ebook img

Embodied abstract semantics The representation of abstract words PDF

66 Pages·2010·0.44 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Embodied abstract semantics The representation of abstract words

Embodiedabstract semantics 1 RUNNINGHEAD: Embodiedabstract semantics Therepresentationof abstract words: Whyemotionmatters Stavroula-ThaleiaKousta* GabriellaVigliocco* DavidP.Vinson* MarkAndrews* ElenaDel Campo^ *UniversityCollege London ^Universita’degli Studi di Padova Embodiedabstract semantics 2 Abstract Althoughmuchis known about therepresentation andprocessingof concreteconcepts, our knowledgeofwhat abstract semantics might beis severelylimited. Inthis paperwefirst address the adequacyofthetwodominant accounts (dual codingtheoryandthe context availabilitymodel)put forwardinorderto explainrepresentationandprocessingdifferences betweenconcreteandabstract words. Wefindthat neitherproposal canaccount forexperimental findings andthat this is, at least partly,becauseabstract words areconsideredto beunrelatedtoexperiential informationinbothof theseaccounts. Wethenaddress oneparticulartypeofexperiential information,emotional content, anddemonstratethat it plays acrucial roleintheprocessingandrepresentationofabstract concepts: statistically,abstract words aremoreemotionallyvalencedthanconcrete words andthis accounts for aresidual latencyadvantagefor abstract words, whenvariables suchas imageability(aconstruct derived from dual codingtheory)and rated context availabilityareheld constant.Weconcludewith adiscussionofournovel hypothesis for embodied abstract semantics. Therepresentationof abstract wordmeanings: Whyemotionmatters Concreteentities exist inspace-time andareindependent ofhumanminds/language; abstract entities, ontheotherhand,donot exist inspace-timebut theirexistencedepends onhumanminds/language (Hale,1988). “Concreteness”,therefore,indexes abasicontological distinction,dividingentities into thesetwokinds. This ontological distinctionis reflectedinourepistemologies, andconcreteness is arguablyanorganizingprincipleofsemanticknowledge.Uptothepresent,researchintosemantic andconceptual representation has focused almost exclusivelyonhowconcretewordmeanings and Embodiedabstract semantics 3 concepts arerepresented andprocessed,totheexclusionofabstract meanings and concepts. However,theabilityto communicatethroughlanguageabout abstract concepts, suchas courage, dignity,revenge, lies at theheart ofwhat it means tobehuman,andnotheoryofsemanticor conceptual representationis completewithout anexplicit account ofhowabstract knowledgeis acquired, represented, andprocessed. Inthis paperwefirst demonstrate,combiningexperiments withlargescaleregressionanalyses of datafrom theEnglish LexiconProject (ELP, Balotaet al.,2007),that thedual codingtheoryandthe context availabilityhypothesis -twoofthemost popularaccounts ofdifferences inrepresentation andprocessingbetween concreteand abstract words -donot exhaustivelyaccount forprocessing (and hencerepresentational)differences between thetwotypes ofwordmeanings. In fact,once imageabilityand context availability(alongwitha largenumberofotherlexical andsublexical variables) are controlled, thereis aresidual advantagefor abstract wordprocessing.Weshowthat this advantagecanbe explainedbydifferences inemotional valencebetweenconcrete andabstract words, andwediscuss a newhypothesis ofhowthesemanticsystem is organizedwithrespect tothe distinctionbetweenconcreteand abstract concepts. Specifically,weproposethat bothconcreteand abstract concepts binddifferent types ofinformation: experiential information(sensory,motorand affective), andalso linguisticinformation.However,concreteand abstract semanticrepresentations differinterms of whethersensory,motororaffectiveinformationhavethe greatest weight,with sensory-motorinformationbeingmorepreponderant forconcreteconcepts andaffectiveinformation playinga greaterrolefor abstract concepts. Thus, acentral andnovel element ofthis proposal is the ideathat experiential informationcontributes totherepresentationofboth concreteand abstract words, however,whereas sensory-motorinformationis statisticallymoreimportant forthe Embodiedabstract semantics 4 representationofconcretewords, emotional content,alargelyneglectedtypeofexperiential informationintheliteratureonsemantic representation/processing, contributes toword representationandprocessing,particularlyfor abstract concepts. TheConcreteness Effect:Dual-CodingTheoryand theContextAvailabilityModel It has been demonstrated repeatedly,and withavarietyofmethodologies,that concretewords have a cognitive advantageover abstract words—anadvantage,labelledthe ‘concreteness effect’.With respect tolexical processing,earlydemonstrations ofaprocessingadvantageforconcreteover abstract words wereprovidedbyJames (1975),Whaley(1978), andRubin(1980).James showed that at least whenlowfrequencywords are considered, concretewords are identifiedas words faster thanabstract words. Whaley(1978)andRubin(1980)adopteda correlational approach,showing that thereis asignificant negativecorrelationbetweenconcreteness ratings andlexical decision reactiontimes forthesameitems. This processingadvantagehas sincethenbeenreplicatedinboth lexical decision(Binder et al.,2005; Bleasdale,1987; deGroot,1989; Howell & Bryden,1987; Kroll &Merves, 1986; Schwanenflugel, Harnishfeger &Stowe,1988; Schwanenflugel &Stowe, 1989)andwordnamingtasks (deGroot,1989; Schwanenflugel andStowe, 1989). Withrespect tomemoryforconcrete andabstract words,it has been againrepeatedlydemonstrated that concretewords have anadvantageoverabstract words inbothlong-term andshort-term memory tasks (e.g.paired-associatelearning(Paivio,Yuille,&Smythe,1966); serial recall (Allenand Hulme,2006; Romani,McAlpine,&Martin,2007; Walker&Hulme,1999),freerecall (Romani, Embodiedabstract semantics 5 McAlpine,&Martin,2007; terDoest &Semin,2005),reconstructionoforder(Neath,1997); and recognitionmemory(Fliessbach,Weis, Klaver,Elger&Weber,2006)). Amongthehandful ofproposals that havebeenput forwardto explainthe‘concreteness effect’,two havebeenparticularlyinfluential: dual codingtheory(Paivio,1971; 1986; 1991; 2007)andthe context availabilitymodel (Schwanenflugel andShoben,1983; Schwanenflugel, 1991). Inbothof theseaccounts, concrete wordrepresentations are assumedtobericherthan abstract word representations (see also Plaut &Shallice,1993). Accordingtodual codingtheory,concretewords arerepresentedintworepresentationallydistinct but functionallyrelatedsystems: averbal,linguistic system and anon-verbal, imagisticsystem.Abstract concepts, ontheother hand,areprimarilyor exclusivelyrepresentedintheverbal system.The cognitive advantagefor words referringto concreteconcepts is attributedtothefact that theyhaveaccess toinformationfrom multiplesystems. Accordingtothe context availabilitymodel,bothconcrete andabstract concepts are representedina singleverbal codeandneithertherepresentations northeprocesses that operateonthese representations differfor thetwotypes ofconcepts.Theargument hereis that comprehensionrelies onverbal context (eithersuppliedbythediscourse orbythecomprehender’s ownsemanticmemory) inorderto beeffective. Accessingthemeaningofawordinvolves accessinganetworkof associated semanticinformationand theadvantagefor concretewords arises becausetheyhavestrongerand denserassociations tocontextual knowledgethan abstract words. Thesetwoproposals have guided researchon concrete/abstract semantics; results, however,havebeeninconclusive.Themajorityof recent workis neuroscientificinnature,employingeitherelectrophysiological orneuroimaging techniques inordertodeterminetheneural bases ofthedistinctionbetween concreteand abstract words. Embodiedabstract semantics 6 Aseries ofstudies usingevent-relatedpotentials (ERPs)suggestedcombiningdual-codingtheory andthecontext availabilitymodel inexplainingtheconcreteness effect (“context-extendeddual codingtheory”—Holcombet al.,1999; West &Holcomb,2000).ERP studies haveidentifiedtwo components associated withconcreteness: theN400andalatenegativecomponent peakingaround 700-800milliseconds. Withrespect tothefirst component,all relevant studies havefoundthat concretewords elicit alargerN400than abstract words (Holcombet al.,1999; Kanske &Kotz, 2007; Kounios &Holcomb,1994; Nittonoet al.,2002; West &Holcomb,2000; vanSchieet al., 2005).Theobservationthat theeffect has ananteriormaximum but is widelydistributedacross the scalp(West &Holcomb,2000)andthefailuretofindanystructural overlapbetween concreteness andvisual object workingmemoryonthat component (vanSchie et al.,2005)has ledtothe suggestionthat theeffect arises withinaverbal semanticsystem that is commontobothconcreteand abstract words. This N400component has beenarguedto reflect postlexical processinginasemantic memorysystem,possiblyinvolvingtheintegration ofsemanticinformation intohigherlevel representations (Osterhout &Holcomb,1995).Accordingtothecontext availabilitymodel,concrete words areassumedtohavestronger anddenserinterconnections withother concepts insemantic memorythanabstract words (Schwanenflugel &Shoben,1983). IntheEEGliterature, concrete words aresaidtoactivate thesemanticnetworkmoreextensivelythan abstract words andthis extensiveactivationis reflectedinan amplifiedN400forconcrete words. Thesecond,later, component is assumedtoreflect thecontributionofmental imageryforconcrete words: it is more sustainedovertime,peakingat around700or800 ms post-stimulus. It is saidtobeassociatedwith theretrieval ofmental imageryassociatedwithconcrete words and thus consistent withdual coding theory.Althoughtheimagery-related component is consistent withdual-codingclaims that imagery Embodiedabstract semantics 7 has alate effect inprocessing,the greaterN400amplitudeforconcretewords is hardertoreconcile withcontext availabilityclaims. Thestrongerinterconnections insemanticmemoryforconcrete words accordingtothemodel leadtofacilitatedintegrationofinformation.TheincreasedN400 amplitudeforconcrete words, however,has beeninterpreted as indexingdifficultyin integrating appropriateinformation(seeKutas, VanPetten& Kluender,2006 forareview).Sotheextent to whichEEGdata actuallysupport thecontext-extendeddual codingtheoryis questionable. Acase foraqualitativedifferencebetweenconcreteandabstract wordmeanings, thus compatible withdual-codingviews, comes from neuropsychological studies where adoubledissociation between concreteand abstract words has beenobserved.Althoughcases where concretewords are betterpreservedinthedamaged/agingbrain arethemost frequentlyreported(e.g.,Coltheart et al., 1980; Franklinet al.,1995; Katz &Goodglass,1990; Martin&Saffran,1992; Roeltgenet al.,1983; Warrington,1975),there are cases reportingbetter performanceon abstract overconcrete words (e.g.,Breedinet al.,1994; Cipolotti andWarrington,1995; Marshall et al,1996; Papagno et al., 2007; Reillyet al.,2007; Sirigu,Duhamel, &Poncet,1991; Warrington,1975; Warrington & Shallice,1984). Intheimagingliterature, althoughabstract wordprocessingseems toinvolveactivations inamore distributednetwork ofbrainregions thanconcrete wordprocessing(Pexmanet al.,2007),thereis convergingevidencethat abstract wordprocessingis associatedwithhigher activationinleft hemisphericareas that areknowntobeinvolvedinsemanticprocessing, e.g,theleft inferior frontal gyrus (LIFG)(Perani et al.,1999; Jessenet al.,2000; Fiebach andFriederici,2003; Noppeneyand Price,2004; Binderet al.,2005)andthesuperiortemporo-lateral cortex (Mellet et al.,1998; Kiehl et Embodiedabstract semantics 8 al.,1999; Wiseet al.,2000; Binderet al.,2005; Binderet al,2009).Withrespect to greater activation forabstract overconcrete words intheleft inferior frontal gyrus, this findinghas been interpreted as indicatingmoreeffortful retrieval ofsemanticinformationforabstract words, a findingthat has beeninterpretedinsomestudies as consistent withcontextavailabilitypredictions. Again,however,themajorityofthestudies useitems matched onfrequency,but not onfamiliarity orotherrelevant variables. Forinstance,inoneof theotherwisebest-controlledstudies inthe imagingliterature(Binderet al.,2005),althoughitems werematchedon frequency,we foundthat concretewords weresignificantlymorefamiliarthanabstract words (average familiarityratings of 534vs. 471respectively, t(98)=3.956,p<.001). It maywell bethat suchdifferences infamiliarity underliesomeofthe effects reportedintheneuroimagingliterature. Whenconcretewords are comparedto abstract words, results havebeenextremelyvariable. Althoughsomestudies havefound activations ofleft hemispheric regions associatedwithhigher levels ofvisual processingsuch as theleft fusiform gyrus (D’Espositoet al.,1997; Mellet et al., 1998; Fiebach & Friederici,2003; Sabsevitz et al.2005),consistent withthedual codingprediction that concretewordmeanings activaterelevant imagisticinformation,anumberofstudies havefailed tofindanyregions at all that are activatedmoreduringconcretewordprocessing(Grossmanet al., 2000; Friederici et al.,2000; Kiehl et al.,1999; Krauseet al.,1999; NoppeneyandPrice,2004; Perani et al.,1999; Pexmanet al.,2007; Tyler et al.,2001).Somestudies havefoundmorebilateral activations duringconcretewordprocessing(Binderet al.,2005; Sabsevitz et al.,2005),whileother studies haveshownthat thereis noright hemisphereinvolvement intheprocessingofconcrete words, andifanything,there aremoreright-hemisphericactivations forabstract ratherthanconcrete words (see Fiebachand Friederici,2003 fora review).Oneofthereasons forthelackof consistency Embodiedabstract semantics 9 intheresults maybethat theconcretewords used withinandacross studies differinterms oftheir featural composition,whichquitereasonablyleads toactivationofdifferent brainnetworks in different studies ortolackofconsistent areas ofactivationwithinthesamestudy.Thus, just as the behavioral andEEGevidence reviewedabove,imagingstudies donot provideclearsupport for eitherdual-codingorcontext availabilitycallingfornewtheoretical directions andfurther empirical investigation. TheConcreteness Effect: TestingDual-CodingTheoryandtheContext-AvailabilityHypothesis Intheliterature,it is invariablyassumedthat thepsycholinguisticconstructs ofconcreteness and imageabilitytapinto the sameunderlyingtheoretical construct,i.e.,theontological distinction between concrete,spatiotemporally-boundconcepts andabstract,non-spatiotemporallybound concepts. Afterall,when nothingelseis takenintoaccount,imageabilityratings explainmorethan 72%ofthevarianceinconcreteness ratings, and uptonowthevariancethat is not explainedby imageabilityhas been consideredtobepurenoise,dueperhaps totheimprecisenatureofsubjective norms. This general assumptionis illustratedinthefollowingquote: “Although imageabilityand concreteness aretechnicallydifferent psycholinguisticconstructs, thecorrelationbetweenthese variables is so strongthat manyauthors usetheterms interchangeably. Herewemakethesame assumptionofsynonymybetweenimageabilityandconcreteness interms oftheory(i.e., concreteness effects =imageabilityeffects).”(ReillyandKean,2007: 158). In fact, concreteness and imageabilityratings have beenusedinterchangeablyinmost oftherecent literatureinthe field(e.g. Binder et al.,2005; Richardson, 2003; Fliessbach et al.,2006; Giesbrecht et al.,2004).However, concreteness andimageabilitytapinto,at least partially,different aspects of semanticrepresentations ifnativespeakerintuitions about them aretakenseriously: our analyses of ratings formorethan Embodiedabstract semantics 10 4,000words intheMRC PsycholinguisticDatabaseshowthat thefrequencydistributionof concreteness ratings is bimodal,withtwodistinctmodes forabstract and concrete words (seealso Cartwright &Nickerson, 1979; Nelson &Schreiber,1992),whilethedistributionofimageability ratings is unimodal (Figure1). Inotherwords, concreteness ratings capture thecategorical ontological distinctionbetweenconcreteandabstract words (andtheirunderlyingconceptual representations),whileimageabilityratings index agradedpropertythat is meant tocapturethe differential association of words withsensory(primarilyvisual)properties. Moreover, from atheoretical point ofview,imageabilityratings areaproxyforconcreteness onlyin thedual-codingtheory,not,forexample,intheContext Availabilityhypothesis becauseonlythe former explains differences betweenconcrete and abstract words interms ofwhether (andtowhat extent)thenonverbal imagisticsystem is engaged (Reilyand Kean,2007; Fliessbach,Weis, Klaver, Elger,&Weber,2006). AccordingtotheContext Availabilityhypothesis, however,imageability wouldnot exhaust thedifferences between concreteandabstract words which,instead,ariseas a consequenceofdifferent degrees ofrichness ofsemanticrepresentationwithinaverbal system. Figure1.Densityplots for concreteness andimageabilityratings for 4,274words fromtheMRC PsycholinguisticDatabase.Usingthediptest (Hartigan &Hartigan,1985)werejectedthe hypothesis of unimodalityfor theconcreteness distribution(dip=.0244,p<.001)but not for the imageabilitydistribution(dip =.0058,n.s.).

Description:
a discussion of our novel hypothesis for embodied abstract semantics. This general assumption is illustrated in the following quote: “Although
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.