汉字形声字声旁的语义加工
心理学报2002,34(1)
ActaPsychologicaSinica
SemanticProcessingofPhoneticRadicals
inReadingChineseCharacters木
ZhouXiaolin
(CenterforBrainandCognitiveSciences,andDepartmentofPsychology,PekingUniversity,Bering,100871)
WilliamMarslen—Wilson
(MRCCognitionandBrainSciencesUnit,Cambridge,UnitedKingdom)
Abstract
Mo8tcomplexChinesecharactersarecomposedofasemanticradicalontheleftandaphoneticradicalontheright,
whichmayprovideinformationconcerningthepronunciationofthewholecharacter.Asemanticjudgementtaskwasusedto
investigatewhethersublexicalprocessingofphoneticradicalsembeddedincomplexcharactersispurelyaphonologic?,d
event,involvingactivationofphonologicalinformationassociatedwiththephoneticradicals,orwhetheritisalsoasemantic
event,involvingactivationofsemanticpropertiesrelatedtotheradicals,whicharemeaningfulcharactersontheirown.
Signifieantinhibitoryeffectswasfoundforcomplexcharacterswhosephoneticradicalsweresemanticallyrelatedtotheother
memberoftheconsecutivelypresentedpairofcharacters.Themagnitudeoftheinhibitoryeffectswasgenerallynot
influencedbytheregularityofphoneticradicalsinprovidingphonologicalinformationforthewholecharacters,norbythe
Dresentationorderofcomplexcharactersandsemanticassociatesofthephoneticradicals.Itisarguedthat,inreading
Chinese,phoneticradicalsembeddedincomplexcharactersaredecomposedfromvisualinputandusedtoactivatetheirown
phonologicalandsemanticproperties,inparalleltotheprocessingofwholecharacters.
KeywordsreadingChinese,Chinesecharacters,sublexicalprocessing,phoneticradicals.
1Introduction
TheChinesewritingsystemisoftendescribedas
logographicormorpho—syllabic,wherethebasic
orthographicunits,thecharacters,corresponddirectly
tomorphemicmeaningsandtosyllablesinspoken
form.Withsomeexceptions,eachcharacterrepresents
onemorphemeandhasonepronunciation,although
differentcharactersmayhavethesamepronunciations.
ModemChinesecharacterscanbebroadly
differentiatedintotwocategories:simpleandcomplex,
bothofwhichaxecomposedofs~okesandarrangedin
squaresofsimilarsize.Simple(orintegrated)
charactersmakeupabout5%ofthetotalcharactersin
ModemChinese.Theyareholisticvisualpatternsthat
cannotbedividedmeaningfullyintosub—lexicalunits.
Complex(orcompound)charactersconstituteabout
95%ofallmodemChinesecharactersandmostof
thesecharactersarecomposedofasemanticradicalon
theleftandaphoneticradicalontheright(e.g.,议yi
[4],discuss,inwhichthephoneticradicalis.义yi
[4],righteousness.Numbersinbracketsrepresentthe
lexicaltonesofsyllables),althoughsomearrangetheir
radicalsinotherways[„,.
收稿日期:2001—03—05.
TheresearchreportedherewassupportedinpartbygrantsfromChinaNationalPandengProgram(95-Sp
ecial-09),NationalScienceFoundation
(3007026)andtheMinistryofEducationofChina(99000127)toXiaolinZhouandinpartbyagranlfrom
UKEconomicandSocialResearchCouncil
t0XiaolinZhouandWilliamMarslen.Wilson.Correspondenceconcerningthispapershouldbeaddress
edtoDr.XiaolinZhou,DepartmentofPsychology,
PekingUniversity,Beijing10087l,China.Electronicmailmaybesentloxzl04@pku.edu.cn.
1
2理学报34卷
Linguistically,whilesemanticradicalshavethe
functionofindicatingthesemanticcategoryof
morphemescorrespondingtothewholecharacters,
phoneticradicalshavethefunctionofpointingtothe
pronunciationsofthesecharacters.However,dueto
theevolutionofthewritingsystem,manycomplex
charactersarenolongerpronouncedinthesamewayas
theirphoneticradicals.Lessthanonethirdofcomplex
characters(“regular”characters)haveexactlythe
samepronunciationsastheirphoneticradicals_3I4],
whileaboutonethirdaretotallyirregularf”irregular”
characters)andtheirpronunciationhasnorelationship
totheirphoneticradicals.Thefinalonethirdshare
someaspectsofphonologywiththeirphoneticradicals.
Mostphoneticradicalscanbeinvolvedinbothregular
andirregularcharacters.Moreover,mostphonetic
radicalsal?ealsomeaningfulcharactersbythemselves,
eventhoughtheirmeaningusuallyhasnothingtodo
withthemeaningofcomplexcharacterscontaining
theseradicals.
Therehavebeenanumberofstudiesonthe
sublexicalorsub?characterprocessingofphonetic
radicalsinreadingChinesecharactersE14].
Withfew
exceptions[?,,
,thesestudieshaveconcentratedon
phonologicalactivationofphoneticradicalsandthe
influenceofthisactivationonphonologicalprocessing
ofwholecharacters.Seidenberg[
,
forexample,
foundthatregularcomplexcharacters—————thatis.
charactershavingthesamepronunciationsastheir
phoneticradicals,_??___._______——werenamedfasterthan
frequency—matchedsimplecharacters.Thiseffect,
however,wasrestrictedmostlytolowfrequency
characters.Usingaprimednamingtask,Zhouand
I?arslen—Wilson[H]demonstmtedmoredirectlvthe
activationofphonologicalinformationassociatedwith
phoneticradicalsembeddedincomplexcharacters.
Highandlowfrequencyirregularcomplexcharacters
(e.g.,猜cai[1],guess,inwhichthephonetic
radicalwas青qing[1],blue)wereusedasprimes,
whiletargetswerecharacters(e.g.,轻qing[1],
坛ht)thatwerehomophonictothephoneticradicals
embeddedinthecomplexcharactersbutnottothe
charactersthemselves.Theauthorsfoundthattargets
precededbylowfrequencycomplexcharacterswere
namedfasterthanwhentheywereprecededbyunrelated
characters.Targetsprecededbyhighfrequency
complexcharacters,however,didnotshowsignificant
facilitationinnaming.
Thesefindingssuggestthatinprocessingcomplex
characters,phoneticradicalsaredecomposedfromthe
visualinputandusedtoaccesstheirownphonological
representationsaswellasrepresentationsofother
characterscontainingtheseradicals,leadingtoa
cooperativeandcompetitiveinteractionbetweenthe
phonologicalactivationofwholecharactersandoftheir
phoneticradicals[9,13,14].Anahernativeaccount,
however.wasalsoproposed[15I.Accordingtothis
account,sublexicalprocessingofphoneticradicalsin
readingcomplexcharactersisessentiallythesameas
theprocessofusinggrapheme?phoneme
correspondencesinreadingalphabeticwords.The
over—-learnedradical—-soundcorrespondencessubserve
therule—likecomputationof”prelexical”or
“nonlexical”phonologyforthewholecharacters.Thus
onthisaccount,sublexicalprocessingofphonetic
radicalsispurelyaphonologicalevent,involvingonly
computationfromorthographytophonology.
Themainpurposeofthisstudyistoinvestigate
whethersublexicalprocessingofphoneticradicals
embeddedincomplexcharactersisrestrictedjusttothe
activationoftheirphonologicalrepresentations.Weuse
asemanticjudgmenttasktoexaminewhetherthe
semanticpropertiesassociatedwiththesephonetic
radicals,sincetheyaremeaningfulcharactersontheir
own,arealsoactivatedintherecognitionofcomplex
characters.Ifweobtainevidenceforsemantic
activationofphoneticradicalsinreadingcomplex
characters,thisevidencewillaHowustorejectthe”
prelexical”or”nonlexical”phonologyaccountof
sublexicalprocessingofphoneticradicals.
ZhouaIldMarslen—Wilson[,,4]haye
investigatedthisissueinaprimednamingstudy.They
observedfacilitatoryprimingeffectsfortargetcharacters
(e.g.,东dong[1],east),whichwerenot
semanticallyrelatedtothecomplexcharacterprimes
themselves(e.g.,洒sa[3],spray)buttothe
phoneticradicals(e.g.,西xi[1],west)embedded
intheprimes.Theyalsofoundaninhibitoryeffectfor
ZhouXiaolin:SemanticProcessingofPhoneticRadicalsinReadingChineseCharacters3
complexcharactertargets,whichwereprecededby
primessemanticallyrelatedtothephoneticradicals
embeddedinthetargetsbutnottothetargets
themselves.Theauthorsarguedthatsublexical
processingofphoneticradicalsisbothaphonological
andasemanticeventandthattherearenofundamental
differencesbetweensublexicalofphoneticradicalsand
lexicalprocessingofsimpleandcomplexcharacters.
Inthepresentsemanticjudgmenttask,
participantswereaskedtodecidewhetherapairof
consecutivelypresentedcharactersweresemantically
related.111ecrucialcomparisonwasforcharacterslike
洒(sa[3],spray)and东(dong[1],east)thatare
notsemanticallyrelatedaswholes,butwherethe
phoneticradical(e.g.,西xi(1),west)embedded
inthecomplexcharacterisrelatedtotheothermember
ofeachpair.Ifsemanticpropertiesoftheembedded
phoneticradicalareactivatedinreadingthecomplex
character.thesemanticrelationbetweenthisradical
andtheothercharactershouldsendapositivesignalto
thedecisionsystemandmakeitdifficulttorejectthe
pairofcharactersasunrelated.If,ontheotherhand,
sublexicalprocessingofthephoneticradicalispurelya
phonologicalevent,thispairofcharacterswillnotbe
relatedateitherthewhole.characterlevelorthe
sublexicalleve1.Theyshouldnotbemoredifficultto
rejectthanappropriatecontrolsinthejudgmenttask.
Animportantpartofthepresentexperimentisthe
useofbothregularandirregularcomplexcharacters.
Thiswasnotonlytoexaminewhetherthepotential
sublexicalsemanticactivationisuniformacross
differenttypesofcomplexcharacters,butalsoto
examinetowhatextentsublexicalphonological
activationcontributestothissublexicalsemantic
activati0n.Forirregularcharacterslike洒(sa[3],
spray),phonologicalactivationofthewhole
charactersshouldcompetewithsublexicalphonological
activationoftheirphoneticradicals(e.g.,西xi[1],
west),hencereducingthelevelofsublexical
phonologicalactivation.Forregularcomplexcharacters
like谓(wei[4],say),however,phonological
processingofthewholecharactersandphonological
processingoftheirphoneticradicals(e.g.,胃wei
[4],stomach)supporteachother.Ifsublexical
semanticactivationofphoneticradicalsismediatedor
atleastinfluencedbyphonologicalactivationofthese
radicals.thelevelofsublexicalsemanticactivationand
hencetheinterferenceeffectinsemanticjudgment
shouldbedifferentforthetwotypesofcomplex
characters.Thususingbothregularandirregular
charactersinthisexperimentcouldshedadditional
lightonthecontroversialissueofwhethersemantic
activationinreadingChinesecharactersis
predominantlymediatedbyphonologicalactivationor
throughdirectmappingbetweenorthographyand
semantics[16..
Thisstudyalsomanipulatedthepositionof
complexcharactersinthecriticalpairs.Theregularor
irregularcomplexcharacterswerepresentedeitheras
thefirstorthesecondcharacter.Ifdecompositionof
phoneticradicalsandaccesstotheirsemantic
propertiesisinducedbytheconsecutivepresentationof
pairedstimuli,thenlargerinterferenceeffectsshould
beobservedwhenthecomplexcharactersarepresented
asthesecondcharactersthanwhentheyarepresented
asthefirstcharacters.Thepresentationofcharacters
whicharesemanticassociatesofthephoneticradicals
embeddedinthecomplexsecondcharactersshould
encouragetheirdecompositionandsemanticactivation.
Thestimulusasynchronyonset(SOA)between
thefirstandsecondcharacterswassetat100ms.This
shortSOAwasusedtominimizepossiblecompetition
effectsthatcouldreducesublexicalprocessingeffects
incomplexcharacterspresentedasthefirstcharacter.
AlongerSOAwouldgivethesemantic(and
phonological,forirregularcharacters)activationofthc
wholecomplexcharactersenoughtimetosuppressthe
semantic(andphonologica1)activationoftheir
phoneticradicals.Ontheotherhand,previousstudies
usingthisandsimilartasks[suggestthatparticipants
couldhavedifficultiesinidentifyingthefirstcharacters
iftheSOAisshorterthan100ms.
2Method
2.1DesignandMaterials
Sixtyregularcomplexcharactersand52irregular
complexcharacterswereselectedasthecritical
stimuli.Fortyoftheirregularcharactersdidnotshare
4理学报34卷
anyoftheirpronunciationwiththeembeddedphonetic
radicals.Aminoritysharedeithertheirconsonants(8
out52)ortheirvowels(4out52)withtheirphonetic
radicals.Allthephoneticradicalshadstrongsemantic
associatesthatwereusedastheothermemberineach
pair.Forexample,theirregularcharacter洒(sa[3],
spray)waspairedwith东(dong[1],east),which
issemanticallyrelatedtotheembeddedphonetic
radical西(xi[1],west);theregularcharacter谓
(wei[4],say)waspairedwith肚(du[4],belly),
whichisasemanticassociateofthephoneticradical胃
(wei[4],stomach).Thesemanticrelatedness
betweenthephoneticradicalsandtheothermember8
werecheckedagainstadatabaseinwhichthesemantic
relatednessbetweenpairsofcharacterswerejudgedby
atleast15undergraduatestudentsona9-pointscale
(1=totallyunrelatedand9=veryrelated).Average
relatednesswas8.3forirregularcharacters(ranging
from7.3to9.0)and8.2forregularcharacters
(rangingfrom7.1to9.0).Theregularorirregular
complexcharacters,asawhole,werenotsemantically
relatedtotheothermemberofeachpair.
Onlycomplexcharacterswithaninternalleft—right
structurewereselected,withtheembeddedphonetic
radicalsalwaysontherightsideofthecomplex
characters.Thisisnotonlybecausetheyconstitutethe
majortypeofcomplexcharactersinChinese,butalso
becausetheyareeasiertodecomposeinperceptual
analysesthancharacterswithotherstructures.
so
thattheeffectsoftheirsublexicalprocessingshouldbe
easiertodetect.Thesecomplexcharacterswerealsoof
relativelylowfrequency.Theaveragefrequencywas85
permillionfortheirregularcharactersand25per
millionfortheregularcharacters.Thereasonfor
selectinglowfrequencycomplexcharacterswasthat
visualdecompositionandsublexicalprocessingare
morelikelytotakeplaceforlowratherthanhigh
frequencycharacters[9.1.
However.duetothe
distributionofcharactersinChinese[2o].wecouldnot
balancethefrequenciesofregularandirregular
complexcharacters.Anyway,wewerenotinterestedin
companng
Thevisual
regularandirregularcharactersdirectly
complexityofthesecharacters,which
correlateswiththeirfrequency,wasnotperfectly
matchedeither.Theaveragenumberofstrokes,which
measuresvisualcomplexity,was8.7percharacterfor
irregularcharactersand10.7percharacterforregular
complexcharacters.Theaveragefrequencyoftheother
pairedcharacters(i.e.,semanticassociatesofthe
phoneticradicals)was1161permillionforthe
irregulargroupand864fortheregulargroup.The
averagefrequencyoftheembeddedphoneticradicals,
whentheyareusedasindependentcharacters,was
1137permillionfortheirregulargroupand756forthe
regulargroup.Clearly,theyweremuchhigherthanthe
frequenciesofthecomplexcharactershavingthemas
phoneticradicals.
Thepositionofeachpairoftestcharacterswas
reversedsothatthecomplexcharacterswereeither
presentedfirst(e.g.,洒(sat3],spray)一东(dong
[1],east)orsecond(e.g.,东(dong[1],east)一
洒(sa[3],spray).Unrelatedcontrolpairswere
createdbyre—pairingthefirstcharacterswiththe
secondcharactersinthesamestimulusgroup,sothat
thesamesetofstimuliwereusedacrosstherelatedand
unrelatedconditions.Forexample,fortwocritically
relatedpairshavingirregularcomplexcharactersasthe
firstcharacters,洒(sa[3],spray)一东(dong[1],
east),and猜(cai[1],guess)一紫(zi[3],
purple),theirunrelatedcontrolpairswere,
respectively,猜(cai[1],guess)一东(zi[3],east)
and洒(sa[3],spray)一紫(zi[3],purple).
Fortheregularandirregulargroupsofstimuli,a
Latinsquaredesignwasusedtoassignthepairsof
charactersinto4counter—balancedtestversions.111is
designwouldallowthesamecharactersappearonly
onceinaversion.Ineachversion,halfofthe112
pairswerecriticallyrelated.Amongtherelated56
pairs,halfofthemtookthecomplexcharactersasthe
firstcharacterandhalfasthesecondcharacter.There
were13irregularcharactersand15regularcharacters
actingasthecriticallyrelatedfirstorsecond
characters.Thirty—fourpairsofcharactersthatwere
neithersemantically,norphonologically,nor
orthographicallyrelatedwereusedasfillersandadded
toeachtestversion.These,togetherwiththecritical
stimuli,required”no”responsesinthesemantic
judgmenttask.Another150pairsofsemantically
ZhouXiaolin:SemanticProcessingofPhonet