citoletitgo原唱

3TheInteractionBetweenMorphologyandPhonologySHARONINKELAS1IntroductionThemorphologyofalanguageconcernsthegeneralizationsaboutformandmeaningthatrelatewordstooneanotherwithinthatlanguage.
Thephonologyofalanguageconcernsthegeneralizationsaboutthesoundpatternsinthatlanguage.
Morphologyandphonologyintersectinsofarasthestatementofmor-phologicalgeneralizationsincludesinformationaboutsoundpatterns,orinsofarasthestatementofphonologicalgeneralizationsincludesinformationaboutmorphology.
2WhenMorphologyAffectsPhonology:ThePhonologicalInterpretationofMorphologicallyComplexWordsTheearliestinuentialgenerativeapproachestotheintimateinteractionbetweenphonologyandmorphology(ChomskyandHalle1968;Kiparsky1982b;Mohanan1986)focusedonthephonologicalinterpretationofmorphologicallycomplexwords,andthisiswherewewillbeginoursurveyaswell,althoughwewillnotrestrictourselvestothephenomenacoveredbyanyparticulartheoryintheprocess.
2.
1MorphologicallyConditionedPhonologyPhonologicalrequirementsinalanguagecanaltertheshapethatindividualmorphemestakeindifferentcontexts,producingallomorphy.
SometimestheseTheHandbookofPhonologicalTheory,SecondEdition.
EditedbyJohnGoldsmith,JasonRiggle,andAlanC.
L.
Yu2011BlackwellPublishingLtd.
Published2011byBlackwellPublishingLtd.
9781405157681_4_003.
indd689781405157681_4_003.
indd6806/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology69alternationpatternsarequitegeneralinthelanguage.
InTurkish,forexample,averygeneralruleofprogressivevowelharmonydeterminesthevalueof[back]forthevowelsofmostsufxes,whichsurfacewithfrontvowelsfollowingrootswhosenalvowelisfront(e.
g.
gyl-ler'rose-pl',anne-ler'mother-pl')butwithbackvowelsfollowingstemswhosenalvowelisback(e.
g.
ok-lar'arrow-pl',elma-lar'apple-pl')(see,forexample,Lewis1967).
Morphologicallyconditionedphonologyariseswhenphonologicalalternationsarenotfullygeneralinthelanguagebutareinsteadspecictoparticularmorphologicalconstructions,suchascompounding,truncation,afxation,orreduplication(foroverviewsatafairlytheory-neutrallevel,seeforexample,Dressler1985;Spencer1998).
InBelhare,forexample,intervocalicvoicingoccursatstem-sufxboundaries(lap>lab-u!
'catchit!
')butnotatprex-stemboundaries(ka-pira!
'giveittome!
'),orinunderivedwords(pipisi'(drinking)straw')(BickelandNichols2007).
InTurkish,thediminutivesufx-cIktriggersthedeletionofstem-nalk(Lewis1967:57):bebek,bebe-cik'baby/baby-dim',kpek,kpe-cik'dog/dog-dim').
Nootherconsonant-initialsufxtriggersthisdeletion(bebek-i'childcareprovider',bebek-lik'infancy',bebek-ten'baby-abl',bebek-ken'whileababy',etc.
).
InDakota,acoronal→velardissimilationruletargetscoronalconsonantclustersthatstraddlethejuncturebetweenthetwocopiesinreduplication(/at/→ag-át-a'curved'),butnotclustersarisinginothermorphologicalcontexts,forexample,compounding(sdod+:hí-ya'know.
I+you-cause=Iknowyou';Shaw1985:184).
Morphologicallyconditionedphonologycanbesegmental,asintheexamplesjustcited,orcaninvolveprosodicpropertiessuchastone,stressorlength.
VeryfamiliarexamplesincludeIndo-Europeanaccentuation(Kiparsky1973b)andJapanese(McCawley1968b;Poser1984;PierrehumbertandBeckman1988;Alderete1999,2001).
InJapanese,morphologicalconstructions,whichincludeprexation,sufxation,zero-derivationandcompounding,comeintwoessentialvarieties:thosewhichpreservelexicalstemaccentandthosewhicheraseit.
Poser(1984)termsthetwotypes"recessive"and"dominant,"respectively,buildingonterminologyintroducedinKiparsky1973b(seealsoKiparskyandHalle1977;HalleandMohanan1985).
Japanesepitch-accentissubjecttostrictdistributionalregularities:eachwordhasatmostoneaccent,andincasesofconictbetweentwolexicallyaccentedmorphemesinthesameword,thegeneralprincipleisthattheleftmostaccentwins(Poser1984).
Recessivesufxes,asshownin(1),behaveaccordingtotheLeftmostWinsprinciple.
Anunaccentedsufx,forexample,pasttense-ta,leavesstemaccentunaffected(1a),whileanaccentedrecessivesufx,forexample,conditional-tára,surfaceswithitsaccentonlyifthestemisnotalreadylexicallyaccented(1b).
Otherwise,LeftmostWinsresultsintheeliminationofsufxaccent(1c).
PagenumbersarefromPoser1984:(1)a.
/yob-ta/→yoNda'called'(49)/yóm-ta/→yóNda'read'(49)b.
/yob-tára/→yoNdára'ifhecalls'(48)c.
/yóm-tára/→yóNdara'ifhereads'(48)9781405157681_4_003.
indd699781405157681_4_003.
indd6906/05/20115:08PM06/05/20115:08PM70SharonInkelasOtherrecessivesufxesarepre-accenting,depositingaccentonthenalsyllableofunaccentedstemsbuthavingnoeffectonlexicallyaccentedstems(2a).
Stillothersareaccent-shifting.
Posertermsthese"dependent";theyshiftstem-accent,ifany,tothestem-nalsyllable,butdonothaveanyeffectonlexicallyunaccentedstems(2b):(2)a.
matumoto-si→matumotó-si'MrMatsumoto'(54)áNdoo-si→áNdoo-si'MrAndo'(54)nisímura-si→nisímura-si'MrNishimura'(54)b.
koná-ya→konáya'ourseller'(55)kúzu-ya→kuzúya'junkman'(55)kabu-ya→kabuya'stockbroker'(55)Incontrasttorecessiveafxes,dominantafxestriggerdeletionofstem-accent.
Accenteddominantsufxes,likeadjective-forming-ppó,erasestemaccentandsurfacethemselvesasaccented(3a).
Unaccenteddominantafxesproducecom-pletelyunaccentedoutputs,likedemonymic-kko(3b).
Stillotherdominantsufxesplaceaccentontheinitialornalstemsyllable,asillustratedby(mostformswith)the"true"prexma(C)-(3c)andfamilynaming-kesufx(3d),orevenonthestem-penultimatesyllable,aswiththegirls'name-forming-ko(3e):(3)a.
abura→abura-ppó-i'oil,fat/oily'(49)yásu→yasu-ppó-i'cheap/cheap,tawdry'(49)adá→ada-ppó-i'charming/coquettish'(49)b.
kóobe→koobe-kko'anindigenéofKobe'(72)nágoya→nagoya-kko'anindigenéofNagoya'(72)nyuuyóoku→nyuuyooku-kko'anindigenéofNewYork'(72)c.
futatu→map-pútatu'two/exactlyhalf'(57)sáityuu→mas-sáityuu'amidst/intheverymidstof'(57)syoozíki→mas-syóoziki'honesty/downrighthonest'(57)d.
nisímura→nisimurá-ke'theNishimurafamily'(55)ono→onó-ke'theOnofamily'(55)hára→hará-ke'theHarafamily'(55)e.
haná→hána-ko'ower/name'(58)kaede→kaéde-ko'maple/name'(59)mídori→midóri-ko~'green/name'(59)midorí-koThusforeachafx,ormoregenerallyforeachmorphologicalconstruction,sincezero-derivationandcompoundingaresubjecttosimilaraccentualparameters,itisnecessarytoknowwhichofseveralpossibleaccentplacementpatternstheafxtriggers(none,stem-initial,stem-nal,stem-penultimate)andwhetherthosepatternspreserveordeletelexicalstemaccent(dominantvs.
recessive).
AmoreunusualcaseoccursintheMayanlanguageMam(England1983;Willard2004),inwhichvowellengthiscontrastivebothinrootsandinsufxes9781405157681_4_003.
indd709781405157681_4_003.
indd7006/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology71andatmostonelongvowelispermittedperword.
Sufxesdivideintotwotypes:thosethattriggershorteningofstemvowels,andthosethatdonot.
Willardtermsthese"dominant"and"recessive"sufxes,respectively,followingtheterminologyusedintheaccentualliterature.
Vowellengthofthesufxitselfisnotapredictorofvowelshortening,asshowninthetable;neitherisstressnormorphologicalfunction.
Whetherornotsufxationcausesstemvowelshorteningisanidiosyn-craticpropertyofeachsufxationconstruction.
Dominantsufxesareshownin(4a);recessivesufxesareshownin(4b):(4)a.
mool-'burn'mol-oloon'easilywilted'(facilitative)juus-'burn'jus-b'een'burnedplace'(resultantlocative)jaaw-'goup'jaw-nax'up'(directional)yuup-'putoutre'yup-na'putout'(participial)b.
iil-'sin'iil-a'scold'(intransitiveverbalizer)ooq'-'cry'ooq'-b'il'somethingwhichcausescrying'(instrumental)Whilecasesofstressandtonereplacementaremorecommonthanvowellengthmanipulation,onthebasisofcurrentknowledgeitseemsreasonabletoassumethatanykindofphonologicalpattern,otherthanthemostlow-levelallophonicalternations,canberestrictedtoamorphologicalcontext,insomelanguageoranother.
Indeedmostphonetically"unnatural"phonologicalalternations(seee.
g.
Anderson1981;Buckley2000,Hyman2001a)aremorphologicallyconditionedinjustthisway,maintainingtheirnicheofproductivityinspecicmorphologicalcontexts(seee.
g.
Pierrehumbert2006b).
HowismorphologicallyconditionedphonologytobehandledCurrentthink-ing,buildingonideasgoingbacktothe1960s,offerstwomainoptions:copho-nologies,whichareco-existingsub-grammarswithinasinglelanguage,eachindexedtoaparticularmorphologicalconstructionorsetofconstructions(e.
g.
Orgun1996;Anttila2002a;ItandMester1995;InkelasandZoll2005);orindexedconstraints,inwhichthelanguagehasjustonephonologicalgrammar,butpar-ticularconstraintswithinitareindexedtospecicmorphemesormorphologicalconstituents(e.
g.
McCarthyandPrince1995;ItandMester1999;Smith1999;Alderete2001;Pater2009).
TohandleJapaneseaccentuation,forexample,acophonologicalapproachwouldsubdividethegrammarintoanumberofcloselyrelatedvariants,andindexeachmorphologicalconstructiontooneofthesevariants(cophonologies).
The"dom-inant"morphologicalconstructionswouldbeassociatedwithcophonologiesinwhichinputstemaccentiseliminated.
Poser1984,usingarule-basedprecursortocophonologies,proposedindexinganaccentdeletionruletoeachdomin-antafx.
InanOptimalityTheory(PrinceandSmolensky2004;McCarthy2008)implementationofcophonologytheory,thesamegoalwouldbeaccomplishedbyvaryingtherankingoftheconstraintscharacterizingaparticularaccentuationpatterneitherbeloworabovethefaithfulnessconstraintpreservingstemaccent.
Forexample,supposestem-nalandstem-initialaccentareimposedbythe9781405157681_4_003.
indd719781405157681_4_003.
indd7106/05/20115:08PM06/05/20115:08PM72SharonInkelasalignmentconstraintsAlign-Right(accent,stem)andAlign-Left(accent,stem),respectively.
Thecophonologyofadominantsufxwouldrankitsaccent-placingconstraintsaboveMax-accent,ensuringthedeletionofstemaccentsthatareinthewronglocation.
ThecophonologyofarecessiveafxwouldrankMax-accenthighest,ensuringthattherelevantaccentuationpatternisimposedonlyasadefault.
AllafxcophonologiesinJapanesesharetheconstraintrankingensuringthatthereisatmostoneaccentpossibleintheoutput.
Generally,incophonologicalmodels,thegreatmajorityofconstraintrankingsaresharedbyallcophonologiesinthelanguage;Anttila(2002a)hasmodeledthissharingusinganinheritancehierarchy,inwhichcophonologiesaregroupedtogetherbytheconstraintrankingsthatunitethem.
Thesuperordinatenodeinsuchahierarchy,orwhatInkelasandZoll(2005)termthe"masterranking,"representstheuniquegeniusofthelanguage,apartialrankingofconstraintstowhicheveryindividualcophonologymustconform.
AsAnttila'sworkmakesclear,itisalsopossibletogroupsmallersubsetsofcophonologiesunderintermediatenodestocapturesubregularities,forexample,overalldifferencesbetweennounsandverbsinJapanese(asdocumentedbyMcCawley1968bandPoser1984)orbetweennounsandadjectivesinFinnish(Anttila2002a).
Constraintindexationisadifferent,contemporaryapproachtomorphologicallyconditionedphonology,developedintheearlydaysofOptimalityTheory.
Theapproachwasoriginallymorpheme-based,indexingconstraintstoparticular(setsof)morphemes.
Forinstance,ItandMester(1999)accountfortheresistanceofrecentlyborrowedrootsinJapanesetonativephonotacticrestrictionssuchasNo-p,thebanon[p],byindexingaspecial,high-rankedfaithfulnessconstrainttoexactlythesetofrelevantroots:FaithAssimilatdForeign>>No-p>>FaithYamato.
Anativerootwithunderlyingillicit/p/wouldhavetogetridof/p/,butanassimilatedforeignroot,asinpato-ka'patrolcar'(p.
63),wouldpreserveit.
Con-straintindexationhasalsobeenappliedtoderivedstems,nearlymergingthedifferencebetweencophonologies(indexedtostem-formingconstructions)andindexedconstrainttheory.
Inhisanalysisofmorphologicallyconditionedaccen-tuationinJapanese,forexample,Alderete(2001)differentiatesdominantandrecessiveafxesbyindexinganti-faithfulnessconstraintstostemsderivedbytheformer.
TheconstraintOODom-MAX-ACCENT("ItisnotthecasethateveryaccentinS1hasacorrespondentinS2")speciesthatinderivedstemscreatedbydominantafxes,aninputstemaccentisnotpreservedinoutput.
InAlderete'smodel,ifadominantafxcausesinputstemaccenttodelete,thenthedefaultaccentuationpatternofthelanguageisimposedinitsplace;itisalsopossibleforalignmentconstraintstolocateaccentattheboundariesbetweenstemsandspecicafxes.
Withregardtothetypesofsubstantivedifferencesthatcanexistbetweenmorpho-phonologicalpatternsinthesamelanguage,thetwoapproachesareverysimilarsubstantively,atleastwhenimplementedinOptimalityTheory;eachusesthesamesetofconstraintsandthuspredictsthesamerangeofpossible9781405157681_4_003.
indd729781405157681_4_003.
indd7206/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology73morphologicallyconditionedphonologicaleffects.
Amoreprobativequestioniswhatdegreeofdifferencecanexistacrossdifferentmorphologicallycondi-tionedpatternsinthesamelanguage.
Proponentsofindexedconstraintshavesuggestedthatthebulkoflanguage-internalvariationcanbedescribedintermsofrelativefaithfulness.
Alderete(2001)hastermedthis"grammardepend-ence,"claimingthateachlanguagehasasinglesetofphonologicalrestrictions(syllablestructure,accentuation,segmentinventory,etc.
);individualmorpho-logicalconstructionsdifferonlyinthedegreetowhichtheyarefaithfultoinputstructureswhichviolatetheserestrictions.
Thus,forexample,inJapanese,stemscreatedbyrecessiveafxesarefaithfultoinputaccent,whilestemscreatedbydominantafxesarenotfaithful,andexhibitthedefaultaccentu-ationpatternofthelanguage.
Thetheoreticalargumentsonthispointaresubtleandcomplex;seeforexample,ItandMester(1999),InkelasandZoll(2007),Pater(2009).
Theempiricalissueiswhetheranylanguageeverimposescom-pletelycontradictorypatternsindifferentmorphologicalenvironments.
Japaneseaccentuationisarguablyacaseofthiskind,sinceevenwithinthesetofdomin-antafxes,atleastfourcontradictoryaccentuationpatternsareobserved,asseenin(3).
Arelatedsubstantivequestionaboutmorphologicallyconditionedphonologyisthenumberofvariants(cophonologies,indexedconstraints)asinglelanguagecanpermit,andthedegreeofdifferencesamongthem.
ThisquestionhasbeenaddressedexplicitlyinworkbyAnttila(2002a),whosehierarchicalcophonolo-gicalmodelpredictsthateveryconstraintrankingpossibilitynotexcludedinthe"masterranking"ofalanguageisexpectedtobeinstantiatedinsomecophon-ology.
Neithercophonologytheorynorindexedconstrainttheoryaddressesthequestionofhowmanydifferentcophonologiesarepossible,or,really,towhatdegreetheycouldpotentiallydiffer.
AsobservedbyItandMester(1999),InkelasandZoll(2007),andPater(2009),theseissuesmayultimatelybelaidatthefeetofthehistoricaloriginsofcophonologicalvariation,whichincludelanguage-internalfactorslikegramaticalizationandanalogyaswellasexternalfactorslikelexicalborrowingormoreextremelanguagecontact,aswellasinu-encesoflanguageacquisition.
Onewayinwhichthecophonologicalandindexedconstraintapproachesclearlydifferisintheirabilitytocapturetheinteractionbetweendifferentmorphologic-allyconditionedpatternsinthesamelanguage,or"layeringeffects.
"2.
2LayeringEffectsIftwomorphologicalconstructionsarepresentinthesameword,andeachisassociatedwithitsownphonologicalpattern,whichpatternprevails,orifbothdo,howdotheyinteractTheevidencesuggeststhatbothpatternsprevail,andthattheyareimposedintheorderinwhichtheassociatedmorphologicalconstructionsarecombined.
Thisisperhapseasiesttoillustrateusingaccentuationpatternsthatareincompatible,9781405157681_4_003.
indd739781405157681_4_003.
indd7306/05/20115:08PM06/05/20115:08PM74SharonInkelassuchthatwhentwomorphologicalconstructionsafliatedwithincompatiblepatternsco-occurinthesameword,onemusttakeprecedenceovertheother.
AgoodcasestudyisTurkish,whichresemblesJapaneseinsomeofitsoverallaccentuationprinciples.
ThedefaultpositionforstressinTurkishwordsisnal(thusarabá'car',araba-lár'cars',araba-lar-dán'fromcar's);thereisexactlyonestressperword,regardlessofmorphologicalcomplexity.
Anumberofmor-phologicalconstructionsassignstress;thesealwaysoverridethedefaultnalstresspattern.
(OnTurkishstress,seee.
g.
Lewis1967;Sezer1981;KabakandVogel2001;InkelasandOrgun2003.
)Inwordswithmorethanonestress-assigningmorphologicalconstruction,orderofmorphologicalcombinationpredictsthestressoutcome.
Forexample,Turkishhasaproductivezero-derivationconstructionformingplacenamesoutofwordsofanypartofspeech;theconstructionismarkedbyadistinctivestresspattern(Sezer1981)whichplacesstressonthepenultimateorantepenultimatesyllable,dependingonsyllableweight:bak-acák'look-fut'~Bakácak(placename),torba-lí'bag-assoc'~Tórbal(placename),,andsoon.
Turkishalsohaspre-stressingsufxeslikepasttensepredicative-(y)DI(torbá-yd'itwasabag'),negative-mE(gel-dí'came'vs.
gél-me-di'didn'tcome'),ormitigative-CE(süt-lü-lér'milk-assoc-pl=themilkyones',vs.
süt-lü-ce'milk-assoc-mit=kindofmilky').
AsdocumentedinInkelas1999,InkelasandOrgun1998,InkelasandOrgun2003,thestresspatternsofthelanguageareallrecessiveinthesensethattheyareimposedonlyiftheinputstemlacksstress.
Inwordslike/torba-lI/,thestressoutcomedependsonwhetheranunstressedroot,forexample,/torba/'bag'isrstconvertedtoastressedplacename(Tórba)andthensufxed(→Tórba-l),retainingitsplacenamestress,orrstsufxed(torbá-l)andthenconvertedtoaplacename(Torbál),retainingthestressassignedbythesufxinsteadofdisplayingtheplacenamestresspattern.
LiketheIndo-EuropeancasesdiscussedbyKiparsky1973b,Turkishrespectsaprincipleof"InnermostWins"(Inkelas1999;InkelasandOrgun2003).
AnotherusefulillustrationoflayeringcanbefoundinHausa,alexicaltonelanguagewhosemorphologicalconstructionseitherpreservestemtone(compar-abletothe"recessive"morphologyofJapanese)orreplaceitwithanewtonemelody("dominant")(Newman1986;2000;Inkelas1998).
Thestructurein(5)illustratesaverbrootwhichcombineswiththedominantventivesufx-o",thenundergoespluractionalreduplication,andisnallyconverted,viazero-derivation,toanimperative.
Boththeventiveandtheimperativeconstructionsaredominant.
Theventiveimposesanall-Hmelody(e.
g.
fìtá"(LH)'goout'→fít-ó"(H)'comeout',gángàrá"(HLH)'rolldown'→gángár-ó"(H)'rolldownhere',andsoon(Newman2000:663).
TheimperativeimposesaLHmelody(e.
g.
ká"mà"(HL)→kà"má"(LH)'catch!
',bíncìké"(HLH)→bìncìké"(LH)'investigate!
';né"mó"(H)→nè"mó"'seek!
',nánné"mó"(H)→nànnè"mó"(LH)'seekrepeatedly!
').
In(5),theventiveoccurshierarchicallyinsidetheimperative.
PredictablyinHausa,theoutermostdominantconstructionistheonewhosepatternsurfaces;inthiscasetheoutermostconstructionistheimperative,andconsequentlythewholewordsurfacesLH.
Zero-derivationconstructionsarerepresentedbynullsufxesforpurelygraphicalconvenience:9781405157681_4_003.
indd749781405157681_4_003.
indd7406/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology75(5)nèn-nè"mó"nén-né"mó"né"mó"-ó"(H)-(LH)nè"má"(LH)CVC-pluract.
-'seek'-ventive-imperativeThiskindofpatternischallengingforindexedconstrainttheory,inwhichallconstraints,morphologicallyindexedandgeneral,existinonexedrankinginthegrammarofthelanguage.
InTurkish,theconstraintsthatrequireplacenamestohavetheSezerstresspatternmustrankeitherbeloworabovetheconstraintsrequiringstresstoimmediatelyprecedesufxeslike/-lI/.
Inawordcontainingbothazero-derivedplacenameandapre-stressingsufx,thehigher-rankedpatternshouldalwaysprevail,regardlessofmorphologicalstructure.
TheproblemisthatbothtypesofembeddingcanoccurinTurkish,withdifferentmeaningsanddifferentstressoutcomescorrespondingtothetwopossiblehierarchicalstructures(InkelasandOrgun1998).
Asingleranking,asinindexedconstrainttheory,cancaptureonebutnottheother,missingtheconnectionbetweenmorphologicalembeddingandconstraintranking.
Bycontrast,incophonologytheoriesthisconnectioniscapturedintrinsically(seee.
g.
Inkelas1993;Orgun1996on"deriv-ingcyclicity");thehierarchicalrelationshipbetweentwoconstructionsdirectlydeterminestheinput-outputrelationshipbetweentheassociatedcophonologies.
Somelayeringtheorieshavebundledlayeringwithadditionalclaims,andhavebeenweakenedinsofarastheadditionalclaimshavenotheldup.
Forexample,thetheoryofLexicalMorphologyandPhonology(Kiparsky1982;Mohanan1986)associatedcyclicity(layering)withstructurepreservationandstrictlevelordering,towhichsubsequentliteraturehasraisedcompellingempiricalobjections.
StratalOptimalityTheory(Kiparsky2000,2008)limitsthenumberofcophonologies(layertypes)inanygivenlanguagetothree,whicharestrictlyordered.
Thevir-tueoflimitingstratainthiswayisthatitdrawsattentiontogeneralpropertiesofstems,words,andphrases,butoftenattheexpenseofbeingabletodescribemore"minorrules.
"Closestudiesofstratainagglutinatinglanguages,forexample,havegenerallyresultedinthepostulationofmorethanthreelevelsbelowthewordlevelalone(seee.
g.
Hargus(1985)onSekani,Mohanan(1986)onMalayalam,Buckley(1994)onKashaya).
BothHargusandMohanan,likeCzaykowska-Higgins(1993,forMoses-ColumbianSalish)andInkelasandOrgun(1998,forTurkish)argueinadditionthatthestratanecessitatedtodescribethemorphophonologicalsubgeneralizationsinthelanguagesinquestioncan-notbecruciallyorderedinthewaythatlevelorderingtheorywouldrequire.
Itisimportant,however,toemphasizethattheessenceoflevelorderingtheoriesisthesameastheessenceofcophonologytheory,namelytheinterleavingof9781405157681_4_003.
indd759781405157681_4_003.
indd7506/05/20115:08PM06/05/20115:08PM76SharonInkelasphonologyandmorphologyisduetotheassociationofmorphologicalconstruc-tionswithparticularphonologicalpatterns.
2.
3ParadigmUniformityApromisingavenueofresearchon"optimalparadigms"seekstoexaminewhetherparadigm-levelconsiderationscouldmotivateorevensupplantcycliccophono-logicalmodels.
Thisisespeciallypromisingincasesofrecessivephonologicalalternationsinwhichstemstructureispreservedundersubsequentafxation.
Theoverallresultisthatparadigmsarekeptlevel,thatis,withphonologicallyuniformstemshape.
Ithasbeenproposedthatratherthanresultingfromcycli-city,stemuniformityeffectsfollowfromparadigmuniformityconstraintswhichkeepthesharedportionsofmorphologicallyrelatedwordsphonologicallyiden-tical;seeforexample,theBase-Identityconstraintsofforexample,Kenstowicz1996.
Whenevaluatedonlywithrespecttothesubconstituentsofasingleword,Base-Identityconstraintsfunctionlikehigh-rankedinput-outputfaithfulnessonacophonologyaccount,causingstructurethatisoptimalfortheinnermostmor-phologicalconstituenttopersistevenifouterlayersofmorphologyrenderitphonologicallyopaque.
ThisoccursinTurkish,asdiscussedearlier:lexicallystressedroots(e.
g.
lokánta'restaurant')andderivedstressedstems(e.
g.
süt-lü-ce'milk-assoc-mit')keeptheirstresswhentheycombinewithwould-bestress-assigningsufxeslikepre-stressingpredicative-(i)di,forexample,lokánta-yd,süt-lü-ce-ydi.
Therecessivecharacterofstress-assigningsufxescanbeattributedtoparadigmuniformity:thederivationalandinectionalparadigmsofalexicallystressednounlikelokánta'restaurant'allshareanidenticallystressedroot(lokánta,lokánta-lar(-pl),lokánta-da(-loc),lokánta-lar-da(-pl-loc),lokánta-yd(-pred),andsoon.
).
Ofcourse,Base-IdentityisnotabsoluteinTurkish;itisonlystressedrootswhosephonologicalstresspatternismaintainedacrosstheparadigm.
Lexicallystresslessrootsalternatesystematically,accordingtowhethertheycombinewithastress-neutralsufx,forexample,araba-yá'car-dat',orastress-assigningsufx,forexample,arabá-yla'by/withcar'.
Itisalsoimportanttonotethatthedenitionof"base"ofaparadigmmustbebroadenedtoincludenotjustrootsbutalsocomplexstems.
Whiletherootarabaisnotinherentlystressed,andthereforevariesinshapedependingonmorphologicalcontext,astressedstemlikearabá-ylakeepsitsstresswhensufxed,forexample,arabá-yla-m1'car-assoc-interrogative=by/withcar'.
Thus"base"isequivalentto"sub-constituent"inalayeringtheory.
Thepredictionsofparadigmconstraintsdivergefromthepredictionsofcopho-nologicallayeringmodelswhenappliedtothesharedstemsofwordsneitherofwhichisasubconstituentoftheother.
Forexample,Kenstowicz(2005)discussesthecaseofSpanishdiminutives,formedbyadding-cito[sito](m.
)/-cita[sita](f.
)whenthebaseendsin[n]or[r]andbyadding-ito/-itawhenthebaseendsinavowel.
1ExamplescitedbyKenstowicz,usinghisphonemictranscription,include[limon]'lemon(m.
)'→[limon-sito],[barko]'ship(m.
)'→[bark-ito],[korona]9781405157681_4_003.
indd769781405157681_4_003.
indd7606/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology77'crown(f.
)'→[koron-ita].
Fornounsthathavefeminineandmasculinegendercounterparts,like[raton]'mouse(m.
)',[raton-a]'mouse(f.
)',thesurfaceconditionsforattachmentofthe[-sita/-sito]diminutiveformativesaremetbythen-nalmasculinebutnotbythea-nalfeminine.
OnKenstowicz'sassumptionthattheformofthenon-diminutivenoundeterminesthedimunitivesufxthatisadded,thefemininediminutiveof'mouse'shouldbe[raton-ita],basedon[ratona],whereasthemasculinediminutiveof'mouse'shouldbe[raton-sito],basedon[ratón].
Infact,however,bothdiminutiveshavethediminutiveformativetrig-geredbyann-nalinput:[ratonsito],[ratonsita].
Kenstowiczproposesaparadigmuniformityanalysis,whichheattributestoAguero-Batista,onwhichmasculineandfemininediminutivesarerequiredtohavethesamesurfacestemshape.
Masculine[ratón]transparentlyselects[-sito]([raton-sito]),andbyparadigmuni-formity,thefeminine[ratona]isrequiredtoselectthe[-sita]allomorphaswell.
Paradigmuniformityfavors[ratonsita],whiletransparencyofsufxselectionfavors[ratonita];paradigmuniformitywinsout.
(Thereisanalternativetoinvok-ingparadigmuniformityinthiscase,namelytreatinggender-unspecied[ratón]astheinputbothto[raton-sit-o]and[raton-sit-a].
Theargumentforparadigmuniformityasaconstraintisonlyasstrongastheargumentthatnounsaregender-markedintheinputtodiminutivization.
Sincethediminutiveendingsthemselvesencodegender,thisassumptioncouldbequestioned.
)AparticularlyinterestingsetofexamplesofparadigmuniformityiscitedbyDowning(2005a:24,130ff.
),inastudyofsufxdoublinginJita(Bantu).
InJitaverbs,thecausativesufx-ytriggersmutation(spirantization)ofanypreceding/r/:/gur-a/'buy-fv'→[gura],vs.
/gur-y-a/'buy-caus-fv'→[gusya].
Jitahasatleasttwootherderivationalsufxeswithwhichthecausativecanco-occur:applicative/-ir/andreciprocal/-an/.
Whenthecausativeco-occurswitheitherofthese,itmustdouble,occurringbothdirectlyaftertherootanddirectlyaftertheothersufx,forexample,/gur-y-ir-y-a/'run-caus-appl-caus-fv'→[gusi:sya]or/gur-y-an-y-a/'run-caus-recip-caus-fv'→[gusyanya].
Inverbswithcausative,reciprocalandapplicativesufxes,thecausativemustoccurthreetimes:/gur-y-ir-y-an-y-a/'run-caus-appl-caus-recip-caus-fv'→[gusi"sya"nya].
Similarmulti-plicationofthecausativeoccursinKinande(MutakaandHyman1990)andCibemba,amongotherBantulanguages(Hyman1994,2003).
AccordingtoDown-ing,themultiplicationoftheJitasufxoccursunderpressurefromparadigmuniformity.
Downingproposesthatthecausativesufxisalwaystheoneaddedmorphologicallyrsttotheroot(thus,foraverbwithallthreesufxes,theabstractunderlyingstructureis/Root-Caus-Appl-Recip-/.
Phonologically,however,thecausative/-y/isalwaysrequiredtobelastinthestem,byaright-alignmentconstraint.
CruciallyonDowning'sanalysis,thephonologicalformofthe(alwaysinnermost)Root-Caussubconstituentisrequiredtobeuniformacrossallcausativeformsofagivenstem.
Theonlywaytosatisfyboththeuniformityandtheright-alignmentrequirementistoaddthecausativemorethanonce.
Intheapplicativ-izedcausative/gur-y-ir-y-a/,forexample,the/gur-y-portionsatisesstemuniformitywhilethe/.
.
.
-y-a/portionsatisesrightwardy-alignment.
Downingargues(p.
128)againstanalternativecyclicaccountofcausativedoublingfacts,9781405157681_4_003.
indd779781405157681_4_003.
indd7706/05/20115:08PM06/05/20115:08PM78SharonInkelassuchastheoneproposedforparallelafxdoublingfactsinCibembaandanum-berofotherBantulangaugesbyHyman(1994,2003),onthegroundsthatthereisnootherevidenceforcyclicityinJita.
2.
4ParadigmContrastAnothermanifestationofparadigmaticconsiderationsisthemorphological(orlexical)needtokeepwordsorstemsphonologicallydistinctfromoneanother;thisneedforparadigmcontrasthasbeenarguedtoinhibitortriggerphonologicaleffects.
Forexample,Crosswhite(1999)arguesonthebasisofevidenceintheTrigraddialectofBulgarianthatotherwisegeneralruleofvowelreductionisblockedjustincaseitwouldcausethemergeroftwowordsinthesameparadigm.
InTrigradBulgarian,unstressed/o/and/Q/surfaceas[a],mergingwithunderlying/a/:(7)/rog-ave/['rogave]'horns'/rog-ave-te/[raga'vete]'thehorns'/sQrp-ave/['sQrpave]'sickles'/sQrp-ave-te/[sarpa'vete]'thesickles'cf.
[a'rala]'plough'Crosswhiteobservesthatunstressed/o/failstoreduceinanumberofsufxes,forexample,the-oendingonnominativemasculineanimatenouns:['ago]'olderbrother(nom.
)',not*['aga].
Crosswhiteobservesthat/o/reductionfailspreciselywhen,asinthesecases,twodistinctsufxes(onewith/o/andonewith/a/)wouldmergeifreductionapplied.
Theaccusativeendingonmasculineanim-atenounsis-a,asin['aga]'olderbrother(acc.
)'.
AccordingtoCrosswhite(andKenstowicz2005),vowelreductionisblockedwhenitwouldmergethenomin-ativeandaccusativeparadigmcellsofmasculineanimatenouns.
Forarecentsurveyoftheseandothereffectsinwhichaneutralizingalternationisclaimedtobeblockedbyaconstraintagainsthomophony,seeIchimura2006.
AccordingtoKurisu(2001),anti-homophonyconsiderationscanalsotriggerdissimilatoryphonologicalalternations.
Kurisuinterpretsanumberofeffectspreviouslydescribedasrealizationalmorphology(seeSection5)asresultingfromtherequirementthatinputandoutputformsbedistinct.
Onthisview,processmorphologyisarepairofwhatwouldotherwisebethenullrealizationofamorphologicalconstruction.
ExamplesincludetheuseofablauttomarkpluralinGerman(Vater~Vter'father(s)',Mutter~Mütter'mother(s)',p.
191),andtheuseofvoweldeletiontoderivedeverbalnounsfrominnitivesinIcelandic(klifra'climb-inf'→klifr'climbing',puukra'conceal(inf.
)'→puukr'concealment',p.
31,citingOrenik1978;Arnason1980;Kiparsky1984;It1986;Benua1995).
Kurisu'sanalysisisthattheseconstructionsconsist,morphologically,ofzero-derivation,butthatanti-homophonyconsiderationscompelthephonologytoaltertheoutputtoavoididentitywiththeinput.
Thefactthatablaut(inGerman)orvoweldele-tion(inIcelandic),arethepreferredoptions,asopposedtoanyotherimaginable9781405157681_4_003.
indd789781405157681_4_003.
indd7806/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology79changes,follows,inKurisu'saccount,fromtherankingoffaithfulnessconstraintspenalizingdeletion,insertion,and/orfeaturalchanges.
AchallengeforKurisu'sviewcomesfromcasesofmorphologicallyconditionedphonologicaleffectsapplyingalongsideafxation,forexample,German:Gast~Gst-e'guest(s)'orGaul~Gul-e'packhorse(s)',withsufxationandablaut(p.
191).
Sinceafxationalonesufcestomaketwoword-formsdistinctinthesecases,whatmotivatestheaccompanyingablauteffectKurisu'sansweristhatthesecasesareinstancesofdoublemorphologicalexponenceresultingfrommorphologicalopacity:theafxesintheseexamplesareessentiallyinvisibletotheanti-homophonyprinciplethatrequiresthesingularandpluralcellsoftheparadigmtobedistinct.
The"rst"layerofmorphologyisnull,andphonologyconspirestomakethezero-markedpluralstem(Gst)distinctfromthesingularstem(Gast).
Thesecondlayerofmorphology,towhichthephonologyisblind,thendouble-marksthepluralwithasufx:Gst-e.
Ofcourse,doubleexponenceisnotlimitedtocasesofthiskindinwhichoneexponentisarguablyapho-nologicalmodicationandtheotherisanovertafx;languagesareknowntousetwoormoreovertafxes,orasuppletivestemplusovertafx(es),tomarkasinglecategoryaswell(e.
g.
Anderson2001;Bobaljik2000;Harris2008a).
Thuswhenablautisoneofthetwoexponentsofamorphologicalcategory,itcouldbeanalyzed,perKurisu,asaphonologicalresolutiontoanti-homophony,oritcouldbeattributedtowhatevermorphologicalfactorsareresponsibleformultipleexponencemoregenerally.
Furtheraeld,WedelandUssishkin(2002)havesuggestedthatneutralizingphonologicalalternationscanbeinhibitedifthewordstheywouldapplytoexistindensephonologicallexicalneighborhoods,thatis,iftherearehighnumbersofphonologicallysimilarwordsinthelexicon.
Ifthishypothesisiscorrect,contrastpreservationmightinhibitphonologicalalternationsnotonlywhenthewordsinquestionareinthesameparadigm,butevenwhentheyaremorphologicallyunrelated.
Dispersionmightthusplayanactiverolesynchronically,notjustthediachronicrolesuggestedbyFrisch,PierrehumbertandBroe(2004)intheirdiscussionofArabicrootconsonants.
Frisch,Pierrehumbert,andBroeshowthatthedistributionofconsonantsinSemiticrootsisskewedtofavortriplesofrootconsonantsthatarephonologicallyinternallydisparateovertriplesofrootcon-sonantsthatareinternallysimilar.
Frisch,Pierrehumbert,andBroesuggestadiachronicpathbywhichdissimilatoryphonologicalpressuresaffectthelexicon.
Whetherthepressuresarepurelydiachronicoralsosynchronicisaquestionthatfutureresearchissuretofocuson.
Whateverthenatureofcontrastpreservationprinciplesturnsouttobe,theprinciplesclearlyplayasubordinateroleingrammars.
Phonologicalalternationsandneutralizationsarerampant,asisthecreationofhomophonyinparadigms.
Evensettingasideallcasesofsystematicsyncretismwithinparadigms(seee.
g.
Baerman2005),westillndnumeroussituationsinwhichphonologicalneutral-izationscreatehomophony.
Totakejustoneexample,inRussiantheneutralizationofunstressed/a/and/o/produceshomophonybetweennominative/accusativeandgenitiveformsofneutero-stems(Baerman2005:809).
Adesinence-stressed9781405157681_4_003.
indd799781405157681_4_003.
indd7906/05/20115:08PM06/05/20115:08PM80SharonInkelasstem,forexample,'wine',hasdistinctnominative/accusative(vin[ó])andgenitive(vin[á])forms,butaroot-stressedstem,forexample,'place',isidenticalinbothcontexts(mést[R]),duetovowelreduction.
2.
5Non-derivedEnvironmentBlocking(NDEB)Ithasbeenwidelyobservedthatneutralizingphonologicalalternationswhicharetriggeredatmorphemeboundariesfailtoapplywhenthesamephonologicalenvironmentoccursmorpheme-internally.
"Derivedenvironmenteffects,"or"non-derivedenvironmentblocking"(NDEB),hasbeengenerallyattributedtocontrastpreservationpressures,althoughformalaccountsofthephenomenonvarywidely.
TheclassicexampleofaderivedenvironmenteffectoccursinFinnish:asnotedbyKiparsky(e.
g.
1993b),theneutralizingassibilationalternationconverting/t/to/s/before/i/and/e/appliesregularlyatstem-sufxboundariesbutdoesnotaffectmorpheme-internal/ti/sequences:tilat-a'order-infinitive'~tilas-i'order-past',but*silat-a,*silas-i.
Itwasthoughtinthe1970sand1980sthatNDEBeffectswereassociatedwiththeclassofcyclic,structure-changingrules;"StrictCycle"principlesproposedbyKiparsky(1982b)andMascaró(1976)formalizedthisapparentcorrelationaspartofthetheoryofLexicalMorphologyandPhonology.
However,sub-sequentndings(e.
g.
Hualde1989a;Kiparsky1993b)underminedtheStrictCyclicitycorrelation,showingthatNDEBeffectswerenotrestrictedtocyclicortostructure-changingrulesandthatnotallcyclicorallstructure-changingrulesexhibitNDEBeffects.
Alaterwaveofproposals,couchedinOptimalityTheory,focusedonthetensionbetweenpreservinginputsubstringsfromalteration,theideabeingthatmorpheme-internalsubstrings(e.
g.
Finnishti)wouldbepre-served,butderivedsubstrings(t-i)wouldbesubjecttoalternation(e.
g.
Burzio1997;ItandMester1996b;McCarthy2003a).
ArelatedapproachistakenbyLubowicz(2002),whosuggeststhatNDEBeffectsarethosewhichapplyonlywheninputfaithfulnesshastobedisruptedforsomeotherreason,forexample,resyllabication.
MorerecentworkhasgonebacktotheintuitionthatwasrstadvancedbyKiparskyinthe1960s,namelythatNDEBeffectspreservecontrast.
Kiparsky's(1968a)AlternationCondition,thoughawedindetailandlaterabandonedbyKiparsky(1982b)infavoroftheStrictCyclecondition,capturedthegeneraliz-ationthatagivenmorphemewillundergoaneutralizingphonologicalalternationonlyifthereisacontrastbetweenmorphologicalcontextsinwhichthealternationisapplicabletothatmorphemeandcontextsinwhichthealternationisnotapplicable,makingitpossiblefortheunderlyingformofthemorphemetoberecoverablebythelearner.
Forexample,inFinnish,theinitialtoftilatisalwaysinthecontextoftheAssibilationtriggeri.
BytheAlternationCondition,itcannotalternate.
Thenaltoftilat,however,sometimesoccursinanAssibilationcontextandsometimesdoesnot.
Asaconsequenceitmayalternatebetweentandswithoutobscuringthelexicalcontrastbetweenstem-nal/t/anad/s/.
9781405157681_4_003.
indd809781405157681_4_003.
indd8006/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology81TheAlternationConditionhasfoundrecentnewlifeinworkbyLubowicz(2003),whoproposesthatneutralizingalternationsbeconstrainedbyagrammaticalpressuretopreservecontrast.
Morphemesthatcontrastunderlyinglyshouldnotbeneutralizedineverypossiblesurfacecontextinwhichtheymightoccur.
Aswithanti-homophony,itisnotclearthattheeffectslabeledbyvariousana-lystsasNDEBareallofthesametype,functionallyorformally.
SomeeffectswhichcouldbeclassiedunderNDEBaremorelikelyduetotherestrictionofthepatterninquestiontoaparticularcophonologywithinthelanguage.
InJapa-nese,forexample,aconditionofbimoraicminimalityisimposedonafxedwords,leadingtovowellengtheningand/orinhibitingthedegreetowhichsufxedstemscanbetruncated;buttherequirementisnotimposedonbareroots,evenwhenusedaswords(It1990).
AsimilarminimalityphenomenoninTurkishisdocumentedinItandHankamer1989andInkelasandOrgun1995.
Thoughthespecicsoftheiranalysesdiffer,theseauthorsessentiallycharacterizetheminimalsizerestrictionsaspropertiesofstemsofaparticularmorphogicaltype.
Rootsarenotstemsofthistype,andevadetheminimalsizeconditionbyvirtueofitsneverbeingimposedonthematall.
ThissamesortofanalysisisgivenbyYu(2000)tothephenomenoninTohonoO'odhamwherebynalsecondarystressisprohibitedexceptinmorphologicallycomplexwords.
Yuprovidesanexplicitlycophono-logicalaccountinwhichtheassignmentofsecondarystresstonalsyllablesispartofthecophonologyofword-formationconstructions,butnotpartofthecophonologyappliedtoroots,eventhoseusedaswords.
CophonologicalaccountssuchasthesehavelittletosayaboutlocalNDEBeffectsofthetypeseeninFinnish;conversely,accountsoflocalNDEBeffectsdonotextendtothemoreglobaleffectsseeninJapanese,Turkish,andTohonoO'odham.
TheAlternationCondition,whetherinitsoriginalformorinLubowicz'smoremodernincarnation,isnotapplicabletoTohonoO'odhamsecondarystress,whichisnotneutralizing.
ItcouldwellbethattherearesimplytwotypesofNDEBeffects,whichcannotbemerged:thoseinvolvingneutralization,whicharetypicallysegmentalandthereforetypicallylocalandforwhichacontrastpreservationapproachisappropriate;andthosewhichinvolveprosody,whicharenotlocalanddonotinvolvecontrastneutralization,forwhichthecophonologicalaccountsaresuited(Inkelas2000).
ThetypologyofNDEBeffectsisclearlyanareaofongoingresearch.
2.
6LocalityandBracketErasureAnimportantquestionforanymodelofthemorphology-phonologyinterfaceiswhetherphonologicalpatternsapplyingtoonesubconstituentofawordcanmakereferencetopropertiesofembeddedstructure.
TheexistenceofNDEBeffectssuggeststhatphonologyneedstodistinguishcomplexfromsimplexstructures.
InFinnishitisnecessaryforthephonology,whenapplyingtoaformlike/tilat-i/,tohaveaccesstotheinformationthattilatand-idifferintheirmorphologicalstatus.
9781405157681_4_003.
indd819781405157681_4_003.
indd8106/05/20115:08PM06/05/20115:08PM82SharonInkelasBeyondNDEBproper,manyotherinteractionsbetweenmorphologyandpho-nologyhavebeenanalyzedusingphonologicalrulesorconstraintsthatdirectlyreferencemorphemeboundariesormorphemeidentity.
Onetypeofevidencethatisfrequentlyadducedisrootprominence.
InTurkish,forexample,thepresenceofalexicallyaccentedmorpheme(rootorafx)inawordoverridesthedefaultassignmentofstresstothenalsyllable.
However,whenbothalexicallyaccentedrootandalexicallyaccentedafxcombineinthesameword,onemustdisappear,sinceTurkishwordshaveonlyonestresseach.
InTurkishitisalwaysafxstresswhichdisappears,givingrisetotheappearanceofwhatMcCarthyandPrince(1995)havecharacterizedasauniversalprincipleofroot-faithfulness:grammarisalwaysmorefaithfultorootstructurethantoafxstructure,insituationswhereitisnecessarytochoose.
Alderete(1999,2001)hasanalyzedsimilarroot-prominenceeffectsinCupeoandJapanese,characterizingthemintermsofroot-faithfulness.
Itdoesnotparticularlymatter,ofcourse,whethertheanalyticaltoolisarootfaithfulnessconstraintorsomethingelse;whatmattersisthatthephonologicalgrammarmustbesensitivetothedistinctionbetweenrootsandafxes.
Incaseslikethese,somecophonologicalaccountshaveadifferentinterpretationofwhatisgoingon.
Itispossible,incophonologytheory,totreatafxationlikerealiza-tionalmorphology,theresultofamorphologically-specicphonologicalmappingthattakesastemasinputandproducesanoutputthatincludeswhatinmoretraditionalmorpheme-basedaccountswouldbecalledtheafx(Orgun1996;Inkelas1998).
Onthisimplementationofcophonologytheory,thephonologicalsubstanceofthe"afx"isnotpresentintheinput;onlythephonologicalsubstanceoftherootispresent.
Theasymmetrybetweenrootandafxonthisaccountdoesnotrequirereferencetomorphemeboundariesortotheidentityofmorphemetypes;itonlyrequiresreferencetoinput.
Themostextremeviewoftherelevanceofmorphemeboundariestophonology,then,wouldbethattherulesorconstraintswithinaparticularcophonologyarecompletelyinsensitivetomorphology,andthatmorphologicalsensivityarisesonlyindirectlybymeansoftheassociationofdifferentcophonologieswithdifferentmorphologicalword-buildingconstructions.
Themostpermissiveviewwouldgrantphonologyaccesstoallkindsofmorpho-logicalinformation.
Thiswastheoriginalassumptioningenerativephonology(ChomskyandHalle1968),inwhichallmorphemeboundarieswerevisibletophonologicalrules,andisstillprevalentintheOptimalityTheoryliterature(e.
g.
McCarthyandPrince1993).
Aviewthatfallssomewhereinbetweenwasdevelopedinthe1980sinthegeneralLexicalMorphologyandPhonologyframework(Kiparsky1982b;Mohanan1986),inwhichitwasassumedthatphonologicalrulesapplyingonaparticularcycle(orstratum,ifnon-cyclic)couldseemorphemeboundariescreatedonthatcycle(orinthatstratum),butthatoncesuchruleshadapplied,theinternalmor-phemeboundarieswouldbe"erased"oronsomeprinciplemadeinvisibletorulesapplyingonasubsequentcycle(stratum)ofmorphology.
(AversionofthisprincipleofbracketerasurecanbefoundinChomskyandHalle1968aswell.
)ThequestionofbracketerasureandtherelevanceofmorphemeboundarieshasdrawnlittledirectattentionsincetheriseofOptimalityTheory,asidefromafew9781405157681_4_003.
indd829781405157681_4_003.
indd8206/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology83workssuchasOrgunandInkelas(2002),ItandMester2002;Shaw2009).
WhilemanyanalysesinOptimalityTheoryallowphonologicalconstraintstodirectlyreferenceallembeddedmorphologicalstructure,itisnotalwaysclearwhetherthisfollowsfromnecessityorfromconvenience.
3WhenPhonologyAffectsMorphology:CombinatoricsThusfarwehavefocusedoncasesinwhichphonologicalpatternsdifferacrossdifferentmorphologicalzonesofcomplexwords.
Inthissectionweexamineadifferentkindofinterface,inwhichword-formationpossibilitiescanthemselvesbeconstrainedbyphonology,eitherbecauseofphonologicalrequirementsoninputstowordformationorbecauseofphonologicalrequirementsontheoutputsofwordformation.
Constraintsonwordformationcanresultinthechoiceofonesuppletiveallomorphoveranother,ortheycanresultinmorphologicalgaps,wherenooutput(oronlyaperiphrasticoutput)ispossible.
Thereareevencasesinwhichitappearsthatafxorderingisphonologicallydetermined.
3.
1SuppletiveAllomorphyThusfarwehavediscussedinteractionsascribabletogrammar.
Suppletiveallo-morphyisatypeofmorphology-phonologyinterfacewhichinvolvesthelexicon.
Suppletiveallomorphyisfamiliartoeverybeginningmorphologystudentasthesituationinwhichagivenmorphologicalcategoryhastwoormoreexponentswhichcannotbederivedfromacommonphonologicalformbutmustbestoredseparately.
Suppletiveallomorphyenterstherealmofthemorphology-phonologyinterfacewhenthechoicebetweenoramongsuppletiveallomorphsisphono-logicallydetermined.
Inanumberofsuchcases,thedistributionofsuppletiveallomorphsappearstoresonatewithphonologicalpatternsinthelanguage,suggestingthatthepho-nologicalgrammarcouldberesponsibleforhandlingtheallomorphy.
InModernWesternArmenian,forexample,thedenitearticletakestheshape-nfollow-ingvowel-nalnouns(e.
g.
katu-n'cat-def')and-Rfollowingconsonant-nalnouns(e.
g.
hat-R'piece');Vaux1998:252.
2SimilareffectsarefamiliarfromKorean,inwhichseveralsufxesexhibitV-andC-initialsuppletiveallomorphswhichoccurafterC-andV-nalstems,respectively.
Thus,thenominative,accusativeandtopic-markedformsofparam'wind'areparam-i,param-landparam-n,vs.
thecorrespondingformsofpori'barley':pori-ka,pori-rl,pori-nn(Paster2006:67,citingOdden1993:133).
AsresearcherssuchasMester(1994),Kager(1996),Anttila(1997a)andothershaveobserved,constraintsoptimizingsyllablestructure(e.
g.
NoCoda,orOnset)wouldautomaticallyentailtheselectionofallomorphswhichproduceCVsyllablesoverthoseresultinginheterosyllabicconsonantclusters(e.
g.
*hat-n,inArmenian)orvowelsequences(e.
g.
*katu-R).
9781405157681_4_003.
indd839781405157681_4_003.
indd8306/05/20115:08PM06/05/20115:08PM84SharonInkelasInabroadcross-linguisticsurveyofsuppletiveallomorphy,Paster(2006)uncoveredacontinuumofcases:somesuppletiveallomorphy(especiallycasesconditionedbysyllableormetricalstructure)iseasytocharacterizeasphono-logicallyoptimizing,whileothercasesofallomorphyseemarbitraryorevennon--optimizing.
Consider,forexample,thecaseofHaitianCreole,inwhichaparticulardeterminertakestheform-afollowingvowels(e.
g.
panié-a'thebasket',trou-a'thehole')and-lafollowingconsonants(e.
g.
pitit-la'thechild',madm-l'thehouse')(Paster2006:86,citingHall1953:32,viaKlein2003).
Thisistheexactoppositedistributionfromwhatisseenin,forexample,Korean,yetoverallthesyllablestructuresofthetwolanguagesaresimilar.
Ifoneallomorphicdistributionmakessensephonologically,theothercannot.
OrtakeArmeniannounpluralization:accordingtoVaux(1998:31),"monosyllabicnounstakethesufx-er.
.
.
andpolysyllabicnounstakethesufx-ner:ta,ta-er'meal(s)',dodo,dodo-ner'toad(s).
'"Forapparentlyarbitraryphonologicallyconditionedallomorphyofthiskind,lexicalsubcategorizationisacommonapproach(e.
g.
Kiparsky1982b;Inkelas1990;Booij2001;Paster2006).
Thelexicalentryincludesallsuppletiveallomorphs,someorallofwhicharelistedwithaselectionalframeidentifyingthephono-logicalenvironment.
InthecaseoftheArmeniannounplural,forexample,atleastoneofthesufxallomorphsmuststipulatethenumberofsyllablesthatthebaseofafxationisrequiredtohave.
Theotheronecanbetheelsewherecase,ifdesired:{[[q]er],[[]ner]}.
Pasterarguesforasubcategorizationapproachinallcasesofsuppletiveallomorphy,eventhosewhichtheallomorphdistributioncouldbeattributedtogrammarratherthanthelexicon.
Herargumentispartlybasedonthefactthatsuppletiveallomorphyisoftenopaque,conditionedbyinputfactorswhichareobscuredintheoutputbyphonologicalalternationsaffectingthederivedstem.
Insuchcases,inputconditioningisnecessaryevenifthedistributionofallomorphsmakesphonologicalsense.
PasterdiscussestheexampleofTurkish,inwhichthethird-personpossessivesufxhastwosuppletiveallomorphs:-I,usedafterconsonant-nalstems(ev-i'his/her/itshouse'),and-sI,usedaftervowel-nalstems(anne-si'his/her/itsmother')(Lewis1967;seealsoPaster2006:99).
Thisdistri-butionisrenderedopaquewhenintervocalicvelardeletionappliestoasufxedstem.
Inthethird-personpossessive,avelar-nalwordlikeinek'cow'combineswiththe-Iallomorph,asexpectedsinceinekisconsonant-nal.
However,theresultofvelardeletionis[ine.
i](orthographicine!
i),withthe"wrong"surfaceallomorph.
Asurfaceoptimizationapproach,giventhechoicebetween[ine.
i]and[inek-si],wouldalmostcertainlybeexpectedtopick[inek-si](oreven[ine-si]);CCclustersacrossmorphemeboundaries,aswouldoccurin[inek-si],arecom-monplaceandneverrepairedbydeletionorepenthesis,whereasVVclustersacrossmorphemeboundariesaretoleratedatnootherstem-sufxjuncturesinthelanguage.
Forthesereasons,Pasteranalyzesthiscasenotasoutputoptimiza-tionbutpurelyasinputselection.
InoneveryinterestingcaseofopaqueallomorphselectioninPolish,Lubowicz(2007)citesphonologicalcontrastpreservationasthemotivationforthechoicebetweensuppletiveallomorphs.
ThelocativeinPolishhastwosuppletive9781405157681_4_003.
indd849781405157681_4_003.
indd8406/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology85allomorphs:-eand-u.
Likeotherfrontsufx-initialvowelsinPolish,-etriggerspalatalizationofastem-nalcoronalconsonant:lis[t](nominative),oli"[c]-e(loca-tive)'letter'.
Exactlythosestemswhosenalconsonantisunderlyinglypalataltake-uinstead:li"[c](nominative),olili"[c]-u(locative)'leaf'.
Lubowiczattributestheselectionofthe-uallomorphtocontrastpreservation.
Exactlywhen-e,thepreferredallomorph,wouldmergethecontrastbetweenunderlyinglyplainandunderlyingpalatalroot-nalcoronalconsonants,-uisselectedinstead.
Itisimportanttonotethatthecontrastbeingpreservedhereisanabstractphono-logicalone.
While"letter"and"leaf"formaminimalpair,thesamedistributionof-eand-uisfoundwithrootsthatareindependentlydistinctinotherways,forexample,9obu[z](nominative),o9obu[z]-e(locative)'troublemaker',butpa[z](nominative),opa[z]-u(locative)'typeofbuttery'.
3.
2PhonologicallyMotivatedMorphologicalGapsInphonologicallyconditionedsuppletiveallomorphy,phonologicalgrammaticalconstraintsorthephonologicalrequirementsofindividualafxescontrolwhichstemscancombinewithwhichafxallomorphs.
Sometimesthephonologicalgrammar,orthephonologicalselectionalrequirementsofindividualafxes,canbesostrictastoblockmorphologicalcombinationaltogether,resultinginphonologicallydrivenmorphologicalgaps.
FormanyspeakersofTurkish,suf-xationisgrammaticalonlyiftheresultingwordisatleastdisyllabic(ItoandHankamer1989;InkelasandOrgun1995).
InDutch,thesuperlativeending-stcannotbeaddedtoadjectivesendingin[is],[sk],[st];thusbruusk'sudden'hasnolexicalsuperlativecounterpart(*bruusk-st[bryskst])butmustenterintoaperiphrasticsyntacticalternative:meestbruusk(Booij2005).
InTagalog,inxationoftheagentivefocusmarker-um-isimpossibleifthestembeginswith/m/or/w/,creatingparadigmgapsforsuchwords(SchachterandOtanes1972;OrgunandSprouse1999).
AnumberofsimilarcasesaresurveyedbyCarstairs-McCarthy1998.
Phonologicallyconditionedgapsdifferfromsuppletiveallomor-phyinthatthereisno"elsewhere"allomorph;withoutthisalternative,thewordsimplycannotbeformed,resultinginagap.
Oftenthereisasyntacticalterna-tive;forexample,inEnglish,forexample,thecomparativesufx-erattachesonlyto(looselyspeaking)monosyllabicstems;thusvast-erbut*gigantic-er,forcingspeakerstoresorttotheperiphrasticcomparative,forexample,moregigantic(Poser1992).
3.
3HaplologyEffectsMennandMcWhinney(1984)drawattentiontoacommoncross-linguisticpatternofprohibitingsequencesofhomophonousmorphemes,whichtheytermtheRepeatedMorphConstraint(RMC).
Awell-knownexampleoccursinEnglish,wherethepossessiveending/z/isnotaddedto–oratleastnotrealizedon–wordsendinginthehomophonouspluralsufx/z/;thusdogsanddogs'arepronouncedidentically([dagz])inthephrasesthedogshatetheircollarsandthe9781405157681_4_003.
indd859781405157681_4_003.
indd8506/05/20115:08PM06/05/20115:08PM86SharonInkelasdogs'collarsdrivethemcrazy.
Irregularpluralstakethepossessive(children's)andsodowordsendinginthestringshomophonouswithallomorphsoftheplural(i.
e.
[()z]or[s]),forexample,Katz's[ktsz]orcats[kts].
TheRMCmay,accord-ingtoMennandMcWhinney,resultininhaplology,asintheEnglishexampleofcats',whereasinglephonologicalexponent[s]standsforwhatappeartobetwomorphemes.
InothercasestheRMCcanalsotriggersuppletiveallomorphyoravoidance,inwhichaperiphrasticalternativeispreferred.
Forexample,theEnglishadverbial-lyending(quick(adj.
),quickly(adv.
))cannotcombinewiththoseadjectivesalreadyendingin-ly,forexample,manlyorheavenly:*manlily,*heaven-lily(adv.
).
Thelexicalgapwhich*manly,andsoon.
cannotllmustbeapproxi-matedbyaphraselikeinanmanlyfashion.
Sinceitaffectswordformandcreatesmorphologicalparadigmgaps,theRMCgeneralizationwouldseemtobeaclearcaseofphonologyinterferingwithmorphology.
TheRMCis,however,clearlynotuniversal,evenwithinalanguage.
Thereexistmanyunperturbedsequencesofhomophonousmorphs;therearealsoinstancesofsuppletionandavoidanceinwhichmorphrepetitionisnotanissue.
English,forexample,permitssequencesofthepluralorpossessivefollowedbythehomophonousreducedformofis,forexample,oneofthecats's[ktsz]trappedinthecloset!
orWhoseguacamoledoyoulikebestJohn'sis[dLnzz]clearlythewinner.
TheRMCappliesonlytosequencesofpluralandpossessive,nottoallsequencesof/z/morphemes.
AnotherexampleoccursinTurkish,whichusesthesamesufx(/-I/~/-sI/)bothtomarkthird-personsingularpossessors(aile'family',aile-si'his/her/itsfamily';araba'car',araba-s'his/her/itscar')andalsoasamarkerattheendofhead-modiercompounds,inwhichthepossessionrelation,ifany,isquiteabstract(Lewis1967:42):ailearaba-s'familycar'.
Thepossessivesufxcannotoccurtwiceinsuccession;therefore,inisolation,compoundslikeailearaba-sareactuallyambiguousbetweenapossessed(e.
g.
'his/her/itsfamilycar')andunpossessedreading(Lewis1967:46).
Theungrammaticalityofadoublyafxedpossessivecompound,forexample,*ailearaba-s-s,cannot,however,sim-plybeattributedtotheRMC.
Otherpossessivesufxes,forexample,rstpersonpossessive/-m/,"associative"-/lI/and"occupational"/-CI/,arealsoincomple-mentarydistributionwiththecompound-markingpossessivesufxeventhoughtheyarenothomophonouswithit(Lewis1967:49–50):ailearaba-m'myfamilycar',not*ailearaba-s-m,etc.
)Thuseventhisapparentlytransparentcaseofarepeatedmorphprohibitiononpossessive/-I/~/-sI/turnsouttobepartofamoregeneralpatternofmorphemeco-occurrence.
How,then,arewetoknowwhetherafxco-occurrencerestrictionsbetweenhomophonousafxesareadis-tinctsubtypeofafxco-occurrencerestrictionsgenerallyFurtherresearchisrequiredinthisarea,butanswersarelikelytobeoftwotypes.
Oneisstatistical:ifhomophonousafxpairsformalargerthanexpectedsubsetoftheclassofmorphemepairsthatcannotoccurnexttoeachother,theRMCwouldbesup-ported,thoughthistaskwouldbehardtoaccomplishgivencurrentdata.
Asecondpossibleanswerwouldbetoshowthatthelexicalgapsorlexicalambi-guitiesresultingfromRMCeffectspatterndifferentlyfromthoseresultingfromother,morearbitrarymorphemeco-occurrencerestrictions.
9781405157681_4_003.
indd869781405157681_4_003.
indd8606/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology873.
4LinearOrderAnumberofcaseshavebeendescribedinwhichphonologyconstrainsthelinearorderofmorphemes.
Mortensen(2006)hascollectedaveryinterestingsetofexamplesinwhichconstituentsincoordinatecompoundingareorderedaccord-ingtotheirphonologicalproperties,principallyvowelqualityandtone.
InonedramaticcasefromJingpho,whichMortensendrawsfroma1990monographinChinesebyQingziaDai,theorderofelementsincompoundswithcoordinatesemanticsfollowsfromtheheightofthetonic(root)vowels:thestemwiththehighervowelalwaysprecedesthestemwiththelowervowel.
Thuslù.
-á'drink-eat=food'isagrammaticalcompound,while*á-lù.
,withthesamepre-sumedmeaning,wouldbeungrammatical(Mortensen2006:222–223).
Mortensendocumentsmanysuchcompoundingcases,mainlyinvolvingvowelqualityand/ortone,inwhichtheorderofelementsfollowsascale.
Sometimesthescaleisphoneticallytransparent,asintheJingphocaseofvowelheight,andsometimesnot,whenhistoricalchangeshaveobscuredtheoriginalphoneticorphonologicalbasisforthescale.
Anotherareainwhichphonologydetermineslinearorderisfoundwith"mobileafxes,"discussedbyFulmer1991,Noyer1994,andKim2008.
Thesevaryfreelybetweenprexalandsufxalattachment,withphonologicalconsiderationsbeingthedecidingfactors.
IntheSanFranciscodelMardialectofHuave,forexample,thesubordinatemarkermattachesasaprextovowel-initialbases(m-[u-ty]'sb-TV-eat=(that)s/heeats')butasasufxtoconsonant-initialbases([mojk-o]-m'face.
down-v-sb=(that)s/heliesfacedown').
Similarbehaviorisexhibitedbyotherafxes,includingthestativen:n-[a-kants]'st-tv-red=red'vs.
[pal-a]-n'close-v-st=closed'(Kim2008:332).
AsproposedbyKim2008and,forsimilarfactsinSanMateoHuave,byNoyer1994,suchcasescanbemodeledinOptimalityTheorybythegeneralschemaproposedbyMcCarthyandPrince(1994a)inwhichphonologicalconsiderations("P")outrankmorphologicalconsiderations,forexample,afxordering("M").
InHuave,accordingtoKim,mobileafxesarepreferentiallysufxing(the"M"condition),butwillprexifsufxationwouldproduceconsonantclustersthatwouldrequireepenthesis(the"P"condition)(pp.
340–341).
Thusforabaselike[a-rang]'tv-do',m-prexation(m-a-rang)ispreferredoverm-sufxation(*a-rang-m,*a-rang-am),sincethelatterwouldproduceanunsyllabiableclusterrequiringrepair.
Incaseswherebothprexationandsufxationoptionswouldrequireepenthesis,sufxationispreferred:rst-personscombinesasasufxwithbaset-a-rang'cp-tv-do=did(it)'toyieldt-a-rang-as,withepenthesis,ratherthanasaprex(*s-tarangor*sa-tarang)(pp.
340,342).
Ingeneral,however,theeffectthatphonologyhasbeenarguedtoplayintheorderingofmorphologicalelementsisfairlylimited.
Thegreatbulkofafxorderingisdeterminedbythemorphology,notbythephonology.
TotakeaverysimpleexamplefromTurkish,considertheinteraction[FIXIPA]ofthe"occupational"sufx/-CI/andcaseendings,forexample,thedative/-E/.
Bothcanattachtoroots.
Turkishepenthesizesvowelstobreakuptriconsonantal9781405157681_4_003.
indd879781405157681_4_003.
indd8706/05/20115:08PM06/05/20115:08PM88SharonInkelasclusters,andepenthesizesglidestobreakupvowel-vowelsequencesatstem-sufxboundaries.
Thuswendalternationslikethese:/jeni-CI/→[jenidi]'new-prof'(yenici),/lm-CI/→[limdi]'lm-prof=lm-maker'(limci);/jeni-E/→[jenije]'new-dat'(yeniye),/lm-E/→[lme]'lm-dat'.
Whenboth/-CI/andacasesufxoccurinthesameword,afxorderisxed.
/-CI/,asaderivationalsufx,alwaysprecedescase:/jeni-CI-E/→[jenidije]'new-prof-dat'(yeniciye;/lm-CI-E/→[limdije]'lm-maker(dative)'(limciye).
Inthelatterexample,twoepenthesisoperationsarerequiredtobringthesyllablestructureoftheresultingwordintoconformitywithTurkishrequirements.
Bycontrast,thealter-nativeafxorderingwouldproduceperfectlywell-formedsyllableswithnoneedforepenthesis:/lm-E-CI/→[l.
me.
di](*lmeci).
But*lmeciiscompletelyimpossibleinTurkish;phonologicalconsiderationsdonottrumpmorphologicalconstraintsonrelativeafxorder.
Paster(2005)exploresonewell-knownapparentcase,fromtheFuutaToorodialectofPulaar(Fula),inwhichphonologyhasbeenclaimedtoorderafxes.
InastudyoftheGombedialect,Arnott(1970)observedthatanumberofCorCVsufxesinthesamegeneral"zone"ofthewordappeartooccurinaphonologic-allydeterminedorder:allsufxeswith"t"precedeallsufxeswith"d,"whichprecedeallsufxeswith"n,"whichprecedeallsufxeswith"r.
"Pastercitessimilarexamplesofthis"TDNR"templatefromFuutaTooroPulaar,forexample,jat-t-id-ir-an-ii'take-intensive-comprehensive-modal-dative-past'andyam-{-it-in-ir-ii'healthy-denominative-repetitive-causative-modal-past'(Paster2005:164).
AsPasterargues,however,theorderofafxesinbothdialectsofPulaarconformstosemanticorderingprinciplesofthekindarticulatedbyBybee(1985)andRice(2000);thereisnocaseinwhichthephonologicalTDNRtemplatecon-travenesanorderingthatonemightotherwiseexpectonmorphologicalgrounds,andthusnoclearevidencethatphonologyisinterferingwithmorphology.
PasteralsoobservesthattheTDNRtemplate,inwhichconsonantsonorityincreasesfromlefttoright,isnotcompletelyconvincingasaphonologicalphenomenon,sinceinactualPulaarwords,vowelstypicallyseparatetheconsonantswhichcorrespondtotheelementsoftheTDNRtemplate.
Asonority-basedtemplatelikeTDNRwouldmakemoresense,Pasterargues,ifconsonantswerebeingorderedbysonorityinordertotintoasinglesyllableonsetorcoda,butthisisnotthecaseinPulaar.
Insum,Pasterconcludes,thePulaarpatternissignicantforcom-ingcloserthananyotherexampletobeingacaseofphonologically-drivenafxsequencingbutstillnotfullymeetingthatdescription.
Themostsubstantialinuenceofphonologicalconsiderationsonthelineariza-tionofmorphemesisfoundininxation,whichisgenerallyviewedasbeingjustlikeafxationexceptthattheafxisphonologicallypositionedwithinthesteminsteadofperipheraltoit(seee.
g.
Moravcik1977,2000;McCarthyandPrince1993;Yu2007).
Theinterestofinxationforthephonology-morphologyinterfaceliesinphonologicalgeneralizationsaboutwhereinawordaninxcanappearandaboutwhat,ifanything,motivatesinxationsynchronically.
SurveysofinxationfromMoravcsik1977toYu2007havefoundasmallandprincipledsetofrecurringsitesforsegmentalinxes:nexttotheinitialornal9781405157681_4_003.
indd889781405157681_4_003.
indd8806/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology89consonantorvowel,or(inlexicalstresslanguages)nexttoametricalprominence.
AsYuobserves,thesesitesaredenedintermsoftypesofelementsthatallwordsintherelevantlanguagecontain.
Allwordscontainconsonants,vowels,and(instresslanguages)stress.
Bycontrast,thereisapparentlynoevidenceofinxationtosyllableswithparticulartones,ortosyllablescontainingparticulartypesofsegments(e.
g.
fricativesorejectives)oreventoheavysyllables(e.
g.
thosewithlongvowelsorconsonantclusters).
Theseareelementsthatlanguagesdonottypicallyrequireallwordstopossess.
Onlydependentmorphemes(afxes,andrarelyclitics;Harris2000)haveeverreliablybeenshowntoinx;inxationisapparentlynotapossiblepropertyofcompoundingorphrasalcombination,asidefromthesuggestiveexampleofexpletiveinxationbeforeastressfootinEnglishwords(amálgamàted→amálga-bloody-màted,Kàlamazóo→Kàlama-fuckin-zóo,andsoon.
;McCarthy1982).
Astimulatingtheoryofinxationwasintroducedintheearly1990s,withintheframeworkofOptimalityTheory,byMcCarthyandPrince(1993),whoobservedthatmanycasesofinxationcouldbeinterpretedasimprovingtheprosodicstructureofthederivedwordincomparisontothestructurethewordwouldhaveiftheinxedelementwereinsteadadxed.
InxationwasakeymotivatorinMcCarthyandPrince'sproposalthatatleastsomephonologicalconstraints"P"canoutrankmorphologicalconstraints"M,"particularlythosehavingtodowiththeedge-alignmentofafxes.
Themostconvincingexamplesbroughtforthforthisview,termedthe"Pho-nologicalReadjustment"viewinYu2007,involvesyllablestructure.
McCarthyandPrince'soriginalexampleconcernstheagentivefocusmarker,-um-,inTaga-log,whichprecedestheinitialvowel:bilih'buy'→b-um-ilih,gradwet'graduate'→gr-um-adwet,andsoon.
AccordingtoMcCarthyandPrince,-um-isaprex,subjecttothe"M"constraintAlign-L(um-,Stem),whichisoutrankedbya"P"constraintbanningclosedsyllables(NoCoda).
Inconsideringthepossibleloca-tionsfor-um-incaseofgradwet,theprexingcandidateum.
grad.
wethastwoclosedsyllables,whiletheinxingcandidategr-u.
m-ad.
wethasonlyoneclosedsyllable,satisifyingNoCodabetter.
InxationispreferredbecauseoftheP>>Mranking.
Bycontrasttonumerousexamplesinwhichinxationcanbeinterpretedasimprovingsyllablestructure,veryfewexampleshavebeenfoundinwhichinxationcanbeconstruedasimprovingsegmentstructure(seee.
g.
Yu2007:Chapter6).
Thisasymmetryraisesdoubtsaboutthegeneralityofthepotentialfor"P"constraintstooutrankmorphologicalalignment.
AsecondmajorissueconfrontedbytheP>>Mmodelislocality.
Ifinxationismisalignmentwiththeaimofavoidingbadstructures,theP>>Mmodelpre-dictswhatYucalls"hyperinxation"orunboundedinxation,apointmadealsobyOrgunandSprouse(1999)andMcCarthy(2003b).
InSection3.
2wementionedthattheTagalog-um-inxisprohibitedfromcombiningwithm-orw-initialstems.
Modelingthisprohibitionasaphonologicalconstraint(e.
g.
OCP,followingOrgunandSprouse),aP>>Mrankingwouldpredictthat-um-couldinxfurtherintothewordthanusualtoavoidtheundesiredmuorwusequence.
Thisdoes9781405157681_4_003.
indd899781405157681_4_003.
indd8906/05/20115:08PM06/05/20115:08PM90SharonInkelasnothappeninTagalog,nordoesthecomparablesituationappeartoariseinanyotherlanguage:inxesaretightlyrestrictedtoappearnearedges.
McCarthy(2003b)addressesthisproblemforP>>Mbymodifyingalignmentconstraintssothattheyarecategorical;Tagalogumissubjectbothtoaviolableconstraintforbiddingitfrombeingseparatedbyasegmentfromthebeginningofthewordandtoaninviolableconstraintforbiddingitfrombeingseparatedbyasyllable(ormore)fromthebeginningoftheword(p.
95ff.
).
AradicallydifferentapproachtoinxationisofferedbyYu(2007),whoobservesthatmanycasesofinxationfalloutsidetheP>>Mmodelinthesenseofnotbeingprosodicallyimprovinginanydiscernibleway,yetstillconformtothelocalitygeneralizationsthatwerepotentiallyproblematicfortheP>>Mmodel.
YupointsoutthatsomecasesofinxationneitherimprovenorworsensyllablestructureForexample,theHuanegativeinx-.
a-,whichisCVinshape,inxesbeforethenalsyllable:harupo→haru-.
a-po'(not)slip',zgavo→zga-.
a-vo'(not)embrace',eventhoughadxing(.
a-harupo,harupo-.
a)wouldhaveproducedequallygoodsyllables(Yu2007:30,citingHaiman1980).
Othercasesofinxationargu-ablymakesyllablestructureworse.
Forexample,thenominalizing-ni-inxinLetifollowstherstconsonant(e.
g.
kaati'tocarve'→k-ni-aati'carving',polu'tocall'→p-ni-olu'actofcalling,call'),producingmarkedconsonantclustersandvowelsequencesthatwouldbeavoidedbysimpleadxation(e.
g.
ni-polu,polu-ni)orinxationtoadifferentposition(e.
g.
po-ni-lu)(Yu2007:28,citingBlevins1999).
Yuconcludesthatlocalityandgenerality,ratherthanphonologicaloptimization,arethemaingeneralizationsthatasynchronicmodelofinxationshouldcapture,andproposesalexicalsubcategorizationapproachbuildingon,forexample,BroselowandMcCarthy1983,McCarthyandPrince1986,Inkelas1990.
Onthisapproach,eachinxisassociatedwithalexicalstatementdeningitspositionrelativetooneorbothedgesofthestemitcombineswith.
Phonologicalentitiestowhichsuchstatementsarepermittedtorefercomefromasmalllistof"pivots"thatcross-linguisticallyareshowntoseparateinxesfromstemedges:segments,syllables,andstressedconstituents(Yu2007:52).
4WhenPhonologyAffectsMorphology:FormSomemorphologicalconstructionsarephonologicallycompositional,inthesensethatthemorphologycombinestwoormoreelementswithxedphonologicalshapesandthe"regular"rulesofthephonologyapplytogivethecombinationitssurfacephonologicalform,whichvarieswiththeshapesoftheinputmor-phemes.
Forexample,prexationofpre-inEnglish(pre-register,pre-ordained)isasimplematterofconcatenatingthexedstring[pri]withabase.
Butinsomemorphologicalconstructions,extrinsicconsiderationsconstrainordetermineout-putphonologicalshape,withinputmorphemesconformingtooutputshaperequirementsinsteadofdeterminingoutputshapethemselves.
Wewillsurveytwosuchphenomenahere:templaticmorphologyandreduplication.
Thesearecommonlytermed"prosodicmorphology,"becauseineachcaseamorphemeis9781405157681_4_003.
indd909781405157681_4_003.
indd9006/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology91expressedphonologicallyinawaythatisnotconstantacrossthesetofstemsformedfromthatconstructionbutispredictablefromconstruction-specicmetri-calorsyllabicconstraintsonthephonologicalshapeofthecomplexstem.
Prosodicmorphologyisoftendescribedasatrio,withinxationasthethirdmember;however,theconsiderationsdrivinginxationareratherdifferent,asseenabove.
4.
1TemplatesTemplatesaremorphologicalconstructions,typicallyassociatedwithspecicderivationalorinectionalmorphologicalcategories,whichdirectlyconstrainthephonologicalshapeofthederivedstem.
McCarthy(1979a,1981)brokenewgroundbyanalyzingthexedshapeofspecicderivationalsubtypesofArabicverbsascomposedoftemplatesconsistingofCVtimingunits.
Thesetemplates,eachexpressingaspecicmorphologicalcategory,combinewithothermorphemeswhichconsistofconsonants,andwithothersconsistingofvowels,toformcom-plexwords.
Forexample,theconsonantalroot/ktb/'write'combineswiththe"perfectivepassive"vocalicmorpheme/ui/andthe"causative"templateCVC-CVCtoformkuttib.
Intheirseminal1986paper,McCarthyandPrinceshowedtheroleofprosodicstructureindeningthevarioustemplatesnotjustintheArabicrootandpatternmorphologicalsystembutmoregenerallyinprosodicmorphologycross-linguistically.
AccordingtoMcCarthyandPrince,templatesarealwaysdenedintermsoftheuniversallyaccessibleunitsofmora,syllable,andfoot,ratherthanintermsoftheC,V,orXtimingunitsproposedinearlierworkbyMcCarthy(1979a,1981),Leben(1980),Hyman(1985),andothers.
Sometimestemplatesconstraintheshapeofstemsorwordswithoutcontribut-inganyparticularsemanticorsyntacticfunctionoftheirown.
Asimplecaseofthisoccurswithminimalwordsizeconstraints,whichcancompelepenthesisorotherphonologicalaugmentationstrategiesinshortwords.
InSwati,asinmanyotherBantulanguages,adisyllabicminimalityrequirementonwordscompelstheuseofadummysufx-niinverbsthatwouldotherwisebemonosyllabic,asituationwhicharisesinimperatives,theonemorphologicalenvironmentwithnoprexes.
Thus,whiletheinnitiveofthestem/dlá/'eat'isk-dlá,disyllabicbyvirtueofcontainingtheinnitiveprex,theunprexedsingularimperativeisdlá-ni,withaugmentationthatisnotrequiredforverbsformedfromlongerstems,suchas/bóna/'see':k-bóna'inf-see=tosee',bóna'see(singularimpera-tive)'(Downing2006:3).
InLardil,uninectednounsaresubjecttoapocope,seeninalternationslikewiwala-n'bushmango-nonfut.
acc',wiwala-'bushmango-fut.
acc',butwiwal'bushmango,from/wiwala/;karikari-n'buttersh-nonfut.
acc',karikari-wu'buttersh-fut.
acc',butkarikar'buttersh'(from/karikari/)(Hale1973:424).
Apocopeisblockedwhentheresultwouldhaveonlyoneshortvowel,forexample,kela-n'beach-nonfut.
acc',kela-'beach-fut.
acc',kela(*kel)'beach',from/kela/(Hale1973:421).
Uninectednounswithonlyoneshortvowelareevensubjecttoaugmentation,sothattheyachievebimoraicsize:óer-in'thigh-nonfut.
acc',óer-u'thigh-fut.
acc',butóera'thigh',from/óer/(Hale1973:427).
UninectednounsinLardilareclearlyaffectedbytheprosodiclimitation9781405157681_4_003.
indd919781405157681_4_003.
indd9106/05/20115:08PM06/05/20115:08PM92SharonInkelasonwordsize.
(ForfurtherdiscussionofLardil,seeforexample,KenstowiczandKisseberth1979,It1986,Blevins1997,andBye2006).
Amoreinvolvedcase,inwhichprosodictemplatesconstrainstemshape,occursinYawelmani(Archangeli1983,1991,basedonNewman1944).
Intheverbalsystem,eachrootandafxislexicallyassociatedwithoneofthreeprosodictem-plates:alightsyllable,aheavysyllable,andaniambicfootconsistingofalightsyllablefollowedbyaheavysyllable.
Thesetemplatesdeterminetheformoftheroot.
When,inthesameword,anafxandrootareassociatedwithconictingtemplates,theoneassociatedwiththeafxprevails,leadingtorootalternationsinrelatedstems.
Forexample,theroot"walk"hasadefaultiambictemplate,asinhiweet-en'willwalk',butshortenstoaheavysyllableandtoalightsyllablewhencombiningwithsufxesassociatedwiththecorrespondingtemplates,asinheütihni'onewhoisroaming'()ihni)andhiütiay'whilewalking'()iay)(Archangeli1991:232,citingNewman1944:101,110,136).
Prosodictemplatescanevenspecifysegmentalcontent.
InTieneverbs,deriva-tionalstemsareconstrainedbyaCV(C)VC-templatewhoseconsonantsaresub-jecttotwomajorrestrictions(HymanandInkelas1997andHyman2006a,basedonEllington1977).
InCVCVC-stems,themiddleconsonant(Cmed)mustbecoronalandthenalconsonant(Cn)mustbenon-coronal,thatis,labialorvelar.
Theserestrictionscanforcethechoiceofinxalallomorphsofsufxessuchasthestative,whichhasbothaninxalallomorph(withcoronal/l/)andasufxalallomorph(withnon-coronal/k/):kab-'bedivided'~ka-la-b-(stative);yat-'besplit'~yat-ak-(stative);sQn-'write'~sQn-Q-(stative).
C2andC3mustalsoagreeinnasality,leadingtonasal~oralalternations:vwu-'bemixed'+-ek-(stative)→vwu-e-;dim-'becomeextinguished'+-se-(causative)→di-se-b-.
InrecentworkinOptimalityTheory,startingwithMcCarthyandPrince1994a,researchershavearguedthattemplatesareemergentartifactsofconstraintinter-action,ratherthanabstractstructuresmanipulatedbygrammar.
InArabicandTiene,forexample,itmightbepossible,insteadofstipulatingbaldlythatthe(derivational)verbstemmustbeCV(C)VC-inshape,orevensimplythatitmustbebimoraic,toletthisproleemergefromconstraintslikeFt-Bin(feetarebinary)andAll-Ft-Left(everyfootmustbeinitial),rankedhighinthemorphologicalenvironmentoftheverbstem.
TheemergenttemplateapproachhasbeenappliedfruitfullytomanycasesofreduplicationbyMcCarthyandPrince(1995)aswellasGafos(1998a)andHendricks(2001),amongothers,andhasbeenextendedbeyondreduplicationbyDowning(2006).
Themotivationforderivingratherthanstipulatingtemplatesistwo-fold:rst,derivingtemplatesfromindependentlyneededmarkednessconstraintsshouldyieldamorelimited,principledsetofpossibletemplatesthanwhatitispossibletostipulate,andsecond,templatesconstrainforminwaysotherthansimpleprosodicsize.
Markednessconstraintscanconstrainsegmentalformaswell.
TheexibilityofemergenttemplatesisusefulincharacterizingcaseslikeTiene,inwhichtherestrictionaboutconson-antalplaceofarticulationinverbscannotbeexpressedbyannotatingparticularprosodicpositionsforsegmentalfeatures.
CnisunrestrictedinCVVCstems;Cnisconstrained,bydissimilatoryprinciples,onlyifCmedispresent.
ThiskindofcontingentrestrictionisbettersuitedtoconstraintsofthesortpositedbyHyman9781405157681_4_003.
indd929781405157681_4_003.
indd9206/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology93andInkelas,inwhichCmedandCnmustdifferinplaceofarticulation(andCmedmustbecoronal).
4.
2ReduplicationReduplicationisthedoublingofsomepartofamorphologicalconstituent(root,stem,word)forsomemorphologicalpurpose.
Totalreduplicationduplicatestheentireconstituent.
Itisoften,thoughnowherenearalways,semanticallyiconic,asintheduplicationofnounswithhumanreferencetoformpluralsinWarlpiri(kurdu'child',kurdu-kurdu'children';wirriya'boy',wirriya-wirriya'boys')(Nash1986:130).
Partialreduplication,whichexhibitsaverywiderangeofmeanings,usuallyinvolvesaprosodicallycharacterizedtemplateforthereduplicant.
Forexample,McCarthyandPrince(1986)analyzethereduplicatingprexmarkingprogressiveaspectinMokileseasabimoraicsyllable:poadok[pQdok]'toplant'→poad-poadok[pQdpQdok]~poah-poadok[pQQpQdok]'tobeplanting';piload[pilQd]'topickbread-fruit'→pil-piload[pilpilQd]'tobepickingbreadfruit';kohkoa[kookQ]'togrindcoconut'→koh-kohkoa[kookookQ]'tobegrindingcoconut'(HarrisonandAlbert1976:60,220).
Typicallythebaseexceedsthereduplicanttemplateinsizeandthustheresultingreduplicationis,asintheseMokilesedatacited,partial.
Occa-sionally,areduplicanttemplatewillbebiggerthanthebase;inwhichcasereduplicantaugmentationoccurs.
Forexample,theMokileseformpa'weave',whichismonomoraic,reduplicatesaspah-pa[paapa],withabimoraicreduplicant(HarrisonandAlbert1976:60).
Whiletypicalreduplicantshapesaredescribedintheprosodicunitsofmora,syllableandfoot,Moravcsik(1977)iscreditedwiththeobservationthatreduplic-ationrarelyunambiguouslycopiesanexistingmora,syllableorfootfromthebase.
Rather,asmodeledbytheorieslikeProsodicMorphology(McCarthyandPrince1986)and,subsequently,approachestoreduplicationwithinOptimalityTheory(McCarthyandPrince1994a,1995),templaticrequirementsseemtobeoutputrequirementsonthereduplicant.
InMokilese,whatcopiesisenoughbasemater-ialtoeshoutaheavysyllablereduplicant,evenifthecorrespondingstringisnotitselfasyllableinthebase,(e.
g.
reduplicant[pil],frombase[pilQd]).
InOptimalityTheorythisoutputorientationcanbemodeledbystatingreduplicantshapeastheoutputrequirementRed=q[[.
McCarthyandPrince(1994a)andUrbanczyk(1996),workinginGeneralizedTemplateTheory,and,fromadifferentangle,Downing2006,havepursuedthegoalofderivingtemplatesratherthanstipulatingthem.
Indifferentways,theseresearchersproposethatreduplicantsassumethecanonicalphonologicalformofwhatevermorphologicalconstituenttype(afx,stem,morphologicallycomplexstem)theyinstantiate.
Thisformdoesnothavetobestipulatedspeciallyforthereduplicantbutismotivatedmoregenerallyforthelanguage,orevencross-linguistically.
Recentliteraturehassug-gestedthatsomereduplication,particularlywhenlimitedtoconsonants,maynothaveevenanindirectprosodictemplaticcharacterizationatall.
Hendricks(1999,2001)pointstocasessuchasexpressivereduplicationinSemai,whichcopiestherstandlastconsonantofthebase:p-paya'appearanceoflargestomachcon-stantlybulgingout',cw-cruha"w'soundofwaterfall,monsoonrain'(Difoth1976).
9781405157681_4_003.
indd939781405157681_4_003.
indd9306/05/20115:08PM06/05/20115:08PM94SharonInkelasPartialreduplicationusuallyduplicatesthatedgeofthestemtowhichthereduplicantisclosest,butopposite-edgereduplication(notincludingthedual-edgeversionfoundinSemai)occursaswell.
AdozenorsocasesaredocumentedinsurveysbyNelson2003,2004;Kennedy2003;andRiggle2003;alltargetthebeginningportionofabase,forexample,KoryakCVCreduplicationmarkingabsolutecase:mXtqa'oil'→mXtqa-mXt;qanga're'→qanga-qan.
Partialreduplicationisalsocommonlyinxing,asinChamorro(Topping1973:183),wherehabitual/continuativeCVreduplicationtargetsstressedsyllables(hátsa'lift'→há-ha-tsa,hugándo'play'→hugá-ga-ndo)andintensifyingCVreduplicationtargetsthenalsyllable(métgot'strong'→métgo-go-t'verystrong',álang'hungry'→ála-la-ng);seeforexample,BroselowandMcCarthy1983:55–56).
Internalreduplicationusuallyduplicatesadjacentmaterial,asintheseexamples,buttherearesomeexceptionstothis.
InWasho,forexample,pluralreduplicationinxesamorainthevicinityofthestressedsyllable.
Incasethestressedsyllableisclosed,asinnén.
t'u'oldwoman(nom.
)'or.
éw.
i.
'father'sbrother',reduplica-tioncopiesanon-adjacentCVstring:ne.
t'ún.
t'u-u'oldwomen(nom.
)',.
e.
íw.
i.
'father'sbrothers'(Yu2005:440,citingJacobsen1964).
Creekpluralsareformedbyinxingacopyofthestem-initialCVbeforethestem-nalconsonant(Riggle2003,citingBooker1980;Haas1977andMartinandMauldin2000):holwak-í"'ugly,naughty'→holwa"-ho-k-í",falápk-i"'crooked'→falap-fa-k-í".
4.
2.
1IdentityEffectsinReduplication:OverandUnderapplicationSinceWilbur'sinuential(1973)dissertation,researchershavepaidspecialattentiontophonologicalopacityarisinginreduplicationconstructions.
Forexample,considerJavanesetotalreduplication,whichhaspluralizingsemanticsandcanapplytoverbsandadjectives.
Whensufxed,forexample,bydemonstrative-e,redupli-catedformsexhibit"overapplication"ofintervocalich-deletionandunderapplica-tionofclosedsyllablelaxingandstem-nalconsonantdevoicing(InkelasandZoll2005:146,148,citingDudas1976:207–208):(7)GlossStem-demonstrative'broken'bRìahbRìa-enormalapplicationofintervocalich-deletionbRìah-bRìahbRìa-bRìa-e"overapplication"ofh-deletioninrststemcopy(demonstrative)'cylindrical'gilXkgilig-enormalapplicationofintervocalicvoicinggilXk-gilXkgilig-gilig-e"underapplication"ofstem-naldevoicinginrststemcopy(demonstrative)9781405157681_4_003.
indd949781405157681_4_003.
indd9406/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology95Althoughopacityoccursoutsideofreduplicationaswell,itsappearanceinredu-plicativeexamplesliketheseisinterpretedbyWilbur(1973),whopositsaredu-plicativeIdentityPrinciple,andinBase-ReduplicantCorrespondenceTheory,whichpositsBase-Reduplicantfaithfulnessconstraints,asdrivenbythefunctionalneedtokeepthetwopartsinreduplication–baseandcopy–segmentallyiden-tical.
Thisimperative,whileapparentlyobeyedinJavanese,isnotsatisedineverycase.
Forexample,Urbanczyk(1996)andStruijke(2000)drawattentiontoareduplicativeconstructioninLushootseed,illustratedbyexampleslikewális'typeoffrog'→wá-w'lis'littlefrog',caq'(a)'spear'→ca-cq''actofspearingbiggameonwater',andsoon.
(Urbanczyk1996:167).
UrbanczykandStruijkeanalyzethispatternasCVprexingreduplicationaccompaniedbysyncopeinthebase.
Onthisaccount,underapplicationofsyncopewouldbetterpreservereduplicant-baseidentity(wá-walis);however,syncopeappliestransparentlyany-waywithoutimpedancefrombase-reduplicantidentityconstraints.
Somecasesofreduplicativeopacitycanbeattributedtolayeringorstratalaspectsofthephonology-morphologyinteraction.
WithregardtotheJavanesecaseabove,InkelasandZoll(2005:Chapter5)arguethatdemonstrativesufxationoccurspriortoreduplication,triggeringh-deletionandconsonantvoicingandpreventingclosed-syllablelaxingfromoccurring;thesufxedstem(forexample,bRìa-eorgilig-e)istheninputtoreduplication,whichcopiestherootasis,pre-servingtheeffectsofthestem-levelphonologicalalternations.
WhatportionofreduplicativeopacitywillyieldtolayeringaccountsasproposedbyInkelasandZoll2005andKiparsky2010,andwhatportionrequiresidentityprinciples,isstillanopenquestion.
Itmaybeimportant,indecidingthisquestion,tofactorapartmorphologicallydrivenreduplication,suchastheJavaneseexample,fromphonologically-drivensegmentduplication.
Thelatterclearlyrequiresphonologicalidentityprinciples(copyingorcorrespondence,asappropriatetothetheoreticalframeworkinuse).
SeeHendricks(1999,2001),Yu(2005),Riggle(2006),Inkelas(2008a),andPulleyblank(2009)fordiscussionrelevanttothedistinctionbetweenmorphologicalreduplicationandphonologicalcopying.
4.
2.
2FixedSegmentisminReduplicationItisoftenthecasethatoneofthetwocopiesinmorphologicalreduplicationcontainssomexedmaterialwhicheitherco-occurswithorsupplantsmaterialthatwouldotherwisebeexpectedtocopy.
AnexampleoftheformeroccursinKhasi,whereiterativeverbreduplica-tionconnectsthetwocopiesoftheverbwiththelinkeri,forexample,iaid-i-iaid'togoonwalking',leh-i-leh'keeprepeating',kren-i-kren'keeptalking'(Abbi1991:130,citedinInkelasandZoll2005:36).
AnexampleofthelatteroccursinEnglish,whereanironic/derisivetotalreduplicationconstructionassigns"shm"tobetheonsetofthesecondcopy,replacinganexistingonset,ifany:fancy-shmancy,handsome-schmandsome,OT-shmOT,andsoon;seeforexample,Aldereteetal.
(1999).
Thisphenomenonhasbeentermed"MelodicOverwriting"(McCarthyandPrince1986;Yip1992;Aldereteetal.
1999).
OnefunctionalmotivationthathasbeenofferedbyYip(1997)forMelodicOver-writingisthatitmakesthetwocopiesdifferent.
Supportforthisinterpretation9781405157681_4_003.
indd959781405157681_4_003.
indd9506/05/20115:08PM06/05/20115:08PM96SharonInkelasisfoundin(a)thepredominanceinofMelodicOverwritingintotal,asopposedtopartial,reduplicationand(b)thefactthatconstructionsinvolvingMelodicOverwritingsometimesblockwhenthetwocopieswouldbeidenticalphono-logically,or,perhapsmorecommonly,exhibitdissimilatoryallomorphywhichguaranteesthatthecopywiththexedmaterialisdifferentfromtheintactcopy.
InTurkish,asdescribedbyLewis(1967:237–238),aconstructionmeaning"andsoon,andsuchlike"doublesawordandimposestheonsetmonthesecondcopy,replacinganexistingonsetifthereisone(e.
g.
dergi'journals',dergimergi'journalsorperiodicalsormagazines').
AccordingtoLewis,thisconstructioncannotbeusedifthewordbeginswithmalready(e.
g.
müfetti%ler'inspectors'),andaperi-phrasticconstructionwithfalanorlnisusedinitsplace(müfetti%lerfalan'inspec-torsandallthatlot').
InAbkhaz,suppletiveallomorphycomestotherescueinthecomparablesituation.
Bruening(1997),citingVaux(1996),describesanAbkhazecho-wordconstructionwhichreplacestheonsetofthesecondcopywith/m/(gaá-k''fool'→gaák'-maák');however,ifthewordalreadybeginswith/m/,/:'/isusedinstead(gaá-k''secret'→maá-k'-:'aá-k').
ThiskindofrequireddissimilationseeninMelodicOverwritingisinsomewaysreminiscentoftheanti-homophonymorphologicaleffectsdescribedinSection2.
4,whichrequireinputsandoutputs,ormembersofthesameparadigm,todiffer.
Yip(1997,1998)likensdissimilatoryMelodicOverwritingtothekindofconventionalpoeticrhymeinwhichidentityisrequiredinoneprosodiclocation(e.
g.
thesyllablerhyme)butnon-identityisrequiredelsewhere(e.
g.
theonsetofthatsamesyllable);thusrhyme-timeisagoodrhymebutrhyme-rhymeisnot.
Aldereteetal.
(1999)havearguedthatsomecasesofxedsegmentismare,ratherthaninstancesofMelodicOverwriting,insteadtheresultofreductiondrivenbyemergentunmarkedness,aphenomenonobservedbySteriade1988tocharacterizepartialreduplication.
InNupegerundives(Downing2004:90,citingAkinlabi1997;Smith1969),aninitialCVreduplicanthasaxed[+high]vowelandmidtone,regardlessofwhatisfoundinthebase:kpi-kpà'drizzling',ji-jákpe'stooping',bi-bé'coming'.
Insofaras[+high]andmidaretheunmarkedvaluesforvowelsandtoneinNupe,asarguedbyAkinlabi,thexedvaluesinthereduplicantcanbederived,ratherthanstipulated.
Reductioninpartialreduplica-tionisconsistentwiththehypothesisthatpartialreduplicationtypicallyderiveshistoricallyfromerosionoftotalreduplication(e.
g.
Niepokuj1997).
Totalredupli-cation,however,virtuallyneverdisplaysphonologicalreductioninonecopy,asobservedinInkelas(2008a:379–380).
4.
2.
3MorphologicalCharacterofReduplicantItistempting,basedonform,tocharacterizepartialreduplicationasafxationandtotalreduplicationascom-pounding.
However,thereislittlemorphologicalevidenceforthisdistinction.
Indeed,somephonologistshaverecruitedtheafxation/compoundingdistinctiontoaccountforphonologicalsizedifferenceswithinpartialreduplication,termingreduplicantswhicharesyllable-sizedorsmaller"afxes"andthosewhicharefoot-sized"roots"(GeneralizedTemplateTheory;forexample,McCarthyandPrince1994,Urbanczyk1996).
Thisdistinctionisgenerallymotivatednotby9781405157681_4_003.
indd969781405157681_4_003.
indd9606/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology97morphologicalcriteriabutbythedesiretoavoidmorpheme-specicreferencetoprosodictemplates.
Takingadifferentview,InkelasandZoll(2005)observethatreduplicationconstructionsdosometimesimposemorphologicalrestrictionsthatareindependentofprosodicones,arguingagainstconatingthetwotypesofrestriction;theypoint(inChapter2)tothedistinctionbetweenconstructionswhichspecicallydoubleafxes,regardlessofsize,vs.
thosethatdoublerootsorstemsasevidencethatreduplicationtargetsmorphologicallydenedconstitu-entsandthenimposesphonologicalshaperequirementsontheoutputofdoubling.
5WhenPhonologyisMorphology:RealizationalMorphologyandMorphologicallyConditionedPhonologyMorphologicallyconditionedphonologyoverlapssignicantlywithwhathasbeencalled"realizational"or"process"morphology,anobservationmadeby,amongothers,FordandSingh(1983,1985),Poser(1984),Dressler(1985),Singh(1996),S.
Anderson(1992),Bochner(1992),Orgun(1996),Inkelas(2008b).
Toillustratethesephenomenaandtheiroverlap,considertwocasesofnalconsonantdeletion.
Therstisawell-knownprocessofsubtractivemorphologyinTohonoO'odham,discussedbyS.
Anderson(1992),citingZepeda1983,andYu(2000),citingZepeda1984.
InTohonoO'odham,perfectiveverbsarederivedfromimperfectivesthroughthedeletionofanalsegment(síkon'hoeobject'→síko(-perf);híwa'rubagainstobject'→híw(-perf)(Yu2000:129–130).
Thiststhestandarddescriptionofrealizationalmorphologybecausethereisnoothermorphologicalexponentoftheperfective.
Nowconsiderthediminutivesufx/-CIk/inTurkish,discussedinSection2.
1,whichtriggersanoptionalprocessofstem-nalvelardeletion(/bebek-CIk/→bebecik'baby-dim',/kpek-CIk/→kpecik'dog-dim')(Lewis1967:57)thatappliesbeforenoothersimilarsufx.
Thiswouldstandardlybedescribedasamor-phologicallyconditionedphonologicalrule,becausethemorphologicalcategoryofdiminutiveismarkedovertlybythesufx.
Theoperativeintuitionisthatthesufx-CIkmarksdiminutivemorphology,whiletheconsonantdeletionisjustincidental.
Thispracticaldistinctionbetweenphonologyasprimaryexponentandphonologyassecondaryconcomitantdoesnotalwaysholdup.
Sometimesitisdifcult,evenunproductive,givenseveralexponentsofagivenmorphologicalconstruction,toidentifywhichistheprimary(morphological)exponentandwhicharethephonologicalaccompaniments.
InHausa,forexample,tonemelodyreplacementcanserveasthesolemarkofamorphologicalconstruction(8a),andsocanovertafxation(8b).
Whenbothco-occur(8c),istonemelodyreplace-mentconsideredrealizationalmorphology,suchthatthewordsin(8c)exhibittwomorphologicalexponentsofpluralization,oristonemelodyreplacementsubjugatedin(8c)tomorphologicallyconditionedphonologyPagenumbersrefertoNewman2000:9781405157681_4_003.
indd979781405157681_4_003.
indd9706/05/20115:08PM06/05/20115:08PM98SharonInkelas(8)a.
Noafxation;tonereplacement(imperativeformation)ká"mà"→kà"má"'catch(!
)'(267)bíncìké"→bìncìké"'investigate(!
)'(267)nánné"mó"→nànnè"mó"'seekrepeatedly(!
)'(Overtsufxation,notonereplacement(various)dáfà"→dáfà"-wá'cook-vbl.
n'-LH(699)gàjé"ré"→gàjé"r-ìyá"'short-fem'-LH(212)hù"lá"→hù"l-r'hat-def'-L(144)c.
Overtafxationandtonereplacement(variouspluralclasses)má"làm→mà"làm-ái'teacher-pl.
'-LH(434)hù"lá"→hú"l-únà"'cap-pl.
'-HL(444)tàmbáyà"→támbáy-ó"yí"'question-pl.
'-H(432)Onepossiblewaytoavoidanalyticalambiguityinthecaseof(8c)wouldbetoreduceallphonologicaleffectsotherthanovertsegmentalafxationtomor-phologicallyconditionedphonology,reanalyzingapparentcasesofrealizationalmorphologyaszeroderivationaccompaniedbymorphologicallyconditionedphonology.
Alternatively,onecouldtrytoreduceallmorphologically-specicphonologicaleffectstorealizationalmorphology,analyzingthedatain(8c)asinstancesof"extendedexponence,"themultiplemarkingofamorphologicalcategory(forexample,Matthews1972;Stump1991).
Multipleexponenceofovertmorphologyisacommonenoughphenomenon;inHausa,forexample,theformationofclass13nounpluralsinvolvessufxation(of-e),(LH)tonereplace-ment,andreduplication,forexample,kwánà"'corner,curve'→kwàné-kwàné(pl)(Newman2000:458).
Harris(2002,2008b)hasarguedthatcircumxesresultdia-chronicallyfromanearlierstageofmultipleafxation,ormultipleexponence.
5.
1TheoreticalApproachestoRealizationalMorphologyandMorphologicallyConditionedPhonologyTheliteratureonmorphologicallyconditionedphonology,primarilyrepresentedbyitem-basedapproaches,hashadlittletosayaboutrealizationalmorphology,despitetheobviousformalsimilaritiesbetweenthetwophenomena.
Onereasonforthisisthatmuchofthemostinuentialliteratureonmorphologicallycondi-tionedphonology,goingbacktoKiparsky's(1982b)theoryofLexicalMorphologyandPhonology(LMP;seealsoKiparsky1984;Mohanan1986),focusesonphono-logicalpatternscommontothemorphologyofacertainstratum.
BothLMPanditssuccessor,StratalOptimalityTheory(Kiparsky2000,2008a,toappear),assumeagrammaticalarchitectureinwhichthemorphologicalconstructionsofeachlan-guageclusterintoasmall,possiblyuniversallyxednumberofsets("levels,""strata"),eachinternallyuniforminitsphonologicalpatterning,whicharetotallyordered.
Instrataltheoriesliketheseitisnecessarytoknowonlythestratumtowhichamorphologicalconstructionbelongstopredictwhichphonologicalpat-ternsitwillconformto;whichstratumaconstructionbelongstoispredictable9781405157681_4_003.
indd989781405157681_4_003.
indd9806/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology99fromwhetheritsplaceinthemorphology,thatis,whetheritisan"early"or"inner"afxasopposedtoa"late"or"outer"one.
Becausestrataltheoriesfocusoncommonalities,theyarenotsuitedtothedescriptionofphonologywhichisuniquetoaparticularmorphologicalcategory;verynarrowlyconditionedphonologicaleffectshavetobesetasideandtreatedasexceptionswithinastratum,ratherthanconstitutingtheirownindividualstratum.
Forexample,inEnglishboth-ibleand-ivetriggerspirantizationonaprecedingconsonant,anunambiguouslystratum1effect(divide,divis-ible,divis-ive).
However,-ibletriggersvoicingwhile-ivedoesnot.
Thisdistinctioncannotbecapturedbystratalassignmentbutmustbetiedtoindividualsufxesusingexceptionfeaturesorothermechanismsbesidesstrata.
Because,byitsnature,realizationalmorphologyisalsonarrowlytiedtoindi-vidualmorphologicalcontexts,stratalorderingtheoriesdonotlendthemselvestothedescriptionofrealizationalmorphologyanymorethantheyaresuitedtocapturingidiosyncraticmorphophonology.
Amiddlegroundwhichcancapturethosemorphophonologicalgeneralizationssoughtbystrataltheoriesbutalsodescribehighlymorphologically-specicpho-nologicalpatternsisrepresentedbycophonologicalmodels.
These,asdiscussedinSection2,associateeachmorphologicalconstruction(afxation,compounding,zero-derivation)withitsownphonologicalmapping.
Thecophonologicalapproacheliminatesthe"toomanyanalyses"problembyusingexactlythesamemechanismtohandlerealizationalmorphologyandmorphologicallyconditionedphonology.
Aphonologicalalternationspecictoaparticularafxisincludedinthecopho-nologythatisuniquetothatafx.
Realizationalmorphologyisaccomplishedbythecophonologyofwhatmightotherwisebedescribedasphonologicallynullmorphologicalconstructions.
Whetheraconstructionis"null"ornot,thatis,whetherornotitisassociatedwithanovertafx,isincophonologytheoryalmostincidental.
InthiswaycophonologytheoryresemblestheapproachofBochner(1992),inwhichphonologicalpatternsarepartandparcelofthedescriptionoftherulesrelatingwordsinaparadigm.
Cophonologytheoryisnotlimitedtoenumeratingidiosyncracies;asAnttila(1997a,2002a)hasdemonstrated,itsinher-itancearchitecturealsogivesittheabilitytopositmeta-constructionslike"word,""stem,"or"stratum,"withassociatedcophonologicalrestrictionsinheritedbythememberconstructions,tocapturegeneralizationsholdingacrossallconstituentsofacertaintype.
6WhenPhonologyandMorphologyDiverge:NonparallelismBetweenPhonologicalandMorphologicalStructureThedomainsofword-internalphonologicalpatternsaregenerallycoextensivewiththemorphologicalsub-constituentsofaword;forthisreason,phonologyprovidesstrongevidenceaboutthemorphologicalstructureofaword.
However,9781405157681_4_003.
indd999781405157681_4_003.
indd9906/05/20115:08PM06/05/20115:08PM100SharonInkelastherecanbemismatches,thatis,situationsinwhichphonologicaldomainsarenotmatchedwithmorphologicalsub-constituents.
Insomecasesthephonolo-gicaldomain–prosodicroot,orstem,orword–isasub-portionofaword(seee.
g.
Booij1984;Sproat1986;Inkelas1990;BooijandLieber1993,amongmanyothers).
Threesituationsstandoutinthisregard:compounding,thedistinctionbetweencoheringandnon-coheringafxes,andreduplicationofaninternalprosodicstem.
Theliteratureonthephonologyofcompoundingconstructionshasoftendrawnattentiontoadistinctionbetweencompoundsthatbehavephonologicallylikeonewordandthosethatbehavephonologicallyliketwowords.
Inthe1980sthisdifferencewasattributedtoprosodicstructurewhichislooselyrelatedtobutexistsindependentlyofmorphologicalandsyntacticstructure.
NesporandVogel(1986)proposedthatwhileGreekcompoundsformasingleprosodicwordandtherebyreceiveonestress,forexample,kúkla'doll'+spíti'house'→[kuklóspito]'doll'shouse'(p.
112),themembersofHungariancompoundsformseparateprosodicwordsandretaintheirownlexicalstresses:[knyv]_[tár]_'bookcollec-tion'(p.
123).
InMalayalam,simplesub(ordinate)compounds,withhead-modiersemantics,formasingledomainforaccentuation,whereasthemembersofsimpleco(ordinate)compounds,withcoordinationsemantics,formseparatedomainsforaccentuation.
Sproat(1986)andInkelas(1990)proposedthatthisdifferenceinbehaviorcouldbeattributedtodifferentprosodicstructure,thoughMohanan(1995)latercounteredthisargumentwithevidencefromcomplexcompoundswiththreeormoremembers.
InadetailedstudyofIndonesian,Cohn(1989)documentsastressdifferencebetweentwoconstructionsthatconcatenatestems.
Head-modiercompoundsimposestressreductionononemember(polúsi'pollution'+udára'air'=polùsiudára'airpollution',p.
188),suggestingthattheyarecompetingforprominencewithinasinglephonologicalword,whiletotalre-duplicationconstructionsmaintaintwoequalstresses(minúman'drink',minúman-minúman'drinks',p.
184).
Cohnattributesthelatterpatterntothefactthattotalreduplicationconsistsoftwoprosodicwords.
ItandMester(1996a)pointtoasimilardistinctioninJapanese,inwhichstem-stemcompoundsformoneprosodicdomain,word-wordcompoundsformtwodomains,andstem-wordcompounds"type-shift,"bymeansofaprincipleofProsodicHomogeneity(p.
38),topatternlikeword-wordcompounds(seealsoHan1994onKorean).
Perhapsevenmoreinterestingthanprosodicdifferencesacrosstypesofcom-poundsarecomparabledifferencesinafxedwords.
Booij(1984)wasoneofthersttohighlightthedistinctionbetween"cohering"and"non-cohering"afxesandtomodelthedistinctionusingprosodicstructure:coheringafxesformasingleprosodicwordwiththebaseofafxation,whilenon-coheringafxesformaseparateprosodicdomain.
InDutch,forexample,sufxesareeithercohering,meaningtheysyllabifywithandjoinintoaprosodicdomainwiththestemtheycombinewith,ornon-cohering,meaningtheycreateaseparateprosodicdomain.
NonnativesufxesinDutchareallofthenon-coheringtype.
ThedifferencebetweenthetwotypesofsufxisillustratedwiththisminimalpairofsufxesbothofwhichareequivalenttoEnglish"-ish":rood-achtig[ro"t.
αx.
tRx]androd-ig9781405157681_4_003.
indd1009781405157681_4_003.
indd10006/05/20115:08PM06/05/20115:08PMTheInteractionBetweenMorphologyandPhonology101[ro".
dRx]'reddish'(Booij1984:152).
AsBooij(1984,2001,2002:172)observes,non-coheringDutchsufxeslike-achtigallowdeletionunderidentity:iftwowordsendingin-achtigarecoordinated,thesufxisomissiblefromtherstconjunct:storm-achtigenregen-achtig'stormyandrainy'canalsoberealizedasstormenregen-achtig(1984:151).
Deletionunderidentityisalsofoundincompoundingconstructions:wespen-stekenenbije-steken~wespenenbije-steken'waspandbeestings'(1984:146).
Non-coheringafxesandmembersofcompoundsformindi-vidualprosodicwords,explainingtheirparallelbehavior.
Bycontrast,coheringsufxeslike-igcannotbeomittedunderidentity:blau-igenrod-ig'blueishandreddish',but*blauenrod-ig(1984:149).
Ondeletionundermorphologicalidentityinotherlanguages,seeforexample,VigárioandFrota(2002),Orgun(1996).
Evidencefortheaccessibilityto"later"processesofword-internalprosodicstemsisfoundinreduplication.
Inanumberofcases,alatemorphologicalprocessofreduplicationtargetstheroot,eveniftheroothasalreadyundergonesignicantafxation.
Aronoff(1988)referstotheseas"headoperations,"andBooijandLieber(1993)proposethattheyinvolvereferencetoaprosodicstem,whichcorrespondscloselyifnotexactlytothemorphologicalroot.
InkelasandZoll(2005)citetheexampleofChumash,whichhaswhatApplegate(1972:383–384)characterizesasaverylateprocessofreduplication,conferringthemeaningofarepetitive,distributive,intensive,orcontinuativeforce.
Chumashreduplicationtargetsasub-constituentofthewordwhichInkelasandZolltermtheprosodicstem.
Theprosodicstemalwayscontainstheroot,alongwithanyprecedingprexesofthetypeApplegate(1972)identiesasreduplicating,andwhichInkelasandZoll(2005)analyze,inBooij'sterms,ascohering.
Forexample,theroot-adjacentprexink-sili-{pil-wayan}'Iwanttoswing',inwhichcurlybracketsdemarcatethepros-odicstemandtherootisunderlined,iscoheringandparticipatesinreduplication:ksili{piw-piwayan}(Applegate1972:387).
Bycontrast,theprexesins-am-ti-{lok'in}'theycutitoff'arenon-coheringanddonotreduplicate:samti{lok-lok'in}(Applegate1972:387).
Evidencethatwhatreduplicatesisaprosodicstem,occupiedbytherootandjoinedbycoheringprexes,isthattheprosodicstemissubjecttoatypicalstem-shapeconstraint;itmustbeconsonant-initial.
OnsetconsonantsarenotrequiredofChumashrootsorprexes,manyofwhicharevowel-initial.
Butprosodicstemsmustbeconsonant-initial.
Asaresult,evenanotherwisenon-coheringprexwillcontributeitsnalconsonanttoafollowingprosodicstem,asshownbyreduplicatedformssuchass-iy-ak{t-aqu-smon}→siyak{taq-taqusmon}'theycometogatherit'(Applegate1972:388).
Parallelphenomena,documentedinInkelasandZoll(2005),occurinTagalog(seealsoBooijandLieber1993)andEasternKadazan(Hurlbut1988).
AnalternativeanalysisoftheChumashandTagalogphenomenaisofferedwithinBase-ReduplicantCorrespondenceTheory(BRCT)byMcCarthyandPrince(1995),whoproposethatthereduplicantisnotinxingbutisinsteadprexeddirectlytothematerialthatiscopied.
Ontheiraccount,thecopyingofthenalconsonantofaprexprecedingthereduplicantistheresultofmorphologicalfusionbetweentheprexconsonantandtheVCreduplicantand"back-copying"oftheresulttothebaseofreduplication:s-i-RED-expe:→s-i-exRED-expe:BASE(withbackcopyingofthei-naltothebase).
9781405157681_4_003.
indd1019781405157681_4_003.
indd10106/05/20115:08PM06/05/20115:08PM102SharonInkelasInkelasandZoll(2005)argueagainstthisaccountofChumash,inparticular,onlanguage-internalmorphologicalgrounds;McCarthyandPrince(1995)have,however,identiedotherapparentcasesofbackcopyinginotherlanguages,andbackcopyingingeneralremainsaviableanalysiswithinBRCT.
Returningtomismatchesbetweenmorphologicalconstituentstructureandprosodicstructure,thereisalsostrongevidencethatword-sizedprosodicdomainscanincludematerialoutsideofthemorphologicalorlexicalword,cliticsbeingthemostobviousexample.
Ithasbeenwidelyarguedthatcliticsarephonologic-allydefectivesyntacticterminalelements,havingtojoinwithanother(non-clitic)syntacticterminalelementtoformasingleprosodicword(e.
g.
Inkelas1990;Halpern1992;Booij1996).
Aquestionofconsiderablecurrentinterestiswhetherprosodicwordstructurecanberecursive;seePeperkamp(1996),ItandMester(2003a),andKabakandReviathidou(2009),amongothers.
7SummaryThephonology-morphologyinteractionshedslightonword-internalstructureandontheabilityforrelativelyunnaturalphonologicalalternationstobeproduc-tive,atleastwithinagivenmorphologicalniche.
Bothrealizationalmorphologyandmorphologicallyconditionedphonologyoperateinthesamedomainsandmanipulatethesamestructuralelements.
Themanyrelatedphenomenacon-stitutingthephonology-morphologyinterfacearecentraltoword-formationinvirtuallyalllanguages,andmustthereforebetakenseriouslybymorphologistsandphonologists,especiallythoseseekingtoreducesynchronicmorphologicalpatternstosyntax,orsynchronicphonologicalpatternstouniversalphoneticmotivations.
NOTES1Thisistheapproximategeneralization,asstatedbyKenstowicz;theactualpictureismoredetailed,inwaysnotmaterialtothepointmadehere.
Seeforexample,ButtandBenjamin2008.
2Vaux(p.
252)analyzesthedenitesufxasunderlyingly/-n/andattributestheschwaallomorphtorulesofepenthesisandconsonantdeletion;however,asthen~Ralterna-tionisspecictothedenite,andmostresearcherswouldprobablyclassifythisassuppletiveallomorphy.
9781405157681_4_003.
indd1029781405157681_4_003.
indd10206/05/20115:08PM06/05/20115:08PM