JournalofMemoryandLanguage
www.elsevier.com/locate/jml
Rethinkinginterferencetheory:Executivecontrol
andthemechanismsofforgetting
MichaelC.Anderson*DepartmentofPsychology,1227UniversityofOregon,Eugene,OR97403-1227,USA
Received21April2003;revisionreceived13August2003
Abstract
Interferenceprovidesanaccountofoneofthemostbasicproblemsinthescienceofmemory:forgetting.Histori-cally,theoriesofthisprocesswereshapedbymodelsofassociativelearningprevalentwheninterferenceresearchbegan.Inthisarticle,Iarguethatweshouldreconsiderthelong-standingconceptualizationofinterferenceasalearningphenomenonandreframeinterferenceasarisingfromsystemsthatachievementalandbehavioralcontrol.Specifically,itisarguedthatforgettingisnotapassivesideeffectofstoringnewmemories,butresultsfrominhibitorycontrolmechanismsrecruitedtooverrideprepotentresponses.Insupportofthisidea,Idiscusstwocontrolsituationsinwhichresponseoverrideisnecessary—selectionandstopping—andshowhowthesesituationshavedirectparallelsinretrieval.Ithenreviewevidencethatinbothofthesesituations,theneedtooverrideprepotent,distractingmemoriesissupportedbyinhibitorymechanismsthatultimatelycauseforgetting.Thetheoreticalpropertiesoftheseinhibitoryeffectsareoutlined,alongwithcriticalfactorsknowntomodulateormaskinhibition.Therelationbetweenthisexecutivecontroltheoryofforgettingandclassicalaccountsofinterferenceisdiscussed.Ó2003PublishedbyElsevierInc.
Overacenturyago,G.E.MuellerandPilzeckerre-portedoneofthefirstempiricaldemonstrationsoffor-€ller&Pilzecker,1900).Ingettingduetointerference(Mu
thisclassicwork,MuellerandPilzeckerfoundthatpeoplewerelesslikelytorecallamemoryitemifintheinterimtheretrievalcuethatwasusedtotestthatitemhadbecomeassociatedtoanothermemory.Theynamedthiseffectretroactiveinhibition,highlightingthemannerinwhichthestorageofnewexperiencesinterfereswithmemoriesencodedearlierintime.MuellerandPilzeckerbelievedthatthismemoryimpairmentoccurredbecausetheprocessofstoringnewmemoriesdisruptedtheconsolidationprocessthatwouldhaveordinarilystrengthenedthetracesthatsubjectshadacquiredear-lier.Bythisview,allwould-bememoriesperseverateforabriefperiodaftertheyareencoded,asevidencedbythe
Fax:1-541-346-4911.
E-mailaddress:mcanders@darkwing.uoregon.edu.
0749-596X/$-seefrontmatterÓ2003PublishedbyElsevierInc.doi:10.1016/j.jml.2003.08.006
*tendencyofrecentmemoriestopopintoconsciousnessunbiddenbyanyparticularcue.Perseverationwasthoughttobenecessarytomorefirmlyfixatraceintolong-termstorage.Ifanothereffortfulactivityintervened(suchaslearningasecondlistofitems),thepersevera-tiveprocessfortheearliermemorieswasthoughttobedampened,ultimatelypreventingthetracesfrombeingwovenintothefabricofmemory.
Althoughthedisruptedconsolidationtheorywaslargelyabandonedasanaccountofretroactiveinter-ference(seeMcGeoch&Irion,1952,forarguments),thephenomenonitselfandthemethodMuellerandPilzec-kerintroducedtostudyithaveplayedacentralroleinshapingthehistoryofmemoryresearch.Theirworksetofftheclassicalinterferenceera(1900–1970)inmemoryresearch.Inthisera,considerableenergywasdevotedtounravelingthemechanismsofinterference—afocusdeemedworthybecauseitaddressedthefundamentalproblemofforgetting.Howisitpossibleforanexperi-encethatisvividandlivelyinourmemoriestodayto
416M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
ultimatelyfade?Whydomemoriesgrowlessaccessibleovertime?Accordingtoclassicalinterferencetheory,MuellerandPilzeckerÕsdiscoveryprovidedananswer:itwasnotthepassageoftimethatcausedforgetting,asmightseemintuitive,butrather,changescorrelatedwithtime,suchasthestorageofnewexperiencesintomem-ory—inparticular,highlysimilarexperiences—thatwereattherootofmemoryfailures.Youcanrememberwhatyouhadfordinneryesterdayeveningnow,yetinafewmonthsyouwillnotbeableto,notbecausetimehaspassed,butbecausetheroutinenatureofourlivesulti-matelycluttersmemorywithmanyhighlysimilardinnerevents.Thiscluttermakesanyparticularmemoryverydifficulttoretrieve.Thus,whenweforget,itisnotbe-causememoriesdecay,butbecausewearevictimsoftheever-changingstructureofourmemoryandofbasiclimitationsinourabilitytodifferentiatesimilartraces.Thisviewhasstoodthetestoftime:after70yearsofresearchandaftertensofthousandsofpapersonthetopic,therecanbelittledoubtthatinterferenceisapowerfulcauseofforgetting.
Whatcanbedoubted,however,isthemannerinwhichinterferencecausesforgetting.Ononehand,for-gettingmaybeadirectconsequenceofaddingnewtracesintomemory.Bothclassicalandmoderntheorieshaveemphasizedthisapproach.Forinstance,McGe-ochÕsinfluentialresponsecompetitiontheory(McGeoch,1942)attributedinterferenceeffectstoheightenedcom-petitionarisingfromtheassociationofadditionaltracestoaretrievalcue(ortothestrengtheningofanexistingcompetitor);inhisframework,forgettingwasaconse-quenceofaddingnewassociativestructure.Moderntheoriessuchasthoseembodiedinrelativestrengthorratio-rulemodelsofretrieval(Anderson,1983;Mensink&Raajimakers,1988)aretheconceptualdescendantsofthisviewintheiremphasisonhowretrievalofagivenitemisimpededbycompetingassociations(seeAnder-son&Bjork,1994,forareview).Structure-basedtheo-riessuchasthesedonotrequirespecialmechanismsofforgettingandhavethevirtueofparsimony.Ontheotherhand,theyde-emphasizeabasicprobleminhowweuseourmemory:howdoweovercomeinterferencebetweencompetingtracestoretrievethememorywewant?Whataretherepercussions,ifany,ofresolvingcompetitionforthetracesthatinterfered?Giventhatourcognitivegoalsoftenrequiretherecallofspecificeventsinlong-termmemory,someprocessmustexistforresolvinginterference.
Inthisarticle,Ipresentaviewofhowinterferenceleadstoforgettingwhichemphasizeshowinterferencegetsre-solved.Iarguethatatheoryofinterferenceshouldbeframedinthelargercontextofhoworganismscontrolthedirectionoftheiractionsandthoughts.Bythisview,memoryretrievalpresentsaspecialcaseofabroadclassofsituationsthatrecruitexecutivecontrolprocesses;itistheexecutivecontrolmechanismthatovercomesinter-
ference—inhibition—thatcausesustoforget,notthecompetitionitself.Thisviewdepartsfromthecommonassumptionthatforgettingisapassivesideeffectoftheever-changingstructureofmemory.Themerestorageofinterferingtracesisnotwhatcausesmemoriestogrowlessaccessiblewithtime.Rather,forgetting,whetherincidentalorintentional,isproducedasaresponsetointerferencecausedbyactivatedcompetitorsinmemory.Ireviewtheevidencesupportingthisfunctionalview.Thisreviewfocusesonretrieval-inducedforgettingandmorerecentworkwiththethink/no-thinkparadigmandprovidesaspecificationofthetheoreticalpropertiesofthosephenomena,someoftheirboundaryconditions,andempiricalchallengestomeasuringinhibition.Inthefinalsection,Icontrasttheproposedviewwithclassicaltheoriesofinterference.Beforebeginning,however,Ielaborateonthetheoreticalperspectivead-vancedhere.
ExecutivecontrolandthemechanismsofretrievalThecurrentperspectivebeginswithasimpleobser-vationabouthumanbehavior:Actions,oncestarted,canusuallybestopped.Thissimplefactwasimpresseduponmeoneeveningwhileopeningthekitchenwindow.Asthewindowslidalongitstrack,itpushedasmallcactusofftheedgeofthesill.Myhanddartedreflexivelytocatchthefallingcactus.Merecentimetersfromit,IstoppedmyhandfromclutchingthecactusÕsneedle-densebody.Theplantdroppedtothefloorandwasruined,butIwashappytohaveavoidedpiercingmyhandwiththousandsoflittleredneedles.Thislastminutesavewasmadepossiblebymyabilitytotermi-natephysicalaction—anabilitysopervasivethatitgoesnearlyunnoticedindailylife.
Theprecedingcaseisaclassicexampleofasituationinwhichweneedtoovercomeastronghabitualre-sponse—asituationwidelyregardedasrequiringexecu-tivecontrol.Thisissometimesreferredtoasresponseoverride,andisillustratedinFig.1.Inresponseover-ride,onemuststopaprepotentresponsetoastimulus(suchasafallingobject).Thismayeitherbebecausethecircumstancerequiresthattheresponsebewithheld,orbecausealesscommonresponseismorecontextuallyappropriate.Forexample,itismorecontextuallyap-propriatetosay‘‘Hola’’whensomeonewavestoyouwhileyouareinSpain,eventhoughyourhabitualre-sponsemaybetosay‘‘Hello.’’Thecapacitytoeitherstoporredirectactioninthiswayiscrucialtodailylife.Withoutit,wewouldloseessentialflexibilitytoadaptbehavioraccordingtochangesinourgoals,ortochangesintheenvironmentitself.Wewouldbeslavestohabitorreflex.
Akeytheoreticalquestionthatthisproblemraisesis‘‘Howdowekeepfrombeingautomaticallycontrolled
bythehabitualaction?’’Onewidelydiscussedanswertothisquestionisthatresponseoverrideisaccomplishedbyinhibitingtheundesiredaction.Accordingtothisview,thepresentationofastimulusactivatesarepre-sentationcorrespondingtothatstimulusinlong-termmemory.Activationthenspreadsfromthatrepresenta-tiontoassociatedresponsesinproportiontohowstronglyassociatedtheyaretothestimulus.Whenaresponsebecomessufficientlyactivated,itwillbeemit-ted.Ifastimulusisassociatedtomultipleresponses,theonethatachievesthresholdmostquicklywillgenerallybeemitted,pre-emptingotherresponses.However,ifaweakerresponseismorecontextuallyappropriate,in-hibitioncanberecruitedtosuppressthestrongerone.Inhibitionisthoughttoreducethelevelofactivationforagivenresponse,preventingitfromachievingthreshold.Insodoing,thisprocesspermitsweaker,butmorecontextuallyappropriateresponsestobeexpressed,en-ablingflexible,context-sensitivebehavior.Thisisknownasinhibitorycontrol.
Giventheputativeimportanceofinhibitorycontrolindirectingovertbehavior,itisreasonabletoaskwhe-therinternalactionsmightalsobethetargetofsuchmechanisms.Clearparallelsexistbetweenthecontrolofactionandthecontrolofmemory.Justasastimulusmayspreadactivationtoaprepotentmotorresponse,aretrievalcuemayspreadactivationtoastronglyassociatediteminmemory,leadingittoberetrieved.
Theretrievalofassociatedmemoriesisnotalwaysde-sirable;sometimes,wemaywishtoretrieveanothermemoryitemthatisassociatedtothecuedrivingre-trieval,butthatismoreweaklyassociatedtothatcue;othertimes,wemaywishtoavoidretrievalaltogethereitherbecausetheassociatedmemoryisunpleasantorsimplybecausewewishtomaintainconceptualfocusontheconceptthatisactingasacue.Althoughwesome-timesretrievethingsthatwedonotintend,weoftenareabletoexercisecontroloverthistendency;wecanre-collecttheparticulareventweareseekingdespitein-terferencefromstrongercompetitors,andwecanstopourselvesfromthinkingaboutunwantedmemories.Giventhesefunctionalparallelsbetweenmotorbehaviorandmemoryretrieval,itispossiblethatresponseover-ridemechanismsarerecruitedtocontrolunwantedmemoriesthatintrudebyvirtueofspreadingactivation(forrelatedarguments,seeShimamura,1995).Ifso,weshouldfindevidenceforinhibitorycontrolinmemorysituationslikelytoinvolveresponseoverride—situationssuchastheneedtoselectaweaker,yetmorecontextu-allyappropriateresponse,giveninterferencefromoneormoreprepotentcompetitors,ortheneedtostopare-sponsealtogether.
Acoreclaimofthisarticleisthatstrongevidencefortheseparallelsexists,andthatinhibitoryprocessesre-cruitedduringthecontrolofmemoryretrievalprecipi-tatetheforgettingassociatedwithinterference.Insupportofthisexecutivecontrolperspective,Ireviewevidenceforaroleofinhibitoryprocessesinmemoryselectionandmorebriefly,inmemorystopping.Mem-oryselectionisrequiredduringretrievalwhenourgoalistorecallaneventorfactfromlong-termmemoryinthefaceofinterferencefromrelatedtracesthatbecomeactivatedbycuesguidingretrieval.Theneedtostopretrievalariseswhenweconfrontacueorreminderandwewishtopreventanassociatedmemoryfromenteringawareness.Inbothsituations,attemptstolimitthein-fluenceofactivatedanddistractingmemorieshavebeenfoundtoimpairtheirlateraccessibility,highlightinganimportantlinkbetweenforgettingandthecontrolofretrieval.Inbothcases,thememoryimpairmentisbetterexplainedbyinhibitionthanbyconventionalassociativeinterferencemechanisms.Theforgettinginducedbyin-hibitionisoftenadaptive,limitingthetendencyforoutdatedorintrusivememoriestodisruptperformance(Bjork,1989;seealsoAnderson,2001;Anderson&Green,2001).
Inhibitorycontrolinselectivememoryretrieval
Theneedtoselectaweakerresponsetoastimulusinthefaceofinterferencefromaprepotentcompetitorfindsanaturalparallelinmemoryinthesituationofselectiveretrieval.Here,theaimistorecallaparticulartargeteventorfactwhenprovidedwithoneormore
retrievalcues.Typically,agivenretrievalcuewillbeassociatedwithothermemoriesaswell—andsomemaybemorestronglyassociatedtothecuethanthetargetitem.Sincetheclassicalinterferenceera,wehaveknownthatwhenmultipletracesareassociatedtothesamecue,theytendtocompeteforaccesstoconsciousawareness(seePostman,1971;seeAnderson&Neely,1996forreviews).Thisformofcompetitionpresentsaproblemofcontrolbecausetheretrievalcuebyitselfcannotbere-liedupontoaccessthetargetitem—infact,thepresenceofastrongcompetitorcouldinprincipleperpetuallydivertusfromthattargetmemory.Ifinhibitorycontrolmechanismsarerecruitedtooverrideprepotentre-sponses,itseemsreasonablethattheymightalsobeusedtooverrideprepotentmemories.Totheextentthattheeffectsofinhibitorycontrolpersist,thensituationsde-mandingtheselectiveretrievalofatargetitemshouldcauselong-lastingmemoryimpairmentforsuppressedcompetitors.Thus,theveryactofrememberingshouldcauseforgettingofrelatedmemories.
Overthelastdecade,wehaveexploredtheforegoingpredictionwithaparadigmwedevelopedtoexaminetheeffectsofretrievalonrelatedmemories:theretrievalpracticeparadigm(Anderson,Bjork,&Bjork,1994).Inthetypicalexperiment,subjectsstudylistsofcategory–exemplarpairs(e.g.,fruit—orange,drinks—scotch,fruit—banana).Theythenperformretrievalpracticeonhalfoftheexemplarsfromhalfofthecategoriesbycompletingcuedstemrecalltests(e.g.,fruit-or_____).Eachpracticeditemistestedthreetimesduringthere-trievalpracticephasetoincreasethemagnitudeoftheeffectonrelateditems.Aftera20-minretentioninterval,subjectsaregivenafinalcuedrecalltestforalltheex-emplars.Performanceonthistestcanbemeasuredforthreeitemtypes:practiceditems(e.g.,orange),unprac-ticeditemsfromthepracticedcategories(e.g.,banana),andunpracticedbaselineitemsfromunpracticedcate-gories(e.g.,scotch).Fig.2Aillustratesourinitialfind-ingswiththisparadigm,whicharequitetypical.Ascanbeseen,recallofthepracticedexemplarswasimprovedonthefinaltestrelativetoperformanceonbaselinecategories,demonstratingthewelldocumentedbenefitsofretrieval-practiceonthepracticeditemsthemselves(Allen,Mahler,&Estes,1969;Bjork,1975;Carrier&Pashler,1992;Gardiner,Craik,&Bleasdale,1973).However,recallfortheunpracticedexemplarsfromthepracticedcategories(e.g.,banana)wassignificantlyworsethanfortheitemsfrombaselinecategories(e.g,drinks).Thus,rememberingsomeitemsduringthere-trievalpracticephasecausedsubjectstoforgetotherthingsthatwererelatedtothemonadelayedretentiontest20minlater.Wehavereferredtothisfindingasretrievalinducedforgetting(Andersonetal.,1994),tohighlightthecentralrolethatretrievalisbelievedtoplayingeneratingtheeffect.Researchonretrieval-inducedforgettingbuildsonclassicworkonthephenomenonof
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445419
outputinterference(Arbuckle,1966;Dong,1972;Roediger,1973;Roediger&Schmidt,1980;Smith,1971,1973;Smith,DÕAgostino,&Reid,1970;Tulving&Ar-buckle,1963,1966)inwhichitwasshownthattheprobabilityofrecallingapreviouslystudieditemde-clineswiththeserialpositionofthatiteminthetestingsequence.However,workonretrieval-inducedforget-tingestablishesthatretrieval-relatedmemoryimpair-mentscanbelong-lasting,andarenotlimitedtodynamicsoccurringinasingle‘‘output’’session.Cru-cially,retrieval-inducedforgettingisconsistentwiththeviewthatinhibitorycontrolmechanismsarerecruitedtoovercomeinterferenceduringretrievalpractice,withinhibitionmanifestingasrecallimpairmentforcom-petitorsonthefinalretentiontest.
Althoughretrieval-inducedforgettingcouldbepro-ducedbyinhibition,theprecedingresultsdonotclearlyestablishinhibitionasthemechanism.Thebasicfindingofretrieval-inducedforgettingiscompatiblewithMcGeochÕs(1942)classicalresponsecompetitiontheoryofinterference.AccordingtoMcGeochÕstheory,thelikelihoodofrecallingatargetresponseshouldgodowneitherwhenanewresponsegetsassociatedtothecuenormallyusedtoretrieveit,orwhenanexistingalter-nativeresponseisstrengthened.Ineithercase,thetargetitemwillsufferincreasedcompetitionfromthealterna-tiveresponse—competitionthatwillblockaccesstothattarget.Thesecompetitivedynamicshavebecomefor-malizedinseveralmodernmemoryarchitecturesthatpositrelativestrengththeoriesofretrieval(e.g.,Ander-son,1983;Raaijmakers&Shiffrin,1981).Inthesemodels,theprobabilityofrecallingatargetisdeter-minedbythatitemÕsstrengthofassociationtoacue,relativetothestrengthsofassociationofallitemsre-latedtothatcue.Thus,whenanalternativeresponseisstrengthened,saybyretrievalpractice,therelativestrengthofallothernonpracticeditemsdeclinesonsubsequenttests.Later,whenthesubjecttriestorecallthetarget,thestrengthenedcompetitorwillhaveare-trievaladvantagethatwillleadittointrudesopersis-tentlythatsubjectswillabandontheireffortstorecalltheunpracticedexemplars(seealso,Rundus,1973).Importantly,thisapproachdoesnotrequireinhibition;rather,practiceditemsbecomesostronglylinkedtothepracticecuethattheyblocktheretrievalofotherex-emplars.Thisblockingaccountisplausible,giventhesubstantialstrengtheningthatpracticeditemstypicallyenjoy(howeverseelatersectiononstrengthindepen-dence).Othernoninhibitorymechanismsmayalsocon-tributetoretrievalinducedforgetting.Forexample,retrievalpracticemaydamagetheassociationlinkingthecategorytotheaffectedexemplaroralterinsteadthemeaningofthepracticedcategorycue(e.g.,bybiasing‘‘Fruits’’towards‘‘Citrusfruits’’)sothatthecategorylabelisnolongerafunctionalcueforretrievingtheunpracticedcompetitor.Allofthesemechanismshave
beenproposedastheoriesofinterference(forareviewofnon-inhibitorysourcesofmemoryimpairment,seeAn-derson&Bjork,1994;seedescriptionofasubsetoftheseinthelaterdiscussion).Althoughitmightseemdifficulttodistinguishthesealternatives,focusedempiricalre-searchhasyieldedevidencefavoringtheinhibitorycontrolview.Idiscussthisnext.Propertiesofretrieval-inducedforgetting
Workonretrieval-inducedforgettinghasrevealedpropertiesthatuniquelysupporttheinhibitorycontrolhypothesis,andthatsuggestthatalternativestrength-basedmodelsmaynotbecorrect.Theseincludecue-inde-pendence,retrieval-specificity,interference-dependence,andstrength-independence.Idiscussthesepropertiesnext,alongwithotherfindingsthatareoftheoreticalinterest.
Cue-independence
Manytheoriesofinterferencepredictthatforgettingshouldbestronglycue-dependent—thatis,observationsofforgettingshouldbetiedtoaparticularcue.Forexample,theblockingtheoryassertsthatstrengtheningsomeexemplarsthroughretrievalpractice(e.g.,FruitOrange)impairstherecallofotherexemplars(e.g.,FruitBanana)onadelayedtestbecausethepresentationoftheirsharedcueattestleadsthestrongerresponse(or-ange)tointrudepersistentlyandblocktheweakeritem.If,however,oneweretotrytorecalltheweakeritemthroughanindependenttestcuenotassociatedtothepracticeditem(e.g.,MonkeyB___),associativecompe-titionshouldbecircumvented.Thus,whetheroneob-servesforgettingofBananashoulddependonwhetheroneusestheoriginalretrievalpracticecuetotestthecriticalitemoranindependentcue.Theoriesthatpro-posethatinterferencederivesfromunlearningofthecue-targetassociationbetweenthepracticedcategoryandthecriticalitem,diversionofactivationalresources,orbiasesinthemeaningoftheretrievalpracticecueallsharethissamefeatureofpredictingcue-dependentforgetting(seeAnderson&Bjork,1994;Anderson&Spellman,1995,fordiscussion).
Theinhibitorycontrolperspective,bycontrast,pre-dictsthatretrievalinducedforgettingshouldexhibitcue-independence—thatis,atendencyfortheimpairmenttogeneralizetonoveltestcuesnotinvolvedintheretrievalpracticeeventsthatcausedimpairment.Thispredictionfollowsbecauseimpairmentisthoughttoarisefromsuppressionofthecompetingmemoryitself,ratherthanfromdamagetoanyparticularassociation.Thus,per-formingretrievalpracticeonFruit-Orangeshouldre-duceactivationfortheitemBanana.IfBananaislessactive,itshouldnotmatterwhethertheitemistestedfromtheoriginalretrievalpracticecue(Fruit)orfromanoveltestcue(e.g.,MonkeyB__).Totestthis,Anderson
420M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
andSpellman(1995)modifiedtheretrievalpracticeparadigmforusewithnewmaterials(Fig.2B).Asintheoriginalretrieval-inducedforgettingexperiment,subjectsstudiedcategoriescontainingsixexemplarseach,butunlikeinthatstudy,thecategorieswererelated.Forexample,althoughsubjectsstudiedBloodandTomatoundertheRedcategory,TomatoisalsoaFood;andalthoughRadishandCrackerswerestudiedasFoods,RadishisalsoaRedthing.Thekeyquestionwaswhe-therretrievalpracticeonitemssuchasRed-Bloodwouldnotonlyimpaircompetitorsexplicitlystudiedunderthesamecategorycue,likeRed-Tomato,butalsoredthingslikeRadishthatwerestudiedandtestedunderasepa-ratecategorycue.
Accordingtotheresponsecompetitionview,retrievalpracticeonRed-BloodshouldnotimpairdelayedrecallforFoodRadish,evenifretrievalpracticestrengthenstheRed-BloodassociationandweakenstheRed-Radishassociation.RadishshouldremainunimpairedbecauseitistestedwiththeFoodcategory—adifferentretrievalcuethatcircumventsthosefactors.However,ifretrievalpracticeonRed-BloodinitiallyactivatesalloftheReditems,bothTomatoandRadishshouldbecomeacti-vated,causinginterferencethattriggersinhibitorycon-trol.TheresultingsuppressionofRadishshouldbeobservablelaterwhenitistestedwithFood.AsFig.2Bshows,therecallofFood—Radishwasimpaired.
Thesedatashowthatinhibitoryprocessescontributetoretrieval-inducedforgetting,renderingcompetingmemorieslessaccessibleregardlessofwhichcueisusedtotestthem.Evidenceforcue-independentforgettinghasnowbeenfoundmanytimeswithstimulivaryinginbothtypeandcomplexity(Anderson&Bell,2001;An-derson&Green,2001;Anderson,Green,&McCulloch,2000;Andersonetal.,submitted;Anderson&Shivde,inpreparationa;Anderson&Shivde,inpreparationb;Anderson&Spellman,1995;Johnson&Anderson,inpress;Levy,Reinholz,&Anderson,inpreparation;Miyamoto&Anderson,inpreparation;Radvansky,1999;Shivde&Anderson,2001;however,seeWilliams&Zacks,2001).Takenasawhole,thesefindingsshowthatcue-independenceisageneralpropertyofretrieval-inducedforgettingandthatcuedependentmechanismssuchasblockingarenotadequatetoaccountfortheeffect.
Retrievalspecificity
Althoughcue-independencearguesthatinhibitioncausesretrieval-inducedforgetting,cue-dependentfor-gettingmechanismsmayneverthelesscontributeinsomecases.Inparticular,whenevertheretrievalpracticecueisusedduringlatertestsofsubjectsÕmemory,bothinhib-itoryandnoninhibitorymechanismsmaycauseforget-ting.Forinstance,theimpairmentofRed-Tomatointheprecedingexample(anitemthatwasbothstudiedandtestedwiththeretrievalpracticecue)mayreflecta
mixtureofblockingandinhibitioneffects.AscanbeseeninFig.2B,however,thisseemsunlikely.IftwosourcesofforgettingcontributedtoimpairmentonitemssuchasRed-Tomato,butonlyonesource(inhibition)onFood-Radish,wewouldexpecttoseemoreretrieval-inducedforgettingintheformerinstancethaninthelatter.Thefailuretofindsuchadifferencecastsdoubtontheroleofblockinginretrieval-inducedforgettingevenwhenthepracticedcategoryisusedasatestcue.Nev-ertheless,amoredirecttestoftheroleofblockinginretrieval-inducedforgettingwouldbedesirable.
Accordingtotheblockinghypothesis,presentingtheretrievalpracticecategoryonthefinaltestleadsprac-ticeditemstointrudeperseveratively,blockingrecalloftheunpracticedcompetitors.Ifso,thenstrengtheningpracticeditemsinanywayshouldimpairrelatedcom-petitors.Impairmentshouldbefound,forexample,evenifitemsarestrengthenedwithrepeatedstudyexposuresinsteadofretrievalpractice.Severalstudieshavead-dressedthispossibility.Forexample,usingAndersonandSpellmanÕscross-categoryinhibitionparadigm,AndersonandShivde(inpreparationa)manipulatedwhethertheto-be-practiceditemswerestrengthenedbyretrievalpracticeorrepeatedstudyexposures.There-trieval-practiceconditionreplicatedboththewithinandcross-categoryimpairmentobservedbyAndersonandSpellman(1995).Extrastudyexposures,however,failedtoimpairrelateditems.Noinhibitionwasfounddespitethefactthatbothstrengtheningmethodsfacilitatedthepracticeditemstothesamedegree,asevidencedbythesubstantialincreaseintheirrecallonthefinaltest.Severalinvestigatorshavefoundthispattern,usingavarietyofdifferenttypesofmaterialsanddifferentpar-adigms(Anderson&Bell,2001;Anderson,Bjork,&Bjork,2000;B€auml,1996,1997,2002;Blaxton&Neely,1983;Ciranni&Shimamura,1999;Shivde&Anderson,2001).Takentogether,thesefindingsarguethatasso-ciativeblockingdoesnotcontributesubstantiallytowithin-categoryretrieval-inducedforgetting,norper-hapstointerferenceeffectsmorebroadly.Rather,inhi-bitionisdrivenbytheneedtooverrideinterferencefromcompetingmemoriesduringtheselectiveretrievaloftargetitems.
Interferencedependence
Retrievalmaybenecessarytoinduceinhibition,butitisnotsufficient.Accordingtotheexecutivecontroltheory,retrievalinducedforgettingshouldonlyarisewheneverarelatedmemoryinterfereswiththeretrievalofatargetitemandtriggersinhibitorycontrol.Ifare-lateditemdoesnotinterfere,itshouldnotbeinhibitedevenwhenatargethasbeenretrieved.
Severalstudiesfavortheviewthatretrieval-inducedforgettingismoderatedbytheamountofinterferencecausedbyacompetingitem.Forinstance,Andersonetal.(1994)foundthatretrievalpracticedidnotalways
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445421
impairthelaterrecallofrelatedexemplars.Retrievalpracticeprimarilycausedimpairmentwhenrelatedcat-egoryexemplarswerehighintaxonomicfrequency(e.g.,FruitBanana).Lowfrequencycompetitors(e.g.,FruitGuava)werealwayslessimpairedandoftenexhibitednomeasurableimpairmentatall,evenwhensubjectsperformedretrievalpracticeonexactlythesameitems.Andersonetal.(1994)foundthatthemaindeterminantoftheamountofretrieval-inducedforgettingwasnei-therthetaxonomicfrequencyofthepracticeditems,northedegreetowhichpracticeditemswerestrengthenedonthefinaltest,butratherthefrequencyofthecom-petitors.Themorestronglyassociatedtothecategoryanunpracticedcompetitorwas,themoreimpairmentwasfound.Thesamepatternhasbeenfoundinanoutputinterferencedesign:high,butnotlowtaxonomicfre-quencyexemplarsexhibitwithincategoryoutputinter-ference(B€auml,1998;seealso,Andersonetal.,1994,Experiment2;however,seeAndersonetal.,1994,Ex-periment3).Thesefindingsareconsistentwiththeideathatinhibitorycontrolismostneededwhenarelatediteminterferesduringretrieval,asmightbeexpectedofthemostdominantexemplars.
Interferencedependencehasbeendemonstratedinotherways,aswell.Forexample,retrievalinducedforgettingcanbeeliminatedsimplybymanipulatingtheinterferencedemandsoftheretrievalpracticetask.ThiswasdemonstratedbyAndersonetal.(2000).Intheircompetitiveretrievalpracticegroup,subjectsweregiventhecategoryandthefirsttwolettersofanexemplarascues(e.g.,FruitOr___forOrange)duringeachpracticetrial(asistypicallydone),andsubjectswereaskedtorecalltheitemtheyhadstudied.Inthenon-competitivepracticecondition,subjectsalsoperformedretrievalpractice,butonthecategoryname.Specifically,subjectsweregiventhefirsttwolettersofthecategoryname,withanexemplar(e.g.,Fr___OrangeforFruitOrange),andwereaskedtorecallthecategoryname.Andersonetal.(2000)arguedthatrelatedexemplarswereunlikelytointerferewiththeretrievalofthecategorynamebe-causeapracticedexemplaritself,whichwasassociatedtothecategoryandnotwiththeotherexemplars,servedasaretrievalcue.Aspredicted,Andersonetal.(2000)foundinhibitioninthecompetitive,butnotinthenoncompetitivecondition.Thisdifferencewasfounddespitethepresenceofretrievalinbothconditionsanddespitesignificantandcomparablestrengtheningofpracticeditems.Thus,whentheretrievaltaskitselfdoesnotrequireinterferencetoberesolved,littleretrieval-inducedforgettingisfound,evenwhenthenatureofthecompetitorisheldconstant.
Intheforegoingstudies,thedegreetowhichprac-ticeditemswerestrengthenedwasnearlyidenticalinboththeconditionsthatshowedanddidnotshowre-trieval-inducedforgetting.Theseresultssuggestthatinadequatestrengtheningofpracticeditemsisunlikely
tobethecauseofdifferentialimpairment.However,onemightstillbeconcernedthatthedegreeofstrengtheningwasnotmanipulatedstronglyenoughtorevealimpair-ment.Toaddressthis,ShivdeandAnderson(2001)performedaparametricmanipulationofthenumberofretrievalpracticetrialsgiventoapracticeditem,toseewhetherimpairmentmightemergeforweakercompeti-tors.Tomanipulatethedegreeofinterference,AndersonandShivdeusedasymmetrichomographs,pairingeachonewithonewordrelatedtoitsdominantsense(e.g.,ArmShoulder)andanotherrelatedtoitssubordinatesense(e.g.,ArmMissile).Subjectswerethenaskedtoperformretrievalpracticeeither0,1,5,or20timesoneitherthedominantorthesubordinatewordassociate.Followingretrievalpractice,subjectsweretestedonthealternateassociatethattheydidnotpractice,witheithertheoriginallytrainedcueoranindependentcuethatwasalsoencodedpreviously.Theresultswereclear:Per-formingretrievalpracticeonthedominantsense(e.g.,ArmShoulder)didnotimpairthelaterrecallofthesubordinatesense(e.g.,ArmMissile)atall,eventhoughretrievalpracticeyieldedsubstantialretrieval-basedstrengtheningforthepracticeditem(seeFig.3).Practiceonthesubordinatesense,however,causedretrievalin-ducedforgettingofthedominantsense.Similarresultswereobtained,regardlessofwhethersubjectsweretestedontheunpracticedcompetitorwiththehomograph(ArmM___)ortheindependenttestcue(e.g.,Target-M___formissile).Thus,evenwhensubjectsperformedasmanyas20retrievalpracticetrialsonthedominantsense,littleretrieval-inducedforgettingwasobserved.Takentogethertheseresultsargueagainstassociativeblockingaccountsofretrieval-inducedforgetting,butareconsistentwithideathatthisphenomenondependsontheneedtooverrideprepotentmemories,aswouldbeexpectedifinhibitoryprocessesarerecruitedtosuppressthosememories(seeConway&Engle,1994forarelateddiscussionoftheroleofinhibitoryprocessesinresolvinginterferenceinmemoryspantasks;seealso,Lustig,Hasher,&Toney,2001,forarecentreviewofworkoninhibitoryprocessesincognitiveaging).
Strengthindependence
Ourearlyworkonretrieval-inducedforgettingwasinitiallypremisedontheclassicalviewthatstrengthen-ingsomeitemswouldimpairlaterretrievalofotherassociates(Andersonetal.,1994).However,wequicklydiscoveredthatthedegreetowhichpracticeditemsarestrengtheneddoesnotpredicthowmuchretrievalin-ducedforgettingwasobserved.Infact,ashighlightedintheprecedingsections,practiceditemscanbesig-nificantlystrengthenedwithoutcausingimpairment:Retrievalpracticeontargetitemsdoesnotimpairlowtaxonomicfrequencycompetitors,subordinatemeaningsofambiguouswords,norevenhighfre-quencyexemplars,providedthatretrievalpracticeis
422M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
Fig.3.ResultsofastudybyShivdeandAnderson(2001):(A)plotsthepercentageofpracticeditems(e.g.,ArmShoulder)thatsubjectsrecalledonthefinaltestasafunctionofthenumberofretrievalpracticesitreceived;(B)showsthatevenafterextensivepracticeonthepracticeditems,recallofweakercompetitors(e.g.,ArmMissile)isnotimpairedonthefinalrecalltest(whentestedwithArmM___);(C)showstherecalldatawhentheunpracticedcompetitorwastestedwithanindependentprobe(e.g.,TargetM___forMissile),againrevealingnoimpairmentasaresultofpractice.noncompetitive.Nordoesinhibitionoccurifpracticeditemsarestrengthenedbyrepeatedstudyexposuresin-steadofretrievalpractice.Inallofthesecases,strengtheningoccurred,withnoresultantinhibition,evenwhentheamountofstrengtheningwasidenticaltoorevengreaterthanthatobservedinotherconditionsinwhichretrieval-inducedforgettingwasfound.Whenanalysesarerestrictedtocasesinwhichretrieval-inducedforgettingisfound,themagnitudeoftheim-pairmentbearslittlequantitativerelationshiptothedegreeoffacilitationonpracticeditems.Together,thesefindingsarguethatimpairmentisindependentofthestrengthofthepracticeditem.
Thepropertyofstrengthindependenceissurprising,giventhehistoricalemphasisontheroleofcompetitioninproducinginterference(e.g.,McGeoch,1942;Melton&Irwin,1940;seealso,Anderson,1983;Mensink&Raajimakers,1988).However,giventhefrequentco-occurrenceofstrengtheningandimpairmentacrossavarietyofparadigms(e.g.,retroactiveandproactivein-terference,part-setcuing,list-strengtheffects),theem-phasisonstrengthasacauseofforgettingmakessense.Andersonetal.(1994)notedhoweverthatnearlyallparadigmsthatappeartoprovideevidenceforstrength-dependentcompetitionhaveconfoundedstrengtheningwithsomeformofretrieval-inducedforgetting.Instudiesofretroactiveinterference,forexample,onetypicallycannotdisentangletheeffectsofstrengtheningwordpairsfromthesecondlist(e.g.,Dog-Sky)fromthesuppressionoffirstlistresponses(e.g.,Dog-Rock).Thisambiguityarisesbecausewordpairsfromthesecondlistaretypicallystrengthenedbyrepeatedstudy/testcycles,aprocedurewhichconflatesstrengtheningofthosepairswithretrievalpractice.Inlist-strengtheffectstudies,onecannotdisentangletheeffectsofstrengtheningonehalfofthelistofwordsfromtheheightenedoutputinter-ferencethatthosestrengtheneditemscauseforthere-mainingnonstrengthenedwordsonlaterfreerecalltests.Iflefttorecallitemsinanyorder,subjectstypicallybeginwiththestrengtheneditems,whichislikelytoinhibittheremainderofthelist.Similarproblemsoccurinpart-setcuingstudies,whichoftendonotadequatelycontrolforoutputinterferencebiases(overtorcovert)createdbyprovidingpart-setcues(seeAnderson&Neely,1996;Nickerson,1984;Roediger&Neely,1982,forreviewsofpart-setcuingandrelatedresearch).Thus,althoughstrengtheningsomeitems(evenwithoutre-trievalpractice)appearstoimpairnonstrengthenedcompetitors,sucheffectsmaybebetterexplainedintermsofretrieval-inducedforgetting.
Thedifferencebetweentheforegoingstudiesandre-centdemonstrationsofstrengthindependenceliespri-marilyintheattempttoseparatetheprocessofstrengtheningfromretrieval-inducedforgetting.Forexample,byusingextrastudyexposures,wehavebeenabletostrengthenpracticeditemswithoutretrievalpractice,sothatwecanseewhethertheaddedstrengthforthoseitemswouldimpairthedelayedrecallofcompetitors.Alsoimportanthasbeenourefforttocontroltheorderinwhichsubjectsrecallstudieditems,soastopreventbiasesinoutputordertypicallycreatedbystrengtheningmanipulations.Towardsthisend,wehaveusedletterstemcuedrecalltasks(e.g.,FruitB___)toforcesubjectstorecallnonstrengtheneditemsbeforestrengtheneditems,andreducetest-timeretrievalin-ducedforgetting.Bycontrollingthesefactors,wehavefoundthatstrengtheningdoesnotcauseforgettingofcompetitors.Inasimilarvein,Baumlandcolleagues(B€auml,1996,1997,1998)havealsoattemptedtosep-aratetheinfluencesofstrengtheningandinhibitioninotherexperimentalproceduressuchastheretroactiveinterferenceandliststrengtheffectparadigms.B€auml(1996)foundthatstrengtheningasecondlistofwordsbyincreasingstudytimedidnotincreaseretroactiveinterferenceonthefirststudylist.B€auml(1997)showedthattheliststrengtheffectvirtuallydisappearsifbiasesinoutputorderareeliminated:strengtheninghalfofastudylistthroughextrastudydoesnotimpairthelaterrecalloftheotherlisthalfaslongasthosenon-strengtheneditemsaretestedfirstintherecallsequence.Thus,avarietyofinterferenceeffectsthathavebeenattributedtostrength-dependentcompetitionmayarisefromtherecruitmentofinhibitorycontrolprocessesduringretrieval.
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445423
Delaydependence?
Thereissomeevidencethatretrieval-inducedforget-tingmayrecoverovertime.Inonestudy,MacLeodandMacrae(2001)hadsubjectsperformretrievalpracticeimmediatelyafterencodinghadended.Afterretrievalpractice,subjectsweretestedeitherimmediatelyoronthefollowingday.Interestingly,retrievalinducedforgettingwasobservedontheimmediatetest,butnotforthosesubjectstestedafter24h.Inafollow-upstudy,MacLeodandMacrae(2001)replicatedthisfinding,butalsoshowedthatwhenretrievalpracticewasperformedaftera24hdelay,significantretrieval-inducedforgettingwasobservedonatestgivenimmediatelyafterwards.Thus,althoughtheimpairmentmaydissipatewithin24h,in-troducingalongdelaybetweenstudyandretrievalprac-ticedidnotinsulatesubjectsfromretrieval-inducedforgetting.Inatleastsomecircumstanceswithsometypesofmaterials,theinhibitoryeffectsofretrievalpracticerecoverovertime,afindingsimilartospontaneousre-coveryobservedinretroactiveinterference(e.g.,Post-man,Stark,&Fraser,1968),directedforgetting(Wheeler,1995),andtheverbalovershadowingparadigm(Finger&Pezdek,1999).Interestingly,thisrecoveryfrominhibitionoccurseventhoughpracticeditemsstillexhibitsignificantfacilitationafterthesamedelay,againsuggestingthatdifferentialstrengthdoesnotcauseimpairment.
Itisnotclear,however,whetherMacLeodandMac-raeÕsparticulardelayisneededforpeopletorecoverfromretrieval-inducedforgetting,orevenwhetherrecoveryalwaysoccurs.AlthoughMacLeodandMacraeÕsfindingssuggestaparticularrecoveryinterval,thisfindingmaynotgeneralizetoothermaterialsortrainingprotocols.Forinstance,retrievalinducedforgettingmaybequitelonglastinggivendifferentparametersforretrievalpractice.Considerlearningthenewtelephonenumberofafriendwhomyoucallfrequently.Initially,theiroldnumberwillintrudeintoconsciousnesswhenyouwanttodialtheirnewnumber.Butafterdialingthenewnumberenoughtimesoveraprotractedperiod,theoldnumbereventuallystopsintruding.Givenenoughprac-ticewiththenewnumber(overmonthsorayear),onemaybecomecompletelyunabletorecalltheoldtele-phonenumber.Thisinabilitywilllikelypersistindefi-nitely,evenwhenyouhaveperiodsduringwhichyoudonotcallyourfriend(canyourememberyouroldphonenumber3residencesago?).Thissuggeststhatifretrievalpracticeoccursfrequentlyandisdistributedoverlongtimeperiods,inhibitioneffectsmaybelong-lasting,althoughthisatpresentremainsanempiricalissue.
Theoretically,recoveryneednotoccuratall,how-ever,evenifinhibitoryprocessesproduceretrieval-in-ducedforgetting.Infact,impairmentofnearlyanydurationmaybepossible,dependingonthemechanismsbywhichinhibitoryeffectsproducememoryfailure(Anderson&Spellman,1995).Onecanseparatethetheoreticalmechanismthatinducesretrieval-induced
forgettingfromtheonethatsustainsit.Forexample,inhibitoryprocessesmayinitiallydeactivatetheunitsinvolvedinrepresentingacompetingmemorytrace,andthisdeactivationmaybeshort-lived.However,thisde-activationmayitselfcauseastructuralchangethatpersistswellbeyondtheinitialperiodofinhibition.Forinstance,theconstituentfeaturesofthetracemaybe-comelesstightlyboundoraconsolidationprocessthatmighthaveotherwisebeenongoingmaybeterminated
(e.g.,M€u
ller&Pilzecker,1900).Inhibitionofasubsetoffeaturesinatracemayleadtoreductionsintheasso-ciationsofthosefeaturestoothersthatremainactive,viathemechanismsofhebbianlearning.Inanentirelydifferentapproach,inhibitionmightbesustainedtoni-cally,evenafterretrievalpracticehasended.Theper-sistingstrengthofpracticeditemsmightsustaininhibitiononcompetitors,vialateralinhibition.
Theforegoingtheoreticalpossibilitiesarenotin-tendedtobestrongclaimsaboutthemechanismsun-derlyingretrievalinducedforgetting.Theyaredescribedmerelytoillustrateatheoreticallycrucialpoint:notenoughispresentlyknownabouthowinhibitionismanifestmechanisticallytostronglyconstrainpredic-tionsaboutwhetherinhibitionshouldrecoverovertime.Inhibitorytheoriesexistthatcanareconsistentwithshortandlong-lastinginhibition(seeAnderson&Spellman,1995forsimilararguments).Whichofthesemechanisticapproachestoretrieval-inducedforgettingprovidesthebestaccountremainstobeestablished.Generality
Manyofthestudiesdiscussedsofarhaveusedverbalcategoriestostudyretrieval-inducedforgetting.How-ever,thisphenomenonhasnowbeenobservedwithavarietyofstimulusclasses.Forexample,CiranniandShimamura(1999)foundthatwhensubjectslearnedthelocationsofcoloredobjects(e.g.,squares,circles,trian-glesorodd,difficulttonameshapes),recallinginfor-mationaboutoneoftheobjects(e.g.,itÕscolororshape)ledsubjectstoforgetpropertiesofotherobjectswiththesameshape.Usingvariationsofthisprocedure,theyin-ducedsubjectstoforgetthecolor,location,andshapeoftheotherobjects,andfoundthatthisimpairmentonlyoccurredwithretrieval-practiceandnotwithextrastudyexposures.Studiesoffactlearninghavefoundthatre-trievingsomefactsaboutatopicimpairsrecallforotherfactsaboutthattopic(e.g.,Anderson&Bell,2001;Macrae&MacLeod,1999;Radvansky,1999).Infact,retrievingsomefactsaboutatopic(e.g.,Theactorislookingatthetulip)notonlyimpairsotherfactsthatdirectlycompetewithit(e.g.,Theactorislookingattheviolin),butalsofactsthatshareconceptswiththecom-petingfacts(e.g.,Theteacherisliftingtheviolin),repli-catingandgeneralizingthecue-independentimpairmentobservedbyAndersonandSpellman(1995).Similarcue-independentimpairmentoccursinthefaninterference
424M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
paradigm(Radvansky,1999;however,seeAnderson&Reder,1999).Koutstaal,Schacter,Johnson,andGal-luccio(1999)foundthatreviewingphotographsofnovelactionsthatsubjectshadperformedtwodaysearlier(e.g.,‘‘tracetheoutlineofthisboomerang’’),impairedtheirlaterrecalloftheotheractionstheyhadperformed.Thus,memorycanbeimpairedforoneÕsownphysicalactions.Usinganeyewitnessmemoryparadigm,studieshaveshownthatinterrogatingsubjectsaboutsomede-tailsofamockcrimesceneimpairsmemoryforotherrelateddetails(MacLeod,2002;Shaw,Bjork,&Handal,1995),afindingthatmayhavesignificantpracticalap-plications.Inarelatedvein,recentworkhasshownthatthetendencyforpeopleÕsmemoryforaneyewitnesseventtobedistortedbymisleadingpost-eventinformationmayrelytosomedegreeonretrieval-inducedforgetting(Saunders&MacLeod,2002).SaundersandMacLeodfoundthatpeoplewerefarmorelikelytoinappropriatelyremembermisinformationonalatertestwhentheyhadearlierperformedretrievalpracticeonotheraspectsoftheeventforwhichthemisinformationwasintroduced.Thissuggeststhatvulnerabilitytomisinformationac-ceptanceisheightenedwhenaccesstotheoriginalmemoryisweakenedbyinhibition.
Implicationsofretrievalinducedforgettingforsocialpsychologicalphenomenahavealsobeenexplored.Forinstance,MacraeandMacLeod(1999)demonstratedthatrecallingsometraitsofapersonimpairsthere-trievaloftheirotherpersonalitytraitslater.DunnandSpellman(2003)recentlydemonstratedthatwhenpeoplerepeatedlyretrieveindividuatingtraitsofapersonaboutwhomtheyrecentlylearned,stereotypictraitsofthatpersonwereinhibited.Interestingly,thetendencyforstereotypictraitstobeinhibitedwasmoderatedbysubjectsÕpriorbeliefinthestereotype:subjectswhoweremorepronetobelieveinthestereotypeshowedgreaterresistancetoinhibition.MacraeandRoseveare(2002)foundthatself-relevantencodingmayalsorenderin-formationresistanttoinhibitoryeffects.Subjectswerepresentedwithalistofwordsandtoldthattheitemsonthelistweregiftsthatwerepurchased.Somesubjectswereaskedtoimaginethattheseweregiftsthattheyhadpurchasedthemselves;othersubjectswereaskedtoimaginethatthegiftswerepurchasedbyabestfriendorbyanunspecifiedother.Followingthisencodingphase,thestandardphasesoftheretrieval-practiceparadigmwereemployed.MacraeandRosevearefoundthatwhensubjectsimagedpurchasingthegiftsthemselves(selfrelevantencoding),retrieval-inducedforgettingwascompletelyeliminated,whereastheinhibitioneffectre-mainedrobustintheotherencodingconditions.Theprotectiveeffectofself-relevantencodingmaybeaninstanceoftheprotectiveeffectsofintegration(seelatersection‘‘Integrationasamoderatingfactor’’).
Someevidencesuggeststhatretrievalimpairsrecog-nitionmemoryforperceptualexperiences.Researchonverbalovershadowinghasshownthatdescribingare-centlyviewedfaceimpairslaterrecognitionofthatface(Schooler&Engstler-Schooler,1990).Similarly,de-scribingtheflavorofawineimpairsitslaterrecognition(Melcher&Schooler,1996).Schooler,Fiore,andBrandimonte(1997)suggestedthatdescribingaper-ceptualmemorymaybeaformofretrievalpractice.Subjectsmayretrievethoseaspectsofthepercepteasiesttoverbalize,eschewingotherinformationthat,whilenotaseasytodescribe,iscrucialtorecognition.Forex-ample,subjectsmightfocusonverbalizablecharacter-isticsofaface,suchasbasicfeatures(nose,mouth),ratherthanconfiguralinformationaboutthespacingoffeaturesinrelationtoeachother.Selectivelyretrievingfeaturesmaysuppressconfiguralinformation.Althoughverbalovershadowinghasalsobeenfoundwhentheneedforretrievaliseliminated(e.g.,whensubjectsre-ceiveadescriptiongeneratedbyanotherperson),thoseeffectsgoawayifsubjectsareaskedtobasetheirrec-ognitionjudgmentssolelyonmemoryforthephoto-graphandnottheverbaldescription.Incontrast,subjectsgeneratingtheirowndescriptionsarenothelpedbysuchinstructions(Dodson,Johnson,&Schooler,1997;Meissner&Brigham,2001).Thesefindingssug-gestthatactiveretrievalisimportanttogeneratingarobusteffect,justaswithretrieval-inducedforgetting.Dodsonetal.alsofoundthatdescribinganotherfaceimpairedrecognitionfortheoneoriginallystudied,in-dicatingageneralizedsuppressionoffacememories(seeAnderson&Spellman,1995,forapotentiallyrelatedfindingtermed‘‘second-orderinhibition’’).Thus,re-trievingspecificfeaturesofaperceptualmemorywhiledescribingitmayinhibitotheraspectsofthememory.Ifinhibitorycontrolmechanismsresolveinterferenceinmemoryretrievalgenerally,wewouldalsoexpectthemtobeatworkinsemanticmemory.Consistentwiththis,B€aumlÕs(2002)foundthatepisodicmemoryforseveralstudiedexemplarsofacategorywasimpairedifsubjectsgeneratednewexemplarsofthesamecategoryfromsemanticmemoryduringtheintervalbetweenstudyandtest.However,episodicrecallwasunimpairedwhenthis‘‘semanticgenerationpractice’’wasreplacedbystudyexposuresofthesamenovelexemplars,show-ingthatimpairmentderivedspecificallyfromsemanticretrieval.Inarelatedstudy,BlaxtonandNeely(1983)foundthatsubjectswereslowertogenerateacriticaltargetexemplar(FruitA___)fromsemanticmemoryaftertheyhadgeneratedfourotherprimeexemplarsfromthatsamecategory.Incontrast,subjectswerefastertogeneratethesametargetwhentheprimeitemswerepresentedintacttosubjectsforspeedednaming.Inrecentwork,JohnsonandAnderson(inpress)haveshownthatrepeatedlygeneratingassociatestothesubordinateverbmeaningofahomographfromgeneralknowledge(e.g.,PruneT__MforPruneTrim)re-ducedtheavailabilityofitsdominantnounmeaning,as
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445425
measuredbyanindependentprobetestinwhichsubjectsfreeassociatedtonoveltestcues(e.g.,YogurtF___for‘‘fruit,’’whichisrelatedtothenounsenseofPrune).Thisfindingbuildsonworkonlexicalambiguityreso-lutionsuggestingthatthecontextuallyinappropriatesenseofahomographmaybesuppressed(Gernsbacher&Faust,1991;Simpson&Kang,1994),byestablishingthatimpairedaccesstohomographmeaningsiscue-independent.Parallelfindingshavebeenobservedinepisodicmemoryexperimentsusinghomographs,whichestablishthatinhibitoryeffectsarerecallspecific,con-sistentwithpropertiesofretrieval-inducedforgettingmoregenerally(Shivde&Anderson,2001).Finally,re-searchusingtherare-wordparadigmhasfoundthatdifficultsemanticretrievalsrecruitinhibitoryprocesses:Whensubjectsstruggletorecallthemeaningofanun-usual,infrequentlyencounteredwordthatisweaklyrepresentedinmemory,relatedconceptsappeartobeimpaired(Barnhardt,Glisky,Polster,&Elam,1996;Dagenbach,Carr,&Barnhardt,1990;seealsoThomp-son-Schill,1997foraninterestingdiscussionoftheroleoftheleftprefrontalcortexincontrollingselectivere-trievalfromsemanticmemory).Takentogether,theseresultsarguethatretrievalinducedforgettingisnotlimitedtoepisodicretrieval,ortotaxonomiccategories;rather,itisageneralconsequencearisingwheneverin-hibitorymechanismsarerecruitedtoguideselectioninthefaceofcompetitionfromdistractingmemories.Moderatingandmaskingfactorsinretrieval-inducedforgetting
Theprecedingreviewdescribesevidenceshowingthatwheneverwetrytoselectivelyretrieveatargetitemfromlong-termmemory,othercompetingmemoriesassoci-atedtothecueguidingretrievalwillbesuppressed.Al-thoughthisisgenerallytrue,itisperhapsnotsurprisingthattherearefactorsthatcaneithermoderateormasktheeffectsofinhibition.Moderatingfactorsarethosethatgenuinelyalterthemagnitudeofinhibitionthatthecompetitorsofmaininterestsufferduringretrievalpractice;maskingfactorsarethosethatalterthelaterbehavioralmeasureofinhibitionwithoutaffectingthemagnitudeofinhibitionthatactuallytranspiredduringretrievalpractice.Appreciatingthesefactorsisafun-damentalpartofunderstandingthebehavioralcondi-tionsunderwhichinhibitorycontrolleadstoforgetting.Thesefactorscanbedividedbroadlyintothosecon-cerninghowmemoriesarerepresented,howretrievalpracticeisperformed,andhowmemoryisultimatelyassessedafterinhibitionhasbeeninduced.Wediscusstheseinturn.
RepresentationalfactorsthatmoderateormaskinhibitionWhenpredictinghowmuchinhibitionwilloccurinagivenpopulationorinaparticularcondition,onecannotconsiderthecharacteristicsoftheinhibitorymechanisminisolation.Inhibitorymechanismsactonmemoryrepresentations.Theserepresentationsmayvaryintype,structure,content,orstrength,andthesevariationsmaymoderatetheimpactofinhibitorypro-cessesoreventhenecessityofinhibition.Concernoverthispossibilityhasalonghistoryinresearchoninter-ference,andisreflectedinclassicworkonverbalme-diation(seeHorton&Kjeldergaard,1961;Jenkins,1963;Kjeldergaard,1968;Postman,1971,forreviews)andsimilarityeffects(seeOsgood,1949,forareview)onretroactiveinterference,integrationeffectsonfaninter-ference(see,e.g.,Radvansky,1999;Radvansky&Zacks,1991;Smith,Adams,&Schorr,1978),andtheeffectsofleveloflearningonthemagnitudeofretro-activeinterferenceorfaneffectsthatareobserved(see,e.g.,Postman,1971forareviewforretroactiveinter-ference;seeHayes-Roth,1977,forareviewconcerningfaninterference).Itisthusnotsurprisingthatthesefactorsarealsoinvolvedinmoderatingretrievalinducedforgettingaswell.Ireviewtheevidencefortwosuchmoderatingfactors:integrationandsimilarity.Ialsodescribearepresentationalfactorthatmasksinhibi-tion—baselinedeflation.
Integrationasamoderatingfactor.Theamountofretrieval-inducedforgettingdependsstronglyonhowwellintegratedtheto-be-retrievedmemoriesarewiththepracticedcompetitors.Althoughthereissomevariationinhowthetermintegrationhasbeenusedinthelitera-ture,wehaveusedittorefertotheexistenceofinter-connectionsbetweenitemssharingacommonretrievalcue—connectionsformedeitheronthebasisofpre-ex-perimentalrelationships,ornovelinterrelationshipsdiscoveredduringthecourseofanexperiment.Forin-stance,supposesubjectsstudied‘‘Animals’’suchasDeer,Dog,Bear,Canary,Goat,andCow.Inadditiontostudyingtheseitemsinrelationtotheirsharedcate-gorylabel,subjectsmightforminter-relationshipsbe-tweenitemssuchasDeerandBear(Wildthingsthatyouhunt),GoatandCow(farmanimals),DogandCanary(pets),orDogandDeer(animageofadogchasingadeer).Theseinter-relationshipscouldbebasedonse-manticsimilarity(e.g.,Dog,Wolf),associativerelated-ness(DogBone),orevenonmoreelaborateencodingofrelations(e.g.,interactiveimagery).
Ingeneral,whensubjectsintegratetheassociatesofacue,itinsulatesnonpracticedexemplarsfromretrieval-inducedforgetting(Anderson&McCulloch,1999).AndersonandMcCullochdemonstratedthisusingtheretrieval-inducedforgettingdesignofAndersonetal.(1994),butwithonechange:atthetimeofencoding,subjectseitherwereorwerenotencouragedtofindinter-relationshipsamongtheexemplarsofacategory.Sub-jectswhowereaskedtointegrateexemplarsshowedasignificantreductioninretrievalinducedforgetting(andinsomecases,itwascompletelyeliminated).
426M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
Interestingly,evensomesubjectswhowerenotaskedtointegratereportedthattheyhaddonesoontheirown,asmeasuredbyapost-experimentalquestionnaire.Thesesubjectsshowedthesamereductioninretrieval-inducedforgettingasthegroupinstructedtouseintegration.Themorestudytimesubjectswereallowed,thehigherthereporteduseofspontaneousintegrationstrategies,andthelowertheamountofretrievalinducedforgetting.Similarintegrationeffectswereobservedinretrieval-in-ducedforgettingexperimentsusingpropositionalmate-rials(Anderson&Bell,2001):whensubjectsreportedintegratingmultiplefactsaboutatopic(e.g.,Theantcrawledontherock,Theantcrawledonthetable),re-trieval-inducedforgettingwaseitherreducedorelimi-nated.Theselattereffectswereevenfoundinanincidentalencodingtaskinwhichsubjectswereaskedtoformvividmentalimagesofthesituationsrepresentedbythesentence;whensubjectsreportedincorporatingmultiplefactsintoasingleimage,inhibitionwassig-nificantlyreduced.Thus,althoughretrievingsomeas-sociatesofaretrievalcuegenerallyimpairsotherassociatesthatbecomeactivatedintheprocess,inte-grationposesastrongboundaryconditiononwhenthisimpairmentoccurs.
Similarityasamoderatingfactor.Theamountofinhibitionthatretrievalwillcausealsodependsonse-manticsimilaritybetweentheassociatesofacue.Thenatureofthisrelationshipiscomplex,however,asil-lustratedbythestudiesofSmithandHunt(2000)andB€aumlandHartinger(2002).SmithandHunt(2000)adaptedtheretrievalpracticeproceduresothatthede-greeofwithin-categorysimilaritymightbevaried.Spe-cifically,theyalteredthestudyphasetoencouragetheencodingofeithersimilaritiesordifferencesbetweenexemplarsofacategory.Forthesimilarityencodinggroup,subjectsviewedallsixexemplarsofthecategoryatonceandwereaskedtofindawaythattheitematthetopofthelistwassimilartoalloftheremainingitems.Sharedfeatureswerethengeneratedinturnfortheotherfiveexemplars.Afterencodingthecategoriesinthisway,subjectswentthroughtheremainingphasesofthere-trievalpracticeprocedure.Thedifferenceencodinggroupfollowedthesamesteps,butwasaskedinsteadtofindonefeaturethatmadethetopitemdifferentfromalloftheremainingitems.SmithandHuntfoundthaten-codingdifferencesbetweenexemplarsabolishedretrievalinducedforgetting,butencodingsimilaritiesyieldedrobustimpairment.Theyconcludedthatinter-itemsimilarityincreasesretrieval-inducedforgetting.However,B€aumlandHartinger(2002)foundapat-ternthatappearstodirectlycontradictthatobservedbySmithandHunt.Theseauthorsalsosoughttomanip-ulatetheinter-itemsimilaritybetweentheexemplarsofacategory,buttheymanipulatedsimilaritybyvaryingwhetherornottheunpracticedcompetitorsinacate-gory(e.g.,FruitLemon)weredrawnfromthesamesubcategory(e.g.,Citrus)astheto-be-practiceditems(e.g.,FruitOrange),oradifferentsubcategory(e.g.,FruitCherry).Aftertheencodingphase,subjectsen-gagedinretrievalpracticeintheusualfashion,andthenweregivenafinalcategory-plus-letterstemcuedrecalltest.IncontrasttoSmithandHunt(2000),BaumlandHartingerfoundthatincreasinginteritemsimilarityre-ducedretrieval-inducedforgetting.BaumlandHartingerreplicatedthispatternusinganoutputinterferenceparadigminsteadofretrievalpractice,andarguedthatsimilarmechanismsunderliethesetwophenomena.TheSmithandHunt(2000)andB€aumlandHartin-ger(2002)findingsarenotnecessarilycontradictory.Inrecentwork,Andersonetal.(2000)exploredwhethersemanticsimilaritymighthavedifferenteffectsonre-trievalinducedforgettingdependingonwhetheroneisconcernedwithwhattheytermedtarget–competitorsimilarityorcompetitor–competitorsimilarity.Asillus-tratedinFigs.4AandB,theunpracticedcompetitorsinacategoryundergoingretrievalpracticecaneithervaryin(a)howsimilartheyaretothetargetitemsreceivingretrievalpractice(i.e.,target–competitorsimilarity),or(b)howsimilartheyaretoeachother,independentofhowsimilartheymaybetotheretrievalpracticetargets(i.e.,competitor–competitorsimilarity).Andersonetal.(2000)arguedthatthesetwodimensionsshouldhaveverydifferenteffectsonretrieval-inducedforgetting,basedonthedistributedrepresentationapproachpro-posedbyAndersonandSpellman(1995).Accordingtothisapproach,increasingtarget–competitorsimilarityfromamoderatelevel(topofFig.4A)toaveryhighlevel(bottomofFig.4A)shoulddiminishretrievalin-ducedforgetting.Lessimpairmentshouldbeobservedbecause,accordingtothemodel,therecallprobabilityofanitemreflectsthesummedactivationofallofitsfea-tures.Becausehightarget–competitorsimilarityleadsmanyofacompetitorÕsfeaturestooverlapwiththere-trievalpracticetarget,thefacilitatoryeffectsofretrievalpracticeonsharedfeatureswillcompensatefororpos-siblyevenoutweightheinhibitionsufferedbythecom-petitorÕsdistinctivefeatures.Ontheotherhand,increasingcompetitor–competitorsimilarityfromamoderatelevel(topofFig.4B)toahighlevel(bottomofFig.4B)shouldmagnifytheamountofretrieval-inducedforgetting.Moreimpairmentshouldbeobservedbe-causeinthehighsimilaritycondition,theimpactofsuppressingasinglefeaturethatoverlapstwodifferentexemplarswillberealizedthroughtheimpairmentoftwoitems,notjustone;thus,thebehavioraleffectofapplyingtheinhibitiontohighlyoverlappingrepresen-tationswillbeexaggerated,evenifthesameamountofinhibitionisapplied.
Andersonetal.(2000)testedthesehypothesesbyseparatelymanipulatingthedegreeoftarget–competitorandcompetitor–competitorsimilarity.FollowingSmithandHunt,theyheldthestudymaterialsconstantand
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445427
Fig.4.Illustrationoftwodifferentdimensionsofsimilarity,asdistinguishedinthetwo-factormodelofAnderson,Green,andMcCulloch(2000).Memoryitems(largercircles)arerepresentedhereassetsofsemanticfeatures(smallcircles).Similaritemsoverlapinfeaturespace(asrepresentedbyoverlappinglargercircles).Retrievalpracticeisassumedtoincreasetheactivationofpracticedfeatures(darkenedcircles)andtoinhibitsomeofthefeaturesofthecompeting,similarpattern(lightercircleswithXesinthem),butnototherfeaturesofthecompetitor(smallwhitecircles).(A)Anillustrationofhowtarget–competitorsimilarity(similaritybetweenthepracticeditemandanunpracticedcompetitor)canbelow(tophalf)orhigh(bottomhalf)andhowthisinfluencesinhibition.Withhightarget–competitorsimilarity,agreaterproportionofacompetitorÕsfeaturesoverlapwiththepracticeditemandarestrengthened,compensatingforinhibitionontheremainingfeatures.(B)Anillustrationofhowcompetitor–competitorsimilarity(similarityamongstthecompetitorsthemselves)canbelow(top)orhigh(bottom),andhowthismayinfluenceinhibition.Withhighcompetitor–com-petitorsimilarity,inhibitingthesamenumberofunitshasagreaterimpactonthetwocompetitors,becausetheinhibitionaffectsfeaturessharedbythetwoitems.manipulatedsimilaritybyaskingsubjectstoidentifysimilaritiesordifferencesbetweenexemplars.However,insteadofhavingsubjectsdothisforallpairwisecom-parisonswithinacategory,subjectswerepresentedwitheithertarget–competitororcompetitor–competitorpairings,tocontrolthedimensionofsimilaritythatwasmanipulated.Followingthissimilarity(ordifference)encodingphase,theremainingstepsoftheretrievalpracticeprocedureweredoneinthetypicalfashion.Theresultswerestriking:inthetarget–competitorcondition,significantlylessinhibitionwasfoundwhensubjectswereaskedtofindsimilaritiesthanwhentheywereaskedtofinddifferencesbetweenitemsduringencoding.Infact,subjectswhowereaskedtofindtarget–com-petitorsimilaritiesshowedsignificantretrieval-inducedfacilitationofcompetingitems,notinhibition.Inthecompetitor–competitorcondition,however,moreinhi-bitionwasfoundwhensubjectswereaskedtoencodesimilaritiesthanwhentheywereaskedtoencodedif-ferences.Indeed,thedifferenceencodingconditionyiel-dednosignificantinhibition.Thesefindingsstronglysupporttheideathatcompetitor–competitorsimilarityhasanoppositeeffectoninhibitionthantarget–com-petitorsimilarity,assuggestedbytheAndersonandSpellman(1995)distributedapproach.Andersonetal.(2000)arguedthatthesefindingshelptoreconcilethe
conflictingfindingsofBaumlandHartingerandSmithandHunt,aswellasanalogousinconsistenciesintheliteratureontheroleofsimilarityinclassicalinterfer-encestudies(seeAndersonetal.,2000foradiscussion).Baselinedeflationasamaskingfactor
Whenconsideringhowrepresentationalvariablesthatmightaffectinhibition,itisalsoimportanttoattendtotherepresentationofbaselineitems.Retrieval-in-ducedforgettingmaybemaskedifthebaselineusedtoassessinhibitionisalsoaffectedbyretrievalpractice.Such‘‘baselinedeflation’’mayariseintwoways.First,asAndersonetal.(1994)noted,practicedandbaselinecategoriesarerepresentedinacommonepisodiccon-text.Retrievalpracticemaythereforesuppressitemsinbaselinecategoriesbecausetheysharecontextualfea-tureswithitemsundergoingretrievalpractice.Totheextentthatbaselinecategoriesarealsosuppressedbyinhibitoryprocesses,theabilitytodeterminehowmuchinhibitionhastakenplaceonwithin-categorycompetingexemplarsiscompromised.Thispossibilityisarguablyconsistentwithseveralfindingsintheoutputinterfer-enceandretrievalpracticeliteratures.Forexample,recallprobabilitydeclinesforcategoriesorpairedassociatesthataretestedlaterinatestingsequence,evenwhenthosecategoriesorpairedassociatesarenot
428M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
apparentlysimilaranddonotexplicitlyshareanycues(e.g.,Roediger&Schmidt,1980;Smith,1971).Fur-thermore,TsukimotoandKawaguchi(2001)foundthatbaselinecategoriescanbeimpairedbyretrievalpractice,ascomparedtoperformanceonthosesamebaselinecategoriesinacontrolgroupwhodidnotperformre-trievalpractice.Thesefindingssuggestthatsharedcon-textualfeaturesmayprecipitatebaselinesuppression,reducingmeasuredinhibition.Thisfactorthusmasksinhibitionbecauseitaffectsestimatesoftheamountofinhibitiononcompetingitemswithoutalteringthetruelevelofinhibitionthatactuallytookplaceforthoseitems.
Asecondsourceofbaselinedeflationcanarisewhenbaselineandpracticedcategoriesaresimilar.Totheextentthatbaselinecategoriessharesemanticfeatureswithitemsthatareinhibitedbyretrievalpractice,re-trieval-inducedforgettingmaygeneralizetothosecate-gories.AndersonandSpellmanÕs(1995)crosscategoryinhibitionfindingsprovideacaseinpoint:PracticingreditemssuchasRed-Bloodnotonlysuppresseditemsex-plicitlystudiedundertheRedcategory(e.g.,RedTo-mato),butalsootherreditemsthatwerestudiedandtestedunderaseparatecategory(e.g.,FoodRadish).Evennon-redFooditems(e.g.,FoodBread)werein-hibitedaftersubjectspracticedRed-Blood,suggestingthattheinhibitionofitemsthatdirectlycompetedwiththeretrievalpracticetarget(e.g.,Red-Tomato)seman-ticallygeneralizedtootheritemsthatoverlappedwiththeminsemanticfeatures.Inasimilarvein,AndersonandBell(2001)foundthatpracticingsomefactsaboutatopic(e.g.,Theactorisplayingtheguitar)impairednotonlyotherfactssharingthattopic(e.g.,Theactorisplayingtheoboe),butalsofactsstudiedunderadifferenttopicbutsharingthesamerelationandcategory(e.g.,Theteacherisplayingthedrum).Thus,impairmentgeneralizedacrosstopics,basedonsemanticsimilarity.AndersonandBell(2001)wereabletomeasurethein-hibitionofthelatteritemsbecausetheyincludedaddi-tionalbaselinetopicsthatdidnotsharethesamerelationandcategorywithpracticeditems(e.g.,Theboxisinthewarehouse,Themopisinthepub).Thegen-eralizedsuppressionwascircumscribedtoitemswithspecificoverlapinsematicrelationswithitemsstudiedwiththepracticedtopic,andcouldnothavebeenpro-ducedbyaglobalcontextualsimilarity.Thesefindingsstronglysuggestthatstudiesofretrievalinducedfor-gettingneedtotakeappropriatemeasurestoensurethatwithin-subjectsbaselineconditionsareasdissimilarfrompracticedcategoriesaspossible.Baselinesuppres-sioneffectssuchasthesemightbeonereasonwhyAndersonandReder(1999)failedtofindevidenceforcue-independentimpairmentintheirfaneffectpara-digm:Alloftheirpropositionswereconstructedusingthesamesemanticrelationandobjectclass(allwere‘‘isin’’facts,suchas‘‘Thelawyerwasinthepark’’).
Retrieval-practicefactorsthatmoderateinhibition
Theamountofinhibitionthatoccursisalsolikelytodependontheamountofattentiongiventothedifferentcuesprovidedforretrievalpractice.Inparticular,anytypeofretrievalpracticethatminimizestheneedtore-solveinterferencebetweencompetingitemsisunlikelytoproduceinhibition.Consider,forexample,thestudybyAndersonetal.(2000)reviewedearlier.Whensubjectsweregiventhecategoryandaskedtorecalltheexemplarbasedonstemcues(e.g.,FruitOr___),significantre-trieval-inducedforgettingwasobserved;however,whensubjectsweregiventheexemplar,andaskedtorecallthecategory(e.g.,Fr___Orange),therewasnoimpairment.Thispatternislikelytohavearisenbecausethecueinthelattercase—Orange—wasassociatedtothecategory,butnottootherexemplarsinthecategory,eliminatingcompetitionthatwouldleadtoimpairment.Similarly,ifsubjectswereaskedtoperformretrievalpracticewithoutthecategorylabel(e.g.,Or_n_eforOrange),otherex-emplarsinthecategoryareunlikelytointerfereandthusmaynotbeimpaired.Subtlercasesmayalsobepossible.Forexample,evenwhensubjectsaregiventhecategoryandafragmentcueforretrievalpractice,subjectsmightfocustheirattentiononthefragmentcue—thatis,theymaysolvetheretrievalpracticetaskbycircumventinginterferencecausedbythesharedcategory.Thisseemsespeciallylikelywhenthefragmentcueishighlyinfor-mativeordrawsattention.Forexample,ifmultiplelettersareprovided(e.g.,FruitB_n_n_),subjectsmightspendmoreoftheirtimefocusingonthedistinguishingletterfeatures,tryingtosolvethefragmentby‘‘soundingthewordout.’’Ingeneral,anyfactorthatreducesat-tentiongiventothesharedcueandfocusesitonthedistinguishingcueislikelytoreduceactivationofcom-petitorsandthereforereduceinhibition.
Testfactorsthatmoderate,mask,orexaggerateinhibitionInourinitialstudiesofretrieval-inducedforgetting,wemeasuredsubjectsÕfinalmemoryperformancewithacategorycuedrecalltest.Subjectswereprovidedwitheachstudiedcategorynameinturn,andaskedtorecallallofthestudiedexemplarsinanyorder.Inhibitionhasbeenfoundconsistentlywiththistypeoftest(Anderson&Bell,2001;Andersonetal.,1994;Anderson&McCulloch,1999;Anderson&Spellman,1995;Butler,Williams,Zacks,&Maki,2001;Macrae&MacLeod,1999;Nader,Coles,Brigidi,&Foa,2001;Smith&Hunt,1999),evenwhenthesharedcueisnotcategoricalinnature(e.g.,Anderson&Bell,2001;Macrae&Ma-cLeod,1999).Othertestshavealsobeenused,however,notonlytocharacterizetherangeofconditionsunderwhichretrieval-inducedforgettingoccurs,butalsotoinfervariouspropertiesoftheeffect.Inthissection,Idiscusssomeoftheworkthathasbeendonewithal-ternativetestingformats,withanemphasisonfactorsthatmaymoderateormaskinhibitoryeffects.
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445429
Outputinterferenceeffects.Dependingonthetesttypethatoneuses,theamountofretrieval-inducedforgettingmayreflectatleasttwosources:impairmentarisingfromtheearlierretrievalpracticephase,andimpairmentproducedbythefinalrecalltest.Thefinaltestcontributesasecondsourceofimpairmentbecausethestrengtheningofpracticeditemsduringtheearlierretrievalpracticephaseleadsthoseitemstoberecalledearlierinthefinaltestsequence.Becausethiswilldelayunpracticedcompetitorsuntillaterintheoutputse-quence,theseitemsaresubjecttoadditionalretrievalinducedforgetting—thatis,theyaresubjecttoexagger-atedoutputinterference,relativetobaselinecategories.Thisbiasintheretrievalofpracticeditemsisinterestingbecauseitmayprovideonemechanismbywhichtheinhibitoryeffectsofretrievalcanbereinstatedonare-curringbasis,evenwheninitialinhibitoryeffectshavedissipated(Anderson&Bell,2001).However,thecon-tributionoftest-basedsourcesofimpairmentcanimpaircleartheoreticalinferencesabouttheconditionspro-ducinginhibition,andsoitisnecessarytoconsiderthisfactorinassessinginhibition.Suchinferentialdifficultiesaremostlikelytoariseintestformatsthatallowsubjectstoreportitemsinanyordertheywish,althoughtheyarenotrestrictedtothosetypesoftest.Theoretically,ex-aggeratedoutputinterferenceisneitheramoderatingnoramaskingfactor,becauseitdoesnotaltertheamountofinhibitionthatactuallytookplaceduringretrievalpractice,nordoesitpreventusfromseeingthiseffect;itdoes,however,alterthemeasuredestimateofretrieval-practicebasedinhibition.
Thecontributionofoutputinterferenceisofgreatestconcernintwovarietiesofexperiment:whenonewantstoestablishtheretrievalpracticephaseastheprimarysourceofimpairment,andwhenoneisconcernedwithvariationsintheamountofinhibitionthathaveoc-curredacrossdifferentconditionsorgroups.Knowingwhetherinhibitionprimarilyreflectseventsintheprac-ticephaseisimportant,forexample,indeterminingwhetherextrastudyexposurescauseinhibition.Re-trievalpracticeandextrastudyexposuresbothstrengthenthepracticeditems,sothatonadelayedre-calltest,thoseitemsarelikelytoberecalledearlyintherecallsequence.Ifsubjectsarefreetorecallitemsinanyorder,unpracticedcompetitorsinbothofthesecondi-tionswillbesubjecttogreateroutputinterference(testbasedretrieval-inducedforgetting)thancorrespondingitemsinbaselinecategories.Thus,evenifextrastudyexposuresproducednoinhibitionduringthepracticephase,significantimpairmentmightbeobservedonthefinaltest,leadingonetoconcludethatextrastudyex-posurescausedinhibition.Similarly,ifoneisconcernedwithhowlongretrieval-inducedforgettinglasts,onemustbesurethattest-basedsourcesofimpairmentdonotcontributetothemeasureofinhibition,oronemightbeledtobelievethatinhibitionlastslongerthanittrulydoes.Finally,whenonewantstocomparetherelativeamountofinhibitionacrosstwoconditionsorgroups,itisessentialtodisentangledifferentsourcesofinhibition.Greaterinhibitionmayoccurinonecondition,forex-ample,merelybecausegreaterassociativestrengtheningforpracticeditemsproducedgreateroutput-basedef-fectsinthatcondition.
Themoststraightforwardwaytodistinguishthepracticeandtest-basedcontributionstoretrieval-in-ducedforgettingistouseatestsuchascategory-plus-stemcuedrecallthatenablesonetocontrolrecallorder(Andersonetal.,1994).Inatypicalstudy,thefinaltestiscomposedoftrialsinwhicheachexemplariscuedwithitscategorynameandaoneletterstemfortheexemplar.Importantly,subjectsarefirstcuedtorecallalloftheunpracticeditemsfromacategory,thenallofthepracticeditems,orviceversa.Comparisonsarethenmadetobaselineitemstestedinthecorrespondinghalvesoftheirrespectivecategories.Itistypicallyas-sumedthatrecallimpairmentobservedwhenallun-practicedexemplarsaretestedbeforepracticeditemsmustreflectthelingeringeffectsoftheretrievalpracticephase,forthesimplereasonthatpracticeditemshaveyettoberecalled.Whenunpracticedexemplarsaretestedfirstinthisway,significantretrieval-inducedforgettingistypicallyfound(Anderson&Bell,2001;Andersonetal.,1994;Andersonetal.,2000;Andersonetal.,2000;Anderson&McCulloch,1999;B€auml,2002;B€auml&Hartinger,2002),althoughsometimesitisre-ducedinmagnitudefromtheeffectsobservedwithcat-egorycuedrecallwithoutletterstems.Thisfindingmakessensegiventheeliminationofoutputinterferencefromtheeffect.Incomparingtherecallofitemstestedinthefirsthalfoftheircategoriestothosetestedinthesecondhalf,outputinterferenceistypicallyobservedonthiskindoftest,reinforcingtheimportanceofisolatingthetwosourcesofimpairment.Byusingthistypeoftestingprocedure,severalstudieshavefoundthatstrengtheningcompetitorsdoesnotreliablyimpairre-lateditemswhenoutputinterferenceiscontrolled(forretrieval-inducedforgetting,seeAndersonetal.,2000;forlist-strengtheffects,seeB€auml,1997;seealsoB€auml,1996foraconceptuallysimilarfindingforretroactiveinterference).
However,usingcategory-plus-stemcuedrecallisnotsufficienttoensurethatoutputinterferencehasbeenadequatelymatchedacrossbaselineandpracticedcate-gories.Therearecasesinwhichoutputinterferencedifferencescanariseevenwhenrecallorderisfixed.Inparticular,category-plusstemcuedrecalltestsinwhichthepracticedandunpracticedexemplarsofacategoryarerandomlyinterspersedintherecallorderdonotadequatelycontrolforoutputinterference.Forexample,supposethatsubjectsstudytheitemsOrange,Banana,Lemon,Cherry,Apple,andGrapeasmembersoftheFruitscategory,andthenperformretrievalpracticeon
430M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
OrangeBananaandLemon.Onthefinalrecalltest,therecallsequenceGrape,CherryandApple,BananaOr-angeLemonwouldcontrolforoutputinterferencebi-ases,whereasApple,Orange,Cherry,Banana,Grape,andLemonwouldnot.Althoughthelatterformattestsitemsinafixedorderthatisconstantacrosspracticedandbaselinecategories,outputinterferenceisnotmat-ched.Thisisbecausewhenthecategoryispracticed,thereisamuchgreaterlikelihoodofrecallingtheprac-ticeditems(Orange,Banana,andLemon)thantheitemsinthecorrespondingpositionsforthebaselinecategory.Thus,moretest-basedoutputinterferencewillbeexertedonunpracticedcompetitorsinthepracticedcategorywhenpracticedandunpracticeditemsareinterspersed(atleastonthosethatfollowpracticeditems).Ifitisimportanttoensurethatinhibitioneffectsarenotbeingproducedatthetimeofoutput,weakeritemsshouldbetestedbeforestrengtheneditems.
Cueprimingasamaskingfactor.Whensubjectsperformretrievalpractice,theyaretypicallypresentedwithacategorynameandthefirsttwolettersoftheexemplarthattheyaretoretrieve.Ifretrievalissuc-cessful,thepracticeditemisfacilitated,andcompetingitemsaresuppressed.However,retrievalpracticeintro-ducesanotherfactoraswell.Giventhatthepracticedcategoryistypicallypresentedninetimesinthestandardretrievalpracticesession(3exemplarsarepracticedthreetimeseach),thecategorynameenjoysasubstantialboostinaccessibility.Insomecircumstances,thiscueprimingcanreducetheamountofinhibitionthatismeasured,withoutactuallyinfluencingthelevelofin-hibitionthattakesplace.
Theeffectofcueprimingonmeasuresofinhibitioncanbeseeninourfirstexperimentonretrieval-inducedforgetting(Anderson,1989).Thisexperimentemployedthebasicretrievalpracticeparadigm,exceptthatweusedfreerecallasourfinaltestinsteadofcategorycuedrecall.TheresultscanbeseenintheleftsideofFig.5Aforthepracticed,competitor,andbaselineconditions.Asmightbeexpected,retrievalpracticefacilitatedthedelayedrecallofpracticeditems;moresurprising,however,retrievalpracticealsofacilitatedunpracticedcompetitors,relativetoitemsinunpracticedcategories.AcomparableresultcanbeseenintherightsideofFig.5A,whichdepictsthefindingsofahighlysimilarbutindependentexperimentbySmithandHunt(2000).ThefindingsofAnderson(1989)andHuntandSmith(1998)appearinconsistentwiththenotionthatretrievalpracticesuppressesrelateditems,ashasbeenarguedthroughout.Indeed,fromabehavioralstandpoint,theseresultsindicatethatundersometestingconditions,re-trievalpracticeenhancestherecallofrelateditems.However,toconcludethatnoinhibitionoccurredintheseexperimentswouldbeamistake.Bothexperimentsusedatestingformat(freerecall)thatpermitscueprimingtoinfluencehowmuchinhibitionismeasured.Specifically,withcategorizedwordlists(andorganizedlistsingeneral),itiswidelybelievedthatsubjectsadoptahierarchicalretrievalschemeforrecallingstudyitems;retrievalprogressesfirstfromarepresentationoftheepisodiccontextinwhichitemsarestudied,tothecat-egoriesonthelist,andnextfromthecategoryrepre-sentationstotheparticularexemplars(see,e.g.,Rundus,1973).Giventhismulti-stageprocess,thelikelihoodofrecallinganexemplarisinfluencedbytwoprobabilities:theprobabilityofrecallingthecategorylabel,giventhecontextasacue,andtheprobabilityofrecallingtheexemplar,giventhatthecategorylabelhasbeenrecalled.Thecombinationofthesefactorsdetermineshowwellpracticeditems,unpracticedcompetitors,andbaselineitemswillberecalled.Ordinarilywhencategorycuedrecallisused,theprobabilityofrecallingthecategorylabelsisconstantat1.0,becausethelabelsareprovided.However,onfreerecalltests,biasesincategoryrecallacrossconditionsbecomeanissue,particularlywhenmorethanjustafewcategoriesareusedandsubjects
Fig.5.ExamplesofcueprimingeffectsinfreerecallinstudiesbyAnderson(1989)andHuntandSmith(1998).SubjectsunderwentthestandardretrievalpracticeprocedureofAndersonetal.(1994)andweretestedwithfreerecallinsteadofcategory-cuedrecall.(A)Percentageofpracticeditems,unpracticedcompetitorsandbaselineitemsrecalledonthefinalfreerecalltest.Practicefacilitatedthepracticeditemsaswellastheunpracticedcompetitors.(B)Thesamedataasin(A),countingonlythoseitemsfromcategoriesforwhichsubjectsrecalledatleastoneitem(ensuringcategoryaccess).Conditionalizingrecallinthiswayrevealsasignificantretrievalinducedforgettingeffectthathadbeenmaskedbyprimedaccesstocategorylabels.M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445431
maythusforgetcategories.Giventhatpracticedcate-gorylabels(e.g.,Fruits)areprimed,theyarequitesa-lienttosubjects,leavingbaselinecategoriesatarecalldisadvantage.Theendresultisthatsubjectsaremorelikelytoforgetwholebaselinecategories,andmisstheopportunitytorecallexemplarsfromthosecategories.Thus,suppressionofunpracticedcompetitorsbyre-trievalpracticemaybemaskedbytheoveralladvantageintheaccessibilityofpracticedcategories.
ToillustratehowcueprimingmaskedatruedeficitinexemplaraccessintheAnderson(1989)andHuntandSmith(1998)studies,thefreerecalldatawerereanalyzedtofocusononlythosecategoriesfromwhichsubjectsrecalledatleastoneexemplar.Ourassumptionwasthatsubjectswhorecalledatleastoneexemplarfromacat-egoryhadaccessedthecategory.Restrictingtheanalysistothosecategorieswouldallowustoexamineexemplaraccessforthepracticedandbaselineconditions,un-contaminatedbydifferencesincategoryaccess.AscanbeseeninFig.5B,thisanalysisrevealedapatternofretrievalinducedforgettingquitesimilartowhatisor-dinarilyobserved,withunpracticedcompetitorsbeingrecalledmorepoorlythanbaselineitems.Additionalanalysesconfirmedthattheprobabilityofforgettingwholecategories(i.e.,category‘‘dropout’’)wasmuchhigherforbaselinecategories(20%)thanitwasforpracticedcategories(1%).HuntandSmith(1998)foundaverysimilarpattern,ascanbeseenintherightsideofFig.5B.Thesefindingsillustratehowcueprimingcanmaskinhibitioneffectswhenamulti-stagerecallprocessislikely,asitisonfreerecalltests.
However,cueprimingeffectsarenotlimitedtofreerecall,nortocategorizedwordlists.Considerthestudyofpropositionalretrieval-inducedforgettingbyAnder-sonandBell(2001).Whensubjectsperformedretrievalpracticeonpreviouslylearnedfactssuchas‘‘Theantcrawledontherock,’’thelaterrecallofotherfactssharingthattopicsuchas‘‘Theantcrawledontheta-ble,’’wasimpairedrelativetobaselinefactssuchas‘‘Theactorlookedatthepainting.’’However,AndersonandBellcuedsubjectsontheirfinaltestwiththetopicandtherelation(e.g.,‘‘Theantcrawledonthe____,’’and‘‘Theactorlookedatthe_______’’),sometimestogetherwithaletterstem.Ifwehadinsteadsimplygivensubjectsthecue‘‘Theant’’and‘‘Theactor,’’thefinaltestwouldlikelyhavebecomeamulti-stagerecalltest,eventhoughfreerecallwasnotused.Thisisbecauseweusedmanydifferenttopicswithdifferentsemanticrelation-ships(e.g.,iscrawlingon,islookingat,isin,iseating,etc),mostofwhichcouldbepairedwithanytopicandsocouldnotbeeasilyguessed.Givenonlythetopicasacue,subjectswouldhavehadtorecalltheactivityorrela-tionshipthatthetopicwasengagedinfirst,followedbytheobjectsofthatactivity.Becausesubjectspracticedthreefactsforeachpracticedtopicthreetimeseach(e.g.,threethingsthattheantcrawledon),thesemanticrelationshipsforthepracticedtopicwouldhavebeenfarmoreaccessiblethanthesemanticrelationshipusedinthebaselinetopics.Thissuggeststhateveniftheobjectoftheunpracticedcompetitor‘‘Theantcrawledonthechair’’wassuppressedduringthepracticeof‘‘theantcrawledontherock,’’thissuppressionwouldbemaskedbyheightenedavailabilityofthesharedrelationalcon-ceptunlessitwasprovidedasafinaltestcue.
Thus,thecomplexitiesintroducedbycueprimingarenotlimitedtocategorizedwordlistsortofreerecall.Indeed,cueprimingmayeveninfluencesimplepairedassociatesteststotheextentthatsubjectslinkthestimulusandresponsememberswitharelation(e.g.,whenencodingthepairBirdWorm,therelation‘‘eats’’islikelytobeinferred)thatmaybeforgottenindepen-dentlyofmemoryfortheresponse.Thisfactormakesitcrucialtoconsiderwhetherthestructureofthematerialsusedinagivenparadigm,whencoupledwiththetestformat,mightmaskinhibitionthroughcuepriming.Maskingthroughtransfer-inappropriatetestingeffects.Whetherinhibitoryeffectswillbeobservedshouldde-pendonthedegreetowhichthememorytracetappedbytheretrievaltestmatchesthetracethatwasinhibitedbyretrievalpractice.Toillustratethis,supposethatasubjectencodesthepairsTree-Prune,Tree-Rock,andTrimPruneandthendoesretrievalpracticeonTree-Rock.Lateron,supposethatsubjectsÕmemoryforPruneistestedeitherbycuingwithTree-P___orTrimP___.Ifretrieval-inducedforgettingisfoundwithTree-P__,butnotTrimP___,woulditmeanthatimpairmentiscue-dependent?Ifso,woulditmeanthat‘‘Prune’’wasneverinhibited?Atfirstglance,itmightseemso,totheextentthatcue-independenceisanessentialfeatureofinhibition.Afterall,thesetestsvaryinthecuesthattheypresenttosubjects,soifimpairmentdependsonwhichcuesareused,itmustobviouslybecuedependent.Thiswouldappeartocontradictthepropertyofcue-inde-pendence.However,thisconclusiondoesnotnecessarilyfollow.
Theproblemisthattheforegoingargumentfailstoconsiderthedistinctionbetweenthenominalformofastimulus,anditsfunctionalrepresentationbysubjects.Althoughfromthestandpointoftheexperimenter,thewordPruneisidenticalwhenpresentedinthepairsTree-PruneandTrim-Prune,theunderlyingrepresentationsformedbysubjectsmaynotbe.WhenstudyingTree-Prune,subjectsmighthaveencodedpruneÕsfruitsense,butwhenstudyingtrim-prune,theycertainlywouldencodeitsverbmeaninginstead.WhenretrievalpracticewasperformedonTree-Rock,anepisodicrepresenta-tionincludingthefruitsenseofprunemayhavebeensuppressed,makingitlessaccessiblewhentestedwith‘‘TreeP___.’’SuchinhibitionwouldnotbeexpectedtomaterializeonthetestTrimP____,however,becausethistesttapssubjectsÕepisodicmemoryforanentirelyunrelatedconceptthatwasneverinhibited(abetter
432M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
independentprobe,inthiscircumstance,wouldhavebeenFruitP___).Inessence,theindependentprobeTrimP___issimplynottestingthesameepisodicrep-resentationthatwasinhibited,eventhoughitmayseemthesamefromtheexperimenterÕsstandpoint.Thisex-ampleillustrateshowthecue-independencepropertyofinhibitionpertainstotheparticularfunctionalrepresen-tationthatisformedbythesubject:giventhatarepre-sentationisinhibited,itsrecallshouldbeimpaired,andthisimpairmentshouldbeobservablefromavarietyofcuesthattapthatparticularrepresentation.Forthesereasons,whendesigningteststodeterminewhetherornotinhibitionispresent,itisessentialtoensurethatthetestmightreasonablybeexpectedtotaptherepresen-tationthatwasinhibitedbythesubject.Ifnot,transfer-inappropriatetestingmaymasktheinhibitionthatac-tuallyoccurred.
Transfer-inappropriatetestingeffectsmaynotbelimitedtostimulithathavedifferentmeanings,ortotheuseoftheindependentprobemethod.Theseeffectsmayalsoarisewhenmultiplelevelsofrepresentationarepossible.Forinstance,duringwordencoding,ortho-graphic,phonological,andconceptualrepresentationsmayeachbeformed(dependingontheorientingtask),andtheserepresentationsmaybefunctionallyandan-atomicallydistinct(seeBalota,1994,forareview;seealsoRoskies,Fiez,Balota,Raichle,&Petersen,2001foradiscussionofanatomicallocalizationofthesedifferentlinguisticcodes).Ifdifferentlevelsofrepresentationareformedforthesamenominalverbalstimulus,thereispotentialfortransfer-inappropriatetestingtoattenuateormaskinhibition.Toseethis,supposethatperformingretrievalpracticeusingacategoricallydrivencued-recalltestsuchasFruitsOr___(forFruitsOrange),inhibitsconceptuallybasedepisodicrepresentationsofcompet-ingfruitssuchasBanana.Ifthisconceptuallybasedrepresentationisstructurallydistinctfromthephono-logicalandorthographicrepresentationsformedduringtheinitialprocessingofBanana,littleinhibitionwouldbeexpectedforBananaonorthographicorphonologi-callyorientedtests.Retrievalmaysimplyfailtomakecontactwiththerepresentationthatwasinhibited.Theunderlyingprinciplebehindthisideareceivessomesupportfromfindingsinthelevelsofprocessinglitera-ture:Manipulatingthelevelofprocessingofwordsatencodinghasdramaticeffectsonlaterrecallandrecog-nitiontests,but,theseeffectscandisappearorevenre-versewhenthefinalexplicitmemorytestfocusessubjectsonthelexicalandphonemicpropertiesofwords(e.g.,Fisher&Craik,1977;McDaniel,Friedman,&Bourne,1978;Morris,Bransford,&Franks,1977).Onpercep-tuallydrivenimplicitmemorytaskssuchaswordfrag-mentcompletion,wordstemcompletion,andperceptualidentification,levelsofprocessinghaslittleeffect(e.g.,Jacoby&Dallas,1981;Roediger,Weldon,Stadler,&Rieger,1992;seeRoediger&McDermott,1993,forareview).Iforthographicorphonologicaltestsarelesssensitivetoincreasesintheaccessibilityofconceptuallycodedinformationaboutaword(asisshownbylevelsofprocessingdissociations),itseemspossiblethattheymightalsobelesssensitivetodecreasesintheaccessi-bilityofthosecodesproducedbyinhibition.
Theideathatcategoricallydrivenretrievalpracticeprimarilyinhibitsconceptualasopposedtoortho-graphicorphonologicallevelsofanalysisreceivessomesupportfromarecentstudybyButleretal.(2001).Theseinvestigatorsemployedtheretrievalpracticepar-adigm,butvariedthenatureofthefinalrecalltest.Differentgroupsweretestedwiththestandardcategorycuedrecalltest(e.g.,presentationofthecategory‘‘Bird’’),orwithoneofseverallexicallyorientedimplicitandexplicitrecalltestssuchaswordfragmentcomple-tion(e.g.,cuingsubjectswith_p_r_owfortheword‘‘Sparrow’’withafreecompletioninstruction),wordfragmentcuedrecall(e.g.,_p_r_owwithanexplicitre-callinstruction),category-plus-fragmentcuedrecall(e.g.,Bird,_p_r_ow)orcategory-plusstemcuedrecall(e.g.,BirdSp_____).Withtheexceptionofcategorycuedrecall,thesetestsfocussubjectsÕattentiontovaryingdegreesontheorthographicandphonologicalfeaturesofthecuedwords.Subjectsarelikelytocom-pletethefragment_p_r_ownotprimarilythroughcon-ceptuallydrivenepisodicrecall,butby‘‘soundingout’’theanswerbasedongeneralknowledgeofwordforms.Ifso,retrieval-inducedforgettingmaybeattenuatedbecausethetestweightsalevelofrepresentationdiffer-entfromtheonethatisinhibited.Consistentwiththis,Butleretal.foundnoretrieval-inducedforgettingonanytestsinvolvinglettercuing.Theseresultsarecom-patiblewiththeideathatretrieval-inducedforgettingprimarilyaffectsconceptuallybasedrepresentations.Unfortunately,itisdifficulttoattributetheseeffectstotransfer-inappropriatetestingbecauseButleretal.Õsex-perimentsarelikelytobecontaminatedbyintegrationstrategiesduringencoding.Subjectsweregiven8stostudyeachexemplarinsteadoftheusual4–5s,apro-cedurelikelytoincreaseintegration(Anderson&Bell,2001;Anderson&McCulloch,1999).Thisseemsespe-ciallyplausible,giventheunusuallysmallamountofretrieval-inducedforgettingthattheyfoundintheircategorycuedrecallcondition(5%,comparedtothetypical9–20%).
Thenotionoftransferinappropriatetestingispar-ticularlyimportanttoconsiderinconnectionwithexperimentsexaminingwhetherretrieval-inducedfor-gettingaffectsperformanceonimplicitmemorytests.Onemightarguethatifretrievalpracticetrulyinhibitscompetitors,effectsshouldbeobservedonindirectmemorytests.Cautioniswarrantedhere,however,be-causenotallindirectmemorytestsarethesame.Manyofthemostcommontestsareperceptuallyoriented,suchaswordfragmentcompletion,lexicaldecision
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445433
(somevarieties)andperceptualidentification.Totheextentthattheseteststapperceptuallybasedrepresen-tations,theywouldnotbeexpectedtoyieldevidenceforinhibition,regardlessoftheirimplicit/explicitstatus.Abetterstrategywouldbetouseconceptuallydrivenin-directtestssuchasfreeassociation,semanticfluencyandperhapscategoryverification,whichwouldbesensitivetovariationsintheaccessibilityofasemanticrepre-sentation.Consistentwiththispossibility,recentstudieshavefoundretrieval-inducedforgettingonconceptuallydriven,butnotperceptuallydrivenimplicittests(Per-fect,Moulin,Conway,&Perry,2002;seealsoMoulinetal.,2002forfurtherevidenceofimpairmentonaconceptualimplicittests).
Theforegoingdiscussiondoesnotimplythatper-ceptuallyorientedmemoryrepresentationscannotbeinhibitedbyretrievalpractice.Indeed,thetypeofrep-resentationaffectedbyinhibitionshouldbedrivenbywhichrepresentationscausecompetitionduringre-trieval.Thisshouldbedeterminedinpartbythenatureofthecuesguidingretrievalpractice,andbysubjectsÕretrievalgoals.IfthesubjectisaskedtoretrieveastudiedwordthatbeginswiththelettersAc,ortho-graphicallysimilarcompetitorsmaybemoreinhibitedthansemanticallyrelatedcompetitors.Althoughthishasnotbeentested,relatedresearchonimplicitmemoryisconsistentwiththispossibility.Forinstance,Rajaram,Srinivas,andTravers(2001)foundthattheamountofrepetitionprimingexhibitedforawordoneitherawordfragmentorstemcompletiontestwassignificantlyre-ducedwhensubjectshadtoignorethatwordÕsidentityduringencoding.Whensubjectswerepresentedwithawordcoloredinred,blue,green,oryellow,andaskedtoquicklyidentifythecoloroftheword,subjectsexhibitedlessprimingthanwhentheysimplyhadtonametheworditself.Althoughonemightattributereducedprimingtoreducedencodinginthecolornamingcon-dition,Rajarametal.establishedthatthewordshadbeenidentifiedsufficientlytocausecompetitionwithcolornaming;thereactiontimetonamethecolorofawordwassignificantlyslowerthanthetimetonameaneutralstimulus(e.g.,arowofXes).Rajarametal.arguedthatthediminishedrepetitionprimingreflectstheinhibitionoftheworditself,drivenbytheneedtofocusattentiononthecolorattributeofthewordduringthecolornamingtrial—aprocesstheyrefertoasdese-lection.Ifcorrect,thisviewsuggeststhatretrievaldrivenbyoneperceptualattributeofastimulus(e.g.,color)maysuppressotherperceptualaspectsofthatstimulusthatcauseinterference(e.g.,visualwordform).Thiseffectmaylaterbeobservedonaperceptuallydrivenimplicittestthatreliesontherejectedattribute.Analo-gousdynamicsmaybepartiallyresponsibleforcertaincasesofimplicitmemoryblocksdrivenbyorthographyofaword(Smith&Tindell,1997).Thestandardretrieval-practiceexperimentwithcategoricallydrivenretrievalpracticemayonlyrevealinhibitiononcon-ceptuallydriventestsbecauseretrievalpracticeiscon-ceptuallyoriented.
Evenifconceptuallydrivenimplicittestsdidnotshowinhibition,however,itwouldnÕtbyitselfimplythatinhibitioneffectsdidnotoccur.Hereagain,itremainspossiblethatthelackofimpairmentontheimplicittestmaybeduetotransfer-inappropriatetestingeffects.Theoretically,itseemsreasonabletodistinguishbetweenageneralsemanticrepresentationofanitem(e.g.,Ba-nana)andanepisodicrepresentationofthatitemasitappearedonastudylist.Theepisodicrepresentationoftheitemmaybecomposednotonlyofdistinctivecon-textualfeatures,butalsoinstantiationsofsemanticfeaturesgenerallyusedtorepresenttheiteminsemanticmemory.Totheextentthatsuchanepisodicrepresen-tationisatleastpartiallystructurallydistinctfromthegeneralsemanticrepresentationoftheitem(theepisode-specificcomponentresidingperhapsasaboundsetoffeaturesinthehippocampus,asopposedtoneocortex;see,e.g,Norman&OÕReilly,inpress),wemustconsiderthepossibilitythattheepisodecanbesuppressedwith-outaffectingthegeneralconceptofBanana(Anderson&Bell,2001).Thisformofepisode-specificinhibitionmaybeparticularlylikelywhenepisodicretrievalprac-ticeisperformed,asinmoststudiesofretrieval-inducedforgetting;becauseretrievalpracticeisguidednotonlybyacategoryandaletterstem,butalsobyacontextualrepresentationofthestudylist,theepisodicrepresen-tationofacompetingitemmaybetheprimarysourceofcompetition,notthesemanticrepresentationofanitem.Infact,researchhasdemonstratedthatepisodicrepre-sentationscanindeedbeinhibited:CiranniandShi-mamura(1999)foundevidencethatnovelvisuo-spatialrepresentationscanbeinhibitedbyretrievalpractice,eventhoughtheserepresentationsclearlydonothavewelllearnedsemanticcounterparts.Itmaythereforebepossibletoobserveepisodespecificinhibitioninmoretraditionalretrieval-inducedforgettingexperimentsinwhichthematerialsalsohappentohaveacorrespondingrepresentationinsemanticmemory.Ifepisodespecificinhibitionispossible,suchinhibitioneffectsshouldgeneralizetoindependentretrievalcuesusedtotestac-cessibilityofthatepisode(onanexplicittest),evenwheneffectsdonotappearonimplicittests.
AlthoughtheCiranniandShimamurafindingsindi-catethatepisode-specificinhibitionmayoccur,anum-berofconsiderationssuggestthatthismaynotprovideageneralaccountofretrieval-inducedforgetting.First,thereisevidencethatsemanticandepisodicretrievalcompetitionarenotsocleanlyseparable,atleastinstudiesofinhibition.Forinstance,semanticretrievalpracticehasbeenshowntoimpairepisodicrepresenta-tionsofsimilaritems(B€auml,2002),andpart-setcuingofepisodicallypresenteditemsappearstoimpairaccesstosemanticallyrelatedcompetitors(Kimball&Bjork,
434M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
2002;seelatersectiononrecognitiontestingforfurtherdiscussion).Second,toadoptepisode-specificinhibitionasatheoryforallvarietiesofepisodicretrieval-inducedforgettingignoresaveryplausiblefeatureofretrieval:subjectscanweightthedifferentcuesthattheyuseflexibly,dependingonthetask.Insometasks,episodiccontextmaybethemostimportantcuetoweight,whereasinothers,thesemanticcategorymaybemorediagnosticinguidingretreival.Ifso,whetheroneob-servesepisode-specificinhibition,orjointeffectsofin-hibitiononbothepisodicandsemanticrepresentationsmayhingeontherelativeweightingofattentiononcontextualversuscategoricalcues.Finally,therela-tionshipbetweenepisodicandsemanticrepresentationsisatpresentnottheoreticallyresolved:episodesmayormaynotbestructurallydistinguishablefromtheirse-manticcounterparts.Theseissuesremaintobeexploredingreaterdepth.Nevertheless,inanystudylookingatwhetherepisodicallyinducedinhibitionmaybeobservedonconceptualimplicitmemorytests,itwouldbepru-denttoentertainepisode-specificinhibitionasatheo-reticalmechanismthatmaycontributetoperformance.Theforegoingexamplesillustratethecentralimpor-tanceofconsideringthenatureoftherepresentationthatislikelytobetappedbyaparticularvarietyoftest,andhowthisrepresentationmayrelatetotheonelikelytobesubjecttoinhibition.Failuretofindevidenceofinhibitiononagiventestmaynotindicatealackofinhibitioningeneral;itmaysimplyreflectamismatchinthetypeofrepresentationtappedbythetestandthataffectedbyinhibition.Nevertheless,althoughindirecttestsmaynotbediagnosticofinhibition,suchexperi-mentsdoservetodefinethescopeofinhibitoryeffectsinducedbyepisodicretrievalpractice,andthenatureoftherepresentationsaffected.
Maskingthroughcovertcuingeffects.Asdescribedearlier,inhibitiontendstogeneralizetonoveltestcuesthatareunrelatedtotheitemsreceivingretrievalprac-ticeortothepracticedcuesthemselves—apropertyknownascue-independence.However,whethercue-in-dependentforgettingwillbeobservedmaydependonwhethersubjectsusecovertcuingstrategiestoaugmenttheirrecallonthefinalmemorytest.Consider,forex-ample,astudybyAndersonetal.(2000).Inthisstudy,subjectsstudieditemssuchasRed-BloodandRed-To-mato,andlaterdidretrievalpracticeonRed-Blood.Onadelayedrecalltest,subjectswerecuedtorecallTomatowithanextra-listcategorylabelandaletterstem(e.g.,Food-T___)toseewhetherornotanyinhibitionthatwasinducedbyretrievalpracticewouldgeneralizetothenovelextralisttestcue(seeAnderson&Green,2001;Johnson&Anderson,inpress;Levy&Anderson,2002;forotherstudiesusingextralistcuing).Aspredicted,significantinhibitionwasfound,suggestingcue-independentimpairment.However,whenasked,onapost-experimentalquestionnaire,whethertheytriedtoaugmenttheirmemorysearchbyrecallingearlier-stud-iedcategories,somesubjectsreportedusingthis‘‘covertcuing’’strategy(theaverageratingwas2.68ona5pointscale).Thus,whengiventheextralistcategorycueFoodT___,somesubjectsmayhavecovertlyrecalledthecategory‘‘RedThings,’’andusedthesetwocategoriesjointlytorecallitems.Subjectswhoreportedusingthisstrategyshowedmodestlyreducedinhibitioneffects,comparedtosubjectswhodidnotusethisstrategy(areductionoftheinhibitioneffectby3%inExperiment1,andby7%inExperiment2).Giventhattheusefulnessofcovertcuingmayhavebeenlimitedbythetimingcon-straintsusedinthetestofthatexperiment(4spercue),thesefindingssuggestthatcovertcuingmayacttore-duceinhibitionunderlessconstrainedconditions.
Theforegoingfindingsmaybeunderstoodbycon-sideringtheeffectsofcueprimingdiscussedearlier.Totheextentthatpracticedcategoriesaremadehighlyac-cessiblebyretrievalpractice,subjectswhoengageincovertcuingaremorelikelytocovertlygeneratethepracticedcategoriesthantheyarethebaselinecatego-ries.Asaresult,whentryingtorecallinhibiteditems,subjectsusingthisstrategyshouldbemorelikelytohavenotone,buttwocategorycuesattheirdisposal,con-ferringacuingadvantagetothoseitems,relativetobaselineitems.Thus,inhibitionmaybecompensatedbythedifferentialavailabilityofcompoundcuing.Suchcompensationwouldleadtoaninaccuratemeasureoftheamountofinhibitionthathadinitiallytakenplace(masking),andperhapsevenundothatinhibitionfortheitemsretrievedbycompoundcues.
Toreducethelikelihoodofcovertcuingcontami-natingrecallperformanceinstudiesusingtheindepen-dentprobemethod,severalstrategiesappeareffective.First,subjectsarelesslikelytousecovertcuingwhentheextralistcuesare,ingeneral,stronglyrelatedtothetargetitem;ifmostcuesarepoorlyrelated,subjectsmaylookforadditionalinformationtosupplementtheirre-call.Second,providinganitemspecificcuesuchasaletterstemfocusessubjectsonrecallingaparticularitem.Third,limitingtheamountoftimethatsubjectshavetorecallthecriticalitemdiscouragestheuseofcomplexsearchstrategiessuchascovertcuing.Fourth,usingalargenumberofstudiedcategoriesmakesitunlikelythatsubjectswillbeabletorecalltherelevantstudiedcategory,eveniftheytry.Finally,administeringpost-experimentalquestionnairestoobtainsubjectivereportsofcovertcuingcanhelptoassesswhethertheforegoingstrategieswereeffective.Inusingtheinde-pendentprobemethodtoestablishthetheoreticalpropertyofcueindependence,itisvitaltoconsiderhowsuchstrategiesmayaffectperformance.
Specialissuesinrecognitiontesting.Initially,webe-lievedthatretrieval-inducedforgettingwouldnotoccuronrecognitionmemorytests(Anderson&Bjork,1994).Thisexpectationwasbasedonanalogiestootherin-
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445435
hibitoryphenomenasuchasdirectedforgettingandretroactiveinterference,whichexhibitlittleimpairmentonrecognitiontests,andontheideathatpresentationoftheitemitselfwouldreleaseitfromitsinhibitedstate(Anderson&Bjork,1994).Thisperspectivehasproventobemistaken.Significantretrievalinducedforgettinghasbeenfoundonrecognitionmemorymeasures,bothintheretrievalpracticeparadigmandincloselyrelatedprocedures.
Thefirstdemonstrationofretrieval-inducedforget-tingonarecognitiontestusingtheretrievalpracticeparadigmwasreportedbyAnderson,DeKok,andChild(1997).Subjectsparticipatedinthestandardre-trievalpracticeprocedureexceptthatafterthe20mindelay,subjectsweregivenayes/norecognitionmemorytestforalloftheexemplarstheyhadstudiedinsteadofcuedrecall.InExperiment1,subjectsweretestedwithcategory–exemplarpairs,oneatatime,andexemplarsofagivencategoryweretestedinblocksof12(sixtar-getsandsixhighlysimilardistractorsintermixed).AscanbeseeninFig.6,significantretrieval-inducedfor-gettingwasobserved,regardlessofwhethertheun-practicedcompetitorsweretestedbeforepracticeditemsintheircategory(tested1st)orafterthem(tested2nd).Subsequentexperimentsprovidedevidencethatthisimpairmentalsooccurredwhenexemplarswerepre-sentedwithouttheircategorylabels,andinrandomizedtestsinsteadoftestsusingcategoryblocks.Thus,re-trieval-inducedforgettingcanbeobservedevenwhensubjectsaretestedwiththeinhibiteditempresentedin-tact,anddonothavetogeneratetheitemfromin-completecues.Andersonetal.(1997)alsoobservedwithin-categoryoutputinterferenceontheirrecognitiontests(toseethis,comparetestedfirsttotestedsecondinFig.6foreachcondition),consistentwithotherstudiesthathavereportedoutputinterferenceonrecognitiontests(Smith,1971).Similarinhibitioneffectshavebeen
Fig.6.Retrieval-inducedforgettinginrecognitionmemory(Andersonetal.,1997).Onafinalcategory–exemplarpairrecognitiontest,subjectswereimpairedintheirabilitytorec-ognizeunpracticedcompetitors,asmeasuredbycorrectedrec-ognition(hits-falsealarms).Thiseffectoccurredregardlessofwhetherunpracticedcompetitorsweretestedinthefirsthalfoftheirrespectivecategories,orinthesecondhalf.reportedintworecentexperimentsbyHicksandStarns(inpress)thatusedanitemrecognitiontest.(seeDop-kins&Ngo,2002,forapotentiallyrelatedinhibitioneffectinducedbyincidentalretrievalofanearlierpre-sentationofanitemduringitsrepetition).Radvansky(1999)alsofoundevidenceforinhibitiononaspeededrecognitionmemorytestusingthefaneffectprocedure.Inadditiontogeneralizingtheseeffectsbeyondcate-goricalmaterials,RadvanskyÕsstudydemonstratedthatimpairmentiscue-independent,aspredictedbythein-hibitionview.
Giventheevidenceforimpairmentonrecognitiontests,thequestionarisesastowhysucheffectswouldoccurforretrieval-inducedforgettingandnototherphenomenasuchasdirectedforgettingandretroactiveinterference.Althoughitispossiblethatretrieval-inducedforgettingmaybeproducedbyqualitativelydifferentmechanisms,otherexplanationsexist.Onepossibilityisthatrecognitiontestsmightbemostsen-sitivetoretrieval-inducedforgettingwhentherecogni-tionjudgmentsrequireactiverecollectionratherthanamereassessmentoffamiliarity.IntheAndersonetal.studyjustdiscussed,subjectswereaskedtoclaimthattheyrecognizedanitemonlyiftheywereveryconfidentthatithadoccurredintheearlierstudyphase.Theseinstructionsshouldhaveencouragedagreaterweightonrecollectiveprocesses.Ifso,perhapsdirectedforgettingmightalsobefoundonrecognitiontestsiftestsrequiredactiverecollection.Consistentwiththisidea,directedforgettingdoescauseimpairmentonrecognitiontestsrequiringsubjectstomakesourcememoryjudgments(e.g.,Geiselman,Bjork,&Fishman,1983).Thus,al-thoughsimpleyes/norecognitiontasksappeartobeinsensitivetodirectedforgetting,judgmentsthatrequireactiveretrievalofaparticularepisodictraceshowtheeffect,asinrecallparadigms.Althoughthisaccountreconcilesthepatternsofinhibitoryeffectsonrecogni-tiontestsacrossthetwoparadigms,itleavesunex-plainedwhyfamiliarity-basedjudgmentsmightfailtoexhibitinhibition.
Anotherdifficultythatmayariseisthepotentialforthedistractoritemsonarecognitiontesttobesup-pressed.Forexample,practicingFruitOrangemaysuppressnotonlyotherstudieditemssuchasFruitBanana,butalsononstudieditemssuchasFruitStrawberry.Becausenonstudiedexemplarsaretheveryitemsthatwouldbemostusefultoemployasdistrac-tors,bothtargetsanddistractoritemsmaybeimpaired.Sucheffectsoughttomakeitdifficulttousesignalde-tectionmethodologytomeasureinhibition.ConsidertheidealizedfamiliaritydistributionsinFigs.7A–C.Fig.7Arepresentsthesituationbeforeretrievalpracticehasta-kenplaceandshowsfamiliaritydistributionsforbase-lineitemsandtheirdistractors.Baselineitemsareassumedtobemorefamiliarthandistractorsduetotheirrecentpresentationonthestudylist,andsothe
436M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
Fig.7.Anillustrationofwhyitmaysometimesbedifficulttodetectinhibitiononrecognitionmemorytests,intermsofsignalde-tectiontheory.Eachfigurerepresentsacontinuumoffamiliarityvaluesforitemsstoredinmemory,withdistributionsforstudiedtargetsandnon-studieddistractoritemspresentedontherecognitiontest.(A)Beforeretrievalpractice,allstudieditemsarepresumedtobemorefamiliartosubjectsthanaredistractors.(B)Familiaritydistributionsforunpracticedcompetitorsandtheirdistractors,afterretrievalpracticehasbeenperformed,accordingtoModelA.InModelA,retrievalpracticeispresumedtoselectivelysuppresstheunpracticedcompetitorsandNOTtheircorresponding,highlysimilardistractorsinsemanticmemory.Thisleadstargetitemstobelessfamiliar,shiftingtheoverallfamiliaritydistributionforthoseitemstotheleft,closertothedistributionfordistractors,reducingd0.(C)Thesamefamiliaritydistributionsasplottedin(B),butplottedaccordingtoModelB.InModelB,retrievalpracticeispresumedtosuppressboththeunpracticedcompetitorsandthehighlysimilardistractors.Thisleadsbothtargetanddistractoritemstobelessfamiliar,shiftingthedistributionsforbothtotheleft,leavingd0unaffectedbysuppression.Thus,ifretrievalpracticesuppressesbothunpracticedcompetitorsandtheirdistractors,impairmentmaynotbeobservedonarecognitionmemorytest,becaused0willremainconstantforbaselineitemsandunpracticedcompetitors.baselinedistributionisshiftedtotheright.Figs.7BandCrepresentthesituationafterretrievalpractice,ac-cordingtotheviewthatretrievalpractice:(1)suppressesonlyotherepisodicallystudiedcompetitorsandnotdi-stractors,or(2)suppressesbothepisodicandsemanti-callyrelatedcompetitorsthatserveasdistractors.Ifinhibitionisrestrictedtoepisodicallyrelatedcompeti-tors(7B),impairmentshouldbemeasurableusingd0becauseretrievalpracticeselectivelyshiftsthedistribu-tionforunpracticedcompetitors,butnotthoseoftheirdistractors(notetheleftwardshiftofthetargetdistri-butioninFigs.7BandC).Nosuchshiftoccursforbaselinecategories(Fig.7A),soadifferenceind0shouldemerge.However,ifinhibitionalsoaffectsunstudiedsemanticallyrelatedcompetitors,bothdistributionswillbeshifted(Fig.7C).Becaused0onlyprovidesameasureoftherelativediscriminabilityoftargetsanddistractors,inhibitionmaybequitedifficulttomeasure,relativetobaselinecategoriesthathavenotshifted(seeSamuel,1996,forananalogoussignaldetectionanalysisinthecontextofspeechperception;asimilarpointwasalsomadeinthecontextoftherevelationeffectbyHicks&Marsh,1998).Thus,inhibitionmaybedifficulttodetectonrecognitiontestsnotbecauseinhibitionhasbeenre-leasedordoesnotaffectfamiliarity,butbecausethenatureofthetestrequirestheuseoffoilsthatarethemselvessuppressed.Hereagain,thewayinwhichthetestisadministeredyieldsaninaccuratemeasureofhowmuchinhibitiontrulytookplace,maskingthoseeffects.Thereisgoodreasontosuspectthatinhibitorypro-cessesrecruitedduringepisodicretrievalsuppresscom-petingitemsinsemanticmemory.First,retrievalinducedforgettingisageneralphenomenonthatoccursonbothsemanticandepisodicretrievaltests(e.g.,Blaxton&Neely,1983;Johnson&Anderson,inpress),showingthatsemanticrepresentationsaresusceptibletoinhibition.Second,inhibitioneffectshavebeenprevi-ouslyshowntospanepisodicandsemanticmemory.Retrievinganexemplarofacategoryfromsemanticmemorycansuppressepisodicmemoryforotherex-emplarsthatwerestudiedpreviously(B€auml,2002).Ifsemanticretrievalcansuppressepisodicmemory,itseemslikelythatepisodicretrievalmightalsosuppresssemanticallyrelatedcompetitorsthatarenotstudied.
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445437
Finally,inarecentstudy,KimballandBjork(2002)foundthatpresentingpart-setcuesduringarecalltestnotonlyimpairedremainingitemsthatwerestudiedinthatset(asexpected),butalsoreducedtheintrusionrateforcriticalnonstudiedsemanticitemsthattendtobemistakenlyrecalledwiththosesamematerials.Takentogether,theseresultssuggestthatthesensitivityofrecognitionteststoinhibitoryeffectsmaybemaskedbysuppressionofrelatedsemanticdistractors.
Summary
Theforegoingreviewhighlightsthecorepropertiesofretrieval-inducedforgettingandsomeofitsboundaryconditions.Takentogether,thesepropertiesargueforastrongparallelbetweenselectiveretrievalandthemoregeneralsituationofresponseoverride.Inparticular,theneedtoselectivelyretrieveatargetiteminthefaceofinterferencefromoneormoreprepotentmemoriesleadstothesuppressionofthosememories,andthissup-pressionunderlieslaterforgettingofthoseitems.Al-thoughinhibitoryeffectsaresometimesmoderatedormaskedbyrepresentationalortestingfactors,thebasicfindingisquitegeneralandlikelytounderliemanycasesofforgettingassociatedwithinterference.Theexperi-enceofforgettingismorelikelytobecausedbyinhib-itorycontrolprocessesthathelptofocusretrievalthanbythestrengtheningofcompetingassociationsinmemory.
Stoppingretrievalthroughinhibitorycontrol
Intheprecedingreview,wediscussedevidenceforinhibitoryprocessesinselectiveretrievalsituations,whichwearguedarelikelytorequireresponseoverride.However,responseoverrideisinvolvedinothersitua-tionsaswell,suchaswhenweneedtostoparesponsefromoccurringatall.Inmemoryretrieval,thisabilitycouldproveusefultopreventaparticularmemoryfromcomingintoconsciousness.Indeed,wesometimescon-frontremindersofthingsthatwewouldprefernottothinkabout:thesightofacarmayremindusofanaccidentwehad,orofaformersignificantotherwhodrovethattypeofcar;orthesightoftheworldtradecenterinanoldmoviemayleadustoterminatethenaturalprogressionfromcuestomemories.Othertimes,wemaywishtofocusonaparticularthoughtorideawithoutlettingthemindwander.Caninhibitorycontrolmechanismsbeengagedtoservethesegoals?Caninhi-bitionhalttheretrievalprocess?Ifso,how?AndersonandGreen(2001)recentlylookedatthisissuebyex-amininghowstoppingretrievalaffectedthememoriesthatweretoberetrieved.Tostudythis,theydevelopedanewproceduremodeledafterthewidelyusedGo/No-Gotask,whichhasbeenusedtomeasuretheabilitytostopaprepotentmotorresponseandtostudyitsneuralbasisinbothhumans(e.g.,Caseyetal.,1997;deZubicaray,Andrew,Zelaya,Williams,&Dumanoir,2000;Gara-van,Ross,&Stein,1999)andmonkeys(e.g.,Sakagami&Niki,1994).InoneversionoftheGo/No-Gotask,lettersarepresentedoneatatimeandsubjectsmustpressabuttonasquicklyaspossiblewhenevertheyseealetter,exceptwhentheletterisanX.WhentheyseeanX,theyaresupposedtoavoidpressingthebutton.Themajorityoftrialsaredesignedtorequireabuttonpress,sothatwhenanXoccurs,subjectshavedifficultywithholdingtheirmotorresponse.Theabilitytowith-holdtheresponseistakenasameasureofinhibitorycontrol.
Toexplorewhetherpeoplecanstopretrieval,An-dersonandGreen(2001)adaptedthego/no-gotasktocreatethethink/no-thinkparadigm.Inthisprocedure,subjectsstudiedpairsofweaklyrelatedwords(e.g.,flag—sword,ordeal—roach)andwerethentrainedtoprovidethesecondword(e.g.,roach;hereinafterre-ferredtoastheresponseword)whenevertheyweregi-venthefirstwordasacue(e.g.,ordeal).Subjectsthenenteredthethink/no-thinkphase,whichrequiredthemtoexertexecutivecontrolovertheretrievalprocess.Formostofthetrialsinthisphase,thetaskwasthesameasithadbeenduringtraining—torecallandsayaloudthecorrespondingwordasquicklyaspossibleatthesightofitsretrievalcue.Forcertaincues,however,subjectswereadmonishedtoavoidthinkingoftheresponseword.Itwasemphasizedthatitwasnotenoughtoavoidsayingtheresponseword—itwascrucialonthosetrialstopreventtheassociatedmemoryfromenteringconsciousawarenessatall.Thus,subjectshadtooverridenotonlyavocalmotorresponse,butalsothecognitiveactofretrieval.Couldsubjectsrecruitinhibitorycontrolmechanismstostopthememoryfromenteringcon-sciousness?
Ofcourse,AndersonandGreencouldnotdirectlymeasurewhethersubjectsstoppedthememoryfromenteringconsciousness,butifinhibitorymechanismswererecruited,laterrecalloftheexcludedmemoryshouldbeimpaired.Toexaminethis,immediatelyafterthethink/no-thinkphase,subjectsweregiventhecuesforallofthepairs,buttheywerenowaskedtorecalltheresponseforeachofthem.Asexpected,forgettingoc-curred:responsewordsthatsubjectstriedtokeepoutofawarenesswereimpairedcomparedtobaselinepairstheyhadstudiedinitiallybuthadnotseenduringthethink/no-thinkphase.Themoreoftensubjectstriedtostopretrieval,theworserecallfortheexcludedmemorybecame(seeFig.8A).Interestingly,avoidedwordswerehardertorecalleventhoughsubjectshadencounteredasmanyas16reminders(i.e.,cues)duringthethink/no-thinkphase.Undernormalcircumstances,remind-erswouldbeexpectedtofacilitatetheremindedmem-ory,muchasitdidfortheitemstowhichsubjectscontinuedtorespond(Fig.8A).AndersonandGreen(2001)furtherestablishedthatthisimpairmentwascue
438M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
Fig.8.FinalrecallperformanceinfourexperimentsreportedbyAndersonandGreen(2001)usingthethink/no-thinkpro-cedure.Eachplotrepresentsthepercentageofitemsthatsub-jectsrecalledonthefinalrecalltestasafunctionofthenumberoftimesthattheysuppressedtheitem(suppressioncondition),ortriedtorecallit(respond).Theleftpanelineachrowrep-resentsfinalrecallperformancewhentestedwiththeoriginallytrainedretrievalcue(i.e.,the‘‘Sameprobe’’),whereastherightpanelineachrowrepresentsfinalrecallperformancewhentestedwithanovel,independent,extralistcategorycue.(AandB)depictsperformanceinExperiment1;(CandD)depictsperformanceinanexperimentofferingmonetaryincentives,andencouragingguessingonthefinaltest;(EandF)depictsperformancewhensubjectsweremisledregardingtheexpectedoutcomeofthestudyjustbeforethetest;(G)depictsfinaltestperformancewhensubjectsaresimplyasked,duringthethink/no-thinkphasetosimplynotsaytheresponseword(withhold)insteadoftonotthinkaboutit;finalmemoryisnotimpaired.
independent,echoingtheresultsofAndersonandSpellman(1995):forgettingoccurredregardlessofwhe-thersubjectsweretestedwiththeoriginallystudiedcueword(e.g.,ordeal)orwithanovelindependentcueneverstudiedintheexperiment(e.g.,insectr____forroach;Fig.8B).Thiscue-independencearguesthattheforgettingisnotcausedsolelybyassociativeinterfer-ence;rather,impairmentreflectsactivesuppressionoftheexcludedmemoryitself.
AndersonandGreen(2001)ruledoutthepossibilitythatsubjectsmighthavedeliberatelywithheldanswersonthefinaltestduetoconfusionortoexpectationsaboutthepurposeoftheexperiment.Inoneexperiment,subjectsweretoldthattheywouldbepaidforallcorrectanswersandwereurgedtorespondtoeverycue,eveniftheywereguessing.Anothergroupwasmisledtobelievethattheexperimentersexpectedthattheirmemorywouldbebetterforwordstheyhadavoidedthinkingabout.Bothmanipulationslefttheinhibitionpatternunchanged(seeFigs.8C–Dforrecallperformanceintheoriginalcueandindependentcueconditionsrespectivelyforthemonetaryincentivesexperiment;seeFigs.8EandF,forthesameconditionsforperformancebymisledsubjects),demonstratingthatsubjectswereneitherconfusednorpurposefullywithholdingresponses.Inafinalexperiment,subjectsweremerelyaskedtoavoidsayingtheresponseoutloudandallmentionofpre-ventingitfromenteringawarenesswaseliminated.Noinhibitionwasobserved(Fig.8G),indicatingthattherecalldeficitsintheprecedingexperimentswerenotmerelyduetosuppressionofthevocalresponseforavoidedwords.Theseresultsisolateforgettinginthethink/no-thinkparadigmtoprocessesdirectedatkeep-ingtheunwanteddeclarativememoryoutofawarenessanddemonstratethatthiscognitiveacthaspersistingconsequencesfortheavoidedmemories.
TheimpairedmemoryobservedbyAndersonandGreen(2001)suggeststhatinhibitorycontrolmecha-nismsmayberecruitedwhenweseektoregulateawarenessofunpleasantorintrusivememories.Inpar-ticular,whenevertheenvironmentissuchthatitpresentsunavoidablereminderstosomethingthatwewouldprefernottothinkabout,peoplemayresorttocon-trollingtheirmemoriesinstead.Theendresultmaybeimpairedmemoryforthethingsthatpeopleavoidthinkingabout.Thissuggeststhatthethink/no-thinkparadigmofAndersonandGreen(2001)mayprovideausefullaboratorymodelofthevoluntaryformofre-pression(suppression)proposedbyFreud(Freud,1966).Ifso,resultsfromthisparadigmandotherrelatedparadigmssuchasthedirectedforgettingproceduremayhaveimplicationsforunderstandingclinicalphenome-nonrelatingtomotivatedforgetting(Anderson,2001;Anderson&Green,2001;Bjork,Bjork,&Anderson,1998;Conway,Harries,Noyes,Racsmany,&Frankish,2000;Deprince&Freyd,2001;Myers,Brewin,&Power,
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445439
1998;seeGolding&MacCleod,1998forareviewofdirectedforgetting).
RelationshiptoclassicalinterferencetheoriesofforgettingAlthoughtheexecutivecontrolviewisarelativelynewapproachtointerference,manyofitsaspectsre-semblecomponentsofclassicalinterferencetheory.Inthissection,Idiscusssomeofthespecificrelationsbetweenthisviewandfourmechanismsdiscussedinclassicalinterferencetheory:responsecompetition,un-learning,reciprocalinhibition,andresponse-setsup-pression.Theexecutivecontrolapproachvalidatesmanyoftheintuitionsbehindtheseclassicalproposals,whileatthesametimequestioningthehistoricalemphasisthathasbeenplacedonassociativelearningasasourceofforgetting.
McGeoch’sresponsecompetitiontheory
AccordingtoMcGeochÕsclassicalresponsecompe-titiontheory,attachingmorethanoneresponsetoaretrievalcueleadsthoseresponsestocompetewithoneanotherwhenthecueispresentedlateron.Themorecompetingresponses,orthestrongeracompetingre-sponsebecomes,themoredifficultitshouldbetorecallagivenitem.McGeochÕsemphasisontheimportanceofsharingaretrievalcueasaconditionofinterferencewas
inheritedfromM€u
llerandPilzecker(1900),andcon-tinuestodayintheformofrelativestrengthorratio-rulemodelsofretrieval(e.g.,Anderson,1983;Mensink&Raajimakers,1988).Inessence,thesetheoriespositthattheadditionofnewstructureintomemoryleadstotheocclusionorblockingofatargetevent.
ManyofthebasicassumptionsofMcGeochÕsre-sponsecompetitiontheoryareacceptedinthecurrentexecutivecontrolapproach.Forinstance,thepresenta-tionofaretrievalcueispresumedtoactivateallas-sociatedresponsesaccordingtotheirstrengthsofassociationtothecueandtheseresponsesarethoughttocompetewithoneanotherforaccesstoconsciousawareness.Itisthisretrievalcompetitionthatprecipi-tatestheneedforexecutivecontrol.Accordingtotheexecutivecontrolapproach,however,thiscompetitionisusuallynotenoughbyitselftoimpairmemoryrecallforatargetbecauseinhibitoryprocessesmaybedeployedtoovercomethecompetition.Furthermore,theempiricalrelationshipbetweenthenumberofcompetingresponsesandtheprobabilityofrecallingatargetitemisalsoac-ceptedbythetheory,alongwiththenotionthatstrengtheningacompetingresponseisempiricallyasso-ciatedwithadecrementinrecallforatarget.
WheretheexecutivecontrolapproachadvancedherediffersfromMcGeochÕstheoryisintheunderlyingmechanismthatproducestheserelationships.According
totheexecutivecontrolapproach,theprobabilityofrecallingatargetitemdoesnotautomaticallydecreaseasaconsequenceofaddingnewassociations,orasaconsequenceofstrengtheningacompetingassociation.Structuralchangesmayimpairthelaterrecallofatargetitemiftheyincreasethechancesthatnontargetitemswilloccasionallyberetrievedbeforethecriticaltarget.Totheextentthatcompetitorsareretrievedearlier,thetargetwillbesuppressedatoutput.Theprobabilitythatthissuppressionwillimpairtargetperformanceshouldgoupwiththenumberofcompetitorsbecausethiswillleadmorecompetitorstoberecalledbeforethetarget,onaverage.Bythisviewthen,strengtheningacompet-itorshouldnotimpairtargetrecallprovidedthatthetargetcanbeensuredtobetestedbeforethecompetitor,afindingthathasbeenobservedmanytimesnowinexperimentsevaluatingthehypothesisofstrength-de-pendentforgetting.Thus,itisnottheadditionofnewassociations,northeirstrengtheningthatimpairsmem-ory,butrathertheincreasedlikelihoodofsuppressioncorrelatedwiththosestructuralchanges.
Thereareseveralcircumstances,however,inwhichresponsecompetitionmightimpairmemory.First,wheneveracueispresentedthatisassociatedtoastrongerandaweakerresponseandthesubjectistoldtoonlyreportthefirstthingthatcomestomind,responsecompetitionmightunderlieinterferenceeffects.Natu-rally,ifthesubjectistoreportthefirstthingthatcomestomind,thestrongerresponsewilltypicallyprevailovertheweakerone,causingtheomissionofthelatter.Thiswilllendtheappearanceofinaccessibilityoftheweakerresponsewhenitmaynotbeinaccessibleatall.Second,whenthesubjectisgivenaveryshorttimetomakememoryresponsestoacue,interferencemaybepro-ducedbyblocking.Hereagain,strongerresponseswillleaptomindmostreadilyandpotentiallyuseupallthetimethatthesubjecthastoexpresstheirknowledgeoftheassociatedmemories.Evenifallresponsesareofequalstrength,theadditionofnewresponsesmightin-creasethechancesthatsomenontargetitemwillbere-portedtotheexclusionofatargetinalimitedtimewindow.Inbothofthesecases,interferenceeffectsmayreflectsomecombinationofsuppressionarisingfromtheprioroutputofnontargetitemsandblockingproducedbyinsufficienttimetoexpressavailableknowledge.Infact,muchoftheearlyworkoninterferencetheoryupuntilthelate1950semployedthemodifiedfree-recalltest(i.e.,theMFRtest),whichrequiredthatthesubjectprovideonlyasingleresponseinalimitedtimewindow.Withtheadventofthemodified-modified-freerecalltest(MMFR),subjectswereaskedtorecallallavailablere-sponsesandweregivenalongerperiodtorecallthem(Barnes&Underwood,1959),aprocedurethatwasthoughttoprovideabettertestofthetrueavailabilityofresponsesinmemory.Third,whenthemeasureofin-terferenceisreactiontime,thepresenceofmultiple
440M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
competitorsorasinglestrongcompetitorshouldslowtherecallofatarget;again,thisretrievalinterferenceisthoughttobeanessentialstepintriggeringinhibitorycontrol.Finally,specialpopulationswithdeficitsinex-ecutivefunction(e.g.,olderadults,children,frontal-lobedamagedpatients)maybesufficientlychallengedinin-hibitingcompetitorssothatretrievalcompetitioncannotbeeffectivelymanaged.Underthesecircumstances,competitorsmayblocktheretrievalofatargetandcauseimpairmentthatdoesnotreflectinhibitoryprocesses.MeltonandIrwin’sunlearningtheory
Accordingtotheunlearninghypothesis,interferenceeffectsarecausedinpartbytheunlearningofassociativeconnectionslinkingaretrievalcuetoaresponse.Spe-cifically,whenapersonistryingtorecallanewlylearnedresponse(e.g.,thenewphonenumberforafriend),pre-viouslylearnedresponsestothatsamecue(e.g.,theoldphonenumberforthatfriend)maysometimesbeelicitedaccidentally.Elicitationcouldtaketheformofanovertorcovertintrusionoftheunwanteditem.Totheextentthattheolderresponseisincorrect,itwasthoughttogo‘‘unreinforced,’’andthereforesufferextinctioneffectsanalogoustothoseexhibitedbyanimalsinconditioningexperiments.Associativeunlearningwasacriticalcom-ponentofMeltonandIrwinÕsclassicaltwo-factortheoryofinterference(Melton&Irwin,1940),whichalsoin-corporatedresponsecompetition.Themoderndescen-dantsofthisviewincludethemanyconnectionistlearningsystemsthatmightattributeforgettinginparttothealterationofweightsbetweenrepresentationalunits.Thecurrentapproachsharesmuchwiththeun-learninghypothesis:itfocusesontheintrusionofunwantedmemoryresponsesduringretrievalasacon-ditionleadingtotheforgettingoftheintrudingitems;itpositsaprocessthatrespondstointrusionsinsuchawayastorenderthemlesslikelyinthefuture—changingsomeaspectoftheintrusionÕsrepresentation.Thus,aspecialforgettingprocessisproposed.Itdiffers,how-ever,bothinitstheoreticalorientation,andinthenatureoftheforgettingmechanism.Theunlearningideawasatheoreticalanalogyinspiredbythebehavioristlearningapproach.Simple,automaticprocesseswereproposed:learningwasthepositiveadjustmentofassociations,forgetting,thenegativeadjustment.Theexecutivecon-trolapproach,however,isconcernedwiththemoment-by-momentcontrolofbehaviorwithrespecttoflexiblegoals.Itassumesmechanismsbywhichmentalrepre-sentationsareadjusteddynamicallyincontextsinwhichtheirongoingaccessibilitymightdisruptouraims.Themechanismsthatachievethisadjustmentarenotthoughtofasgenerallearningprocesses,butaspro-cessesthatcontroltheoperationalstateofasystem.Thesedifferentorientationsleadtodifferentconcep-tualizationsofhowintrudingmemoriesbecomeim-
paired:whereasunlearningpositsadecrementintheassociativebondlinkingacuetoatarget,theexecutivecontrolapproachattributesimpairmenttoasuppressionofthetargetitself.Thus,thecurrentapproachpredictscue-independentimpairment,whereasunlearningdoesnot.Theexistenceofcue-independentimpairmentofcoursedoesnotruleoutthepossibilitythatassociativeunlearningmightalsooccurandcontributetotheim-pairmentobservedinbothretrieval-inducedforgettingandclassicalinterferenceparadigms.Osgood’sreciprocalinhibitionhypothesis
AnoftenoverlookedtheoryisOsgoodÕsreciprocalinhibitionapproachtointerference.Accordingtothistheory,strengtheningtheassociationbetweenastimulusandaresponsealsostrengthensaninhibitoryassocia-tionbetweenthestimulusandsemanticallyantagonisticresponsesthatareattachedtoit(Osgood,1946,1948).Forinstance,ifsubjectslearnthepairTree-Elated,apositiveassociationisformedbetweenthetwowords,butaninhibitoryoneisalsoestablishedbetweenTreeandtheantagonisticresponseDejected.Inessence,subjectsnotonlylearntomakethecorrectresponse,butalsotoNOTmaketheoppositeresponse—anotionborrowedfromHullÕsbehavioraltheory(Hull,1943).Boththeexcitatoryandinhibitoryassociationswerethoughttogeneralizesemantically,sothatintermediateresponsessuchasLow,wouldalsosufferinhibition,byvirtueofitssimilaritytoDejected.Osgoodprovidedsomesupportforthistheory,showinggraduallyin-creasingretroactiveinterferenceacrosssimilar,neutral,andantagonisticresponsestostimuli,asaresultofin-terpolatedassociativelearning.
OsgoodÕstheoryisperhapsthefirsttheoryofretro-activeinterferencethatattributedimpairmenttoaninhibitorymechanism.InOsgoodÕsframework,impair-mentwasthoughttobeadirectresultofinhibitingthepotentiallyintrusiveresponse,andsothetheorycanexplaincueindependentimpairment.Hereagain,thehypothesisbearssomeresemblancetothecurrentexec-utivecontroltheory.However,OsgoodÕsassertionthatinhibitionisadirectfunctionofsemanticantagonismbetweentworesponsesisnotafeatureofthecurrentapproach,noristhereanycommitmenttothedevel-opmentofaninhibitoryassociationbetweenastimulusandanunwantedresponse,asOsgoodproposed.Inthecurrentperspective,ifacueactivatesamemorythatisunwanted—eitherbecauseitinterfereswitharetrievalattempt,orbecauseitisdistractingorunpleasant—in-hibitorycontrolmechanismscanberecruitedtosup-presstheitem.Consistentattemptstosuppressamemorymayormaynotresultintheformationofan‘‘inhibitoryhabit’’foragivenitem,asOsgoodproposes,butthispossibilityisbeyondthescopeofthepresenttheory.
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445441
Postman’sresponse-setsuppressionhypothesis
Neartheendoftheclassicalinterferenceera,Postmanandcolleagues(Postmanetal.,1968)proposedatheoryofinterferencethatdepartedsubstantiallyfromap-proachespreviouslyproposed.Ashighlightedintheprecedingsections,mostclassicalaccountsofinterferencewereembeddedwithinlargerscaletheoriesofassociativelearningthathadtheirconceptualrootsinbehavioristlearningtheory.Forgettingwasassumedtoreflecttheeffectsofcompetitionbetweenalternateresponses,orthedegradationofassociationsbygenerallearningmecha-nisms.However,Postmanproposedmechanismsthatwentwellbeyondthesomewhatlimitedconceptualarsenalofmostlearningframeworks.Accordingtohisresponse-setsuppressionhypothesis,retroactiveinter-ferencewascausedbytheactivesuppressionofresponsemembersfromtheinitiallist.SuppressionwasthoughttooccurduringtheacquisitionofthesecondlistofpairsbywhatPostmanreferredtoasa‘‘selectormechanism.’’Thefunctionoftheselectormechanismwastobothenhancetherepresentationsofresponsesthatwereintendedtobepartofthecurrentresponsesetandtosuppressoutdatedresponsesets.Thesuppressionprocesshelpedtoreduceproactiveinterferencecausedbytheinitiallist,andtoeffectively‘‘shift’’intoa‘‘responseset’’moreappropriatetothecurrenttask.
Theresponse-setsuppressionhypothesiscanbeseenanearlyexampleoftheexecutivecontrolapproach.Liketheexecutivecontroltheory,thishypothesisattributedforgettingtoamechanismthatdirectlysuppressedtheresponserepresentationsofitemsfromthefirstlistofpairs.Thismechanismwasclearlylinkedtoresponseoverride:ithelpedtheorganismto‘‘select’’current,morecontextuallyappropriateresponsesetsinthefaceofinterferencefromprecedingresponsesets.Thus,thishypothesisacknowledgedtheneedtocontrolmemoryinaccordancewithcurrentgoals,andadvocatedaspecialprocesstoachievethatcontrol.Nevertheless,thecurrenthypothesisdiffersfromPostmanÕstheoryinseveralre-spects.First,accordingtotheresponse-setsuppressionview,theselectormechanismwasthoughttoactonentire‘‘responserepertoires’’andnotatthelevelofin-dividualresponses.So,ifasubjectlearnedalistoftenpairs,followedbyasecondlistoftenpairs,allresponsesfromtheinitiallistwouldbesuppressed,irrespectiveofwhetherornotthestimulusmemberforagivenfirst-listitemwasalsousedinthesecondlist.Thesetoffirstlistresponseswassuppressedasawhole,andthesetofsecondlistresponses,facilitated.Thecurrentapproachismoreflexible,permittingforsuppressionofspecificcompetingresponses.Accordingly,itshouldbe(andis)possibletosuppressonlyselectitemsfromalist,basedonhowmuchinterferencetheycauseduringretrievalofsecondlistitems,asisevidentinstudiesofretrieval-in-ducedforgetting.Second,theresponse-setsuppression
viewdrewatightconnectionbetweentheneedtofacil-itateanewresponsesetinordertosuppressaprecedingset.Thecurrentapproachentertainstheideathatsup-pressioncanbedirectlyappliedtoanunwantedmemorywithouttheneedtofacilitateacompetingresponseorresponseset.Workwiththethink/no-thinkparadigm,forexample,suggeststhatsuppressionisdirectlyappliedtodistractingmemories.However,furtherworkneedstobedonetodeterminewhethersuchdirectsuppressionistrulypossible.Finally,PostmanÕstheorymadeavarietyofspecificassumptionsintendedtoexplaintheconditionsunderwhichspontaneousrecoveryfromretroactiveinterferenceshouldoccur.Althoughtheseassumptionsmaybecorrect,theyarenotanintrinsicpartofthecurrenttheory,asitispresentlyspecified.Despitethesedifferences,thepresenttheorymightberegardedasamoderncousintoresponse-setsuppressionthatdecouplesitfromtheparticularparadigmwithinwhichthetheorywasdeveloped.Theresponse-setsup-pressionviewhasbeenoverlookedasanapproachtoin-terferenceinpartbecausethetheorywasdevelopedtowardstheendoftheclassicalinterferenceera,whenthefieldbecamecaptivatedbycognitivetheory.Theshiftawayfrominterferenceresearchledtotheabandonmentofthetheory,andofresearchoninterferencegenerally.Ironically,totheextentthatinterferencewasdiscussedafterthecognitiverevolution,theoriesbecamefarmoreassociationisticthanPostmanÕs—moreintheveinofMcGeochÕsresponsecompetitiontheory(e.g.,Anderson,1983;Mensink&Raajimakers,1988;Rundus,1973).Thedevelopinginterestinexecutivecontrolfunctionsinthelast15years,andworkoninhibitoryprocesseshasmadeitpossibletoviewPostmanÕstheoryinadifferentlight.
Concludingremarks
Researchoninterferencehasoccupiedacentralroleinthescienceofmemorysincethebeginningsofex-perimentalpsychology.Throughoutmuchofthislonghistory,theoreticaldiscussionsofinterferencehavebeendominatedbyideasthatwereeitherdirectlyborrowedfrom,orinspiredbyclassicalassociativelearningtheo-ries.Inmanyways,thisconceptualinfluencepervadesthinkingaboutinterferenceeventodaynotonlyinhowthisphenomenonisdescribedinmoderntextbooks,butalsoinhowitisexplainedwithincurrenttheoreticalframeworks.Inmoderntextbooks,retroactiveinterfer-enceisoftendefined,forexample,astheforgettingthatarisesasaresultofnewlearning,andproactiveinter-ference,astheforgettingthatarisesasaresultofpre-viouslearning.Ifatheoryisdescribedatall,itisoftentheclassicaltwo-factortheoryofMeltonandIrwin(1940).Despitemanydifferencesinterminologyandconstructsparticulartocognitivepsychology,currenttheoreticalaccountsofinterferencehaveessentially
442M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
returnedtoMcGeochÕsassociativeinterferencetheory.Thereareexcellentreasonsforthecontinuinginfluenceoftheseclassicalideasaboutlearning:interferenceef-fectsarehighlycorrelatedwiththestorageofnewtracesintomemory,andwiththemodificationofexistingones.Theactoflearninganewlistdoesimpairmemoryforapreviousone,andstrengtheningacompetingassociationisoftenassociatedwithimpairedrecallofrelatedtraces.Theseempiricalrelationshipslendforcetotheideathatforgettingultimatelyderivesfromtheeverchangingcontentsofmemory,andourinabilitytocopewiththecompetitioncreatedbythosechanges.
Inthisarticle,Ihavearguedthatdespitetheseem-piricalrelationships,weshouldrethinkourviewofhowinterferenceleadstoforgetting.Ihavearguedthatatheoryofinterferenceshouldbeframedinthelargercontextofhoworganismscontroltheirownthoughtsandactions.Memoryretrievalisjustaspecialcaseofabroadclassofsituationsthatrecruitexecutivecontrolprocesses,anditistheseprocesses—particularlyinhibi-tion—thatcauseforgetting.Bythisview,theempiricalrelationshipbetweenassociativelearningandforgetting
thatemergedwithM€u
llerandPilzecker(1900)andthatdrivestheorizingtoday,shouldnotbeconstruedasproofthatnewlearningimpairsmemory,aspositedinmanyclassicalandmodernmodels.Newlearningsetsthestageforthemechanismthatactuallycausesfor-getting:inhibition.Inhibitionistriggeredasadirectresponsetothecompetitioncausedbyrelatedtracesandthegoaltoselectivelyretrieveatarget,or—inthecaseofmotivatedforgetting—inresponsetothegoaltopreventawarenessofadistractingmemory.Theforgettingthatresultsisnotapassivesideeffectofthenewlearning,butaconsequenceofthemechanismsthathaveevolvedtoalloworganismstooverrideprepotentresponses.Thesemechanismsareessentialtoourabilitytobehaveinaflexible,context-appropriatemanner.
Ihavearguedthatthisviewvalidatesmanyoftheinsightsofferedbyclassicaltheories,whilequestioningthewidespreadassumptionthatforgettingistiedinadirectwaytonewassociativelearning.Rather,ourex-periencesofforgetting—ofpastexperiences,ofourfriendsÕnames,orofideaswithwhichwewereonceadept,areseenascostsoftheverymechanismsthatenableustodirectcognitiontointernalthoughtsandtotheexternalenvironment.
Acknowledgments
IwouldliketothankDouglasHintzman,LarryJacoby,BenLevy,BriceKuhl,SarahJohnson,andtwoanonymousreviewersforcommentsonanearlierversionofthemanuscript.Portionsofthisarticleevolvedfromaplannedpapertobeentitled‘‘Howtonotfindinhibitioninmemorywithoutreallytrying’’by
thepresentauthorandB.A.Spellman.PreparationofthispaperwassupportedbyagrantfromtheNationalInstituteofMentalHealth.References
Allen,G.A.,Mahler,W.A.,&Estes,W.K.(1969).Effectsof
recalltestsonlong-termretentionofpairedassociates.JournalofVerbalLearningandVerbalBehavior,8,463–470.Anderson,J.R.(1983).Thearchitectureofcognition.Cam-bridge,MA:HarvardUniversityPress.
Anderson,J.R.,&Reder,L.M.(1999).Thefaneffect:New
resultsandnewtheories.JournalofExperimentalPsychol-ogy:General,128,186–197.
Anderson,M.C.(1989).Similarity-basedretrievalinhibitionand
mnemonictechniques.UnpublishedMasterÕsThesis,Univer-sityofCalifornia,LosAngeles,LosAngeles.
Anderson,M.C.(2001).Activeforgetting:Evidencefor
functionalinhibitionasasourceofmemoryfailure.JournalofAggression,Maltreatment,andTrauma,4,185–210.
Anderson,M.C.,&Bell,T.(2001).Forgettingourfacts:The
roleofinhibitoryprocessesinthelossofpropositionalknowledge.JournalofExperimentalPsychology,General,130,544–570.
Anderson,M.C.,Bjork,E.L.,&Bjork,R.A.(1994).
Rememberingcancauseforgetting:Retrievaldynamicsinlong-termmemory.JournalofExperimentalPsychology:Learning,Memory,andCognition,20,1063–1087.
Anderson,M.C.,Bjork,E.L.,&Bjork,R.A.(2000).
Retrieval-inducedforgetting:Evidenceforarecall-specificmechanism.PsychonomicBulletin&Review,7,522–530.Anderson,M.C.,&Bjork,R.A.(1994).Mechanismsof
inhibitioninlong-termmemory:Anewtaxonomy.InD.Dagenbach,&T.Carr(Eds.),Inhibitoryprocessesinattention,memoryandlanguage(pp.265–326).SanDiego:AcademicPress.
Anderson,M.C.,DeKok,D.,&Child,C.(1997).Retrieval-inducedforgettingonatestofrecognitionmemory[Abstract].AbstractsofthePsychonomicSociety,2,2.
Anderson,M.C.,&Green,C.(2001).Suppressingunwanted
memoriesbyexecutivecontrol.Nature,410,131–134.
Anderson,M.C.,Green,C.,&McCulloch,K.C.(2000).
Similarityandinhibitioninlong-termmemory:Evidenceforatwo-factormodel.JournalofExperimentalPsychology:Learning,Memory,andCognition,26,1141–1159.
Anderson,M.C.,&McCulloch,K.C.(1999).Integrationasa
generalboundaryconditiononretrieval-inducedforgetting.JournalofExperimentalPsychology:Learning,Memory,andCognition,25,608–629.
Anderson,M.C.,&Neely,J.H.(1996).Interferenceand
inhibitioninmemoryretrieval.InE.L.Bjork,&R.A.Bjork(Eds.),Memory.Handbookofperceptionandcognition(2nded.,pp.237–313).SanDiego,CA:AcademicPress.Anderson,M.C.,Ochsner,K.,Kuhl,B.,Cooper,J.,Robert-son,E.,Gabrieli,S.W.,Glover,G.,&Gabrieli,J.D.E.(submitted).Neuralsystemsunderlyingthesuppressionofunwantedmemories.
Anderson,M.C.,&Shivde,G.S.(inpreparationa).Inhibition
inepisodicmemory:Evidenceforaretrieval-specificmech-anism.
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
443
Anderson,M.C.,&Shivde,G.S.(inpreparationb).Strengthis
notenough:Evidenceagainstablockingtheoryofretrieval-inducedforgetting.
Anderson,M.C.,&Spellman,B.A.(1995).Onthestatusof
inhibitorymechanismsincognition:Memoryretrievalasamodelcase.PsychologicalReview,102,68–100.
Arbuckle,T.Y.(1966).Intratrialinterferencein‘‘immediate’’
memory.DissertationAbstracts,26,5554.
Balota,D.A.(1994).Visualwordrecognition:Thejourney
fromfeaturestomeaning.InM.A.Gernsbacher(Ed.),Handbookofpsycholinguistics(pp.303–358).SanDiego,CA,US:AcademicPress.
Barnes,J.M.,&Underwood,B.J.(1959).‘‘Fate’’offirst-list
associationsintransfertheory.JournalofExperimentalPsychology,58,97–105.B€auml,K.(1996).Revisitinganoldissue:Retroactiveinterfer-enceasafunctionofthedegreeoforiginalandinterpolatedlearning.PsychonomicBulletin&Review,3,380–384.B€auml,K.(1997).Thelist-strengtheffect:Strength-dependent
competitionorsuppression.PsychonomicBulletin&Review,4,260–264.B€auml,K.(1998).Strongitemsgetsuppressed,weakitemsdo
not:Theroleofitemstrengthinoutputinterference.PsychonomicBulletin&Review,5,459–463.B€auml,K.(2002).Semanticrecallcancauseepisodicforgetting.
PsychologicalScience,13,356–360.B€auml,K.,&Hartinger,A.(2002).Ontheroleofitemsimilarity
inretrieval-inducedforgetting.Memory,10,215–224.
Barnhardt,T.M.,Glisky,E.L.,Polster,M.R.,&Elam,L.
(1996).Inhibitionofassociatesandactivationofsyn-onymsintherare-wordparadigm:Furtherevidenceforacenter-surroundmechanism.Memory&Cognition,24,60–69.
Bjork,E.L.,Bjork,R.A.,&Anderson,M.C.(1998).Varieties
ofgoal-directedforgetting.InJ.M.Golding,&C.M.MacLeod(Eds.),Intentionalforgetting:Interdisciplinaryapproaches(pp.103–137).Mahwah,NJ:Erlbaum.
Bjork,R.A.(1975).Retrievalasamemorymodifier.InR.
Solso(Ed.),Informationprocessingandcognition:TheLoyolasymposium(pp.123–144).Hillsdale,NJ:Erlbaum.Bjork,R.A.(1989).Retrievalinhibitionasanadaptive
mechanisminhumanmemory.InH.L.RoedigerIII,&F.I.M.Craik(Eds.),Varietiesofmemoryandconscious-ness:EssaysinhonourofEndelTulving(pp.309–330).Hillsdale,NJ:Erlbaum.
Blaxton,T.A.,&Neely,J.H.(1983).Inhibitionfrom
semanticallyrelatedprimes:Evidenceofacategory-specificretrievalinhibition.Memory&Cognition,11,500–510.Butler,K.M.,Williams,C.C.,Zacks,R.T.,&Maki,R.H.
(2001).Alimitonretrieval-inducedforgetting.Memory&Cogntion,27,1314–1319.
Carrier,M.,&Pashler,H.(1992).Theinfluenceofretrievalon
retention.Memory&Cognition,20,633–642.
Casey,B.J.,Trainor,R.J.,Orendi,J.L.,Schubert,A.B.,
Nystrom,L.E.,Giedd,J.N.,Castellanos,F.X.,Haxby,J.V.,Noll,D.C.,Cohen,J.D.,Forman,S.D.,Dahl,R.E.,&Rapoport,J.L.(1997).AdevelopmentalfunctionalMRIstudyofprefrontalactivationduringperformanceofaGo-No-Gotask.JournalofCognitiveNeuroscience,9,835–847.Ciranni,M.A.,&Shimamura,A.P.(1999).Retrieval-induced
forgettinginepisodicmemory.JournalofExperimental
Psychology:Learning,Memory,andCognition,25,1403–1414.
Conway,A.R.,&Engle,R.W.(1994).Workingmemoryandretrieval:aresource-dependentinhibitionmodel.JournalofExperimentalPsychology:General,123,354–373.
Conway,M.A.,Harries,K.,Noyes,J.,Racsmany,M.,&Frankish,C.R.(2000).Thedisruptionanddissolutionofdirectedforgetting:Inhibitorycontrolofmemory.JournalofMemoryandLanguage,43,409–430.
Dagenbach,D.,Carr,T.H.,&Barnhardt,T.M.(1990).Inhibitorysemanticprimingoflexicaldecisionsduetofailuretoretrieveweaklyactivatedcodes.JournalofExperimentalPsychology:Learning,Memory,andCogni-tion,16,328–340.
Deprince,A.P.,&Freyd,J.J.(2001).Memoryanddissociativetendencies:Therolesofattentionalcontextandwordmeaninginadirectedforgettingtask.JournalofTraumaandDissociation,2(2),67–82.
deZubicaray,G.I.,Andrew,C.,Zelaya,F.O.,Williams,S.C.R.,&Dumanoir,C.(2000).Motorresponsesuppressionandtheprepotenttendencytorespond:AparametricfMRIstudy.Neuropsychologia,38,1280–1291.
Dodson,C.S.,Johnson,M.K.,&Schooler,J.W.(1997).Theverbalovershadowingeffect:Whydescriptionsimpairfacerecognition.Memory&Cognition,25,129–139.
Dong,T.(1972).Cuedpartialrecallofcategorizedwords.JournalofVerbalLearningandVerbalBehavior,11,654–661.
Dopkins,S.,&Ngo,C.T.(2002).Inhibitionofverbalmemoryretrievalasaconsequenceofpriorretrieval.JournalofMemoryandLanguage.,46,606–621.
Dunn,E.W.,&Spellman,B.A.(2003).Forgettingbyremembering:Stereotypeinhibitionthroughrehearsalofalternativeaspectsofidentity.JournalofExperimentalSocialPsychology,39,420–433.
Finger,K.,&Pezdek,K.(1999).Theeffectofcognitiveinterviewonfaceidentificationaccuracy:Releasefromverbalovershadowing.JournalofAppliedPsychology,84,340–348.
Fisher,R.P.,&Craik,F.I.M.(1977).Interactionbetweenencodingandretrievaloperationsincuedrecall.JournalofExperimentalPsychology:HumanLearningandMemory,3,701–711.
Freud,S.(1966).InJ.Strachey(Ed.),ThestandardeditionofthecompletepsychologicalworksofSigmundFreud(pp.117–128).Hogarth.
Garavan,H.,Ross,T.J.,&Stein,E.A.(1999).Righthemisphericdominanceofinhibitorycontrol:Anevent-relatedfunctionalMRIstudy.ProceedingsoftheNationalAcademyofSciencesUSA,96,8301–8306.
Gardiner,J.M.,Craik,F.I.,&Bleasdale,F.A.(1973).Retrievaldifficultyandsubsequentrecall.Memory&Cognition,1,213–216.
Geiselman,R.E.,Bjork,R.A.,&Fishman,D.L.(1983).Disruptedretrievalindirectedforgetting:Alink.JournalofExperimentalPsychology:General,112,58–72.
Gernsbacher,M.A.,&Faust,M.E.(1991).Themechanismofsuppression:Acomponentofgeneralcomprehensionskill.JournalofExperimentalPsychology:Learning,Memory,andCognition,17,245–262.
Golding,J.M.,&MacCleod,C.M.(Eds.).(1998).Intentionalfor-getting:Interdisciplinaryapproaches.Mahwah,NJ:Erlbaum.
444M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
Hayes-Roth,B.(1977).Evolutionofcognitivestructuresand
processes.PsychologicalReview,84,260–278.
Hicks,J.L.,&Starns,J.(inpress).Retrieval-inducedforgetting
occursintestsofitemrecognition.PsychonomicBulletin&Review.
Hicks,J.L.,&Marsh,R.L.(1998).Adecrement-to-familiarity
interpretationoftherevelationeffectfromforced-choicetestsofrecognitionmemory.JournalofExperimentalPsychology:Learning,Memory,andCognition,24,1105–1120.
Horton,D.L.,&Kjeldergaard,P.M.(1961).Anexperimental
analysisofassociativefactorsinmediatedgeneralizations.PsychologicalMonographs,75,26.
Hull,C.L.(1943).Principlesofbehavior:Anintroductionto
behaviortheory.NewYork:Appleton-Century.
Hunt,R.R.,&Smith,R.E.(1998,November).Retrievalmay
ormaynotcauseforgetting.PaperpresentedattheannualmeetingofthePsychonomicSociety,Dallas,TX.
Jacoby,L.L.,&Dallas,M.(1981).Ontherelationship
betweenautobiographicalmemoryandperceptuallearn-ing.JournalofExperimentalPsychology:General,110,306–340.
Jenkins,J.J.(1963).Mediatedassociations:Paradigmsand
situations.InC.N.Cofer,&B.S.Musgrave(Eds.),Verbalbehaviorandlearning:Problemsandprocesses(pp.210–257).NewYork:McGraw-Hill.
Johnson,S.,&Anderson,M.C.(inpress).Theroleof
inhibitorycontrolinforgettingsemanticknowledge.Psy-chologicalScience.
Kimball,D.R.,&Bjork,R.A.(2002).Influencesofintentional
andunintentionalforgettingonfalsememories.JournalofExperimentalPsychology:General,131,116–130.
Kjeldergaard,P.M.(1968).Transferandmediationinverbal
learning.InT.R.Dixon,&D.L.Horton(Eds.),Verbalbehaviorandgeneralbehaviortheory(pp.1071–1075).EnglewoodCliffs,NJ:PrenticeHall.
Koutstaal,W.,Schacter,D.L.,Johnson,M.K.,&Galluccio,
L.(1999).Facilitationandimpairmentofeventmemoryproducedbyphotographreview.Memory&Cognition,27,478–493.
Levy,B.,Reinholz,J.,&Anderson,M.C.(inpreparation).
DirectcuingofinhibitoryprocessesintheThink/no-Thinkprocedure.
Levy,B.J.,&Anderson,M.C.(2002).Inhibitoryprocessesand
thecontrolofmemoryretrieval.TrendsinCognitiveScience,6,299–305.
Lustig,C.,Hasher,L.,&Toney,S.T.(2001).Inhibitorycontrol
overthepresentandthepast.EuropeanJournalofCognitivePsychology,13,107–122.
MacLeod,M.(2002).Retrieval-inducedforgettingineyewit-nessmemory:Forgettingasaconsequenceofremembering.AppliedCognitivePsychology,16,135–149.
MacLeod,M.D.,&Macrae,C.N.(2001).Gonebutnot
forgotten:Thetransientnatureofretrieval-inducedforget-ting.PsychologicalScience,12,148–152.
Macrae,C.N.,&MacLeod,M.D.(1999).Onrecollections
lost:Whenpracticemakesimperfect.JournalofPersonalityandSocialPsychology,77,463–473.
Macrae,C.N.,&Roseveare,T.A.(2002).Iwasalwaysonmy
mind:Theselfandtemporaryforgetting.PsychonomicBulletin&Review,9,611–614.
McDaniel,M.A.,Friedman,A.,&Bourne,L.E.(1978).
Rememberingthelevelsofinformationinwords.Memory&Cognition,6,156–164.
McGeoch,J.A.(1942).Thepsychologyofhumanlearning:An
introduction.NewYork:Longmans.
McGeoch,J.A.,&Irion,A.L.(1952).Thepsychologyofhuman
learning(2nded.).NewYork:Longmans,Green,&Co.Meissner,C.A.,&Brigham,J.C.(2001).Theinfluenceof
retrievalprocessesinverbalovershadowing.Memory&Cognition,29,176–186.
Melcher,J.M.,&Schooler,J.W.(1996).Themisremembrance
ofwinespast:Verbalandperceptualexpertisedifferentiallymediateverbalovershadowingoftastememory.JournalofMemoryandLanguage,35,231–245.
Melton,A.W.,&Irwin,J.M.(1940).Theinfluenceofdegreeof
interpolatedlearningonretroactiveinhibitionandtheoverttransferofspecificresponses.AmericanJournalofPsychol-ogy,3,173–203.
Mensink,G.J.M.,&Raajimakers,J.G.W.(1988).Amodelof
interferenceandforgetting.PsychologicalReview,95,434–455.
Miyamoto,A.,&Anderson,M.C.(inpreparation).Suppress-ingawareness:Evidenceforalinkbetweentheregulationofconsciousnessandinhibitoryphenomenainmemory.PsychologicalScience.
Moulin,C.J.A.,Perfect,T.J.,Conway,M.A.,North,A.S.,
Jones,R.W.,&James,N.(2002).Retrieval-inducedforgettinginAlzheimerÕsdisease.Neuropsychologia,40,862–867.
Morris,C.D.,Bransford,J.P.,&Franks,J.J.(1977).Levelsof
processingversustransferappropriateprocessing.JournalofVerbalLearningandVerbalBehavior,16,519–533.
M€u
ller,G.E.,&Pilzecker,A.(1900).ExperimentalleBeitragezurLehrecomGedachtnis.ZeitschriftfurPsychologie,1,1–300.
Myers,L.B.,Brewin,C.R.,&Power,M.J.(1998).Repressive
copingandthedirectedforgettingofemotionalmaterial.JournalofAbnormalPsychoogy,107,141–148.
Nader,A.,Coles,M.E.,Brigidi,B.,&Foa,E.B.(2001).The
effectofpracticeonrecallofemotionalinformationinindividualswithgeneralizedsocialphobia.JournalofAbnormalPsychology,110,76–82.
Nickerson,R.S.(1984).Retrievalinhibitionfrompart-set
cuing:Apersistingenigmainmemoryresearch.Memory&Cognition,12,531–552.
Norman,K.A.,&OÕReilly,R.C.(inpress).Modeling
hippocampalandneocorticalcontributionstorecognitionmemory:Acomplementarylearningsystemsapproach.PsychologicalReview.
Osgood,C.E.(1946).Meaningfulsimilarityandinterferencein
learning.JournalofExperimentalPsychology,36,277–301.Osgood,C.E.(1948).Aninvestigationintothecausesof
retroactiveinterference.JournalofExperimentalPsychol-ogy,38,132–154.
Osgood,C.E.(1949).Thesimilarityparadoxinhuman
learning:Aresolution.PsychologicalReview,56,132–143.Perfect,T.J.,Moulin,C.J.A.,Conway,M.A.,&Perry,E.
(2002).Assessingtheinhibitoryaccountofretrieval-inducedforgettingwithimplicit-memorytests.JournalofExperi-mentalPsychology:Learning,Memory,andCognition,28,1111–1119.
M.C.Anderson/JournalofMemoryandLanguage49(2003)415–445
445
Postman,L.(1971).Transfer,interference,andforgetting.InJ.
W.Kling,&L.A.Riggs(Eds.),WoodworthandSchlos-bergÕs:Experimentalpsychology(3rded.,pp.1019–1132).NewYork:Holt,Rinehart&Winston.
Postman,L.,Stark,K.,&Fraser,J.(1968).Temporalchanges
ininterference.JournalofVerbalLearningandVerbalBehavior,7,672–694.
Raaijmakers,J.W.,&Shiffrin,R.M.(1981).Searchof
associativememory.PsychologicalReview,88,93–134.
Rajaram,S.,Srinivas,K.,&Travers,S.(2001).Theeffectsof
attentiononperceptualimplicitmemory.Memory&Cog-nition,29,920–930.
Radvansky,G.A.(1999).Memoryretrievalandsuppression:
Theinhibitionofsituationmodels.JournalofExperimentalPsychology:General,128,563–579.
Radvansky,G.A.,&Zacks,R.T.(1991).Mentalmodelsand
thefaneffect.JournalofExperimentalPsychology:Learning,Memory,andCognition,17,940–953.
RoedigerIII,H.L.(1973).Inhibitioninrecallfromcueingwith
recalltargets.JournalofVerbalLearningandVerbalBehavior,12,644–657.
RoedigerIII,H.L.,&McDermott,K.B.(1993).Implicit
memoryinnormalhumansubjects.InF.Boller,&J.Grafman(Eds.),HandbookofNeuropsychology(Vol.8,pp.63–131).
RoedigerIII,H.L.,&Neely,J.H.(1982).Retrievalblocksin
episodicandsemanticmemory.CanadianJournalofPsy-chology,36,213–242.
Roediger,H.L.,&Schmidt,S.R.(1980).Outputinterference
intherecallofcategorizedandpairedassociatelists.JournalofExperimentalPsychology:HumanLearningandMemory,6,91–105.
Roediger,H.L.,Weldon,M.S.,Stadler,M.A.,&Rieger,G.
H.(1992).Directcomparisonofwordstemsandwordfragmentsinimplicitandexplicitretentiontests.JournalofExperimentalPsychology:Learning,Memory,andCogni-tion,18,1251–1269.
Roskies,A.L.,Fiez,J.A.,Balota,D.A.,Raichle,M.E.,&
Petersen,S.E.(2001).Task-dependentmodulationofregionsintheleftinferiorfrontalcortexduringsemanticprocessing.JournalofCognitiveNeuroscience,13,829–843.Rundus,D.(1973).Negativeeffectsofusinglistitemsas
retrievalcues.JournalofVerbalLearningandVerbalBehavior,12,43–50.
Sakagami,M.,&Niki,H.(1994).Spatialselectivityofgo/no-go
neuronsinthemonkeyprefrontalcortex.ExperimentalBrainResearch,100,165–169.
Samuel,A.(1996).Doeslexicalinformationinfluencethe
perceptualrestorationofphonemes.JournalofExperimen-talPsychology:General,125,28–51.
Saunders,J.,&MacLeod,M.D.(2002).Newevidenceonthe
suggestibilityofmemory:Theroleofretrieval-inducedforgettinginmisinformationeffects.JournalofExperimentalPsychology:Applied,8,127–142.
Schooler,J.W.,&Engstler-Schooler,T.Y.(1990).Verbal
overshadowingofvisualmemories:Somethingsarebetterleftunsaid.CognitivePsychology,22,36–71.
Schooler,J.W.,Fiore,S.M.,&Brandimonte,M.A.(1997).At
alossfromwords:Verbalovershadowingofperceptualmemories.InD.Medin(Ed.),Thepsychologyoflearning
andmotivation(Vol.37,pp.293–334).SanDiego:AcademicPress.
Shaw,J.S.,Bjork,R.A.,&Handal,A.(1995).Retrieval-inducedforgettinginaneyewitness-memoryparadigm.PsychonomicBulletin&Review,2,249–253.
Shivde,G.,&Anderson,M.C.(2001).Theroleofinhibitioninmeaningselection:Insightsfromretrieval-inducedforget-ting.InD.S.Gorfein(Ed.),Ontheconsequencesofmeaningselection:Perspectivesonresolvinglexicalambiguity(pp.175–190).Washington,DC:AmericanPsychologicalAsso-ciation.
Shimamura,A.P.(1995).Memoryandfrontallobefunction.InM.S.Gazaniga(Ed.),Thecognitiveneurosciences(pp.803–813).MITPress.
Simpson,G.B.,&Kang,H.(1994).Inhibitoryprocessesintherecognitionofhomographmeanings.InD.Dagen-bach,&T.Carr(Eds.),Inhibitoryprocessesinattention,memory,andlanguage(pp.359–381).SanDiego,CA:AcademicPress.
Smith,A.D.(1971).Outputinterferenceandorganizedrecallfromlong-termmemory.JournalofVerbalLearningandVerbalBehavior,10,400–408.
Smith,A.D.(1973).Inputorderandoutputinterferenceinorganizedrecall.JournalofExperimentalPsychology,100,147–150.
Smith,A.D.,DÕAgostino,P.R.,&Reid,L.S.(1970).Outputinterferenceinlong-termmemory.CanadianJournalofPsychology,24,85–89.
Smith,E.E.,Adams,N.E.,&Schorr,D.(1978).Factretrievalandtheparadoxofinterference.CognitivePsychology,10,438–464.
Smith,R.E.,&Hunt,R.R.(2000).Theinfluenceofdistinctiveprocessingonretrieval-inducedforgetting.Memory&Cog-nition,28,503–508.
Smith,S.M.,&Tindell,D.R.(1997).Memoryblocksinwordfragmentcompletioncausedbyinvoluntaryretrievaloforthographicallyrelatedprimes.JournalofExperimen-talPsychology:Learning,Memory,andCognition,23,355–370.
Thompson-Schill,S.(1997).Roleoftheleftinferiorprefrontalcortexintheretrievalofsemanticknowledge:Areevalua-tion.ProceedingsoftheNationalAcademyofSciencesUSA,94,14792–14797.
Tsukimoto,T.,&Kawaguchi,J.(2001).Retrieval-inducedforgetting:Isthebaselineintheretrieval-practiceparadigmtrue?PostersessionpresentedattheInternationalConfer-enceonMemory,Valencia,Spain.
Tulving,E.,&Arbuckle,T.Y.(1963).Sourcesofintratrialinterferencesinpaired-associatelearning.JournalofVerbalLearningandVerbalBehavior,1,321–334.
Tulving,E.,&Arbuckle,T.Y.(1966).Inputandoutputinterferenceinshort-termassociativememory.JournalofExperimentalPsychology,72,145–150.
Wheeler,M.A.(1995).Improvementinrecallovertimewithoutrepeatedtesting:Spontaneousrecoveryrevisited.JournalofExperimentalPsyvchology:Learning,Memory,andCognition,21,173–184.
Williams,C.C.,&Zacks,R.T.(2001).Isretrieval-inducedforgettinganinhibitoryprocess.AmericanJournalofPsychology,114,329–354.
因篇幅问题不能全部显示，请点此查看更多更全内容