroup502,00050NoGroup10002,0001,000NoGroup20002,0002,000NoGroup1000-ITI1,0001,000NoGroup2000-IS2,0002,000Yes Fig.2Experiment1:Meanreactiontimesandstandarderrorsintheacquisition-consistentand-inconsistenttestblocksinthevecontiguitygroups.indicateasignicantconsistencyeectwithinagroup(.05) Newman-Keulstests,thecriticaldierencebetweentheconsistencyeectsoftwogroupsis32.1ms(of.05ad-justedbyaBonferronicorrectionfor10possibletests=.005).Thus,theconsistencyeectsofthefollowinggroupsdieredsignicantly:Group50vs.Group2000(37.0msvs.-14.5ms),Group50vs.Group2000-IS(37.0vs.-19.0ms),Group1000vs.Group2000-IS(14.3vs.-f19.0ms),andGroup1000-ITIvs.Group2000-IS(13.6vs.-19.0ms).ThereliableconsistencyeectsinGroup50,Group1000,andGroup1000-ITIdidnotdier,andneitherdidthenon-reliableeectsinGroup2000andGroup2000-IS.Incorrectkeypresseswererare(2.2%and2.6%intheacquisition-consistentand-inconsistentblocksrespec-tively)anddidnotproduceanyeectinanANOVA.DiscussionTheresultsofExperiment1meetourexpectationthatthetemporalcontiguitybetweenactionsandtheireectsisanimportantprerequisiteforacquiringassociationsbetweentheircognitiverepresentations.Whenpartici-pantshaveexperiencedthatakeypressmovementisfollowedbyacertaintonewithadelayof50or1,000ms,theyperformthatkeypressfasterafterhear-ingtheformereecttonethanafterhearingthealter-nativetone.However,whenthetoneoccurs2,000msafterthekeypress,hearingtheeecttoneinthetestphasedoesnotleadtoreliableperformancedierencesunderacquisition-consistentand-inconsistenteect-ac-tionmapping.TheconsistencyeectoftheGroup50isonlynumericallyhigherthan,butnotstatisticallydif-ferentfrom,theeectofthetwogroupswitha1,000-msmovement-eectinterval.Althoughthismaybecausedbyalackofstatisticalpowerduetosmallgroupsizes=8),ourdatapointmoretoacriticaltimewindowfortheintegrationofsimplekeypressmovementsandfollowingsimpletonesthantoagradualdecreaseinthestrengthoftheacquiredaction-eectassociationswithdecreasingcontiguity.Beingexposedtoaparticularaction-eectmappinginthelearningphaseinuencedsubsequenttestperformanceonlyifeectsofkeypressesweredelayedbynomorethan1s,whereasa2-sdelaywasapparentlytoolongtocreatesucientlystableac-tion-eectassociations.Analternativeexplanationforthelackofthecon-sistencyeectinGroup2000isthattheparticipantshadproblemsindierentiatingthe2,000-msmovement-ef-fectintervalfromthe2,000-msintertrialinterval(Kap-lan&Hearst,1982).However,ifthishadbeenthecase,theconsistencyeectshouldalsobelackinginGroup1000-ITI,whichalsofacedaR-EintervalandITIofequallength,anditshouldbepresentinGroup2000-IS,inwhichtheinterveningstimulusfacil-itatedthediscriminationoftheR-EintervalandITI.ButthedatashowedasignicantconsistencyeectinGroup1000-ITIthatwassimilartothatofGroup1000,butdieredfromthatinGroup2000-IS.Thus,thetemporalcontiguityofmovementandeectseemstobemoreimportantfortheacquisitionofaction-eectrelationsthanthediscriminabilityofthemovement-eectintervalandintertrialinterval.Theimportanceofmovement-eectcontiguityisfurthersupportedbythecomparisonofGroup2000andGroup2000-IS.ThefactthatneitherofthesegroupsshowedareliableconsistencyeectcontradictstheexpectationthatanirrelevantstimuluspresentedintheR-Eintervalcancompensateforlowtemporalcontigu-ity.Inviewoftheevidenceofapositiveimpactofinterveningstimulioninstrumentallearninginhumansandotheranimals(Einhorn&Hogarth,1986;Gruber,Fink,&Damm,1957;Reed,1992,1999)thisisasome-whatsurprisingnding.However,themoreconvincingdemonstrationsofISbenetscomefromstudiesofani-malconditioningratherthanstudiesonhumancausaljudgments.AndeventhoughReeds(1992,1999)datasupporttheassumptionthatllingthegapbetweenactionandeectfacilitatesthedetectionofacausalrelationbetweenthetwo,Davey(1983)showedthatwhenachainofstimuliispresented,participantsacquireknowledgeoftheassociationbetweencontiguousstimuli(i.e.,betweenactionandIS,orbetweenISandeect),butlittleknowledgeoftherelationshipbetweenmoredistalstimuli(i.e.,betweenactionandeect).Thefailuretondamediatingeectoftheinter-veningstimulusmayalsohavesomethingtodowithourdependentvariable.AccordingtoReed(1999),theim-pactoftheISisbasedonsecondaryreinforcement.Thus,thedetectionofarelationshipbetweenactionandeectisfacilitatedbylearningprocessesthatassociatetheISwithboththeresponseandtheeect,resultinginaR-IS-Eassociation.Inthepresentstudy,theoccurrenceofaconsistencyeectinthetestphaseisaratherindirectmeasureoftheacquiredaction-eectassociations.ThelackoftheconsistencyeectinGroup2000-ISmayeitherbeduetothefactthatnoR-IS-Eassociationshavebeenlearned,ortothefactthattheactivationspreadingbackwardsonR-IS-Eassociationsistooweaktoallowanactivationoftheaction(i.e.,Associate1)byactivatingtherepresentationoftheeect(i.e.,Associ-ate3).Ifso,thelackofabehavioraleectinGroup2000-ISwouldnotbeduetotheabsenceoflearning,butinsteadduetolearningthatisinsucienttoaectovertperformance.Furtherstudiesarerequiredtoclarifythisissue.Takentogether,theresultsofExperiment1showthatthetemporalproximityofmovementandeectisacriticalfactorfortheacquisitionofaction-eectknowledge.Onlyifthedelaybetweenkeypressesandeectswasnolongerthan1s,theacquiredaction-eectassociationshadanimpactontheselectionandexecu-tionofsubsequentactions.ThissupportsandextendstheassumptionofElsnerandHommels(2001)two-stagemodelclaimingthatvoluntaryactioncontrolisbasedonacquiredaction-eectassociations.Indeed,suchassociationsseemtobeformedontheoccasionofmereco-occurrencesofamovementandasubsequent sensoryevent,providedthatthelatterisperceivedwithinawindowofabout1saftermovementonset.Thissuggeststhattraceconditioningisatworkhere,i.e.,associationsbetweenthecodesofamovementandaperceptualeectarecreatedwhentheiractivationsoverlapintime(Hommel,2003).Ifweassumethattheactivationofcodesdecaysovertime(Hommel,1994)suchanoverlapislesslikelythemoretheeectisde-layed.Inviewofourpresentresults,itwouldneedtobeassumedthatthedecayofthecodesofsimplekeypressestakeslongerthan1butshorterthan2s.Experiment2:ContingencyAlthoughtemporalcontiguityisanimportantmech-anisminhumanandanimalassociativelearning,amechanismrelyingsolelyonthecontiguityofmove-mentandeectwouldoftennotobtainvalidresults.Letusimagineatonethatisproducedbypressingapianokeyandatelephoneringthathappenstoco-occurwithahitonthecomputerkeyboard.Bothtonesmaybeperfectlycontiguouswiththeactionandyetonlytheformerisactuallyproducedbyit.Hence,validjudgmentsoftherelationbetweenactionandeectwouldrequireanorganismtoexperienceseveralsituationsinwhichamovementandafollowingsen-soryeventhavethechancetoco-occurfrequently,whichwouldpointtoacausalrelation,ortoappearseparately.Thereisampleevidencethathumans(Al-lan&Jenkins,1980;Chatlosh,Neunaber,&Wasser-man,1985;Shanks&Dickinson,1987)andotheranimals(Dickinson&Charnock,1985;Hammond,1980)useinformationaboutprobabilisticcontingen-ciestodetectrelationsbetweenmovementsandeects.Contingencydetectionisusuallyinvestigatedinthefree-operantparadigm(Dickinson&Charnock,1985;Hammond,1980;Shanks&Dickinson,1991;Wasserman,1990),inwhichparticipantsmayperform(R)ormaynotperform(R)aparticularresponse.Iftheydoperformtheresponse,aneectdoes(E)ordoesnot(E)occuraccordingtoaxedprobability(E|R).Iftheydonotperformtheresponse,theeectappearswithanotherxedprobability(E|-R).Usingthisparadigm,DickinsonandShanks(1985)notedthatforagiven(E|R),increasing(E|-R)decreasesattributionsofcausaleectivenessregardingthere-sponseinhumans.IntheclassicaldenitionbyRescorla(1967),thedegreeofcontingencybetweenresponseandeectisexpressedbythedierencebetweenthetwoprobabilities(i.e.,Delta-),whichagainarecalculatedfromtherel-ativefrequencies()ofthepresenceorabsenceofre-sponseandeect: FðÞFððÞ ApositiveDelta-valueexpressesthattheeectoc-cursmoreofteninthepresencethanintheabsenceofagivenresponse.Underpositivecontingency,animalslearnexcitatoryassociations,andhumansjudgethere-sponseasthecauseoftheeect(Shanks&Dickinson,1987;Wasserman,1990;Young,1995).Typically,thestrengthofthelearnedassociationincreaseswithincreasingpositivecontingency.Undernon-contin-gency,i.e.,whenDelta-equalszero,theeectappearsequallyofteninthepresenceandintheabsenceofaresponse,andthus,actionandeectareindependentofeachother.Animalsrespondtonon-contingencieswithlearnedirrelevance(Seligman,Maier,&Solomon,1971).Theybecomeinactivebecausetheirbehaviorhasnoimpactontheappearanceofareinforcer.Forhumans,non-contingenciesaremorediculttojudgethanpo-sitivecontingencies(Shanks,1993).Rule-basedorstatisticalmodelsofinstrumentallearning(seeAllan,1993;Shanks,1993;Waldman&Holyoak,1992;Wasserman&Miller,1997)taketheinterdependencebetweenbehaviorandcontingencyasevidencethatorganismsadapttheirbehavioraccordingtotheDelta-rule.However,theapplicationoftherulemakesstrongdemandsoncognitiveabil-ities.Firstly,thecalculationofrelativeprobabilitiesre-quiresaretrospectiveevaluationofthepresenceorabsenceofallactionsandsensoryeventsinallrelevantsituations,andthuscallsforhighmemoryandatten-tionalcapacities.Secondly,thecalculationrequiresanaccurateperceptionoftheco-occurrenceofthetwoevents,whichis,amongotherthings,dependentonthetemporallimitsofthecontingencyanalysis(Wasser-mann,1990).Forinstance,ifaneectappears5safterthemovement,a6-sanalysisintervalwouldresultinadding1to(E|R),buta2-sintervalwouldresultinadding1toE|R),1toR),and1toGiventhehighdemandsofDelta-calculation,someresearchersdoubtthathumansandotheranimalsactu-allyusethisruletodeterminetherelationshipsbetweenevents,andfavorassociativemodelsinstead(e.g.,Allan,1993;Shanks,1993;Wasserman&Miller,1997).Sha-kleeandWasserman(1986)showedthatcausalityjudgmentsthatapparentlymatchedtheDelta-rulecouldwellresultfromother,simplerlearningprocesses.Shakleeandcolleaguesfoundthattheprocesspartici-pantsusetodeterminethedependenceoftwoeventsisdetermined,amongotherthings,bytheirageandmemoryload.Elementary-schoolchildrenseemtorelyonfrequency-basedassociativemechanisms,whereasuniversitystudentsprefertheDelta-rule(Shaklee,Holt,Elek,&Hall,1988).However,adultsalsotendtoshifttosimplerprocessesunderhighmemoryload(Shaklee&Mims,1982).DeviationsfromtheDelta-rulearealsocommoninnon-contingentconditions(Shanks&Dickinson,1987;Wasserman&Shaklee,1984).Chatloshetal.(1985)reportedthattheoverallfrequencyoftheeect(E)hasanimpactoncausalityjudgments,especially innon-contingentsituationswithaversiveeects.Thus,theparticipantsmadehighercausalityratings(E|R)and(E|-R)wereequallyhigh(e.g.,.80and.80)thanwhentheywereequallylow(e.g.,.20and.20),althoughDelta-waszeroinbothcondi-tions.Theoverallfrequencyoftheeectisalsoimportantwhenparticipantshavetojudgethecau-salityofseveralactionsatthesametime,suchaswhentwokeyscausetheappearanceofalightwithdierentcontingencies(Jenkins&Ward,1965).Probably,theconcentrationonoverallfrequencyoftheeectmakeslowerdemandsonmemoryandattentionandisthereforeusedinsituationswithhighcognitiveload.Takentogether,studiesoninstrumentallearninginhumansandotheranimalsimplythatapositivecontin-gencyfacilitatesthedetectionofarelationshipbetweenactionandeect.However,thesameobservablebehaviormaybebasedondierentprocesses,suchasthecalculationofrelativeprobabilities(i.e.,Delta-),ortheconcentrationontheoverallfrequencyoftheeect.Takingasidetheprocessesthatunderlietheobservablebehavior,theappearanceofaneectintheabsenceofaresponseisimportantforthecausalpowerofaresponse,andmodelsthatrelysolelyonthetemporalcontiguityofactionandeectarenotwellequippedtotakethisinformationintoaccount.Iftheacquisitionofactioneectsisbasedonasso-ciativelearningmechanisms,itshouldbeinuencednotonlybythetemporalcontiguityofmovementandef-fect,asinvestigatedinExperiment1,butalsobyac-tion-eectcontingency.Therstevidenceoftheimpactofcontingencyonlatentaction-eectlearningcomesfromstudiesusingserialresponsetasks(Homannetal.,2001;Ziessler,1998).However,thesestudiesdidnotdirectlyaddresswhetheragivenaction-eectassociationisweakeriftheeectoccursfrequentlyintheabsenceofthemovement.ThiswastestedinExperiment2,inwhichtheaction-eectcontingencyinthelearningphasewasvariedinvegroups.BecausecomputingDelta-requiresthatresponsesaresome-timesperformed(R)andsometimesnot(R),wedesignedago/no-goversionofourtask.Wepresentedgreenorredarrowsandparticipantswereinstructedtorespondonlytoagreen,andnottoaredarrow(Fig.3).Wheneverparticipantsrespondedtoagreenarrow,thecorrespondingactioneectappearedwiththexedprobability(E|R),andwheneverthesubjectdidnotrespondtoaredarrow,theeectappearedwiththexedprobabilityR).Ifcontingencyhasanimpactonaction-eectlearning,theacquisition-consistencyeectinthetestphaseshouldbemorepronouncedthehighertheDelta-inthelearningphase,i.e.,themoretheprobabilityofaneectgiventheaction(E|R)exceedstheprobabilityoftheeectgiventheabsenceoftheactionInthreegroups,Delta-contingencieswerevariedbykeeping(E|R)constantandincreasingtheproba-bilityofR):Group.60,Group.30,andGroup.00/.80(Table2).Totestwhethertheoverallfrequencyofaneect(E)hasanimpactonthetestbehaviorundernon-contingency,weaddedtwofurthergroupsinwhichDelta-equalszero:Group.00/.50andGroup.00/.20.InGroup.00/.80,theeectoc-curredin80%ofthelearningtrials,inGroup.00/.50in50%,andinGroup.00/.20in20%.ToallowforacomparisonoftheresultsofExperiment2withthedataoncausalityjudgments(cf.Shanks&Dickinson,1991),andtotestforthepossibleimpactofperceivedcausalityontheconsistencyeect,wealsoaskedtheparticipants,afterthelearningphase,toratetheamounttowhichtheoccurrenceofthetoneshasbeendependentonthekeypresses.ThetestphasewasidenticalforallparticipantsandcorrespondedtothatofExperiment1.Again,performancedierencesbetweentheacquisition-consistentand-inconsistenttestblockweretakenasevidencethataction-eectassociationshadbeenacquiredinthelearningphase,andgroupdierencesintheacquisition-consistency Fig.3Experiment2:Possiblecombinationofthepresence(R)orabsenceofakeypress(R)andofthepresence(E)orabsence(ofaneectforoneofthefourresponsesinthelearningphase.Frequency()ofeverycombinationin100learningtrials,probability()andcontingency(Delta-)foronecontingency Table2Experiment2:Frequencies()andprobabilities()ofthepresence)orabsence(E)ofaneecttonegiventhepresence()orabsence(R)ofakeypressinthelearningtrialsofthevecontingencygroupsR)DeltaGroup.604010.801040.20.6050Group.304010.802525.50.3065Group.00/.804010.804010.80.0080Group.00/.502525.502525.50.0050Group.00/.201040.201040.20.0020 eectsweretakenasevidencethatthecontingencybetweenmovementandeectinuencedaction-eectlearning.MethodFortyadults(28female,12male;33right-handed,6left-handed,1ambidextrous)werepaidtoparticipateintheexperiment.Theiraverageagewas28years.TheyfullledthesamecriteriaasinExperiment1.Theparticipantswererandomlyassignedtovegroupsofeightparticipantseach.StimuliandapparatusThesewereasinExperiment1,withoneexception.Thearrowthatservedasimperativestimulusinthelearningphasehadeitheraredoragreenoutline.Theexperimentconsistedofasinglesessionofabout45min,di-videdintoalearningphaseandatestphase.LearningphaseTheprocedureofthelearningphasewasidenticaltothatofExperiment1,withthefollowingexceptions:Afteragreenarrow,participantsweretopressthecorrespondingkey,justasinExperiment1.Afteraredarrow,however,theyweretoleavetheirngersonthehomekeys(key2andkey5).SimilartoExperiment1,eachkeywasmappedontooneoffourtones.However,thetonesdidnotalwaysandnotonlyoccurafterakeypress,hence,sometimesakeypressdidnotproduceatone,andsometimesatoneappearedwithoutakeypress.Inthevecontingencygroups,therelativefrequenciesofthepresenceorabsenceofthetonesgiventhepresenceorab-senceofthecorrespondingkeypressvariedasshowninTable2.Withineachgroup,thecontingencywasconstantforallfouraction-eectpairings.Inthreegroups,(E|R)wassetto.80,andthecontingencywasvariedbyincreasingtheprobabilityofR).AsdisplayedinTable2,thisresultedinthreegroupswithdierentDelta-contingencies:Group.60(R)=.80.20),Group.30(.80.50),andGroup.00/.80.80).IntwofurthergroupswithaDelta-contingencyof.00,theoverallfrequencyofaneect(E)wasvaried.InGroup.00/.80,theeectoccurredin80%ofthelearningtrials,inGroup.00/.50(.50.50)in50%,andinGroup.00/.20(.20.20)in20%.Eachlearningtrialstartedafteranintertrialintervalof1,000mswiththepresentationofthexationpointandthearrowasdescribedinExperiment1.Afteragreenarrow,theprogramwaitedupto1,000msforakeypress.Whenthecor-rectkeywaspressedwithinthisinterval,thecorrespondingtonewaseitherpresentedfor200ms,withanaction-eectintervalof50ms,orwasnotpresented.Inthelattercase,thekeypresstriggeredthenextITIdirectly.Afteraredarrow,theprogramwaitedupto700msforaresponse.Ifnokeywaspressedinthisinterval,thecorrespondingtonewaseitherpresentedfor200ms,orwasnotpresented.Errors,anticipations,andomissionsweredenedandtreatedasinExperiment1.Additionally,keypressesfollowingaredarrowwerecountedasfalsealarms,andweretreatedliketheothererrors.Theparticipantsworkedthrough24practicetrialsand400validlearningtrials(fourR-Epairswith100replications).Participantswereinformedthatatonemayormaynotappearafterakeypress,andthattonesmayormaynotappearwhennokeyhadbeenpressed.Weaskedtheparticipantstoattendtothedependencyoftheeectonthekeypressinthelearningphase,andaftercompletingthisphasetheyweretojudgethisdepen-dencyonascalesimilartothatusedbyShanksandDickinson(1991),whichrangedfrom0(thekeypresswasneverfollowedbyatone)to100(thekeypresswasalwaysfollowedbyatone).Toeasethejudgmenttask,thelearningphasewasdividedintosevenblocksinwhicheitherone,two,orallfourofthekeypresseswererequired.Thesuccessionoftheblockswasbalancedacrosspar-TestphaseTheprocedureofthetestphasewasidenticaltothatofExperiment1.ResultsLearningphaseResponseomissions(0.3%),anticipations(0.02%),andfalsealarmresponses(0.01%)wereexcludedfromtheanalysis.TheremainingdataweretreatedasinExperiment1.Again,thereactiontimesofthecon-tingencygroupsdidnotdiersignicantly(meanRTandstandarderror:Group.60:445.3[12.2],Group.30:434.2[11.5],Group.00/.80:428.3[19.6],Group.00/.50:448.3[20.1],Group.00/.20:439.2[15.9]).CausalityjudgmentsThecausalityjudgmentswerecalculatedforeachgroupandeachkey.A5(contingencygroup)4(key)ANOVArevealedonlyasignicantmaineectofcon-tingency((4,35)=7.57,.001).ThemeancausalityjudgmentsforeachgroupareshowninFig.4.Onthegivenscale,anodependencyjudgmentwouldresultinaratingof50.ThecausalityjudgmentsofGroup.60(7)=3.45,=.005),Group.30((7)=3.45,.015),andGroup.00/.80((7)=5.95,)weresignicantlyhigherthanthisvalue,thejudgmentsofthe Itmaybeaconcernthataskingparticipantstoattendtotheaction-eectrelationshipmayhaverenderedthelearningmoreexplicitthaninExperiment1and,thus,mayhaveengagedadierentlearningmechanism.Wehavenoevidencethatwouldsupportthisconsideration.Firstly,Hommel,Alonso,andFuentes(2001)investigatedaction-eectlearningunderanimplicitinstructioninalargesampleandsortedtheirparticipantsaccord-ingtowhethertheywereableorwerenotabletorecallthere-sponse-eectmappingofthelearningphaseafterthetestphase.Therewasnoevidencethatlearningwasanydierentinthesetwogroups,suggestingthattheimpactofaction-eectassociationsdoesnotdependon,andisnotevenmodiedby,explicitknowl-edge.(Wemadethesameobservationinanumberofpilotstudieswherewesystematicallymanipulatedtheemphasisgiventotheaction-eectrelation.)Secondly,ifanexplicitinstructionwouldreallyengageadierent,moreintentionallearningmechanism,thiswouldbeexpectedtoincreasethecorrelationbetweentheconsis-tencyeectsandthecausalityjudgments;andyet,wewillseethatthiscorrelationisanythingbutimpressive.Thirdly,wewillseethattheeectiveconditionsinExperiment2producedeectsthatareverycomparableinsizetotheeectsobtainedinExperiment1,whichdoesalsonotsupporttheassumptionthatdierentlearningmechanismswereatwork. twoothergroupsdidnotdierfrom50(s�.08).Newman-Keulstestsyieldedacriticaldierencebetweenthemeancausalityjudgmentsoftwogroupsof17.4(adjustedfor5tests=.01),andthus,thejudgmentsofGroup.00/.80,Group.60,andGroup.30didnotdierfromeachother(72.5vs.65.3vs.59.8),butthejudg-mentsofGroup.00/.80andGroup.60weresigni-cantlyhigherthanthoseofGroup.00/.50andGroup.00/.20,whichalsodidnotdierfromeachother(44.9vs.44.8).TestphaseResponseomissions(0.9%)wereexcludedfromtheanalysis,andanticipationsdidnotoccur.Fortheremainingtrials,meanRTsanderrorrateswerecalcu-latedforeachgroupandforbothblocks(i.e.,acquisi-tion-consistentand-inconsistentmappingoftoneandkeypress).AsshowninFig.5andconrmedbyseparate-tests,Group.60andGroup.00/.80respondedsignicantlyfasterintheacquisition-consistenttestblockthanintheacquisition-inconsistenttestblock,s(7)=2.24(.03)and3.08(=.009).Inallothergroups,thecon-sistencyeectdidnotreachthesignicancelevel(Group.30:(7)==.16;Group.00/.50:=.79,=.23;Group.00/.20:(7)==.06).FortheRTs,a5(contingencygroup)2(consistencyblock)ANOVAyieldedasignicantmaineectofcontingencygroup((4,35)=2.58,=.05),whichwasduetoasignicanthigheroverallRTinGroup.60thaninGroup.00/.50.Additionally,theANOVAyieldedasignicantinteraction((4,35)=2.70,=.05),show-ingthattheconsistencyeectdiersbetweenthecon-tingencygroups.Aseparate3(contingencygroup)(consistencyblock)ANOVAforthegroupsinwhich(E|R)wassetto.80andthecontingencywasvariedbyincreasingtheprobabilityofR)(i.e.,Group.60,Group.30,andGroup.00/.80)alsorevealedasigni-cantinteraction((2,21)=3.93,=.04).ThecriticalNewman-Keulsdierencebetweentheconsistencyef-fectsoftwogroupsis28.2ms(adjustedfor3tests=.017)andthustheconsistencyeectsofGroup.00/.80andGroup.60(23.1vs.18.7ms)didnotdiersigni-cantly,butbothgroupsdieredfromGroup.30(-17.2ms).Forthenon-contingentgroups(i.e.,Group.00/.80,Group.00/.50,andGroup.00/.20),anotherseparate3(contingencygroup)2(consistencyblock)ANOVAyieldedasignicantmaineectofcontingencygroup(2,21)=4.44,=.03),causedbythefactthatGroup.00/.50respondedfasterthantheothergroups.TheANOVAalsoshowedasignicantinteraction(2,21)=3.27,=.05).Thecriticaldierencebe-tweentheconsistencyeectsoftwogroupsis32.3ms(adjusted=.017),andthustheconsistencyeectsofGroup.00/.80andGroup.00/.50(23.1vs.13.5ms)didnotdiersignicantly,butbothgroupsdieredfromGroup.00/.20(-19.9ms).Errorswererare(2.5%vs.2.8%inacquisition-con-sistentand-inconsistentblocks)anddidnotproduceareliableeect.Totestforapossibleimpactofperceivedcausalityontheacquisition-consistencyeectwecalculatedindividualconsistencyeects(i.e.,acquisition-incon-sistentmeanRTminusacquisition-consistentmeanRT)andcorrelatedthemwiththeindividualmeancausalityjudgments.However,asshowninFig.6,andindicatedbytheperfectlyatregressionline,therewasnohintofanydependencybetweenthesemea-sures.Indeed,thePearsoncorrelationacrossallgroupswasverysmall(=.15)andfarfromsig-nicance(=.35),andthispicturedidnotchangewhenthecorrelationswererunforeachgroupsepa-rately(s�.49). Fig.4Experiment2:Meansandstandarderrorsofthecontin-gencyjudgments,averagedoverthefourkeys,ineachofthevecontingencygroups.Thejudgmentswereobtainedonascalerangingfrom0(thetoneappearedafteraresponse)to100(thetoneappearedafteraresponse) Fig.5Experiment2:Meanreactiontimesandstandarderrorsintheacquisition-consistentand-inconsistenttestblocksinthevecontingencygroups.indicateasignicantconsistencyeectwithinagroup(.05) DiscussionExperiment2yieldedthreeimportantoutcomes.Firstofall,asexpected,theacquisition-consistencyeectinthetestphasewasaectedbythecontingencyofactionandeectinthelearningphase.Thus,ourstudiesconrmtheresultsofHomannetal.(2001)andZiessler(1998)inshowingthataction-eectlearningissensitivetoboththepresenceandtheabsenceofagivenactionanditsperceivedeects.Whichactioneectsareacquiredde-pendsonbothtemporalcontiguityof,andprobabilisticcontingencybetween,actionsandtheireects.Secondly,consistencyeectsandcausalityjudgmentsshowedbothcommonalitiesanddierences.Ontheonehand,bothmeasuresseemtobeaectedbythecontingencymanipulations,andtheyseemtobeaectedinsimilarways,apointwewillgetbacktobelow.Ontheotherhand,however,commonalitieswererestrictedtogroupmeansonly,whereastheindividualcausalityjudgmentsdidnotpredicttheindividualsizeoftheconsistencyeect.Thismeansthattheperceivedcausalityofanaction-eectrelationandtheimpactofanacquiredrelationonlaterperformanceareunlikelytobepuremeasuresofthesameunderlyingfactor,suchasanac-tion-eectassociation.WewilltakeupthisissueintheGeneraldiscussionsection.Thirdly,thepatternsofRTeectsandcausalityjudgmentsacrossthecontingencygroupsindicatethatnotallgroupsbehavedaccordingtotheDelta-rule.Iftheyhad,wewouldhaveexpectedconsistencyeectsandcausalityjudgmentstobemostpositiveinthegroupwiththehighestpositiveDelta-contingency(i.e.,Group.60),somewhatlesspro-nouncedinthegroupwithlowbutstillpositivecon-tingency(i.e.,Group.30),andsmallandcomparableinthenon-contingentgroups(i.e.,Group.00/.80,Group.00/.50,andGroup.00/.20).Instead,theresultsindicatethattwocontingencygroupsstickoutbyshowingbothsignicantconsistencyeectsandhighcausalityjudgments:Group.60andGroup.00/.80.Letusdiscusstheresultsfortheseparategroupsinturn.Theresultsforthetwogroupswithpositivecontin-gency(i.e.,Group.60andGroup.30)maybetakentoreecttheDelta-rule.Givenequalprobabilitiesofaneectgivenaresponse(E|R),increasingtheprobabilityofaneectintheabsenceofaresponseR)seemstoworkagainstcreatinganassociationbetweenactionandeect,sothatacquisition-consistencyeectsde-crease.EventhoughGroup.30showsnoconsistencyeect,themeancausalityjudgmentinthisgroupisstillquitehighanddoesnotdierfromthatofGroup.60.Apparently,underpositiveDelta-contingency,theparticipantsperceiveeecttonestobemoredependentontheiractionsthanundernon-contingencyandlowoverallfrequencyoftheeect(i.e.,inGroup.00/.50andGroup.00/.20).Thus,theparticipantsoffouroutofourvegroupswereabletojudgetherelativerelationshipbetweenthepresenceorabsenceofaneectgiventhepresenceorabsenceofaresponsequiteappropriately,andthisperceptionmirrorstheacquisitionofaction-eectassociations.IfGroup.00/.80isexcludedfromtheanalysis,theimpressionmayindeedbegiventhatpar-ticipantsperformanceinthetestphaseisbasedontheHowever,theresultsofthenon-contingentGroup.00/.80indicatethattheparticipantsofthisgrouphaveacquiredsucientlystrongaction-eectassociationsinthelearningphase,inspiteofDelta-non-contingency,thusproducingareliableconsistencyeectinthetestphase.Thisisalsoreectedbythecausalityjudgmentofthisgroup,whichisthenumeri-callyhighestofallthegroups.Theparticipantsofthisgroupdidnotdetectthenon-contingencybutinsteadhadtheimpressionofactivelyproducingthetones.ButifthebehaviorofGroup.00/.80isnotconsistentwiththeDelta-rule,whatotherlearningprocessesmayitreect?AccordingtoChatloshetal.(1985),theoverallfrequencyoftheoccurrenceofaneect(E)hasanimpactonparticipantsbehaviorundernon-contingentconditions.Indeed,Group.00/.80experiencedthehighest(E)rateinExperiment2(cf.Table2).In80%ofthelearningtrials,oneoftheeecttonesappearedwhetheranactionwasperformedornot.InGroup.30,aneectappearedin65%ofthelearningtrials,inGroup.60andGroup.00/.50in50%,andinGroup.00/.20in20%.Mostprobably,thehighoverallfrequencyoftheeectinGroup.00/.80ledtotheacquisitionofstrongaction-eectassociationsinthelearningphasethatmediatedthehighcausalityjudg-mentsandthereliableRTeectsinthetestphase.In-deed,ifGroup.30isexcludedfromtheanalysis,theimpressionmaybegiventhatparticipantsperformanceinthetestphaseisbasedontheoverallfrequencyoftheeect,orTheimpactof(E)maybeduetothefactthatassociationsbetweenkeypressesandeectsarenot Fig.6Experiment2:Relationbetweenindividualperceptionsofcausalitybetweenkeypressesandtheirauditoryeects(meanpercentage,averagesacrossthefourkeys)andthesizeoftheacquisition-consistencyeect(meanRTininconsistentblockminusmeanRTinconsistentblock).Numbersrefertothevecontingencygroups:Group.60;Group.30;Group.00/.80;Group.00/.50;Group.00/.20 onlylearnedinthegotrials,butalsointheno-gotrials.Thiscouldhappeniftheparticipantsintheno-gotrialsplannedtheresponsetoacertaindegree,sothattheeecttonecouldbeintegratedwiththeactionplan.SuchlearningwasdemonstratedbyZiesslerandNattkemper(2002)inanexperimentinwhichpartici-pantshadtoprepareamovementtoawhitestimulusS1,buttopostponemovementexecutionuntilS1changeditscolortoyellow(GOsignal).Inhalfofthetrials,S1turnedred(NOGOsignal),andthepreparedmovementwasnottobeperformed.Inallgoandno-gotrials,eectletterswerepresented.Theresultsshowedclearresponse-eectlearningintheno-gotrialswhenthetimebetweentheonsetofS1andthecolorchangewaslongenoughtoallowfullresponseprepa-ration.However,thislearningeectwasdramaticallyreducedwithapreparationtimeofonly100ms.Thus,intheno-gotrials,participantsrelatedtheeectstotheprepared,butnotexecuted,response,butonlywhentherewaselaborateresponseplanning.Inthepresentstudy,theno-gosignalappearedatthesametimeastheimperativestimulus,anditthusseemsunlikelythatparticipantshavepreparedkeypressesintheno-gotrials.Still,thesetrialsmayhavecontributedtoaction-eectlearninginthenon-contingentgroups,leadingtotheimpactofAdditionally,theparticipantstendencytorelyontheoverallfrequencyoftheeectmaybefueledbythehighattentionalandmemorydemandsofthetaskofmoni-toringthecontingenciesoffourkeypressesandfoureects.Althoughwetriedtofacilitatethedetectionofcontingenciesbyintroducingthefouraction-eectpairsstepbystep,thetaskmaystillhavebeentoodicultfortheparticipants,leadingthemtoshifttosimplerprocessessuchasthoseinthestudyofJenkinsandWard(1965).Ifthefactthatparticipantsusuallyexperiencenon-contingentconditionsasmoredemand-ingthanpositivecontingentconditionsistakenintoaccount(Allan,1980;Allan&Jenkins,1980;Seggie,1975),theresultsofExperiment2resemblethoseofShakleeandMims(1982).Especiallyundermoredemandingnon-contingencies,theparticipantsconcen-tratedontheoverallfrequencyoftheeecttodetectthedependenciesbetweenactionandeect.Takentogether,theresultsofExperiment2indicatethataction-eectlearningisinuencedbytheoccurrenceoftheeectintheabsenceofanaction.However,thequestionwhetherparticipantsrelyonthecalculationofDelta-oronanevaluationoftheoverallfrequencyoftheeectwhendeterminingthedependenciesofactionsandeectscallsforfurtherinvestigation.GeneraldiscussionThepresentstudywasconductedtotestthehypothesisthatthelearningofrelationshipsbetweenactionsandtheirperceptualconsequencesisaccomplishedbyasso-ciativelearningprocesses.ThisisacentralassumptionintherststageofElsnerandHommels(2001)two-stagemodeloftheemergenceofactioncontrol.Thebasicideaisthatperforminganactionrequiresanactionplanthatconsistsofcodesthatspecify,amongotherthings,thatscharacteristics(Hommel,1997;Stoet&Hommel,1999).Ifactivationofthesecodesfrequentlyoverlapsintimewithactivationofcodesrepresentingtheperceivedconsequencesoftheaction,action-relatedandeect-re-latedcodesbecomeassociatedwitheachotherduetoasortoftraceconditioning(seeHommel,2003).Ifso,learningshoulddependonthetemporalcontiguityandtheprobabilisticcontingencyofactionandeect,becausethesefactorsdeterminetheactivationoverlapoftheMoreover,themodelclaimsthattheemergingasso-ciationsarebi-directional,sothatbeingexposedtoanacquiredactioneectwillprimetheassociatedaction.Asthedegreeofresponseprimingshouldvarywiththestrengthoftheacquiredassociation,theamountofprimingshoulddecreasewithdecreasingaction-eectcontiguityandcontingency.Andthisiswhatthepresentexperimentsshow.Evidenceofresponsepriming,asindicatedbyacquisition-consistencyeects,wasmostpronounced:1.Whenactionandeectwereseparatedbylessthan2.Whentheeectonlyrarelyoccurredintheabsenceoftheaction3.WhentheoverallfrequencyofaneectwashighThus,ourresultsindicatethatthefrequentco-occurrenceofanactionandaneectisacriticalfactorfortheacquisitionofaction-eectknowledge,andhenceforthedegreetowhichthisknowledgeaectstheplanningandexecutionoffutureactions.Somemorespecicimplicationsofourndingsrelatetothemechanismofaction-eectacquisitionandtheutilizationofaction-eectknowledge,whichwenowdiscussinturn.Themechanismofaction-eectacquisitionWeassumethatbeingexposedtothefrequentco-occurrenceofanactionandafollowingsensoryeventinducesthecreationandincrementalstrengtheningofanassociationbetweenthemotorpatterndrivingtheactionandthecodesrepresentingtheperceivedevent.Notethat,takeninisolation,thepresentdatadonotunequivocallyprovethisassumptiontobecorrect.Inparticular,ourdatadonotaddresstheissueofhowdirecttheassociationbetweenactioncodesandeectcodesreallyis.Forinstance,theaction-contingenttonesmaywellhavebeenassociatedwiththearrowstimuliprecedingtheminthelearningphase(astimulus-eectassociation),whichagainbecameassociatedwiththeactionstheysignaled(stimulus-responseassociation). Then,inthetestphase,presentingatonemayhavere-trievedthememoryofthecorrespondingarrow(asar-rowsdidnotappearinthetestphase),whichagainprimedtheassociatedresponse.Hence,theconsistencyeectsweobservedmayreecttheoperationofE-S-RassociationchainsratherthanthemoredirectE-Rassociationsweassume.Alternatively,theemergingassociationsmaynothavelinkedeectcodesandactioncodesdirectlybut,instead,consistofaclusterofasso-ciationsbetweeneecttonesandresponselocations,andresponselocationswiththeresponsesproper.Eventhoughthesepossibilitiescannotberuledoutonthebasisofthepresentresultswedonotconsiderthistoundermineourarguments.Firstofall,anyofthosepossiblescenariosassumesthat,onewayortheother,actionsandtheireectsdobecomepartofthesamecognitivestructure,whichisexactlywhatwehaveclaimedhereandelsewhere(Elsner&Hommel,2001;Hommel,1997).Howcomplexthesestructuresareandwhichelementstheyincludeisaninterestingissue,butitdoesnotaectthegeneralideathatactionsandeectsareintegrated.Secondly,thereisnoevidencethatusingthesametargetstimuliinthelearningphaseandthetestphasewouldincreasetheimpactofactioneects;ifanything,theoppositeseemstobetrue(cf.Hommel,1996,andElsner&Hommel,2001).Inourview,thisatleastrulesoutanaccountintermsofE-S-Rassocia-tions.Thirdly,wehaveprovidedneurophysiologicalevidencethatmerelyperceivingalearnedactioneectinataskthatrequiresnoresponseleadstotheactivationofthehumansupplementarymotorarea(SMA;Elsneretal.,2002).Moreover,theonlyotherbrainareawhoseactivitycovariedwiththefrequencyofpreviousactioneectswastherighthippocampus,whichdoesnotsup-porttheideathattheretrievalofthe(inthiscasevisual)targetstimulifromthelearningphaseorofthelocationsoftheresponsekeyswerenecessarymediatorsofactionpriming.Allinall,wethinkthatitmakessensetoas-sumethatperceivinganaction-eectsequenceresultsintheintegrationofaction-relatedandeect-relatedcodesintoastructurethatallowsforbi-directional(i.e.,ac-eectandeectaction)priming.Buthowmightsuchstructuresemerge?Inmoregen-eralterms,Hebb(1949)proposedthatthefrequentco-activationoftwoormoreneuralstructuresleadstochangesinthesynapticalconnectionsbetweenthem,whichagainarethoughttofacilitatethetransmissionofelectricalimpulsesfromonetotheother.Appliedtoac-tion-eectlearningthiswouldimplylinkingneuralpat-ternsinmotorareastoneuralpatternsofsensoryareasinthebrain.Althoughadirectassociationofprimarymotorcodestoprimarysensorycodesseemsunlikely,thereisincreasingevidencethatcertainbrainareasintegratesensoryandmotorcodes,suchastheintraparietalsulcus(Murata,Gallese,Luppino,Kaseda,&Sakota,2000)ortheareascontainingmirrorneurons,e.g.theprefrontalcortex(Gallese,Fadiga,Fogassi,&Rizzolatti,1996).TogetherwiththeobservationsofElsneretal.(2002),thesendingsbolstertheviewthatactionsandtheireectsareintegratedintosensorimotor(orperception-action)structuresthatcanbecharacterizedasactionconcepts(Elsner&Hommel,2001;Hommel,1997).Ifintegrationdependsonoverlappingactivationofperception-andaction-relatedcodes,itshouldbecon-strainedbycodedecay.Inthecontextofthisassump-tion,theresultsofExperiment1canbetakentoindicatethattheactivationofaction-relatedcodesdecaysratherquickly.Indeed,afteronly2stheactivationofthecodesofamanualkeypressseemstohavedecreasedtoalevelthatnolongercreatesorstrengthensaction-eectasso-ciations,atleastnottoadegreethatwouldinuenceperformanceonalateroccasion(i.e.,inthetestphase).ThisobservationtswiththendingsofStoetandHommel(1999),whichshowedthatmaintaininganac-tionplanimpairstheplanningofanotherfeature-over-lappingaction.However,theinterferencedisappearslessthan1saftertheexecutionofthemaintainedplan,whichsuggeststhatthecodesmakinguptheplan(andproducingtheinterference)decayedtoalevelthatdoesnotinuenceconcurrentprocessesanylonger.Inotherwords,usedactionplansareshort-lived.TheassumptionofaHebbianlearningmechanismstressestheimportanceoftemporalcontiguityandprobabilisticcontingencyforaction-eectlearning.Asthereisevidenceofassociativeaction-eectlearningnotonlyinadults,butalsoinhumaninfants(seeElsner&Aschersleben,inpress;Gergely&Watson,1999),cats(Brogden,1962),rats(Meck,1985;Rescorla,1992),andpigeons(Urcuioli&DeMarse,1996),theunderlyinglearningmechanismisunlikelytobeoverlycomplexortoputhighdemandsonmemorycapacities.Ifso,itmaybedoubtedwhetherthecalculationofrelativeprobabilitiesaccordingtotheDelta-rulewhichrequiresthestorageofinformationaboutalloccurrencesandnon-occur-rencesofanactionanditseectsisareasonablecan-didatetounderlieaction-eectlearning.Incontrast,contiguity-basedlearningwouldonlyrequirethatthecognitivesystemregisterstheco-occurrenceoftwoeventson-line,andautomaticallyformsanassociationbetweentheircodes.Byincreasingorupdatingthestrengthoftheemergingassociationateverysubsequentexperienceoftheco-occurrence(ornon-occurrence)ofthesameevents,randomco-occurrencesofanactionandaneventmayeventuallybelteredout.However,thedatareportedheredonotprovidesucientevidenceforanalanswertotheissuewhetheraction-eectlearningisbasedontemporalcontiguityoronprobabilisticcontingency.Theutilizationofaction-eectknowledgeThepresentstudywasmainlyconcernedwiththerststageofElsnerandHommels(2001)two-stagemodel,i.e.,withthequestionofhowknowledgeaboutactionsandtheireectsislearned.Inordertodemonstratethatlearningactuallytookplace,wemadearatherstrongassumption,namely,thatpresentinganeectstimulustosomeonewhohadacquiredaparticularaction-eect associationprimestheassociatedactiontoadegreethatfacilitatestheselectionofthecorrespondingactionand/orhamperstheselectionofanotheraction.Onlyifthisstrongassumptionholds,acquisition-consistencyeectscanbeexpectedtoshowupintheRTsofthetestphase.Ontheonehand,thereareacoupleofreasonstobelievethatactionscanindeedbeactivatedbypresent-ingpeoplewiththeperceptualeectsofthataction.Forinstance,Hommel(1996)instructedparticipantstoperformchoicereactionstovisualstimuliwhilebeingpresentedwithpreviouslylearnedauditoryactionef-fects.Eventhoughtheauditorystimuliwereentirelyirrelevanttothetask,performancewasbetteriftheyhadpreviouslybeenassociatedwiththecurrentlycorrectkeypressthanwiththealternativekeypress.Likewise,Beckersetal.(2002)foundthatthetask-irrelevantaectivevalueofstimuluswordsfacilitatestheselectionofresponsesthathadpreviouslybeenfollowedbyavalue-congruentelectrocutaneousfeedback(presenceorabsenceofamildelectro-shock).ElsnerandHommel(2001,Experiments2 4)instructedparticipantstoper-formfree-choiceresponsestimedbyatonesignalthathadbeenpresentedasanactioneectinapreviouspartofthesession.Eventhoughparticipantswereencour-agedtoignorethepitchofthetones,theyweremorelikelytochoosethemovementthathadpreviouslyproducedthetone.TheperhapsmostdirectevidenceoftheactivationofactiontendencieswasobtainedbythestudybyElsneretal.(2002)alreadydiscussed,wherethepresentationoflearnedactioneectsinalisteningtaskwasfoundtoincreaseactivationoftheSMA.Ontheotherhand,however,thereissomeevidencethatacquisition-consistencyeectsarenotnecessarilyapuremeasureofthepresenceandstrengthofaction-eectassociations.Onepieceofevidencecomesfromourownobservationthatindividualcausalityjudg-mentsarenotavalidpredictoroftheindividualsizeoftheacquisition-consistencyeect.Firstofall,thisimpliesthatthesetwomeasuresdonotreectexactlythesamething.Indeed,itwouldbeaplausibleassumptionthatexperiencingthefrequentoccurrenceofanactionandaneectinducesanassociationbe-tweenaction-andeect-relatedcodes,andthatthisassociationdrivesboththecausalityjudgmentsandtheresponseprimingresponsiblefortheconsistencyeect.Giventheclosecorrespondenceofndingsfrominstrumentalconditioninginanimalsandcausaljudg-mentsofaction-eectrelationsinhumans(Shanks&Dickinson,1987;Wasserman,1990;Young,1995)itisinfactnotunreasonabletoassumethatthestrengthofaction-eectassociationsandthedegreeofcausalityattributedtothecorrespondingaction-eectrelationaremirror-imagesofeachothereventhoughlearningtheoristsandcognitiveresearchersmayargueaboutthecausalrelationbetweenthesetwomeasures(i.e.,whetherjudgmentsaredrivenbyassociationsorwhetherlearningdependsoncausalperception).Butifso,thediscrepancybetweenindividualcausaljudg-mentsandconsistencyeectswouldsuggestthattheformerrepresentapurermeasureofassociationstrengththanthelatter.Theavailableevidencesug-geststwo,innowayexclusivereasonsforwhythismaybethecase.Onehastodowithinter-individualdierencesinrelativeandabsoluteprocessingspeed.Hommel(1996)studyprovidesevidencethatthedegreetowhichanirrelevantaction-eectstimulusaectsresponseperformancetoatargetstimulusdependsontherela-tiveprocessingspeedforthefeaturesofthetwostimuli.Firstly,primingwasincreasediftheactioneectstimulusprecededthetargetstimulusby100ms(Experiment5b),suggestingthatpriminganactiontakessometime.Secondly,primingwasalsoincreasedbydecreasingthediscriminabilityoftargetstimuli(Experiment5a).Asthismanipulationmadethetaskmoredicultand,thus,increasedtheRTlevel,itcanbeassumedthatresponsepriminghadmoretimetounfoldbeforeresponseselectionwascompleted.In-deed,thirdly,analysesoftheRTdistributionsrevealedthattheprimingeectwasrestrictedtotheslowtailofthedistribution.Thersttwooftheseobservationsarebasedondatathatwereaveragedacrossparticipants,buttheirprocessinglogicalsoappliestointer-individ-ualdierences.Mostimportantly,thetimeneededtoprocessstimuliofparticularmodalitiesortocompleteresponseselectionmaydierbetweenindividuals,whichwouldintroducevarianceintheintervalinwhichresponse-primingcanunfold.Becausethesefactorsarelikelytoinuencethesizeofacquisition-consistencyeectsbutnotthejudgmentsofcausalitybetweenac-tionandeects,theymayprovideatleastpartofanexplanationwhythelatterareimperfectpredictorsoftheformer.Asecondreasonwhyacquisition-consistencyeectsmaynotalwaysmirrorthestrengthofaction-eectassociationshastodowiththecurrenttaskgoals.ConsiderthestudyofHommel(1993),whoinvestigatedtheimpactofirrelevantspatialstimuluslocationonthespeedofselectingspatialresponses(i.e.,theso-calledSimoneect).Underthestandardinstructiontopressaleftandarightkeyinresponsetoaparticularstimulus(e.g.,tothecolorofavisualpatch),peopleareknowntoshowbetterperformancewhenthespatiallocationsofstimulusandresponsekeycorrespond(e.g.,Simon&Rudell,1967).Hommel(1993)rstshowedthatthisisalsothecaseifthetwokeysproducelightashesontheoppositeside,i.e.,ifpressingaleftkeyproducesaright-sideashandviceversa.However,whenthesametaskwascarriedoutundertheslightlydierentinstructiontoashtherightandleftlightinresponsetothetargetstimulus,performancewasbestifthelocationsofstimulusandlightcorrespondedeventhoughthisim-pliednon-correspondenceofstimulusandresponsekey.Thismeansthattaskintentionscanselectparticularactioneects(here:referringtokeyorngerlocationversuslightlocation)andtherebyinuencethewayagivenactioniscognitivelycodedinagivensituation.Thisdoesnotnecessarilyeliminatenon-selectedaction eectsfromactioncontrolentirelybutitdrasticallyre-ducestheirimpactonovertbehavior(Hommel,1993,1996).ThisassumptionisfosteredbytherecentndingsofHommel(2003b).Heinstructedparticipantstorespondmanuallytothecolorofwordswithacongruentorincongruentmeaning(i.e.,Stroop-typestimuli).Intwogroups,eachresponseproducedavisualpatchofaparticularcoloronascreen.Inonegroup,theeectwasalwaysofthesamethecolorasthestimulus,intheothergroup,theeectwasofadierentcolor.Inathirdgroup,theresponsesproducednoeects.Iftheimpactofactioneectsonresponseselectionwereentirelyautomatic,thebenetofthecompatible-map-pinggroupovertheno-eectgroupshouldbemirroredbyacomparabledisadvantageintheincompatible-mappinggroup.However,overallperformanceinthecompatible-mappinggroupwasbetterthanintheothergroups,whichdidnotdier.Becausethecompatible-mappinggroupwastheonlygroupinwhichthestimulusandeectcolorwereidentical,themembersofthisgroupmayhaveprotedfromintentionallychangingthecodingoftheiractionsfromlocationtocolor.Thus,forexample,theyperformedtheresponseafterthestimulus,whereastheothergroupsperformedtheresponse(whichcausedablueeectornoeect).Accordingtothisinterpreta-tion,thecurrentintentiondetermineswhichactionef-fectsareselectedtocognitivelyrepresentaparticularaction.Ifso,theautomaticimpactofactioneectsthatareunrelatedtothecurrenttaskgoalisnecessarilyweak,anditmaybeunreliableorevenabsentinovertbehavior.Moreover,theautomaticimpactmayvarydependingontheparticipantsstrategies,onthesub-jectivesalienceoftheactioneects,andsoforth.Inotherwords,acquisitionofactioneectsmaywellbeautomatic,buttheimpactofacquiredactioneectsonactioncontrolmaydependonsubjectiveutility,whichagaindependsontherelevanceoftheactioneectstothetaskathand.Toconclude,thetwoexperimentsofthisstudypro-videevidencethataction-eectacquisitiondependsonthetemporalproximityofactionandeectandonthecontingencyoronthefrequencyoftheirco-occurrence.Theseobservationsandtheircorrespondencewithnd-ingsfrominstrumentalconditioninginanimalsandcausalperceptioninhumanssupporttheassumptionthatacquiringactioneectsisaccomplishedbywell-knownassociativelearningmechanisms.However,whetherornottheemergingaction-eectassociationsproduceameasurableimpactonactioncontrolislikelytodependontaskintentionsandonindividualprocessingcharacteristics.ThisresearchwassupportedbyagrantoftheDeutscheForschungsgemeinschaft(CollaborativeResearchCenter462:SensorimotorProcesses).WethankKathrinMuMarkStaiger,andSilkeWendelsforrunningtheexperiments,andFiorelloBanciandKarl-HeinzHonsbergfortheirassistanceinconstructingtheapparatus.Allan,L.G.(1980).Anoteonmeasurementofcontingencybe-tweentwobinaryvariablesinjudgmenttask.BulletinofthePsychonomicSociety,147 149.Allan,L.G.(1993).Humancontingencyjudgments:RulebasedorPsychologicalBulletin,435 448.Allan,L.G.&Jenkins,H.M.(1980).Thejudgmentofcontingencyandthenatureoftheresponsealternatives.CanadianJournalof,1 11.Beckers,T.,DeHouwer,J.,&Eelen,P.(2002).Automaticin-tegrationofnon-perceptualactioneectfeatures:ThecaseoftheassociativeaectiveSimoneect.PsychologicalResearch/PsychologischeForschung,166 173.Brogden,W.J.(1962).Contiguousconditioning.JournalofExperimentalPsychology,172 176.Chatlosh,D.L.,Neunaber,D.J.,&Wasserman,E.A.(1985).Response-outcomecontingency:Behavioralandjudgmentaleectsofappetitiveandaversiveoutcomeswithcollegestu-LearningandMotivation,1 34.Davey,G.C.L.(1983).Anassociativeviewofhumanclassicalconditioning.InG.C.L.Davey(Ed.),Animalmodelsofhuman(pp.95 114).Chichester:Wiley.Dickinson,A.,&Charnock,D.J.(1985).Contingencyeectswithmaintainedinstrumentalreinforcement.QuarterlyJournalofExperimentalPsychology,397 416.Dickinson,A.,&Shanks,D.R.(1985).Animalconditioningandhumancausalityjudgment.InL.-G.Nilsson&T.ArcherPerspectivesonlearningandmemory(pp.167 191).Hillsdale,NJ:Erlbaum.Einhorn,H.J.,&Hogarth,R.M.(1986).Judgingprobablecause.PsychologicalBulletin,3 19.Elsner,B.,&Aschersleben,G.(inpress).DoIgetwhatyouget?Learningabouttheeectsofself-performedandobservedac-tionsininfancy.Consciousness&CognitionElsner,B.,&Hommel,B.(2001).EectanticipationandactionJournalofExperimentalPsychology:HumanPerceptionandPerformance,229 240.Elsner,B.,Hommel,B.,Mentschel,C.,Drzezga,A.,Prinz,W.,Conrad,B.,&Siebner,H.R.(2002).Linkingactionsandtheirperceivableconsequencesinthehumanbrain.Gallese,V.,Fadiga,L.,Fogassi,L.,&Rizzolatti,G.(1996).Actionrecognitioninthepremotorcortex.,593 609.Gergely,G.,&Watson,J.S.(1999).Earlysocial-emotionalde-velopment:Contingencyperceptionandthesocialbiofeedbackmodel.InP.Rochat(Ed.),Earlysocialcognition(pp.101 136).Hillsdale,NJ:Erlbaum.Grice,G.R.(1948).Therelationofsecondaryreinforcementtodelayedrewardinvisualdiscriminationlearning.JournalofExperimentalPsychology,1 16.Gruber,H.E.,Fink,C.D.&Damm,V.(1957).Eectsofex-perienceonperceptionofcausality.JournalofExperimental,89 93.Haggard,P.,Aschersleben,G.,Gehrke,J.,&Prinz,W.(2002).Ac-tion,binding,andawareness.InPrinz,W.,&Hommel,B.(Eds.),Commonmechanismsinperceptionandaction:Attention&Per-formanceXIX(pp.266 285).Oxford:OxfordUniversityPress.Hammond,L.J.(1980).Theeectofcontingencyupontheappetitiveconditioningoffreeoperantbehavior.JournaloftheExperimentalAnalysisofBehavior,297 304.Hazeltine,E.(2002).Therepresentationalnatureofsequencelearning:Evidenceforgoal-basedcodes.InW.Prinz&B.Hommel(Eds.),Commonmechanismsinperceptionandaction:Attention&PerformanceXIX(pp.673 689).Oxford:OxfordUniversityPress.Hebb,D.O.(1949).Theorganizationofbehavior.NewYork:Wiley.Homann,J.,Sebald,A.,&Stoecker,C.(2001).Irrelevantre-sponseeectsimproveseriallearninginserialreactiontimeJournalofExperimentalPsychology:Learning,Memory,&Cognition,470 482. Hommel,B.(1993).InvertingtheSimoneectbyintention:Determinantsofdirectionandextentofeectsofirrelevantspatialinformation.PsychologicalResearch,270 279.Hommel,B.(1994).Spontaneousdecayofresponse-codeactiva-PsychologicalResearch,261 268.Hommel,B.(1996).Thecognitiverepresentationofaction:Auto-maticintegrationofperceivedactioneects.PsychologicalRe-,176 186.Hommel,B.(1997).Towardanaction-conceptmodelofstimulus-responsecompatibility.InB.Hommel&W.Prinz(Eds.),Theoreticalissuesinstimulus-responsecompatibility(pp.281 320).Amsterdam:North-Holland.Hommel,B.(1998).Perceivingonesownactionandwhatitleadsto.InJ.S.Jordan(Ed.),Systemstheoriesandaprioriaspectsof(pp.143 179).Amsterdam:North-Holland.Hommel,B.(2003a).Acquisitionandcontrolofvoluntaryaction.InS.Maasen,W.Prinz&G.Roth(Eds.),Voluntaryaction:Brains,minds,andsociality(pp.34 48).Oxford:OxfordUni-versityPress.Hommel,B.(2003b).Coloringanaction:IntendingtoproducecoloreventseliminatestheStroopeect.PsychologicalResearch(inpress).Hommel,B.,Alonso,D.,&Fuentes,L.(2001).Acquisitionandgeneralizationofactioneects.Talkpresentedatthe12thConferenceoftheEuropeanSocietyforCognitivePsychology.Edinburgh,Scotland,September2001.Hume,D.(1739/1964).Treatiseofhumannature.London:OxfordUniversityPress.James,W.(1890).Theprinciplesofpsychology.NewYork:Dover.Jenkins,H.M.,&Ward,W.C.(1965).Judgmentofcontingencybetweenresponsesandoutcomes.PsychologicalMonographs:GeneralandApplied,1 17.Kaplan,P.S.,&Hearst,E.(1982).BridgingtemporalgapsbetweenCSandUSinautoshaping:Insertionofotherstimulibefore,during,andafterCS.JournalofExperimentalPsychology:AnimalBehaviorProcesses,187 203.Kunde,W.,Homann,J.,&Zellmann,P.(2002).Theimpactofanticipatedactioneectsonactionplanning.ActaPsychologi-,137 155.Meck,W.H.(1985).Postreinforcementsignalprocessing.ofExperimentalPsychology:AnimalBehaviorProcesses,52 MurataA.,GalleseV.,LuppinoG.,KasedaM.,&SakataH.(2000).Selectivityfortheshape,size,andorientationofobjectsforgraspinginneuronsofmonkeyparietalareaAIP.Journalof,2580 2601.Pavlov,I.P.(1927).Conditionedreexes.London:OxfordUni-versityPress.Piaget,J.(1952).Theoriginsofintelligenceinchildren.NewYork:InternationalUniversitiesPress.Prinz,W.(1997).Perceptionandactionplanning.EuropeanJournalofCognitivePsychology,129 154.Reed,P.(1992).Eectofasignaleddelaybetweenanactionandoutcomeofhumanjudgementofcausality.QuarterlyJournalofExperimentalPsychology,81 100.Reed,P.(1999).Roleofastimulusllinganaction-outcomedelayinhumanjudgmentsofcausaleectiveness.JournalofExperi-mentalPsychology:AnimalBehaviorProcesses,92 102.Reed,P.,&Reilly,S.(1990).Contextextinctionfollowingcon-ditioningwithdelayedrewardenhancessubsequentinstru-mentalresponding.JournalofExperimentalPsychology:AnimalBehaviorProcesses,48 55.Rescorla,R.A.(1967).Pavlovianconditioninganditspropercontrolprocedures.PsychologicalReview,71 80.Rescorla,R.A.(1982).EectofastimulusinterveningbetweenCSandUSinautoshaping.JournalofExperimentalPsychology:AnimalBehaviorProcesses,131 141.Rescorla,R.A.(1992).Response-outcomevs.outcome-responseassociationsininstrumentallearning.AnimalLearningand,223 232.Rochat,P.(1998).Self-perceptionandactionininfancy.mentalBrainResearch,102 109.Seggie,J.(1975).TheempiricalobservationofthePiagetianconceptofcorrelation.CanadianJournalofPsychologySeligman,M.E.P.,Maier,S.F.,&Solomon,R.L.(1971).Un-predictableanduncontrollableaversiveevents.InF.R.BrushAversiveconditioningandlearning(pp.347 400).SanDiego,CA:AcademicPress.Shaklee,H.,&Mims,M.(1982).Sourcesoferrorinjudgingeventcovariations:Eectsofmemorydemands.JournalofExperi-mentalPsychology:Learning,Memory,andCognition,208 Shaklee,H.,&Wasserman,E.A.(1986).Judgingintereventcon-tingencies:Beingrightforthewrongcauses.BulletinofthePsychonomicSociety,91 94.Shaklee,H.,Holt,P.,ElekS.,&Hall,L.(1988).Covariationjudgment:Improvingruleuseamongchildren,adolescents,andChildDevelopment,755 768.Shanks,D.R.(1989).Selectionalprocessesincausalityjudgment.MemoryandCognition,27 34.Shanks,D.R.(1993).Humaninstrumentallearning:Acriticalreviewofdataandtheory.BritishJournalofPsychologyShanks,D.R.,&Dickinson,A.(1987).Associativeaccountsofcausalityjudgment.ThePsychologyofLearningandMotivation,229 261.Shanks,D.R.,&Dickinson,A.(1991).Instrumentaljudgmentandperformanceundervariationsinaction-outcomecontingencyandcontiguity.MemoryandCognition,353 360.Shanks,D.R.,Pearson,S.M.,&Dickinson,A.(1989).Temporalcontiguityandthejudgmentofcausalitybyhumansubjects.QuarterlyJournalofExperimentalPsychology,139 159.Simon,J.R.,&Rudell,A.P.(1967).AuditoryS-Rcompatibility:Theeectofanirrelevantcueoninformationprocessing.JournalofAppliedPsychology,300 304.Stock,A.,&Homann,J.(2002).Intentionalxationofbeha-viourallearning,orhowR-OlearningblocksS-Rlearning.EuropeanJournalofCognitivePsychology,127 153.Stoet,G.,&Hommel,B.(1999).Actionplanningandthetemporalbindingofresponsecodes.JournalofExperimentalPsychology:HumanPerceptionandPerformance,1625 1640.Thorndike,E.L.(1927).Thelawofeect.AmericanJournalof,212 222.Tolman,E.C.,Hall,C.S.,&Bretnall,E.P.(1932).Adisproofofthelawofeectandasubstitutionofthelawsofemphasis,motivationanddisruption.JournalofExperimentalPsychology,601 614.Trapold,M.A.(1970).Areexpectanciesbasedupondierentreinforcingeventsdiscriminablydierent?LearningandMoti-,129 140.Urcuioli,P.J.,&DeMarse,T.(1996).Associativeprocessesindierentialoutcomediscriminations.JournalofExperimentalPsychology:AnimalBehaviorProcesses,192 204.Waldmann,M.R.,&Holyoak,K.J.(1992).Predictiveanddiag-nosticlearningwithincausalmodels:AsymmetriesincueJournalofExperimentalPsychology:GeneralWalker,E.L.(1969).ReinforcementTheonering.InJ.T.Tapp(Ed.),Reinforcementandbehavior(pp.47 62).NewYork:AcademicPress.Wasserman,E.A.(1990).Detectingresponse-outcomerelations:Towardanunderstandingofthecausaltextureoftheenvi-ThePsychologyofLearningandMotivation,27 Wasserman,E.A.,&Miller,R.R.(1997).Whatselementaryaboutassociativelearning?AnnualReviewofPsychologyWasserman,E.A.,&Shaklee,H.(1984).Judgingresponse-out-comerelations:Theroleofresponse-outcomecontingency,outcomeprobability,andmethodofinformationpresentation.MemoryandCognition,270 286.Young,M.E.(1995).Ontheoriginofpersonalcausaltheories.PsychonomicBulletin&Review,83 104. Ziessler,M.(1998).Response-eectlearningasamajorcomponentofimplicitseriallearning.JournalofExperimentalPsychology:Learning,Memory,&Cognition,962 978.Ziessler,M.,&Nattkemper,D.(2001).Learningofeventsequencesisbasedonresponse-eectlearning:Furtherevidencefromaserialreactiontask.JournalofExperimentalPsychology:Learning,Memory,&Cognition,595 613.Ziessler,M.,&Nattkemper,D.(2002).Eectanticipationinactionplanning.InW.Prinz&B.Hommel(Eds.),Commonmechan-ismsinperceptionandaction:Attention&PerformanceXIX(pp.645 672).Oxford:OxfordUniversityPress.