Lexical Bundles in Discourse Structure--A Corpus-B

AppliedLinguistics2013:34/3:369–388OxfordUniversityPress2012doi:10.1093/applin/ams045AdvanceAccesspublishedon27September2012LexicalBundlesinDiscourseStructure:ACorpus-BasedStudyofClassroomDiscourseENIKOCSOMAYDepartmentofLinguisticsandAsian/MiddleEasternLanguages,SanDiegoStateUniversity,5500CampanileDrive,SanDiego,CA92182-7727,USAE-mail:[e-mail]scorpus-basedmethodstodocumentthedistributionalpatternsofpreviouslyreportedlexicalbundlefunctionsastheyrelatetodis-coursestructure.Specifically,84lexicalbundlesandtheirdiscoursefunctions(Biberetal.2004a)weretrackedin1,176discourseunitsextractedfromtheinitialphasesof196universityclasssessions.Thefindingsshowthatintheopeningphaseofclasssessionsbundlesconveyingstancearemostfrequentwhilebundlesarticulatingreferenceareleastinnumber.Withthestartoftheinstructionalphase,however,stancebundlessharplydropandreferentialbun-dlesstartrising.Discourseorganizingbundlesareleastprominentoverall,al-thoughwithsomeriseattheverybeginningoftheinstructionalphase.Thesefindingsindicatethatlexicalbundlefunctionscorrespondwelltothecommu-nicativefunctionsindiscoursestructurefoundthroughlinguisticvariation(Csomay2005)andsuggestthatthereisastrongrelationshipbetweengrammarandlexisonthediscourselevel.BACKGROUNDTheinquiriesofthepastfourdecadesontherelationshipbetweenlexicalpatternsanddiscourseleadtodescriptionsofvocabularypatternsintwomajorways:(i)howindividualwordsrelatetooneanotherastheyformcol-locationsorotherfrequentlyco-occurringpatterns(e.g.lexicalbundles);and(ii)howpatternsofvocabularyrelatetothetextualaspectsofdiscourse.Studiesinthefirstgrouptypicallytrackeitherfrequencypatternsofindividualwordsorreportonwaysinwhichtwoormorewordssystematicallyappeartogether(Sinclair1991;Biberetal.1999;Hunston2002;Schmitt2004).Scholarsrelyonlargecollectionsoftexts(acorpus)toidentifysuchpatterns,andapplyautomatic,computationaltechniquestoextractthem.Toanalyzesuchpatternsfurther,theydescribetheinternalgrammaticalmake-upof,forexample,multi-wordcombinations(Biber2009),reportontheirfunctionalpropertiesintext,anddocumenttheirdistributionacrosstextstohighlightregisterdifferences(Biberetal.1999).Althoughmulti-wordsequencesareatShanghaiInternationalStudiesUniversityonNovember14,2013(HallidayandHasan1976)orthehierarchicalstructureofdiscourse(MannandThompson1988;NattingerandDeCarrico1992;NesiandBasturkmen2006),anddocumentinformationflow(Prince1981;Givo´n1983)ortopicshifts(Hansen1994)asthesediscoursefeaturesaremarkedbylexicalpatternsintexts.Whilestudiesinthepastofferedin-depthanalysesofasmallnumberoftextsatatime,workinthenewmillenniumhascombinedquantitativeandqualitativeinvestigations.Toprovidegeneralizablefindingsontherelationshipbetweenintra-textuallinguisticvariationanddiscoursestructure,mostrecentstudies(e.g.Csomay2005)haveappliedcorpus-basedtechniquestotheirana-lyses.Thatis,computationaltoolsweredevelopedtoautomaticallysegmentalargenumberoftextsintosmallercoherentunits,andtherelationshipbe-tweenthepatternsofco-occurringgrammaticalfeaturesinthoseunitsandthemacro-structureofdiscoursewasanalyzed.Whatismissinginthislineofresearch,however,isadescriptionoftherelationshipbetweenpatternsoflexis(e.g.lexicalbundles)inthoseunitsanddiscoursestructure.Thepresentresearchtakeslexicalbundlespreviouslyidentifiedacrosstextsinclassroomdiscourse(Biberetal.2004a)andlooksatthedistributionalpat-ternsoftheirfunctionsinthepreviouslyidentifiedsmallerdiscourseunits(Csomay2005)toexploretheirrelationshiptodiscoursestructure.Inthissection,morebackgroundisprovidedinthetwoareasoutlinedaboveastheyframetheneedforthisstudy.First,studiesontheidentificationandclassificationoflexicalbundlesareintroduced.Second,studiesonthestruc-tureofclassroomdiscoursearediscussed,includingabriefintroductiontocorpus-baseddescriptionsoftherelationshipbetweengrammaticalpatternsanddiscoursestructureinthatcontext.Inthenextsection,thedetailsofthemethodologyusedinthisstudyaredescribed,focusingonhowlexicalbundlesweretrackedinthepreviouslyidentifiedandgrammaticallycharacterizeddis-courseunits.Thefinalsectionsdiscussthefindingsandimplications.LexicalbundlesLexicalbundleshavebeendefinedasthemostfrequentrecurringlexicalse-quencesinaregister(Biberetal.1999).1Theyaresequencesofwordswithvaryinglength:threeormorewordsatatime.Somesequencesappearinmultipleregisters,whileothersareuniquetoparticularregisters.Bundlessuchas,whatdoyoumean,Idon’tknowwhyappearmoreoftenineverydayconversation,whileasaresultof,inthecaseof,ontheotherhandoccurmoreofteninacademicprose.370LEXICALBUNDLESINDISCOURSESTRUCTUREatShanghaiInternationalStudiesUniversityonNovember14,2013