The formal semantics of grammaticalization

NELS25Proceedings(GLSA,UMassAmherst),Volume2:PapersfromtheWorkshopsonLanguageAcquisition&LanguageChange,pp.175-189TheFormalSemanticsofGrammaticalizationKaivonFintelMassachusettsInstituteofTechnologyGrammaticalizationisthegradualhistoricaldevelopmentoffunctionmorphemesfromcontentmorphemes.Amongthecommonlyidentifiedcharacteristicsofthisprocessiswhatisoftencalled“semanticbleaching”:whilebecomingmoreandmorefunctionalthemorphemelosesmostofitsmeaning.Inthispaper,Iaskwhatsensewecanmakeof“semanticbleaching”fromaformalsemanticperspective.Iconsiderseveralwaysofcharacterizingthesemanticsoffunctionalmorphemesanddiscusswhathastohappentothemeaningofamorphemethatisbeinggrammaticalized.Mymainclaimsare:•Functionalmorphemesarenotmeaningless.•Theirmeaningsarespecialhowever:theyarepermutation-invariant,havehightypes,andaresubjecttouniversalconstraints.•Grammaticalizationenrichestheinventoryoffunctionalmeaningsofalanguagebycomposingpre-existingfunctionalmeaningswith“bleached”lexicalmeanings.Iwritethispaperwithalotoftrepidation,sinceIhaven’tevenbeguntogetwellacquaintedwiththevastliteratureongrammaticalizationandthemanydetailedandsophisticatedanalysesofparticularpathsofgrammaticalization.Mydiscussionthroughoutwillthereforestayatashockinglyabstractlevel.Ifthereisanyjustificationforthispaper,itisthatitpresentsasketchofadifferentperspectivethatmightbeinterestingbothtoscholarsofgrammaticalizationandtolinguistsinthegenerativetradition.Inanycase,whatfollowsismorearesearchproposalthanaresearchreport.175176KAIVONFINTEL1.FunctionalCategoriesinTheoreticalLinguisticsThedistinctionbetweenfunctionalmorphemesandcontentmorphemesisofcoursewell-establishedinmanylinguisticframeworks.Oneofthecrucialpropertiesoffunctionalmorphemesisthat,inanygivenlanguage,theirinventoryislimited,asopposedtotheopen-endedlexiconofcontentitems.Alistofsomeimportantkindsoffunctionalmorphemesmaygiveanideaofwhatwearedealingwith:(1)NounClass-Gender-Number-Determiner-Quantifier-Case-VerbClass-Voice-Aspect-Tense-Modality-Negation-Complementizer-Conjunction-‘Wh’-Elements-DegreeWords-Comparative-SuperlativeThenotionoffunctionalcategorieswasbroughtintomainstreamgenerativegrammarbyFukui(1986)andAbney(1987).Fromlookingatthelistin(1),itseemsthatonecouldessentiallywriteagrammarofawholelanguagebyjustdescribingthefunctionalmorphemesinthatlanguage.Thisintuitionisoftenframedasaprincipleofnaturallanguage:grammaticallyrelevantcross-linguisticvariationisconfinedtothepropertiesoffunctionalmorphemes(Borer1983;Fukui1986;Chomsky1991).Ofcourse,languageswilldifferintheinventoryofcontentwords,butthatseemsofcomparativelylittletheoreticalimportance.Wethusendorsethefollowingclaim:(2)Functionalcategoriesarewhatgrammarisallabout.Butwhatexactlyarefunctionalcategories?2.TwoExamplesofGrammaticalizationThereisnowaywecandojusticeheretotherichnessoftheworkbeingdoneintheframeworkofgrammaticalization.UsefulsurveyscanbefoundinHeine,Claudi&Hünnemeyer(1991)andHopper&Traugott(1993).WhatIwilldoisbrieflydiscuss(i)theoriginofauxiliariesand(ii)theoriginofdeterminersandquantifiers.Thesearetwowell-wornpathsofgrammaticalization(or“clines”asHopper&Traugottcallthem),whichIwilllaterconsiderfromaformalsemanticpointofview.1TheOriginofAuxiliaries.Auxiliaries(andultimatelyverbaffixesderivedfromauxiliaries)havetheirorigininfullverbs.Onecelebratedexampleconcernsfutureauxiliaries/inflectionsthatoriginateinvariouskindsofcomplementationstructures(Fleischmann1982;Bybee&Pagliuca1985;Pinkster1987).Romancefuturesdevelopedfromaconstructionwhereaverbofobligationembeddedaninfinitivecomplement:(3)haeccantarehabeothesesing-INFhave-1sg“Ihavethesetosing”or“Ihavetosingthese”TheFrenchfutureinflectionasinchanterai(“Iwillsing”)isdevelopedfromthiscollocation.Similarchangescanbefoundagainandagaininlanguagesaroundtheworld.1Iwillnotspendanytimeondiscussingwhethertheveryphenomenonofgrammaticalizationpresentsathreattothelegitimacyofgenerativegrammar.Itseemstomethatitdoesn’teventhoughsomeauthorsdoseetheapparentlygradualandsquishynatureofgrammaticalizationasanembarrassmentfortheclear-cutlexical/functionaldichotomybuiltintogenerativegrammar.ButseeVincent(1993),Haspelmath(1992,1994)forsomereconciliatorymoves.Similarly,anallegedthreattogenerativegrammarfromthephenomenonoficonicityisshowntobelargelyillusoryinNewmeyer(1992).THEFORMALSEMANTICSOFGRAMMATICALIZATION177TheOriginofDeterminersandQuantifiers.Determinersoftendevelopoutofparticularkindsofadjectives:definitedeterminersoriginateindemonstrativeadjectives(Greenberg1978),whileindefinitedeterminerscomefromnumeraladjectivesmeaning‘one’(Givón1981).Haspelmath(forthcoming)showsthatuniversalquantifiersoftencomefromadjectivesaswell.Iwouldliketonotesomeimportantfactsaboutsuchpathsofgrammaticalization,factsthatwillberelevantlater.First,theoriginsoffunctionalitemsdonotlieinanyoldlexicalconcepts.Thesourceitemsareusuallyfoundinfairlyrestrictedsemanticfields.Twomajorsourcesoffuturemarkershavebeenidentified:(i)agent-orientedmodalitiesofdesire,obligation,andability,and(ii)verbsofmotiontowardsagoal.Hook(1974,1991)identifie