Rule Based Models of Phonology: Interfaces Modularity and ... · Rule-Based Models of Phonology:...
32
1 Rule-Based Models of Phonology: Interfaces Modularity and Derivation at the Syntax-Phonology Interface Heather Newell, UQAM Abstract This chapter investigates some of the nuanced distinctions between the predictions of a cyclic, modular, Rule-Based Phonological account of the Morphosyntax-Phonology Interface, and those of a cyclic, modular, Constraint-Based account (Stratal Optimality Theory). The focus here is on the theoretical distinctions between these two proposed systems that pertain to the construction of a general theory of linguistic computation. The discussion is centred around questions of modularity and derivation/cyclicity. Specific topics include (i) investigating the necessity of postulating both derivational and representational accounts of phonological domains, (ii) detailing how the two theories differ regarding how cycles are triggered, and in which module (the syntax or the phonology) they are controlled, and (iii) questioning whether the effects of constraint re-ranking across strata can be captured by a theory of structural underspecification. Data from English, Portuguese, Malayalam, and Ojibwe are analysed, and it is argued that the patterns we see are best accounted for within a Rule-Based system. 1. INTRODUCTION This chapter focuses on two issues that are important to the construction of an explanatory theory of phonology; modularity and derivation. Modularity encapsulates the proposal that the vocabulary used by each linguistic sub-domain (phonology, syntax, semantics) is module- specific. It is therefore also concerned with how information is passed from one module to another. Here we concentrate on how morphosyntactic information is transferred from the syntax to the phonology. It is argued that syntactic features have no place in the phonological module, and that this is an important factor for distinguishing between theories of phonology, and of the morphosyntax-phonology interface. The discussion of modularity then leads us to the question of derivation, or more precisely, of cyclicity. If morphosyntactic information is illicit in the phonological module then the fact that phonological outputs display evidence of cyclic domains must be due either to properties specific to the phonological module, or to the existence of a derivational engine that sends information to the phonology in pieces. I support the latter view here. How this cycling of the computational system works, and how it affects the phonology is the second topic of the current chapter. These two issues are integral to constructing a theory of the morphosyntax-phonology interface. They are also two domains of inquiry where traditional, serial, Rule-Based Phonology (RBP) and parallel Optimality Theory (OT), or Constraint-Based Phonology (CBP), have diverged, sometimes significantly. Within OT, which was traditionally entirely parallelist (and still is in many, if not most sub-camps) the re-introduction of cyclicity to the derivational computation has closed the gap between the two frameworks. Serial OT frameworks such as Lexical Phonology and Morphology Optimality Theory (LPM-OT) (Kiparsky 2000) or Stratal Optimality Theory (SOT) (Kiparsky 2007; Bermúdez-Otero 2014) are fundamentally quite
Transcript of Rule Based Models of Phonology: Interfaces Modularity and ... · Rule-Based Models of Phonology:...
1
Rule-Based Models of Phonology: Interfaces
ModularityandDerivationattheSyntax-PhonologyInterfaceHeatherNewell,UQAM
Abstract
Thischapterinvestigatessomeofthenuanceddistinctionsbetweenthepredictionsofacyclic,modular, Rule-Based Phonological account of the Morphosyntax-Phonology Interface, andthoseof a cyclic,modular,Constraint-Basedaccount (StratalOptimalityTheory).The focushereisonthetheoreticaldistinctionsbetweenthesetwoproposedsystemsthatpertaintotheconstructionofageneraltheoryoflinguisticcomputation.Thediscussioniscentredaroundquestionsofmodularityandderivation/cyclicity.Specific topics include(i) investigatingthenecessity of postulating both derivational and representational accounts of phonologicaldomains,(ii)detailinghowthetwotheoriesdifferregardinghowcyclesaretriggered,andinwhichmodule(thesyntaxorthephonology)theyarecontrolled,and(iii)questioningwhetherthe effects of constraint re-ranking across strata can be captured by a theory of structuralunderspecification.DatafromEnglish,Portuguese,Malayalam,andOjibweareanalysed,anditisarguedthatthepatternsweseearebestaccountedforwithinaRule-Basedsystem.
1. INTRODUCTIONThischapterfocusesontwoissuesthatareimportanttotheconstructionofanexplanatorytheoryofphonology;modularityandderivation.Modularityencapsulatestheproposalthatthevocabulary used by each linguistic sub-domain (phonology, syntax, semantics) is module-specific. It is thereforealsoconcernedwithhow information ispassed fromonemodule toanother.Hereweconcentrateonhowmorphosyntactic information is transferred from thesyntaxtothephonology.Itisarguedthatsyntacticfeatureshavenoplaceinthephonologicalmodule,andthatthisisanimportantfactorfordistinguishingbetweentheoriesofphonology,andofthemorphosyntax-phonologyinterface.Thediscussionofmodularitythenleadsustothequestionofderivation, ormoreprecisely, of cyclicity. Ifmorphosyntactic information isillicitinthephonologicalmodulethenthefactthatphonologicaloutputsdisplayevidenceofcyclicdomainsmustbedueeithertopropertiesspecifictothephonologicalmodule,ortotheexistenceofaderivationalenginethatsendsinformationtothephonologyinpieces.Isupportthelatterviewhere.Howthiscyclingofthecomputationalsystemworks,andhowitaffectsthephonologyisthesecondtopicofthecurrentchapter. Thesetwoissuesareintegraltoconstructingatheoryofthemorphosyntax-phonologyinterface.Theyarealsotwodomainsofinquirywheretraditional,serial,Rule-BasedPhonology(RBP) and parallel Optimality Theory (OT), or Constraint-Based Phonology (CBP), havediverged,sometimessignificantly.WithinOT,whichwastraditionallyentirelyparallelist(andstill is inmany, if notmost sub-camps) the re-introduction of cyclicity to the derivationalcomputationhasclosedthegapbetweenthetwoframeworks.SerialOTframeworkssuchasLexicalPhonologyandMorphologyOptimalityTheory(LPM-OT)(Kiparsky2000)orStratalOptimality Theory (SOT) (Kiparsky 2007; Bermudez-Otero 2014) are fundamentally quite
2
similar to a cyclic, modular, rule-based framework, but there are still some significantdifferences.Manyoftheseparticulardifferences(ex.strata-internalruleordering)havebeendetailedinVaux(2008)andintheprevioustwochapters(Vaux,Myler(thisvolume),Purnell(this volume)), and will therefore not be the focus of the current discussion. The recentexplosion of modifications to CBP theories (ex. Stochastic OT, Boersma 1997; HarmonicSerialism,McCarthy2000;CandidateChainTheory,McCarthy2007;inadditiontoLPM-OTandSOT)makesageneraloverviewofthedistinctionsbetweenRule-BasedandConstraint-Basedtheories at the interface beyond the scope of this chapter. I therefore focus here on theargumentthatanunderstandingofthelinkbetweensyntacticandphonologicalderivationsisa crucial precursor to developing the correct theory of purely phonological alternations. IemphasizesomepertinentdistinctionsbetweenRBPandSOTastheyarecurrentlypresented,astheyhavesimilarpremisesregardingthearchitectureofthegrammarandthereforeofferanopportune environment in which to discuss detailed nuances of theory-construction andtesting.Iofferargumentsthataserial,cyclicderivation,whereaffixesarenotassignedtostrata-particularconstraintrankingsbestcapturestheinner-workingsofthegenerativephonologicalsystem. In§2wewillbrieflydiscusssomehistoricalperspectivesandtheoreticalassumptionspertinenttotherestofthischapter.In§3criticaldistinguishingissuesspecifictomodularityandderivationgermanetotheOT-RBTdebateareexpounded.Inthesetwosectionsthereaderwillfindextensivereferencespointingthemtomorein-depthdiscussionsofthefoundationalissues.Somerelevantcurrentcontributionsandresearchwillbepresentedin§4.ThismainsectionwilldiscussaspectsoftwospecificissuesrelevanttotheSOT-RBTdebate;(i)whetherthedelimitationofcyclesisaccomplishedviarepresentationalorderivationalmeans,anditslinktoOTalignmentconstraintsandtheProsodicHierarchy,and(ii)therelationship,orlackthereof, between themorphological status of affixes and phonological cycles. §5 will thenconcludewithaneyetofuturepotentialresearchdirections,discussingtheimplicationsofthedatain§4fortheoriesofphonologybelowtheProsodicWord,wheretheinterfaceisnotatissue.2. HISTORICALPERSPECTIVESTosituateanydiscussionofthemorphosyntax-phonologyinterfaceitisnecessarytodiscussone’s foundational assumptions regarding (i) themorphosyntax (and themorphology andsyntaxseparatelyiftheyaretakentobeseparatemodules),(ii)thephonology,and(iii)theirmannerofinteracting.Thesefoundationalsuppositionscannotbeseparatedfromthehistoryoflinguistictheorizingthattheyemergefrom,leadingtotheirinclusioninthepresentsection.Thepresentationbelow isbrief,ofnecessity, andshouldnotbe takenasadefinitive listofresourcesorviewpointspertinenttothedebate,butratherasanoverviewofthetheoreticalunderpinningsassumedinthefollowingsections. Proposalsinthemorphosyntacticliteraturehavevacillatedoverwhetherthesyntaxandmorphology are one single or two separate generative engines. Originally compounded(ChomskyandHalle1968),theemergenceofinvestigationsintomorphologicalirregularityvsmorpho-syntacticregularityinspiredtheirdivision,instigatingtheemergenceoftheoriessuchasLexicalPhonologyandMorphology(LPM)(Kiparsky1982;Mohanan1986).Thereitwasproposed that a pre-syntactic word-formation module existed wherein affixes andphonological rules were organized into Levels, or Strata (ex. irregular inflection, regular
3
inflection, various types of derivation, compounding). This was taken to explain the linkbetween certain kinds of phonological phenomena and particular constraints on affixpositioning,andisthetheoreticalprecursortoStratalOptimalityTheory,tobediscussedinmoredetailbelow.IssueswithLPM,suchasitsinabilitytocorrectlypredictaffixalpatterning(Fabb1988),thenecessaryintroductionofaloopingmechanismbetweenlevelswhichledtoalackofpredictivepower(Kaisse&Shaw1985),anditsoverlookingoftheparallelsbetweenword-internalandsyntactic configurations (Marantz1997), amongotherarguments, led totheories that reunified the twomodules. This chapter assumes a particular version of thisreunification,namelya realizational theoryofmorpho-syntax (a laDistributedMorphology(Halle&Marantz1993,1994)orNano-Syntax(Starke2010))wherethebuildingblocksofthesyntacticgenerativeenginearemorphemes,andthesemorphemesconsistofabstractfeaturebundleswhicharegivenphonologicalformpost-syntactically. On the strictly syntactic side, it is a historically robust stance that the derivationproceeds insteps,orcycles(SeeBoeckx&Grohmann2007forahistoryof thecycle in thesyntacticliterature.).Thecurrenttheoryofcycles,namelyPhases,isassumedhere.DevelopedinitiallybyChomsky(1999,2001a)(seealsoUriagereka1999).Phasesaresub-structuresthataresenttoSpell-OutatbothPhonologicalForm(PF)andLogicalForm(LF).Phaseshavebeenundergoingtheoreticalmodificationssincetheirinception,andrecentlyproposedvariationsandmodificationstouchontheirsizeandmutability(Embick2010,2014;Marvin2002;Adger2006; Newell 2008; Bobaljik &Wurmbrand 2014; Svenonius 2004; Epstein & Seely 2002;Boskovic2014,amongothers).Understandingtheexactsizeandpropertiesofsyntacticcyclesiscrucialtothedevelopmentofanytheoryofthemorphosyntax-phonologyinterface,butwillnotbeinvestigatedhere.Itisassumedhereinthatphasesmaybelarge(EP(traditionallyv*P),CP,DP,andperhapsPP),orsmall(triggeredbycategorydefiningderivationalheadsthatprojectvP,nPandaP).ThemainfocusofthischapterbearsonthequestionofthemannerinwhichthephonologicalmoduleinterfaceswiththemorphologyandsyntaxatthepointofSpell-Out,andwhatimplicationsthishasforcurrentRBPorCBPproposals. The phonological foundation of the current chapter is one where, in the words ofMcCarthy (1988: 84), “…if the representations are right, then the rules will follow.” It isproposedherethatthephonologicaloutputisorganizedbothsub-andsupra-segmentallyintoa version of feature geometric and syllabic (or CV) representations, respectively. The exactspecificsoftheorganizationofthesephonologicalsubcomponentsaresubjecttodebate.FordiscussionsoffeaturegeometryseeClements(1985),Sagey(1986),McCarthy(1988),Avery&Rice (1989) and Uffmann (2011, and references therein). For discussions of syllabicorganization,orthelackthereof,seeBosch(2011andreferencestherein),andScheer(2004,andreferencestherein).Animportantpointofdebatethatwillbediscussedhereinconcernsthe relative roles of representation (phonological structure) and derivation (proceduralcomputation)toatheoryofphonology.FollowingNewell&Scheer(2007)andScheer(2010)the existence of prosodic organization above the foot level is questioned (the ProsodicHierarchyofSelkirk1984;NesporandVogel1986),butwewilloccasionallyresorttomakinguseofitbelowforeaseofexposition. Thischapterthereforeassumesacompletelymodularlinguisticcomputationalsystem,wherethestructuregeneratedbythemorphosyntaxisinterpretedcyclicallyattheinterfaces,andwherecyclesaregovernedbypurelysyntacticmeans.AttheinterfacewithPhonologicalForm(PF)morphemesarereplacedwithvocabularyitems,andtheirformsareunderlyinglystructured at the melodic level. Underspecified segmental structure and supra-segmental
4
structure are both projected from these phonological underlying representations uponinterpretationatPF,andtheseunderlyingandsurfacerepresentationsmaybesubjecttotheapplicationofphonologicalrules.Inviolablephonologicalconstraints,inthepre-OTsense(ex.theObligatoryContourPrinciple,Leben1973)mayalsobeactive,andarenotconsideredabasicpartofRule-Basedphonologicalmodels.
3. CRITICALISSUESANDTOPICSThissectionfocusesonthemainissuestobediscussedinthischapter,namelymodularityandderivation.Somegeneralaspectsoftheseissuesarediscussed,andin§4wedelveintosomespecific subjects that will tease apart certain differing predictions of SOT and the RBPsupportedhere.
Modularity speaks to what kind of information is processed in each of themorpho-syntactic, semantic, and phonological computational systems. Within a strictlymodular framework, themorpho-syntactic and phonological systems are non-overlapping.Morpho-syntacticinformationmustundergotranslationintophonologicalprimesbeforethephonologicalsystemcanactuponit.Inotherwords,thesyntaxmanipulatesstrictlysyntacticfeaturesandrepresentations,whilethephonologymanipulatesstrictlyphonologicalfeaturesandrepresentations(Zwicky&Pullum1986).Thisentailsthatsyntacticinformationsuchas3rd personor XP is invisible to, or rather, non-existent in the phonological domain,while aphonological element suchasnasaloronset canplaynopart in the syntacticdomain.Thisentailmentholdsuniversally. Inno languagedoesaperson feature trigger,say, lenition,nordoesaphonologicalfeaturetriggermovement.iThisnotionofmodularitybearsnotonlyonthedistinctionbetweenphonologicalandmorpho-syntacticprimes,or features,butalsoonthequestionofthetranslationofcyclicdomainedgesintophonologicalterms.Itisclearthatthesyntacticderivationeffectsthesizeofthedomainsforphonologicalruleapplication,buthowthisinfluenceisexertedisopentodebate.Itwasproposedinthepost-SPEerathatthe#and+boundarysymbolscharacteristicofearlygenerativederivationswerediacritics(notnativephonologicalobjects)andthereforehadnoplaceinatheoryofphonology.Thisbroughtaboutthe rise of indirect-reference theories, the most popular and widespread of which is theProsodicHierarchy(Selkirk1984,2011;Nespor&Vogel1986).RecentlyScheer(2008)hasquestionedtheexistenceoftheProsodicHierarchyasaphonologicalobject,andreopenedthedebateovertheexistenceofdiacriticsinthephonology,proposingthatstructuressuchastheProsodicWord(PWd)havesuchastatus.Thisbringsustoanoverlappingpointpertinenttothe spheres of modularity and derivation. Cyclicity in the phonology can potentially bedeterminedintwodifferentways;derivationallyorrepresentationally.AdetaileddiscussionofthisdistinctioncanbefoundinBermudez-Otero&Luis(2009),Newell&Scheer(2007),andScheer(2010).In§4.1wewillexaminehowsyntactically-drivencyclicinterpretationmakesindependentlyverifiablepredictions,andwilldiscusswhethertherepresentationalcyclicityintroduced in the (morpho)phonology via the Prosodic Hierarchy is therefore redundant.Argumentsforandagainstthenecessityofbothrepresentationalandderivationalexplanationsfor the cycle are assessed in §4.2. It is argued that a purely derivational phonology is thetheoreticallycleaneroption.This,inturn,callsintoquestionOTconstraintslikeALIGNthatmakereference to theProsodicHierarchy. It isofnote thatmanycurrentphonologicalproposalsinclude rules or constraints that make simultaneous reference to both phonological andmorpho-syntacticinformation.Thismixed-moduleapproachiscommonwithinOTandinRBP,
5
but isnotspecific to,or fundamentallyrequiredbyeither. In§3.1we investigate the issuesraisedbythemixingofmodules,anditisdemonstratedthatthismixingisproblematicforanexplanatorytheoryofphonology.Acompletelymodulartheoryofphonologyisproposedtooffersuperiorexplanations.
Thediscussionofthestatusofderivationwithinthephonologicalmoduleisinextricablyentangledwiththequestionofmodularityinanadditionalway.Questionsofderivationrelatetowheninformationisprocessed,andwhattriggersinterpretation.Thefirstcentralquestionregardingphonologicalderivationistodeterminewhetherphonologicalstringsareprocessedincyclesorinonefellswoop.Asmentionedin§2,forthemostpart,historicallyandcurrently,cyclic derivation has been the standard assumption (with the notable, and ongoing, hiatustakenbytraditional,parallelOptimalityTheory(McCarthy&Prince1993,2008;andmanyofitsdescendants).RBPhastraditionallypresupposedcyclicity,alongwithserialapplicationofphonologicalrules,andthesewillbetakenasbasicoperationsofphonologicalcomputationinthischapter.Asmentionedabove,phonologybeingprocessedincyclesleadstothequestionofwhether these cycles are determined phonologically, morphologically (lexically), orsyntactically. This brings us to the second question discussed in this chapter relevant toderivation, central only once one accepts that cyclic interpretation is a property of thegrammar;whatdrivesphonologicalcyclicity?In§3.2wepreviewthefundamentalquestionspertinenttohowparticularphonologicalcyclesaretriggered.In§4.3wecarefullyexaminethecaseofcyclicphonologyinOjibwe(Newell&Piggott2014)andwilldemonstratethatcyclesmustbedeterminedinthenarrowsyntax,withoutreferencetomorpho-lexicalinformation. Acentralconcernofthischapteristhereforeametadiscussionofthearchitectureofthegrammarassumedgenerallybysomeproponentsofrule-basedandconstraint-basedsystems.Issueswiththeunderlyingassumptionsofthetheorieswillberaised,andmotivationsforthereturntoaderivational,rule-basedphonologywillbeoffered.Itiscrucialtonote,however,thatneithermodularitynorderivationisincompatiblewitheitherRBPorCBP.Wewillthereforefocusonafewinstanceswheretheunderlyingassumptionsofcurrenttheoriesdiffer.3.1 Modularity:ThedomaindivisionOne of the most important distinctions between a classical OT conception of linguisticprocessingandthatofRBPisthattheformerisnotstrictlyatheoryofphonology.Regardlessofthetrendthathasemerged,wherephonologistsarethestrongestsupportersofconstraint-basedgrammars,aclassicalOTgrammarisarankingofgrammaticalconstraintsthatmayalsoincludeconstraintsonmorphologicalandsyntacticstructuresandderivations(Legendreetal.1993,Aissen1999,Hayes2000).Aproperty of standardOptimalityTheoretic proposals isthereforethatthegrammarisnotnecessarilymodular.Morphologicalandsyntacticconstraintscan,inprinciple,berankedbothhigherandlowerthanphonologicalconstraints.Also,singleconstraintsintheliteraturerefersimultaneouslytobothphonologicalandextra-phonologicalproperties,asinthefollowingconstraintthatreferencesbothlinearorderandsemanticscope:(1) SCOPE:Makescopetransparentons-structure (Jager1999)This non-separation of domains predicts thatmany non-occurring patterns could emerge;wherephonologyandsyntaxarepredictedtobeabletointeractinwaysthatareunattested.Remember that phonological features never impactwhether a verb raises fromT to C, for
6
example, justasno formalsyntactic feature influenceswhetherVowelHarmonymayoccur.Withinanon-modularfreerankingofconstraints,therestrictionoflanguagestotheattestedpatternsisunexpected.Admittedly,thephonologicalOTliteraturemakesverylittlereferencetonarrow-syntacticconfigurationsorfeatures,andamodularOTatthatlevelisnotdifficulttoconceive.ParticularlyrelevanttothischapteristhatamodularOTis,infact,activelypromotedwithinStratalOptimalityTheory(Bermudez-Otero2012).Yet,itisimportanttonoteherethattheinstanceswherethephonologyandmorpho-syntaxdoco-referareanintegralpartoftheOTphonological system,anddononethelessbleedover into theSOT framework.Themostbasic non-modular constraints are those like ALIGN (McCarthy& Prince 1993a,b;McCarthy2003),WRAP(Truckenbrodt1999),orMATCH(Selkirk2011).Considerthefollowing:(2) MatchPhrase(Match(Phrase,φ)):
Aphraseinsyntacticconstituentstructuremustbematchedbyacorrespondingprosodicconstituent,callitφ,inphonologicalrepresentation.
(Selkirk2011)Thisconstraintisoneofthree(theothersbeingspecifictowordsandclauses)proposedbySelkirk togovern the interfacebetweenthesyntaxandthephonology.Now, the translationfrom syntax to phonology must be effected somehow, but the formulation of the class ofalignment constraints underpins a more general issue for the computational system, andspecifically for the phonology. The above constraint is not part of a subset of translationoperations(dubbedtheTranslator’sOfficebyScheer),butrather is interleavedwithpurelyphonologicalconstraintsinthephonologicalsystem,asin(3).(3)a.
clause[[verb[noun]NP]VP]clause BinMin(φ,ω) Match(Phrase,φ) i.ι(φ(verbφ(noun)φ)φ)ι * F ii.ι(φ(verbnoun)φ)ι *
b.
clause[[verb[nounadj]NP]VP]clause BinMin(φ,ω) Match(Phrase,φ)F i.ι(φ(verbφ(nounadj)φ)φ)ι ii.ι(φ(φ(verbnoun)φadj)φ)ι *
(Selkirk2011:447)The above tableaux demonstrate blatant non-modularity. Not only is the Match Phraseconstraint ranked directly in relation to BinMin, a constraint that regulates phonologicalbinaritywithinaprosodicdomain,butalsoboththeinputtoGenandtheoutputcandidatesareajumbleofphonological(φ,ι),morpho-syntactic(verb,noun,adj)andsyntactic(clause,NP,VP) information.Thismix in formalvocabularystronglyentails thepresenceofother,non-discussed, constraints never listed in the OT syntax-phonology interface tableaux in theliterature,namely thosepreventing theemergenceof syntactic features in thephonologicaloutput,suchasthefollowing:(4) a. MAX,CLAUSE:EveryinstanceofCLAUSEintheinputmustemergeintheoutput.
7
b. *CLAUSE:NoinstanceofCLAUSEintheinputmayemergeintheoutput.RichnessoftheBase,combinedwiththesyntacticfeaturesintheunderlyingstructuresin(3),impliesthatthenon-emergenceofsyntacticfeaturesinthephonologicaloutputmustbeduetouniversallyundominatedconstraintssuchas*CLAUSE(itisunclearwhattheemergenceofthefeatureCLAUSEinaphonologicaloutputwould look like).Yet,anytimeaconstraintmustbeproposedtobeuniversallyundominated,orafeatureinaninputmustbeexcludedfromalloutputs,werunupagainstthepredictionsofthegeneralOTframework,whereinconstraintrankingis freelymutable. Isuggestherethatthefactthatconstraints like*CLAUSEwouldbeinviolablefindsabettersolutioninastrictlymodulargrammar.Nosyntacticandmorphologicalfeaturesplayaroleinthephonology,andthereforethetableauin(3)doesnotrepresentanypossible derivationwithin the bounds ofUniversal Grammar. This has implications for thestatusoftheProsodicHierarchy,tobediscussedbelowin§4.1and§4.2. NotethattheabovecriticismcannotberestrictedtoCBPderivations,butthataruleinanRBPframeworkliketheoneassumedhereinmustalsoapplytoaphonologicalstringorstructureafterthetranslationofsyntacticstructureintophonologicalvocabularyhasoccurred.Inotherwords,thetranslationofsyntacticdomainsintophonologicaldomainslikePWdandPPh(if theyare indeedphonologicalobjects) isnotconsideredtobepartof thephonologyproperinaradicallymodulartheory(Scheer2012).Atthepointwherephonologicalrulesareapplied, reference to the syntax is impossible. Whether this radical stance is correct hasimplications for any theory of phonology. It is, however, simplermechanically to eliminatereferencetomorpho-syntaxfromanRBTtheorythanfromanOTframeworkwhereconstraintslikeALIGN(alongwithfaithfulnessandmarkednessconstraints)formoneofthefundamentalconstraintclasses.3.2 Derivation:TheSourceofCyclicityAsstatedabovein§2,itisassumedinthischapterthatcyclesinthephonologyareparasiticonacyclicsyntacticgenerativeengine.ItfollowsthatsyntacticfeaturesrelevanttocyclicSpell-Outwilldeterminethesizeandnumberofphonologicaldomains.Awordlikeungrammaticalityinsuchasystemiscomputedinfourcycles,indicatedin(5).(5) [[un]3[[grammatical]1ity]2]4Thestatusofun-asseparatephonologicallyfromitsbasewillbediscussedfurtherin§4.1,butitisclearfromthelackofnasalassimilationevidenced,anditsstatusasaseparate(secondary)stress domain that the phonology has a means to keep this affix separate from its base;grammaticality. In Phase-based terms, the set ofmorphemes {un-} comprises a separatelyprocessedsubsetoflexicalitemsinthederivation(anumeration)andthereforeconstitutesanindependentSpell-Outdomain.Then,wehaveevidencethatgrammaticalityisprocessedin2cycles,asindicatedbythefactthatthestressongrammaticalfromcycle1carriesoverintocycle 2. The stress pattern of grammàticálity is distinguished from the stress onmonomorphemic,singlecycle,longwordsinEnglishlikeLòllapalóoza,wheresecondarystressgenerallyfallsonthefirstsyllable(Pater2000).Wethenmustpostulateacyclewhereun-andgrammaticalityarelinearized,givingusthefourcyclesin(5).
8
WithinaLPM-OTor SOTaccount, it ispostulated that thereareonly threepossiblestrata;stem,word,andphrase.Aswewilllimitourdiscussioninthischaptertoelementsthatareconsideredtobewordswewillrestrictourdiscussiontothefirsttwostrata.Thisadherencetoalimitednumberofstratadoesnotprecludeafour-cycleanalysisof(5),wherethestemstratummayiterate(cycles1and2)asmaytheword(cycles3and4),butthisadherencedoescoincidewithtwootherpertinentpropositions(Butsee§3.4foradiscussionoftheproposalinBermudez-Otero(2012)thatword-levelphonologycannotbesubjecttocyclicreapplication).Thefirstistheproposalthattheparticularstratumthataphonologicalstringissubmittedtoisdetermined by the particular affix(es) added during a cycle. Aswith its predecessor, LPM,affixesinSOTareaffiliatedwithstrata,andthisproposalisthereforesubjecttomanyoftheargumentsagainstLPMintheliterature(seereferencesin§2,andthediscussionbelowin§4.3).Yet,anadvantageoftheproposalthatphonologicalcyclesaremorphologicallydeterminedisthatitaccountseasilyforthere-applicationofstem-levelrules,likethemainstressrule,in(5)in a way that a phase-based system like the one assumed here appears to have troubleexplaining.Inaphase-basedsystemwithoutmorphologicalaffiliationweexpectthatonlythemorphemespresentinthefirst,innermost,phasewillaffectthephonologicaldomainoftheroot(heregrammar)(Marantz2001).Cycle1shouldthereforedelimitthedomainformain-stressassignment,contrarytotheattestedoutput(cf.thelesscommon[[grammátical]1ness]2).Thatthisisnotthecaseisclearlycausedbyanidiosyncraticpropertyoftheaffix–ity;regardlessofthemorphosyntacticmakeupofthestructuretowhichitattachesitwillbeinterpretedaspartofthemainstressdomain.WithinanSOTframeworkthisfallsoutdirectly.InaPhase-basedRBPaccountanextra,lexically-sensitive,processmustbepositedthatmayinsertanaffixintoapreviously interpretedphonologicaldomain toaccount for these facts (Lowenstamm2010).Fortunately,apossiblesolutiontothisproblemcanbeproposed,andanindependentlymotivatedclassofoperations(whichmaybemorphologicallyorphonologicallymotivated),dubbedPhonologicalMergerinNewell&Piggott(2014:353),maybeco-optedtoaccountforthisbehaviour.(6) PhonologicalMerger [X[......PWd]]®[X[..X....PWd]],whereXisanaffix.Thisoperationcausesanouteraffix to incorporate into thephonologicaldomainbuiltonapreviouscycle.Twooftheclearestcross-linguisticmanifestationsofthistypeofoperationareinfixation (7a), or Selkirk’s (1996:207) internal clitics (7b), where the clitic, which issyntacticallyseparate fromthenountowhich itcliticizes,emergesas internal to thestressdomainthatincludesthenoun.(7) a. absobloodylutely b. ugraad ‘tothecity’(Serbo-Croation) PhonologicalMerger,asseenin(7b)isnotrestrictedtooperationsthatmodifylinearorder,aswillbeseenagainbelowinthediscussionofOjibwe.Thestem-levelaffiliationof–ityin(5)maytherefore be ascribed to its phonological (dis)position, rather than its morphosyntacticfeatures.Hereitwouldbethecasethataffixeslike–itytargetthesub-footstructureinthebasetheyattachto,triggeringsomenecessaryre-syllabificationwhichwouldaffectthepositionofstress.iiIncomparisonwiththebehaviourofaffixeslike-ity,§4.3demonstratesthatasingle
9
affixinmanylanguagesmayhavemultiplephonologicalbehaviours.Itwillbearguedthatthebehaviourofthese‘multi-stratal’affixesisbestdeterminedsyntactically,ratherthanmorpho-lexically.Thenotionthatmorphemesbelongtoparticularstrataisthereforenotthedecidingfactorintheirphonologicalbehaviour. ThisisinpartduetothesecondissuethatcoincideswiththeSOTproposalthatthereare a limited number of phonological strata, namely that morphosyntactic features thatcorrespondtothenotionof‘stem’or‘word’donotexist.VolumessuchasJulien(2002),Newellet.al.(inpress)expoundupontheimpossibilityofdeterminingthestatusoftheterm‘word’syntactically.Amorphosyntacticnotionof‘stem’isevenmoreelusive.AsSOTismodulartheory,wherenomorphosyntacticfeatures(besidesthosepickedoutbyalignment)arepermittedinthe phonology, the imposition on the syntax to determine stratal affiliation of affixes istheoretically unmanageable. Within a phase-based account, where the only relevantmorphosyntactic features determining phonological cycles are those proposed to beindependentlynecessarytodeterminesyntacticcycles,thisproblemisnotencountered.Thatsaid,itisnonethelessthecasethatdifferentphonologicalcycleshaveparticularcharacteristics,andthisfactdoesnotfalloutofanyparticularsyntacticaspectofphases.PotentialdirectionstotakeinindependentlymotivatingcyclicphonologicaldifferenceshavealonghistorybasedinnotionssuchastheStrictCycleCondition(Kean1974:179).Restrictionsontheapplicationofphonologicalrulesandthepotentialwaysinwhichtodeterminethephonologicalbehaviourofdifferentcycleswillbediscussedin§5. Thissectionhaspresentedanoverviewoftheissuesthatarecriticaltothediscussionoftheparticularproblemstobepresentedin§4.WewilldiscusshowtheparticularbehaviourofadjunctshasbeenusedtomotivatethenecessarypresenceoftheProsodicHierarchyinthephonologicalcomponent,andhowaderivationalaccountmayeliminatethisnecessityin§4.1and §4.2. In §4.3 we will examine multi-cyclic affixes in Malayalam and Ojibwe, and willconcludethatthecross-linguisticevidencesupportstheconclusionthataffixesarenotlexicallyaffiliatedwithparticularphonologicalstrata.Althoughthemorphosyntax-phonologyinterfacepredictionsofSOTandtheRBPframeworkpresentedhereagreeinmanyaspects,itwillbeseenthattherearedomainswhereintheirimplicationsarenon-overlapping.4. CURRENTCONTRIBUTIONSANDRESEARCH
HerewewillexamineparticularanalysesofphonologicalphenomenaforwhichRBTpremisesarearguedtoofferasimpler,moremodularsolutionthanthatofferedwithinatheorysuchasSOT.§4.1willbeginwithadiscussionofEnglishnegativemorphemesthathasbeenrecurrentintheliteraturesinceatleastBooij&Rubach(1984).Thisdiscussionispre-emptedinNewell(2008),andScheer(2012),andrestsonthepropositionthat,duetothesyntacticgovernanceofcyclicity,wordsthatappeartobestructurallyindistinctmaynonethelessdisplaydivergentderivations.Thisleadsintoadiscussionin§4.2ofdataraisedinBermudez-Otero&Luis(2009)from European Portuguese (EP). Bermudez-Otero & Luis contend that the EP data arguedefinitivelyforthenecessityofarepresentationaldistinctionattheword-level.AsolutionalongthelinesofNewell&Scheer(2007),andNewell(2008)willbearguedtoaccountforthisdatawithinaderivationalmodelaswell, eliminating theneed fora solution thatappeals to theProsodicHierarchy.§4.3willthendiscussthenotionofstrataorlevelsinSOT,anddiscusshowthepropositionthatmorphemesarelinkedtospecificstratadoesnotholdwhenwelookatlanguageswherethesamemorphemeshavedivergentphono-syntacticbehaviour.Wewilllook
10
indetailattheproposalinNewell&Piggott(2014)thatthereisnopossibleOTaccountforHiatus Resolution in Ojibwe. This will lead to a short discussion of affixes whose varyingphonological behaviour must be due to modular interface effects rather than to Stratalaffiliation.Theconclusiontobereachedhere is that, for thedataathand,determinationofstrataorcyclesiseffectedwithoutreferencetotheidentityofparticularmorphemes,butratherthroughsyntacticcomputationalmechanisms.4.1 ThePhonologicalvsSyntacticDerivationofEnglishun-/in-Thereisawell-knownphonologicaldistinctionbetweenthenegativeprefixes in-andun- inEnglish.ThenasalconsonantintheformerassimilatesphonologicallyinPlace(andManner,forsonorants)toafollowingconsonant,whilethenasalinthelatterdoesnot.(8) a. i[m]possible b. u[n]balanced i[n]tolerable u[n]timely i[ŋ]congruous u[n]kempt i[l]licit u[n]lovable i[ɹ]refutable u[n]roundedBooij&Rubach(1984)propose,withintheframeworkofLexicalPhonology,thatthedistinctionbetweenthetwoisrepresentational:in-isincorporatedintothePWdofitsbase(ItisaLevel1affix),whileun-projectsitsownPWd(Level2).(9) a. [in-possible]PWd b. [[un]PWd[balanced]PWd]AssumingthatAssimilationoccursonlywithinaPWd,thedistinctionin(8)iscaptured. It is quite simple to translate the above into an SOT framework. Here, a constraintrequiringassimilationmustberankedhigherthanaconstraintrequiringfaithfulnesstoPlacefeaturesintheStemstratum(Level1),wherein-isintroduced(10),whiletheoppositerankingmustbetrueattheWordstratum(Level2),whereun-isinserted(11).(10) in-possible ASSIMILATE FAITHPLACE inpossible * F impossible *(11) un-balanced FAITHPLACE ASSIMILATE umbalanced * F unbalanced * Theabove,liketherepresentational/LPaccountsuccessfullycapturesthephonologicalpatternobserved.Itdoes,however,leavesomequestionsunanswered.Thefirstiswhethertheabove (re-)rankingsare indicativeof a larger, cross-linguisticpattern. Ithasbeenobservedsincethebeginningofthegenerativephonologicalenterprisethattheoutputofafirstcycle
11
tends to be preserved on a subsequent cycle (ex. Kean 1974). In versions of OT thatcountenancemultipleconstraintrankings/levelsthishasbeentranslatedastheproposalthatre-rankingacrossstratacanonlyleadtothepromotionofFaithfulnessconstraintsinrelationtoMarkednessconstraints.(12) Constrainttypologyandthelimitsofreranking:Thecore-peripheryorganizationofthe
lexiconistheconsequenceofthefactthat, inthetypicalcase,rerankingis limitedtoFaithfulness constraints (PARSE and FILL), within an otherwise invariant constraintsystem.
(Ito&Mester1995:183)The above is not, however, assumed in current OT analyses generally. Bermudez-Otero &Trousdale (2012),alternately,proposes that constraint re-ranking isdue to the life cycleofphonetico-phonological processes, leading to the diachronic demotion of Faithfulnessconstraintsinsmallerdomains.Ineithercase,thedistinctionbetweenun-andin-isnotlinkedreliablytoatenetoftheOTsystem. This brings us to the second question of whether this re-ranking is purelylexico/phonological,orwhetherthedifferinglevelsoffaithfulnessacrosscyclesislinkedtothesynchronicmorpho-syntacticderivationofparticularconstructions.InNewell(2005a,b)itisdemonstratedthattheprefixesin-andun-notonlyhavedistinctphonologicalbehaviour,butalsodivergemorpho-syntactically.(13) (a)phonology (b)morphosyntax:features (c)morphosyntax:structurein- assimilation projectionofadjectival
featuresnoparticipationinbracketingparadoxes
un- noassimilation noprojectionoffeatures participationinbracketingparadoxes
(13a)hasalreadybeendemonstratedin(8).(13b)isdemonstratedbythefactthatallwordsprefixedwith in-areadjectives,whilewordsprefixedwithun-maybeadjectives,nouns,orverbs.(14) a. inept,impossible,*intie,*imbirthday b. unhappy,undo,unbirthdayiiiWhat(14)indicatesisamorphosyntacticdistinctionbetweenthetwoaffixes.Eitherin-selectsforanadjectivalbase,orin-isprojectingadjectivalfeatures.WithinatheorysuchasDistributedMorphologyeachfeaturelessrootmorphememerges(directlyorindirectly)withacategory-defining head (but seeWiltschko&Dechaine 2010 and Borer 2013 for alternate views oflexicalizationandthesyntax-phonologyinterface).Thisproposal,pairedwiththefactthatin-andnotun-attachestoboundroots(ex.inept,withthenotableexceptionofunkempt),aswellasthefactthatinemergesinthesamephonologicaldomain/cycleasitsbase,arguesforthelatter, seen in (15a). The fact that in- cannot attach to nouns and verbs indicates that thismorphememayonlyselectforrootsoradjectives.Asun-doesnotattachtoboundroots,and
12
never affects the category of the base towhich it attaches,we can conclude thatun-, as ischaracteristicofsyntacticadjuncts,projectsnocategoryfeatures(15b).(15) a. a b. a 2 2 in√una possible 2 a√ xhappy Adjuncts,whichdonotprojectmorphosyntacticfeaturesupthetree,havebeenarguedin the syntactic literature (i) to merge a-cyclically, counter to the Extension Condition ofChomsky(2001b) (Lebeaux1988,Nissenbaum2000,Stepanov2001,Ochi1999), (ii) tobeinterpreted phonologically prior to their syntactic merger (Uriagereka 1999), and (iii)participateinbracketingparadoxes(Pesetsky1985,Nissenbaum2000,Newell2005a,b,2008).The divergent syntactic behaviour of adjuncts and non-adjuncts leads to the phonologicaldistinction we see in (8) without calling for additional machinery like prosodic or stratalaffiliationinthephonologicaldomain.Thederivationfor(15a)occursinonecycle,orphase,in-beinganadjectivalphasehead:(16) a 2 à in-possible in√ possibleAssumingforthemomentthattheoutputofcyclicinterpretationtothephonologyincludestheprojectionofprosodicstructure,thePFoutputof(15a)willbeasingledomain.LetusassumethisdomaintobeaProsodicWord.PlaceassimilationtakesplaceatPFinterpretation.(17) [impossible]PWd Thederivationof(15b),however,iscomputedinthreesteps,thefirsttwoofwhich,giventheadjunctnatureofun-,mayoccur inparallel.Anulladjectivalcategorizingheadtriggers thephasein(18a).In(18b)un-isthesolememberofitsnumeration/cycle.(18c)isthestructureafterthea-cyclicmergerofun-,whichmustbeinterpretedatPFtoensureproperlinearizationofitspieces.(18)(a) a 2 √a à [happy]PWd happyx (b) un à [un]PWd
13
(c) a 2 √ x à [[un]PWd[happy]PWd]PWd 2 un√ happy AseachPFinterpretationleadstoprosodicprojection,theoutputof(18)willbeasfollowsthearrowin(18c).Note that theStrict-LayerHypothesis,a tenetofclassicProsodicPhonology(Selkirk1981,1984,NesporandVogel1986),hasbeenabandonedandthereforethenestedPWdstructuresseenherearepermittedbythetheory(followingSelkirk2011,Ito&Mester2013). Thespecificsofthephonologicalruleresponsibleforassimilationwillbeexpandedonbelow.What is important here is that the morpho-syntactic distinctions between the twonegativeprefixesinEnglish,alongwithacyclicaccountofsyntacticinterpretation,leadstothedistinction in phonological structure seen in (17) and (18c). This is the exact structuraldistinction that was proposed in Booij & Rubach (1984). The advantage of the presentproceduralderivationisthatthedistinctionbetweenin-andun-isnon-stipulative.Syntaxfeedsphonology.AssimilationhereapplieswithinaPWd,andthemorpho-syntaxofun-resultsinitsphonologicalseparatenessfromthebasetowhichitattaches. Beforemovingontothenextdataset,itisimportanttorememberthediscussionin§3.1,above.ThePWdstructureassumedintheabovederivationsisinlinewiththestandardlyheldviewinthephonologicalliteraturethattheProsodicHierarchyisthe(indirect)phonologicallinktosyntacticstructure.TheproposalthattheassimilationruleinEnglishissensitivetothisprosodicstructureisthereforealsostandard.(19) Nà[αPLACE]/____C[αPLACE] (*Nà[αPLACE]/____]PWdC[αPLACE])The rule in (19) will not apply if a PWd boundary intervenes between the nasal and thefollowingconsonant,astheenvironmentforassimilation(directlinearadjacency)willthennotbe met. This type of explanation is an illustration of a representational account of thephonological derivation (Newell & Scheer (2007) and Scheer (2010, 2012)). But, thederivationsin(16-18)allowforanalternate,purelyprocedural,accountofthemannerinwhichphonologicalrulesareapplied.Ifweassumethattheassimilationrulemustbeappliedatthepoint in the derivation where the nasal consonant is first interpreted, then reference toprosodic structure becomes superfluous. Let us assume that the nasals in both affixes areunderspecifiedforPlace.In(16)thenasalisinterpretedinthefirstcycleatPFatthesametimeastherootpossibleisinterpreted.Thenasalandtheconditioningsegmentforassimilationarethereforeprocessedinthesamecycle,andassimilationmayoccur.In(18),however,un-isfirstinterpreted alone in (18b). As there is no conditioner for assimilation at this point in thederivation,thenasalwillemergewithdefaultPlacefeatures,herecoronal.Theproposalhereis that these underspecified nasal segments in the input must be fully specified uponinterpretationatPF,eitherbytheapplicationofphonologicalrules,orbydefaultprojectionoffeatures.Atalaterpointinthederivationassimilationwillnotapply(18c),asthenasalisnolongerunderspecified.Giventhisaccountofthedistinctioninassimilationpatternshere,thenreferencetothePWdisnotnecessaryandthereforenottheoptimaltooltoexplainthenasalassimilation patterns. If similar procedural accounts of phonological derivations can be
14
motivatedcross-linguisticallythiswillsupporttheproposalofScheer(2008)thattheProsodicHierarchyisnotthecorrectwaytoaccountforphonologicalcycliceffects.Thistypeofanalysisisdiscussedfurtherin§5. The conclusion here is that a Rule-Based, underspecification account of nasalassimilation inEnglishfollowsfromaphase-based, fullymodularaccountof thederivation.This accounts for both the morpho-syntactic and phonological characteristics of thesederivationswithoutrecoursetoeitherdifferentstrata-specificphonologiesordiacriticlexicalaffiliation. A framework like SOT, where Richness-of-the-Base holds only at the input(Bermudez-Otero2007),canalsoaccountforthis,butitisofnotethattheRBPproposalneednotappeal todifferent constraint rankings/rules toachieve the correct results,making it asimpleraccount.4.2 EuropeanPortugueseandDiminutiveAdjunctionIntheabovesectionwemanagedtoeliminatetheneedtoappealtotheProsodicHierarchyinourphonologicalruleswhilecapturingtheapparentlevel-distinctionsbetweentwoprefixes.Bermudez-Otero&Luis(2009),workingwithanSOTframework,arguesthatthephonologicalcycles in European Portuguese not only support a stratal account of the phono-syntacticderivation,butalsooffercrucialevidencefortheexistenceoftheProsodicHierarchy.Itisofnoteherethattheseauthorsareexplicitlyawareofthepotentialduplicationofexplanations,representationalvsprocedural,exemplifiedbythediscussionofthenegativeprefixesaboveandthattheyattempttopushthepossibilityofaproceduralaccountasfaraspossiblewithintheboundariesofSOT.WithinSOT,remember,thereareafinitenumberoflevelscountenanced;Stem,Word,andPhrase(asopposedtotheWord,Phrase,andClausedomainsofMatchTheory(Selkirk2011)).Inthissectionweofferanalternativeaccountthatagainappealstosyntacticadjunctionandsupportsanaccountofthedatathatneednottakerecoursetorepresentationaldomaindistinctions. Interestingly,Bermudez-Otero&LuisshowEuropeanPortugueseaffixes/cliticstohave(atleast)4differentbehaviouralpatterns.CharacteristicbehaviourofsuffixesinEPisthattheyformasingledomainforstressassignmentwiththeroot,and,ifasuffixbeginswithahighfrontsegment, itwill trigger lenitionof certain root-final /t/, /k/, or /g/s. Suffixes are thereforeclearlystem-levelaffixesaccordingtoBermudez-Otero&Luis,orfirst-PhaseaffixeswithinacyclicRBPaccount.ProcliticsinEPcanbeseparatedfromtheverbbyotherwords,andcanscopeoverconjoinedverbphrases.Theyareclearlyseparatewordsfromtheverb,asopposedtoprefixes,andthereforemustbeintroducedatthephrasalstratumwithinanSOTframework.Enclitics, on the other hand, are not separable from the verb and cannot scope overcoordination,soareclearlyaffixal.But,encliticsneveraffectthepositionofmainstressintheword, and never trigger root-final lenition. It therefore appears that these enclitics aremembersofthewordstratum.Theproblemhereisthebehaviourofthediminutivesuffixes–inhoand–ito,whichfallsbetweenthatoftheproposedstemandwordlevelmorphemes.Thediminutive affixesmay be stressed, like stem-level affixes, but do not trigger lenition, likeenclitics.(20) a. profet-iaàprofecıa (c.f.profeta) ‘prophecy’
15
b. profet-inhaàprofet-ınha ‘littleprophet’
ItisthereforenotpossiblefortheabovepatternstoemergewithinapurelyproceduralSOTsystem.(21) a. [proclitics[[[root-suffixes]diminutives]enclitics]] [phrase[[[stem]word?]word?]]
(Bermudez-Otero&Luis2009) If SOT only allows for 3 strata, then how can the 4-way distinction in affix/cliticbehaviour be accounted for? The solution proposed by Bermudez-Otero & Luis is thatdiminutivemorphemes in the language are representationally distinct from enclitics. Bothdiminutivesandencliticsareintroducedatthewordstratum,butwhilethediminutivesformpartofthePWdwiththestemaffixes,theencliticsarephonologicallyadjoinedtothePWd.(22) a. PWd b. PWd 2 2 profetiɲha PWdlhes @ lev-a-va-mosTheauthorsthereforeconcludethattheProsodicHierarchyisempiricallymotivated,asitisthesolewayinwhichtocapturethis4-waydistinctioninphonologicalbehaviourwithinasinglelanguage. In other words, their conclusion is that the phonology-syntax interfacemust bemoderated both procedurally (stratally) and representationally (prosodically). The authorsadmitthatthisconclusionislinkedtotheSOTproposalthatthestratabelimitedinnumbertothree.If4levelswereallowedfor,wewouldnotneedtoproposearepresentationaldistinctionbetween the diminutives and the enclitics, allowing for a purely procedural account of theinterface.
Each of these proposals,where SOT is either augmented representationally (by theProsodic Hierarchy) or procedurally (by adding an extra stratum) is, however, equallystipulative.Afourthstratum,liketherepresentationaldistinctionin(22)ismotivatedonlybytheneedtoaccountforthe4-waydistinctionin(21).Bermudez-Otero&Luis,andproponentsofSOTingenerallimitthenumberofstratato3inordertoconstrainthepredictivepowerofthe theory (among other motivations). It is unclear, however, how allowing for lexically-specifieddistinctionsinrepresentationalprosodicstructuredoesnotproliferatethenumberof cycles in exactly the way restricting SOT to three levels is meant to prevent. But, if aproliferationofstrataweretobeallowedinstead,wewouldbefacedwiththeproblemthatplaguedSOT’stheoreticalpredecessor,LexicalPhonologyandMorphology(LMP),discussedin§3. LMP had to resort to adding levels, and loops between levels, to explain the manyphonologicaldistinctionsthatneededtobeaccountedfor,bothwithinandacrosslanguages(Kiparsky1982,Mohanan1986).Theseissuesleadustolookforadeeperexplanationofthedistinctionsseen.ThedatainEuropeanPortugueseareindicativeofadistinctionthatisreal,justlikeinthecaseoftheEnglishnegativeprefixes.Wecanthereforeaskourselvesifthereisananalogousphono-syntacticaccountoftheaffixaldivisionsin(21).
16
To this end, consider the analysis of themorpho-syntax of the diminutive affixes inBrazilianPortuguese.ItisarguedinBachrachandWagner(2007)thattheeffectdiminutiveshaveonthepositionofstressinaworddiffersfromthatofotherderivationalaffixes.Bachrach&Wagnerproposethatdiminutiveaffixesbehavelikemembersofco-compounds(dvandva)ratherthanlikederivationalaffixes.Itisarguedthattheyconstituteaseconddomainforstress,rather than constituting part of the stress domain of the root. This is supported by theinteractionofdiminutiveaffixationwithmid-vowelraisinginunstressedpositions(asnotedandreferencedbyBermudez-Otero&Luis).
(23) a. bɛloàbel-ez-a ‘beautiful,beauty’ b. bɔlaàbɔl-ıɲ-a ‘ball,(small)ball’ c. bɔlaàbɔla-zıɲ-a ‘ball,(small)ball’
(Bachrach&Wagner2007:2)
Non-diminutivederivationalsuffixesshiftstressawayfromtherootvowel,leadingtoitsraising(23a).Thediminutiveaffixmaycausedestressingoftherootvowel(23b)ordemotionofrootstresstosecondary(23c).In(23b)therootvowelisarguedtohavebeenstressedonapreviouscycle, and stress clashhasbeen resolved throughdeletion. Interestingly, stress clash is notresolvedinalldialects(ex.cafèzínhoCegalla2008:38).Inboth(23a&b)itisapparentthatthederivationincludestwocyclesofphonologicalinterpretation,wheretherootvowelsreceivestressinthefirst.Whyshouldthesediminutiveaffixesconstituteseparatecyclicdomainsfromtheirhosts?AccordingtoBachrach&Wagner,diminutiveaffixesgenerallyareadjuncts,andarecomposed of a diminutive root and a category-defining head. Non-adjunct suffixes areproposedtobe,ontheotherhand,monomorphemic.(24) a. /iɲ/-Adjunction b. /ziɲ/-Adjunction # #1 4 n1 #1 #2 ti n1 n2 n1 n22 2 2 2 Stemn1Dimn2# Stemn1#1Dimn2#2g | g | g g | g g | g amigxiɲ xo amigxoziɲxo [amigıɲo]‘smallfriend’ [amıgozıɲo]‘smallfriend’
17
c. Derivationalsuffix # n Suffix 2 StemSuffixn# g g | g amig ad xa [amigada]‘concubine’
(Bachrach&Wagner2007:5) The positions of adjunction in (24a,b) are argued to explain other agreement anddistributionalfactsregardingthediminutiveaffixesthatwouldtakeustoofarafieldfromthediscussion at hand. For our purposes it is sufficient to note that the adjunct status of thediminutive affixes brings their derivational status in line with that of un-; they will beinterpretedphonologicallypriortomergerwiththeirbase.Thephonologicalstructureofthewordsin(24)thereforeparallelsthestructuresin(16)and(18c),butnowthisstructurehasmorpho-syntacticmotivation.Thedistinctionhereisthat,contrarytoBermudez-Otero&Luis’analysis,itisthediminutiveaffixesthatare‘PWd’adjoiners.(25) a.[[amig]PWd[ıɲo]PWd]PWd b.[[amıgo]PWd[zıɲo]PWd]PWd c.[amigada]PWdIfthediminutivemorphemesareadjunctstothePWd,aswellasbeingadjunctsinthesyntax,thenwemustaskourselveswhatthederivational/phonologicalpropertiesofthecliticsarethatcausethemtobephonologicallyfartherfromtheirbase.ThesyntaxofPortuguesecliticsandtheirrelationtophonologicalspell-outwillnotbesolvedhere,butwillundoubtedlybetiedtothefunctionalnatureofthesemorphemesandtheresistancefunctionalheadsshowtobearingstress.Inanycase,itwasthedistinctionbetweenthetwo‘word-level’suffixes(diminutivesandenclitics)thatwasthemotivationfortheproposalthattheProsodicHierarchyisnecessarytoaccountfortheEPdata.AssumingthatEPdiminutivesareadjuncts,asinBP,thisentailmentnolongerholds.PhonologicalrulesapplyatPFinterpretation,andadjunctsareinterpretedpriortotheir incorporationintotheNarrowSyntacticstructure.Theenclitics,assuffixes,maybeinterpretedaftermergertotheirbase(see(22b).Inthistypeofstructuretheencliticwillnotbring its own stress into the derivation, and stress determined on the first cycle of PFinterpretationwillpersist. Theentailmentisthenthatitisthetimingofinterpretationdeterminedbythesyntacticstructurethatdistinguishestheproblematiclevelsseenin(21)thatdistinguishesthesurfacestresspatternsinEP,andwecanagaindoawaywithanyreferencetotheProsodicHierarchy.EPsuffixeswillbeinterpretedatthesametimeastheroottowhichtheyattach,whileencliticsmergetoadomainthathasalreadybeenstressed,andsyllabified.Thefinalconsonantsthatdonotundergo lenition inthepresenceofencliticsarethereforestructurallydistinct fromthe
18
consonantsthatdoundergolenitionundersuffixation,justasinthecaseofnasalassimilationinEnglish.Inthecaseofdiminutives,theythemselvesconstituteseparatephasesfromthebaseto which they attach. They therefore, as discussed by Bachrach and Wagner, have thephonological behaviour of a kind of compound. Their syntactic structure predicts theirphonological differentiation from the enclitics. A fully procedural account is mechanicallysimpler than the Strata+PH proposal, and accounts for both the morphosyntactic andphonologicalbehaviourofthemorphemesinvolved. ThissectionhasarguedthatapurelyproceduralaccountofphonologicaldomainsinEPcanbecountenanced, leading toamoreexplanatoryaccountofdata thathaveappeared torequirearepresentationalaccount.Ithasnotbeenshown,however,thatanSOTaccountcannotaccountforthedataathand.AnSOTframeworkcouldbeadaptedtotheaccountabove.Thefollowing section, however, elaborates a pattern that has been argued in Newell & Piggott(2014) tobe the result of cyclic rule application, and, importantly, tonotbederivable in aparallel,CBPframework.4.3 Multi-CyclicAffixesIt is not uncommon cross-linguistically for a single affix to display different phonologicalbehaviour depending on the construction in which it emerges. An example of this is thecausativemorphemeinMalayalam(Michaels2009).Thesamemorpheme,-ikk,marksbothlowand high (traditionally lexical and syntactic) causative constructions. The low causativemorphemeeithercoalesceswiththeroot-finalconsonant(26a),orresolvesanemergenthiatusbydeletionofavowel(26b),whilethehighcausativemorphemedoesnotcoalescewitharootfinalconsonant(27a),andresolveshiatusthroughepenthesis(27b).(26) a. /aaʈ +ikk/ [aaʈʈ] ‘YshakesX’ shake+cause b. /nana +ikk/ [nanakk] `YwatersX' water+cause(27) a. /paaʈ +ikk/ [paaʈikk] `YmakesXsing' sing +cause b. /kaɹa +ikk/ [kaɹajikk] `YmakesXcry' cry +causeThistypeofpatterniscruciallyproblematicforanon-StratalParallelistmodelsuchasclassicOT,astheinputphonologicalstringsinthe(a)and(b)examplesareindistinguishableinallrelevantrespects.Anon-modularsolutiontothisproblemwouldbetoreferencethedifferentsyntacticstructuresof(26)and(27)intheconstraintranking(root-attachedandnon-root-attachedcausativemorphemes,respectively),butthisthenopensuptheissuediscussedin§3,where the mixing of syntactic and phonological constraints predicts a sensitivity to thephonologybythesyntaxthat isunattestedat thesegmental level.Toaccount for theabovepatterninSOT,theaffix–ikkmustbepermittedtobeamemberofmorethanonestratum.Thisisproblematicinthatstrataarethenunderdeterminedbythemorphology,callingintoquestionthepremisethatstrataaremorphologicallydriven(affixesbeingeitherstemorwordlevel).Although SOT is a realizational theory of phonology, where cycles are determined in the
19
syntacticstructure, it isnonethelessproposed,asseenabove, thatasingleaffixatdifferentpositionswill trigger the cycle to which it is affiliated, as seenwith the behaviour of –itydiscussedin§3.2.(Bermudez-Otero2014,butc.f.Bermudez-Otero2015).Taketheexampleoftheaffix–alinEnglish,asdiscussedbyBermudez-Otero(2014).–aldisplaysstemlevel,stress-affecting,behaviourregardlessofwhetheritisaffixedtoaroot(28a)ortoaword-levelaffix(28b).iv(28) a. affıxal (c.faffix) b. governmental (c.f.government)Comparingthebehavioursof–aland–ikk, itisapparentthattheirdifferentbehaviourmusthavedifferentsources,andthereforecannotbothbeduetostratalaffiliationbeinglinkedtoparticularsyntacticpositions.Thatbeingthecase,itissimplertoproposethatthedifferentphonologicalbehavioursofthecausativemorphemearedeterminedbynon-lexicalmeans,andthatitisnotthemorphologicalaffiliationoftheseaffixesthatdeterminestheirphonologicalbehaviour.Notethatthepatternseenin(26)and(27)isnotanisolatedcase.Productivemulti-stratalaffixesarefoundcross-linguistically inunrelatedlanguagessuchasMalagasy,Acholi,Berber, and Ojibwe (Newell 2014). In addition, Bermudez-Otero (2013: 12) proposes thatstem-levelphonologyisnon-analyticalandthereforeliabletoidiosyncraticbehaviour,yetitisclearly not the case that all multi-level affixes display quirky behaviour within the firstphonologicalcycle.Storingpredictablephonological formsisarguablyundesirable(seealsoEmbick&Halle2005,Myler2015forargumentsagainststem-levelstorage). Settingtheseproblemsasideforthemoment,atechnicalaccountoftheMalayalamispossible in SOT, if the causative morphemes are treated as accidentally homophonous (Apropositionthatisalsoproblematic,ashomophonyofthistypeisarguedtobedispreferredorimpossiblebyLeu2015.).AlthoughKilborne-Ceronetal(toappear)arguethatbothinstancesofthecausativemorphemeareheadsofanEventPhrase(followingTravis2010)itispossiblethatthesesyntacticpositionsaredistinguishableinawaythattriggersinterpretationateitherthe stemorword level. Here the domain defined by the low causativemorpheme (E1P) issubjecttoaStem-levelconstraintranking(29),andthedomaindefinedbythehighcausative(E2P)toaWord-levelranking(30).(29) nana+ikk(E1) *HIATUS DEP[J] FAITHV nanajikk * F nanakk * nanaikk * (30) kaɹa-ikk(E2) *HIATUS FAITHV DEP[J]F kaɹajikk * kaɹakk * kaɹaikk *
20
If allmulti-stratal affix behaviour could be accounted for in thisway, then the question ofwhetherRBPandCBPareequivalentlycapableofbridgingtheinterfacebetweensyntaxandphonologywouldbeindeterminateinthissphere. The following Ojibwe data demonstrate, however, that this equivalency does notuniversallyhold.Newell&Piggott(2014)arguethattheOjibwefactsareunderivableinSOT,and thereforewewill expandon their presentationhere.Hiatus inOjibwe (Algonquian) isresolvedbydeletionwithintheEventPhrase(EP)inaverbalconstruction(likeseenaboveforMalayalam)orwithinnPinanominalconstruction(31).BetweenTenseandEP,hiatusisnotresolvediftheprefixisbi-moraic,andisresolvedbyepenthesisiftheprefixismonomoraic(32). Between a person prefix and its base, hiatus is resolved by epenthesis in verbal andalienable possession constructions (33a,b), but by deletion in inalienable possessionconstructions(33c).(31) HiatusresolutionintheEPdomain a. [niɡi:[wa:biwe:ʒi:na:na:niɡ]EP]CP ‘wepaintedthemwhite' ni-ɡi:-wa:bi-we:ʒi:-in-a:-ina:ni-Ø-aɡ '1-PAST-WHITE-PAINT-FINAL-TS(3THEME)-1PLURAL-IND-3PLURAL b. [o[name:miwa:n]nP]DP ‘theirsturgeon(s)' o-name:-im-(i)wa:-an '3-sturgeon-possessive-3plural-obviative’(32) HiatusResolutionbetweenTenseandEP a. [ɡi:[a:ɡamose:]EP]CP 'hewalkedinsnowshoes' ɡi:-a:ɡam-ose: 'past-snowshoe-walk' b. [niɡa[da:ɡamose:]EP]CP 'Iwill(probably)walkinsnowshoes' ni-ɡa-a:ɡam-ose:-Ø '1-future-snowshoe-walk-Fin'(33) HiatusResolutionbetweenPersonMarkeranditsbase a. [ni[da:ɡamose:EP]CP 'Iwalkinsnowshoes’ ni-a:ɡam-ose: '1-snowshoe-walk’ b. [ni[dakwe:mnP]DP 'mywife' ni-akwe:-im '1-woman-possessive' c. [no:komis]DP 'mygrandmother' ni-o:komis '1-grandmother'Giventhefactthatthedatain(31-33)consistuniquelyofwords,andnotphrases,letusagainrestrictdiscussionofanSOTaccountto2strata(stem,word).Asthesuffixesin(31)emergeinsyntactic positions that are closer to the root than the prefixes (inside EP and nP), hiatusresolutionbydeletionmustbeeffectedatthestemlevel,asseenbelowforthefirstcycleof(31b).
21
(34) name:-im-(i)wa:-an *HIATUS DEP[d] FAITHVF name:miwa:n ** name:imiwa:an ** name:dimiwa:dan ** Are-rankingofconstraintsat theword-levelwillgivethecorrectoutput fornon-resolutionbetweenthetensemorphemesasin(32a),herein(35).Anadditionalrestrictiononmono-moraicprefixes suchas (36a)will forceepenthesis in cases like (32b (and33a,b)), seen in(36b).(35)ɡi:-a:ɡam-ose: FAITHV DEP[d] *HIATUSF ɡi:a:ɡamose: * ɡi:ɡamose: * ɡi:da:ɡamose: * (36) a.*HIATUS(μ): Amonomoraicvowelmaynotbefollowedbyavowel. b. ni-ɡa-a:ɡam-ose: FAITHV *HIATUS(μ) DEP[d] *HIATUS niɡaa:ɡamose: * * niɡa:ɡam-ose: * F niɡada:ɡamose: * Theproblemariseswhentryingtoaccountforthedistincthiatusresolutionstrategiesin(33b)and (33c). Unlike in the case of Malayalam causatives, the person prefixes in the Ojibwepossessionconstructionscannotbestacked.ItisarguedinNewell&Piggott(2014)thattheyemergeinDP;intheidenticalsyntacticpositionineachconstruction.Relevantstructuresforalienableandinalienablederivationsareseenin(37)and(38),respectively(strikethroughsindicateelementsthathavemoved).
22
(37) DP 4 [+1]Dprox 4 niD nP 4 n’ 4 n AGRP2 4 AGR2’ 4 AGR2 AGRP1 4 4 AGR1 AGR2 AGR1’ 4 aɡ 4 POSS AGR1 AGR1 POSSP 4 ina:ni 4 n POSS [+1] … 4 im prox √ x ni DUCK ʒi:ʃi:b (Newell&Piggott2014:347)(38) DP 4 [+1] D' pro 5 niD nP4 4 n D n’%Ø 4GRANDMOTHER n AGR2Po:komis-ina:ni 4 AGR2’ 4
AGR2 AGR1P 4 AGR1’ 4 AGR1 nP 4 n’ 4 n √
(Newell&Piggott2014:350)It is therefore impossible to distinguish the two prefixes based on either their featural orsurfacedistributionalproperties.Newell&Piggottarguethatthederivationin(33c)seenin
23
(38)hastheattestedoutputduetothefactthatininalienablepossessionconstructiontheroot(hereo:komis)raisestoDtocheckitsargumentstructurefeatures(Inalienablenouns,unlikealienablenouns,areineffable inOjibwewithoutapossessiveprefix).Thederivationin(37)showsthattheroot,hereʒi:ʃi:b,doesnotraiseoutofnP.Therootanditssuffixesarethereforeinterpretedinthefirstphase(nP),whiletheprefixisnotintroducedorinterpreteduntilthesecondphase(DP),paralleltotheEPencliticderivationintheprevioussection.In(38)bothni-ando:komisareinterpretedinthesamecycle,astheyemergeinthesamephase(DP).Itisthisfact,ratherthananyfactabouttheparticularpropertiesofthepersonprefix,thatdeterminesthathiatuswillberesolvedthroughdeletion.Notethatthepersonprefixesarenotamenabletoanaccountlikethatfor–al,astheydonotalwaysbehavelikestem-levelaffixes(infact,stem-levelbehaviourisexceptionalfortheseaffixes),noraretheyamenabletoamulti-levelaccount,as their syntactic positions upon PF interpretation are invariable. The Ojibwe derivationsthereforeevidencemultipleissuesforastratal,constraint-basedaccount.First,asseenabove,morphologicalaffiliationtoaparticularstratacannotaccountforthebehaviourofthepersonprefix.Secondly,non-analyticlistingofstem-levelexpressionswouldforcethestorageofalargenumberofpredictableinflectedformsinthelanguage,aproposalthatisclearlyundesirable. Furthermore,Ojibweshowscyclicreapplicationofprocessesthatmustbeconsideredword-levelinSOT,aphenomenonthatisproposedtobeunattestedinBermudez-Otero(2012):§3.1.IfboththeTenseandpersonprefixesareword-levelaffixes,andtheword-levelcannotiterate,thispredictsthateitherword-levelphonologywillonlyapplytotheinnermostword-levelaffix,orthatword-levelphonologywillapplytobothprefixeswithinoneapplicationofthecycle.Neitherofthesederivationsgivesthecorrectresult.Newell&Piggottproposethatthereisamorenuancedexplanationforwhyepenthesisoccursthanstatedintheconstraintin(36a).ItisproposedthereinthatdegeneratefeetareillicitattheleftedgeofaPWd.Amono-moraicprefixwillthereforemoveintothePWdtoitsrighttopermittheconstructionofalicitprosodicstructure,anotherinstanceofPhonologicalMerger,repeatedhere.(39) PhonologicalMerger [X[......PWd]]®[X[..X....PWd]],whereXisanaffix.In linewith the attempt in this chapter to avoid a dependency on the upper levels of theProsodicHierarchy,notethattheseprefixescannotprojectalicitfoot.Thisalonecanmotivatetheir incorporationintothedomaintotheirright,erasingtheneedtoreferencethePWdin(39).Consider(40);areiterationof(32b)showingfootstructureandamoredetailedsyntacticbracketing,whichcontainsbothatenseandapersonprefix:(40) [(niɡa)(da:)(ɡamo)(se:)]PWd 'Iwill(probably)walkinsnowshoes' [ni[-ɡa-[a:ɡam-ose:-Ø]EP]TP]CP '1-future-snowshoe-walk-Fin'Twothingsarepertinenthere.First,notethathiatusisresolvedbyepenthesisbetweenthetensemorphemeandtheverb.ThisindicatesthatPhonologicalMergerhasoccurred.Secondly,note that the person prefix and tensemorphemes are footed together, as indicated by thesecondarystressonga-andnotonni-.Ojibwestress isexhaustive,anddegeneratefeetarepermittedasalastresort,attherightedgeofadomain.Itfollowsthatthephonologyoftheword in (40) is computed in three cycles. The first is the PF interpretation of the EP;
24
[(a:)(gamo)(se:)].Ifhiatusemergedinthisdomainitwouldberesolvedbydeletion,asin(31).The next cycle of interpretation is the complement of C; TP. Here ga-, beingmonomoraic,undergoes Phonological Merger into the domain projected on cycle 1, givingga[(gada:)(gamo)(se:)].Herehiatusisresolvedthroughepenthesis.Notethatitcannotbethecase thatni- andga- are interpreted in the same cycle. If theywere, they could be footedtogetheratSpell-Out,andwewouldpredictthatthey,likethebi-moraictenseprefixes,wouldneitherundergoPhonologicalMergernorresolvehiatus,as in(24a). InthefinalcycleCPisinterpretedandni-alsoundergoesPhonologicalMerger,givingni-gi[(nigi)(da:)(gamo)(se:)].Notethatwerethetensemarkernot there, thepersonprefixwouldalsotriggerepenthesis(nida:gamose:). That each of the prefixes prompts the same phonological processesindependentlyleadstotheconclusionthattherearethreephonologicalcycleswithintheword,andthattwoofthemareword-level.Therefore,theword-levelphonology,inthetermsofSOT,may iterate. The particulars of the phonological rule that can account for the distinctionbetweendeletionandepenthesisinOjibwewillbetakenupinthefollowingsection. Thissectionhasarguedthatphonologicalruleapplicationisnotgovernedbyalimitednumberofstrata,andthatneithermorphemesnorparticularmorpho-syntacticdomainscanbelinkedwithcycle-specificphonologicalprocesses.Thisispredictedwithinacomputationalsystemwherenotionslikestemorwordarenotprimesinanymodule.ItisthereforearguablysimplerheretodoawaywiththenotionofstrataandtoremainwiththenotionthatcyclicSpell-Outistriggeredatcertainpointsinthesyntacticderivation.AnymorphemesituatedwithinaSpell-Out domainwill undergo interpretation, leading to the expectation thatwemay findmorphemes that display varying phonological behaviour dependent on their syntacticconfigurationinaparticularderivation;anexpectationthatisborneout.5. FUTUREDIRECTIONSIn this chapterwe have focused on two notions crucial to any theory of the phonology-morphosyntax interface;modularity andderivation.Wehave examined twovery closely-relatedtheoriesandseenthatthedistinctionsbetweenthemarequitenuanced.Wehavenotdelvedtoodeeply,however,intotheexactformofrulesinaRBPthatwillgiveusthekindsofoutputsthatweseeinthischapterandcross-linguistically.Atheorylike(S)OThashadasone of its foci an investigation of how constraints may rerank across strata and acrosslanguages.ThisisduetothefundamentalpremiseofOTthatallconstraintsarepotentiallyactiveinthegrammarofeachlanguage.Rule-basedtheories,however,astheyarenotbasedon the contention that all rules are present in all languages, have been based upon thepremisethatrulesmayormaynotbeactiveinaparticulargrammar,orinaparticularcycle.Anyconsistentcycliceffectsontheformoroutputofrules,orpatternsintheoutputformsatdifferentcyclesarenotderivablefromthebasictenetsofeitherOTconstraintrankingorruleconstructioninandofthemselves.Eachofthesetheoriesmustthereforeworkatmotivatingthepatternswedosee.OneevidentpatternissomethingliketheStrict-CycleCondition(41),whichisakintothePhase-theoreticnotionofPhaseImpenetrability(42).(41) “<an> ...association created in the inner domain cannot be undone in an external domain.» (Kaye1995:307)
25
(42) “[thephonologicalcomponent] isgreatlysimplifiedif itcan ‘forgetabout’whathas been transferred to it at earlier phases; otherwise, the advantages of cyclic computationarelost”(Chomsky2004:107)Both of these formulations are stipulative or descriptive rather than explanatory. Theunderlyingexplanation forwhy it shouldbe thecase that cyclicoutputs shouldpersist isthereforeadomain thatneeds further investigation.Wehaveseenherein thatBoth Itô&Mester(1995)andBermúdez-OteroandTrousdale(2012)haveproposedthatthedifferentrankings of faithfulness constraints in different parts of the grammar can explain cyclicphonologicalpersistence.Thesetheoriestakethemotivationfortheserankingdistinctionsto be controlled at least partially by extra-phonological elements like restrictions ondiachronic change or the existence of co-phonologies. Here I have offered a purelyphonology-internalmotivation forpersistence;structure-building(seeNewell2014,2015forfurtherdetails).Phonologicalinterpretationaltersthetargetofarule,butwillnotalwayseradicatethestructuralenvironmentfortheapplicationofarule,leadingtodivergencesinruleapplication(seealsoHoneybone2005foradiscussionofhowtheamountofmelodicstructureintherepresentationofasegmentaffectsphonologicalruleapplication). TakethehiatusresolutionstrategiesinOjibwe;deletionandepenthesis.Atthefirstcycleof interpretationsegmentsmaybeunderspecified,andwillnotyethaveundergonesyllabification,norwilltheyhavebeenorganizedintofeet.Hereabanonhiatusisresolvedthroughadeletionrule(Rememberthatconstraints,orbans,oncertainstructuresarenotdisallowedinanRBPframework).Breakingdownthetimelineof,say,theconstructionofsyllables,we are confrontedwith a point in the derivationwhere itmust be determinedwhichsegmentsinastringhavethepropertiesthattheycouldformlicitnuclei(e.g.sonority).Therefore,beforeprojectionofanysyllabicstructure,thederivationiscognizantofwhetherastringcontainsaVVsequence.Atthispoint,deletioncanoccurbeforesyllabification,andconsequentlynosuprasegmentalinformationisdestroyedbythisoperation(VVàV).Inthecaseofhiatusresolutionacrosscycles,afterPhonologicalMerger,thesequenceofvowelsisonewhereeachofthetwosegmentsisenvelopedinsuprasegmentalinformationstructure(Theyhavebeensyllabified).NotethatPhonologicalMergerinOjibweonlyoccursifaprefixcannotprojectalicitfoot.Thisinabilitycanonlybedeterminedafterprojectionofstructureat the syllabic level. The sequence (V)σ(V)σ is crucially structurally distinct from thesequenceVV,but similar enough in that the constraint againsthiatus is still triggered. Inorder toconserve thepreviously-builtsyllabicstructureandsatisfy*Hiatus,epenthesis iseffected.Now,thismayappeartobeveryclosetoanOTaccount,whereconstraintsarepittedoneagainsttheother.Itiscontendedhere,however,thatitisnotthestructureofaruleoraconstraintthatcruciallydistinguishesCBPsfromRBPs.Itisratherthatthemotivationfortheapplicationornon-applicationofoperationsinanRBPaccountmustbepurelyphonological(features, structures), where the overt application of an OT constraint is due to extra-phonologicalconsiderations(ranking).AsbothRBPandCBPframeworksrefertothesamephonologicalprimes(features,syllables,etc.) thenconstraintrankingbears theburdenofrequiring independent justification. If accounts like the ones herein, where cyclicinterpretation paired with structural underspecification can account for the cross-cyclicpatternsattested,thenthisweakensthesupportforanexplanationthatcallsonconstraintre-ranking. Teasing apart the nuances of these types of accounts is a fertile area ofinvestigation.
26
Otherfutureresearchdirectionspertinenttothequestionsraisedinthischapterarethefollowing:
(1) Doweneed theProsodicHierarchy/representationaldomaindelimitation?Does itperform phonological duties that cannot be subsumed by cyclic, proceduralderivation?Relatedly,canSOTfunctionintheabsenceofAlignmentconstraints?
(2) What are the restrictions on the destruction/modification of previously computedphonological structure? It is uncontroversial that deletion and feature-changingoperationsexist.Thisbeingthecase,whatconstrainstheirapplication?
(3) Is the distinction between rules and constraints the issue here, or are the crucialdistinctionsbetween these frameworks in thedifferent assumptions regarding theorganizationofthegrammar?Howfundamentalarenotionsofstructure-buildingvs.constraints-on-structure,andwhatarethedistinctionsbetweenthem?
(4) What are the pertinent cycles that determine the timing of phonologicalinterpretation,andhowaretheydefined?
ThisisobviouslynotanexhaustivelistoffutureresearchdirectionsrelatedtotheRBP/CBPdebate at the interfaces. These questions can however help guide us to a deeperunderstandingof theorganizationof theentiregrammatical computational system.Theiranswerscanleadusnotonlytoabetterunderstandingofthephonologicalmodule,butalsoofthesyntacticmodulethatunderliesit,andofthetranslationoperationsbetweenthetwo.Wewill finishherewithashort listof theadvantagesofanRBPtheorypresented in thischapter. First, eliminating reference to the ProsodicHierarchy is desirable on theoreticalgrounds.ThePHisnotfundamentaltotheRBPtoolboxinthesamewaythatALIGNisclaimedtobeinCBP.Secondly,phonologicalcyclesarenotrestrictedinnumberinanRBP.WehaveseenthatoncewenolongertakerecoursetoaProsodicHierarchythatthisproliferationofcyclesisunavoidableand,crucially,necessary.Thedifferentphonologicaltendenciesateachcyclethenneedtobegovernedsolelybyphonologicalfeaturesandstructures.Finally,cyclesinanRBPdon’tmakereferencetoelementsthatareundefinableinthesyntacticcomponent(stem,word,andevenphraseinsomeframeworks).Syntacticandphonologicalcyclesaredeterminedbasedonidenticalelements;thosewhichdefineasyntacticcycle,orPhase.
6. FURTHERREADINGNespor,M.&I.Vogel(1986/2007)“ProsodicPhonology/Prosodicphonology:withanewforeword”ThisclassicworkontheProsodicHierarchyoffersadetaileddiscussionofthe motivations for the theory, as well as a trove of data relevant to the study of themorphosyntax-phonologyinterface.Selkirk, E. (2011) "The syntax-phonology interface" This is one of themore recentupdatesoftheTheoryoftheProsodicHierarchy(bytheoriginalproponentofthetheory)thatdemonstrateshowtheviewofstructure-phonologyrelationshasevolved.Scheer,T.(2010)“Aguidetomorphosyntax-phonologyinterfacetheories:howextra-phonologicalinformationistreatedinphonologysinceTrubetzkoy’sGrenzsignale”
27
Thisbookcoversthetheoreticalpaththattheoriesoftheinterfaceingenerativelinguisticshavetakeninminutedetail,andoffersanin-depthanalysisofthetheoreticalpremisesandimplicationsofeachadvance.Bermúdez-Otero(toappear)“StratalOptimalityTheory” Currentlypostedonhiswebsiteasalibraryofseparatepapers,thisbookwillcoverthetheoryofStratalOptimalityTheory, offering a detailed look at the data that support a cyclic account of themorphosyntax-phonologyinterface.Newelletal.(toappear)“TheStructureofWordsattheInterfaces”Thisisacollectionofworks,manyofwhichfocusoninteractionsatthemorphosyntax-phonologyinterfacewithaneyetodetermininghowtheelementsthatwepre-theoreticallydescribeas‘words’emergeatthePFinterface.
7. RELATEDTOPICSThetwootherchaptersintheRBPsection,theOTsectionchapterontheinterfaces,CharlesReiss’chapteronSubstance-FreePhonology,andBermudez-Otero,thisvolume.
8. REFERENCES Adger,D(2006)Stressandphasalsyntax,ms.QueenMaryCollege,UniversityofLondon.Aissen, J. (1999)Markednessandsubject choice inOptimalityTheory.NaturalLanguage& LinguisticTheory17(4).p.673-711.Avery, P. andRice, K. (1989) Segment structure and coronal underspecification.Phonology, 6(02).p.179-200.Bachrach, A. and Wagner, M. (2007) Syntactically driven cyclicity vs. output-output correspondence: the case of adjunction in diminutivemorphology.U. PennWorking PapersinLinguistics10(1).Bakovic,Eric.(2007)Arevisedtypologyofopaquegeneralisations.Phonology24(02).p.217- 259.Bale,A.,Papillon,M.andReiss,C.(2014)Targetingunderspecifiedsegments:Aformalanalysisof feature-changingandfeature-fillingrules.Lingua148.p.240-253.Bermudez-Otero,R.andLuis,A.(2009)Cyclicdomainsandprosodicspansinthephonology ofEuropeanPortuguesefunctionalmorphs.Presentedatthe6thOldWorldConference inPhonology(OCP).Edinburgh.Jan.24.Bermudez-Otero,R.andTrousdale,G.(2012).Cyclesandcontinua:onunidirectionality and gradualness in language change. In T. Nevalainen and E. Closs Traugott (eds). The OxfordhandbookofthehistoryofEnglish.NewYork:OxfordUniversityPress.Bermudez-Otero,R.(2013)Thestem-levelsyndrome.UPennLinguisticsDepartmentSpeaker Series.Philadelphia11.Bermudez-Otero, R. (2007). Marked phonemes vs marked allophones: segment evaluation in Stratal OT. Paper presented at theWorkshop on Segment Inventories, GLOWXXX,Tromsø,Apr.11.
28
Bermudez-Otero,R.(2014)Strataloptimalitytheory.OxfordStudiesinTheoreticalLinguistics. Oxford: Oxford University Press. Partly available at [http://myweb. tiscali. co.uk/bermudez/Stratal_Optimality_Theory.htm]Bermudez-Otero,R.(2015)Paradigmaticeffectswithoutoutput-outputcorrespondence:the roleoflexicalacquisition. Paper presented at Spell-out and the Syntax-Phonology Interface.Tromsø.Apr.24.Bermudez-Otero, R. (2012) The architecture of grammar and the division of labour in exponence.InTrommer,J.(ed.)Themorphologyandphonologyofexponence(Vol.41). OxfordUniversityPress.Boeckx,C.andGrohmann,K.(2007)Remark:Puttingphasesinperspective.Syntax10(2).p. 204-222.Boersma, P. (1997). How we learn variation, optionality, and probability. Proceedings of theInstituteofPhoneticSciences21.p.43–58.UniversityofAmsterdam.Booij, G. (1997) Non-derivational phonology meets Lexical Phonology. In Roca, I. (ed.) DerivationsandConstraintsinPhonology.ClarendonPress,OxfordBooij, G. and Rubach, J. (1984)Morphological and prosodic domains in Lexical Phonology. Phonology1.p.1-27.Borer,H.(2013)TakingForm:StructuringSense,Vol.III.OUP.pp.432.Boskovic,Z� .(2014).NowI’maphase,nowI’mnotaphase:Onthevariabilityofphaseswith
extractionandellipsis.LinguisticInquiry,45(1),27-89.Cegalla, D. P. (2008) Novíssima gramática da língua portuguesa. 48.ed. Sao Paulo: CompanhiaEditoraNacional.Chomsky,N.andM.Halle.(1968)ThesoundpatternofEnglish.Harper&Row.Chomsky,N.(1999)DerivationbyPhase.MITOccasionalPapersinLinguistics,18.MIT WorkingPapersinLinguistics.DepartmentofLinguisticsandPhilosophy,Cambridge, Mass.Chomsky,N.(2001a)Derivationbyphase.InKenstowicz,M.(ed.)KenHale,alifeinlanguage. MITPress.Chomsky,N.(2001b)BeyondExplanatoryAdequacy.MITOccasionalPapersinLinguistics20.
p.1-28.Chomsky,N.(2004)BeyondExplanatoryAdequacy.InBelletti,A.(ed.)StructuresandBeyond. TheCartographyofSyntacticStructure(3).Oxford:OUP.Clements,GeorgeN.(1985)Thegeometryofphonologicalfeatures.Phonology2(01).p.225- 252.Embick,D,andM.Halle.(2005)Onthestatusofstemsinmorphologicaltheory.InT.Geerts,I.
van Ginneken, and H. Jacobs (eds) Romance languages and linguistic theory 2003:Selected papers from ‘Going Romance’ 2003, Nijmegen. p. 37–62. Amsterdam: JohnBenjamins.
Embick,D.(2014)Phasecycles,φ-cycles,andphonological(in)activity.Theformofstructure,thestructureofforms:essaysinhonorofJeanLowenstamm,271-86.
Embick,David.(2010)Localismversusglobalisminmorphologyandphonology.Vol.60.MIT Press.Epstein, S. and Seely, T. D. (2002) Rule applications as cycles in a level-free syntax. In Derivationandexplanationintheminimalistprogram.Blackwell.
29
Fabb, N. (1988) English suffixation is constrained only by selectional restrictions.Natural Language&LinguisticTheory6(4).p.527-539.Halle,M.andMarantz,A.(1994)SomeKeyFeaturesofDistributedMorphology.MITWorking PapersinLinguistics,21.p.275-288.Halle,M.andMarantz,A. (1993)DistributedMorphologyand thePiecesof Inflection. InK. HaleandS.J.Keyser(eds.)TheViewFromBuilding20.Cambridge,Mass.:MITPress.Hayes, B. (2000) Gradient well-formedness in Optimality Theory. In J. Dekkers, F. van der Leeuw, and J. van de Weijer (eds.) Optimality Theory: Phonology, syntax, and acquisition.Oxford:OxfordUniversityPressHoneybone,P.(2005)Sharingmakesusstronger:processinhibitionandsegmentalstructure.
InPhilipCarr,JacquesDurandetColinJ.Ewin(eds.)Headhood,elements,specificationand contrastivity: phonological papers inhonourof JohnAnderson. Amsterdam: JohnBenjaminsPublishingCompany.p.167-192.
Ito, J. andMester,A. (1995)Thecore-peripherystructureof the lexiconandconstraintson reranking.InJ.Beckman,S.Urbanczyk and L. Walsh (eds.) Papers in Optimality Theory,Amherst:GLSA.Ito,J.andMester,A.(2013)ProsodicsubcategoriesinJapanese.Lingua124.p.20-40.Jager,G.(1999)Optimalsyntaxandoptimalsemantics.PaperpresentedatDIP-colloquium. Julien,M.(2002)Syntacticheadsandwordformation.OxfordUniversityPress.Kaisse,E.M.,andShaw,P.A.(1985)OnthetheoryofLexicalPhonology.Phonology2(01)p.1- 30.Kaye, J. (1995) Derivations and Interfaces. In J. Durand and F. Katamba (eds.) Frontiers of Phonology.London&NewYork:Longman.Kean,M-L.(1974)Thestrictcycleinphonology.LinguisticInquiry5(2).p.179-203.Kilborne-Ceron,O.,Newell,H.,Noonan,M. andTravis, L. (to appear) Phase domains at PF: RootSuppletionanditsImplications.InH.HarleyandD.Siddiqi(eds).Morphological Metatheory.Benjamins.Kiparsky,P. (1982)Fromcyclicphonology to lexicalphonology. InH.vanderHulstandN. Smith(eds.)Thestructureofphonologicalrepresentations1. Dordrecht,Foris.Kiparsky,P.(2000)OpacityandCyclicity.TheLinguisticReview17.p.351-367.Lebeaux, D. (1988) Language acquisition and the form of the grammar. Ph.D. dissertation. UniversityofMassachusetts,Amherst.Leben,W.(1973).Suprasegmentalphonology.Ph.D.dissertation.MIT,Cambridge,MA.Legendre,G.,Raymond,W.andSmolensky,P.(1993)AnOptimality-theoreticTypologyofCase andGrammaticalVoiceSystems.IntheProceedingsoftheNineteenthAnnualMeetingof theBerkeleyLinguisticsSociety.Berkeley.p.464–478.Leu, T. (2015) Is there No homophony? Paper presented at the meeting of the Canadian LinguisticSociety.UniversityofOttawa.May30.Lowenstamm,J.(2010).Derivationalaffixesasroots(phasalspelloutmeetsEnglishstress
shift).Ms.,UniversitéParis-DiderotandCNRS.Marantz,Alec.(1997)Noescapefromsyntax:Don'ttrymorphologicalanalysisintheprivacy ofyourownlexicon.UniversityofPennsylvaniaworkingpapersinlinguistics4(2).Marantz,Alec.(2001)Wordsandthings.Ms.MIT.Marvin,T.(2002)TopicsintheStressandSyntaxofWords.Doctoraldissertation.MIT.McCarthy, J. (2000) Harmonic serialism and parallelism. Linguistics Department Faculty PublicationSeries.Paper98.
30
McCarthy, J. (2007) Hidden Generalizations: Phonological Opacity in Optimality Theory. London:Equinox.McCarthy,J.(2003)OTconstraintsarecategorical.Phonology20.p.75–138McCarthy,J.(1988)Featuregeometryanddependency:Areview.Phonetica45.p.84-108.McCarthy,J.andPrince,A.(1993a)Generalizedalignment.InG.BooijandJ.vanMarle(eds.) Yearbookofmorphology.p.79–153.Dordrecht:Kluwer.McCarthy, J. and Prince, A. (1993b) Prosodic morphology: Constraint interaction and satisfaction.Ms.UniversityofMassachusetts,AmherstandRutgersUniversityMichaels,J.(2009).Toalternateornottoalternate:Whatistheboundary?InKahn,S.,Moore- Cantwell,C.,andStaubs,R.,(eds).ProceedingsofNorthEastLinguisticSociety40,p.93- 106.Mohanan,K.(1986)Thetheoryoflexicalphonology.Dordrecht:D.ReidelPublishingCompany.Myler, N. (2015) Stem Storage? Not Proven: A Reply to Bermudez-Otero 2013. Linguistic Inquiry.46(1).p.173-186Newell,H.,Noonan,M.,Piggott,G.,andTravis,L.(eds)(inpress)TheStructureofWordsatthe Interfaces.OUP.pp.352.Newell, H. and Piggott, G. (2014) Interactions at the syntax–phonology interface: Evidence fromOjibwe.Lingua150.p.332-362.Newell,H.andScheer,T.(2007).Proceduralfirst.Papergivenatthe38thPoznan LinguisticMeeting,Gniezno,Sept.13-16.Availableat [http://www.unice.fr/dsl/tobias.htm.]Newell,H.(2008)AspectsoftheMorphologyandPhonologyofPhases.Doctoraldissertation. McGillUniversity.Newell,Heather.(2005a)Bracketingparadoxesandparticleverbs:alateadjunctionanalysis. InBlaho,S.,Vicente,L.andSchoorlemmer,E.(eds.)ProceedingsofConsoleXIII.p.249- 272.Newell,Heather. (2005b)A lateadjunctionsolutiontobracketingparadoxes.Proceedingsof the35thannualmeetingoftheNorthEastLinguisticSociety(NELS).No.2.Newell, H. (2014) Phonological Persistence. Paper presented at the North American PhonologyConference(NAPhC).ConcordiaUniversity.May9-11.Newell, H. (2015) Structural Sensitivity in Phonology: Phonological Persistence. Paper presentedat the12thOldWorldConferenceinPhonology(OCP).Barcelona.Jan.30.Newell,H.(2016)TheOntologyofEnglishmorpho-phonology.Presentedatthe24thManchester
PhonologyMeeting(MfM).UniversityofManchester.26-28mai.Newell,H.(2016)EngishLexicalLevelsarenotLexical.Ms.UQAM.Nespor,M.andI.Vogel.(1986)ProsodicPhonology.Dordrecht.Foris.Nespor,M,andI.Vogel.(2007)Prosodicphonology:withanewforeword.Vol.28.Walterde
Gruyter.Nissenbaum,J.(2000)Investigationsofcovertphrasemovement.Doctoraldissertation.MIT.Ochi,M. (1999).Multiple Spell-Out and PF adjacency. In theProceedings of the North East LinguisticSociety29.UMass.Amherst.Pater, J. (2000)Non-uniformity inEnglishsecondarystress: theroleofrankedand lexically specificconstraints.Phonology17(02).p.237-274.Pesetsky,D.(1985)MorphologyandLogicalForm.LinguisticInquiry16.p.193-246.Prince,A.andSmolensky,P. (1993)OptimalityTheory:Constraint interaction inGenerative Grammar.ms.
31
Prince, A. and Smolensky, P. (2008)Optimality Theory: Constraint interaction in generative grammar.JohnWiley&Sons.Sagey,E.(1986)Therepresentationoffeaturesandrelationsinnon-linearphonology.Doctoral dissertation.MIT.Scheer,T.(2004)Alateraltheoryofphonology.Vol.1.MoutondeGruyter.Scheer, T. (2008)Why the prosodic hierarchy is a diacritic andwhy the interfacemust be direct.InHartmann,J.,Hegedus,V.andvanRiemsdijk,H.(eds.)Soundsofsilence:empty elementsinsyntaxandphonology.Vol.63.BRILL.p.145-192.Scheer, T. (2010) A guide to morphosyntax-phonology interface theories: how extra- phonological information is treated in phonology since Trubetzkoy’s Grenzsignale. WalterdeGruyter.Scheer,T. (2012)Chunkdefinition inphonology:prosodic constituencyvs.phasestructure. ModulesandInterfaces.p.221-253.Scheer, T. (2013) Melody-free syntax. Paper presented at the Workshop on complexity at UQAM,Montreal.Feb.8.Selkirk,E.(1981)Onthenatureofphonologicalrepresentation.AdvancesinPsychology7.p. 379-388.Selkirk,E.(1984)Phonologyandsyntax.MITPress.Selkirk,E.(1996).Theprosodicstructureoffunctionwords.InJ.L.MorganandK.DeMuth(eds)
Signaltosyntax:Bootstrappingfromspeechtogrammarinearlyacquisition.p.187-214.LawrenceEribaumAss.Inc.
Selkirk,E.(2011)Thesyntax-phonologyinterface.InGoldsmith,J.A.,Riggle,J.andAlan,C.L. (eds.)Thehandbookofphonologicaltheory.Vol.75.JohnWiley&Sons.Starke,M.(2010)Nanosyntax:Ashortprimertoanewapproachtolanguage.Nordlyd36(1). p.1-6.Stepanov, A. (2001) Late adjunction and minimalist phrase structure. Syntax 4(2). p. 94- 125.Svenonius, P. (2004)On theEdge. In:D. Adger, C. de Cat andG. Tsoulas, (eds.)Peripheries. SyntacticEdgesandtheirEffects.Kluwer,Dordrecht.Travis, L. (2010) Inner aspect: The articulation of VP. Vol. 80. Springer Science & Business Media.Truckenbrodt,H.(1999)Ontherelationbetweensyntacticphrasesandphonologicalphrases. LinguisticInquiry30.p.219–255.Uffmann,C.(2011)TheOrganizationofFeatures.InvanOostendorp,M.,Ewen,C.J.,Hume,E. andRiceK.(eds).TheBlackwellCompaniontoPhonology.BlackwellPublishing. BlackwellReferenceOnline. [http://www.blackwellreference.com/subscriber/tocnode.html?id=g9781405184236 _chunk_g978140518423629]Uriagereka, J. (1999) Multiple Spell-out. In S. Epstein and N. Hornstein (eds.) Working Minimalism.Cambridge,Mass.:MITPress.Vaux,B.(2008)Whythephonologicalcomponentmustbeserialandrule-based. InB.Vaux andA.Nevins (eds.)Rules, Constraints, andPhonological Phenomena.Oxford:Oxford UniversityPress.Wiltschko, M., and Dechaine, R.-M. (2010). Interface Syntax. Ms. University of British Columbia.
32
Zwicky,A.M.,andPullum,G.K.(1986)Theprincipleofphonology-freesyntax:introductory remarks. OhioStateUniversityWorkingPapersinLinguistics32.p.63-91.
i Scheer(2013)notesthatthisnon-communicationbetweenthesyntaxandthephonologymayberestrictedtomelodicprimes. ii ForamoredetailedaccountofthephonologicalmotivationforPhonologicalMergerinthecaseofEnglishaffixesseeNewell2016a,b. iiiThattheadjectivalandverbalsemanticsofun-seemtobedifferentbutthisisarguablyduetothesemanticsofverbsandadjectives,ratherthanthesemanticsofun-.un-impliesreversal;of direction of timeflow in the case of verbs, and of positive/negative scale in the case ofadjectives. In the case of nouns un- indicates the opposite (reversal) of reference (yourunbirthdayisadaythatisnotyourbirthday).iv Thatthestem-levelaffiliationof–aloverridestheword-levelaffiliationof–mentisaseparateissuethatwewillignorehere(butseeNewell2016a,b).Thisabilityofanouteraffixtotriggerthe reassignment of stress on an interiorword-level affix is problematic for CBP and RBPaccounts.
![An Introduction to Historical Phonology 1syllabification . fiːt. . fiːl. .fiː.lɪŋ. l-velarisation — fiːɫ — HVB — fɪəɫ — SR [fiːt] [fɪəɫ] [fiːlɪŋ] In Rule-Based](https://static.fdocuments.net/doc/165x107/60b181a922c4ae54e5177073/an-introduction-to-historical-phonology-1-syllabification-fit-fil-fil.jpg)
