primary怎么读

InternationalJournalofCorpusLinguistics15:1(2010),56–10.1075/ijcl.15.1.03liu

issn1384–6655/e-issn1569–9811©JohnBenjaminsPublishingCompany

Isitachief,main,major,primary,or

principalconcern?

Acorpus-badbehavioralprofilestudyofthe

near-synonyms

DilinLiu

UniversityofAlabama

UsingtheCorpusofContemporaryAmericanEnglishasthesourcedataand

employingacorpus-badbehavioralprofile(BP)approach,thisstudyexamines

theinternalmanticstructureofatoffivenear-synonyms(chief,main,major,

primary,andprincipal).1Byfocusingontheirdistributionalpatterns,especially

thetypicaltypesofnounsthattheyeachmodify,thestudyhasidentifiedveral

importantfine-grainedmanticandusagedifferencesamongthefivenear-syn-

onymsandproducedameaningfuldelineationoftheirinternalmanticstruc-

thefindingsofthestudychallengeveralexistingunderstandings

oftheadjectives’rmore,theresultsof

thestudyhaveaffirmed(i)thetheoryandapplicabilityoftheBPapproachfor

studyingthemanticandusagepatternsofsynonymsinat,and(ii)previous

rearchfindingsabouttheco-occurrentsofadjectivesthatbestcapturethees-

nceofthemanticsofadjectives,especiallyattributiveadjectives.

Keywords:behavioralprofile,distributionalpattern,internalmanticstructure,

near-synonym

uction

Withitsubiquityinlanguage,synonymyornear-synonymyisacommonyetcom-

ynonymxpressbasicallythesameconcept,

theyoftendosoindifferentfashions,fordifferentcontexts,and/orfromdifferent

rwords,synonymsareoftennotentirelyidenticalinmeaning

uisticterms,they“areneither

infreevariation,norincomplementarydistribution”(Divjak2006:21).Assuch,

Behavioralprofilestudyoffivesynonymousadjectives57

synonymsareachallengingandsimultaneouslyimportantlexicalcategorybecau

theyareesntialforexpressingshadesofmeaningtohelpusconveyourideasand

feelingsprecilyforeffectivecommunication(Edmonds&Hirst2002;Hatch&

Brown1995).Yettheissueofsynonymyhasnotreceivedtheattentionitderves,

especiallyincomparisonwiththeissueofpolymy(Edmonds&Hirst2002;Di-

vjak2006;Divjak&Gries2006,Taylor2003).Additionalsystematicrearchis

neededtoshedmorelightonthisintriguinglinguisticphenomenonand,moreim-

portantly,togainabetterunderstandingofhowvariousspecifictsofsynonyms

tthisbackdrop,theprent

studyexaminesthemanticandusagedifferencesamongfivenear-synonymous

adjectives:chief,major,main,primary,andprincipal,usingacorpus-badbehav-

ioralprofileapproach.2Therationalesforexaminingthistofnear-synonymsby

usingthesaidapproacharethreefold:(i)thecomplexityoftheinternalmantic

structureofthissynonymt,(ii)inadequaciesintheexistingdescriptionsofthe

synonymsinvolved,and(iii)theneedtotesttheeffectivenessofcorpus-badbe-

havioralprofilerearchondifferentiatingsynonymousadjectives.

plexityoftheinternalmanticstructureofthenear-synonym

tandtheinadequaciesintheexistingdescriptionsofthesynonyms

Thesinglemostimportantreasonforexaminingthistofnear-synonymsisits

rathercomplexinternalmanticstructureandtheinadequaciesintheexisting

nehand,thefiveadjectivessharethe

samebasicdictionarydefinitions:“mostimportant”and“main”(LongmanDic-

tionaryofAmericanEnglish2002;TheNewburyHouDictionaryofAmericanEng-

lish2000;TheOxfordDictionaryofAmericanEnglish2005)andtheyappeartobe

interchangeableinmodifyingcertainnouns,suchasconcernandreason,asshown

indictionaryexamplesandevidencedbythefactthatwecanindeedsayachief,

main,major,principal,orprimaryconcern,reason,therhand,when

udwithsomeothernouns,theadjectivesbecomecompletelynon-interchange-

able,nchiefexecutiveandmaininmaindishcannotbereplacedbyany

tentirelyclearforwhichtypesofnouns

allfiveadjectivesmayfunctionasmodifiers,towhatextenttheyaresynonymous

insuchaca,forwhichtypesofnounsonlyoneorsomeoftheadjectivescanbe

udasmodifiers,andwhatthemotivationsareforsuchusagedifferences.

Badonmyscrutinyoftwelvereferencesources(6dictionaries,4thesauri,

and2synonymdictionaries),noclearinformationabouttheissuesappearsto

sthethreeaforementioned

dictionaries,theotherninereferencesareTheAmericanHeritageDictionaryofthe

58DilinLiu

EnglishLanguage(2006),TheAmericanHeritageCollegeThesaurus(2004),The

CambridgeThesaurusofAmericanEnglish(1994),TheConciOxfordAmerican

Thesaurus(2006),LongmanSynonymDictionary(1986),TheOxfordAmericanCol-

legeDictionary(2002),OxfordEnglishDictionary(henceforthOED;online2008),

Webster’sNewDictionaryofSynonyms(1984),andWebster’sNewWorldDiction-

aryandThesaurus(1996).Myexaminationshowedthatapartfromofferingthe

uniquecoremeaningsoftheadjectives(ndprincipalexpressthehighest

inrankofauthorityandpowerandprimaryindicatesfirstinorderororigin),

mostofthereferencematerialsurathergeneralandevencirculardefinitionsfor

theadjectives,which,however,isnotreallysurprisingduetothemainfunction

,Churchetal.(1994:156–158)alsofound

similarinadequatetreatmentofsynonymsintheirstudyofthesynonymsaskfor,

request,welvereferencematerialsIexamined,TheAmerican

HeritageDictionaryoftheEnglishLanguage(2006),Webster’sNewDictionaryof

Synonyms(1984)andtheOED(2008)aretheonlythreethatofferanydetailed

informationregardingtheuniquemeaningsandusagepatternsofeachofthead-

informationisratherlimitedandinsomecaven

erroneous,astheresultsoftheprentstudywillshow.

gtheeffectivenessofthecorpus-badBPapproachinexamining

thesynonymt:Theoreticalfoundationsandmethodologicalissues

Developingamoreadequatedescriptionofthefivesynonymscallsforaclo

corpus-badstudyofthenear-synonymsbecaucorpus-baddescriptionsof

languagehavebeenshowntobemuchmoreaccurateandinformativethantradi-

mplecorpus-badgrammarbooks

(tal.1999;Carter&McCarthy2006)andcorpusstudiesoflexicalus-

agessuchasphrasalverbs,idioms,andlinkingadverbials(r&Davies

2007;Liu2003,2008;Moon1998)havereportedalargeamountofinformation

aboutlexicogrammaticalusageissuesthatchallengestraditionaldescriptions.A

corpus-badstudyofthefivenear-synonymswillbeespeciallymeaningfulbe-

cau,whiletherehavebeensomecorpus-badstudiesonEnglishlexicalitems,

fewarefocudonEnglishsynonyms,especiallyongroupsofsynonymsandtheir

ak(2006:33)correctlynotesinherstudyofagroup

ofsynonymousverbs,theissueoftheinternalstructureofatofsynonymshas

“hithertoremainedlargelyundiscusdintheliterature”.

Thecorpus-badBPapproachhasbeenshowntobeespeciallyeffectivein

thestudyofnear-synonyms(Divjak2006;Divjak&Gries2006;Gries2001;Gries

&Otaniinpress;Hanks1996).Thisapproachisbuiltlargelyonthetheorythat

Behavioralprofilestudyoffivesynonymousadjectives59

themeaningofalexicalitemcorrelatesclolywithitsbehavioralprofileordis-

ghHanks(1996)isbelievedtohavecoinedtheterm

“behavioralprofile”fordescribingthedistributionalpatternsoflexicalitems,the

approachthatadvocatedfocusingondistributionalpatternsinlexicalanalysis

originatedwiththeworkbyFirthandhisfollowersHallidayandSinclair,who

togetherhelpedestablishthestudyoflexisasalinguisticlevelandtheexamina-

ing

toFirth(1957:7,11),“thecompletemeaningofawordisalwayscontextual”and

we“knowawordbythecompanyitkeeps”.Badonthistheory,“[i]nalexi-

calanalysis,”asHalliday(1966:156)explains,“itisthelexicalrestrictionwhichis

underfocus:theextenttowhichanitemisspecifiedbyitscollocationalenviron-

ment”.Inotherwords,themajortaskoflexicalanalysisis,asSinclair(1966:411)

putsit,todescribe“thetendenciesofitemstocollocatewitheachother.”Sinclair

(1987,2004)distinguishestwotypesoftendencyinlexis:“thephraologicalten-

dency”and“theterminologicaltendency”.TofullyappreciateFirth,Halliday,and

Sinclair’stheoryaboutlexicalmanticsandcollocation,onehastogainagood

understandingofthetwotendenciesSinclairdifferentiates.

“Theterminologicaltendency[…]isthetendencyforawordtohaveafixed

meaninginreferencetotheworld”while“thephraologicaltendency”referstothe

factthat“wordstendtogotogetherandmakemeaningsbytheircombinations”

(Sinclair2004:29).Theformeroccursinlanguageuunderthe“openchoiceprin-

ciple”conditionwherelanguageurslectindividualwordstoconstructphras

whomeaningsarederivedfromtheindividualwords(compositionalmeaning);

thelattertakesplaceunderthe“idiomprinciple”conditionwherespeakers/writ-

ersmakeuof“prefabricated”or“mi-prefabricatedphrasthatconstitutesin-

glechoices”withmeaningsthatareoftennotcompositionalbutphraologicalor

unitbad(Sinclair1987:319–320).Asound,completetheoryoflexicalmantics

requiresanunderstandingofbothtypesoflexicaltendencies/meaningsandan

understandingthat“[t]hepreponderanceof[language]usageliesbetweenthetwo

[types]”(Sinclair2004:29).

Firth,Halliday,andSinclair’sworkhasprofoundlyinfluencedcontemporary

rearchonlexicalmantics,encouragingempiricalstudiesthathelpedestablish

aclocorrelationbetweenlexicalitems’meaningsandtheirdistributionalpat-

terns(etal.1994;Miller&Charles1991).Inanexperimentalstudyin

whichcollegestudentswhowerenativespeakersofEnglishwereaskedtoratethe

degreeofsimilaritybetweenwordsinpairsandestimatethecontextualsimilarity

ofsomeofthewordmbeddedinntences(contexts)drawnfromtheBrown

Corpus,Miller&Charles(1991:1)foundthat“themoreoftentwowordscanbe

substitutedintothesamecontextsthemoresimilarinmeaningtheyarejudgedto

be”.3Usinga“lexicalsubstitutability”testinacorpusstudyofthenear-synonyms

60DilinLiu

askfor,request,anddemand,Churchetal.(1994:169)attainedbasicallythesame

finding:lexicalitems’“textualsubstitutability”isagoodindicatoroftheirmantic

similarity,yettherearcherscautionthat“distributionalevidencealonecannotbe

udconclusively”correlationbetweenlexicalmantics

anddistributionpatternsthatthescholarshelpedestablishbyfollowingFirth/

Halliday/Sinclair’stheoreticaltraditionlaidthefoundationsforthecorpus-bad

BPstudiesoflexicalitemsthatbeganinthelate1990s(Divjak2006;Divjak&

Gries2006;Gries2001;Gries&Otaniinpress;Hanks1996).

Inwhatappearstobeoneofthefirstcorpus-badBPstudies,Hanks(1996),

usingtheBritishNationalCorpus(BNC)data,producedtheBPsofveralverbs

includingurge,incite,bother,rofilesarebadonthesyn-

tacticandcollocationalpatternsoftheverbs(especiallytheircomplementstruc-

tures)infor-

mationhelpsrevealeachverb’sprimaryandcondarymeaningsanddifferentiate

itfromits(near)synonyms(aofurge,howitsBPdifferentiatesit

fromrelatedverbslikeask,request,order,andcommand).Althoughnotadetailed

studyfocusingononetofsynonyms,Hanks’discussiondemonstratesclearlythe

powerofacorpus-badBPapproachinenablingustounderstandmanticnu-

ancesandusagepatternsofverbsinawaythatwouldnotbepossibleotherwi.

SinceHanks’study,therehavebeenafewmoreinterestingandwell-executed

corpus-badBPstudiesofverbs(2006;Divjak&Gries2006).Divjak

(2006)andDivjak&Gries(2006)eachexaminedatofnear-synonymousRus-

sianverbs(verbsof“intending”intheformerandverbsof“trying”inthelatter)

viaaclocorpus-badanalysisoftheverbs’morphosyntactic,syntactic,and-

manticdistributionpatterns(measuredbymanyvariablesorIDtagswithwhich

theyannotatedthedata).Theythentabulatedthefindingsoftheanalysisbadon

sultedinalucid

internalstructureofthenear-synonymtinquestionintheformofdendrograms

thatmakeitpossibletoobjectivelydifferentiatethefine-grainedmeaningsofthe

entiatingthefine-grainedmeaningsofsynonymsinatis

importantinunderstandingsynonymyornear-synonymybecau,asEdmonds

&Hirst(2002)havedemonstrated,whenlexicalitemsareconsideredsynonyms,

theysharethesamemeaningatthecoar-grainedlevelbutdifferinmeaningat

thefine-grainedlevelinonewayoranother.

Whiletheexistingcorpus-badbehavioralstudiesaremostlyonverbs,there

havebeenafewonadjectivesrecently(Gries2001;Gries&Otaniinpress).Gries

(2001)prentedaninterestingcorpusinvestigationofthepairedEnglish-icand

-icaladjectives(ic/economical).Usingatwo-dimensional‘Estimationof

SignificantCollocateOverlap’(ESCO)methodinexaminingthepairedadjectives’

collocationpatterns(primarilytheirmodifiednouns,sharedandunshared),hewas

Behavioralprofilestudyoffivesynonymousadjectives61

abletoidentifythedegreeofmanticsimilarityanddifferencebetweenthetwo

,

Gries(2001)wasnotthefirstonetoexaminethetypesofnounsthatadjectives

rts(1986)alreadydemon-

stratedhowthemeaningsofpolymousadjectivescanbedifferentiatedbythe

nounstheymodifyusingtheexampleoftheDutchwordvers(“fresh”).Forexample

whenudtomodifythewordwound,theadjectivemeans“fresh,recent”butwhen

modifyingthewordair,itmeans“fresh,pure,untainted”or“optimal”(Geeraerts

1986:283).Gries&Otani(inpress)isasophisticatedcorpus-badBPstudyofa

tofthreesynonymousadjectives,big,great,andlarge,andtheirantonyms,little,

small,rtogainaclearinternalmanticstructureofthesynonym/

antonymt,theyfirstretrievedallthetokensofthelemmasofthesixadjectives

fromtheBritishComponentoftheInternationalCorpusofEnglishandthenhad

thedata“importedintoaspreadsheetsoftwareandthenannotatedforavarietyof

features”suchassyntax(ratokenadjectiveisanattributiveorpredica-

tiveu)andmantics(whetherthenounmodifiedbyatokenadjectiveisabstract

orconcrete).Thentheyudastatisticalprogramcalled“RscriptBP1.0”toana-

lyzethedata,y,theycompared

theadjectives’BPsusingahierarchicalagglomerativeclusteranalysis(thesame

analysisudinDivjak2006andDivjak&Gries2006),yieldingdendrogramsthat

clearlyexhibitthemanticandusagepatternsofthesynonym/antonymt.

Itisimportanttonotethat,despitethesimilarityintheirgoalofidentifying

thedistributionpatternsoflexicalitems,thespecific-orwhatIwouldliketocall

‘micro’-proceduresudbythescholarsintheaforementionedcorpus-badBP

studies(whichdonotincludeGeeraerts1986,asitisnotacorpus-badstudy)are

ctappearstosuggestthattherearenoestablishedmicro-pro-

,theuofdiverproceduresmakes

perfectninthestudyofinternalstructuresofnear-synonymtsbecauthe

waynearsynonymsdifferoftenvariesfromsynonymtosynonymandfromt

ingtoCru(1986,2000)andEdmonds&Hirst(2002),near-syn-

onymsmaydifferinanyaspectoftheirmeaningandtheirdifferencesfallintofour

broaddimensions:‘denotational’,‘stylistic’,‘expressive’,and‘structural’(Edmonds

&Hirst2002:109).Thusarearcherhastodecide,badonacloscrutinyof

thefeaturesofthesynonymsbeingexamined,whatmicro-procedurestoemploy.

Hanks(1996:96)makesthispointindirectlywhenhearguesforthenecessityof

humanjudgmentincorpus-badbehavioralprofilerearch:

Butthe[usage/distribution]patternsdonotspring,untouchedbyhumanhand,

vetobeteadout,oftenpainstakingly

ureshavetobedevelopedfordistinguishingrelevantfeatures

62DilinLiu

riatelevelsofgeneralizationhavetobechonateverystep.

(Underscoringadded.)

Theabovereviewhasshownthatcorpus-badbehavioralprofilestudiesarean

effectiveapproachtounderstandingsynonyms,onethatsuitsthepurpoofthe

rmotivationforusingtheapproachintheprentstudy

isthat,asshownintheabovediscussion,theapproachhasthusfarbeenud

mostlyinexaminingtsofsynonymousverbswithonlyonestudyonatof

synonymousadjectives(Gries&Otaniinpress).Itwillbeinterestinganduful

tofurthertesttheeffectivenessandviabilityoftheapproachinexaminingalarge

s(1996:92)notes,“thereisnoreasonto

believethattheproceduresthathelpustouandunderstandverbsarerelevant

to,say,nounsoradverbs”.

Inshort,usingacorpus-badBPapproach,thisstudyaimstouncoverthe

manticandusagedifferencesamongthefiveadjectivesinquestionbyexamining

primarilytheirvariousdistributionalpatterns,includingtheircollocational,colli-

gational,rocess,thestudywillalsotrytode-

terminewhichco-occurrent(s)besthelpidentifytheadjectives’manticandusage

pus-baddistributionpatternanalysiswillbecomplementedby

ascrutinyofsomespecificconcordanceexamplestohelpbetteruncoverthediffer-

encesamongthenear-synonymsandtoarriveatasounddepictionoftheinternal

pusudforthis

studywasthe360million-wordonlineCorpusofContemporaryAmericanEnglish

(COCA;Davies2008-).4Thecorpuswaschonforitscomprehensiveness,con-

temporariness,varietyofufulur-friendlyarchfunctions,andeasyaccess.

trieval,analysis,anddiscussion

4.1Overallusagepatterns

Tounderstandtheusagepatternsofthefiveadjectives,thefirstandmostbasicin-

e

allfiveadjectivesmayalsofunctionasnouns,toidentifyeachitem’sadjectiveu,

Iudthe“showpos[partsofspeech]”functionavailableintheCOCAarch

ults(reportedinTable1)indicatedthefollowingfrequencyorder

(fromlowesttohighest):principal,primary,chief,main,andmajor,andavery

largevariationinfrequencyamongtheadjectives(withtheexceptionofprimary

andchiefwhichonlyhadaminusculedifferenceinfrequency).Forexamplethe

frequencyofprincipalisonly10%thatofmajor,20%thatofmain,and37%that

ofprimaryandchief.

Behavioralprofilestudyoffivesynonymousadjectives63

ncyofthefiveadjectives

principalprimarychiefmainmajor

7,83721,05421,07139,07676,586

Thetotalfrequencyinformation,thoughimportant,doesnottellusanythingabout

nascrutinyof

thetwelvedictionariesandthesauriIconsulted,itappearsthatacentralmeaning

thatthefiveadjectivesshareisanindicationofamaximuminimportance,value,

etc.,althoughthesourceorthecoremeaningfromwhichthismeaningisderived

mpleinthecaofchiefand,tosomeextent,

principalaswell,itcomesfrombeingthehighestinrankofauthorityandpower;for

primary,itisderivedfrombeingfirstinquenceororigin;formainandalsomajor

toacertaindegree,

differencesinthecoremeaningscanbeexpectedtocaumanticandusagevaria-

tionsamongtheadjectives,particularlyinthetypesofnounstheyeachtypically

cusonthenounstheadjectivesmodifyismotivatedbytheprimary

functionadjectivesperform–efivesynonymsareallsup-

podlyattributiveadjectives,apointmadeclearinsomedictionaries(-

fordDictionaryofAmericanEnglish2005).Examiningthetypesofnounsattributive

adjectivesmodifyisalogicallysoundandproductiveprocedureforunderstanding

themanticandusagepatternsoftheadjectivesaspreviousrearchhasshown

(Geeraerts1986;Gries2001;Justeson&Katz1995).5Afewotherco-occurrentsor

propertiesrelatedtotheadjectiveswillalsobeexaminedlaterinthispaper.

Toidentifythetypicaltypesofnounstheadjectivesmodify,Ifirstarchedthe

COCAfortheadjectives’modifiednounsviatwotypesofqueries:(i)frequency

(nsthattheadjectivesmodifymostfrequently)and(ii)MutualInfor-

mation(MI),firstintroducedtocorpusanalysisoflexisbyChurch&

Hanks(1990:23),“comparestheprobabilityofobrvingx[word]andy[word]

together(thejointprobability)withtheprobabilitiesofobrvingxandyindepen-

dently(chance)”.AnMIscorearound0suggeststhatthetwowordsdonotcol-

locatewhileascoreof3orhighercanbeconsideredevidencethatthetwoitems

ttheonlystatisticalprocedureudformeasuringcolloca-

tions.T-scoreisalsowidelyudbutMIisadoptedherebecauitfavorscollocat-

ingcontentwords,whichisthefocusoftheprentstudy,whereast-scorefavors

functionwords(formoreinformationaboutthetwomeasures,eChurchetal.

1991;Churchetal.1994).6ThefrequencymeasureandtheMIscorecancomple-

menteachotherinbetteridentifyingthenounsthataretypicallymodifiedbythe

quencymeasureprivilegeswordsthathaveanoverallhighfre-

quencyinthecorpus,yetitundervaluesthothathavealowgeneralfrequency

64DilinLiu

ult,thefrequencymeasure

mayendupincludinginitslistofthenounsmosttypicallymodifiedbytheadjec-

tivesthothatdonotactuallycollocatestronglywiththeadjectivesbutwho

sheerhighgeneralfrequencycausthemtohaveafairlyhighrawfrequency

corecanhelpovercomethis

disadvantageofthefrequencymeasurebutithasitsownshortcominginprivileg-

ingwordsthathavealowoverallfrequencybutaveryhightendencytocollocate

withthelexicaliteminquestion(Church1994:168).Forexampleofallthenouns

modifiedbyprimary,hyperparathyroidismhasthehighestMIscore(12.67)simply

becauthereareonly52tokensofthewordintotaland20ofthemcollocatewith

ore,despiteitsveryhighMIscorewithprimary,hyperparathy-

roidismisoflittleimportanceforthepurpoofthisstudyduetoitxtremely

,toensurethereprentativenessofthemostfrequent

nounsmeasuredbytheMIscore,Iexcludedthonounsthathavefewerthan200

hecaofnouns

modifiedbyprincipal,Iloweredthelectionthresholdto50becau,duetothe

verylowoverallgeneralfrequencyofprincipal,onlyonenounhasmorethan200

tokensasacollocateofprincipal,fivehavebetween100and199tokens,andven

rwords,ifthethresholdof200tokenswerealso

appliedtothenounsmodifiedbyprincipal,thetypicalnounlistforthisadjective

wouldcontainonlyoneword.

BadontheresultsofthefrequencyandtheMIqueries,alistofthetopten

nounsmodifiedbyeachadjectiveineachmeasurewascompiled(eTable2).

DespitethedifferentfociofthefrequencymeasureandtheMIscore,thereissub-

stantialoverlapbetweentheresultsofthetwomeasures:80%overlapforprincipal,

70%forchiefandprimary,60%formain,and40%salsooverlap

eofthissubstantialover-

lap,tincludesonly

thetoptennounsineachmeasureforeachadjectivefortworelatedreasons:First

andforemost,iftheinclusionexpandstofifteenortwentyineachmeasure,thelist

wouldcontainsomeverylow-frequencyitems,especiallyinthecaoftheitems

modifiedbyprincipalduetothelatter’,expanding

thelectiontofifteenortwentydoesnotsignificantlyincreatheoverallnumber

ofwordsintheentirelistduetotheaforementionedoverlapbetweenthenounsin

thetwomeasuresandamongtheitemsmodifiedbythefivedifferentadjectives.7

Anotherissueworthdiscussinghereisthatquiteafewofthenouns(-

ecutiveinchiefexecutive,careinprimarycare,andstreetinmainstreet)canbe

consideredcompoundsormulti-wordunits(MWUs),thusraisingthequestion

sonsforincludingthemarethree-

fold:First,despitethefactthattheymaybeconsideredcompounds,theadjectives

Behavioralprofilestudyoffivesynonymousadjectives65

involvedinthemareeachudasadjectivesandtheyareindeedsotaggedinthe

andmoreimportantly,asnounsin“compound”-likeunits,often

odifiedmostfrequentlybyeachofthefiveadjectivesmeasuredby

frequencyandMIscore

principalprimarychiefmainmajor

Byfre-

quency

ByMI

score

Byfre-

quency

ByMI

score

Byfre-

quency

ByMI

score

Byfre-

quency

ByMI

score

Byfre-

quency

ByMI

score

investi-

gator

266*

investi-

gator

11.73**

care

1,081

objec-

tive

8.35

execu-

tive

5,870

execu-

tive

11.31

street

2,918

street

8.40

league

3,151

league

9.98

compo-

nents

190

spon-

sor

10.75

goal

556

color

8.26

justice

1,803

econo-

mist

11.07

reason

1,075

en-

trance

8.37

role

1,088

obstacle

8.47

owner

160

compo-

nents

9.58

source

512

goal

8.25

opera-

officer

940

nego-

tiator

10.70

effect

955

dish

8.18

prob-

lem

958

corpora-

tions

7.54

spon-

sor

134

resi-

dence

8.82

school

457

pur-

po

8.14

econo-

mist

710

opera-

officer

9.70

thing

885

objec-

tive

7.48

cities

885

cities

7.49

reason

96

owner

8.56

focus

451

source

7.98

engi-

neer

336

justice

9.40

road

615

effect

7.35

changes

794

airlines

7.28

source

88

source

6.89

con-

cern

431

con-

cern

7.83

counl

265

curator

9.45

con-

cern

448

char-

acters

7.33

factor

640

factor

7.13

author

64

goal

6.41

pur-

po

430

respon-

sibility

7.62

cor-

respon-

dent

241

engi-

neer

9.30

charac-

ter

438

cours

7.28

source

627

compo-

nent

6.82

cau

61

reasons

6.38

reason

415

focus

7.44

scien-

tist

184

investi-

gator

8.75

cour

434

reason

7.20

con-

cern

557

source

6.57

goal

57

author

6.17

respon-

sibility

317

care

7.41

pros-

ecutor

153

counl

8.52

prob-

lem

417

con-

cern

7.04

issue

538

concern

6.50

resi-

dence

55

cau

5.89

colors

302

reason

6.68

deputy

149

cor-

respon-

dent

8.51

source

387

pur-

po

6.88

part

528

powers

6.42

Note:*Numberoftotaltokens,requencyasnounmodifiedbyprincipal.

**MIscore.

66DilinLiu

theyeachmaytakeonlyoneofthefiveadjectivesasitscollocate(imary

maycollocatewithcareinprimarycare).Thisfactindicatesaclearspecialman-

ticandusageconnectionbetweentheadjectiveandthenouninquestion,anissue

,determiningtheadjectives’typical

nouncollocatesisthekeyprocedureudinthisstudytouncoverthedistribu-

enounsarestrongcollocateswiththeir

respectiveadjectives,itisnecessarytoincludetheminthisstudywhethertheyare

trulycompounds/MWUsornot.

Todeterminethedifferencesamongthenear-synonymadjectiveswithregard

tothetypesofnounstheymodify,Iclassifiedthenounsintosixmajormantic

categories:(a)abstract(nandreason),(b)concrete(ddish),

(c)dual(andcomponent),(d)institution(andcorporation),

(e)position-title(iveandcounl),and(f)non-position-title(-

sorandauthor).8TheclassificationresultsareincludedinTable7intheAppendix

withexplanationsaboutwhysomeoftheitemswereclassifiedthewaytheywere

dentifiedthetypesofnouncollocatesofeachadjec-

tive,andtheresults(reportedinTable3)showclearlythattheyvarysubstantially

fromadjectivetoadjective.

fnounsmodifiedbyeachadjective

principal

(10different

nouns)

primary

(11different

nouns)

chief

(13different

nouns)

main

(14different

nouns)

major

(15different

nouns)

3abstract

3non-position-

title

2dual

1concrete

1position-title

7abstract

2institution

1dual

1concrete

13position-title7abstract

5concrete

2dual

6abstract

5institution

4dual

Thetypesofnounsthatboastthewidestdistributionrangeareabstractanddual

chappearedwithfourofthefiveadjectives,

resultsprovideuswithsomeimportantinformationabouttheusagepatternsof

mpleeventhough

allthemostfrequentnounsunderchiefarepositiontitles,itdoesnotmeanthe

do

notknowforcertainwhetherandhowoftentheadjectiveisudtomodifythe

suchinformationtogainabetterand

morecompleteunderstandingoftheadjectives’usagedistributionsandmeanings.

Thus,IqueriedCOCAregardingthefiveadjectives’frequencieswitheachofthe

50nounsandthentabulatedtheadjectives’totalfrequencieswiththe50nouns

Behavioralprofilestudyoffivesynonymousadjectives67

ultsarereported

Idiscussthestatisticalanalysisofthefrequencydistributions,it

isimportanttonotethattheorderofthetotalfrequenciesoftheadjectives’us

withthe50nounsmirrorsthatoftheoveralltotalfrequenciesoftheadjectives

reportedinTable1:principal

relationoffersindirectevidencethatthe50most-frequentnounsarereprenta-

rderweredifferent

here,thatwouldraiquestionsaboutthereprentativenessofthefiftynouns.

Forexample,ifprincipalwererankedhigherthanprimaryhere,itwouldmeanthe

adjectivehadmoretokensthanprimaryinthefiftynouns,butintheoveralldata

turnwouldindicate

anover-reprentationofprincipalandanunder-reprentationofprimaryinthe

fiftynounsinrelationtotheiroveralldistributionsinthecorpus.

butionsofthetypesofnounsmodifiedbythefiveadjectives*

AbstractConcreteDualInstitu-

tion

Position-

Title

Non-

position-

Title

Total

principal740A70A557T38A421A395T2,221

primary4,001T440A1,124T1,705T55A847,409

chief420A1A116A16A11,465T75A12,093

main7,422T5,157T1,639A161A4A44A14,427

major7,746T199A3,150T6,127T12A67A17,301

Note.*AcellfrequencyfollowedbytheletterTmeansitisa‘type’badonanHCFAtest(discusd

below)whileacellfrequencyfollowedbyanAmeansitisan‘antitype’.Acellfrequencyfollowedbyno

letterisneither.

4.2Semanticandusagepatternsacrossdifferenttypesofnouns:Noticeable

variationsandsimilarities

Todeterminewhetherandinwhatwaythefiveadjectives’distributionsamong

thesixtypesofnounsdiffersignificantly,Iconductedamultifactorialtestcalled

hierarchicalconfiguralfrequencyanalysis(HCFA)withtheadjectives’frequency

numbersinTable4,usingGries’(2004)HCFA3.2forRprogram.9Theresults

isamuchmoreinformative

testthantheChi-squaretestbecau,inadditiontoyieldingtheresultsthata

Chi-squareproduces,italsoshowswhichcellfrequenciesinthecontingencyta-

ble

frequencysignificantlyhigherthanexpectedisconsidereda‘type’whereasacell

frequencysignificantlylowerthanexpectedisan‘antitype’.Acellfrequencythat

68DilinLiu

ishigherorlowerthanexpectedbutitsdifferenceiswhatchancewouldpredictis

consideredneitheratypenoranantitype,markedas“ns”(notsignificant)inthe

“Dec”(decision)columnoftheresultsprintout.

TheChi-squarepartoftheHCFAtestshowedasignificantdifferenceamong

thefiveadjectives’distributionsacrossthesixtypesofnouns:χ2(df=20)=69616.09,

p<0.001,withaneffectsizeofCramer’sV=ailedresultsoftheHCFA

testrevealthatinthreeofthesixnouncategories(rete,position-title,

andnon-position-titlenouns)oneadjectivedominatesthecategorywithafre-

quencysosignificantlyhigherthanexpectedthatitconstitutestheonlytypeinthe

oncretecategory,mainisthesoletypewiththeotheradjectives

allbeingantitypes;intheposition-titlecategory,chiefistheonlytypewiththe

otheradjectivesallbeingantitypes;inthenon-position-titlecategory,thesituation

isslightlydifferent:principalisthesoletype,chief,main,andmajorareantitypes,

andprimaryisneitheratypenoranantitypebecau,althoughitsfrequencyis

lowerthanexpected-asarethoofchief,main,andmajor-,itisnotsignificantly

esultssuggestthatconcretenounsaremodifiedprimarilybymain,

position-titlenounsalmostexclusivelybychief,andnon-position-titlenouns

stheaforementionedthreetypesofnouns,theinstitu-

y

differenceisthatinthiscategorytwoadjectives(majorandprimary)aretypes

rwords,institutionnounsare

,thetwoadjectivesincombination

modify97%ofnounsinthecategory,leavingchief,main,andprincipaltoshare

theremaininginfinitesimal3%.Onemorepointworthnoticingisthat,inallof

thefournouncategories,thetypesaresodominantthatsomeoftheantitypeshave

onlyafewtokens(numberinginthesingleordoubledigits).

Comparedwiththeabovefourcategories,abstractanddualnounxhibita

etwocategories,threeadjectives

ticeableis

thefactthat,inthetwocategories,eventhetwoantitypeadjectiveachhavea

substantialnumber(allinthehundreds).Infact,althoughprincipalintheabstract

categoryisanantitype,thenumberoftokensithasinthecategory(740)ismuch

higherthanitsnumbersinanyoftheothernouncategoriesincludingthenon-po-

sition-titlecategorywhereitistheonlytype(withonly395tokens).Chiefisanan-

titypeinboththeabstractanddualnouncategories,andyetthenumbersoftokens

ithasinthetwocategoriesaremuchhigherthanitsnumbersintheothercatego-

aofmain,

thoughitisanantitypeinthedualcategory,ithasinfactaverylargenumberof

tokens(1,639)init,,intermsofrawfrequencycounts,

abstractanddualnounsaretheonlytwocategoriesinwhicheachadjectivehas

Behavioralprofilestudyoffivesynonymousadjectives69

rmore,asmentionedearlier,abstract/dual

nounsaretheonlytwocategoriesthatappearedinthelistofthemostfrequently

modifiednounsunderfourofthefiveadjectives(exceptchief;3).

Thatallofthefiveadjectivesareudfairlycommonlytomodifyabstract/dual

nounsalsoemstobeafactacceptedbydictionaries,forwhendefiningthefive

adjectivesmostofthedictionariesIexaminedudexamplesshowingtheadjec-

tivesmodifyingabstract/dualnouns,,main,major,primaryandprincipal

concern/reason/,badonthedataanalysissofar,thefiveadjectives’

mainusagepatternsacrossthesixtypesofnounmaybesummarizedandillustrated

difyingabstract/dualnouns,allfiveadjectives

appeartomean“mostimportant”,“rious”,,theremaybefine-grained

meaningandfunctionaldifferencesamongtheadjectives’uwiththetwotypes

ofnouns,animportantissuethatwillbeexaminedinthectionsbelow.

Non-Position

TitleNouns

Modi