Illustrating the prototype structures of parts of speech ...

University of New MexicoUNM Digital Repository

Linguistics ETDs Electronic Theses and Dissertations

5-1-2016

Illustrating the prototype structures of parts ofspeech: A multidimensional scaling analysisPhillip Rogers

Follow this and additional works at: https://digitalrepository.unm.edu/ling_etds

This Thesis is brought to you for free and open access by the Electronic Theses and Dissertations at UNM Digital Repository. It has been accepted forinclusion in Linguistics ETDs by an authorized administrator of UNM Digital Repository. For more information, please contact [email protected].

Recommended CitationRogers, Phillip. "Illustrating the prototype structures of parts of speech: A multidimensional scaling analysis." (2016).https://digitalrepository.unm.edu/ling_etds/28

https://digitalrepository.unm.edu?utm_source=digitalrepository.unm.edu%2Fling_etds%2F28&utm_medium=PDF&utm_campaign=PDFCoverPages

https://digitalrepository.unm.edu/ling_etds?utm_source=digitalrepository.unm.edu%2Fling_etds%2F28&utm_medium=PDF&utm_campaign=PDFCoverPages

https://digitalrepository.unm.edu/etds?utm_source=digitalrepository.unm.edu%2Fling_etds%2F28&utm_medium=PDF&utm_campaign=PDFCoverPages

https://digitalrepository.unm.edu/ling_etds?utm_source=digitalrepository.unm.edu%2Fling_etds%2F28&utm_medium=PDF&utm_campaign=PDFCoverPages

https://digitalrepository.unm.edu/ling_etds/28?utm_source=digitalrepository.unm.edu%2Fling_etds%2F28&utm_medium=PDF&utm_campaign=PDFCoverPages

mailto:[email protected]

i

PhillipGordonRogersCandidateLinguisticsDepartmentThisthesisisapproved,anditisacceptableinqualityandformforpublication:ApprovedbytheThesisCommittee: WilliamCroft,ChairpersonMelissaAxelrodIanMaddieson

ii

ILLUSTRATINGTHEPROTOTYPESTRUCTURESOFPARTSOFSPEECH:AMULTIDIMENSIONALSCALINGANALYSIS

by

PHILLIPGORDONROGERS

B.A.,HISTORY,OHIODOMINICANUNIVERSITY,2009

THESIS

SubmittedinPartialFulfillmentoftheRequirementsfortheDegreeof

MasterofArtsLinguistics

TheUniversityofNewMexicoAlbuquerque,NewMexico

May2016

iii

ACKNOWLEDGMENTS

First,Iwanttothankmyadvisor,Dr.BillCroft,forintellectualinspiration,

thoughtfulguidanceandencouragement,andremarkablepatience.Hisinfluence

permeatesallaspectsofthisthesis.

IamalsogratefulforthesupportofcommitteemembersDr.MelissaAxelrod

andDr.IanMaddieson.Theyhaveneverhesitatedtooffertheirtimeandwisdom—

evenonshortnotice—regardingthisthesisorotherresearch.

AspecialthanksgoesouttoDr.JamesandMariaMondlochfortheirtimeand

willingnesstosharetheirknowledgeofK’ichee’Mayan.Eveninthefaceofabstract

andrelativelyunnaturalelicitationprompts,theyrespondedwithpatienceand

valuableinsight.

IwouldliketoacknowledgefriendsandcolleagueswithwhomIhave

discussedthisresearchorrelatedtopics,includingDanielHieber,Matthew

Menzenski,StefanodePascale,MitchellSances,LoganSutton,MichaelWoods,and

TimZingler.

Last,andmostimportantly,IwouldliketothankmywifeChrista,my

children—Isaac,NoaRose,andMicah—andtherestofourfamilies.Timespenton

thisthesiswastimeawayfromthem.Inparticular,Christahasbeenboththeheart

andthecornerstoneofourhousehold,keepingmegroundedandourkidsinspired.

iv

ILLUSTRATINGTHEPROTOTYPESTRUCTURESOFPARTSOFSPEECH:AMULTIDIMENSIONALSCALINGANALYSIS

by

PhillipGordonRogers

B.A.,History,OhioDominicanUniversity,2009M.A.,Linguistics,UniversityofNewMexico,2016

ABSTRACT

RadicalConstructionGrammar(Croft2001)proposesthatpartsofspeech

canbeexplainedasprototypesthatemergefromtheuseofbroadsemanticclasses

ofwords—objects,properties,andactions—inbasicpropositionalactfunctionsof

discourse—reference,modification,andpredication.Thistheorypredictsthateach

ofthesebroadsemanticclasseswillbetypologicallyunmarkedinitsprototypical

propositionalactfunctionandrelativelymarkedinotherpropositionalactfunctions.

Becausethistheoryspeakstosuchabroadandfundamentalorganizationof

linguisticstructure,therichstructureoftheseprototypeshasnotbeenfully

exploredinacomprehensivemanner.Gradienceisakeycharacteristicof

prototypes(Rosch1978),anditisfoundforpartsofspeechinthecontinuumfrom

objectconceptstopropertyconceptstoactionconcepts.Thisgradience—andthe

semanticprimitivesthatmotivatethecontinuum—areexploredinmoredetail

withineachofthesebroadsemanticclassesthroughadiscussionoftheliteratureon

nounclasses,adjectiveclasses,andverbclasses.

Equippedwithalistofconceptualtargetsthatarepredictedtorepresentthe

rangeofprototypicalityforeachpartofspeech,thisthesissetsouttoillustratetheir

v

prototypestructures.Forelevengenetically,geographically,andtypologically

diverselanguages,lexemesareidentifiedtorepresenttheseconceptualtargets.The

criteriaoftypologicalmarkednessisusedtoidentifyasymmetriesinhowthese

lexemesareformallyencodedrelativetoeachotheracrossthethreepropositional

actfunctions.Thesemarkednessasymmetriesarethencodedforandillustratedina

multidimensionalscalinganalysis.

Theinsightofthespatialmodelistwofold:First,itillustratesthetrue

prototypestructuresofpartsofspeechinwhichmanyconceptsclusteratthe

prototypesandafewendupontheperipheries.Evenmore,itrevealstherelative

strengthsoftheprototypes,confirmingthehypothesisofaweakeradjective

prototype.Second,thespatialmodelshedslightonthesemanticprimitivesthat

motivatetheprototypestructures,bothinternallyandinrelationtoeachother.

vi

TABLEOFCONTENTS

LISTOFFIGURES...................................................................................................................viii

LISTOFTABLES......................................................................................................................ix

LISTOFABBREVIATIONS.......................................................................................................x

1. Introduction.......................................................................................................................1

2. Background........................................................................................................................32.1. Partsofspeechasprototypes...........................................................................................52.2. Prototypestructureandthesemanticcontinuumofpartsofspeech.................62.2.1. Objectsubclassification................................................................................................................82.2.2. Propertysubclassification.......................................................................................................102.2.3. Actionsubclassification............................................................................................................13

2.3. Typologicalmarkedness..................................................................................................152.3.1. Structuralcoding..........................................................................................................................172.3.2. Behavioralpotential...................................................................................................................182.3.3. Motivationsfortypologicalmarkedness...........................................................................21

2.4. Sourcesoftheprototypestructures............................................................................232.5. Cuttinguptheconceptualspace....................................................................................252.5.1. Semanticmaps..............................................................................................................................252.5.2. Multidimensionalscalinganalysis........................................................................................26

3. Dataandmethodology................................................................................................303.1. Selectionofthelanguagesample..................................................................................303.1.1. Geneticandgeographicdiversity.........................................................................................303.1.2. Diversityinpartsofspeechstrategies...............................................................................32

3.2. Datasources.........................................................................................................................333.3. Attestedpatternsofmarkedness..................................................................................363.3.1. Reference.........................................................................................................................................363.3.2. Modification...................................................................................................................................383.3.3. Predication......................................................................................................................................40

3.4. Codingthedata...................................................................................................................443.5. Dimensionality....................................................................................................................46

4. Methodologicalchallenges.........................................................................................48

vii

4.1. Cross-linguisticcomparison...........................................................................................484.2. Scope.......................................................................................................................................494.3. Theoreticalconstraints....................................................................................................494.3.1. Distributionalpotential.............................................................................................................494.3.2. Non-prototypicalinflectionalpotential.............................................................................49

4.4. Practicalconstraints.........................................................................................................514.4.1. Availabilityofdata.......................................................................................................................514.4.2. Generalizabilityanddetail.......................................................................................................514.4.3. Gradienceingrammaticality...................................................................................................524.4.4. Variationthatisnotcapturedinthestudyduetolistofconcepts.........................544.4.5. Precisioninmeaning..................................................................................................................55

4.5. Linguisticconsiderations................................................................................................564.5.1. Morphemecounting...................................................................................................................564.5.2. Choosingamonglexicalcandidates.....................................................................................574.5.3. Compounds.....................................................................................................................................574.5.4. Diachrony........................................................................................................................................584.5.5. Constructionalvariation...........................................................................................................604.5.6. Semanticshift................................................................................................................................624.5.7. Apparentarbitrariness..............................................................................................................66

5. Resultsanddiscussion................................................................................................685.1. Distributionoftheprototypes.......................................................................................715.2. Nounprototype...................................................................................................................755.3. Adjectiveprototype...........................................................................................................785.4. Verbprototype....................................................................................................................825.5. Interpretingthedimensions..........................................................................................865.6. Futureresearch..................................................................................................................885.7. Experience,conceptualstructure,andlinguisticform.........................................89

6. Conclusions.....................................................................................................................91

LISTOFAPPENDICES............................................................................................................93

REFERENCES........................................................................................................................117

viii

LISTOFFIGURES

Figure1.Conceptualspaceforpartsofspeech(Croft2001:92).........................................6

Figure2.Geographicdistributionofthelanguagesample....................................................30

Figure3.MDSanalysisofpartsofspeechwithpointsonly..................................................70

Figure4.MDSanalysisofpartsofspeechwithcuttinglinesonly.....................................71

Figure5.MDSanalysisofpartsofspeechwithconceptlabelsonly.................................72

Figure6.MDSanalysisofpartsofspeechwithpointsandprototypes...........................74

Figure7.Nounprototype.....................................................................................................................76

Figure8.Adjectiveprototype.............................................................................................................79

Figure9.Verbprototype......................................................................................................................83

Figure10.MDSanalysisofpartsofspeechforthree,six,nine,andelevenlanguages(clockwisefromtopleft)............................................................................................................89

ix

LISTOFTABLES

Table1.Semanticpropertiesofprototypicalpartsofspeech(Croft2001:87)............7

Table2.Objectconceptsubclassificationandselectedexemplars....................................10

Table3.Propertyconceptsubclassificationandselectedexemplars..............................13

Table4.Actionconceptsubclassificationandselectedexemplars....................................15

Table5.Prototypicalinflectionalcategoriesforeachpropositionalactfunction(Crofttoappear).............................................................................................................................20

Table6.Summarydataforthelanguagesusedinthisstudy(geneticclassificationandpopulationnumbersarefromEthnologue)...............................................................31

Table7.Selectedinsightsintothenatureofpartsofspeechfromtheauthorsofthereferencematerials.......................................................................................................................33

Table8.TensedistinctionsofMaliVerbsaccordingtoVerbClass,transitivity,andstativity(basedondataanddescriptioninStebbins2011).......................................42

Table9.AspectualdistinctionsinCreekVerbs(fromdataanddescriptioninMartin2011)...................................................................................................................................................44

Table10.Fitnessstatisticsinone,two,andthreedimensionsfortheMDSanalysisofpartsofspeech................................................................................................................................68

Table11.Semanticpropertiesofprototypicalpartsofspeech(Croft2001:87)........81

Table12.Semanticpropertiesmotivatingtheprototypestructuresofpartsofspeech.................................................................................................................................................87

x

LISTOFABBREVIATIONS1 firstperson2 secondperson3 thirdpersonACC accusativeASSOC associativeB nounclass“B”marker(Ingush)CAUS causativeCONTR contrastiveCOP copulaDEM demonstrativeDENOM denominalizerDUR durativeFOC focusFUT futureIII nounclass“III”marker(Mali)INF infinitiveIPFV imperfectiveLOC locativeM masculineNFUT non-futureNPRS non-presentNPST non-pastOBJ objectPERS personalPL pluralPOSS possessivePRED predicatingparticlePRF perfectPRS presentPST pastRDP reduplicationSG singularV nounclass“V”marker(Ingush)VENT venitive

1

1. Introduction

RadicalConstructionGrammar(Croft2001)proposesthatpartsofspeech

(noun,adjective,andverb)1canbeexplainedasprototypesthatemergefromthe

useofbroadsemanticclassesofwords—objects,properties,andactions—inbasic

propositionalactfunctionsofdiscourse—reference,modification,andpredication.

Thistheorypredictsthateachofthesebroadsemanticclasseswillbetypologically

unmarkedinitsprototypicalpropositionalactfunctionandrelativelymarkedin

otherpropositionalactfunctions.

Becausethistheoryaddressessuchabroadandfundamentalorganizationof

linguisticstructure,therichstructureoftheseprototypeshasnotbeenfully

exploredinacomprehensivemanner.Gradienceisakeycharacteristicof

prototypes(Rosch1978),anditisfoundforpartsofspeechinaconceptual

continuumfromobjectstopropertiestoactions.Thisgradiencewillbeexploredin

moredetailwithineachofthesebroadsemanticclassesthroughadiscussionofthe

literatureonnounclasses,adjectiveclasses,andverbclasses.

1Certainlabelingconventionsareusedthroughoutthispapertomakeclearthedistinctionbetweenuniversalcategoriesandlanguage-particularones(followingComrie1976;Bybee1985;Croft2001).Semanticnotionssuchaspropertiesorsemelfactives,aswellaspropositionalactfunctionssuchaspredication,willbewrittenwithalowercasefirstletter.Thesearecross-linguisticconcepts.Language-particularcategories,suchasAdjectivesorParticiples,arecapitalized.Labelsforreoccurringmorphosyntacticstrategies,suchasrelativeclausesorpredicatenominals,arelowercase,eveniftherearelanguagesthatdonotusethatparticularstrategy.Finally,thecategoriesNoun,Adjective,andVerbareoftencapitalized,followingthetheoreticalstanceofthispaperthattheyrepresentlanguage-particularcategories.However,theydoappearinlowercasewhentheyrefer—forsakeofconvenience—totheconceptual-functionalprototypes,evenforpreviousresearchwhichimplicitlyorexplicitlycharacterizesthemdifferently.

2

Equippedwithalistofconceptualtargetsthatarepredictedtorepresentthe

rangeofprototypicalityforeachofthepartsofspeech,thisthesissetsoutto

illustratetheseprototypestructures.Forelevengenetically,geographically,and

typologicallydiverselanguages,lexemeswillbeidentifiedtorepresentthese

conceptualtargets.Iwillusethecriteriaoftypologicalmarkednesstoidentify

asymmetriesinhowtheselexemesareformallyencodedrelativetoeachother

acrossthethreepropositionalactfunctions.Thesemarkednessasymmetrieswillbe

codedforandillustratedinamultidimensionalscalinganalysis.

Thegoalsofthismultidimensionalscalinganalysisaretwofold:First,the

analysisshouldconfirmthecharacteristicsoftrueprototypestructures:many

conceptsclusteredattheprototypesandafewconceptsinhabitingthe

peripheries—particularlybetweentwoprototypes.Evenmore,therelative

strengthsoftheprototypesshouldbeevidentintheanalysis.Ofparticularinterest

istheadjectiveprototype,whichhasbeenclaimedtobeweakeroreven

incomparabletothatofnounsandverbs.Second,themultidimensionalscaling

analysiscanshedlightonthesemanticprimitivesthatmotivatetheprototype

structures,asrelevantsemanticfactorswillcorrelatestronglywithparticular

dimensionsinthespatialmodel.

3

2. Background

Theexactnatureofpartsofspeechhasbeenacentraltopicoflinguistic

inquiryanddebatesinceatleastantiquity.Longbeforenativelanguagesofthe

PacificNorthwestwerepresentedasachallengetotheuniversaldistinction

betweennounsandverbs,Platofamouslydisputedtherealityof‘tableness’and

‘cupness’withDiogenestheCynic(Stanley1687).Atstakeisafundamental

organizationoflanguagestructure,andmanytermshavebeenusedforthis

phenomenonwithgreaterorlessertheoreticalimplications:partsofspeech,lexical

classes,wordclasses,grammaticalcategories,etc.InthispaperIwillmostoftenuse

partsofspeechforthesakeofconsistency;whenanyoftheothertermsareused,

theyaremeanttobesynonymous.Furthermore,partsofspeechcangenerallyrefer

toallthewordclassesinalanguage,fromlarge,lexical,openclassestosmall,

functional,closedclasses—andeverythinginbetween.Thisthesisfocusesonthe

basicpartsofspeechcommonlyreferredtoasnouns,verbs,andadjectives.Adverbs

areoftenincludedwiththesethree,andalthoughtheycanbeexplainedbythe

theorytowhichthispaperascribes,theyareexcludedinthisstudyinorderto

reasonablylimititsscope.

Itisawidely-heldbeliefthatpartsofspeechshouldbeidentifiedina

particularlanguageusingformalcriteria(Schachter&Shopen2007).Thesecriteria

includeaword’ssyntacticdistribution,itssyntacticfunction,anditsabilitytotake

varioussyntacticand/ormorphologicalinflections.Itfollowsfromthisthatpartsof

speecharelanguage-specific,andtherearecountlessexamplesintheliteratureto

illustratehowaparticularlexicalclassinonelanguagedoesnothaveexactlythe

4

samemembersasthepurportedlyequivalentlexicalclassinanotherlanguage.In

fact,thevariouscriteriawithinalanguagedonotalwaysyieldthesame

classifications,andthisleadstoimportanttheoreticalquestions(Schachter&

Shopen2007:4):Aretheretrulydiscretedistinctionsamongpartsofspeech?When

doesaclusterofgrammaticalcriteriaindicateadistinctionbetweenclasses,and

whendoesitindicateadistinctionbetweensubclasses?Whichcriteriaaremost

important?Mostanalysesofpartsofspeechinparticularlanguagesfailtoanswer

thesequestionsexplicitly,resultinginsomeamountofarbitrariness.Itseemsthat

formalcriteriaarenotsufficienttoexplainboththelanguage-particularaspectsand

cross-linguisticsimilaritiesofpartsofspeech.

Althoughthedelineationofpartsofspeechinalanguageisaccomplishedon

formalgrounds,theterminologyappliedtoalexicalclassonceithasbeenidentified

isoftenbasedonthesemanticnatureofitsmembers.Forexample,alexicalclass

identifiedinaparticularlanguagewilllikelybelabeledasNounsifmostofits

membersincludepeople,places,andthings.However,thismethodformatching

language-particularwordclasseswithunifyingsemanticlabelsisnotalways

straightforward.Inalanguagewithtwodistinctclassesofpropertywords,which

onegetsthetraditionalAdjectivelabelandwhatlabelisusedfortheotherclass?

Thesekindsofsemanticconsiderationsarenotenoughtovalidatecomparisonof

AdjectivesinonelanguagetoAdjectivesinanother.Thefactisthatlanguagesmake

differentdecisionsregardinghowmanylexicalclasses,wheretodrawtheline

betweenthem,andbywhichformalmeans.

5

Themissingcomponentinourunderstandingofpartsofspeechthusfaris

discoursefunction.HopperandThompson(1984)arguethatalllinguisticforms

lack‘categoriality’untilitisforcedonthembyuseinaparticulardiscoursefunction.

However—foreshadowingaprototypetheorythatinvokesbothdiscoursefunction

andlexicalsemantics—theywritethat“mostformsbeginwithapropensityor

predispositiontobecomeNounsorVerbs”(1984:747).Morerecentmodels(Croft

1991;Hengeveld1992)invokethepropositionalactfunctionsofreference,

modification,andpredication,whicharebasedonSearle’s(1969)workonspeech

acts.Thesepropositionalactfunctionsrepresentdistinctcommunicativegoals:

referenceservestoestablishareferent,modificationtomodifyanexistingreferent,

andpredicationtoreportonareferent’sstateofaffairs.

2.1. Partsofspeechasprototypes

Bringingtogetherpreviousinsights,Croft(1991,2001)proposesthatparts

ofspeechcanbeexplainedasprototypesthatemergefromtheuseofbroad

semanticclassesofwords—objects,properties,andactions—inthebasic

propositionalactfunctionsofdiscourse—reference,modification,andpredication.

Thistheorypredictsthateachofthesebroadsemanticclasseswillbetypologically

unmarkedinitsprototypicalpropositionalactfunction—objectsinreference,

propertiesinmodification,andactionsinpredication—andrelativelymarkedin

otherpropositionalactfunctions.Typologicalmarkednesswillbedefinedand

discussedinmoredetailinsection2.3,anditplaysanimportantroleinthe

methodologyofthispaper.Fornowitwillsufficetosaythatpatternsoftypological

6

markednessarebasedonformalcriteria,whichaswehaveseenisacritical

componentofanytheoryofpartsofspeech.

Moreover,theprototypetheoryofpartsofspeechalsoaccountsforboththe

semanticnatureofpartsofspeechandtheroleofdiscoursefunction.Figure1

featuresaconceptualspace(thistoolisdiscussedinmoredetailinsection2.5.1)for

partsofspeechthatillustrateshoweachbroadsemanticclasscanbeemployedin

eachpropositionalactfunction.Thedarkersectionsaretheprototypes:eachbroad

semanticclassisaprototypeoftheconstructionsofonepropositionalactfunction.

Figure1.Conceptualspaceforpartsofspeech(Croft2001:92)

2.2. Prototypestructureandthesemanticcontinuumofpartsofspeech

Croft’sconceptualspaceforpartsofspeechillustrateswelltheprototypes

thatresultfromcombinationsofbroadsemanticclassandpropositionalactfunction.

However,afundamentalcharacteristicofprototypestructureistheabsenceof

discreteboundaries(Rosch1978).Instead,prototypeshaveagradientdimension

alongwhichmembersofthecategorycanbeidentifiedasmoreorlessprototypical

objectreferenceOBJECTS

PROPERTIES

ACTIONS

REFERENCE MODIFICATION PREDICATION

objectmodification

objectpredication

propertypredication

propertymodification

propertyreference

actionreference

actionmodification

actionpredication

locationpredication

identitypredication

7

tothecategory.Thisgradiencearisesfromtheintersectionofmanyindividual

attributes,whichinturncanbeeitherdiscreteorgradient.Themoreprototypicala

memberisofacategory,themoreattributesithasincommonwithothermembers

ofthecategoryandthefewerattributesithasincommonwithmembersof

contrastingcategories.

Forpartsofspeech,thecontinuumfromobjectstopropertiestoactionsisa

gradientdimension.Concerningtheconceptualspaceofpartsofspeech,Croftwrites

that,“theverticaldimensioncouldbeasfinelydividedassinglewordconceptsif

necessary”(Croft2001:94).ThiscontinuumhypothesiswasfirstdevelopedbyRoss

(1972),andhasbeenrefinedbyseveralotherssince(e.g.,Comrie1975;Givón1979;

Dixon1982;Croft1991,2001).Givón(1979)introducedtheideathattime-stability

differentiatesnounsfromverbs,andthatadjectivesfallsomewhereinthemiddleof

thiscontinuum.Croft(2001)splitsthisconceptintotransitorinessandstativity,and

heincludestwoothers:relationalityandgradability.Ahypothesisofthispaperis

thatthemultidimensionalscalinganalysiswillreflecttheinfluenceofthese

semanticprimitives.Morespecifically,theyareexpectedtodifferentiateobject,

property,andactionconceptclusters.

Table1.Semanticpropertiesofprototypicalpartsofspeech(Croft2001:87)

Relationality Stativity Transitoriness GradabilityObjects nonrelational state permanent nongradableProperties relational state permanent gradableActions relational process transitory nongradable

ThesemanticprimitivesinTable1wereproposedtomotivatethe

distinctionsamongthethreebroadsemanticclasses.Butwhatabouttheinternal

8

structureofeach?Itisclearfromtheliteraturethattheseclassesarenotinternally

homogeneouseitherformallyorsemantically.Fortunately,muchhasbeenwritten

abouttheformalandsemanticvariabilitywithinthesegroups,althoughmostly

undertheterminologyofnoun,adjective,andverbclasses.Infact,different

semanticfactorsseemtoberelevanttovariabilityandprototypicalitywithineachof

theseclasses.Foreachoneinturn,Iwillbrieflyreviewthepertinentliteratureand

usetheinsightsthereintobuildalistofconceptualtargetsforuseinthisstudy.The

goalisalistofconceptualtargetsthatrepresentthefullrangeofprototypicalityon

whicheversemanticdimensionsareexpectedtoberelevant.Ifthesemanticfactors

discussedintheliteraturearetrulymotivatingprototypicalityforaparticularpart

ofspeech,theprototypestructurewillbeborneoutinthespatialmodel.

2.2.1. Objectsubclassification

Themostimportantdevelopmentforunderstandingobjectconcept

prototypicalityhasbeentheextendedanimacyhierarchyfirstformulatedby

Silverstein(1976;seealsoDixon1979;Comrie1989;Croft2003).Thishierarchy

actuallyincludesthreesemanticdimensions:person,referentiality,andanimacy

(Croft2003:130).Theroleoftheextendedanimacyhierarchycanbefoundinmany

areasofmorphosyntax.Awell-knownexampleistheinflectionofnounsfornumber

(Corbett2000).Ifalanguageinflectsanimatenounsfornumber,itwillalsoinflect

fornumberallthecategoriesthatarehigheronhierarchy(e.g.,humannounsand

pronouns),butnotnecessarilythoseloweronthehierarchy(e.g.,inanimates).Some

languagesinflectmostorallpronouns/nounsfornumber,whileothersinflectonlya

9

verylimitedsubset—eitherway,thehierarchydetermineswhichcategoriesare

includedinthatsubset.Ithasbeensuggestedthatthesesemanticfactorsinteract

withotherfactorssuchasdefinitenessandtopicality(Comrie1989);thisisalmost

certainlytrue,butitisstillunclearhowmuchoftheireffectonmorphosyntactic

patternsisdirectandhowmuchisindirect.

Theroleofpersonisfoundamongpronouns,wherefirstandsecondperson

pronounsgenerallyoutrankthirdpersonpronouns.Forpurposesofscopeandin

considerationofthemethodologicalchallengesthattheywouldpresent,Ichosenot

toincludepronounsinthisstudy.Thereisevidencethatpronounsare

“superprototypical”membersofthenounprototype,lessmarkedthaneventhe

mostprototypicalcommonnouns(Croft1991:126-8).Forexample,Toqabaqita

Pronounsmakeathree-waynumberdistinctionwhileNounscontrastonlysingular

andplural(Lichtenberk2008:325).Similarly,referentialityisrelevantin

distinguishingpronounsfrompropernounsfromcommonnouns.Likepronouns,

propernounswereleftoutofthisstudy,leavingnovariationuponwhich

referentialitymightact.

Thisleavesanimacy,whichishypothesizedforthisstudytomotivate

prototypicalityamongobjectconcepts.ThelistofobjectconceptsinTable2isbased

onwell-knownnounclasses(seeforexampleLehmann&Moravcsik2000;Croft,to

appear),anditstronglyreflectsadimensionofanimacy.Foreachclass,exemplars

werechosenfortheirlikelihoodtobefoundinallormostofthelanguagesofthis

study.Whereitwaspossibletoincludetwoexemplarsfromasingleclass,

complementssuchasmale/female,higheranimate/loweranimate,large/small,

10

manmade/natural,etc.wereincludedtomaximizevariation.Thesameprinciples

forchoosingexemplarswereappliedforthepropertyconceptandactionconcept

subclassificationsdiscussedbelow.

Table2.Objectconceptsubclassificationandselectedexemplars

Class Exemplars

Humans WOMANBOY

Kinship FATHERSISTER

Animals DOGBIRD

Plants/PlantProducts TREESEED

Artifacts HATBED

BodyParts ARMFACE

Places RIVERHOUSE

Material/Substance WATERSTONE

2.2.2. Propertysubclassification

Adjectives(andthepropertyconceptstheyprototypicallyencode)have

receivedagreatdealofattentionovertheyearsfortheircontroversialpartof

speechstatus.Somelinguistsassertthatadjectivesareuniversalandonparwith

nounsandverbsasbasicpartsofspeech,whileothersciteevidencefromparticular

languagesinwhichadjectivesappeartobeincomparableornon-existent.Whilethis

studydoesfindevidenceforthevalidityofauniversaladjectiveprototype,this

findingdoesnotnecessitatethateverylanguagehasasolitary,dedicated,andeasily

11

delineatedcategoryofAdjectives.Rather,itisclearthatlanguage-particular

Adjectivecategoriescomeindifferentshapesandsizes,sotospeak.

Whilethevariationinadjectiveclassesissubstantial,itisnotunconstrained.

Inhisfamousarticle/chapter,“Wherehavealltheadjectivesgone?”,Dixon(1982)

identifiedsubclassesofpropertyconceptsbasedontheirsemanticcharacteristics.

Hedemonstratedwithcross-linguisticdatahowsomeofthesesubclassesaremore

centraltothecross-linguisticnotionofadjective,showingupasAdjectivesevenin

languageswithsmallAdjectivewordclasses;theseareage,dimension,value,and

colorconcepts.Othersweremoreperipheral,encodedasAdjectivesonlyin

languageswithlargeAdjectiveclasses.Furthermore,hefoundtendenciesamong

theseperipheralsubclasses:physicalpropertyandspeedconceptsaremorelikely

toshowupasVerbsinalanguage(ifnotAdjectives),whilehumanpropensitywere

morelikelytobeNouns.Dixon’sfindingshavesincebeenconfirmedandhis

subclassificationrefinedbysubsequentresearch.

Therearetwostudiesofadjectivalpredicationthatwereparticularly

importantforourunderstandingofpropertyconcepts.Inhistypologyofadjectival

predication,Wetzer(1996)foundthateveninlanguageswhereAdjectivesare

distinctfromNounsandVerbs,theyshowcharacteristicsthatmakethemeither

“nouny”or“verby”.Heconceptualizedthesefindingsaslanguage-specifictreatment

ofpropertyconceptsonthenoun-verbcontinuum,whereadjectivesaresomewhere

inthemiddle.Interestingly,heconcludesthatadjectivesareverbybydefault.

Inasimilarvein,LeonStassen’sIntransitivePredication(1997)examined

formalpatternsofintransitivepredicationinalargelanguagesample.Semantically,

12

hedividedintransitivepredicationintofourtypes:event,class-membership,

locational,andproperty-concept(representingverbs,nouns,adverbs,and

adjectives,respectively).Notably,hefoundthatconceptpredicatesdonothavetheir

ownprototypicalencodingstrategy,butinsteadtakeonthestrategyofoneofthe

othertypes.Basedontimestability,heestablishedanadjectivehierarchysimilarto

Dixon’s,fromhumanpropensityatoneendtogenderandmaterialattheotherend.

Thiscontinuumcorrelateswiththefrequencyatwhichthoseconceptstakenoun-

likeversusverb-likeencodingstrategies.Furthermore,Stassenintroducesa

tensednessparameterthatpitsmandatorytenseinflectiononVerbsagainstaverby

treatmentofAdjectivesinpredication.Simplyput,alanguagewillnothaveboth,as

thatwouldresultinmandatorilyassigningtensetorelativelytime-stableproperty

conceptsforwhichtenseisnotrelevantorappropriate.Thetensednessparameter

supportsthehypothesisthattemporalsemanticprimitivesareamongthose

motivatingtheprototypestructuresofpartsofspeech.

BothWetzerandStassenfocusonthemorphosyntacticrealizationof

propertiesinpredication.Aswe’veseen,predicationisjustonepropositionalact

functioninwhichpropertyconceptscanbeused,anditisn’teventheprototypical

functionforpropertyconcepts.However,theirworkrefinedourunderstandingof

thewhereparticularpropertyconceptsfallinthesemanticcontinuum.Furthermore,

theirconsiderationofanimpressiverangeofcross-linguisticdataservedto

elucidatethevariouswaysinwhichpropertyconceptscanbeformallyencoded

relativetoobjectconceptsandactionconcepts.Whenweconsiderproperty

13

conceptsinpropositionalactfunctionsotherthanpredication,manyofthesame

patternsemerge.

Justasforobjectconcepts,eightsubclassesofpropertyconceptswere

identifiedforuseinthisstudy.ThesearefoundinTable3,withtwoexemplars

representingeachclass.Thesubclassesareorderedfrommostobject-likeatthetop

tomostaction-likeatthebottom.

Table3.Propertyconceptsubclassificationandselectedexemplars

Class Exemplars

Gender MALEFEMALE

Color WHITERED

Form ROUNDFLAT

Age OLDYOUNG

Value GOODBAD

Dimension BIGSHORT

PhysicalProperties HEAVYSOFT

HumanPropensity HAPPYANGRY

2.2.3. Actionsubclassification

Thesubclassificationofactionconcepts(‘verbclassification’,formany)has

beendominatedbyadiscussionofaspectualstructure.Vendler(1967)identified

fourverbclasses—states,activities,accomplishments,andachievements—and

demonstratedforEnglishhowcertaingrammaticalbehaviorswerepredictable

basedonaVerb’saspectualclassification.Thedistinctionsamongtheseclassescan

14

beattributedtosemanticprimitivessuchasduration,telicity,andstativity.For

example,statesaredurative,atelic,andstative,whileachievementsare

instantaneous,telic,anddynamic.Laterresearchconfirmedthevalidityofthese

aspectualclassesforotherlanguages,andrefinedtheclassification(Comrie1976;

Smith1997;Croft2012;seealsoChafe1970;Hopper&Thompson1980;Langacker

1987;interalia).Forexample,cyclicachievements(alsoknownassemelfactives)

havebeenidentifiedasachievementsforwhichtheendstateisidenticaltothe

beginningstate.Cyclicachievementscanbejuxtaposedwithdirectedachievements,

inwhichtheactionresultsinachangeofstate.Forexample,theconceptHITcanbe

classifiedasacyclicachievement,whileBREAKisclassifiedasadirectedachievement.

Theprimarysubclassificationofactionconceptsforthisstudyisbasedonthese

aspectualclasses.

BuildingontheinsightsofFillmore(1970),BethLevin’s(1993)

comprehensiveinvestigationofEnglishVerbsyieldsamorefinely-grained

classificationthantheaspectualonedescribedabove.ByidentifyingVerbswith

similarsyntacticbehavioracrossarangeofsyntacticalternations,sheisableto

identifydozensofsemanticallyandgrammaticallycoherentsubclasses(e.g.,

“prepare”verbs,“chew”verbs,etc.).Levinandothershaveextendedthesefindings

toincludelanguagesotherthanEnglish(fordiscussionofcross-linguistic

applicabilityandrelevantcontributions,seeLevin,inpress).Ofcourse,notevery

oneofLevin’sVerbclassescouldbeincludedinthisstudy.Iattemptedtoinclude

exemplarsfromasmanydifferentLevin-inspiredverbclassesaspossible,andin

particularthoseclasseswhosemembersarefrequentcross-linguistically.

15

Table4showsthesubclassificationofactionconcepts,withaspectualclasses

representedinthefirstcolumnandLevin’sverbclassesrepresentedinthesecond

column.

Table4.Actionconceptsubclassificationandselectedexemplars

Class Type Exemplars

2-ParticipantStatesCognition KNOWDesire WANTEmotion LOVE

InactiveActionsPerception SEE/LOOKATPosture SITSpatial/Possession WEAR

DurativeProcesses

SocialInteraction FIGHTConsumption EATMotion COMEChangeofState COOKCreationVerbs BUILD

CyclicAchievements(Semelfactives)

Emission COUGHContact HITBodilyMovement WAVE

DirectedAchievementsExistence DIEChangeofPossession GIVEChangeofState BREAK

2.3. Typologicalmarkedness

Theconceptoftypologicalmarkednesswasinvokedinthediscussionofparts

ofspeechasprototypes.Inthissection,Iclarifythenotionofmarkednessasitis

usedinthispaperanddiscussthecriteriathroughwhichitismanifestedin

individuallanguages.

Theterm‘markedness’wasintroducedbythePragueSchooloflinguistic

theory(Trubetzkoy1931,1939/1969;Jakobson1932/1984,1939/1984,citedin

Croft2003),andhassincebeenusedindifferentwaysbydifferentanalyticaland

theoreticaltraditions.Inthispaper,Irefertomarkednessasitappliestocross-

16

linguisticuniversals,alsoknownas‘typologicalmarkedness’(Greenberg1966;Croft

1996).Croft(2003:87)summarizesmarkednessas“asymmetricorunequal

grammaticalpropertiesofotherwiseequallinguisticelements:inflections,wordsin

wordclassesandevenparadigmsofsyntacticconstructions.”Theseasymmetries

aremanifestedinparticularlanguagesthroughdifferentcriteria,whichwillbe

describedbelowandexemplifiedinSection3.3.

Croftmakesitclearthattypologicalmarkednessis“auniversalpropertyofa

conceptualcategory”ratherthanthepropertyofaspecificlanguageorconstruction

(2003:88).Thisistruefortheapplicationofmarkednesstopartsofspeech.The

conceptualcategoriesatstakearethecombinationsofbroadsemanticclassand

propositionalactfunction,andasFigure1shows,threeofthesecombinationsare

typologicallyunmarkedrelativetotheothers.Perhapscontroversially,Iwillalso

use‘marked’/‘unmarked’terminologytorefertolanguage-specificmanifestationsof

theuniversalpattern.Theselanguage-particularasymmetriesarealsoidentifiedby

justthatterm—‘asymmetries’—aswellastheirrelevantcriteria.However,theterm

‘asymmetry’doesnotcapturethedirectionoftheasymmetry,whichisessentialto

theconceptionofmarkednessatboththelanguage-specificanduniversallevels.

Therefore,Iwilloftenrefertoalanguage-particularconstructionas‘marked’

relativetoaparallelconstruction.Itisthesenumerouslanguage-specificdirectional

asymmetriesthatcollectivelyconstitutethepatternoftypologicalmarkednessfor

thatcategoryanditscorrespondingvalues.

Furthermore,markednessisessentiallyarelationbetweentwovalues.A

constructionviewedinisolationcannotbedescribedasmarkedorunmarked;

17

rather,itcanbeidentifiedassuchrelativetoaparallelconstructionaccordingtoone

ofseveralcriteria.ItisthesecriteriathatIturntonext.

2.3.1. Structuralcoding

Onecriterionoftypologicalmarkednessisstructuralcoding,whichentailsan

asymmetryintheformalexpressionoftwoormorevaluesofacategory(Croft

2003).Thisasymmetryisoneofquantity.Forstructuralcoding,thequantityisthe

numberofmorphemesneededtoexpressthevaluesforthecategoryinquestion2.

Structuralcoding:themarkedvalueofagrammaticalcategorywillbeexpressedbyatleastasmanymorphemesasistheunmarkedvalueofthatcategory(Croft2003:92)

Zerocoding—thatis,theformalexpressionofavaluewithoutanovertmorpheme—

isacommonstrategyformanycategoriesandmanylanguages.Alternatively,values

canbeexpressedthroughovertcoding.

Itisimportanttorememberthattheasymmetryisauniversalone,whichis

notnecessarilyrealizedineachlanguage.Thephrasing“atleastasmany”inthe

definitionaboveiskey.Croftusesthecategoryofnumbertoillustratethispoint.

Singularisclaimedtobeanunmarkedvaluefornumberrelativetoplural.Indeed,

forlanguagesinwhichthereisanasymmetryinthenumberofmorphemesneeded

toexpressthesevalues,thepluralisalmostalwaysthevaluethatrequiresmore

morphemes.Yetforsomelanguages,bothvaluessingularandpluralareexpressed

equallyintermsofnumberofmorphemes—eithertheyarebothzero-coded,orboth

2Countingmorphemesanddeterminingwhichmorphemestocountarenotuncontroversialendeavors.ThereaderisreferredtoCroft(2003:ch.4)fordiscussion.

18

overtlycoded.Still,theselanguagesdonotviolatetheclaimthatsingularis

unmarkedrelativetoplural.Evenforcategoriesinwhichthereisnoasymmetryina

particularlanguage,thetypologicalmarkednesspatternremainsvalid.

2.3.2. Behavioralpotential

Thesecondcriterionoftypologicalmarkednessisbehavioralpotential,

whichreferstotheversatilityofavaluerelativetoothervaluesofacategory.This

versatilitycanbeexpressedintwoways,correspondingtotwotypesofbehavioral

potential.First,inflectionalpotentialreferstothemorphologicaldistinctionsthata

valuecanexpress;avalueisrelativelyunmarkedifitcaninflectformorecross-

cuttingcategoriesthanacomparablevalue.

Inflectionalpotential:ifthemarkedvaluehasacertainnumberofformaldistinctionsinaninflectionalparadigm,thentheunmarkedvaluewillhaveatleastasmanyformaldistinctionsinthesameparadigm(Croft2003:97)

Amoresubtleinflectionalpotentialdistinctioncanalsobemadebasedonhowthe

inflectionisaccomplishedmorphosyntactically.Fromleastmarkedtomost,the

inflectionmaybeaccomplishedbyinternalchange,affixation,orperiphrasis.These

assignmentsofmarkednessarebasedontokenfrequency(see2.3.3):suppletive

inflectionsaregenerallyfoundonlexemesofhighfrequency,affixationcoversa

widerangeoflexemefrequencies,andperiphrasticconstructionsareoften

employedforlowfrequencylexemes.

Second,distributionalpotentialreferstothesyntacticcontextsinwhicha

valuecanoccur;avalueisrelativelyunmarkedifitcanoccurinmoresyntactic

contextsthanacomparablevalue.Unfortunately,acomprehensiveinclusionof

19

distributionalpotentialisbeyondthescopeofthisstudy.Astraightforwardinstance

ofdistributionalpotentialforpartsofspeechwouldbealanguagethatdoesnot

allowpropertyconceptstobeusedinreference,forexample.However,Ifoundno

explicitevidenceforthistypeofprohibitioninanyofthelanguagesofthisstudy.3

Therearemoresubtlewaysinwhichdistributionalpotentialinteractswithpartsof

speech,buttheyarenumerousandlanguagedescriptionsarenotconsistentinthe

inclusionandorganizationofthiskindofinformation.Forthesereasons,the

analysisofmarkednesscriteriainthisstudyislimitedtostructuralcodingand

inflectionalpotential.

Asitappliestopartsofspeech,thecriterionofinflectionalpotentialmust

alsospecifywhatcategoriesofinflectionarerelevant.Thegreatestinflectional

potentialisfoundattheprototypes—prototypicalcombinationsofbroadsemantic

classandpropositionalactfunction—anddifferentinflectionalcategoriesshowup

ineachoftheprototypes.Table5liststheprototypicalinflectionalcategoriesfor

eachpropositionalactfunction.Fromtheseprototypes,inflectionalcategoriescan

extendvertically(intermsofCroft’sconceptualspaceinFigure1),applyingtothe

useoftheotherbroadsemanticclassesinthesamepropositionalactfunction;or

theycanextendhorizontally,applyingtotheuseofthesamebroadsemanticclassin

theotherpropositionalactfunctions.Becausethemethodologyofthisstudyisto

compareconcepts(representedbycorrespondinglexemesinparticularlanguages)

toeachotherinidenticalcontexts,onlytheverticalextensionsofinflectional

categoriesdescribedaboveareconsideredinthisstudy.Thecontextforcomparison

3Ifoundonlytheoccasionallackofinformationaboutwhetherornotitispossible.

20

ofconceptsisaparticularpropositionalactfunctioninaparticularlanguage,soonly

theinflectionalcategoriesprototypicaltothatpropositionalactfunctionare

relevant.Horizontalextensionsofinflectionalcategoriesareresidual;alexememay

retaininflectionalpotentialfromitsmorphosyntacticexpressionintheprototype

evenwhenitisusedinotherpropositionalactfunctions.Sincetheseinflectional

categoriesarenotequallyavailabletoallconcepts(becausetheyhavedifferent

prototypes),theyarenotappropriatemeasuresforcomparingacrossconcepts(see

also4.3.2).

Table5.Prototypicalinflectionalcategoriesforeachpropositionalactfunction(Crofttoappear)

Reference Modification Predicationsize

numberdefiniteness

deixiscase

degreeindexationlinker

voiceaspecttense

modalityevidentialityspeechactpolarityindexation

Tosummarize,markedvaluesofacategoryrequireequalormorestructural

codingandexhibitequalorlessbehavioralpotentialthantheirunmarked

counterparts.Inapartsofspeechstudy,whatarethecategoriesandvaluesrelevant

formarkedness?Thecategoryistheconceptualcontinuumwhichwasdescribedin

section2.2.Sinceitisacontinuum,intheoryitsvaluesareinfinite.Forthisstudy,

however,thevaluesaretheselectedconceptualexemplars,andtheseare

representedineachlanguagebythelexemesthatencodethoseconcepts.As

describedinthepreviousparagraph,thecontextsforcomparingthesevaluesare

21

thepropositionalactfunctions.Themarkednessasymmetriesamongtheconceptual

valuesareexpectedtobedifferentforeachpropositionalactfunction,sincea

differentsubsetofconceptsisprototypicalineachpropositionalactfunction.

2.3.3. Motivationsfortypologicalmarkedness

Typologicalmarkednesspatternsaremotivatedbybothtokenfrequency

(Greenberg1966;Bybee1985;Croft2003)andthetug-of-warbetweeneconomy

andiconicity(Croft2003:101-2).4Thestructureoflanguagerepresentsthe

structureoftheworldashumansexperienceit(iconicity).Ontheotherhand,

limitationsontimeandresourcesinhumancommunicationmotivatethe

minimizationoflanguagestructure(economy).Frequencyprovidestheasymmetry

onwhichtheseopposingforcescanact.Givenacategorywithtwovaluesdiffering

infrequency,economycanbemaximizedwhilesimultaneouslyretaininga

distinctionthatisiconicofthedistinctionintheworld:themorefrequentvaluecan

bezero-coded,whilethelessfrequentvalueremainsovertly-coded.Thisisprecisely

thepatternthatisfoundformanycategoriesinmanylanguages(butnottheonly

oneofcourse;inmanyinstances,bothorneitherofthevaluesareexpressed

overtly).Inflectionalpotentialcanbeexplainedinthesameway.Throughcross-

cuttinginflectionalcapabilities,morefrequentvaluesdisplayversatilitythatis

representativeoftheworld.Thesamecross-cuttinginflectionalcapabilitiesareless

usefulforinfrequentvalues,sotheyarelostorfailtocrystallizeinthegrammar.

4Thediscussionhereofthemotivationsfortypologicalmarkednessisacursoryone.Thereaderunfamiliarwiththisliteratureisreferredtotheworkscitedforamoredetailedaccount.

22

Theseideasaresummarizednicelyintheoft-quotedmantra,“grammarsdobest

whatspeakersdomost”(DuBois1985:363).

Thesameprinciplesareatworkshapingpartsofspeechsystemsdespitethe

countlesscomplexitiesandinterveningfactorsthatexistwithinsuchabroad

characterizationoflanguagestructure.Certainconcepts—basedontheirsemantic

make-up—lendthemselvestouseinparticulardiscoursefunctionsasadirect

reflectionofhowhumansexperiencetheworld.Becauseoftheirusefulness,these

conceptsareusedmorefrequentlyincertainpropositionalactfunctionsthanothers.

Iconicandeconomicmotivationsturnthisasymmetryinfrequencyintoan

asymmetryinform.

Nowthatitissufficientlyclearwhatismeantbytypologicalmarkedness,I

mustreiteratethatmarkednessasymmetriesarethevariationuponwhichthe

multidimensionalscalinganalysisisbased.Noteverydifferenceinformqualifiesas

amarkednessasymmetry.Considerthefollowinghypotheticalexample:onelexical

classofalanguagerequiresaparticularderivationalmonomorphemicsuffixin

reference,whileanotherclassrequiresadifferentmonomorphemicsuffixin

reference.Therelevantcriterionhereisstructuralcoding,butthesetwo

constructionshavethesamenumberofmorphemes.Inmarkednessterms,theyare

equal.Therearemanylinguisticanalysesforwhichanydifferenceinformis

significant,butonlymarkednessasymmetriesarerelevantforthisstudy.

23

2.4. Sourcesoftheprototypestructures

Theprototypestructuresofpartsofspeechemergefrommarkedness

asymmetrieswithinandacrosslanguages.

Themorphosyntacticdistinctionsmadewithinaparticularlanguageserveto

divideuplexemesintoclassesandsubclasses.Foralllanguagesinthisstudy,

morphosyntacticevidencewasfoundtomotivateatleastabasicthree-way

distinctionfortheencodingofobjects,properties,andactions.Bothstructural

codingandinflectionalpotentialplayasignificantroleinhowthesebroadsemantic

classesarerealizedineachofthethreepropositionalactfunctionsofreference,

modification,andpredication.Yet,thebasicdistinctionsmadeinjustonelanguage

arenotenoughtofullyillustrateprototypestructures.Sinceeachlexemeis

evaluatedforstructuralcodinginthreepropositionalactfunctions,thereisthe

possibilityforsomestaggeringofcategorizationlinesacrosstheseenvironments.

However,thedistinctionsthatformlexicalclassesandsubclassesinaparticular

languageareoftenfairlypersistentacrosspropositionalactfunctions.Thatis,

NounsofalanguagepatternlikeotherNounsinreference,andtheyalsopatternlike

otherNounsinmodification.ItislessfrequentthattheclassofNounsthatdisplay

particularmorphosyntacticcharacteristicsinreferenceissignificantlydifferent

fromanotherclassofNounsthatdisplayparticularmorphosyntacticcharacteristics

inmodification.

However,inflectionalpotentialplaysanimportantroleinaddingtothe

markednessstructurewithinaparticularlanguage.Evenasmorphosyntactic

evidenceisfoundtomotivatebroadlexicalclasseswithinalanguage

24

(correspondingroughlytoobjects,properties,andactions)inflectionalpotential

oftenprovidesfurtherdelineationswithineachoftheseclasses.Moreprototypical

membersofeachoftheseclassescanexhibitinflectionalcapabilitiesthatdistinguish

themfromlessprototypicalmembers.Unlikethebroadstructuralcoding

distinctionsthatarereinforcedinmorethanonepropositionalactfunction,these

inflectionalpotentialasymmetriesareoftenspecifictoaparticularpropositionalact

functionanditsrelevantcross-cuttinginflectionalcategories.Thus,evenwithina

singlelanguage,majorandminormarkednessdistinctionscanbeobserved.Even

still,fullprototypestructuresarenotreadilyapparentuntilmultiplelanguagesare

considered.

Whendatafrommanylanguagesarecompiled,thegradienceofthe

prototypestructuresemerges.Languagesmakedifferentdecisionsonwhereto

drawthelinesbetweenbothbroadlexicalclassesandsmallersubclasses.Thereisa

setoffunctionalpressuresthatareatworkinalllanguages,andthesefunctional

pressuresmotivatesimilaritiesintheirlexicalclassifications.Yetwiththousandsof

lexemes(oreventheforty-nineincludedinthisstudy),notwolanguageswillcome

topreciselythesamelexicalclassifications.Languagesmoreoftenalignintheir

classificationofsomelexemes,andtheseformtheheartsoftheprototypes;

languageslessoftenalignintheirclassificationofotherlexemes,andtheseformthe

peripheriesoftheprototypes.Itistheaccumulationofalltheselanguage-specific

decisionsfrommanydifferentlanguagesthatresultsintherichprototypestructures

illustratedinthisthesis.

25

2.5. Cuttinguptheconceptualspace

2.5.1. Semanticmaps

Animportanttooldevelopedinlinguisticsforrepresentingcross-linguistic

regularitiesinsemanticstructureisthesemanticmapmodel,exemplifiedinFigure

1.ItcametoprominencethroughearlyworkbyAnderson(1974,1982,1986,1987),

andhassincebeenusedbyvariousscholarstorepresentdatafromawiderangeof

linguisticdomains(e.g.,Croft,Shyldkrot&Kemmer1987;Kemmer1993;Stassen

1997;Haspelmath1997a,1997b,2003;vanderAuwera&Plungian1998;Croft

2003).FollowingCroft(2001:93),Iwillusetheterm‘conceptualspace’forthe

language-universalconceptualstructureand‘semanticmap’forthelanguage-

particularsemanticstructurethatcanbe‘mapped’ontotheconceptualspace.The

strengthsofaconceptualspacelieinitsabilitytocapturesimilaritiesacrossmany

languageswithoutassuminguniversallanguagecategoriesandtoillustratethose

relationshipsbeyondasimplelinearstructure(Croft2001;Haspelmath2003).

Whilesomeapplicationsofthesemanticmapmodelclaimonlyacharacterizationof

theshapeofthedata,conceptualspace—asitsnameimplies—likelyreflectsa

universalconceptualstructureinthemindsofhumanbeings(Croft2001:105,

2003:138-9).

Therelationshipbetweenconceptualspaceandmarkednesspatterns

realizedinstructuralcodingandbehavioralpotentialhasalsobeenexplicitly

framed(Croft2003:140-2).Thiswouldsuggestthatthesemanticmapmodelcould

beappliedtothepartsofspeechdatapresentedhere.However,thesemanticmap

26

modelfacessomeinherentchallenges.First,buildingaconceptualspaceisnoteasily

tractablewithlargeamountsofdata.Second,thesemanticmapmodelisunableto

dealwithexceptions.Currentresearchintypologyhasincreasinglycometo

appreciatetheprevalenceandsignificanceofstatisticaluniversals,sotheabilityto

dealwith‘messy’dataisquicklybecominganessentialrequirementoftypological

models.Third,thesemanticmapmodelisnotmathematicallyformalized.Itsnodes

andconnectionsfailtorepresentthedegreeofsimilaritybetweenrelatedconcepts.

Theexistenceofaconnectionbetweentwoconceptsiscapturedinaconceptual

space,butnotthestrengthofthatconnection.Fortunately,thereisa

mathematically-formalizedalternativetothesemanticmapmodelwhichhasbeen

adaptedforanddemonstratedonlinguisticdata.

2.5.2. Multidimensionalscalinganalysis

Multidimensionalscaling(MDS)ismultivariatemethodthatdirectlymodels

similarityindatabydistanceinaspatialmodel.Ithasonlyrecentlybeenadopted

fromrelateddisciplinesforuseingrammaticalanalysis(Croft&Poole2008;further

backgroundonMDScanbefoundinPoole2005:ch.1).MDSmodelshaveseveral

advantagesoversemanticmapmodels.MDSisamathematicallyformalized

approachthatcapturesnotonlythecategoricalrelationshipsamongdatapoints,but

alsotheirdegreesofdissimilarity.Infact,MDSdeliversatrueEuclideanmodel.This

isanadvantageofMDSoveralternativemethodsfordiscoveringstructurein

multivariatedatasets,suchasfactoranalysisandprincipalcomponentsanalysis,

27

whichonlyattempttocapturemostofthevariationinthedata(foramoredetailed

comparisonoftheseapproaches,seeCroft&Poole2008:12-13;Baayen2008:ch.5).

NotonlycanMDSbeappliedtolargedatasets,themodelactuallyimproves

withtheadditionofdata(Croft&Poole2008:31-2).Thedataonpartsofspeech

corroboratesthisclaim.Aswasmentionedabove,theinclusionofelevenlanguages

andnearlyfiftylexemesforeachlanguageturnedouttobejustenoughdatato

producerecognizableprototypestructuresandillustratetheirinternalstructures.

Thecodeddataforthisstudycomprisea49X169matrix.Bycomparison,

Haspelmath’s(2003)pronoundata(asanalyzedbyCroft&Poole[2008])forma9X

139matrixandDahl’s(1985)tenseandaspectdataforma250X1107matrix.

Finally,MDScanhandleexceptionsinthedata—thatis,itmeasuresthefitofthe

modeltothedata.IturnnowtoabriefdescriptionoftheMDSspatialmodelthat

willbeimportantforinterpretingtheresultsfromthepartsofspeechdata.

AnMDSspatialmodelconsistsofanarrangementofdatapointsandlinear

cuttinglines.Eachcuttinglinesignifiesadiscretebinarycategorizationofthedata,

inthiscaseasinglegrammaticaldistinctionfromalanguage.Thedatapoints

representtheconceptualtargetsusedinthisstudy.Asinglecuttinglinewilldivide

someportionofthedatapointsfromtherest.Thepointsononesideofthislineare

allthelexemesfromaparticularlanguagethatassumeamarkedforminaparticular

propositionalactfunctions,whilethepointsontheothersideareallthelexemes

thatassumetheunmarkedform.Itistheaccumulationofallthesecuttinglinesand

theirdiscretecategorizationsthatforcethedatapointsintoaparticularspatial

arrangementinthemodel.Specifically,thecuttinglinesformpolytopesthatcontain

28

theideallocationforeachdatapoint,butadatapointcanbeanywherewithinthat

polytope.Moredataresultinmorecuttinglines,andmorecuttinglinesresultin

moreaccuratepolytopes.Iftwopointsarecloseinproximityintheanalysis,itis

becausetheyfallonthesamesideofacuttinglinemoreoftenthannot.Iftwopoints

areverydistant,itisbecausetheyfallonoppositesidesofacuttinglinemoreoften

thannot.Inthisway,theaccumulationofbinarycategorizations(cuttinglines)

resultsinsimilarity(ordissimilarity)measuresamongallthedatapoints.Forparts

ofspeech,thistoolprovidesamorenuancedillustrationofcross-linguisticpatterns.

Likethesemanticmapmodel,anMDSanalysisoflinguisticdatais

hypothesizedtocapturemorethanjustpatternsinthedata.Akeydistinctionis

madebetweenthelanguage-andconstruction-specificnatureofthecuttinglines

andtheuniversalnatureoftheresultingconceptualspace.Thisconceptualspace

modeledbyMDSishypothesizedtobethesameforallspeakers.AccordingtoCroft

&Poole,“therelativepositionanddistanceofpointsinthespatialmodelrepresenta

conceptualorganization,presumablytheproductofhumancognitionand

interactionwiththeenvironment,thatconstrainsthestructureofgrammar”(2008:

33).Itcanthenbeinferredthatthesemanticorfunctionalcategoriesreflectedinthe

structureofthisconceptualspacearerelevanttogrammar.Inotherwords,the

dimensionsofanMDSanalysisaremeaningful,andtheyallowforatheoretical

interpretation.

ThespaceintheMDSanalysisofpartsofspeechcouldplausiblyrepresent

“conceptualdiscreteness”(Croft&Poole2008:20-1).However,basedonthe

complexityofthedataandthetheoreticallyunlimitednumberofpossiblelexemes

29

acrosslanguages,itismorelikelythatthespacerepresentsacontinuumintowhich

morelexemescouldfallshouldtheybeincludedintheanalysis.Croft&Poole(2008:

21)writethatwiththeaccumulationofmoredata,“whatatfirstappearstobea

discreteconceptualspaceisrevealedtobeamorecontinuousstructureina

conceptuallymeaningfulspatialrepresentation.”

Thespatialmodelisconstructedbasedonboundariesratherthanprototypes

(Croft&Poole2008:11),andthishasimplicationsfortheoreticalinterpretation.

Prototypestructures—centersofsimilaritygravityaroundwhichpointscluster—

mayshowupinanMDSanalysis,buttheyareanindirectproductofaccumulated

dissimilarities.Therefore,mysubsequentdiscussionsofprototypestructuresinthe

analysisofpartsofspeechdatashouldbeunderstoodinthislight.Thisfitsnicely

withthelargerconceptionoflanguageuniversalsasindirectconstraintson

grammaticalvariationratherthanuniversallinguisticcategories(Croft&Poole

2008:32).Thesemanticcontinuumfromobjecttopropertytoactionconcepts

statisticallyconstrainsthepossibilitiesforlexicalcategorizationwithinalanguage,

eventhoughnotwolanguageschooseexactlythesamecategoriesandboundaries.

30

3. Dataandmethodology

3.1. Selectionofthelanguagesample

3.1.1. Geneticandgeographicdiversity

Figure2.Geographicdistributionofthelanguagesample

Forthisstudy,datawascollectedfromelevengeneticallyandgeographically

diverselanguages.Asconcernsgeneticrelatedness,onlylanguagefamiliesthathave

beenthoroughlydemonstratedandacceptedbyamajorityofthelinguistic

communityareconsideredhere.Accordingtothisstandard,notwolanguagesused

inthisstudybelongtothesamelanguagefamily.However,morecontroversial

claimsaboutlargerlanguagegroupings,suchasAmerind(Greenberg1960,1987;

Ruhlen1991),havethepotentialtosubsumemorethanoneoftheselanguages.

Geographically,threelanguageswereselectedfromeachofthreeregions—the

31

Americas,Africa,andAustralia/Oceania—andtwofromEurasia.Withineachregion

geographicdiversitywaspreferred,withonenoteworthyexception:theinclusionof

bothMali(Baining)fromEastNewBritainandToqabaqitafromtheSolomon

Islands.Eventhoughtheyaregeneticallyunrelated,botharespokenonislandsjust

totheeastofNewGuinea.Ultimately,thecombinationofgeneticandgeographic

diversityyieldsafairlybalancedtypologicalsurvey.

Table6.Summarydataforthelanguagesusedinthisstudy(geneticclassificationandpopulationnumbersarefromEthnologue)

Region Language Classification L1SpeakersAfrica

Kanuri Nilo-Saharan 3,760,500Khwe Khoisan 6,790Kisi Niger-Congo 10,200

Americas

Creek(Muskogee) Muskogean 4,000K’ichee’ Mayan 2,330,000Quechua(Huánuco) Quechuan 40,000

Australia/Oceania

Jingulu Australian 52Mali(Baining) EastNewBritain 2,200Toqabaqita Austronesian 12,600

Eurasia English Indo-European 335,148,868Ingush NorthCaucasian 322,900

Admittedly,theavailabilityofdetailedreferencematerialswasaprimary

factorintheselectionoftheselanguages.Formostofthelanguagesinthisstudy,

descriptivegrammarsandtwo-waydictionarieswithEnglishwererequiredto

collectthenecessarygrammaticalandlexicaldata.Manyofthesewere

recommendationsfromexperiencedtypologists,andafewlanguagesoriginally

includedwerereplacedduetotheinsufficientnatureoftheirreferencematerials.

Typologicaldiversitywasconsideredintheselectionoflanguages,butnotall

typologicalinformationwasexpectedtoberelevant.Forexample,thereisno

expectedcorrelationbetweenwordorderandpartsofspeechstrategies.

32

3.1.2. Diversityinpartsofspeechstrategies

Anotherformofdiversitythatisimportantforthisparticularstudyisthatof

partsofspeechstrategies.Bythis,Iamreferringtoboththeclaimsmadeofsome

languagesthattheydonothaveallthreebasicpartsofspeech,aswellthe

generalizationsmadeaboutotherlanguagesbasedonperceivedsimilarities

betweentwoofthethreemajorpartsofspeech.Languages“withoutadjectives”are

themostcommonexampleofthefirsttype.Additionally,therearelanguagesfor

whichithasbeensuggestedthattheNoun/Verbdistinctiondoesnotexist.The

secondtypeincludeslanguageswith“VerbalAdjectives”or“AdjectivalNouns”(see

section2.2.2).ThegeneralizationhereisthatthelexicalclassofAdjectivesina

particularlanguagesharesasignificantamountofmorphosyntacticpropertieswith

eitherNounsorVerbs,insomecasestotheextentthattheymightbeconsidered

onelargelexicalclasswithtwosubclasses.Bothtypesarerepresentedinthisstudy.

Anothercross-linguisticvariableisthesizeoftheAdjectiveclass(also

discussedin2.2.2).Again,thelanguagesusedinthisstudycovertherangeof

inventorysizes,fromalanguagewithone‘true’Adjectivetootherswithverylarge

Adjectiveclasses.

Itwouldbedifficulttogroupthelanguagesofmystudydefinitivelyintothe

typesdescribedabove.Whileseveralofthegrammarauthorsfortheselanguages

makesuchclaims,thestrengthsoftheclaimsvary,asdoestheamountofevidence

thatisprovidedbytheauthorstosupporttheirclaims.Toillustratethediversityof

thelanguagesampleintermsofpartsofspeechstrategies,Ihaveincludedrelevant

33

excerptsfromthelanguageresourcesinTable7.However,itcannotbeunderstated

thatmarkednessasymmetriesofatleastonetypewerefoundtodistinguishall

threesemanticclasses(objects,properties,andactions)ineverylanguageinthe

study.Thatis,therewasnolanguageforwhichIfailedtofindmorphosyntactic

evidencetodistinguishatleastthreebasicpartsofspeechfromeachother.

Table7.Selectedinsightsintothenatureofpartsofspeechfromtheauthorsofthereferencematerials

Language PartsofSpeechDescriptionKanuri "Wordsthattranslateintootherlanguagesasadjectives…havenoovert

phonological,morphologicalorsyntacticcharacteristicsthatdifferfromthesamecharacteristicsofwordsthattranslateintootherlanguagesasnouns."(Hutchison1981:35)

Khwe "Khwedoesnothaveacoherent,properwordclassofadjectives…Itformallydistinguishesbetweenverbaladjectives,adjectivesderivedfromnouns,adjectivesrelatedtopronouns,andnumerals”(Kilian-Hatz2008:195)

Kisi "Adjectivesareaneasilydefinedclassconsistingofwordsattributiveinmeaningandshowingagreementwithnounstheymodify."(Childs1995:151)

Creek(Muskogee)

"StemsinCreekcanbedividedintoseveralclassesbasedontheirgrammaticalbehavior"(Martin2011:29)

Quechua(Huánuco)

"Thereisinsufficientevidenceofastrictlymorpho-syntacticnatureforconsideringthatnounsandadjectivesformdistinctlexicalclasses."(Weber1989:9)

Jingulu "Australianlanguagesoftendonotshowanymorphosyntacticcontrastbetweennounsandadjectives…InJingulu,however,thereappeartobesomereasonsforconsideringnounsandadjectivestobeatleastpartiallydistinctpartsofspeech."(Pensalfini2003:57)

Mali(Baining) "Itcanbedifficulttodistinguishadjectivesfromverbssinceanadjectivemayheadanintransitivepredicate."(Stebbins2011:95)

Toqabaqita Theclassofadjectivesconsistsofonemember(Lichtenberk2008b:52,230)Ingush Ingushhasthreemajorclassesofwords:nouns,verbsandmodifiers(Nichols2011:

113)

3.2. Datasources

Threedifferentmethodswereusedinthecollectionofdatafromeleven

languages.Fornineoftheelevenlanguages,datawascollectedfromgrammarand

dictionaryreferences.Asmentionedabove,thequalityofreferencematerialswas

importantintheselectionoflanguagesforthestudy.Evenso,challengesarosewith

thismethodofdatacollection,andtheyarediscussedlaterinthispaper.Asanative

34

speaker,self-elicitationandpersonalgrammaticalityjudgmentswereusedfor

English.

TheK’ichee’Mayandatainthisstudywerecollecteduniquelythrough

elicitation,soIdescribethemethodshere.MuchoftheelicitationwasfromJames

Mondloch,ProfessorofK’ichee’attheUniversityofNewMexico.Forcertain

constructions,hedeferredtohiswife,MariaMondloch,whoisanativespeakerof

thelanguage.

Themethodsofelicitationevolvedthroughoutthecollectionofthedata.

Beforeanyelicitation,abriefoverviewofthisstudywaspresentedtoDr.Mondloch.

Itwasexplainedthattheanalysiswasofparticularlexemesandhowtheselexemes

areencodedineachofthetargetedpropositionalactfunctions.Originally,example

sentenceswerepresentedinEnglishwiththelexemeappearinginallcapitalsinits

unmarkedform:

<<WHITErepresentspurity>>(reference)<<TheWHITEdogbarked>>(modification)<<ThecatWHITE>>(predication)

ThegoalofthiselicitationstrategywastoavoidabiastowardEnglishconstructions,

particularlyforthenon-prototypicalpairingsofsemanticclassandpropositionalact

function.Forexample,ifagerundformwasusedintheelicitationpromptforan

actioninreference,itmightbiasthespeakertouseananalogousforminK’ichee’at

theexpenseofotherpossiblemorphosyntacticstrategiesforthiscombination.

Ultimately,thismethodwasinsufficient.Formanycombinationsoflexeme

andnon-prototypicalpropositionalactfunction,thebare(un-derived)formofthe

lexemecreatedconfusionastothemeaningofthesentence.Toclarifytheintended

35

elicitation,thelexemeshadtobegiveninfullygrammatical(derived)forminthe

Englishprompts.ForEnglishpromptsforwhichthereismorethanonepossible

morphosyntacticstrategy,allrelevantstrategiesweregivenintheelicitation.For

example,Englishhasmultiplestrategiesforencodinganactionwordinmodification,

namelyaParticipleformandaRelativeClauseconstruction.Bothofthesewere

giveninEnglishtoalleviatebiasinelicitingthesamecombinationinK’ichee’:

<<TheFIGHTINGmanshouted>>(Participle)<<ThemanWHOFIGHTS/WHOWASFIGHTINGshouted>>(RelativeClause)

Thisrevisedelicitationstrategywasmuchmoreeffectiveandwasresponsiblefor

mostoftheK’ichee’data.

ThedifficultyoftheelicitationprocessforK’ichee’actuallysupportsthe

theoryofpartsofspeechtowhichthispaperascribes.Onthesurface,frequencycan

explainwhycertaincombinationsofsemanticclassandpropositionalactfunction

aredifficulttoartificiallyconstructfortherespondent.Themoreaparticular

constructionispracticed,themoreeffortlessandnaturalitbecomestoproduce.As

wasdiscussedinsection2.3.3,somecombinationsofsemanticclassand

propositionalactfunctionareveryuncommon.Furthermore,theelicitationofa

particularlexemeinoneoftheseuncommoncombinationsmaybeaskingthe

respondenttoproduceaformofthatlexemeneverproducedbefore.Forexample,

it’sverypossiblethatanativespeakerofK’ichee’hasneverheardorproducedthe

abstractreferentialderivedformofROUNDequivalenttothe(ratherawkward)

Englishround-ness.Therespondentwilllikelybeabletoproducesuchaformbased

onanalogywithotherlexemes,buthesitationand/ordoubtastothegrammaticality

ofthenewconstructionwouldbeunderstandable.Infact,thisreactionwaswell

36

attestedinmyelicitationoftheK’ichee’data.Formanyofthenon-prototypical

combinations,therespondentsmadecommentstotheeffectof,‘Idon’tknowifI’ve

everheardthatbefore.’

3.3. Attestedpatternsofmarkedness

Thepurposeofthissectionistopresentexamplesofasymmetryin

markednessfromthelanguagesinthisstudy.Theseasymmetries—capturedinthe

datacodedformultidimensionalscaling—arewhatdifferentiatetheconceptsinthe

spatialmodel.Theasymmetriesarerealizedinbothstructuralcodingand

inflectionalpotential.Theywillbepresentedforeachpropositionalactfunctionin

turn.Notalltheasymmetriesinmarkednessthatwerefoundinthedataare

presentedhere.Rather,thissectionfeaturesasubsetoftherelevantdatato

illustratetheobservationsofmarkednessthatwerefoundamongtheeleven

languagesofthisstudy.5

3.3.1. Reference

AnexampleofstructuralcodinginreferencecomesfromIngush.Abstract

NounscanbederivedfromAdjectiveswiththesuffix–al.Severalexamplesofthis

derivationareillustratedin(1).

5Manyexamplesareincludedinthispapertoillustrateitsfindings.Allincludeaninterlinearmorphemetranslation.Foralltheselanguages,examplesarereproducedhereusingthesamephonetic,phonemic,ororthographicscriptusedinsourcematerial.NoattemptwasmadetoconvertthedatatoastandardtranscriptionsuchastheInternationalPhoneticAlphabet.Tomyknowledge,thisdecisionbearsnoimpactonthemorphologicalandsyntacticvariablesthatareatissue.

37

(1) AbstractNounderivationinIngush(Nichols2011:158) Adjective: AbstractNoun: dika‘good’ dikal‘goodness’ d.aza‘heavy’ dazal‘heaviness,weight,value’ q’eana‘elderly’ q’eanal‘oldage’ q’uona‘young’ q’uonal‘youth’Thisderivationevenworksforthewordeghazal‘anger’fromtheverbeghaz-lu‘get

angry’,eventhoughthereisnocorrespondingadjectiveformfor‘angry’.Therefore,

withregardtostructuralcoding,membersoftheIngushAdjectiveclassaremarked

inreferencerelativetomembersoftheNounclass.

TheinflectionofnumberonSubjectMarkersinToqabaqitaillustratesthe

effectofanimacyonmarkednessofobjectsinreference.Inthethirdperson,higher

animatenounsarealmostalwaysindividuated,anddualandpluralSubjectMarkers

areused.Ontheotherhand,asingularSubjectMarkerisoftenusedforevendual

andpluralmentionsoflesseranimatesandinanimates.Thereareexceptionstothis

ruleinbothdirections.

(2) AnimacydistinctionsinSubjectMarkernumberagreementinToqabaqita(Lichtenberk2008:150)

a. botho baa ki kera oli na=mai pig that PL 3PL.NFUT return PRF=VENT ‘Thepigshavecomeback.’ b. fau neqe ki qe kuluqa stone this PL 3SG.NFUT be.heavy ‘Thesestonesareheavy.’Furthermore,pronounsarenotoftenusedwhenthereferentisnothuman

(Lichtenberk2008:72).Thesedistinctionsininflectionalpotentialserveto

disambiguatesomeobjectconceptsfromothersinToqabaqita.

38

3.3.2. Modification

InthissectionIpresentsomeexamplesofmarkednessdistinctionsin

modification.AstraightforwardexamplecomesfromKanuri,inwhichNounsare

overtlymarkedwithasuffixwhenusedinmodification.Forexample,the-másuffix

in(3)indicatesarelationshipoftrade,profession,ororigin,amongothers.

(3) KanuriNounsinmodification(Hutchison1981:196-7) kâm fə̀r-má person horse-DENOM ‘personowningahorse’

ModificationconstructionsinToqabaqitademonstratehowlabelssuchas

Noun,Adjective,andVerbdonottellthefullstoryofmarkednessdistinctions.

AccordingtoLichtenberk(2008),ToqabaqitahasonlyoneAdjective.Thisanalysisis

basedoninflectionalpotential:thissoleAdjectiveistheonlylexemethathas

differentformsbasedonnumber,animacy,andcount/massnounstatus.Asaresult,

mostpropertywordsinToqabaqitaarecategorizedasVerbs.Yet,thesemantic

distinctionbetweenpropertiesandactionswithinthisclassofVerbsisstillrealized

instructuralcoding.OnlyVerbswith"astativemeaningintheirsemanticrange"

(327)canbeusedasmodifierswithoutanyadditionalstructuralcoding.

(4) ToqabaqitaStativeVerbsinmodification(Lichtenberk2008:328) toqa maamaelia toqa suukwaqi toqa leqa peoplebe.powerful people be.strong people be.good ‘powerfulpeople,strongpeople,goodpeople’AllotherToqabaqitaVerbscanonlymodifyreferentsinarelativeclauseintroduced

bytheparticlena.Theplacementoftherelativeclausealsoservestodifferentiate

thistypeofmodificationfromthestativeVerbmodifiers.

39

Evenmore,ToqabaqitaNounscanmodifyotherNounsinagenitive

relationshipwithapossessivesuffixorsimplejuxtaposition(Lichtenberk2008b:ch.

8).

(4) Toqabaqitagenitiveconstruction(Lichtenberk2008b:385)

kaleko faalu welaclothes be.new child‘thechild’snewclothes’

Insummary,itispossibleforallToqabaqitaobjectandpropertywordsto

assumeamodificationrolewithoutanyaddedstructuralcoding.Onlythenon-

stativeactionwordsrequireadditionalstructuralcodingtodothesame.Despitethe

classificationofalmostallpropertywordsasVerbsinToqabaqita,amarkedness

distinctionremainsbetweenproperty-like(stative)Verbsandaction-like(non-

stative)Verbs.ThesamelargeVerbclassgivesa“verby”impressionoftheproperty

wordsinToqabaqita,yetitisobjectwordsthatpatternwiththepropertywordsas

unmarkedinmodificationwithregardtotheirstructuralcoding.

Havingdemonstratedhowmarkednessdifferencesmanifestinstructural

coding,Iturnnowtoinflectionalpotentialinmodification.TheAdjectiveclassin

Ingushmakesdistinctionswithinthebroadsemanticclassofpropertywordsbased

ontheirinflectionalpotentialinmodification.Onlytwopropertywords—‘big’and

‘small’—agreeinnumberwiththeIngushNoun.Furthermore,lessthantenpercent

ofAdjectivesagreeingenderwiththeNoun.

(5) NumberandgenderagreementinIngush(Nichols2011:219-21) a. v-oaqqa sag V-big man ‘oldman’

40

b. b-oaqq-ii nax B-big-PL people ‘elders’TheinflectionalcapabilitiesoftheseAdjectivessetthemapartfromothermembers

oftheIngushAdjectiveclass.

3.3.3. Predication

Structuralcodinginpredicationistypicallymanifestedintheformofa

copulaforobjectsand/orproperties.Forsomelanguages,suchasKanuri,nocopula

isnecessary.Actions,objectsandpropertiescanbepredicatedbyjuxtapositionto

thesubject.

(6) PredicationofobjectsandpropertiesinKanuri(Hutchison1981:10-1) a. Álì bàrèmà Ali farmer ‘Aliisafarmer.’ b. Fə̂r-nzé kúrà. horse-3SG.POSS big ‘Hishorseisbig.’Forotherlanguages,thecopulaisrequired.InMali,thecoordinatorda‘and’

functionsasacopulaforthepredicationofproperties.Thisconstruction

distinguishespropertiesasmoremarkedthanactionsinpredication.

(7) PredicationofpropertiesinMali(Stebbins2011:49-50)chēvicha da aululkakēvi=ka da aulul=kaCONTR=M.SG.III and tall=3SG.M.III‘Thatguyistall.’

AuniquepatternofmarkednessisfoundinJingulupredicationandit

implicatesstructuralcoding.TheJingulupredicateconsistsofamandatoryLight

41

VerbandanoptionalCoverbalRoot.Theleastmarkedlexemesinpredicationare

theLightVerbsthemselves,sincetheyoccurwithminimalstructuralcoding.The

CoverbalRootsaremoremarkedbecausetheyonlyoccurwithanadditional

morpheme—aLightVerb.

(8) VerbalelementsofJingulu(Pensalfini2003:59-60)a. ya-jiyimi bininja 3SG-come man ‘Themaniscoming.’b. ngaba-nga-ju karnarinymi have-1SG-do spear ‘Ihaveaspear.’c. ngaba-nga-rriyi karnarinymi have-1SG-will.go spear ‘I’lltakeaspear.’

TheuniquewayinwhichMaliVerbsinflectfortenseinpredication

constructionsresultsinseverallevelsofmarkedness.Theinflectionalcapabilityofa

MaliVerbdependsonbothitsclass(A-D)andwhetheritis“StativeIntransitive”or

“ActiveIntransitive”/“Transitive”(Stebbins2011:53-7).TheVerbclassesare

differentiatedbytheirphonologicalformsandtheabilityoftheverbtoinflectfor

tensedirectly:ClassAVerbsencodethreetenses,ClassBandCVerbsencodeonly

thepresent/non-presentdistinction,andClassDVerbsdonotencodeanytense

distinctions.Thestativityandtransitivitycategoriesarerelevantbecausethey

determinewhichConcordialPronounaccompaniestheVerb;StativeIntransitives

requireaClassIIIConcordialPronounwhichdoesnotencodeapast/non-past

distinctionlikeotherConcordialPronounclasses.

42

Sinceboththeabilitytoinflectandthenatureofthatinflection(ablaut,

affixation,orperiphrasis)arerelevantfactorsinmarkedness,MaliVerbsfeaturefive

levelsofmarkedness.

(9) LevelsofmarkednessintenseinflectionofMaliVerbs,fromleastmarkedtomostmarked(Stebbins2011)a. ClassA:Encodepast/present/futureonVerbb. ClassB+C(Transitive/ActiveIntransitive):Encodepresent/non-present

onVerb,andpast/non-pastonConcordialPronounc. ClassB+C(StativeIntransitive):Encodepresent/non-presentonVerb,

butnopast/non-pastdistinctionduetouseofClassIIIConcordialPronoun

d. ClassD(Transitive/ActiveIntransitive):Nopresent/non-present

distinctiononVerb,butencodepast/non-pastonConcordialPronoune. ClassD(StativeIntransitive):Nopresent/non-presentdistinctionon

Verb,andnopast/non-pastdistinctionduetouseofClassIIIConcordialPronoun

ThethirdlevelofmarkednessfeaturingClassesBandCStativeIntransitiveVerbsis

notrepresentedbyanyofthelexemesinthisstudy,buttheotherfourlevelsare

attested.Table8mapstheselevelsontotheintersectionofVerbClassand

transitivity/stativity.

Table8.TensedistinctionsofMaliVerbsaccordingtoVerbClass,transitivity,andstativity(basedondataanddescriptioninStebbins2011)

Transitive ActiveIntransitive StativeIntransitiveClassA PST/PRS/FUTClassB PST/PRS/FUT(partiallyperiphrastic) PRS/NPRSClassCClassD PST/NPST NoTenseDistinctions

Inflectionalpotentialalsoyieldsmarkednessdistinctionsinpredicationin

Creek.AspectandotherdistinctionsaremadeontheCreekVerbviaroot

43

alternations(“grades”)andsuffixes.Creekhasaprimarydistinctionbetween

temporaryevents(Non-Duratives)andstates/prolongedevents(Duratives).All

unmarkedVerbs(includesAdjectives,consideredasubclassofVerbs)inCreekrefer

toanevent,whileVerbswithaDurativesuffixcanbeeithereventiveorstative

(Martin2011:248-51).Manyoftheactionconceptsinthisstudyarespecifically

non-eventive(e.g.,KNOW,WEAR)andwouldrequirethisDurativesuffix—marked

structuralcoding—inordertoaccomplishanon-eventivemeaninginpredication.

Thiswouldindicateamarkednessdistinctionbetweeneventiveandnon-eventive

conceptsinCreek.

However,thereisanotheravailableconstructionforconceptsthatfall

somewherebetweennon-eventiveandprototypicallystative:“Verbsreferringto

dressing,knowledge,perception,andholding…commonlyoccurintheresultative

stativeaspect”(244).TheResultativeStativeiscloseinmeaningtotheDurative,but

therearesubtledifferences;theResultativeStativeisusedforeventsthatare

portrayedasshorterinduration,whiletheDurativeisusedforlongereventsor

thosethataremoreneutralinthisregard(247).TheformoftheResultativeStative

featuresafallingtone,butnosuffixliketheDurativeform.Therefore,theselexemes

areunmarkedinpredicationjustliketheeventiveconcepts.

InTable9,thecellswithboldtextaretheaspectualformsthatmatchthe

conceptualtargetsofthisstudy.Both‘know’and‘wear’representconceptsincluded

inmystudy.Whiletheyarenotconceptsincludedinmystudy,‘ripe’and‘bark’are

usedheretorepresentphysicalpropertiesandcyclicachievements,respectively;

44

thiswasamatterofconvenience,astheCreekformsfortheselexemeswerereadily

available.

Table9.AspectualdistinctionsinCreekVerbs(fromdataanddescriptioninMartin2011)

Non-DurativeEventive(LengthenedGrade)

ResultativeStative(FallingGrade)

Durative(ZeroGrade+DurativeSuffix)

‘ripe’ 'lo:kc-ís'it'sgettingripe'

lókc-i:-s'it'sripe'

‘know’ ki:ll-ís's/heislearningit'

kîll-is's/heknows'

kíll-i:-s's/heknows'or's/heisknowledgeable'

‘wear’ 'a:cc-ís's/heisputtingon'

â:cc-is's/heiswearing'

ácc-i:-s's/hehason'or's/hewould/couldwear'

‘bark’ 'wo:hk-ís'it'sbarking'

wo:hk-í:-s'itbarks(allthetime)'

3.4. Codingthedata

Themultidimensionalscalinganalysisusedinthisstudywasaccomplished

withtheMDSOCRprogram,developedbyKeithPooleandmodifiedslightlyby

JasonTimmsforuseonlinguisticdata.ThisprogramfeaturesPoole’sOptimal

Classificationalgorithm,anonparametricbinaryunfoldingalgorithmwhich

progressivelyapproximatesdatapointsandcuttinglinesuntilanoptimal

classificationisreached(Poole2000,2005).

TheMDSanalysisisconstructedfrombinarydata,andeachcolumnofdatais

responsibleforonecuttinglineintheanalysis.Eachcolumnrepresentsonelevelof

markednessattestedforatleastonelexemeinaparticularlanguageand

propositionalactfunction.Theforty-nineconceptualtargets(representedby

lexemesineachlanguage)constitutetherows,andtheyarecodedinthatcolumn

45

foroneoftwovaluesindicatingwhetherornottheyoccuratthatlevelof

markednessinthatlanguageandpropositionalactfunction.

Iwillillustratethiswithanexample.Onecolumnofmydataislabeledwith

thefollowingshorthand:“KI-R-SC-Affix”.Theshorthandcontainsfourcomponents

separatedbydashes.Thefirstcomponentisatwo-lettercodeindicatingthesource

language,inthiscaseKIforKisi.Thesecondcomponentindicatesthepropositional

actfunctioninquestion(R=reference,M=modification,P=predication).Thethird

componentindicateswhetherthemarkednessdistinctioninquestionpertainsto

structuralcoding(SC)orbehavioralpotential(BP).Thefourthcomponentcan

indicatethetypeofstructuralcoding,or—forthecriterionofinflectional

potential—thecategorybeinginflectedfor.Fortheexampleabove,“Affix”isthe

amountofstructuralcodingnecessaryforthelexemetooccurinthatpropositional

actfunction.Thereareafewcolumnlabelsthatincludeafifthcomponent,

necessitatedbydifferencesinhowinflectionalpotentialisaccomplished,eithervia

internalchange,affixation,orperiphrasis(see2.3.2).

Athirdcodingoptioncanindicatewhenthedataisnotavailable.Thismay

resultfromanyofanumberofscenarios.Forsomelanguages,aparticularlexeme

couldnotbefoundtomatchtheconceptualtarget.Thatmeantthelexemewas

assignedthisvalueacrossallthecolumnsforthatlanguage.Inothercases,the

grammarwriterwasexplicitthats/hehadnoinformationregardingacombination

oflexeme(s)andpropositionalactfunction(s).Finally,thereweresomerelevant

constructionsthatwereleftunclearorunmentionedbythereferencematerial.

46

Fortunately,theMDSalgorithmusedherecanadeptlyhandlethesemissingdata

points.

Inmanyinstances,variousinflectionalabilitiespatterntogether.Forexample,

asetofactionwordsmaytakeonseveralinflectionalabilitiesinpredication,suchas

tense,aspect,mood,etc.Ifalltheseinflectionalabilitiesapplytothesamesetof

lexemes,theyaretreatedtogetherinonecolumnofcodeasalevelofmarkedness

(whichatleastsomeotherlexemesdonotdisplay).Thiswasasteptoavoid

redundancy;representingeachinflectioninitsowncolumnwouldhaveresultedin

manyidenticalcuttinglines.

3.5. Dimensionality

OncethedataiscodedandtheMDSanalysiscomplete,thereisstillanother

theoreticaldecisiontomake.TheappropriatenumberofdimensionsforanMDS

analysisisdependentonthepropertiesofthedata.Theanalysisisperformedinone,

two,andthreedimensions,andfitnessstatisticsareprovidedtoindicatethelevelof

fitforeachnumberofdimensions.Complexdataresultinaninverserelationship

betweeninformativenessandgoodnessoffit:asmoredimensionsareincluded,

goodnessoffitincreasesbuttheinformativenessofthemodeldecreases.Thebest

modelisthelowest-dimensionalmodelthatoffersarelativelyhighdegreeoffit,and

forwhicheachadditionaldimensionoffersonlysmallimprovementsinthefit(Borg

andGroenen1997:ch.3&4,citedinCroft&Poole2008:12).Fordatainwhich

languageuniversalsarepresent,alow-dimensionalmodelwithrelativelygood

47

degreeoffitisexpected.Thelimitednumberofdimensionscanthenbeassociated

withasmallsetoffunctionalfactorsthatareresponsibleforthevariation.

48

4. Methodologicalchallenges

Beforeapresentationanddiscussionoftheresultsofthisstudy,itis

importanttobringattentiontosomeofitsmethodologicalandtheoretical

challenges.Whilethesechallengesdonottakeawayfromthesignificanceofthe

results,theyappropriatelyqualifythem.Forfuturestudiesinthisvein,thissection

offerspotentialareasforcriticalthoughtandcreativesolutions.

4.1. Cross-linguisticcomparison

Apersistentchallengeinthediscipline-widetypologicalendeavoristhatof

cross-linguisticcomparison.Infact,thetheoryoflexicalwordclassesmaintainedin

thisstudyrejectstheuniversalityofformalwordclasses.Aswehaveseen,aword

classsuchasAdjectiveinaparticularlanguageisnotequivalentorcomparableto

theAdjectivewordclassinanotherlanguage.Yetthisstudyattemptstorelatethe

formalwordclassdistinctionsmadeineachlanguage—includingvarious

constructionsandgrammaticalcategories—tothedistinctionsmadeintheother

languagesofthestudy.Thesolutionistocomparelexemesonlytootherlexemesin

thesamelanguageandthesamepropositionalactfunction.Alexemeinaparticular

languageisnevercomparedtoalexemewiththesamemeaninginanotherlanguage.

Onlythelexemeswithinalanguagearecomparedtoeachother,andmeasuresof

dissimilarityaresummedcumulativelyacrossconstructions,propositionalact

functions,andlanguages.Sooveralldissimilaritymeasuresamongconceptsarethe

cumulativedissimilaritiesofmanyintra-languageandwithin-propositionalact

functioncomparisons.

49

4.2. Scope

Theprimarylogisticalchallengeofthisstudyisitscombinationofambitious

scopeandattentiontodetail.Collectionofdatainvolvedcombingthrough

thousandsofpagesofgrammarsanddictionariestofindtherelevantinformation

fornearlyfiftylexemes,inthreepropositionalactfunctions(andfortypicallymore

thanonepossibleconstructionineachofthesecombinations),andforeleven

languages.Yetallthesewerenecessary,andtheinterestingresultsofthisstudy

wouldlikelybecompromisedifnotfortheirinclusion.Theapproximatelyfifteen

conceptsforeachbroadsemanticclassarejustenoughtoexploretheinternal

structureofeachprototype.Allthreebroadsemanticclasseshadtobeincluded,

sinceitisthebehaviorofconceptsbetweentheprototypes—relativetothoseator

neartheprototypes—thatrepresentarguablythemostinterestingdatapoints.

4.3. Theoreticalconstraints

4.3.1. Distributionalpotential

Asdiscussedinsection2.3.2,distributionalpotentialisnotcodeddirectlyin

thisstudy.However,itinfluencesothermarkednesspatternsthatareincluded.

4.3.2. Non-prototypicalinflectionalpotential

Itisimportanttoremindthereaderwhatisnotcapturedinthisstudy.In

section2.3.2,Iindicatedthatinflectionalpotentialcanextendfromtheprototypes

50

totheotherbroadsemanticclassesortootherpropositionalactfunctions.Thefirst

oftheseisincludedinthescopeofthisstudy.Thesecond—inflectionalpotential

thatisnotprototypicalforthatpropositionalactfunction—isnotincluded.

Forexample,tenseinflectionisprototypicalforpredication,andCreekisno

exception:itfeaturesafullsetoftenseinflectionsinmainclauses(actionsin

predication).Relativeclauses(actionsinmodification)havethesametense

distinctionsasmainclauses(Martin2011:398),butcomplementclauses(actionsin

reference)havefewertensedistinctions(390).Asimilarphenomenoncanbefound

inToqabaqita.Whenusedinmodification,ToqabaqitaVerbscanretainsome

inflectionalcapabilitiesprototypicalforpredication;yettherearealsosomeverbal

particlesthattheycannottakeinthisconstruction.In(10),theVerbmae‘bedead’is

modifying‘man’andisinflectedforperfectiveaspect.

(10) Toqabaqita(Lichtenberk2008b:331)kafa qeri, wane mae naqa n=ecomb this man be.dead PRF FOC=3SG.NFUTalu-lu-aown-RDP-3SG.OBJ‘Thiscomb,itwasthedeadman(lit.:thenow-deadman)whousedtoownit.’

Thesepatternsmatchwhatisexpectedbytheprototypetheoryofpartsofspeech;

asthecombinationoflexical-semanticclassandpropositionalactfunctionmove

furtherfromprototypicality,markednesswillstaythesameorincrease(inthiscase,

lostinflectionalpotential).However,thesedistinctionsarenotcapturedinthis

studybecausetenseandaspectarenotprototypicalinflectionalcategoriesfor

modificationorreference.

51

4.4. Practicalconstraints

4.4.1. Availabilityofdata

Evenwithexcellentgrammarsandextensivedictionariesatmydisposal,the

availabilityofdatawasachallengethroughoutthestudy.Afewexampleswillsuffice

toillustratethesechallenges.6InKisi,forexample,thegrammarindicatesthatthe

"o"nounclassisunmarkedonthenoun(nosuffixednounclasspronoun)

(Childs:1988:231),yetthedictionaryforKisidoesnotlistthenounclassforeach

nounentry.TheMaligrammar,despiteitsclarityandorganization,doesnotgive

anyinformationabouthowAdjectivesorVerbsareusedinreference.Some

grammarsweremoreexplicitincertaindomains;theKisireferencegrammarleft

outdiscussionofthegenitiveduetoitsintendedfocusonthephonologyand

morphologyofthelanguage.

4.4.2. Generalizabilityanddetail

Thenatureofgrammarsanddictionariesadditionallypresentstheproblem

ofcategorizationandgeneralization.Forthisstudy,itisoftentheexceptionstothe

rulesthatrepresentthemostinterestingdatapoints.Yetgrammarsanddictionaries

mustnecessarilygeneralizeinordertocaptureasmanyimportantfactsabouta

languageaspossiblewithinreasonableconstraintsoftimeandspacesetby

themselves,publishers,andpracticality.Ifoundoccasionalexamplesinmyresearch6Theseexamplesareinnowayintendedtocriticizetheauthorsofthesourcematerials.Onthecontrary,thetypologicalnatureofthisresearchisonlypossibleduetothediligenceoffieldworkersanddescriptivelinguists.Thegrammarsanddictionariesutilizedwerechosenfortheirhighquality.

52

thatindicateaninconsistencyamonglexemesassignedtothesameclassorcategory.

AninterestingexamplecomesfromK’ichee’.Thesuffixes–iiland–aaloftenmake

AbstractNounsoutofAdjectives,suchasutz-iil‘goodness’fromutz‘good’,nim-aal

‘bigness’fromniim‘big’,andk’a’n-aal‘angriness/anger’fromk’a’n‘angry’.However,

ketekik‘round’yieldsketekik-iil‘spherical’—anAdjectivejustlikeitsstembutwith

anaddedspatialdimensioninitsmeaning.Inthiscase,generalizationsregarding

thesuffixfailtocapturewhatconstitutesasignificantdistinctioninmyanalysis.

Infact,thelanguagesthataremostinterestingforthiskindofstudyarethose

thatmakevariousdistinctionsamongandwithinmajorlexicalclasses.Thosesame

languages,andinparticulartheirvariousgradations,arethemostdifficultforthe

descriptivelinguisttoaccuratelydescribewithoutextremelydetailedattentionto

theuniquepropertiesofindividualwordsandconstructions.

4.4.3. Gradienceingrammaticality

Notsurprisingly,theelicitationofK’ichee’Mayandatabroughtsomeunique

insights.Themostimportantoftheseisthegradientnatureofgrammaticality.AsI

havearguedabove,thebroadsemanticclassesarenotdistinguishedcategorically,

butthelexemesoftheseclassesfallonacontinuumbasedonsemanticprimitives.

ThiscontinuumwasevidentintheelicitationresponsesforK’ichee’.

ThesemanticcontinuumwasperhapsbestexemplifiedintheK’ichee’an

strategiesforpredicatingpropertyandobjectwords.Theresponsesandsubsequent

grammaticalityjudgmentsincludedarangeofcertaintywithmorethanacouple

“awkward”butgrammaticallyacceptableresponses.Muchoftheuncertainty

53

revolvedaroundtheuseofaree,which—inadditiontoitsotherfunctions—serves

essentiallyasacopulaintheseexamples.Generally,thecopulaisnecessaryfor

predicatingNouns,butnotforpredicatingAdjectives.However,theresponsesto

elicitationindicatethatthedistinctionisnotascategoricalasitseems.(11)shows

thepredicationofthreelexemesinK’ichee’.

(11) ObjectandpropertypredicationinK’ichee’Mayan(Mondloch,pers.comm.)a. <<ThewomanisGOOD>>

utzleixoq utz le ixoq good DEM woman ‘Thewomanisgood.’

b. <<Thealligator’shomeisRIVER>> (aree)ja’ro’chleayiin (aree) ja’ r-o’ch le ayiin

(COP) river 3.POSS-home DEM alligator ‘Thealligator’shomeistheriver.’

c. <<ThelazyoneisWOMAN>>

areeixoqleq’or aree ixoq le q’or COP woman DEM lazy(one) ‘Thelazyoneisthewoman.’Theexamplesaboveareorderedsuchthatthelexemebeingpredicatedgoesfrom

mostprototypicallyproperty-liketomostprototypicallyobject-like.While(11a)

and(11c)fitthegeneralizationforcopulausegivenabove,thecopulaareeis

optionalin(11b).

Forthemostprototypicalpropertywords,suchasexample(11a)thecopula

isnotrequired.TheseelicitationspresentednoproblemfortheK’ichee’speaker.

However,forlexemesthatfallsomewhatintermediatetothepropertyandobject

prototypes,itbecamemoredifficulttodeterminewhetherthecopulawasnecessary

54

ornot.Movingclosertotheobjectprototype,‘river’in(11b)elicitedagrammatical

sentencebothwithandwithoutthecopula;however,thissentencewithoutthe

copulawasapproachingungrammaticality.Finally,(11c)showsthatthemost

prototypicalobjectwordsrequirethecopulainpredication—itisungrammaticalto

leaveitout.Moreover,thissentencewasstillconsideredtobesomewhatawkward

evenwhenthecopulawasincluded.

Thistypeofgradienceingrammaticalityislikelypresentinsomeforminall

thelanguagesusedinthisstudy.Ifoundonlyacouplementionsofitinthe

resources,however.Asdiscussedabove,theauthorsoflanguageresourcesare

forcedtogeneralizeandcategorizeforthesakeofeconomyandreadability.Asa

result,muchofthiskindofvariationisnotrepresentedinthisstudy.

4.4.4. Variationthatisnotcapturedinthestudyduetolistofconcepts

Insomecases,thelistoflexical-semanticconceptsthatIchoseforthisstudy

resultedintheomissionoflinguisticvariation.Aclearexampleofthiscomesfrom

Toqabaqita’slonelyAdjectiveclass.ThereisonlyoneAdjectiveinToqabaqita—

‘small’—anditisnotaconceptincludedinthisstudy.Thiswordcomesbeforea

Nounandithasthreeformsthatvaryaccordingtoitsnumber,animacy,and

count/massstatus.

(12) Toqabaqita’sonlyAdjective(Lichtenberk2008:230)kasi biqu faqekwasmall house be.small‘(very)smallhouse’

55

Unfortunately,theinflectionalpotentialthatmakesthislexemeuniquein

Toqabaqitaisnotrepresentedinmyanalysis.

ToqabaqitaalsofeaturestwobasicclassesoftransitiveVerbs(Lichtenberk

2008:51).Inoneclass,onlythird-persondirectobjectsareindexedontheVerb.In

thesecondclass,alldirectobjectsareindexed.Thelexemesusedinthisstudy

includeonlytransitiveVerbsofthefirstclass,sothisdistinctioninmarkednessis

notcapturedinthedataeither.

4.4.5. Precisioninmeaning

Anotherchallengeinthecollectionofdataistheproliferationofsensesthata

wordcantakeon.Thiswasespeciallyapparentamongactionwordsinmany

languages.Mypredictionsformeaningfulconceptualdifferencesamonglexemes

includedfineaspectualdistinctions,andtheseareoftentranscendedbythesenses

ofasingleword.AcommonexampleistheconceptualtargetWEAR,whichis

supposedtorepresentaninactiveactionofpossession.However,inmanylanguages,

thisconceptisrepresentedbyawordthatalsomeans‘puton’,whichissurelynot

inactive.

Thispolysemyisnotinitselfaproblemforthemethodology.Afterall,ifthe

wordcanbeusedfortheinactivemeaning,thenitmustbeincludedinthestudy.

However,withEnglishusedasameta-language,theprecisemeaningofthewordin

thesourcelanguagecangetlostintranslation.Englishdoesnotmakeallpossible

semanticandgrammaticaldistinctions,soEnglishtranslationsdonotalwaysconvey

thesesubtledistinctionsinthesourcelanguage.IfawordistranslatedtoEnglishas

56

‘sit’,itmightrefertotheinchoativeactionofsittingdownorthestateofbeingina

sittingposition.Tobeclear,thisisnotacriticismofthesourcematerials;itwould

beunrealistictoexpectthislevelofdetailinadictionarythatwascreatedwithso

manyotherneedsandprioritiesinmind.Thisrealityjustmeansthatsome

assumptionsmustbemadealongthewayregardingtheexactmeaningsofcertain

words.Onepossiblesolutionistoelicitdatausingimagesorvideos(e.g.,Levinson&

Meira2003),butthatwasnotrealisticforthescopeandtimeframeofthisproject.

4.5. Linguisticconsiderations

4.5.1. Morphemecounting

Amethodologicalchallengethatcanbetakenforgrantedinanalyzing

differencesinstructuralcodingismorphemecounting.Itcanbedifficultto

determinewhetheranadditionalmorphemeispresentinsomecases.Asimple

examplefromKisiissufficienttoillustratethispoint.Considerthederivational

formsof(13).

(13) LexicalderivationinKisi(Childs2000) Word: LexicalClass: lùsùlló‘beheavy’ Verb

lùsùlló‘weight,heaviness’ Nounlùsùl-ɛ́í‘heavy’ Adjective

ThereisnodifferencebetweentheVerbalandNominalforms,whichfeaturea

singleóvowelattheendoftheword.TheAdjectivalform,however,replacesthisó

withatwo-vowel-ɛ́ísuffix.Forthisstudy,Iacceptthemorphologicalanalyses

57

providedbythegrammarauthors,sincetheyarelikelyinformedbytheirexpertise

inthelinguisticanalysisofthelanguage.

4.5.2. Choosingamonglexicalcandidates

Anothercommonchallengeistheexistenceofmultiplewordsinalanguage

foroneconcept.Iwasabletonarrowdownthepossibilitiesinmostcasesby

favoringwordswiththehighestfrequency(oftenoneoftheoptionsismarkedas

rareorarchaic)andthemostgeneralmeaningforthetargetconcept.Insomecases,

morethanonelexemeforaconceptwasretained.Iftheselexemesexhibited

differentmarkednesspatterns,thatdiversitywasincludedintheanalysisinthe

samewayasthediversityofasinglelexemethatcanbeusedinmultiple

constructions.Forexample,KisihasbothanAdjectivekɛ̀ndɛ̀meaning‘good’anda

Verbnàŋɔ̀ɔ́meaning‘begood’(Childs2000).Sincetheselexemesfallintodifferent

classes,theyexhibitdifferentmarkednesspatterns,andthisdiversityhas

consequencesfortheplacementofthisconceptinthemultidimensionalscaling

analysis.

4.5.3. Compounds

Compoundingandlexicalizationpresentachallengetoanysynchronic

analysis,includingthisone.Thisissueistoobroadtoincludeafulldiscussionhere,

butthechallengecanbeillustratedwithexamplesfromthedata.

InhisgrammarofCreek,JackMartinidentifiesthreedistinctconstructions

onthelexicalizationpathwayfromtheparticiplemodifyinganountothefully

58

lexicalizedcompound.Hewritesthat“wordsdescribingcolor,shape,age,orsize

commonlycombinewiththenounstheymodifyinCreek,asinmaifa-lást-i‘that

blackdog’,butthesearedistinctgrammaticallyfromcompounds”(2011:114).The

threeconstructionsareshownin(14a-c).

(14) Noun-ParticipleLexicalizationinCreek(Martin2011:125) a. Noun+Participle: ifá lást-i: dog black-DUR ‘blackdog’ b. Noun+ReducedParticiple(adjoined): ifa-lást-i dog-black-(DUR) ‘blackdog’ c. Noun+ReducedParticiple(compounded): fos-cá:t-i bird-red-(DUR) ‘cardinal’Thekeytothisanalysisissemantic:acompoundtypicallyhasafixedreferencetoa

type.ThemostlexicalizedofthesethreeconstructionsinCreekexhibitsthefixed

reference—inthisexample‘cardinal’ratherthananyredbird.

4.5.4. Diachrony

Diachronyalsopresentsachallengetothekindofsynchronicanalysissought

inthisstudyinmuchthesamewayastheothergradientphenomenadescribed

above.Withgradualprocessesoflexicalizationandgrammaticalization,thequestion

arisesastowhichhistoricalcomponentsofawordarestillrecognizabletothe

speaker.InAmericanEnglish,forexample,thewordhappyderiveshistoricallyfrom

theLatinhap(‘chance,fortune’)plusthecommonadjectivizingsuffix–y.However,a

59

contemporaryanalysislikelywouldnottreathappyasamulti-morphemic,andmost

Englishspeakerswouldnotrecognizethisinternalstructureeither.Thequestionof

whatiscognitivelyrealtospeakersisnotalwayseasytodetermine,particularly

fromagrammarordictionary.

Kanuri’stwoverbclassesareaprimaryexampleofthisdiachronicchallenge.

Class2Verbsincludearootnwhichcomesfromalexememeaning‘say,think’,and

thesemanticcomponentsoftheseClass2Verbsderivefromnon-Verballexemes.

Class1Verbsdonotincludethisroot.

(15) VerbalmorphologyinKanuri(Hutchison1981:90-1)a. Class1: búkìn bù-k-ìn eat-1SG-IPFV ‘Ieat’b. Class2: lěngîn lè-n-k-ìn go-say/think-1SG-IPFV ‘Igo’

ItisnotclearwhetherthemodernspeakersofKanurirecognizetherootnasa

distinctmorphemeoraspartofalargerrootwiththesemanticcomponent.The

analysisofmarkednessdependsonthisdetail,sinceClass2Verbshaveextra

structuralcodingifthisrootisanalyzedasseparate.Forcasessuchasthese,Idefer

tothemorphologicalanalysisofthegrammar.Ifthegrammarpresentsthese

componentsasseparatemorphemes,thenItreatthemassuch.Althoughgrammar

writersmaybebiasedtowardanatleastpartiallyetymologicalanalysisof

60

morphologybecauseitoffersadeeperlevelofexplanation,Iamnotinapositionto

determinewhetherthisisactuallythecaseandforwhichanalyses.

4.5.5. Constructionalvariation

Languagesoftenpresentmultiplesolutionstothesameproblem.Thisistrue

alsoofconstructionalpossibilitiesforagivenlexemeinaparticularpropositional

actfunction—thereisalltoooftenmorethanoneconstructionavailable.These

possibleconstructionsmaybeequallyavailabletoalllexemesofacertainlexical

classorrestrictedtoacertainsubset.Inthesecases,theconstructionalpossibilities

affordedaparticularlexemearerelevantforthisstudy,especiallywhenone

constructionismoreorlessmarkedthantheother.Theyareincludedwhenthe

informationabouttheirapplicabilityisdescribedinsufficientdetail.Themethodfor

codingdatainthisstudyallowsaparticularlexemetobeanalyzedashavingmore

thanonelevelofmarkedness.Infact,thesecasescanresultinaricheranalysis.Each

lexeme/propositionalactfunctioncombinationwithmorethanoneconstructional

possibility—onemoremarkedandonelessmarked—exhibitsbothsimilarityand

dissimilaritywithalexemethatisalwayslessmarked,andexhibitsbothsimilarity

anddissimilaritywithalexemethatisalwaysmoremarked.Thusjusttwolevelsof

markednesscanyieldthreedifferentlexicalclasseswhenonelexicalclasscanbe

usedinbothconstructions.

Insomecases,morethanoneconstructionispossiblebuttherearestrong

correlationsbetweentheconstructionandanotherpredictivefactor.Anexampleof

thiswaspresentedin(2)forToqabaqita.Higher-animatesubjectsusuallytrigger

61

numberagreementintheSubjectMarker,whileinanimateandlower-animate

subjectsarealmostalwaysusedwiththesingularSubjectMarkerregardlessoftheir

number(Lichtenberk2008:51).Thisisjustageneralization,andthereare

exceptionsforbothgroups.Again,forcaseslikethese,Idefertotheanalysis

providedinthegrammar.Thatis,ifthegeneralizationisstrongenoughthatthe

grammarwriterisabletoidentifyaboundary,thenIincludethatdistinctioninmy

analysis.

Someconstructionalalternativesarebasedonotherfactors,suchasthe

particularrelationshipofthatlexemetotheconstructionoranotherelementtherein.

Forexample,Toqabaqitafeaturesseveralpossibleconstructionsforobjectwordsin

modification.Theserepresentrelationshipssuchasgenitive,possession,

part/whole,andassociation(16a-c).

(16) Noun-NounmodificationconstructionsinToqabaqita(Lichtenberk2008b:385,379,408)a. Juxtaposition:

rua araqi loo work mature.man upward ‘God’swork’

b. Personalsuffix:

gwero-na kuukua crest-3.PERS chicken ‘a/thechicken’screst’

c. Associativesuffix:

biqu-i dudualinga house-ASSOC bee ‘bee’snest/beehive’

62

Itisnotcleariftheserelationshipsarerestrictedtoparticularlexemesorsubclasses

ofobjectwords.Thisstudydoesnotaimtocapturemarkednessdifferencesamong

thesekindsofmodifyingrelationships.

4.5.6. Semanticshift

AclassicexampleofsemanticshiftinmydatacanbefoundinQuechua.

Whenpropertywordsareusedinreference,theycanappearwithoutanystructural

coding.However,thereisarecognizableshiftinmeaning.

(17) ObjectsandpropertiesinreferenceinQuechua(Weber1989:35-6)

a. rumi-ta rikaa stone-ACC I.see ‘Iseea/thestone.’b. hatun-ta rikaa big-ACC I.see ‘Iseea/thebig(one).’

Inthesecondexample,theresultingconstructiondoesnotrefertothepropertyof

‘bigness’,butrathersomeoneorsomethingthatexhibitsthatproperty.Although

thispropertywordappearswithoutanystructuralcoding,ithastakenonanobject

meaninginthisconstruction(Croft1991:72-74,2001:70-5).Thisparticular

semanticshiftforpropertywordsinmodificationisverycommoncross-

linguistically.Anyinstanceofsemanticshiftthatchangesthebroadsemanticclassof

alexeme(e.g.,frompropertytoobject)isconsideredtoomuchsemanticshiftinthis

study.

Anotherexampleofunacceptablesemanticshiftisfoundforactionsin

reference.Someactionconcepts,suchasCOUGHandFIGHT,appearasNounsin

63

severallanguages.AsNouns,however,theyrefertoaproductoftheeventrather

thantheeventitself(Croft,pers.comm.).ForFIGHT—likeDANCE—theproductisthe

structureoftheaction.Foremissionevents,theoftenephemeralproductiswhatis

emitted:sound,light,orsubstance.Theseproductsallrepresentnon-prototypical

objectsratherthanderivedactions.

Forsomelanguages,specificderivationalmorphemesencodethe

relationshipbetweenthestemandthederivedlexeme.Thisisatleastgenerallytrue

forthenominalizationofQuechua’sVerbs.Thesuffix–qisan“Agentive

Substantivizer”,bywhichtheresultingNounreferstotheagentoftheaction

denotedbythesourceVerb(Weber1989:53).Similarly,thesuffix-nacreatesa

NounthatreferstothetoolusedtodotheactiondenotedbythesourceVerb,such

asthederivationof‘broom’fromtheVerbmeaning‘sweep’(50-1).Neitherofthese

arerelevantforthisstudy,sincetheyshiftthereferentfromanactiontoanobject.A

thirdQuechuanominalizeristhesuffix–y,an“InfinitiveSubstantivizer”(51-3).

WhenattachedtosomeVerbs,theresultingNounappearstomaintainitsaction

meaning.However,otherexamplesindicateasignificantsemanticshiftonparwith

theotherQuechuanominalizersdiscussedabove.Stillothersareglossedinsucha

waytobesomewhatambiguousastothedegreeofsemanticshift.

(18) InfinitiveSubstantivizingsuffixinQuechua(Weber1989:51-3) a. miku-y eat-INF ‘food’ b. kanan papel-niki-kuna tinku-chi-y-chaw lloqshi-nki alli now paper-2.POSS-PL equal-CAUS-INF-LOC come.out-2 good ‘Nowin(thecircumstanceof)seeingifyourpapermeasuresup,youwill

comeoutgood…’

64

c. aqcha rutu-y ka-n hair cut-INF be-3 ‘Thereishair-cuttinggoingon.’Becauseitisimpossibletodeterminefromtheavailabledatahowthissuffix

interactswitheachlexemeinthisstudy,thisinformationisleftoutoftheanalysis.

Kanurifeaturesasimilarsemanticambiguityforpropertywordsin

reference:

Thewordcìmêforexamplemaybeusedasamodifiermeaning‘red’,orasanounmeaning‘theredone’,oreventomean‘redness’,thoughforthelattermeaningthereisalsothealternativederivednə̀m-cìmê‘redness’formedthroughtheaffixationoftheabstractnominalprefix…Thewordkúràmeans‘big’,‘important’asamodifier,andeither‘boss’,‘leader’,‘head’or‘thebigone’,‘theimportantone’,whenusedasanindependenthead…AnyofthewordsthathavetraditionallybeentranslatedasandreferredtoasAdjectivesinKanuricanfunctionasnounphraseheads,referringbacktoananaphoricallyunderstoodnounphraseantecedent.(Hutchison1981:36)

Thesemanticallyshiftedmeaningfrompropertytoobjectisgeneralizabletoall

membersoftheKanuriAdjectiveclass.However,itisunclearifallofthemcan

renderapropertymeaninginreference,asisdescribedfor‘red’.Sincethelatter

meaningisthetargetofthisstudy,somefactsabouthowKanuri’sAdjectivesare

usedinreferencearesimplymissingfromtheanalysis.

Theexamplespresentedaboveshouldmakeitclearthataprimarychallenge

totheinclusionofderivationalmorphologyinmyanalysisisthevariabilityof

semanticeffectofsomederivationalmorphemes.Combinedwithdifferentlexemes,

thesamemorphemecanoftenderiveresultswithcriticallydifferentlevelsof

semanticshift.Thisvariabilitymakesitverydifficultforthegrammarwriterto

makeusefulgeneralizationswithoutcompromisingsomeofthedetailsthatareso

crucialforthiskindofstudy.

65

Thistopicalsohasimplicationsfortheoverallprototypestructuresofparts

ofspeech.Iwilldemonstrateinlatersectionshowthefullrangeofmarkednessdata

supportsmanyofmyhypothesesaboutwhichsemanticcriteriaaremostinfluential

inshapingtheprototypestructures.Someofthesearesemanticfeaturesthatexert

theirinfluenceattheheartsoftheprototypes.Forexample,theanimacyhierarchy

mostcommonlycreatesdistinctionsininflectionalpotentialwithintheprototypical

combinationofobjectsinreference.Thesederivationalconstructions,however,are

primarilyaffectingstructuralcodinginnon-prototypicalcombinationsofsemantic

classandpropositionalactfunction.Interestingly,thesamesemanticcriteriathat

I’vepredictedtoberelevantattheprototypesandoverallmaybedifferentfromthe

criteriamostrelevanttominordistinctionsinthe“noman’slands”wherethese

derivationalconstructionsareatwork.Thisisbecausesemanticshiftbecomesthe

dominatingfactorinthesenon-prototypicalconstructions.

Whenthesemanticclassdoesnot“match”thepropositionalactfunction,

thereisatug-of-warbetweenthesemanticcontributionofthelexemeandthe

semanticcontributionoftheconstruction.Thisiswhy,forexample,propertywords

inreferencesooftensemanticallyshifttorefertoapersonorthingthatexhibitsthat

property.Thenewlyformedobjectwordfitscomfortablyinreference,asobjects

generallydo.Thenaturalfitandfrequencyofthissemanticallyshiftedderivational

constructionmaycoercealanguageintoderivingadifferentconstructionforthe

less-commonlyusedreferenceofthepropertyitself,andthatalternative

constructionmaybemarkedrelativetothederivationinvolvingsemanticshift.For

thoselanguagesinwhichthederivationalsemanticsareinconsistentandlexically

66

specific,markednesspatternscouldboildowntohowpredictableasemanticshiftis

forthatlexicalconcept.Asdiscussedabove,someactionconceptsrefertoanevent

ofemissionwithanidentifiableproduct.Theproductmaybeanobvioustargetfor

semanticshiftwhenthatlexemeisusedinreference,whileotheractionconcepts

withoutsuchanobvioustargetmayresistasimilarsemanticshift.Thiswouldresult

indifferencesinmarkednessfortheseactionsinreference.Thesemantic

characteristicsthatdeterminewhetheranactionhasanobvioustargetforsemantic

shiftmaybedifferentfromtheaspectualcharacteristicsthatwerehypothesizedto

berelevantforthisstudy.Thiscouldbeinvestigatedwithmoredetailedlanguage

data,particularlyforthoselanguagesinwhichonlysomeactionwordsundergo

semanticshiftwhenusedinreference.

4.5.7. Apparentarbitrariness

Theexamplesfromthedataandthediscussionofchallengestothe

methodologyhaveillustratedthatmarkednessdifferencesariseoutofthe

confluenceofseveralfactors.Historicaldevelopmentsinteractwithfrequency

effectsandprocessesofmorphosyntacticreanalysisandphonologicalfusion.A

comprehensiveunderstandingofthesefactorswouldrequireconsiderationofthe

uniquecircumstancesineachlanguageandintheevolutionofeachconstruction.

Forexample,CreekVerbs(thisclassincludesAdjectives)indexthenumberoftheir

subject(intransitive)orobject(transitive)inthreedistinctways(Martin2011:197).

AsmallsetofcommonVerbs(e.g.,‘sit’,‘come’,‘die’,‘break’)achievethisthrough

suppletion.Alternatively,reduplicationaccomplishesthisinflectionformanystative

67

Verbroots(suchasAdjectives—e.g.,‘white’,‘soft’).StillanothersubsetofVerbs(e.g.,

‘red’)utilizeasuffix.Thesedifferentstrategiesobviouslyreflecttheinfluencesof

frequencyandstativity,butthefullstoryismuchmorecomplex.Howdoes‘red’end

upinadifferentclassthan‘white’?Atleastonthesurface,frequencyandstativitydo

notappeartoanswerthisquestion.

InKanuri,seeminglyminordifferencesinthephonologicalformofaVerb

turnouttohaveamajorimpactonmarkedness(Hutchison1981:157).Verbal

Nouns(basicallyinfinitives)canbederivedfromClass1KanuriVerbsintwo

differentways.ForCVrootsfeaturinganyvowelexcept/i/,atonechange

accomplishesthisderivation(e.g.,tá->ta).Incontrast,thesuffix–oisaddedto

derivetheVerbalNounforalmostallCVrootswith/i/,CVCroots,andpolysyllabic

roots(e.g.,dí->di-o,lad->lad-o).Thissuffixisadditionalstructuralcodinginmy

analysis,andtheresultingVerbalNounformismarkedrelativetothoseVerbal

Nounsderivedfromonlyachangeintone.ThevowelofaKanuriCVrootisnottruly

random,sinceitcomesfromaparticularsetofhistoricalprocessesatworksinceits

inception.Yetitseemstobeafairlytrivialcharacteristicoftherootfromwhicha

significantmarkednessdistinctionhasarisen.

Theseexamplesareillustrativeofthemyriadoffactorsmotivatingthe

prototypestructuresofpartsofspeechandthecomplexrelationshipsamongthese

factors.Thatarecognizablyconstrainedpatternemergesfromthesecomplexitiesis

atestamenttothestrengthanduniversalityofthefunctionalpressuresmotivating

partsofspeech.

68

5. Resultsanddiscussion

Themultidimensionalscalinganalysiswassuccessfulinrepresentingmostof

thevariationinthedatainonlytwoorthreedimensions.

Table10.Fitnessstatisticsinone,two,andthreedimensionsfortheMDSanalysisofpartsofspeech

Dimensions Classification APRE1 0.94030 0.792402 0.98679 0.954063 0.99212 0.97261

Asdiscussedabove,thebestmodelyieldsahighdegreeoffit,andforwhicheach

additionaldimensionoffersonlysmallimprovementstothefit.Forthepartsof

speechdata,atwo-dimensionalmodelclearlyfitsthisdescription.Theclassification

andespeciallytheaggregateproportionalreductionoferrorshowasignificant

improvementfromonedimensiontotwo.However,fromtwotothreedimensions,

thesenumbersonlyshowsmallimprovement.Thetwo-dimensionalmodel

demonstratesanexcellentfitwhileremainingveryinformative—itcharacterizesa

largedatasetwithverylowdimensionality.

AcommonresultofanMDSanalysisintwodimensionsisahorseshoeshape

(BorgandGroenen1997,citedinCroft&Poole2008:17),andthisshapewasfound

forthepartsofspeechdata.Thedataarearrangedlinearly,andthatlineisbentinto

ahorseshoeshapeallowingstraightcuttinglinestoisolatesmallerstringsofdata

thatfallalongtheline.Thelinearnatureofthisdatacorrespondstothesemantic

continuumfromobjectstopropertiestoactions.Ifsomegrammaticalfunctions

applytoasubsetofthedatapoints(lexemes)inthemiddleofthelinear

arrangement—suchasanAdjectivecategoryinalanguagethatincludesproperties

69

butexcludesobjectsandactions—thelinemustbecurvedtoallowastraightcutting

linetoseparatethosepointsfromtherest.Figure3showsthatthehorseshoeshape

pointsupward;i.e.,theopenendofthehorseshoeisatthetopofthefigure.For

moreevenlydistributeddata,pointswillappearallalongthehorseshoeshapewith

theendsdrawingsomewhatneartoeachother.Becausethepartsofspeechdataare

dominatedbythreeclusterscorrespondingtoNouns,Adjectives,andVerbs,the

endsofthehorseshoeshapedonotcometogetherinthisway.Anopenendstill

facesupward,butthepointsformmoreofanopen,“U”shape.

Thereisanotherimportantcharacteristicofthishorseshoeshapethatwillbe

relevantinthediscussionbelow.Althoughthedatapointsaregenerallyalignedin

theshapeofthehorseshoe,somepointsendupclosertotheinsideofthehorseshoe

andsomeendupalongtheouteredgeofthehorseshoe.Thisisnotanaccident,but

ratheranemergentcharacteristicofthespatialmodel.Thepointsalongtheoutside

ofthehorseshoecanmoreeasilybeisolatedfromtherestbyacuttingline,while

pointsalongtheinsideofthehorseshoecannot.Therefore,ifgrammatical

distinctionsseparateoneparticularconceptualtargetfromtherest,thatpointis

likelytoshowupalongtheoutsideofthehorseshoe.Conceptualtargetsthatare

neverorless-oftenindividuatedinthedatawilllikelyendupalongtheinsideofthe

horseshoe.Inthisway,thepositionofapointalongtheinsideoroutsideofthe

horseshoeistheoreticallysignificant,notjustitspositionalongthecurvedlinefrom

oneendofthehorseshoetotheother.

70

Figure3.MDSanalysisofpartsofspeechwithpointsonly

71

5.1. Distributionoftheprototypes

TheprototypestructuresofpartsofspeechcorrespondingtoNouns,

Adjectives,andVerbsareimmediatelyclearinthespatialmodel.Before

investigatingeachoftheseinmoredetail,Ifirstpresentsomeobservationsthatmay

seembasicbutneverthelesshavetheoreticalimportance.

Figure4.MDSanalysisofpartsofspeechwithcuttinglinesonly

72

Figure5.MDSanalysisofpartsofspeechwithconceptlabelsonly

Therearesomepointsthataredirectlyoverlapping,meaningtherewereno

distinctionsamongtheminthedata.Yetmanyofthedatapointsaredifferentiated

intheanalysis.Whyisthisimportant?Firstofall,thereweremanyconceptual

targetsevenwithineachbroadsemanticclass.Eachofthesebroadclasses—objects,

properties,andactions—includedatleast16targets.Someoftheconceptualtargets

fallintothesamesubclass,suchastheageconceptsYOUNGandOLD.Yetalmostallof

theseconceptualtargetsweredifferentiatedfromeachotherbyamarkedness

73

distinctioninatleastoneoftheelevenlanguages.Languagesexhibitconsiderable

variationregardingwheretheymakemeaningful,grammaticaldistinctionsalong

theobjects-properties-actionssemanticcontinuum.

Despitethedifferentiationofnearlyeverydatapoint,thebroadclassification

oftheseconceptsintoobjects,properties,andactionsissupportedinthespatial

model.Therearegreaterdissimilaritiesacrossthesebroadclassesthanwithineach

one.Everylanguageinthestudymakessignificantmarkednessdistinctions

betweenmostobjectsandmostproperties,andbetweenmostpropertiesandmost

actions.Languagesmakesomedifferentdecisionsintheirclassificationof

conceptualtargetsthatareclosetotheboundariesbetweenthesebroadsemantic

classes,buttheyallmakeatleastonemajordistinctionnearthoseboundaries.In

fact,thisistheprimaryreasonforanalyzingeachofthethreeclustersasseparate

prototypes.

Inowturntotheprototypesthemselves.First,strongnounandverb

prototypestructurescanbeobservedintheupperrightandupperleftregionsof

theanalysis,respectively.Asdiscussedinsection(2.6.2.),thespatialmodelisbased

ondissimilaritydata,soprototypeclustersareanindirectresultofgreaterand

lesserlevelsofdissimilarityamongpoints.Thediscussionhereof“stronger”

prototypesreferstodatapointclusterswhichexhibitlessdissimilarity,resultingin

pointsthatareclosertogetherintheanalysis.Inadditiontothenounandverb

prototypes,aweakeradjectiveprototypeisobservableinthelower,middleregion

oftheanalysis.Thesedatapointsrepresentpropertyconcepts,andwhilethey

74

exhibitmoredissimilarityamongthemthanisfoundintheotherprototypes,they

maintainsomeamountofcohesionaswell.TheseclustersareillustratedinFigure6.

Figure6.MDSanalysisofpartsofspeechwithpointsandprototypes

Inthenextsections,Iturntoadiscussionoftheinternalstructureofeach

prototype.Thearrangementofpointsinthespatialmodelwillbeevaluatedto

determinewhethertheysubstantiatemypredictionsregardingtherelevant

semanticcharacteristicsforprototypicality.

AdjectivePrototype

VerbPrototype Noun

Prototype

75

5.2. Nounprototype

Thenounprototypeisprobablythemoststraightforwardinboththespatial

arrangementofitsdatapointsandthesemanticmotivationforthisarrangement.

Figure7offersacloserlookatthedatapoints(representedbytheirlabels)that

constitutethenounprototype.Allthesepointsrepresentobjectconcepts,andthey

nearlyformastraightlineinasteep,positive-slopingdiagonal.Recallthatacurved

lineisnecessaryforcuttinglinestoseparatemedialsegmentsfromtherest.The

objectconceptsaregenerallyalignedonastraightlinebecausethereisoneprimary

functionalmotivationthatdominatestheirarrangement:animacy.

Animacywasresponsibleforseveralmarkednessdistinctionsacrossmore

thanonelanguageinthisstudy.Theseincludedhuman/non-human,higher-

animate/lower-animate,andanimate/inanimatedistinctions.Asaresult,the

uppermostpointsinthisclusterofthespatialmodelaretheobjectconceptsthatare

mostanimate.Allfourhumanconceptsarefoundatthisextremity,andtheseare

followedbynon-humananimates—firstDOGandthenBIRD.WOMANwasalso

differentiatedfromBOYeventhoughtheyarebothhumananimateconcepts.For

example,theEnglishablautpluralformwomenislessmarkedthanthesuffixed

pluralformboy-s.

Atthelowerendoftheobjectscluster,STONEandWATERarestretchedoutin

thedirectionoftheadjectiveprototype.LexemesforSTONEpatternedwithproperty

wordsinseverallanguages,whichexplainsitsspatialorientation.WATERisunableto

inflectfornumberinKisiduetoitsstatusasamassnoun,distinguishingitfrom

76

otherobjectconceptsinthisanalysis.Thissuggeststhatunboundedobjectsare

mostdistantfromthenounprototypeamongobjectconcepts.

Figure7.Nounprototype

Betweenthesetwoconceptsonthelowerendandtheanimatesontheupper

end,asignificantnumberofconceptslandedinthesamespotinthespatialmodel.

ThisgroupincludesTREE,SEED,HAT,BED,ARM,FACE,RIVER,andHOUSE.Nomarkedness

distinctionswerefoundtodisambiguatethesetermsinmyanalysis.The

explanationmaybefoundinthesemanticfeaturesthatwereusedtosubclassify

77

theseobjectconceptsinthefirstplace.Theseconceptsrepresentseveralsubclasses:

plants(andplantproducts),artifacts,bodyparts,andplaces.Theyarenotasreadily

rankedintermsofanimacy.Bodyparts(includespartsofhumans,animals,and

plants)areindirectlybasedonanimacybytheobjects’relationshipswithhumansor

otheranimates.Placesaresubclassifiedseparatelyfromartifactsbasedon

magnitude(amongotherthings),especiallyrelativetohowhumansexperience

them.Itispossiblethatdistinctionsinmarkednesswouldbefoundtodistinguish

thesesubclassesandtheirconceptsifmorelanguageswereaddedtothestudy.Even

so,thisstudysuggeststhatplantsandbodypartsaretreatedlikeinanimateswith

regardtopartsofspeech.Distinctionsamongtheseconceptsmightbereservedto

lowerlevelgrammaticalstructures.Forexample,alienableandinalienable

possessionrelationshipsareencodeddifferentlyinmanylanguages,whichcould

disambiguatebodypartsfromartifacts.Thisgrammaticaldistinctionisless

obviouslyrelatedtotheoccurrenceoftheselexemesinaparticularpropositional

actfunction;rather,ithasimplicationsforhowthatobjectconcepttakespartina

possessor-possesseerelationship.

Insummary,theinfluenceofanimacyisrobustinthenounprototypeinthis

analysis.Unliketheotherprototypeclustersdiscussedbelow,theobjectconcepts

arenearlyarrangedinastraightlinereflectingthisimplicationalhierarchy.

Furthermore,someknownconceptualdistinctionssuchasalienable/inalienable

andplant/inanimatemayonlyberelevantforgrammaticalstructuresunrelatedor

indirectlyrelatedtopartsofspeech.

78

5.3. Adjectiveprototype

Themostobviouscharacteristicoftheadjectiveprototypeisthatitisless

cohesivethanitsnounandverbcounterparts.Thesepropertyconceptsforma

relativelylooseclusterthatcoversalargerareaofthespatialmodel.Furthermore,

thisclusterdoesnotformedthecurvedlinethatwemightexpectatthebottomofa

horseshoearrangement.Instead,thepointsarespreadsignificantlytotheinside

andtheoutsideofthehorseshoe.Figure8givesacloserlookatthissectionofthe

spatialmodel.Afewoverlappingconceptsneedtobeidentified,sincetheyare

difficulttoreadevenintheenlargedimage.Theoverlappingconceptsintheupper

rightcornerofthissectionareMALEandFEMALE.Inthecenterofthesection,just

aboveSHORT,aretheconceptsGOODandOLD.Finally,YOUNG,FLAT,andBIGaretheat

leastpartiallyoverlappingconceptsinthelowerrightsection.

Theanalysisofpropertyconceptsintheliteratureasacontinuumfrom

object-liketoaction-likeispartiallysupportedbytheanalysis.Specifically,those

propertyconceptsthathavebeenplacedontheextremesofthiscontinuumshowup

thiswayintheanalysis.ThegenderconceptsMALEandFEMALEwereexpectedtobe

themostobject-likepropertyconceptsbasedonthisliterature,andtheyappearin

theupperrightcornerofthepropertyconceptcluster,closesttothenounprototype.

Aspredicted,thecolorconceptsWHITEandREDappearintheupperrightcorneras

well,justbelowandtotheleftofthegenderconcepts.Ontheaction-likesideofthe

continuum,humanpropensitiesarefoundintheupperleftcornerofthiscluster,

closesttotheverbprototype.Justalittleclosertothecenterofthecluster,physical

79

propertyconceptsSOFTandHEAVYarerightbehind.Theseobservationsinthespatial

modelreinforcetheclaimsintheliterature.

Figure8.Adjectiveprototype

Theeightpropertyconceptsthatareinthemiddleofthecontinuum—

neitherveryobject-likeoraction-like—arenotarrangedsoneatlyinthespatial

model.Someoftheseconceptsareclusteredveryclosetoeachother,andtheredoes

notseemtobeanobviousstructurefrom“nouny”to“verby”withinthem.Thismay

80

resultfromthelimiteddatasetusedinthisstudy.Datafromadditionallanguages

wouldsurelyteaseoutthedetailsattheheartofthisprototypestructure.Acouple

ofobservationswillhavetosufficeforthispaper.First,ROUNDappearsinthelower

leftcornerofthecluster,withasignificantamountofseparationfromtherestofthe

propertyconcepts.Itisnotclearwhatmotivatesthisposition,orwhetherthe

inclusionofadditionallanguagedatawouldperpetuateitsoutlyingpositioninthe

analysis.Second,BIGappearsinalow,relativelycenterpositionintheproperties

cluster,andthereforeontheoutsideofthehorseshoeshape.Thispositionreflects

thestatusofBIG(andothercommonconceptsrepresentingdimension,suchas

LITTLE)asanextremelyprototypicalpropertyconceptwhichenjoysunique

inflectionalcapabilitiesinmodificationinsomelanguages.

WhatarethesemanticmotivationsbehindtheprototypestructurethatI

havejustdescribedforpropertyconcepts?Althoughtheliteratureidentifies“nouny”

and“verby”propertyconcepts,thesemanticmotivationsforthesecharacteristics

arenotalwaysexplicated.Basedonthemarkednessdifferencesobservedinthis

study,Iwilldiscussthesemanticmotivationsthatseemtobeinfluencingthe

structureoftheadjectiveprototypeinthespatialmodel.

TheconceptsMALEandFEMALEappearintheupperrightcornerofthe

adjectiveprototype,closesttothenounprototype.Inseverallanguagesofthisstudy,

theseconceptsareexpressedwithlexemesclassifiedasNouns.Asaresult,these

lexemesareunabletoparticipateinAdjectivalinflections,suchascomparativesand

superlatives.Thismayreflectaconstrualoftheseconceptsaslackinggradability,

andfurtherevidencecomesfromlanguageswhichencodetheseconceptsas

81

Adjectives.InEnglish,maleandfemalecannottakecomparativeandsuperlative

suffixeslikemanyotherpropertyconcepts;instead,theyinflectperiphrastically

withmore/most.SimilarlyinK’ichee’,theseconceptscannotinflectforcomparative

likeotherpropertyconcepts,despitebeingclassifiedasAdjectivesbasedonother

formalcriteria.

Table11.Semanticpropertiesofprototypicalpartsofspeech(Croft2001:87)

Relationality Stativity Transitoriness GradabilityObjects nonrelational state permanent nongradableProperties relational state permanent gradableActions relational process transitory nongradable Onthe“verby”sideofthepropertyconceptcontinuum,thesemantic

primitivesthatwereclaimedtodistinguishpropertiesfromactionsaregradability,

stativity,andtransitoriness.Gradabilitydoesnotseemtoplaymuchofarolehere.

Infact,someoftheconceptsthatlandedintheverbprototypeclustercanbe

construedasgradable,suchasWANTandLOVE.Incontrast,theroleofstativityand

transitorinesscanbeseeninthemost“verby”propertyconcepts:thoseofhuman

propensity.Firstofall,HAPPYandANGRYareknownthroughhumanexperiencetobe

transitory.AtleastinEnglish,theycanrefertoanoverallpersonalitycharacteristic,

butmoreprototypicallytheyevokeatemporaryattitudeordemeanor.This

transitorinessofhumanpropensitiesistiedupwiththeirconstrualasaprocess,

sincehumanshavetotransitionintoandbackoutofaparticulartemperament.In

fact,thevariouspossibleconstrualsofanexperienceofhumanangerpresenta

challengetothisstudy.Alexemerecruitedtorepresenttheexperiencemighttarget

theresultingstateortheinchoativeprocess,forexample.Itispreciselythese

82

semanticcharacteristicsinherenttotheexperienceofhumanpropensitythatleadto

itsoccasionally“verby”morphosyntactictreatment.

5.4. Verbprototype

Theactionconceptsformarelativelytightclusterinthespatialmodel,

althoughnotquiteastightasthenounprototype.Also,asisfoundintheadjective

prototype,theactionsconceptsnotonlyfollowthecurvedlineofthehorseshoebut

exhibitvariationtotheinsideandoutsideaswell.Thisisindicativeofamore

complexsetofsemanticmotivationsbehindthemorphosyntacticrealizationof

theseconcepts.Inthelowerleft,theoverlappingconceptsareSITandWEAR.Inthe

lowerright,SEE/LOOKAT,LOVE,andWANTarebunchedtogether.Finally,GIVE,COUGH,

andEATaretheoverlappingconceptsnearthecenterofFigure9.

Asdiscussedin2.2.3,aspectualdistinctionswereexpectedtoplayarolein

theprototypicalityofactionconcepts.Thishypothesiswasconfirmedwithregardto

oneparticulardistinction:stativeconceptsversusdynamicconcepts.Two-

participantstates(KNOW,WANT,LOVE)andinactiveactions(SEE/LOOKAT,SIT,WEAR)are

stativeconcepts,andtheseareclearlyseparatedfromtherestoftheactionconcepts

inthespatialmodel.Asexpected,theyarefoundinthelowerandlowerrightareas

oftheverbcluster,closertotheadjectiveprototypethantherestoftheaction

concepts.Asdiscussedinsomeexamplesabove,thestative/dynamicdistinctionwas

foundtobesignificantformarkednesspatternsinsomeofthelanguagesinthis

study,andtheresultsareapparentinthemodel.

83

Figure9.Verbprototype

Withinthissubgroupofstativeconcepts,however,thedistinctionbetween

two-participantstatesandinactiveactionsisnotclearinthespatialmodel.Inactive

actionsSITandWEARarefoundtogetherinthelowestpositionintheverbcluster,

butSEE/LOOKAT—alsoaninactiveaction—isbunchedupwiththetwo-participant

states.TheseparationofSITandWEARfromtheotherstativeconceptsmayreflect

theunintentionalinclusionofmoreactivemeaningsinsomeofthelexemeschosen

84

torepresenttheseconcepts.Forsomelanguages,thesamelexemecanmean‘sit’

and‘sitdown’or‘wear’and‘puton’.Thesealternativemeanings—oftendifficultto

disentanglefromthestativemeanings—mightbemotivatingtheselexemesto

patternmorphosyntacticallyinsomewayswithdynamicactionconcepts.

Theremainingactionconceptsarespreadfromthecenterandleftcenter

sectionsoftheclustertowardtheupperright.Thisgenerallyfollowsthecurvedline

oftheexpectedhorseshoeshapeendingatCOOK.Thesemanticmotivationforthe

arrangementofthesedynamicactionconceptsinthemodelislessclear.Durative

processes,cyclicachievements,anddirectedachievementsareallintermingled

togetherinthispartoftheverbprototype.

Thereis,however,atendencysuggestiveofadurative/punctualdistinction.7

Durativeconceptstendtobefoundattheupperandupperrightportionsoftheverb

prototype.ThisgroupingmightincludeCOOK,WAVE,DIE,FIGHT,andBUILD.WAVEwas

chosenforthisstudytorepresentacyclicachievement,butiterativelyconstrued,it

couldtakeonadurativesense.DIEwouldbeaclearexceptiontothedurativetheme.

Punctualconcepts,ontheotherhand,arefoundclusteredinthecenterandleft

centeroftheprototype.TheseincludeCOME,HIT,BREAK,GIVE,COUGH,andEAT.COMEwas

selectedtorepresentadurativeprocess,butanalternativeconstrualforthis

concept—‘arrive’—couldexplainitsplaceamongpunctualconcepts.However,itis

difficulttoconceiveofastrictlypunctualconstrualofEAT,sothisconceptmustbe

consideredanexceptioninthisgroup.

7IamthankfultoBillCroftforbringingthistomyattentioninourdiscussionoftheresultsoftheMDSanalysis.

85

Ifweacceptthedurative/punctualdistinctionsuggestedinthemodel,why

arethedurativeconceptsfoundattheendofthehorseshoe(upperrightportionof

theactionconceptscluster)whereverbprototypicalityishighest?Ahypothesisof

thisstudywasthatpunctualactions—representedbycyclicachievementsand

directedachievements—wouldbemoreprototypicalamongactionsthandurative

ones.Theanswermaylieinanunexpectedrelationshipbetweenachievementsand

states.Fromaperspectiveofboundednessorcompletion,thesetwoaspectual

classesappeartobeverydifferent.However,thereisevidencethattheyareunited

aspotentialconstrualsofadirectedaspectualcontour(Croft2012:165-71).This

contourcorrespondstooneofBybeeetal.’s(1994)familiesofgrammaticaltense-

aspectcategorieswhichincludesperfectandperfectivesenses.Synchronicallyand

diachronically,verbscanalternatebetweenanachievementconstrualand

(resulting)stateconstrual.(Alternatively,durativeconceptsareconstrualsofan

undirectedaspectualcontour,correspondingtoBybeeetal.’sotherfamilyof

grammaticaltense-aspectcategorieswhichincludesimperfective,progressive,

present,andhabitualsenses.)Thissurprisingrelationshipmaybemotivatingthe

contiguityofpunctualandstativeconceptsinthepartsofspeechmodel.

Thespatialmodelhasprovidedvariousinsightsintothenatureoftheverb

prototype,butitdoesnothavetoexplaineveryaspectofthedata.Severalofthe

languagesinthisstudyfeatureimportantmarkednessdistinctionsamongaction

concepts—particularlywithregardtohowtheyarepredicated—thatdonotseemto

resultfromaspectualdistinctions.Forsomelanguagesthesedistinctionsarethe

resultofverbclasses,forotherlanguagesitisduetotheexistenceofsemantically

86

meaningfulauxiliaryverbs.Itisalsopossiblethatindividuallexicalfrequency

manifestsitselfinmorphosyntacticmarkednessenoughtodisrupttheinfluenceof

aspectualprototypicality.Itisinterestingthatthisdoesnotalsohappenwithin

objectconcepts.Forexample,noteverylexemerepresentingahumanconceptis

frequent,yetincertainlanguagesthemorphosyntactic“advantages”suchas

inflectionalpotentialareappliedtoalllexemesrepresentinghumanconcepts.What

differentiatestheactionconceptsfromtheobjectconceptsinthisregard?Itmaybe

thattheconceptualunityofsubclassesintheanimacyhierarchyisgreaterthanthat

ofaspectualsubclasses.Alternatively,thegrammaticalprocessesbywhich

markednesspatternsevolveinobjectsmaybedifferentthatthoseforactions,soit

maybeonlyindirectlyrelatedtothesemanticcharacteristicsthatmotivatethese

subclasses.

5.5. Interpretingthedimensions

Partsofspeechrepresentoneofthemostbasicandpervasivelevelsof

grammaticalorganization,anditwouldbeextremelysurprisingifonlytwoorthree

functionalorsemanticfactorswereshowntoexplainnearlyallofthedata.Rather,

thesedimensionsrepresentdifferentsemanticmotivationsindifferentareasofthe

spatialmodel.Forexample,withinthenounprototypecluster,onecouldarguethat

ananimacydimensionrunsalongasteep,positive-slopingline;highlyanimate

conceptsarehigherandfurthertotheright,whileinanimateconceptsarelowerand

furthertotheleft.However,ananimacyinterpretationofthisdimensionis

87

inappropriateforotherareasofthemodelwhereothersemanticfactorsappearto

bedrivingthedistributionofconcepts.

Themultidimensionalscalinganalysisconfirmsthehypothesisofthisstudy

thatatleastseveralsemanticfactorsarerelevantforthemorphosyntactic

realizationofobjects,properties,andactionsinthreepropositionalactfunctions.

Table12showstheprimarysemanticfactorsbasedonwheretheireffectisfoundin

thesemanticcontinuumfromobjectstopropertiestoactions.Mostofthese

semanticprimitivesarefamiliarfromCroft(2001:87,seealso1991),andtheywere

introducedinTable1.However,theeffectofanimacyinthisanalysisisstrong

enoughtoincludeitwiththeothersemanticfactors.

Table12.Semanticpropertiesmotivatingtheprototypestructuresofpartsofspeech

ACTIONSPROPERTIESOBJECTSdynamic stative

transitory permanent

non-gradablenongradable gradable nongradable

relational nonrelational

inanimate animate human

WhileTable12impliesthatallofthesesemanticprimitivesareactiveacross

theentirecontinuumfromobjectstopropertiestoactions,thismaynotbethecase.

Stativityandtransitorinessarecomponentsofthelargerconceptoftime-stability,

andtheyappeartoberelevantacrossthecontinuum.Similarly,relationality

motivatesadistinctionbetweenobjectsandproperties,soitseffectsarenot

confinedtoasingleprototype.However,animacymayberelevantonlyamong

nonrelationalconcepts,effectivelylimitingitsdomaintothenounprototype.Itmay

88

beappropriatetosaythatanimacycontributestotheinternalstructureofthe

prototype,butnottothe(non-discrete)boundarybetweenprototypes.Itisless

clear,butgradabilitymaybesimilarlylimitedinitsscope.Itseffectsareprimarily

foundinthemiddleofthecontinuum—thatis,amongpropertyconcepts—andless

sooneitherend.

Theanalysispresentedhereshowsthathumanscanindeedreducethe

complexityoftheworldintojustafewdimensions,butthatthosedimensionsmay

actuallyrepresentvariousfunctionalorsemanticprinciplessimultaneously.

5.6. Futureresearch

Themostobviousextensionofthisstudywouldbeitsapplicationtomore

languages.Theelevenlanguagesincludedinthisstudyarejustenoughtoillustrate

theprototypestructures.ThiscanbeseeninFigure10,whichshowsspatial

analysesusingdatafromthree,six,nine,andelevenlanguages.Ittakesclosetoa

dozentoteaseapartmostoftheconceptsinthespatialmodel.Doublingortripling

thesamplesizewouldgoalongwayinfurtherelucidatingdissimilaritiesandtheir

semanticmotivations.

Additionally,futureresearchcanimproveonthelistofconceptsincludedin

suchastudy.Ofcourse,withmorefine-grainedconceptualdistinctions,amore

informativemodelispossible.However,economicalconsiderationslimitthe

numberofconceptsthatcanbeincluded,sofutureworkmightinsteadreplace

uninformativeconceptualdistinctionswithonesthatarepredictedtoberelevant

basedontheresultshere.

89

Figure10.MDSanalysisofpartsofspeechforthree,six,nine,andelevenlanguages(clockwisefromtopleft)

5.7. Experience,conceptualstructure,andlinguisticform

Thisstudyisaninvestigationintohowthesameconceptualideamaybe

encodedgrammaticallyindifferentwaysfromonelanguagetothenext.Thisisthe

sourceofvariationbetweenthepartsofspeechprototypes.Thereis,however,

anothersourceofvariationinpartsofspeechamonglanguages.Thesamereal

90

worldtarget—whetheranobject,property,oraction—canbeconstrueddifferently

acrosslanguages.This,ofcourse,wouldresultindifferentmorphosyntactic

behaviorsforthelexemesrepresentingthesedifferentconstruals.Forexample,two

languagesmayconstruetheexperienceofangerdifferently,withonefocusingon

thetemporarystateofbeingangryandtheotherfocusingontheprocessofchange

fromanon-angrystatetoanangryone.Astudythatstartsataconceptualbaseline

missesthiskindofvariation.Thisthesisaimedtocapturevariationstartingfroma

conceptualbaseline,butaswehaveseen,alternativeconstrualsfossilizedinlexical

polysemyhavemostlikelyintroducedsomeofthisvariationinconstrual.

Therearemanystudiesthathaveaimedtocapturethevariationinconstrual,

perhapsmostnotablythepioneeringworkofthePearStories(Chafe1980).Theuse

ofvideoallowedparticipantsinthestudytogodirectlyfromexperiencetolinguistic

form,withoutapredeterminedsyntacticorconceptualstructure.Forsometargets

inthefilm,thelinguisticresultsaresurprisinglyhomogeneous;forothers,agreat

dealofvariationresulted.Itwouldbeinterestingtoconductastudyonpartsof

speechfromanexperientialbaseline.Thiswouldrequirevideoorpictureelicitation,

andagreatdealoftimeandresourcestocollectthedatafrommanydiverse

languages.Theresultwouldlikelybemorevariationthanisfoundinthisstudy,but

howmuchmore?Wouldthepartsofspeechprototypestructuresemergeinsucha

study,orwouldtheybehiddenbytheconfoundingeffectsofvariationinconstrual?

91

6. Conclusions

Thisthesishassoughttoillustratetheprototypestructuresofpartsofspeech.

Aftercodingmarkednessasymmetriesinelevenlanguages,Iwasabletoapplya

multidimensionalscalinganalysistothedatatocreateatwo-dimensionalspatial

representationthatcapturesalmostallofthevariation.Themodelshowsclearly

threeprototypestructuresrepresentingnouns,adjectives,andverbsthatresult

indirectlyfromconceptualdissimilarities.Thenounandverbprototypeclustersare

stronger,whiletheadjectiveprototypeclusterisweaker.

Inparticular,itistheinteractionofconceptualclassesandsubclasseswith

thethreebasicpropositionalactfunctionsthatyieldmarkednessdifferences.

Languagesexhibitvariationregardingwhichconceptsbelongtowhichlanguage-

specificclassesandregardinghowtheseconceptsareencodedmorphosyntactically.

Yetthroughthisvariation,theprototypicalityofobjectsinreference,propertiesin

modification,andactionsinpredicationmotivatethestructuresthatareseenclearly

inthespatialmodelpresentedhere.Itissignificantenoughtoemphasizeagainthat

everylanguageinthisstudywasfoundtoexhibitsomemorphosyntactic

distinctionsbetweenobjects,properties,andactions.Thischaracteristicofthedata

iscriticalinmotivatingaspatialmodelwiththreedistinctclusters,ratherthan

continuousandinseparabledatapointsalongthesemanticcontinuumfromobjects

topropertiestoactions.

Oneofthemostimportantcontributionsofthisstudyistoourunderstanding

ofthesemanticfactorsthatmotivatethepartsofspeechprototypes.Theroleof

severalsemanticfactorsidentifiedintheliteratureweresupportedinthisanalysis,

92

suchasrelationality,gradability,transitoriness,andstativity.Theevidencefortheir

roleisseenbothqualitativelyinmarkednessdistinctionsofparticularlanguages

thatreflecttheirinfluenceandquantitativelyinthearrangementofconceptsinthe

spatialmodel.Animacyhasalsobeendiscussedintheliteratureasshapingformal

distinctionsamonglexemesrepresentingobjectconcepts,butwasnotincludedasa

primarysemanticfactorinpartsofspeechprototypestructures.Theanalysishere

hasshownthatitplaysakeyrole,butalsothatthisrolemaybeconfinedtothe

internalshapeofthenounprototype.Thisopensthedoorforfutureresearchto

explorethedomainofeachsemanticfactor—withinaparticularprototype,shaping

theboundarybetweentwoormoreprototypes,orboth.

Theseresultssupportandillustratetheprototypetheoryofpartsofspeech.

Evenmore,theyareconsistentwithcontemporarytypologicalresearchand

specificallymultidimensionalscalinganalysesthatsupportneitheran“extreme

universalist”noran“extremerelativist”perspectiveonlanguage(Croft&Poole

2008:31-2).Thevariationinpartsofspeechstrategiescannotbeexplainedinterms

ofuniversalcategories,yetthepatternsthatemergeareindicativeofuniversal

functionalpressuresthatconstrainthevariationprobabilistically.Thespatialmodel

ofpartsofspeechdatapresentedhereillustratestheuniversalconceptualstructure

underlyingthemostbasicorganizationofgrammar.

93

LISTOFAPPENDICES

AppendixA.Completeconceptsubclassificationandselectedexemplars.....................94

AppendixB.PartsofspeechdatacodedforMDSanalysis....................................................95

AppendixC.MDSoutputtableforpoints...................................................................................112

AppendixD.MDSoutputtableforcuttinglines......................................................................113

94

AppendixA.Completeconceptsubclassificationandselectedexemplars

Class(/Type) Exemplars

Humans WOMANBOY

Kinship FATHERSISTER

Animals DOGBIRD

Plants/PlantProducts TREESEED

Artifacts HATBED

BodyParts ARMFACE

Places RIVERHOUSE

Material/Substance WATERSTONE

Gender MALEFEMALE

Color WHITERED

Form ROUNDFLAT

Age OLDYOUNG

Value GOODBAD

Dimension BIGSHORT

PhysicalProperties HEAVYSOFT

HumanPropensity HAPPYANGRY

2-ParticipantStatesCognition KNOWDesire WANTEmotion LOVE

InactiveActionsPerception SEE/LOOKATPosture SITSpatial/Possession WEAR

DurativeProcesses

SocialInteraction FIGHTConsumption EATMotion COMEChangeofState COOKCreationVerbs BUILD

CyclicAchievements(Semelfactives)

Emission COUGHContact HITBodilyMovement WAVE

DirectedAchievementsExistence DIEChangeofPossession GIVEChangeofState BREAK

95

AppendixB.PartsofspeechdatacodedforMDSanalysis

KH-R-SC-0

KH-R-SC-VarStr

KH-R-BP-PGN

KH-R-BP-0

KH-M-SC-0

KH-M-SC-VarStr

KH-M-BP-SPRL1

KH-M-BP-SPRL2

KH-M-BP-0

KH-P-SC-0

WOMAN 1 6 1 6 6 1 6 6 1 6BOY 1 6 1 6 6 1 6 6 1 6FATHER 1 6 1 6 6 1 6 6 1 6SISTER 1 1 1 1 1 1 6 1 1 1DOG 1 6 1 6 6 1 6 6 1 6BIRD 1 6 1 6 6 1 6 6 1 6TREE 1 6 1 6 6 1 6 6 1 6SEED 1 6 1 6 6 1 6 6 1 6HAT 1 6 1 6 6 1 6 6 1 6BED 1 6 1 6 6 1 6 6 1 6ARM 1 6 1 6 6 1 6 6 1 6FACE 1 6 1 6 6 1 6 6 1 6RIVER 1 6 1 6 6 1 6 6 1 6HOUSE 1 6 1 6 6 1 6 6 1 6WATER 1 6 1 6 6 1 6 6 1 6STONE 1 6 1 6 6 1 6 6 1 6MALE 9 9 9 9 9 9 9 9 9 9FEMALE 9 9 9 9 9 9 9 9 9 9WHITE 6 1 6 1 1 6 6 1 6 1RED 1 1 1 1 1 1 6 1 6 1ROUND 6 1 6 1 1 6 6 1 6 1FLAT 6 1 6 1 1 6 6 1 6 1OLD 6 1 6 1 1 6 6 1 6 1YOUNG 6 1 6 1 1 6 6 1 6 1GOOD 6 1 6 1 1 6 6 1 6 1BAD 6 1 6 1 1 6 6 1 6 1BIG 6 1 6 1 1 6 1 6 6 1SHORT 6 1 6 1 1 6 6 1 6 1HEAVY 6 1 6 1 1 6 6 1 6 1SOFT 6 1 6 1 1 6 6 1 6 1HAPPY 6 1 6 1 1 6 6 1 6 1ANGRY 6 1 6 1 1 6 6 1 6 1KNOW 6 1 6 1 1 6 6 1 6 1WANT 6 1 6 1 1 6 6 1 6 1LOVE 6 1 6 1 1 6 6 1 6 1SEE/LOOKAT 6 1 6 1 1 6 6 1 6 1SIT 6 1 6 1 1 6 6 1 6 1WEAR 6 1 6 1 1 6 6 1 6 1FIGHT 6 1 6 1 1 6 6 1 6 1EAT 6 1 6 1 1 6 6 1 6 1COME 6 1 6 1 1 6 6 1 6 1COOK 6 1 6 1 1 6 6 1 6 1BUILD 6 1 6 1 1 6 6 1 6 1COUGH 6 1 6 1 1 6 6 1 6 1HIT 6 1 6 1 1 6 6 1 6 1WAVE 6 1 6 1 1 6 6 1 6 1DIE 6 1 6 1 1 6 6 1 6 1GIVE 6 1 6 1 1 6 6 1 6 1BREAK 6 1 6 1 1 6 6 1 6 1

96

AppendixB.PartsofspeechdatacodedforMDSanalysis(cont’d)

KH-P-SC-Cop

KH-P-BP-TAM1

KH-P-BP-TAM2

KH-P-BP-0

KS-R-SC-0

KS-R-SC-Aff

KS-R-BP-Class

KS-R-BP-Num

KS-R-BP-0

KS-M-SC-0


97


KS-M-SC-Suff

KS-P-SC-0

KS-P-SC-Cop

KS-P-BP-TAM

KS-P-BP-0

JI-R-SC-0

JI-R-SC-Aff

JI-R-BP-NGC

JI-R-BP-0

JI-M-SC-0


98


JI-M-SC-Aff

JI-M-BP-NGCagr

JI-M-BP-0

JI-P-SC-0

JI-P-SC-Aff

JI-P-SC-LV

JI-P-BP-TAM-abl

JI-P-BP-TAM-peri

JI-P-BP-0

MA-R-BP-NPArt


99


MA-R-BP-SpecArt

MA-R-BP-NClass

MA-R-BP-0

MA-M-SC-REL

MA-M-SC-REL+Pr

MA-M-SC-PossPr

MA-M-BP-Intens

MA-M-BP-0

MA-P-SC-0

MA-P-SC-Cop


100


MA-P-BP-ConcPr

MA-P-BP-T1

MA-P-BP-T2

MA-P-BP-T3

MA-P-BP-T4

TO-R-SC-0

TO-R-SC-Aff

TO-R-BP-Num

TO-R-BP-Pron

TO-R-BP-PLmark


101


TO-R-BP-Subj-IndNum

TO-R-BP-0

TO-M-SC-0

TO-M-SC-Aff

TO-M-SC-RelMar

TO-P-BP-TAM

TO-P-BP-0

QE-R-SC-0

QE-R-SC-Aff

QE-M-SC-0


102


QE-M-SC-Suff

QE-M-BP-SPRL

QE-M-BP-0

QE-P-SC-0

QE-P-SC-Cop

QE-P-BP-TAM

QE-P-BP-TAM-peri

CR-R-SC-0

CR-R-SC-Aff

CR-R-BP-Num


103


CR-R-BP-Dim

CR-R-BP-0

CR-M-SC-Attach

CR-M-SC-Aff

CR-M-SC-AffLgr

CR-M-BP-CMPR

CR-M-BP-0

CR-P-SC-0

CR-P-SC-DURAff

CR-P-SC-Cop


104


CR-P-BP-TAM

CR-P-BP-TAM-peri

KI-R-SC-0

KI-R-SC-Aff

KI-R-BP-Num-Suff

KI-R-BP-Num-Peri

KI-R-BP-Num-Vindex

KI-R-BP-0

KI-M-SC-0

KI-M-SC-GEN


105


KI-M-SC-REL

KI-M-BP-CMPR

KI-M-BP-0

KI-P-SC-0

KI-P-SC-Plus-Noun

KI-P-SC-FOC

KI-P-BP-TAM-IndArg

KI-P-BP-0

EN-R-SC-0

EN-R-SC-Aff


106


EN-R-SC-Peri

EN-R-BP-Num-Abl

EN-R-BP-Num-Suff

EN-R-BP-0

EN-M-SC-0

EN-M-SC-Aff

EN-M-SC-REL

EN-M-BP-CMPR-Abl

EN-M-BP-CMPR-Suff

EN-M-BP-CMPR-Suff2


107


EN-M-BP-CMPR-Peri

EN-M-BP-SPRL-Abl

EN-M-BP-SPRL-Suff

EN-M-BP-SPRL-Suff2

EN-M-BP-0

EN-P-SC-0

EN-P-SC-Cop

EN-P-SC-CopArt

EN-P-BP-TAbl

EN-P-BP-Taff


108


EN-P-BP-TPeri

EN-P-BP-P+N-Suff

EN-P-BP-P+M-Peri

IN-R-SC-0

IN-R-SC-Aff

IN-R-BP-Case-Num

IN-R-BP-0

IN-M-SC-0

IN-M-SC-Aff

IN-M-BP-CMPR


109


IN-M-BP-Gend-Agr

IN-M-BP-Num-Agr

IN-M-BP-Case-Agr

IN-M-BP-0

IN-P-SC-0

IN-P-SC-Comp

IN-P-SC-Cop

IN-P-BP-Iter

IN-P-BP-Aux

IN-P-BP-Ind-Num


110


IN-P-BP-Ind-Gend

IN-P-BP-Sppl

IN-P-BP-Full-Akt

IN-P-BP-Part-Akt

IN-P-BP-Restr-Akt

KA-R-SC-0

KA-R-SC-Aff

KA-R-BP-PL

KA-R-BP-0

KA-M-SC-0


111


KA-M-SC-Aff

KA-M-BP-PL

KA-M-BP-Intens

KA-M-BP-0

KA-P-SC-O

KA-P-SC-Comp-Verb

KA-P-BP-TAM-Dir

KA-P-BP-TAM-Indir

KA-P-BP-0

WOMAN 1 6 6 1 1 6 6 6 1BOY 1 6 6 1 1 6 6 6 1FATHER 1 6 6 1 1 6 6 6 1SISTER 1 6 6 1 1 6 6 6 1DOG 1 6 6 1 1 6 6 6 1BIRD 1 6 6 1 1 6 6 6 1TREE 1 6 6 1 1 6 6 6 1SEED 1 6 6 1 1 6 6 6 1HAT 1 6 6 1 1 6 6 6 1BED 1 6 6 1 1 6 6 6 1ARM 1 6 6 1 1 6 6 6 1FACE 1 6 6 1 1 6 6 6 1RIVER 1 6 6 1 1 6 6 6 1HOUSE 1 6 6 1 1 6 6 6 1WATER 1 6 6 1 1 6 6 6 1STONE 1 6 6 1 1 6 6 6 1MALE 6 6 9 9 1 6 6 6 1FEMALE 6 6 9 9 1 6 6 6 1WHITE 6 6 1 6 1 6 6 6 1RED 6 6 1 6 1 6 6 6 1ROUND 1 6 9 9 6 6 6 6 1FLAT 6 6 1 6 1 6 6 6 1OLD 6 6 1 6 1 6 6 6 1YOUNG 6 6 1 6 1 6 6 6 1GOOD 6 6 1 6 1 6 6 6 1BAD 6 6 1 6 1 6 6 6 1BIG 6 1 1 6 1 6 6 6 1SHORT 6 6 1 6 1 6 6 6 1HEAVY 6 6 1 6 1 6 6 6 1SOFT 6 6 1 6 1 6 6 6 1HAPPY 6 6 1 6 1 6 6 6 1ANGRY 6 6 1 6 1 6 6 6 1KNOW 1 6 6 1 6 1 6 1 6WANT 1 6 6 1 1 1 1 1 6LOVE 1 6 6 1 1 6 1 6 6SEE/LOOKAT 1 6 6 1 1 6 1 6 6SIT 9 9 9 9 9 9 9 9 9WEAR 9 9 9 9 9 9 9 9 9FIGHT 1 6 6 1 1 1 1 1 6EAT 1 6 6 1 1 6 1 6 6COME 1 6 6 1 1 6 1 6 6COOK 1 6 6 1 6 1 6 1 6BUILD 1 6 6 1 1 1 1 1 6COUGH 1 6 6 1 1 6 1 6 6HIT 1 6 6 1 6 1 6 1 6WAVE 1 6 6 1 1 6 1 6 6DIE 1 6 6 1 1 1 1 1 6GIVE 1 6 6 1 1 6 1 6 6BREAK 1 6 6 1 6 1 6 1 6

112

AppendixC.MDSoutputtableforpoints

rank

correctYea

wrongYea

wrongN

ay

correctNay

volume

coord1D

coord2D

WOMAN 49 75 2 0 90 0.006000001 0.861697508 0.428549482BOY 48 74 1 0 91 0.022000002 0.859375265 0.417361899FATHER 46 72 0 0 94 0.021000002 0.834974071 0.405622173SISTER 45 73 0 5 88 0.012000001 0.834398446 0.418269535DOG 47 72 0 0 94 0.030000001 0.853451011 0.384299431BIRD 44 70 0 0 96 0.032000002 0.817363695 0.377763505TREE 39.5 69 0 0 97 0.047000002 0.797914134 0.33058484SEED 39.5 69 0 0 97 0.048000004 0.797914134 0.33058484HAT 39.5 69 0 0 97 0.047000002 0.797914134 0.33058484BED 39.5 64 0 0 93 0.049000002 0.797914134 0.33058484ARM 39.5 69 0 0 97 0.049000002 0.797914134 0.33058484FACE 39.5 69 0 0 97 0.050000001 0.797914134 0.33058484RIVER 39.5 64 0 0 90 0.050000001 0.797914134 0.33058484HOUSE 39.5 69 0 0 97 0.050000001 0.797914134 0.33058484WATER 35 68 1 0 97 0.025 0.792363729 0.321871172STONE 34 69 0 0 97 0.011000001 0.769228641 0.286396204MALE 31.5 42 2 0 67 0.002 0.33050118 -0.467034192FEMALE 31.5 44 2 0 67 0.015000001 0.33050118 -0.467034192WHITE 30 71 0 2 91 0.100000001 0.171691973 -0.508718317RED 33 76 1 3 85 0.018000001 0.339187447 -0.515571356ROUND 18 50 1 1 75 0.003 -0.321879643 -0.817196723FLAT 27 62 0 0 86 0.027000001 0.10800206 -0.853864506OLD 24.5 70 1 1 92 0.150000006 0.001233645 -0.720621782YOUNG 28 68 1 1 95 0.026000001 0.108121724 -0.854013585GOOD 24.5 69 2 2 91 0.149000004 0.001233645 -0.720621782BAD 20 65 5 5 89 0.213000014 -0.101191287 -0.71706093BIG 29 71 1 0 92 0.032000002 0.109490557 -0.868232146SHORT 26 70 0 0 93 0.181000009 0.017681246 -0.747266058HEAVY 21 72 1 2 85 0.045000002 -0.098314302 -0.55680758SOFT 22 73 1 0 91 0.014 -0.08870984 -0.542168107HAPPY 19 62 0 3 90 0.003 -0.284728932 -0.532321416ANGRY 23 67 2 3 88 0.002 -0.08686224 -0.511127201KNOW 14 63 0 2 104 0.005 -0.763423849 0.212456871WANT 16 64 0 3 102 0.018000001 -0.723274315 0.223196688LOVE 17 62 2 4 98 0.003 -0.720340106 0.226282023SEE/LOOKAT 15 65 0 1 103 0.020000001 -0.726102476 0.224426089SIT 4 59 1 0 95 0.006000001 -0.859664189 0.190334549WEAR 3 54 0 0 89 0.007 -0.859962073 0.189018745FIGHT 8 65 2 1 101 0.042000003 -0.808105513 0.379253998EAT 9 64 1 2 99 0.011000001 -0.789116407 0.287659753COME 1 67 7 1 94 0.002 -0.913218145 0.27430276COOK 13 65 5 1 95 0.004 -0.773803421 0.490559158BUILD 7 64 0 0 93 0.017000001 -0.823639237 0.371566355COUGH 10 63 3 3 100 0.012000001 -0.789116407 0.287659753HIT 2 67 4 0 98 0.012000001 -0.899269646 0.282302936WAVE 12 55 2 0 85 0.014 -0.777703801 0.437822889DIE 5 69 2 1 97 0.003 -0.837063773 0.392426895GIVE 11 66 0 3 100 0.010000001 -0.785703816 0.289630448BREAK 6 65 1 0 103 0.037 -0.829541173 0.31672256

113

AppendixD.MDSoutputtableforcuttinglines

correctYea

wrongYea

wrongN

ay

correctNay

PRE

normVector1D

normVector2D

midpoints

KH.R.SC.0 17 0 0 30 1 0.805007101 0.593265174 -0.103897367KH.R.SC.VarStr 31 0 1 15 0.933333333 0.135342226 0.990798911 0.384659862KH.R.BP.PGN 17 0 0 30 1 0.805007101 0.593265174 -0.103897367KH.R.BP.0 31 0 1 15 0.933333333 0.135342226 0.990798911 0.384659862KH.M.SC.0 31 0 1 15 0.933333333 0.135342226 0.990798911 0.384659862KH.M.SC.VarStr 17 0 0 30 1 0.805007101 0.593265174 -0.103897367KH.M.BP.SPRL1 1 0 0 46 1 0.083924075 -0.996472152 0.865474831KH.M.BP.SPRL2 31 1 0 15 0.9375 0.886533834 -0.462663767 0.548464427KH.M.BP.0 16 0 0 31 1 0.759709206 0.650262964 -0.052589374KH.P.SC.0 31 0 1 15 0.933333333 0.135342226 0.990798911 0.384659862KH.P.SC.Cop 17 0 0 30 1 0.805007101 0.593265174 -0.103897367KH.P.BP.TAM1 17 0 0 30 1 0.815400825 -0.578896791 -0.322512825KH.P.BP.TAM2 14 0 1 32 0.933333333 0.61189348 -0.790940181 0.245815811KH.P.BP.0 16 0 0 31 1 0.759709206 0.650262964 -0.052589374KS.R.SC.0 15 0 3 27 0.833333333 0.905904266 -0.42348254 0.554941479KS.R.SC.Aff 28 3 0 15 0.833333333 0.901715792 -0.432329307 0.553038699KS.R.BP.Class 15 0 0 27 1 0.759709206 0.650262964 -0.052589374KS.R.BP.Num 14 1 0 27 0.928571429 0.131411632 0.991327889 0.384872399KS.R.BP.0 27 0 0 15 1 0.759709206 0.650262964 -0.052589374KS.M.SC.0 12 0 1 17 0.923076923 0.833042122 0.553209566 -0.403280106KS.M.SC.Suff 21 1 0 3 0.75 0.069839772 0.997558222 -0.839075576KS.P.SC.0 23 0 2 22 0.909090909 0.858824074 0.512270642 -0.362859967KS.P.SC.Cop 28 1 1 15 0.875 0.798757173 -0.601653537 0.237371388KS.P.BP.TAMext 23 0 2 22 0.909090909 0.858824074 0.512270642 -0.362859967KS.P.BP.0 28 1 1 15 0.875 0.798757173 -0.601653537 0.237371388JI.R.SC.0 16 0 0 15 1 0.815400825 -0.578896791 0.144504531JI.R.SC.Aff 15 0 0 16 1 0.759709206 0.650262964 -0.052589374JI.R.BP.NGC 18 0 0 29 1 0.910031919 0.414538184 -0.151896832JI.R.BP.0 29 0 0 18 1 0.848618008 0.529006122 -0.236404902JI.M.SC.0 14 0 0 15 1 0.815400825 -0.578896791 -0.322512825JI.M.SC.Aff 31 0 0 14 1 0.077290281 -0.997008632 0.132589649JI.M.BP.NGCagr 30 0 0 15 1 0.815400825 -0.578896791 -0.322512825JI.M.BP.0 15 0 0 30 1 0.815400825 -0.578896791 -0.322512825JI.P.SC.0 15 1 0 29 0.933333333 0.392207382 0.919876823 -0.10554094JI.P.SC.Aff 16 0 0 29 1 0.759709206 0.650262964 0.359377534JI.P.SC.LV 14 1 0 30 0.928571429 0.140628632 -0.990062416 -0.307058682JI.P.BP.TAM.abl 1 0 0 44 1 0.903766448 -0.428025943 -0.938371886JI.P.BP.TAM.peri 14 1 0 30 0.928571429 0.140628632 -0.990062416 -0.307058682JI.P.BP.0 30 0 0 15 1 0.815400825 -0.578896791 -0.322512825MA.R.BP.NPArt 17 0 0 31 1 0.966883805 -0.255216983 0.452026892MA.R.BP.SpecArt 29 0 0 19 1 0.871261763 -0.490818643 0.181535394MA.R.BP.NClass 29 0 0 19 1 0.871261763 -0.490818643 0.181535394MA.R.BP.0 19 0 0 29 1 0.871261763 -0.490818643 0.181535394MA.M.SC.REL 31 0 0 17 1 0.966883805 -0.255216983 0.452026892MA.M.SC.REL.Prep 17 0 0 31 1 0.966883805 -0.255216983 0.452026892MA.M.SC.PossPron 17 0 0 31 1 0.966883805 -0.255216983 0.452026892MA.M.BP.Intens 12 1 0 35 0.916666667 0.105814136 -0.994385925 0.499294263

114

AppendixD.MDSoutputtableforcuttinglines(cont’d)

correctYea

wrongYea

wrongN

ay

correctNay

PRE

normVector1D

normVector2D

midpoints

MA.M.BP.0 35 0 1 12 0.916666667 0.105869247 -0.994380059 0.499292599MA.P.SC.0 19 0 0 29 1 0.871261763 -0.490818643 0.181535394MA.P.SC.Cop 29 0 0 19 1 0.871261763 -0.490818643 0.181535394MA.P.BP.ConcPron 31 0 0 17 1 0.966883805 -0.255216983 0.452026892MA.P.BP.T1 3 1 1 43 0.5 0.996973736 -0.077739114 -0.867734984MA.P.BP.T2 11 4 0 33 0.636363636 0.148875981 -0.988855875 -0.319300761MA.P.BP.T3 1 1 1 45 0 0.906657705 -0.421867048 -0.929610446MA.P.BP.T4 31 0 0 17 1 0.815400825 -0.578896791 -0.322512825TO.R.SC.0 15 0 0 30 1 0.759709206 0.650262964 -0.052589374TO.R.SC.Aff 30 0 0 15 1 0.759709206 0.650262964 -0.052589374TO.R.BP.Num 15 0 0 30 1 0.759709206 0.650262964 -0.052589374TO.R.BP.Pron 4 0 0 41 1 0.551009719 0.834498825 0.79789426TO.R.BP.PLmark 5 0 0 40 1 0.72049683 0.69345823 0.868118196TO.R.BP.SubjIndN 5 0 0 40 1 0.72049683 0.69345823 0.868118196TO.R.BP.0 30 0 0 15 1 0.759709206 0.650262964 -0.052589374TO.M.SC.0 31 0 0 16 1 0.815400825 -0.578896791 -0.322512825TO.M.SC.Aff 17 0 0 30 1 0.832039125 0.55471695 0.003333235TO.M.SC.RelMark 30 0 0 17 1 0.832039125 0.55471695 0.003333235TO.P.BP.TAM 30 0 0 17 1 0.832039125 0.55471695 0.003333235TO.P.BP.0 17 0 0 30 1 0.832039125 0.55471695 0.003333235QE.R.SC.0 16 0 0 18 1 0.906241601 -0.422760169 0.011581476QE.R.SC.Aff 21 0 0 16 1 0.759709206 0.650262964 -0.052589374QE.M.SC.0 31 0 0 18 1 0.845659021 -0.533723543 0.104845334QE.M.SC.Suff 37 0 0 12 1 0.439234795 -0.898372303 0.499757669QE.M.BP.SPRL 16 0 0 33 1 0.657924726 -0.753083697 0.286430384QE.M.BP.0 36 0 0 13 1 0.710980511 -0.703211712 0.342388845QE.P.SC.0 21 0 0 28 1 0.746366578 -0.665535071 0.302672077QE.P.SC.Cop 31 0 0 18 1 0.845659021 -0.533723543 0.104845334QE.P.BP.TAM 21 0 0 28 1 0.746366578 -0.665535071 0.302672077QE.P.BP.TAMperi 31 0 0 18 1 0.845659021 -0.533723543 0.104845334CR.R.SC.0 16 0 0 31 1 0.759709206 0.650262964 -0.052589374CR.R.SC.Aff 31 0 0 16 1 0.759709206 0.650262964 -0.052589374CR.R.BP.Num 2 0 0 45 1 0.758010496 0.652242354 0.91607827CR.R.BP.Dim 16 0 0 31 1 0.759709206 0.650262964 -0.052589374CR.R.BP.0 31 0 0 16 1 0.759709206 0.650262964 -0.052589374CR.M.SC.Attach 14 0 0 33 1 0.077290281 -0.997008632 0.132589649CR.M.SC.Aff 36 0 0 11 1 0.778488432 -0.627658953 -0.790927946CR.M.SC.AffLgr 11 0 0 36 1 0.778488432 -0.627658953 -0.790927946CR.M.BP.CMPR 20 0 0 27 1 0.218296833 0.975882418 0.064665113CR.M.BP.0 27 0 0 20 1 0.218296833 0.975882418 0.064665113CR.P.SC.0 17 0 0 30 1 0.815400825 -0.578896791 -0.322512825CR.P.SC.DURAff 14 0 0 33 1 0.077290281 -0.997008632 0.132589649CR.P.SC.Cop 30 0 0 17 1 0.815400825 -0.578896791 -0.322512825CR.P.BP.TAM 31 0 0 16 1 0.759709206 0.650262964 -0.052589374CR.P.BP.TAMperi 30 0 0 17 1 0.815400825 -0.578896791 -0.322512825KI.R.SC.0 18 0 0 26 1 0.910031919 0.414538184 -0.151896832KI.R.SC.Aff 26 0 0 18 1 0.910031919 0.414538184 -0.151896832

115


correctYea

wrongYea

wrongN

ay

correctNay

PRE

normVector1D

normVector2D

midpoints

KI.R.BP.Num.Suff 2 0 0 42 1 0.758010496 0.652242354 0.91607827KI.R.BP.Num.Peri 14 2 0 28 0.857142857 0.134894412 0.990859979 0.384684387KI.R.BP.Num.Vind 6 0 0 38 1 0.726753379 0.686898483 0.851460168KI.R.BP.0 28 0 0 16 1 0.759709206 0.650262964 -0.052589374KI.M.SC.0 15 0 0 33 1 0.108156838 -0.994133843 0.498570448KI.M.SC.GEN 16 0 0 32 1 0.759709206 0.650262964 0.359377534KI.M.SC.REL 17 0 0 31 1 0.845659021 -0.533723543 0.104845334KI.M.BP.CMPR 14 0 0 34 1 0.03816815 -0.999271331 0.493397837KI.M.BP.0 35 0 0 14 1 0.03816815 -0.999271331 0.493397837KI.P.SC.0 32 1 0 15 0.9375 0.905462452 -0.424426376 0.57581991KI.P.SC.PlusNoun 0 0 1 47 0 0.995343772 0.096388671 0.903267054KI.P.SC.FOC 15 0 0 33 1 0.750974118 0.66033164 0.778518423KI.P.BP.TAMIndArg 17 0 0 31 1 0.845659021 -0.533723543 0.104845334KI.P.BP.0 31 0 0 17 1 0.845659021 -0.533723543 0.104845334EN.R.SC.0 19 2 0 28 0.894736842 0.836046157 0.548659115 -0.360463655EN.R.SC.Aff 30 3 0 16 0.842105263 0.894682439 -0.446702734 0.549675992EN.R.SC.Peri 0 0 2 47 0 0.997660913 0.068357168 0.893144248EN.R.BP.NumAbl 1 0 0 48 1 0.080235799 0.996775911 0.493979893EN.R.BP.NumSuff 15 1 0 33 0.933333333 0.13658569 0.99062826 0.38459134EN.R.BP.0 33 0 0 16 1 0.759709206 0.650262964 0.359377534EN.M.SC.0 28 0 1 20 0.95 0.705045934 -0.70916164 0.301908103EN.M.SC.Aff 36 2 0 11 0.846153846 0.004674728 -0.999989073 0.530793426EN.M.SC.REL 17 0 0 32 1 0.815400825 -0.578896791 -0.322512825EN.M.BP.CMPRAbl 0 0 2 47 0 0.994291391 0.106698784 0.906817239EN.M.BP.CMPRSuff 10 3 0 36 0.7 0.101960971 -0.9947884 0.500063737EN.M.BP.CPRSuff2 0 0 1 48 0 0.148835634 0.988861949 -0.858103443EN.M.BP.CMPRPeri 1 0 0 48 1 0.122816239 0.992429429 -0.849346925EN.M.BP.SPRLAbl 0 0 2 47 0 0.994291391 0.106698784 0.906817239EN.M.BP.SPRLSuff 11 2 0 36 0.818181818 0.105237598 -0.994447107 0.499311583EN.M.BP.SPRLSuff2 0 0 1 48 0 0.148835634 0.988861949 -0.858103443EN.M.BP.0 35 0 0 14 1 0.03816815 -0.999271331 0.493397837EN.P.SC.0 17 0 0 32 1 0.815400825 -0.578896791 -0.322512825EN.P.SC.Cop 18 0 0 31 1 0.157747437 -0.987479491 -0.19994926EN.P.SC.CopArt 14 0 0 35 1 0.746614471 0.665256966 0.812687789EN.P.BP.TAbl 11 4 0 34 0.636363636 0.35504321 -0.934849891 -0.467402502EN.P.BP.Taff 3 0 3 43 0.5 0.759653708 -0.650327797 -0.872386957EN.P.BP.TPeri 32 0 0 17 1 0.815400825 -0.578896791 -0.322512825EN.P.BP.P+N.Suff 17 0 0 32 1 0.815400825 -0.578896791 -0.322512825EN.P.BP.P+N.Peri 32 0 0 17 1 0.815400825 -0.578896791 -0.322512825IN.R.SC.0 16 0 2 31 0.888888889 0.888541909 -0.458795461 0.546715708IN.R.SC.Aff 33 0 0 16 1 0.759709206 0.650262964 0.359377534IN.R.BP.CaseNum 32 0 0 17 1 0.815400825 -0.578896791 -0.322512825IN.R.BP.0 17 0 0 32 1 0.815400825 -0.578896791 -0.322512825IN.M.SC.0 16 0 0 33 1 0.077290281 -0.997008632 0.132589649IN.M.SC.Aff 34 0 1 14 0.928571429 0.679071109 -0.734072496 0.314391842IN.M.BP.CMPR 14 0 0 35 1 0.03816815 -0.999271331 0.493397837IN.M.BP.GendAgr 1 0 1 47 0.5 0.112080302 0.993699153 -0.84931586

116


correctYea

wrongYea

wrongN

ay

correctNay

PRE

normVector1D

normVector2D

midpoints

IN.M.BP.NumAgr 1 0 0 48 1 0.083924075 -0.996472152 0.865474831IN.M.BP.CaseAgr 33 0 0 16 1 0.759709206 0.650262964 0.359377534IN.M.BP.0 16 0 0 33 1 0.759709206 0.650262964 0.359377534IN.P.SC.0 15 2 0 32 0.866666667 0.146680095 -0.989183982 -0.306770662IN.P.SC.Comp 1 0 1 47 0.5 0.840059075 -0.542494931 -0.916119457IN.P.SC.Cop 32 0 0 17 1 0.815400825 -0.578896791 -0.322512825IN.P.BP.Iter 1 1 1 46 0 0.953009227 0.302941269 -0.76640673IN.P.BP.Aux 3 1 1 44 0.5 0.941653131 -0.33658488 -0.896301325IN.P.BP.IndNum 2 0 0 47 1 0.840928597 -0.541146093 -0.916224298IN.P.BP.IndGend 12 2 0 35 0.833333333 0.26358795 -0.964635368 -0.408055624IN.P.BP.Suppletive 3 1 2 43 0.4 0.950644026 -0.310283639 -0.892136433IN.P.BP.FullAkt 15 2 0 32 0.866666667 0.146680095 -0.989183982 -0.306770662IN.P.BP.PartAkt 1 0 1 47 0.5 0.840059075 -0.542494931 -0.916119457IN.P.BP.RestrAkt 32 0 0 17 1 0.815400825 -0.578896791 -0.322512825KA.R.SC.0 24 1 2 6 0.571428571 0.145426729 -0.989369024 -0.403838328KA.R.SC.Aff 30 0 0 16 1 0.759709206 0.650262964 0.359377534KA.R.BP.PL 31 0 0 15 1 0.815400825 -0.578896791 -0.322512825KA.R.BP.0 15 0 0 31 1 0.815400825 -0.578896791 -0.322512825KA.M.SC.0 30 0 1 16 0.9375 0.807219254 -0.590251706 0.228067302KA.M.SC.Aff 31 0 1 15 0.933333333 0.609497902 -0.792787682 0.2454613KA.M.BP.PL 1 0 0 46 1 0.083924075 -0.996472152 0.865474831KA.M.BP.Intens 13 0 0 31 1 0.077290281 -0.997008632 0.132589649KA.M.BP.0 31 0 0 13 1 0.077290281 -0.997008632 0.132589649KA.P.SC.O 41 1 1 4 0.6 0.919650781 0.392737115 -0.61577581KA.P.SC.CompVerb 6 1 2 38 0.625 0.940998987 -0.338409378 -0.882509908KA.P.BP.TAMDir 11 3 0 33 0.727272727 0.137246678 -0.9905369 -0.318698564KA.P.BP.TAMIndir 6 1 2 38 0.625 0.940998987 -0.338409378 -0.882509908KA.P.BP.0 32 0 0 15 1 0.815400825 -0.578896791 -0.322512825

117

REFERENCESAnderson,LloydB.1974.Distinctsourcesoffuzzydata:waysofintegrating

relativelydiscreteandgradientaspectsoflanguage,andexplaininggrammaronthebasisofsemanticfields.InRogerW.Shuy&Charles-JamesN.Bailey(eds.),Towardstomorrow’slinguistics,50–64.Washington,D.C.:GeorgetownUniversityPress.

Anderson,LloydB.1982.The‘perfect’asauniversalandasalanguage-particular

category.InPaulHopper(ed.),Tense-aspect:Betweensemanticsandpragmatics,227–64.Amsterdam:JohnBenjamins.

Anderson,LloydB.1986.Evidentials,pathsofchange,andmentalmaps:

Typologicallyregularasymmetries.InWallaceChafe&JohannaNichols(eds.),Evidentiality:Thelinguisticencodingofepistemology,273–312.Norwood:Ablex.

Anderson,LloydB.1987.Adjectivalmorphologyandsemanticspace.InBarbara

Need,EricSchiller&AnnBosch(eds.),Papersfromthe23rdAnnualRegionalMeetingoftheChicagoLinguisticSociety,PartOne:TheGeneralSession,1–17.Chicago:ChicagoLinguisticSociety.

Auwera,Johanvander&VladimirA.Plungian.1998.Modality’ssemanticmap.

LinguisticTypology2.79–124.Baayen,R.Harald.2008.Analyzinglinguisticdata:Apracticalintroductionto

statisticsusingR.Cambridge:CambridgeUniversityPress.Bickel,Balthasar.2010.Capturingparticularsanduniversalsinclauselinkage:A

multivariateanalysis.InIsabelleBril(ed.),Clause-hierarchyandclause-linking:Thesyntaxandpragmaticsinterface,51-101.Amsterdam:JohnBenjamins.

Borg,Ingwer&PatrickGroenen.1997.Modernmultidimensionalscaling:Theoryand

applications.NewYork:Springer.Bybee,JoanL.1985.Morphology:Astudyoftherelationbetweenmeaningandform.

Amsterdam:Benjamins.Bybee,JoanL.,ReverePerkins&WilliamPagliuca.1994.Theevolutionofgrammar:

Tense,aspectandmodalityinthelanguagesoftheworld.Chicago:UniversityofChicagoPress.

Chafe,WallaceL.1970.Meaningandthestructureoflanguage.Chicago:Universityof

ChicagoPress.Chafe,WallaceL.(ed.).1980.Thepearstories:Cognitive,cultural,andlinguistic

aspectsofnarrativeproduction.Norwood,NewJersey:Ablex.

118

Childs,GeorgeTucker.1995.AgrammarofKisi:aSouthernAtlanticlanguage.Berlin:

deGruyter.Childs,GeorgeTucker.2000.AdictionaryofKisi.Cologne:RudigerKoppe.Comrie,Bernard.1975.Politepluralsandpredicateagreement.Language51(2).

406-18.Comrie,Bernard.1976.Aspect.Cambridge:CambridgeUniversityPress.Comrie,Bernard.1989.Languageuniversalsandlinguistictypology.2ndedn.Chicago:

UniversityofChicagoPress.Corbett,GrevilleG.2000.Number.Cambridge:CambridgeUniversityPress.Croft,William.1991.Syntacticcategoriesandgrammaticalrelations:Thecognitive

organizationofinformation.Chicago:UniversityofChicagoPress.Croft,William.1996.‘Markedness’and‘universals’:fromthePragueschoolto

typology.InKurtR.Jankowsky(ed.),Multipleperspectivesonthehistoricaldimensionsoflanguage,15-21.Münster:Nodus.

Croft,William.2001.Radicalconstructiongrammar.Oxford:OxfordUniversityPress.Croft,William.2003.Typologyanduniversals.2ndedn.Cambridge:Cambridge

UniversityPress.Croft,William.2012.Verbs:Aspectandcausalstructure.Oxford:OxfordUniversity

Press.Croft,William(toappear).Morphosyntax.Cambridge:CambridgeUniversityPress.Croft,William,HavaBat-ZeevShyldkrot&SuzanneKemmer.1987.Diachronic

semanticprocessesinthemiddlevoice.InAnnaGiacoloneRamat,OnofrioCarruba&GuilianoBernini(eds.),Papersfromthe7thInternationalConferenceonHistoricalLinguistics,179–192.Amsterdam:JohnBenjamins.

Croft,William&KeithT.Poole.2008.Inferringuniversalsfromgrammatical

variation:Multidimensionalscalingfortypologicalanalysis.TheoreticalLinguistics34(1).1-37.

Cyffer,Norbet&JohnHutchison.1990.DictionaryoftheKanurilanguage.

Providence:Foris.

119

Cysouw,Michael.2007.Newapproachestoclusteranalysisoftypologicalindices.InReinhardKöhler&PeterGrzbek(eds.),Exactmethodsinthestudyoflanguageandtext,61-76.Berlin:MoutondeGruyter.

Dahl,Östen.1985.Tenseandaspectsystems.Oxford:BasilBlackwell.Dixon,RobertM.W.1979.Ergativity.Language55(1).59-138.Dixon,RobertM.W.1982.Wherehavealltheadjectivesgone?Berlin:Moutonde

Gruyter.DuBois,JohnW.1985.Competingmotivations.InJohnHaiman(ed.),Iconicityin

syntax,343-65.Amsterdam:JohnBenjamins.Fillmore,CharlesJ.1970.Thegrammarofhittingandbreaking.InRoderickA.Jacobs

&PeterS.Rosenbaum(eds.),ReadingsinEnglishTransformationalGrammar,120–133.Waltham,MA:Ginn.

Givón,Talmy.1979.Onunderstandinggrammar.NewYork:AcademicPress.Greenberg,JosephH.1960.ThegeneralclassificationofCentralandSouthAmerican

languages.InAnthonyWallace(ed.),MenandCultures:Selectedpapersofthe5thInternationalCongressofAnthropologicalandEthnologicalSciences,791-94.Philadelphia,UniversityofPennsylvaniaPress.

Greenberg,JosephH.1966/2005.Languageuniversals:Withspecialreferenceto

featurehierarchies.NewYork:MoutondeGruyter.Greenberg,JosephH.1987.LanguageintheAmericas.Stanford:StanfordUniversity

Press.Greenberg,JosephH.1990.Someuniversalsofgrammarwithparticularreferenceto

theorderofmeaningfulelements.InKeithDenning&SuzanneKemmer(eds.),Onlanguage:SelectedwritingsofJosephH.Greenberg,40-70.Stanford:StanfordUniversityPress.

Haspelmath,Martin.1997a.Indefinitepronouns.Oxford:OxfordUniversityPress.Haspelmath,Martin.1997b.Fromspacetotime:Temporaladverbialsintheworld’s

languages.München:LincomEuropa.Haspelmath,Martin.2003.Thegeometryofgrammaticalmeaning:Semanticmaps

andcross-linguisticcomparison.InMichaelTomasello(ed.),Thenewpsychologyoflanguage,vol.2,211–42.Mahwah,NJ:LawrenceErlbaumAssociates.

120

Hengeveld,Kees.1992.Non-verbalpredication:Theory,typology,diachrony.Berlin:MoutondeGruyter.

Hopper,PaulJ.&SandraA.Thompson.1980.Transitivityingrammaranddiscourse.

Language56(2).251-299.Hopper,PaulJ.&SandraA.Thompson.1984.Thediscoursebasisforlexical

categoriesinuniversalgrammar.Language60(4).703-752.Hutchison,John.1981.TheKanurilanguage:Areferencegrammar.Madison:

UniversityofWisconsin,Madison.Jakobson,Roman.1932/1984.StructureoftheRussianverb.InLindaR.Waugh&

MorrisHale(eds.),RussianandSlavicgrammar:Studies,1931-1981(JanuaLinguarum,SeriesMaior106),1-14.TheHague:Mouton.

Jakobson,Roman.1939/1984.Zerosign.InLindaR.Waugh&MorrisHale(eds.),

RussianandSlavicgrammar:Studies,1931-1981(JanuaLinguarum,SeriesMaior106),151-160.TheHague:Mouton.

Kemmer,Suzanne.1983.Themiddlevoice.(TypologicalStudiesinLanguage,23.)

Amsterdam:JohnBenjamins.Kilian-Hatz,Christa.2003.Khwedictionary.Cologne:RüdigerKoppe.Kilian-Hatz,Christa.2008.AgrammarofmodernKhwe(CentralKhoisan).Cologne:

RüdigerKoppe.Lehmann,Christian&EdithMoravcsik.2000.Nouns.InGeertBooij,Christian

Lehmann&JoachimMugdan(eds.),Morphology:Aninternationalhandbookoninflectionandword-formation,vol.1,732-57.Berlin:deGruyter.

Langacker,RonaldW.1987.Nounsandverbs.Language63(1).53-94.Levin,Beth.1993.Englishverbclassesandalternations:Apreliminaryinvestigation.

Chicago:UniversityofChicagoPress.Levin,Beth(inpress).Verbclasseswithinandacrosslanguages.InBernardComrie

andAndrejMalchukov(eds.),Valencyclasses:Acomparativehandbook.Berlin:DeGruyter.

Levinson,StephenC.&SérgioMeira.2003.'Naturalconcepts'inthespatial

topologicaldomain—adpositionalmeaningsincrosslinguisticperspective:Anexerciseinsemantictypology.Language79(3).485-516.

121

Lichtenberk,Frantisek.2008a.AdictionaryofToqabaqita.Canberra:PacificLinguistics.

Lichtenberk,Frantisek.2008b.AgrammarofToqabaqita.Berlin:deGruyter.Martin,JackB.2011.AgrammarofCreek(Muskogee).Lincoln:Universityof

NebraskaPress.Martin,JackB.&MargaretM.Mauldin.2000.AdictionaryofCreek/Muskogee.

Lincoln:UniversityofNebraskaPress.Nichols,Johanna.2004.Ingush-EnglishandEnglish-Ingushdictionary.NewYork:

RutledgeCurzon.Nichols,Johanna.2011.Ingushgrammar.Berkeley:UniversityofCaliforniaPress.Pensalfini,Robert.1997.Jingulugrammar,dictionary,andtexts.Cambridge:

MassachusettsInstituteofTechnologydoctoraldissertation.Pensalfini,Robert.2003.AgrammarofJingulu.Canberra:PacificLinguistics.Poole,KeithT.2000.Non-parametricunfoldingofbinarychoicedata.Political

Analysis8.211–237.Poole,KeithT.2005.Spatialmodelsofparliamentaryvoting.Cambridge:Cambridge

UniversityPress.Rosch,Eleanor.1978.Principlesofcategorization.InEleanorRosch&BarbaraB.

Lloyd(eds),Cognitionandcategorization,27–48.Hillsdale,NJ:LawrenceErlbaum.

Ross,JohnR.1972.Thecategorysquish:EndstationHauptwort.InPaulM.

Peranteau,JudithN.Levi&GloriaC.Phares(eds.),ProceedingsoftheEighthRegionalMeetingoftheChicagoLinguisticSociety,316-28.Chicago:ChicagoLinguisticSociety.

Ruhlen,Merritt.1991.TheAmerindphylumandtheprehistoryoftheNewWorld.In

SydneyM.Lamb&E.DouglasMitchell(eds.),Sprungfromsomecommonsource,328-50.Stanford:StanfordUniversityPress.

Schachter,Paul&TimothyShopen.2007.Partsofspeechsystems.InTimothy

Shopen(ed.),Languagetypologyandsyntacticdescription,vol.1:Clausestructure,2ndedn.,1-60.Cambridge:CambridgeUniversityPress.

Searle,JohnR.1969.Speechacts:anessayinthephilosophyoflanguage.Cambridge:

CambridgeUniversityPress.

122

Silverstein,Michael.1976.Hierarchyoffeaturesandergativity.InRobertM.W.

Dixon(ed.),GrammaticalcategoriesinAustralianlanguages,112-171.NewJersey:HumanitiesPress.

Smith,CarlotaS.1997.Theparameterofaspect.Dordrecht:KluwerAcademic

Publishers.Stanley,Thomas.1687.Thehistoryofphilosophy.GarlandPub.Stassen,Leon.1997.Intransitivepredication.Oxford:ClarendonPress.Stebbins,TonyaN.2011.Mali(Baining)grammar.Canberra:PacificLinguistics.Stebbins,TonyaN.2012.Mali(Baining)dictionary.Canberra:Asia-PacificLinguistics.Trubetzkoy,Nikolai.1931.DiephonologischenSysteme.TravauxduCercle

LinguistiquedePrague4.96-116.Trubetzkoy,Nikolai.1939/1969.Principlesofphonology.(TranslationofGrundzüge

derPhonologie,Prague,1939).BerkeleyandLosAngeles:UniversityofCaliforniaPress.

Vendler,Zeno.1967.Linguisticsinphilosophy.Ithaca:CornellUniversityPress.Weber,DavidJ.1989.AgrammarofHuallaga(Huánuco)Quechua.Berkeley:

UniversityofCalifornia.Weber,DavidJ.,FélixC.Zambrano,TeodoroC.Villar&MarleneB.Dávila.2008.

Rimaycuna:QuechuadeHuánuco(SerieLingüísticaPeruana48).2ndedn.Lima:InstitutoLingüísticodeVerano.

Wetzer,Harrie.1996.Thetypologyofadjectivalpredication.Berlin:Moutonde

Gruyter.

Illustrating the prototype structures of parts of speech ...

Documents

Transcript of Illustrating the prototype structures of parts of speech ...