Corpus-analysis-and-linguistic-theory

1CorpusanalysisandlinguistictheoryWhenthefirstcomputercorpus,theBrownCorpus,wasbeingcre-atedintheearly1960s,generativegrammardominatedlinguistics,andtherewaslittletoleranceforapproachestolinguisticstudythatdidnotadheretowhatgenerativegrammariansdeemedacceptablelinguisticpractice.Asacon-sequence,eventhoughthecreatorsoftheBrownCorpus,W.NelsonFrancisandHenryKuˇcera,arenowregardedaspioneersandvisionariesinthecorpuslinguisticscommunity,inthe1960stheireffortstocreateamachine-readablecorpusofEnglishwerenotwarmlyacceptedbymanymembersofthelinguisticcommunity.W.NelsonFrancis(1992:28)tellsthestoryofaleadinggenera-tivegrammarianofthetimecharacterizingthecreationoftheBrownCorpusas“auselessandfoolhardyenterprise”because“theonlylegitimatesourceofgrammaticalknowledge”aboutalanguagewastheintuitionsofthenativespeaker,whichcouldnotbeobtainedfromacorpus.Althoughsomelinguistsstillholdtothisbelief,linguistsofallpersuasionsarenowfarmoreopentotheideaofusinglinguisticcorporaforbothdescriptiveandtheoreticalstudiesoflanguage.Moreover,thedivisionanddivisivenessthathascharacterizedtherelationshipbetweenthecorpuslinguistandthegenerativegrammarianrestsonafalseassumption:thatallcorpuslinguistsaredescriptivists,interestedonlyincountingandcategorizingconstructionsoccurringinacorpus,andthatallgenerativegrammariansaretheoreticiansunconcernedwiththedataonwhichtheirtheoriesarebased.Manycorpuslinguistsareactivelyengagedinissuesoflanguagetheory,andmanygenerativegrammarianshaveshownanincreas-ingconcernforthedatauponwhichtheirtheoriesarebased,eventhoughdatacollectionremainsatbestamarginalconcerninmoderngenerativetheory.Toexplainwhycorpuslinguisticsandgenerativegrammarhavehadsuchanuneasyrelationship,andtoexploretheroleofcorpusanalysisinlinguistictheory,thischapterfirstdiscussesthegoalsofgenerativegrammarandthethreetypesofadequacy(observational,descriptive,andexplanatory)thatChomskyclaimslinguisticdescriptionscanmeet.Investigatingthesethreetypesofade-quacyrevealsthesourceoftheconflictbetweenthegenerativegrammarianandthecorpuslinguist:whilethegenerativegrammarianstrivesforexplanatoryadequacy(thehighestlevelofadequacy,accordingtoChomsky),thecorpuslinguistaimsfordescriptiveadequacy(alowerlevelofadequacy),anditisar-guablewhetherexplanatoryadequacyisevenachievablethroughcorpusanal-ysis.However,eventhoughgenerativegrammariansandcorpuslinguistshave1©CambridgeUniversityPressCambridgeUniversityPress0521808790-EnglishCorpusLinguistics:AnIntroduction-CharlesF.MeyerExcerptMoreinformation2Corpusanalysisandlinguistictheorydifferentgoals,itiswrongtoassumethattheanalysisofcorporahasnothingtocontributetolinguistictheory:corporacanbeinvaluableresourcesfortestingoutlinguistichypothesesbasedonmorefunctionallybasedtheoriesofgram-mar,i.e.theoriesoflanguagemoreinterestedinexploringlanguageasatoolofcommunication.Andthediversityoftexttypesinmoderncorporamakessuchinvestigationsquitepossible,apointillustratedinthemiddlesectionofthechapter,whereafunctionalanalysisofcoordinationellipsisispresentedthatisbasedonvariousgenresoftheBrownCorpusandtheInternationalCorpusofEnglish.Althoughcorporaareidealforfunctionallybasedanalysesoflan-guage,theyhaveotherusesaswell,andthefinalsectionofthechapterprovidesageneralsurveyofthetypesoflinguisticanalysesthatcorporacanhelpthelinguistconductandthecorporaavailabletocarryouttheseanalyses.1.1LinguistictheoryanddescriptionChomskyhasstatedinanumberofsourcesthattherearethreelevelsof“adequacy”uponwhichgrammaticaldescriptionsandlinguistictheoriescanbeevaluated:observationaladequacy,descriptiveadequacy,andexplanatoryadequacy.Ifatheoryordescriptionachievesobservationaladequacy,itisabletode-scribewhichsentencesinalanguagearegrammaticallywellformed.SuchadescriptionwouldnotethatinEnglishwhileasentencesuchasHestudiedfortheexamisgrammatical,asentencesuchas*studiedfortheexamisnot.Toachievedescriptiveadequacy(ahigherlevelofadequacy),thedescriptionortheorymustnotonlydescribewhetherindividualsentencesarewellformedbutinadditionspecifytheabstractgrammaticalpropertiesmakingthesentenceswellformed.Appliedtotheprevioussentences,adescriptionatthislevelwouldnotethatsentencesinEnglishrequireanexplicitsubject.Hence,*studiedfortheexamisungrammaticalandHestudiedfortheexamisgrammatical.Thehighestlevelofadequacyisexplanatoryadequacy,whichisachievedwhenthedescriptionortheorynotonlyreachesdescriptiveadequacybutdoessousingabstractprincipleswhichcanbeappliedbeyondthelanguagebeingconsideredandbecomeapartof“UniversalGrammar.”Atthislevelofadequacy,onewoulddescribetheinabilityofEnglishtoomitsubjectpronounsasaconsequenceofthefactthat,unlikeSpanishorJapanese,Englishisnotalanguagewhichper-mits“pro-drop,”i.e.theomissionofasubjectpronounthatisrecoverablefromthecontextordeduciblefrominflectionsontheverbmarkingthecase,gender,ornumberofthesubject.WithinChomsky’stheoryofprinciplesandparameters,pro-dropisaconse-quenceofthe“null-subjectparameter”(Haegeman1991:17–20).Thisparame-terisoneofmanywhichmakeupuniversalgrammar,andasspeakersacquirealanguage,themannerinwhichtheysettheparametersof