Why-frequency-can-no-longer-be-ignored-in-ELT-中英文翻

WhyfrequencycannolongerbeignoredinELT?GeoffreyLeechLancasterUniversity1.IntroductionIfaskedwhatistheonebenefitthatcorporacanprovideandthatcannotbeprovidedbyothermeans,Iwouldreply“informationaboutfrequency”.Frequencyisalsoathemewhichhasrecurredinlanguagelearningalthoughithasalsosufferedfromneglect.Hencethereisneedforareappraisalofthelinksbetweenfrequency,corporaandlanguagelearning.Followingthisintroduction,thispaperisdividedintothreemainsections:section2:Abriefglanceathistory;section3:Recentprogressinfrequencystudiesrelevanttolanguagelearning;section4:Newdirectionsinappliedlinguisticsfavorabletofrequency.Tobeginwith,itisaswelltomakeclearthattherearethreeusagesoffrequencythatmightbeconfused.(A)Thereisrawfrequency,whichissimplyacountofhowmanyinstancesofsomelinguisticphenomenonXoccurinsomecorpus,textorcollectionoftexts.(B)Thenthereisnormalizedfrequency(sometimescalledrelativefrequency),whichexpressesfrequencyrelativetoastandardyardstick:e.g.tokenspermillionwords).(C)ThereisalsowhatIwillcallordinalfrequency,wherethefrequencyofXiscomparedwiththefrequenciesofY,ofZ,etc.Thusarankfrequencylist,inwhichwordsarelistedinorderoffrequency,istheclassicexampleofordinalfrequency.Although(A)istherawmeasurefromwhich(B)and(C)arederived,itisoflittleornouseinitself.(B)Normalizedfrequencyisofcourseessentialifwearetomakecomparisonsbetweencorpora,texts,etc.ofdifferentsizes.Butmyviewisthat(C)ordinalfrequencyisthemostusefulmeasuretousewhenweareconsideringlanguagelearning.Itisofnouseforthelanguageteachertobetoldthatshalloccurs175timespermillionwordsinacorpus.Buttobetoldthatwillismuch(15times)morefrequentthanshallmaywellbepedagogicallyuseful.2.AbriefglanceathistoryThehistoricalsketchIamabouttogiveroughlydividesintothreeepochs:(a)earlyfrequencystudies;(b)therejectionoffrequency;(c)thecomputerageandtheresurgenceoffrequencystudies.2.1EarlyfrequencystudiesFormypresentpurpose,itisenoughtorefertooneortwolandmarksintheprovisionofwordfrequencyinformationonEnglish.Thorndike,Thorndike&Lorge,andWestarenotedexamplesofwordfrequencylistsproducedbycountingandcalculatingwordfrequenciesbyhand.Bypresentdaystandards,thecorporausedwerepitifullysmall,andtheselectionoftextstheycontainedincludedsomechoiceshardlyidealforlearnersofthecurrentlanguage.However,theimportantpointhereisthatwordfrequencywastakenseriouslyasaguideforlanguageteachinginthosedays,andinspiteoftheenormousamountofunrewarding“slavelabor”involved,buildingfrequencylistswasfelttobeaworthwhileexercise.Thesimplepostulatejustifyingthiseffortwas:morefrequent=moreimportanttolearn.OfgreaterinterestfromthetheoreticalpointofviewwasthemathematicalworkofZipf.Zipf’sLawheldthatthefrequencyofanywordisinverselyproportionaltoitsrankinthefrequencylist,suchthatthenthwordhasafrequencyofapproximately1/nXthefrequencyofthewordofhighestrank.Zipf’sLawgaveamoreheavilyweightedimportancetothemostfrequentwordsthanwouldbeexpectedaccordingtonormaldistribution.Languageissuchthatthemostfrequent50words(i.e.wordtypes)accountfor50%ofwordtokensinacorpusoftexts;themostfrequent3,000wordsaccountfor85%ofwordtokens;andthemostfrequent10,000wordsaccountfor92%ofwordtokens.ForpracticalpurposeswecansaythatthewordstockofEnglishisbothverylargeandopenended.2.2TherejectionoffrequencyInlinguistics,thesecondhalfofthetwentiethcentury,atleastuptothe1990s,wasdominatedbythegenerativeschoolofNoamChomsky,whorejectedthevalueoffrequencyinthestudyandunderstandingoflanguage.ChomskyfamouslyusedtheillustrationofIliveinDayton,OhioandIliveinNewYorktoshowthatthegreaterfrequencyofthelattersentenceascomparedwiththeformerwasofnolinguisticrelevanceorinterest.Ofcourse,thishadmoretodowiththedifferencesofpopulationbetweenDayton,OhioandNewYork--fromChomskyspointofview,amatterofperformance(andhenceofnovaluetolinguistics)ratherthancompetence.Heconcludedthat“probabilisticconsiderationshavenothingtodowithgrammar”--usinggrammarinabroadChomskyansensetoincludethewholelanguagesystem.Fromthattimeuntil(roughly)theendofthecentury,itwasdifficulttofindanyseriousreferencetofrequencyinpublicationsaboutthelearningoflanguages,andwherefrequencywasdiscussed,itwasdealtwithperfunctorilyandsometimesnegatively.ThewellknownauthoritativehandbookbyRodEllis,TheStudyofSecondLanguageAcquisition(1994),haslittletosayaboutfrequency,andoffersverylittleextrainitssecondeditionofoverathousandpages,publishedasrecentlyas2008.Theonlysubstantialreferencetofrequencyisinthesectionheaded“Thefrequencyhypothesis”,inwhichtheemphasisiswhollyonthelearner’sinputfrequency.Forcorpuslinguistics,amorerelevantquestionis:Howcanboththelearner‘sinputandoutputbeadjustedtothefuturelikelyneedsofthelearnerasrevealedincorpora?2.3ThecomputerageandtherevivaloffrequencystudiesThecorpusrevolutioninlinguisticsbeganwiththecompletionanddistributionoftheBrownCorpusin1964.Shortlyafter,Kucera&FrancisusedthistocreatethefirstwordfrequencylistsforEnglishbasedoncorpusdata.Later,inFrancis&Kucera,theypublishedlemmatizedfrequencylists,basedonthepartofspeech[POS]taggedversionofthecorpus.FurtherwordfrequencylistswerederivedfromtheLancasterOslo/Berge