Professional Documents
Culture Documents
DistributedMorphology:
FrequentlyAskedQuestionsList
Clickonasubtopicorjustbeginscrollingdown.
General|Categories|Structures|Meaning|SpellOut|Allomorphy|Impoverishment|Morphological
Merger|Clitics|Miscellaneous|Bibliography
General
WhatisDistributedMorphology?
HowisDMdifferentfromothertheoriesofthearchitectureofgrammar?
WhathappenedtotheLexicon?
Categories
Whataremorphemes?
WhatareVocabularyitems?
Whatkindsofmorphemesarethere?
DoesDMuseconventionalsyntacticcategorieslikeNounandVerb?
Structures
Howarethepiecesofwordsputtogether?
Howarethemorphemesofanexpressionputtogether?
Howarethephonologicalpiecesofanexpressionputtogether?
Isthereasyntacticterminalforeverypieceofaword?Don'twehavetoomanyfunctional
projectionsalready?
Meaning
WhatareidiomsinDM?
WhatistheEncyclopedia?
IfVocabularyinsertiondoesnotoccuruntilaftersyntax,andVocabularyisnotpresentat
LF,howisthemeaningofexpressionsdetermined?
HowdothetarolesfigureinDM?
SpellOut
HowdoesSpellOutwork?
HowdoesSpellOutoffmorphemeswork?
HowdoesSpellOutoflmorphemeswork?
WhichVocabularyitemwinsifthefeaturesoftwoVocabularyitemscompetingfor
insertionintothesamemorphemearenotinasubset/supersetrelation?
WhatisFission?
Allomorphy
Howisallomorphyobtained?
WhatcriteriadifferentiatebetweenSuppletionandMorphophonologicalAllomorphy?
IfDMis'piecebased'howis'process'morphologyhandled?
Whatkindoflanguagewouldbepossibleinaprocessbasedmorphologybutimpossiblein
DM?
http://www.ling.upenn.edu/~rnoyer/dm/#DM 1/14
6/4/2017 DistributedMorphology
Impoverishment
WhatisImpoverishment?
WhatkindsofImpoverishmentrulesarethere?
DoesImpoverishmenteverinvolverulesthatchangemorphosyntacticfeaturevalues?
MorphologicalMergerandClitics
WhatisMorphologicalMerger?
HowarecliticsanalyzedinDM?
Miscellaneous
HowdoparadigmsfigureintheDMmodel?
WhatisSeparationism?
IsthereanydifferencebetweeninflectionalandderivationalmorphologyinDM?
HowcanIfindoutmoreaboutDM?
WhatisDistributedMorphology?
DistributedMorphology(DM)isatheoryofthearchitectureofgrammarfirstproposedintheearly1990s
atMITbyMorrisHalle,AlecMarantzandtheirstudentsandcolleaguesincludingEulaliaBonet,Rolf
Noyer,JimHarris,HeidiHarley,AndreaCalabrese,DavidEmbickandothers.ThelocusclassicusforDMis
Halle&Marantz1993,1994.ForfurtherliteratureseeHowcanIfindoutmoreaboutDM?
HowisDMdifferentfromothertheoriesofthearchitectureofgrammar?
AlthoughtherearenumeroushypothesesanddirectionsincurrentDMresearch,threecorepropertiesdefine
thetheory:LateInsertion,Underspecification,andSyntacticHierarchicalStructureAlltheWayDown.
LateInsertionreferstothehypothesisthatthephonologicalexpressionofsyntacticterminalsisinallcases
providedinthemappingtoPhonologicalForm(PF).Inotherwords,syntacticcategoriesarepurely
abstract,havingnophonologicalcontent.Onlyaftersyntaxarephonologicalexpressions,calledVocabulary
Items,insertedinaprocesscalledSpellOut.
UnderspecificationofVocabularyitemsmeansthatphonologicalexpressionsneednotbefullyspecifiedfor
thesyntacticpositionswheretheycanbeinserted.Hencethereisnoneedforthephonologicalpiecesofa
wordtosupplythemorphosyntacticfeaturesofthatwordratherVocabularyitemsareinmanyinstances
defaultsignalsinsertedwherenomorespecificformisavailable.
SyntacticHierarchicalStructureAlltheWayDownentailsthatelementswithinsyntaxandwithin
morphologyenterintothesametypesofconstituentstructures(suchascanbediagrammedthroughbinary
branchingtrees).DMispiecebasedinthesensethattheelementsofbothsyntaxandofmorphologyare
understoodasdiscreteinsteadofas(theresultsof)morphophonologicalprocesses.
WhathappenedtotheLexicon?
ThereisnoLexiconinDMinthesensefamiliarfromgenerativegrammarofthe1970sand1980s.Inother
wordsDMunequivocallyrejectstheLexicalistHypothesis.ThejobsassignedtotheLexiconcomponentin
earliertheoriesaredistributedthroughvariousothercomponents.ForlinguistscommittedtotheLexicalist
Hypothesis,thisaspectofDMmaybethemostdifficulttounderstandortoaccept,butitisneverthelessa
centraltenetofthetheory.
http://www.ling.upenn.edu/~rnoyer/dm/#DM 2/14
6/4/2017 DistributedMorphology
BecausethereisnoLexiconinDM,thetermlexicalitemhasnosignificanceinthetheory,norcananything
besaidto'happenintheLexicon',norcananythingbesaidtobe'lexical'or'lexicalized.'Becauseofthe
greatmanytaskswhichtheLexiconwassupposedtoperform,theterms'lexical'and'lexicalized'areinfact
ambiguous(Aronoff1994).
Lexical(ized)=Idiomatized.BecausetheLexiconwassupposedtobeastorehouseforsoundmeaning
correspondences,ifanexpressionissaidtobe'lexicalized'theintendedmeaningmaybethattheexpression
islistedwithaspecializedmeaning.InDMsuchanexpressionisanidiomandrequiresanEncyclopedia
entry.
Lexical(ized)=NotconstructedbySyntax.Theinternalstructureofexpressionsisnotalwaysaproductof
syntacticoperations.InDMstructurecanbeproducedbothinsyntaxandaftersyntaxinacomponentcalled
Morphology(seeHowarethepiecesofwordsputtogether?).Nevertheless,becauseof'Syntactic
HierarchicalStructurealltheWayDown',operationswithinMorphologystillmanipulatewhatareessentially
syntacticstructuralrelations.
Lexical(ized)=Notsubjecttoexceptionlessphonologicalprocesses,i.e.partof'lexical'phonologyinthe
theoryofLexicalPhonologyandMorphology(Kiparsky1982etseq.).
InDMthedistinctionbetweentwotypesofphonology'lexical'and'postlexical'isabandoned.All
phonologyoccursinasinglepostsyntacticmodule.WhileLexicalPhonologyandMorphologyproduced
manyimportantinsights,DMdeniesthattheseresultsrequireanarchitectureofgrammarwhichdivides
phonologyintoapresyntacticandpostsyntacticmodule.Rather,postsyntacticPhonologyitselfmayhavea
complexinternalstructure(Halle&Vergnaud1987).
Whataremorphemes?
InDM,thetermmorphemeproperlyreferstoasyntactic(ormorphological)terminalnodeanditscontent,
nottothephonologicalexpressionofthatterminal,whichisprovidedaspartofaVocabularyitem.
Morphemesarethustheatomsofmorphosyntacticrepresentation.Thecontentofamorphemeactivein
syntaxconsistsofsyntacticosemanticfeaturesdrawnfromasetmadeavailablebyUniversalGrammar.
WhatareVocabularyitems?
AVocabularyitemis,properlyspeaking,arelationbetweenaphonologicalstringor'piece'andinformation
aboutwherethatpiecemaybeinserted.Vocabularyitemsprovidethesetofphonologicalsignalsavailable
inalanguagefortheexpressionofabstractmorphemes.ThesetofallVocabularyItemsiscalledthe
Vocabulary.
Vocabularyitemschema
signal<>contextofinsertion
(phonologicalexponent)
Examplevocabularyitems
/i/<>[___,+plural]
ARussianaffix(Halle1997)
/n/<>[___,+participant+speaker,plural]
AcliticinBarceloniCatalan(Harris1997a)
/y/<>elsewhere
AnaffixintheUgariticprefixconjugation(Noyer1997)
http://www.ling.upenn.edu/~rnoyer/dm/#DM 3/14
6/4/2017 DistributedMorphology
zero<>2plu
AsubpartofacliticinIberianSpanish(Harris1994)
NotethatthephonologicalcontentofaVocabularyitemmayanyphonologicalstring,includingzeroor
'null'.Thefeaturalcontentorcontextofinsertionmaybesimilarlydevoidofinformation:insuchcaseswe
speakofthedefaultorelsewhereVocabularyitem.
Whatkindsofmorphemesarethere?
InearlyworkinDM,Halle1992proposedadistinctionbetweenconcretemorphemes,whosephonological
expressionwasfixed,andabstractmorphemes,whosephonologicalexpressionwasdelayeduntilafter
syntax.CanonicalDMhoweverendorsesLateInsertionofallphonologicalexpression,soHalle'searlier
concretevs.abstractdistinctionhasbeenmodifiedinrecentwork.
Harley&Noyer1998proposethatmorphemesareoftwobasickinds:fmorphemesandlmorphemes,
correspondingapproximatelytotheconventionaldivisionbetweenfunctionalandlexicalcategories.
FmorphemesaredefinedasmorphemesforwhichthereisnochoiceastoVocabularyinsertion.Inothe
words,fmorphemesarethosewhosecontentsufficestodetermineauniquephonologicalexpression.The
spelloutofanfmorphemeissaidtobedeterministic.
Incontrast,anlmorphemeisdefinedasoneforwhichthereisachoiceinspellout.Forexample,inanl
morphemecorrespondingtowhatwouldbepretheoreticallycalleda'noun'theremightbeinsertedthepieces
dog,cat,fish,mouse,tableetc.Becausethelabelsnoun,verb,adjectiveetc.arebyhypothesisnotpresentin
syntax,thewidelyadoptedhypothesisthatProsodicDomainconstructionshouldbeoblivioustosuch
distinctions(Selkirk1986,Chen1987)followsautomatically.
DoesDMuseconventionalsyntacticcategorieslikeNounandVerb?
Arelatedhypothesis(Marantz1997a,Embick1997,1998a,1998b,Harley&Noyer1998,toappear)
contendsthatthetraditionaltermsforsentenceelements,suchasnoun,verb,adjectivehavenouniversal
significanceandareessentiallyderivativefrommorebasicmorphemetypes(seealsoSapir1921,ch.5).
Marantz1997acontendsthattheconfigurationaldefinitionofcategorylabelsisalreadyimplicitinChomsky
1970,andthatChomsky'sclassicargumentsin'RemarksonNominalization'donotinfactentailthe
LexicalistHypothesisaslaterinterpreted.
Specifically,thedifferent'partsofspeech'canbedefinedasasinglelmorphemetype,calledRoot
(Pesetsky1995),incertainlocalrelationswithcategorydefiningfmorphemes.Forexample,a'noun'ora
'nominalization'isaRootwhosenearestccommandingfmorpheme(orlicenser)isaDeterminer,a'verb'is
aRootwhosenearestccommandingfmorphemesarev,AspectandTensewithoutTensesuchaRootis
simplya'participle'.
Thus,thesameVocabularyitemmayappearindifferentmorphologicalcategoriesdependingonthesyntactic
contextthattheitem'slmorpheme(orRoot)appearsin.Forexample,theVocabularyitemdestroyappears
asa'noun'destruct(ion)whenitsnearestlicenserisaDeterminer,butthesameVocabularyitemappearsasa
'participle'destroy(ing)whenitsnearestlicensersareAspectandvifTenseappearsimmediatelyabove
Aspect,thenthe'participle'becomesa'verb'suchasdestroy(s).
However,itisprobablythecasethatmanytraditionalpartsofspeechlabelscorrespondtolanguagespecific
featurespresentaftersyntaxconditioningvariousmorphologicaloperationssuchasImpoverishmentand
VocabularyInsertion.
Howarethepiecesofwordsputtogether?
http://www.ling.upenn.edu/~rnoyer/dm/#DM 4/14
6/4/2017 DistributedMorphology
Thisisreallytwoquestions,dependingonwhatismeantbya'pieceofaword.'InDManygiven
expressionacquiresatleasttwostructuraldescriptionsduringitsderivation.Inamorphophonological
description,anexpression'sphonologicalpieces(itsVocabularyitems)andtheirconstituentstructureare
displayed.Inamorphosyntacticdescription,anexpression'smorphemesandtheirconstituentstructure
aredisplayed.
Theexpressioncows:
Morphosyntacticdescription:[Root[+plural]]
Morphophonologicaldescription:[kaw+z]
Howarethemorphemesofanexpressionputtogether?
Themorphosyntacticstructureofanexpressionisgeneratedbyseveralmechanisms.Syntax,includingthe
operationofheadmovement,playsamajorroleinconstructingmorphosyntacticstructures,including'word'
internalstructure.Butinaddition,DMemploysseveraladditionalmechanisms.
DissociatedMorphemes.First,morphemessuchas'passive'or'case'(insomeinstances,seeMarantz1991)
which,byhypothesis,donotfigureinsyntaxproper,canbeinsertedaftersyntaxbutbeforeSpellOut.
Thesemorphemes,whichonlyindirectlyreflectsyntacticstructures,arecalledDissociatedmorphemes
(Embick1997).
Second,theconstituentstructureofmorphemescanbemodifiedbyMorphologicalMerger,whichcaneffect
relativelylocalmorphemedisplacements.
Howarethephonologicalpiecesofanexpressionputtogether?
SpellOutinsertsVocabularyitems(phonologicalpieces)intomorphemes.Intheunmarkedcase,the
relationbetweenvocabularyitemsandmorphemesisonetoone,butseveralfactorsmaydisruptthisrelation
(Noyer1997),includingFissionofmorphemesandlocaldisplacementsofVocabularyitemsby
MorphologicalMerger.
Isthereasyntacticterminalforeverypieceofaword?Don'twehavetoomany
functionalprojectionsalready?
Intheearly1990ssomelinguistslookedonwithapprehensionatthe'explosion'ofINFLandtheproposalof
moreandmorenewsyntacticprojections.InDM,becausedissociatedmorphemescanbeinsertedafter
Syntax,noteverymorphemecorrespondstoasyntacticterminal.Ratheritremainsasalwaysanopen
questionwhatthesetofsyntacticterminalsisandhowtheserelatetothemorphophonologicalformofan
utterance.Inaddition,FissionofmorphemesduringSpelloutinsomecasesallowsmultiplephonological
piecestocorrespondtosinglemorphemes,furtherobscuringthemorphosyntacticstructure.Nevertheless,
thesedeparturesareconsideredmarkedoptionswithinagrammar,andthereforeareassumedtorequire
(substantial)positiveevidenceduringacquisition.
WhatareidiomsinDM?
Thetermidiomisusedtorefertoanyexpression(evenasinglewordorsubpartofaword)whosemeaning
isnotwhollythepredictablefromitsmorphosyntacticstructuraldescription(Marantz1995,1997a).F
morphemesaretypicallynotidioms,butlmorphemesarealwaysidioms.
Someidioms
http://www.ling.upenn.edu/~rnoyer/dm/#DM 5/14
6/4/2017 DistributedMorphology
cat(afuzzyanimal)
(the)veil(vowsofanun)
(rain)catsanddogs(alot)
(talk)turkey(honestdiscourse)
Foranalternative,nonDManalysisofidioms,seeJackendoff1997.
WhatistheEncyclopedia?
TheEncyclopediacontainsEncyclopediaentriestheserelateVocabularyitems(sometimesinthecontextof
otherVocabularyitems)tomeanings.Inotherwords,theEncyclopediaisthelistofidiomsinalanguage.
IfVocabularyinsertiondoesnotoccuruntilaftersyntax,andVocabularyisnotpresent
atLF,howisthemeaningofexpressionsdetermined?
Themeaningofanexpressionisinterpretedfromtheentirederivationofthatexpression.LFdoesnot
expressorrepresentmeaningLFismerelyalevelofrepresentationwhichexhibitscertainmeaningrelated
structuralrelations,suchasquantifierscope.SeeMarantz1995.
DothetarolesfigureinDM?
MostworkinDMdoesnotrecognizeasetofdiscretethematicroles.Instead,followingtheinsightsofHale
&Keyser1993,1998,thematicrolesarereducedtostructuralconfigurations.Forexample,Harley1995
proposesthatAgentistheinterpretationgiventoanyargumentprojectedintothespecifierofEventPhrase
(seealsoTravis1994on"EventPhrase",andKratzer1996forrelatedideas).Themeistheinterpretation
giventoanyargumentprojectedasasisterofRoot.UnlikeHale&Keyser1993,however,DMdoesnot
differentiatebetweenan'lsyntax'occuringintheLexiconandaregular'ssyntax.'Botharesimplyone
module,Syntax.SeealsoMarantz1997a.
HowdoesSpellOutwork?
SpellOut,alsocalledVocabularyInsertion,worksdifferentlydependingonwhattypeofmorphemeis
beingspelledout,fmorphemesorlmorphemes.Regardlessofthetypeofmorpheme,however,SpellOutis
normallytakentoinvolvetheassociationofphonologicalpieces(Vocabularyitems)withabstract
morphemes.Halle1992construesSpellOutastherewritingofaplaceholder'Q'inamorphemeas
phonologicalmaterial.Thisoperationisnormallyunderstoodascyclic,suchthatmoredeeplyembedded
morphemesarespelloutfirst.
HowdoesSpellOutoffmorphemeswork?
EarlyworkinDMwasfocusedprimarilyonthespelloutoffmorphemes.InsuchcasessetsofVocabulary
itemscompeteforinsertion,subjecttotheSubsetPrinciple(Halle1997).
SubsetPrinciple.'ThephonologicalexponentofaVocabularyitemisinsertedintoa
morpheme...iftheitemmatchesallorasubsetofthegrammaticalfeaturesspecifiedinthe
terminalmorpheme.InsertiondoesnottakeplaceiftheVocabularyitemcontainsfeaturesnot
presentinthemorpheme.WhereseveralVocabularyitemsmeettheconditionsforinsertion,the
itemmatchingthegreatestnumberoffeaturesspecifiedintheterminalmorphememustbe
chosen.'
Example(Sauerland1995).
Dutchstrongadjectivaldesinences
http://www.ling.upenn.edu/~rnoyer/dm/#DM 6/14
6/4/2017 DistributedMorphology
[neuter][+neuter]
[pl]ezero
[+pl]ee
VocabularyItems
zero<>[___,+neuterplural]/Adj+____
e<>Adj+____
InDutch,aftersyntax,adissociatedmorphemeisinsertedasarightadjunctofmorphemeswhichare
conventionallylabeled'adjectives.'TheVocabularyitemsabovecompeteforinsertionintothismorpheme.
Inthespecificenvironmentoftheneutersingular,zeroisinserted.Intheremainingorelsewhere
environmenteisinserted.Theinsertionofzerointhespecificenvironmentbleedstheinsertionofe
because,undernormalcircumstances,onlyasingleVocabularyItemmaybeinsertedintoamorpheme.Note
thattheVocabularyitemsabovearenotspeciallystipulatedtobedisjunctiveexceptinsofarastheycompete
forinsertionatthesamemorpheme.
HowdoesSpellOutoflmorphemeswork?
ForlmorphemesthereisachoiceregardingwhichVocabularyitemisinserted.Forexample,aRoot
morphemeinanappropriatelylocalrelationtoaDeterminermightbefilledbycat,dog,house,tableorany
otherVocabularyitemwewouldnormallycalla'noun.'Harley&Noyer1998proposethatsuchVocabulary
itemsarenotincompetition,asaretheVocabularyitemsinsertedintofmorphemes.Rather,these
VocabularyitemscanbefreelyinsertedatSpellOutsubjecttoconditionsoflicensing.Licensersare
typicallyfmorphemesincertainstructuralrelationstotheRootwheretheVocabularyitemisinserted.
'Nouns'arelicensedbyDeterminerdifferentverbclasses,suchasunergatives,unaccusatives,and
transitives,eacharelicensedbydifferentstructuralconfigurationsandrelationstovarioushighereventuality
projections.SeealsoDoesDMuseconventionalsyntacticcategorieslikeNounandVerb?
WhichVocabularyitemwinsifthefeaturesoftwoVocabularyitemscompetingfor
insertionintothesamemorphemearenotinasubset/supersetrelation?
Insomecasesitwouldbepossibletoinserttwo(ormore)Vocabularyitemsintothesamemorpheme,and
theSubsetPrincipledoesnotdeterminethewinner.TwoapproacheshavebeenproposedinDMforsuch
cases.Halle&Marantz1993suggestthatsuchconflictsareresolvedbyextrinsicordering:oneVocabulary
itemissimplystipulatedasthewinner.Alternatively,Noyer1997proposesthatsuchconflictscanalwaysbe
resolvedbyappealtoaUniversalHierarchyofFeatures(cf.alsoLumsden1987,1992,Zwicky1977and
Silverstein1976).Specifically,theVocabularyItemthatuniquelyhasthehighestfeatureinthehierarchyis
inserted.
FragmentoftheHierarchyofFeatures
1person>2person>dual>plural>otherfeatures
SeealsoHarley1994.
Howisallomorphyobtained?
DMrecognizestwodifferenttypesofallomorphy:suppletiveandmorphophonological.
SuppletiveallomorphyoccurswheredifferentVocabularyitemscompeteforinsertionintoanf
morpheme.Forexample,Dutchnounshave(atleast)twopluralnumbersuffixes,enands.The
conditionsforthechoicearepartlyphonologicalandpartlyidiosyncratic.Sinceenandsarenotplausibly
relatedphonologically,theymustconstitutetwoVocabularyitemsincompetition.
http://www.ling.upenn.edu/~rnoyer/dm/#DM 7/14
6/4/2017 DistributedMorphology
MorphophonologicalallomorphyoccurswhereasingleVocabularyitemhasvariousphonologically
similarunderlyingforms,butwherethesimilarityisnotsuchthatPhonologycanbedirectlyresponsiblefor
thevariation.Forexample,destroyanddestructrepresentstemallomorphsofasingleVocabularyitemthe
latterallomorphoccursinthenominalizationcontext.DMhypothesizesthatinsuchcasesthereisasingle
basicallomorph,andtheothersarederivedfromitbyaruleofReadjustment.TheReadjustmentinthis
casereplacestheRimeofthefinalsyllableofdestroywithuct.
WhatcriteriadifferentiatebetweenSuppletionandMorphophonologicalAllomorphy?
Traditionallyitisoftenthoughtthatthereisagradientbetweensuppletionandothertypesofmore
phonologicallyregularallomorphy,andthatnoreasonablegroundscanbegivenforhowtodividethetwoor
iftheyshouldbedividedatall.Marantz1997bhasrecentlyproposedthattruesuppletionoccursonlyfor
Vocabularyitemsincompetitionforfmorphemes,sincecompetitionoccursonlyforfmorphemes(How
doesSpellOutoflmorphemeswork?).Animmediateconsequenceisthatundeniablysuppletivepairslike
go/wentorbad/worsemustactuallyrepresentthespellingoffmorphemes.Theclassoffmorphemesisasa
resultconsiderablyenriched,butsincetheclassoffmorphemesiscircumscribedbyUniversalGrammar,it
isalsopredictedthattruesuppletionshouldbelimitedtouniversalsyntacticosemanticcategories.Moreover,
giventhatsomeindependentgroundsmightinthiswaydividesuppletivefromReadjustmentdriven
allomorphy,atheoryoftherangeofpossibleReadjustmentprocessesbecomesmorefeasible.
IfDMis'piecebased'howis'process'morphologyhandled?
DMispiecebasedinasmuchasVocabularyitemsareconsidereddiscretecollectionsofphonological
materialandnot(theresultof)phonologicalprocesses(asinAnderson1992).NeverthelessReadjustment
canaltertheshapeofindividualVocabularyitemsinappropriatecontexts.TwofactorsthusdistinguishDM
fromprocessonlytheoriesofmorphology.
First,sinceReadjustmentcanaffectonlyindividualVocabularyitemsandnotstringsofthese,itispredicted
that'process'morphologyisalwaysakindofallomorphy(seealsoLieber1981).Forexample,Marantz
1992showsthattruncationappliesto(Papago)O'odhamverbstemstoproduceaseparatestemallomorphit
doesnotaffectmorethanoneVocabularyitematonce.
Second,sinceprocessesproduceallomorphsbutdonotdirectly'discharge'features,itiscommonforan
allomorphtohaveseveralcontextsofuse.Forexample,inPapagothetruncatedverbstemallomorphhas
severalfunctions,includingbutnotlimitedtoitsuseintheperfectiveform,andthepropertyofperfectivity
isprimarilyexpressedinanothermorpheme,namelyanaffixonthesyntacticauxiliary.Itistherefore
incorrecttodirectlyequatetruncationandtheperfectiverather,truncationappliestoverbstemswhich
appearintheperfective.
Whatkindoflanguagewouldbepossibleinaprocessbasedmorphologybutimpossible
inDM?
Sinceprocessmorphologycaninprincipleapplytoanystring,regardlessofitsmorphologicalderivation,it
ispredictedthatalanguagecouldmarkthecategoryPluralbydeletionofafinalsyllable,regardlessof
whetherthatsyllableconsistedofoneorseveraldiscretephonologicalpieces.Consider'Martian'below:
Singularandpluralnounsinthepseudolanguage'Martian'
singularplural
takatatakaearthling
takataritakataearthlinggenitive
laamilaaantenna
jankapjanflyingsaucer
jankaprijankaflyingsaucergenitive
http://www.ling.upenn.edu/~rnoyer/dm/#DM 8/14
6/4/2017 DistributedMorphology
zuuklorpcanal
zuukrizuucanalgenitive
yuuniyuuantennawaving
(cf.yuun'waveantennas')
mergimercanaldigging
(cf.merg'digacanal')
mergirimergicanaldigginggenitive
In'Martian',nominalizationscanbeformedfromnounsstemsbyadditionofthesuffix(i)andgenitives
withthesuffix(ri).Regardlessofthederivationofanoun,thepluralisalwayseitheratruncationofthelast
syllableofthesingular,orsuppletive(zuuk~lorp).Thetruncatedformneveroccursanywhereelseexcept
inplurals.Numbermarkinghasnootherexpressionthantruncation.
The'Martian'ruleofpluralformationiseasytoexpressinaprocessmorphology:insteadofaddinganaffix,
onesimplydeletesthefinalsyllable.InDMhowever,thislanguagecouldneverbegenerated,because
processeslike'deletethefinalsyllable'couldonlybeexpressedasReadjustmentswhichaffectindividual
Vocabularyitems.
WhatisMorphologicalMerger?
MorphologicalMerger,proposedfirstinMarantz1984,wasoriginallyaprincipleofwellformedness
betweenlevelsofrepresentationinsyntax.InMarantz1988:261Mergerwasgeneralizedasfollows:
MorphologicalMerger
Atanylevelofsyntacticanalysis(dstructure,sstructure,phonologicalstructure),arelation
betweenXandYmaybereplacedby(expressedby)theaffixationofthelexicalheadofXto
thelexicalheadofY.
WhatMergerdoesisessentially'trade'or'exchange'astructuralrelationbetweentwoelementsatonelevel
ofrepresentationforadifferentstructuralrelationatasubsequentlevel.
Mergerhasdifferentconsequencesdependinguponthelevelofrepresentationitoccursat.WhereMerger
appliesinsyntaxproperitistheequivalentofHeadMovement,adjoiningazerolevelprojectiontoa
governingzerolevelprojection(Baker1988).SyntacticLoweringmaybeatypeofMergeraswell,
presumablyoccuringaftersyntaxproperbutbeforeVocabularyInsertion.SeeBobaljik1994.
ThecanonicaluseofMergerinMorphologyistoexpresssecondpositioneffects.Embick&Noyer(in
progress)hypothesizethatwhereMergerinvolvesparticularVocabularyitems(asopposedtomorphemes),
theitemsinquestionmustbestringadjacentsuchcasesofMergerarecalledLocalDislocation.
SchematicallyLocalDislocationlookslikethis:
LocalDislocation:
X[Y...]>[Y+X...
InLocalDislocation,azerolevelelementtradesitsrelationofadjacencytoafollowingconstituentwitha
relationofaffixationtothelinearhead(peripheralzeroelement)ofthatconstituent.
LocalDislocationhasalsoreceivedconsiderableattentionoutsideofDMfromresearchersworkingin
AutolexicalSyntax(Sadock1991).
Forexample,Latinqueisasecondpositioncliticwhichadjoinstotheleftofthezerolevelelementtoits
right(*representstherelationofstringadjacencyQrepresentsdissociatedmorphemes):
[[AQ][NQ]][cl[[AQ][NQ]]]Mophologicalstructure
[[boni][pueri]][que[[bonae][puellae]]]Vocabularyinsertion
[[bon*i]*[puer*i]]*[que*[[bon*ae]*[puell*ae]]]Linearization
[[bon*i]*[puer*i]]*[[[bon*ae]*que]*[puell*ae]]]Localdislocation
http://www.ling.upenn.edu/~rnoyer/dm/#DM 9/14
6/4/2017 DistributedMorphology
goodnom.plboynom.plgoodnom.plandgirlnom.pl
'Goodboysandgoodgirls'
Byhypothesis,ProsodicInversion(Halpern1995)isadistinctspeciesofMergeratthelevelofPF,and
differsfromLocalDislocationinthattheaffectedelementsareprosodiccategoriesratherthanmorphological
ones.
Forexample,Schuetze1994,expandingonZec&Inkelas1990,arguesthattheauxiliarycliticjeinSerbo
CroatianissyntacticallyinC,butinvertswiththefollowingPhonologicalWordbyProsodicInversionatPF
(parenthesesbelowdenotePhonologicalWordboundaries):
je[[UovojsobiPP]klavirVP]syntacticstructure
je(Uovoj)(sobi)(klavir)parseintoPhonologicalWords
((Uovoj)+je)(sobi)(klavir)ProsodicInversion
InthisAUXroompiano
'Inthisroomisthepiano'
Byhypothesis,thepositioningofthecliticcannotbestatedintermsofa(morpho)syntacticconstituent,since
Uovoj'inthis'doesnotformsuchaconstituent.Izvorski&Embick1995specificallyarguethatsyntactic
explanations,includingthoseinvolvingremnantextraposition,cannotreasonablybeheldaccountablefor
thispattern.
However,itshouldbeemphasizedthattheextenttowhichLocalDislocationandProsodicInversionare
distinctdevicesinthemappingtoPFremainscontroversial.
WhatisImpoverishment?
Impoverishment,firstproposedinBonet1991,isanoperationonthecontentsofmorphemespriortoSpell
Out.InearlyworkinDM,Impoverishmentsimplyinvolvedthedeletionofmorphosyntacticfeaturesfrom
morphemesincertaincontexts.Whencertainfeaturesaredeleted,theinsertionofVocabularyitems
requiringthosefeaturesforinsertioncannotoccur,andalessspecifieditemwillbeinsertedinstead.Halle&
MarantztermedthistheRetreattotheGeneralCase.
Example(Sauerland1995).
TheadjectivalsufixesinNorwegian.
STRONG[neuter][+neuter]
[pl]zerot
[+pl]ee
WEAK[neuter][+neuter]
[pl]ee
[+pl]ee
InNorwegian,thereisathreewaydistinction(t~e~zero)inadjectivalsuffixesina'strong'syntactic
position,butintheweakpositiononefindsonlye.Byhypothesis,itisnotaccidentalthatthesameaffixe
appearsintheweakcontextasistheelsewhereorlessmarkedaffixinthestrongcontext.Sauerland1995
proposesthefollowingsetofVocabularyitems:
/t/<>[___,pl+neut]/Adj+____
zero<>[___,plneut]/Adj+____
/e/<>elsewhere/Adj+____
Intheweaksyntacticposition,aruleofImpoverishmentapplies,deletinganyvaluesforgenderfeatures:
[neuter]>null
http://www.ling.upenn.edu/~rnoyer/dm/#DM 10/14
6/4/2017 DistributedMorphology
ImpoverishmentthusguaranteesthatboththeVocabularyitemstandzerocannotbeinserted,sinceboth
requireexplicitreferencetoavaluefor[neuter].Insertionofthegeneralcase,namelye,follows
automatically.
WhatkindsofImpoverishmentrulesarethere?
InBonet'soriginalproposal(1991)andinseveralsubsequentworks(Harley1994,Harris1997a,Ritter&
Harley1998),morphosyntacticfeaturesarearrangedinafeaturegeometrymuchlikephonologicalfeatures,
andImpoverishmentisrepresentedasdelinking.Consequently,thedelinkingofcertainfeaturesentailsthe
delinkingoffeaturesdependentonthem.Forexample,ifpersonfeaturesdominatenumberfeatureswhichin
turndominategenderfeatures,thentheImpoverishment(delinking)ofnumberentailsthedelinkingof
genderaswell:
2>2
|
=
|
pl
|
f
Noyer1997rejectstheuseofgeometriesofthissortastoorestrictive,andproposesinsteadthat
Impoverishmentsarebetterunderstoodasfeaturecooccurrencerestrictionsorfiltersofthetypeemployed
byCalabrese1995forphonologicalsegmentinventories.Forexample,theabsenceofafirstpersondualin
Arabicisrepresentedasthefilter*[1dual],andaUniversalHierarchyofFeaturesdictatesthatwherethese
featurescombine,because[dual]isanumberfeatureand[1]isa(hierarchicallyhigher)personfeature,
[dual]isdeletedautomatically.Calabrese1994and1996furtherexpandthisidea.
TheuseoffeaturegeometriesinDMremainsanunresolvedissueatthistime,butFeatureHierarchies,
whethergeometricornot,ensurethatnormallymoremarkedfeaturevaluespersistincontextsof
neutralization.
DoesImpoverishmenteverinvolverulesthatchangemorphosyntacticfeaturevalues?
FeaturechangingImpoverishment,whichasadevicehasapproximatelythesamepowerasRulesof
Referral(Zwicky1985b,Stump1993),hasingeneralbeeneschewedinDM.However,Noyer1998
discussescaseswherefeaturechangingreadjustmentsseemnecessary.Itisproposedthatsuchcasesalways
involveachangefromthemoremarkedvalueofafeaturetothelessmarkedvalueandneverviceversa.
WhatisFission?
FissionwasoriginallyproposedinNoyer1997toaccountforsituationsinwhichasinglemorphememay
correspondtomorethanoneVocabularyitem.Inthenormalsituation,onlyoneVocabularyitemmaybe
insertedintoanygivenmorpheme.ButwhereFissionoccurs,VocabularyInsertiondoesnotstopaftera
singleVocabularyItemisinserted.Rather,Vocabularyitemsaccreteonthesisterofthefissionedmorpheme
untilallVocabularyitemswhichcanbeinsertedhavebeenorallfeaturesofthemorphemehavebeen
discharged.AfeatureisdischargedwhentheinsertionofaVocabularyitemisprimarilyconditionedbythat
feature.
Forexample,intheprefixconjugationofTamazightBerber,theAGRmorphemecanappearasone,twoor
threeseparateVocabularyitems,andthesemayappearasprefixesorassuffixes:
TamazightBerberPrefixConjugation.dawa'cure'
singularplural
http://www.ling.upenn.edu/~rnoyer/dm/#DM 11/14
6/4/2017 DistributedMorphology
3midawadawan
3ftdawadawant
2mtdawadtdawam
2ftdawadtdawant
1dawagndawa
Vocabularyitems
/n/<>1pl
/g/<>1
/t/<>2
/t/<>3sgf
/m/<>plm(2)
/i/<>sgm
/d/<>sg(2)
/n/<>pl
/t/<>f
Inafissionedmorpheme,Vocabularyitemsarenolongerincompetitionforasinglepositionofexponence,
i.e.forthepositionofthemorphemeitself.Rather,anadditionalpositionofexponenceisautomatically
madeavailablewheneveraVocabularyitemisinserted(seeHalle1997foraslightlydifferentview).
Inaformliketdawant'you(fempl)cure'hasthreeaffixes,t,n,andt.Theaffixesareaddedinanorder
determinedbytheFeatureHierarchy.Hencet'2'isaddedfirst,thenn'plural',andfinallyt'feminine.'
Inaformlikendawa'wecure'thereisbutoneaffix.Bydischargingthefeature'1',theinsertionofn'1pl'
preventsthesubsequentinsertionofg'1'.ThisillustratesthattwoVocabularyitemscanbedisjunctivenot
bycompetingforthesamepositionofexponence,butratherbycompetingforthedischargeofthesame
feature.SuchcasesaretermedDiscontinuousBleeding.
SomefeaturesintheaboveVocabularyitemlistareinparentheses.Thisnotationdenotesthatthe
Vocabularyiteminquestioncanbeinsertedonlyiftheparenthesizedfeaturehasalreadybeendischarged,
whereasthefeatureswhicharenotinparenthesescannotalreadyhavebeendischargedifinsertionisto
occur.Forexample,mcanbeinsertedonlyonaverbtowhicht'2'hasalreadybeenattached.Thus
featuresconditioningtheinsertionofaVocabularyitemcomeintwotypes.AVocabularyitemprimarily
expressesthenonparenthesizedfeaturesinitsentry,butitsecondarilyexpressestheparenthesizedfeatures
(oranyfeaturesbelongingtoothermorphemeswhicharerequiredforinsertionoftheitem).Inotherwords,
theparenthesizedfeaturesmustbeprimarilyexpressedbysomeotherVocabularyitem.Thisdistinction
corresponds(approximately)tothedistinctionbetweenprimaryandsecondaryexponence(Carstairs
1987).
HowdoparadigmsfigureintheDMmodel?
Aparadigmisacollectionofrelatedwordsinsometheoriescertainsuchcollectionshaveaprivileged
statusandcanbereferredtobystatementsofthegrammar.InDMparadigms,likecollectionsofrelated
phrasesorsentences,donothaveanystatusastheoreticalobjects,althoughcertainregularitiesobtaining
overparadigmsmayresultfromconstraintsoperatingduringlanguageacquisition.
WhatisSeparationism?
Separationismcharacterizestheoriesofmorphologyinwhichthemechanismsforproducingtheformof
syntacticosemanticallycomplexexpressionsareseparatedfrom,andnotnecessarilyinasimple
correspondencewith,themechanismswhichproducetheform('spelling')ofthecorrepondingphonological
http://www.ling.upenn.edu/~rnoyer/dm/#DM 12/14
6/4/2017 DistributedMorphology
expressions.DMendorsesavarietyofSeparationism,asdoesLexemeMorphemeBaseMorphology
developedbyRobertBeard(e.g.Beard1995).
TheoriesendorsingSeparationismareattractivebecause(a)theyallowsimilarsyntacticosemanticformsto
berealizedinquitedifferentwaysphonologicallyand(b)theypermitpolyfunctionalityofphonological
expressions:asinglephonologicalpiece(e.g.theEnglishaffixs)mightcorrespondtoasetofdistinctand
unrelatedsyntacticosemanticfunctions.
TheoriesendorsingSeparationareunattractiveforexactlythesamereasonsasabove:whenunconstrained,
theyfailtomakeanyinterestingpredictionsaboutthedegreetowhichsyntacticosemanticandphonological
formcandiverge.SeeEmbick1997,1998a,1998b.
IsthereanydifferencebetweeninflectionalandderivationalmorphologyinDM?
ThiscontroversialdistinctionhasnoexplicitstatusinDM.However,DMdoesdistinguishbetweenf
morphemesandlmorphemes(notallfmorphemeswouldnormallybeconsidered'inflectional'however)as
wellasbetweensyntacticandnonsyntactic(dissociated)morphemes.
HowarecliticsanalyzedinDM?
'Clitic'isnotaprimitivetypeinDMbutratherabehaviorwhichanelementmaydisplay.Conventionally,
cliticsaresaidto'lean'ona'host'thissortofdependencyrelationofoneelementonanothermanifestsitself
differentlydependingonwhattheelementisandwhereitsdependencyrelationmustbesatisfied.Hence
thereisnocoherentclassofobjectswhichcanbetermedcliticsinsteadmorphemesandVocabularyitems
mayshowarangeofdependencies.
Leaners(Zwicky1985a)areVocabularyitemswhichcannotformPhonologicalWordsbythemselvesbut
whosemorphemeshavenootherspecialdisplacementproperties.Forexample,theEnglishreduced
auxiliarys(fromis)'promiscuously'attachestoanyphonologicalhosttoitsleft(Zwicky&Pullum1983):
ThepersonIwastalkingto'sgoingtobeangrywithme.
Anyanswernotentirelyright'sgoingtobemarkedasanerror.
Selkirk1996analyzesprosodicallydependentfunctionwordsaseitherfreeclitics(adjunctstoPhonological
Phrases),affixalclitics(adjunctstoPhonologicalWords)orinternalclitics(incorporatedintoPhonological
Words).Theseoptionsareshownschematicallybelow:
Typesofphonologicalclitics
{=phraseboundary,(=wordboundary
{...freeclitic{(host)...}}
{...(affixalclitic(host))...}
{...(internalclitic+host)...}
Englishleanersaretypicallyfreeclitics,accordingtoSelkirk,butotherlanguagesexploitotheroptions.For
example,Embick1995showsthat,dependingontheirsyntacticprovenance,Polishcliticsbehave
phonologicallyaseitheraffixalclitics(allowingtheirhosttoundergoworddomainphonology),oras
internalclitics(preventingtheirhostfromundergoingworddomainphonologyonitsown).Alexicalist
accountofthesamefactsisshowntobehighlycumbersomeandunexplanatory.
SecondpositioncliticsareVocabularyitemswhichundergoeitherLocalDislocationorProsodicInversion
withahost.
Syntacticclitics.Finally,theterm'clitic'issometimesusedtodescribesyntacticallymobileheads,
typicallyDeterminers,suchascertainRomancepronominalsonsomeaccounts.Insuchcasesthe
dependencyrelationorspecialbehaviorisasyntacticpropertyofamorpheme(syntacticcategory).Inmany
http://www.ling.upenn.edu/~rnoyer/dm/#DM 13/14
6/4/2017 DistributedMorphology
casestheVocabularyitemswhichareinsertedintothesemorphemesalsoshoweitherphonological
dependencyasleanersoradditionalpeculiaritiesofpositionviaLocalDislocationorProsodicInversion.
SeeHarris1994,1997aandEmbick1995forcasestudies.
HowcanIfindoutmoreaboutDM?
Clickhereforalistofrelevantliterature.
PleasedirectquestionsorcommentsaboutthispagetoRolfNoyer.
http://www.ling.upenn.edu/~rnoyer/dm/#DM 14/14