UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Syntax, prosody, and metrical structure in Blackfoot Weber, Natalie 2020

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


24-ubc_2020_may_weber_natalie.pdf [ 2.71MB ]
JSON: 24-1.0389893.json
JSON-LD: 24-1.0389893-ld.json
RDF/XML (Pretty): 24-1.0389893-rdf.xml
RDF/JSON: 24-1.0389893-rdf.json
Turtle: 24-1.0389893-turtle.txt
N-Triples: 24-1.0389893-rdf-ntriples.txt
Original Record: 24-1.0389893-source.json
Full Text

Full Text

Syntax, prosody, and metrical structure in BlackfootbyNatalie WeberB.A., Rice University, 2009A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFDoctor of PhilosophyinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Linguistics)The University Of British Columbia(Vancouver)April 2020c© Natalie Weber, 2020The following individuals certify that they have read, and recommend to the Faculty of Graduate andPostdoctoral Studies for acceptance, the dissertation entitled:Syntax, prosody, and metrical structure in Blackfootsubmitted by Natalie Weber in partial fulfillment of the requirements for the degree of Doctor ofPhilosophy in Linguistics.Examining Committee:Douglas Pulleyblank, LinguisticsCo-supervisorRose-Marie Déchaine, LinguisticsCo-supervisorGunnar Ólafur Hansson, LinguisticsSupervisory Committee MemberAnne-Michelle Tessier, LinguisticsUniversity ExaminerStefan Dollinger, EnglishUniversity ExamineriiAbstractThis dissertation investigates the correspondences between syntactic, prosodic, and metrical constituentsin Blackfoot (Algonquian), a polysynthetic language. I propose that the syntax-prosody correspondenceis distinct from the alignment of prosodic and metrical structure. In a parallel constraint-based modelof phonology, this predicts that a language might satisfy isomorphic syntax-prosody correspondenceat the expense of prosodic and metrical alignment, or vice versa. To determine the generalizations inBlackfoot, I gathered data by conducting fieldwork with speakers and consulting published referencematerials. Some arguments in the dissertation are based on original morphological and phonologicalanalyses of Blackfoot stems.For the syntax-prosody correspondence, I hypothesize that each syntactic phase corresponds to a par-ticular prosodic constituent by default. Specifically, the vP phase (the predicate of events), matches to aProsodic Word (PWd) constituent, and the DP and CP phases match to Phonological Phrase (PPh) con-stituents. I model these relationships using a modified version of Match Theory (Selkirk 2011), wheremismatches between syntactic phases and prosodic structure only occur in order to satisfy prosodicwellformedness constraints. For the relation between prosody and metrical structure, I hypothesize thatthe edges of metrical constituents align to different prosodic constituents (prosodic word, phonologicalphrase, or intonational phrase).Regarding structure in Blackfoot, I argue that a constraint which requires sister nodes within theprosodic structure to be of the same type outranks the syntax-prosody MATCH constraints. This forceseach DP argument and also the remainder of the CP (e.g. the verbal complex) to be matched to a PPhconstituent. The vP phase and every higher vP projection corresponds to a PWd constituent, which isdistinct from the PPh. I argue that the metrical constituents in Blackfoot align to PPh edges, and thatsyllables frequently span PWd edges. This is a predicted outcome, given that the MATCH and ALIGNconstraints are violable. The model I propose accounts for the correspondence relations in Blackfoot,and leads to a typology of predicted language types, with implications for extending Match Theory toaccount for polysynthetic languages.iiiLay SummaryThis dissertation investigates the structure of words in Blackfoot in terms of the (1) grammar (syntax),(2) speech sounds, and (3) stress assignment on syllables. I focus on verbs elicited in original fieldworkand taken from published resources, with several results. First, verbs are grammatically complex, andcan convey the same information as some entire sentences do in a language like English. Second, soundsexhibit unique patterns and processes within a smaller unit which is contained in the larger sentence-sized word. Third, each verb contains a single syllable with primary stress, much like a single word ina language like English. I conclude that the word in Blackfoot emerges from the combination of thesethree properties. I propose a model of the correspondence relations between representations of grammar,speech sounds, and stress assignment in Blackfoot. This research contributes to our understanding oflinguistic structure in words of all languages.ivPrefaceThis dissertation is an original intellectual product of the author, Natalie Weber. The fieldwork reportedin Chapters 2–4 was covered by UBC Ethics Certificate numbers H07-01365 and H16-00986. The mapin Figure 1.2 was created by Eric Leinberger and used with permission. The map in Figure 1.3 wascreated by Kevin McManigal and used with permission.vTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiAbbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiiSymbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Proposal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.2.1 Spell Out and the interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.2.2 Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.2.3 Phonology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91.2.4 Correspondence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131.3 Language overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231.4 Data and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271.4.1 Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271.4.2 Presentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271.4.3 Transcription . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29vi1.5 Outline of the dissertation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312 Blackfoot phonology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.1 Consonants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.1.1 Phonemic inventory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.1.2 Allophony . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372.2 Vowels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412.2.1 Phonemic inventory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412.2.2 Allophony . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442.3 Suprasegmentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 452.4 Syllable structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472.4.1 Open syllables: contrastive vowel length . . . . . . . . . . . . . . . . . . . . 482.4.2 Closed syllables: vowel length neutralization . . . . . . . . . . . . . . . . . . 542.4.3 Syllabic fricatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 712.4.4 Degenerate syllables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 762.5 Chapter summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 843 Correspondence of CP phrases and phonological phrases . . . . . . . . . . . . . . . . . 863.1 Syntax of the CP phrase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 883.1.1 The verbal complex contains a C0 . . . . . . . . . . . . . . . . . . . . . . . . 893.1.2 The verbal complex has the distribution of a CP . . . . . . . . . . . . . . . . . 1043.1.3 Syntax of nominal expressions . . . . . . . . . . . . . . . . . . . . . . . . . . 1103.1.4 Interim summary: CP syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . 1133.2 Phonology of the PPh constituent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1143.2.1 Glides are prohibited at the left edge of the PPh . . . . . . . . . . . . . . . . . 1143.2.2 Primary stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1563.2.3 Interim summary: PPh phonology . . . . . . . . . . . . . . . . . . . . . . . . 1773.3 Mapping the CP to prosodic structure . . . . . . . . . . . . . . . . . . . . . . . . . . 1784 Correspondence of vP/VP phrases and prosodic words . . . . . . . . . . . . . . . . . . . 1854.1 Syntax of the vP/VP phrase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1884.1.1 Theoretical assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1884.1.2 Stems with XP-adjoined roots only . . . . . . . . . . . . . . . . . . . . . . . 1944.1.3 Stems with X0-adjoined roots . . . . . . . . . . . . . . . . . . . . . . . . . . 2134.1.4 Roots have non-uniform morphosyntax . . . . . . . . . . . . . . . . . . . . . 2204.1.5 The first phase is a predicate of events . . . . . . . . . . . . . . . . . . . . . . 2304.1.6 Comparison with Algonquian templatic stem . . . . . . . . . . . . . . . . . . 231vii4.1.7 Interim summary: vP/VP syntax . . . . . . . . . . . . . . . . . . . . . . . . . 2324.2 Phonology of the PWd constituent . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2334.2.1 Low√ROOT prosodification: PWd-internal epenthesis . . . . . . . . . . . . . 2344.2.2 High√ROOT prosodification: PWd-initial allomorphy and epenthesis . . . . . 2604.2.3 Suffixal domain: outside the PWd . . . . . . . . . . . . . . . . . . . . . . . . 2814.2.4 Interim summary: PWd phonology . . . . . . . . . . . . . . . . . . . . . . . 2834.3 Mapping the vP/VP to prosodic structure . . . . . . . . . . . . . . . . . . . . . . . . . 2855 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2885.1 PWd ̸= PPh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2885.1.1 Left edge restrictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2895.1.2 Domain-restricted generalizations . . . . . . . . . . . . . . . . . . . . . . . . 2925.1.3 Summary and future research . . . . . . . . . . . . . . . . . . . . . . . . . . 2945.2 Prosodic and metrical misalignment . . . . . . . . . . . . . . . . . . . . . . . . . . . 2956 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303A Minimal pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326A.1 Consonants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326A.1.1 Contrast based on place and manner . . . . . . . . . . . . . . . . . . . . . . . 326A.1.2 Contrast based on length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332A.1.3 Assibilants and contrast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337A.2 Vowels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338A.2.1 Vowel quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338A.2.2 Vowel quantity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339A.2.3 Non-derived lax mid vowels . . . . . . . . . . . . . . . . . . . . . . . . . . . 342B X-adjoined root alternations: supplemental evidence . . . . . . . . . . . . . . . . . . . . 346B.1 Roots beginning with [i1] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346B.2 Roots beginning with [i2] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347C Neutralization in the independent clause type . . . . . . . . . . . . . . . . . . . . . . . . 349C.1 Roots beginning with long vowels . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349C.2 Roots beginning with short vowels . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350C.3 Roots beginning with /j/ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352C.4 Roots beginning with {m, n} ∼ [j] alternations . . . . . . . . . . . . . . . . . . . . . 354viiiC.5 Roots beginning with [ij] ∼ [j] alternations . . . . . . . . . . . . . . . . . . . . . . . 355C.6 Roots beginning with obstruents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356C.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357ixList of TablesTable 2.1 Blackfoot phonemic consonant inventory . . . . . . . . . . . . . . . . . . . . . . . 33Table 2.2 Blackfoot phonemic vowel inventory . . . . . . . . . . . . . . . . . . . . . . . . . 41Table 2.3 Positional realization of prominence . . . . . . . . . . . . . . . . . . . . . . . . . 47Table 2.4 Distribution of consonants by position . . . . . . . . . . . . . . . . . . . . . . . . 77Table 3.1 Person proclitics in Blackfoot clause types . . . . . . . . . . . . . . . . . . . . . . 90Table 3.2 Dependencies between C0 and IP features . . . . . . . . . . . . . . . . . . . . . . 92Table 3.3 Exponents of C0 in intransitive independent clauses (Frantz 2009) . . . . . . . . . . 93Table 3.4 Partial neutralization of Agree C0 in independent clauses (after Bliss 2013: 234) . . 97Table 3.5 Exponents of C0 in intransitive unreal clauses (Frantz 2009) . . . . . . . . . . . . . 98Table 3.6 Exponents of C0 in conjunctive clause types (Frantz 2009) . . . . . . . . . . . . . . 100Table 3.7 Exponents of C0 in subjunctive clause types (Frantz 2009) . . . . . . . . . . . . . . 100Table 3.8 Exponents of C0 in imperative clause types (Frantz 2009) . . . . . . . . . . . . . . 103Table 3.9 Dependencies between C0 and IP features (repeated) . . . . . . . . . . . . . . . . . 104Table 3.10 Dependencies between C0 and embeddedness . . . . . . . . . . . . . . . . . . . . 105Table 3.11 Dependencies between C0 and embeddedness (repeated) . . . . . . . . . . . . . . . 109Table 3.12 Relation of Blackfoot C0 to embedding . . . . . . . . . . . . . . . . . . . . . . . . 109Table 3.13 Indexicality vs. embedding in Plains Cree (after Cook 2008: 3) . . . . . . . . . . . 110Table 3.14 Internal syntax of complements to transitive verbs (Weber and Matthewson 2017) . 112Table 3.15 External syntax of complements to transitive verbs (based on Bliss 2013) . . . . . . 112Table 3.16 Semantic properties of complements to transitive verbs (Weber and Matthewson 2017)112Table 3.17 Summary of CP properties of the verbal complex . . . . . . . . . . . . . . . . . . . 113Table 3.18 Distribution of onsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117Table 3.19 Allomorphy of roots with initial heavy syllables . . . . . . . . . . . . . . . . . . . 127Table 3.20 Vowel hiatus resolution strategies: V2 is [-high] and parsed to a heavy syllable . . . 137Table 3.21 Co-occurrence restrictions between root-initial /j/ and a following vowel; first sylla-ble must be heavy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143xTable 4.1 Diagnostic properties for the Merge strategies of roots . . . . . . . . . . . . . . . . 194Table 4.2 Pairs of AI and II stative stems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195Table 4.3 Transitive paradigms with XP-adjoined roots only by v0 . . . . . . . . . . . . . . . 199Table 4.4 Theme suffixes (Frantz 2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209Table 4.5 Transitive paradigms with XP-adjoined roots only by TI V0 . . . . . . . . . . . . . 210Table 4.6 Transitive paradigms with XP-adjoined roots only by AI V0 . . . . . . . . . . . . . 211Table 4.7 Transitive paradigms with “instrumental” X0-adjoined roots . . . . . . . . . . . . . 216Table 4.8 Diagnostic properties for√ROOT vs. head . . . . . . . . . . . . . . . . . . . . . . 220Table 4.9 Realization of vowel sequences in Blackfoot (Elfner 2006b: 97) . . . . . . . . . . . 238Table 4.10 Realization of vowel sequences in Blackfoot (revised) . . . . . . . . . . . . . . . . 239Table 4.11 Underlying short vowels within the PWd . . . . . . . . . . . . . . . . . . . . . . . 252Table 4.12 States built on classificatory medials (subset) . . . . . . . . . . . . . . . . . . . . . 257Table 4.13 Person proclitics in Blackfoot clause types (repeated) . . . . . . . . . . . . . . . . 262Table 4.14 Segments allowed at the left edge of roots in PPh-initial and PPh-medial positions . 272Table 5.1 Phonological generalizations of the PPh and PWd constituents . . . . . . . . . . . 289Table C.1 Initial vowel mutation for stems with initial long vowels . . . . . . . . . . . . . . . 351Table C.2 Initial vowel mutation for stems with initial short vowels . . . . . . . . . . . . . . . 352Table C.3 Initial vowel mutation for stems with initial /j/ . . . . . . . . . . . . . . . . . . . . 353Table C.4 Initial vowel mutation for stems with initial {m, n} ∼ /j/ . . . . . . . . . . . . . . 355Table C.5 Initial vowel mutation for stems with initial [ij] ∼ /j/ . . . . . . . . . . . . . . . . 356Table C.6 Initial vowel mutation for stems with initial obstruents . . . . . . . . . . . . . . . . 357xiList of FiguresFigure 1.1 Model of grammar within the Minimalist Program . . . . . . . . . . . . . . . . . 4Figure 1.2 Location of Blackfoot in relation to the Algonquian and Algic languages. Map byEric Leinberger. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25Figure 1.3 Locations of Blackfoot reserves. Map by Kevin McManigal. . . . . . . . . . . . . 26xiiAbbreviations1 first person21 first person inclusive2 second person3 third person3′ third person obviative0 third person inanimateX unspecified subjectAI animate intransitiveAN animateAPPL applicativeASSOC associativeBERRY small, round, solid objectBP body part markerCAUS causativeCMD command clauseCNJ conjunctive orderCOM comitativeCONJ conjunctionCORD one-dimensional flexible materialDEG degree markerDEM demonstrativeDEP dependent clauseEVID evidentialEXT medial extensionxiiiFUT futureGET acquire, gather, make (light verb)IC initial changeII inanimate intransitiveIMP imperativeIN inanimateIND independent orderINS instrumentalINV inverseINVS invisibleIPFV imperfectiveLOC locativeMANNER mannerMEANS means/instrumentalMETAL metal, bladeMOOD mood suffixNEG negativeNMLZ nominalizerNONAFF nonaffirmativeOBJ objectOBV obviativePL pluralPOSS possessivePRF perfectPRX proximatePST pastQ question particleREFL reflexiveSAP speech act participantSBJ subjunctive orderxivSG singularSHEET two-dimensional flexible materialSTICK stick, woodSUB subjectSUPPORT one-dimensional rigid materialTA transitive animateTI transitive inanimateTI1 transitive inanimate (Class 1)TI2 transitive inanimate (Class 2)TI3 transitive inanimate (Class 3)TIME temporal linkerUNR unreal orderWARD direction/locative nominalWHEN temporal anchor (‘when’)xvSymbols– affix separator= clitic separator/ head-adjoined root separator\ non-concatenative morphology( ) optional% dialectal variant* unattested or unacceptable! fatal constraint violation/ actual pronunciation but non-optimal candidate optimal candidate∗ constraint violation markerv light v0 headv*P first verbal phaseIPH intonational phrasePPH phonological phrasePWD prosodic wordC consonantG glideN nasalV syllable nucleus (vowel or moraic [s])xviAcknowledgmentsCompleting this dissertation is the hardest thing I have ever done in my life. No academic project iscompleted in a vacuum, and in my case I have many people to thank for helping me get to this point.If we take the beginning of the project to be the beginning of my journey as a linguist, then there arepeople at (at least!) four institutions to thank.At Rice, I have Claire Bowern and Laura Robinson to thank for introducing me to fieldwork anddocumentation in linguistics—I was hooked from the beginning. Katherine Crosswhite was my veryfirst phonetics and phonology teacher, and probably was the first to set me on this path. Two years outof undergrad when I was considering applying to graduate school, Robert Englebretson suggested I takea look at the program at UBC, and it’s lucky for me that he did.At UBC, I worked quite closely with Gunnar Ólafur Hansson for my first several years of gradschool, as a student, a TA, and an RA. Although he did not end up supervising me on this project, hismark is on my work quite strongly if you know how to see it and on my thoughts around phonology ingeneral. My advisor, Rose-Marie Déchaine, rather comfortably took me on as a student, even though Iknew little syntax at the time, and guided me through the maze of Algonquian literature. Her mentorshipwent far beyond theory, and her lessons on how to student and how to survive academia have beeninvaluable in more ways than I can express. My co-advisor, Doug Pulleyblank, joined my committeelater and never ceased to question the hidden assumptions in all of my work. He also sent commentsback on every draft within days—an act that now as a professor I see clearly as true mentorly love.At Lethbridge, Inge Genee really served as an advisor away from home for the eight months I spentthere as an RA and a Sessional Instructor. She was invaluable with helping to connect me with Blackfootspeakers, making sure I had enough paid work to keep a roof over my head, and in offering advice on allmanner of interpersonal issues. Don Frantz was truly inspirational to meet! How wonderful to be ableto share an office and to chat about Blackfoot together. Judith Lapadat and Rob Sampson let me house-and petsit for them on a second trip to Lethbridge. Graduate research often hangs by a financial thread,and if they had not given me this opportunity I don’t think I could have returned. I’m very grateful.xviiAt Yale, I must thank my excellent colleagues here for being amazing, welcoming, helpful, and greatto talk shop with. The graduate and undergraduate students have also been amazing. I look forward toeach mentoring meeting every week.Beyond the ivory tower, there were many more people who supported me. First and foremost, thisdissertation would not have been possible without the support from many Blackfoot speakers, espe-cially Totsinámm (Beatrice Bullshields), Ááhsaikamo’sáákii (Natalie Creighton), and Rod Scout. Mycolleagues of Algonquian linguistics have been very supportive as well. I especially have Sarah MurrayandMonicaMacaulay to thank for mentoring me as a junior colleague. More generally, Richard Rhodes,David Pentland, Ives Goddard, Conor Quinn, and Taylor Miller all come to mind as Algonquianists whohave gone out of their way to support me.My cohort and housemates have been invaluable, especially Ella Fund-Reznicek, Andrei Anghe-lescu, and Heather Burge. To Vica, Nico, Benny, Anna and Yarrow, and many others. Finally, tomultitudes of dancers in Houston, San Francisco, Vancouver, Lethbridge, Canberra, and elsewhere Isay: thanks for the dance!xviiiChapter 1Introduction1.1 OverviewThis dissertation is an empirical and theoretical study of the relationship between syntactic structure andphonological form (prosodic and metrical structure) in Blackfoot, an Algonquian language spoken insouthern Alberta, Canada and northern Montana, USA. As such, it addresses questions at the interfacebetween syntax and phonology. In particular, this is a study of structural correspondences across severalgrammatical domains. To a lesser extent, I also look at the phonological exponence of morphemes(allomorphy) and how this is conditioned by prosodic structure. Blackfoot is an ideal case for thestudy of “word” level correspondences between syntax, prosody, and metrical constituents, because itis a polysynthetic and strongly head-marking language. This means there is likely to be syntactic andphonological structure “below the word”.Consider the biclausal examples in (1) and (2). These illustrate the puzzle of Blackfoot correspon-dence, which can be informally stated as follows. In terms of syntax, each verbal complex has the syntaxroughly of a CP phrase. In terms of prosody, each verbal complex exhibits phonological generalizationssimilar to prosodic words in other languages. For example, syllabic prominence (expressed via a pitchpeak, and bolded below) is obligatory over the entire verbal complex but not over any smaller portion,and each verbal complex is a single domain of syllabification. The question is how to model the factthat the verbal complex is both a syntactic CP phrase and a prosodic word.(1) [ [ejP.kI´s.ts":.sa:.ki:.ni.ki ]CPAi’kísstsssaakiinikia–iksist–[ssi–aki]–in–ik–iIPFV–finish–[wipe–AI]–SBJ–MOOD–DEPni.>tsI.tO´:.tois.si:.s>ts ]CPnitsitáótoissiistsnit–it–a–oto–[ssi–iist/i]–(hp)1–LOC–IPFV–go.to.do–[wipe–by.water/AI]–(IND)‘When I finish doing my dishes, I go and take a shower.’ (BB; 2012-06-28, Daily Routine)1(2) CP[ i.ta.n´i.s>tsi.>ksi.ms":.ta.jaItanístsiksimsstayait–anist–iksim–[sst–aa]–yi=aawaLOC–MANNER–secret–[wish–AI]–PL=PRX.PLCP[o.max.ks":.>tso´:.toxw.pUm.max.saˆ:omaahksstsóótohpommaxsáao–m–aahk–sstsi–oto–[ohpomm–a]–hsi=aawa3–3–might–town–go.to.do–[buy–AI]–CNJ=PRX.PLsa´.>tsa´:.pi.njo.w2nsátsáápiniowanpisatsaapiniowancandy]CP ]CP‘They decided to go to town to buy some candy.’ (BB; 2013-02-13, Old Woman in the Cold)I propose that linguistic utterances are arranged into three types of representational structure, (3):syntactic, prosodic, and metrical. Each structure is a hierarchical arrangement of a universal set ofcategories which is unique to that type of structure.(3) STRUCTURE CATEGORIESa. Syntax Complementizer Phrase (CP), . . . verb phrase (vP/VP), . . .b. Prosody Intonational Phrase (IPh), Phonological Phrase (PPh), Prosodic Word (PWd)c. Metrical Foot, Syllable (), Mora ()I propose there is a direct correspondence relationship between prosodic and syntactic structure andbetween prosodic and metrical structure, but that there is no direct relationship between syntactic andmetrical structure. First, there is a universal ‘default’ correspondence between syntactic phases and auniversal set of prosodic constituents. Specifically, the first phase (v*P) corresponds to a Prosodic Word(PWd) constituent, while the second phase (CP) corresponds by default to a Phonological Phrase (PPh).Second, I propose that the edges of prosodic and metrical constituents align by default. The differentalignment strategies are encoded via ranked, violable constraints which interact with language-specificconstraint rankings in different ways.In this dissertation I develop independent diagnostics to determine syntactic, prosodic, and metri-cal constituents within the Blackfoot verbal complex. I then show how the particular correspondencerelationships in Blackfoot can be accounted for with the model that I propose in the next section. InBlackfoot, metrical constituents like feet, syllables, and moras must align to PPh edges but are not re-quired to align to PWd boundaries, resulting in misalignments within the PPh between syllable edgesand PWd boundaries. A schematic representation of these correspondence relations is in (4).2(4) SYNTAX-PROSODY-METRICAL CORRESPONDENCES IN BLACKFOOTSyntaxCPv*PProsodyPPhPWdMetricalFt, , Any model of syntactic and prosodic correspondence must also account for mismatches betweenthe three structures. There are two points of flexibility in my model. The first point of flexibility isthat I conceive of the syntax-to-prosody correspondence relations as a family of ranked but violableconstraints within the broader phonological grammar. Specifically, I modify the MATCH constraints inMatch Theory (Selkirk 2009, 2011) to refer to syntactic phases. Prosodic structure differs from syntacticstructure when markedness constraints on prosodic wellformedness dominate the MATCH constraints.A full factorial typology of the MATCH and prosodic markedness constraints would make predictionsabout possible and impossible languages. The second point of flexibility are the Alignment constraints(McCarthy and Prince 1993a) between prosodic and metrical structures. These constraints can align anyone of the prosodic and metrical constituents together, which also make predictions about possible andimpossible languages when ranked freely.In the following section I discuss my assumptions about linguistic architecture and the relationbetween the syntactic and phonological components of grammar. I conceive of the syntax-prosodyinterface as a translation from syntactic structure to prosodic structure. Only certain types of syntacticphrases have some kind of correspondent in prosodic structure, and prosodic structure itself is subjectto prosodic wellformedness constraints that syntactic structure is not. These facts account for whysome aspects of prosodic structure may track syntactic constituency quite closely while others are non-isomorphic.A final note before laying out the proposal: throughout this dissertation, I use the words ‘prosodic’and ‘prosody’ to refer to phonological domains which correspond to syntactic units, such as the ProsodicWord (PWd), Phonological Phrase (PPh), or Intonational Phrase (IPh), and their hierarchical structure.I do not use ‘prosodic’ or ‘prosody’ to refer to suprasegmental properties such as pitch, stress, andduration. However, suprasegmental properties and prosody do bear some relation in the sense that pitchor stress assignment may be delimited by one of the prosodic constituents.31.2 ProposalIn this section I discuss the elements needed for my proposal of the syntax-prosody interface and theconnection between prosodic domains and metrical structure. Some of these elements simply clarifymy assumptions about syntax, prosody, and metrical constituents. Others are novel contributions whichare intrinsic to the model. I discuss my views of the syntax-prosody interface in Section 1.2.1, syntaxin Section 1.2.2, phonology (prosodic and metrical structure) in Section 1.2.3, and the correspondenceretlationships between syntax, prosody, and metrical structure in Section Spell Out and the interfaceI adopt a “Y-model” of the cognitive system of language, whereby the syntactic derivation is the inputto the phonological and semantic components of grammar. This model is laid out in more detail in theMinimalist Program (Chomsky 1993, 1995b, 2000), as well as earlier works. The path of derivationis schematized by Figure 1.1. At a certain point during the derivation called “Spell Out”, the syntac-tic derivation is submitted to Phonological Form (PF), an interface system which interprets the inputsyntactic representation into a phonological representation.LexiconSpell-outPF LFFigure 1.1: Model of grammar within the Minimalist ProgramI assume that a mix of syntactic and phonological elements are simultaneously visible at PF. Fol-lowing Selkirk (2011), I assume that PF is responsible for at least three aspects of the interface betweensyntax and phonology: (1) the correspondence of syntactic and prosodic representations, (2) phono-logical exponence of abstract syntactic feature bundles and lexical items, and (3) the linearization ofelements. Because syntax and phonology are copresent at PF, phonological constraints are able to influ-ence each of these three interface issues.Regarding representational correspondence, my proposal is couched within a modified version ofMatch Theory (Selkirk 2011), which assumes that Spell Out to PF occurs only once during the syn-tactic derivation. However, the syntactic elements which are relative to structural correspondences aresyntactic phases, of which there are several in any given clause. I argue that particular syntactic phasescorrespond to particular prosodic categories, which means that syntactic phases and prosodic categories4must both visible to PF. (See Section 1.2.4 for details on the correspondence relations between syntacticand prosodic structure.) I take the first phase (v*P) to be a predicate of events, which is the vP/VP phraseat which point the predicate of events is complete. (For a similar interpretation of the first phase, seeRamchand 2008. Other research which equates the vP to a predicate of events include Davidson 1967;Higginbotham 2000; Kratzer 1996; Pustejovsky 1995.) In addition to v*P, I also assume that a CP is aphase, following Chomsky (2001).Regarding phonological exponence, I assume that the underlying phonological form of each lexi-cal item is determined relatively “late”, once agreement has taken place. This can be modeled in aninterpretive theory of morphology (e.g. Anderson 1992; Halle and Marantz 1993) – for example, aspost-syntactic context-free or context-sensitive Vocabulary Insertion in Distributed Morphology (Halleand Marantz 1993), with the options given in (5). In (a), the terminal element X has a single phono-logical form /A/ which occurs in all contexts. In (b), the terminal element X has one form /B/ whichoccurs in the presence of a feature [F] on X, and a second form /A/ which occurs elsewhere. In (c), theterminal element X has one form /C/ which occurs in the presence of a feature [G] on a locally adjacenthead Y0, and a second form /A/ which occurs elsewhere. Allomorphs with more specific contexts takeprecedence over forms with less specific contexts. In this way, Vocabulary Insertion can account forsyntactically-conditioned allomorphy in phonological exponence.(5) VOCABULARY INSERTIONa. /A/↔ X0 context-freeb. /B/↔ X0 / [F] conditioned by a feature [F] on X0/A/↔ X0 defaultc. /C/↔ X0 / Y0[G] conditioned by a feature [G] on adjacent Y0/A/↔ X0In my model, allomorphy can be conditioned by syntactic or phonological contexts, which followsin a model where syntactic features and phases as well as prosodic boundaries are both visible at PF.Observationally, there is a divide between how allomorphy is conditioned for an XP-adjoined√ROOTversus a syntactic X0 head or an X-adjoined√ROOT in Blackfoot. (For a description of root syntacti-cization, see Section XP-adjoined√ROOT allomorphy is phonologically optimizing (Mascaró2007) and is often conditioned by phonological restrictions at prosodic boundaries. X-adjoined root andX0 head allomorphy is conditioned by syntactic features, as in (5). I do not have an explanation for whythis divide occurs, and I leave a full discussion of the typology of conditioning factors on phonologicalexponence for future research.This thesis does not directly address issues of linearization. However, since syntactic and phonolog-ical elements are both available at PF, my model predicts that linearization could be affected by syntactic5or phonological properties. Indeed, linearization has been argued to follow purely from syntactic prece-dence (see Embick 2010 for one formal model) as well as phonological properties, such as minimalityconstraints, interaction with stress or tone, or syllable shape (see Bennett, Elfner and McCloskey 2016;Elfner 2012; Harizanov 2014; McCarthy and Prince 1993a; Werle 2009, among others).1.2.2 SyntaxIn this section I lay out my assumptions about the mechanics of phrase structure and the syntacticizationof roots. Both elements are necessary for the proposal of syntax-prosody correspondence I discussbelow. Phrase structureFollowing Bare Phrase Structure (Chomsky 1995a), I take an explicitly derivational approach to syn-tactic construction. A derivation is simply the pairwise combination of items drawn from the lexicon tobuild up a constituent structure using the syntactic operations Select, Merge, and Move. I further assumethat phrases and words are constructed via ‘syntax all the way down’, such that stems and words arebuilt via the same syntactic operations as phrases. As such, I explicitly reject the Lexicalist hypothesis(Chomsky 1970; Halle 1973) and adopt a constructionist view of syntax. A constructionist approachis used in a variety of other frameworks, including Distributed Morphology (Embick and Noyer 2007;Hale and Keyser 1993; Halle and Marantz 1993; Harley and Noyer 1999), Borer’s Exoskeletal Model(Borer 2013), and Nanosyntax (Starke 2009, 2011).Bare Phrase Structure eliminates bar levels as formal elements, and syntactic features instead projectdirectly. In (6) below, x is a complex category which first Merges with y and then subsequently Mergeswith z. Unlike in X-bar theory (Chomsky 1970; Jackendoff 1977), rules cannot refer specifically to x0,x′, or x′′, because these are not formal elements. However, certain relational properties of categories canbe determined by structure. For any complex category, X, an X which does not dominate another x is theminimal projection of that category, XMIN, and an x which is not dominated by another x is the maximalprojection of that category, XMAX. The tree in (7) includes the same structure as (6), with maximal andminimal projections explicitly labelled.6(6) xy xz x(7) xMAXy xz xMINMaximal and minimal projections are not syntactic primitives, but structurally determined, whichmeans that in a derivational framework these properties may change after each step of the derivation. Forexample, in the first step of the derivation for the tree above, xMerges with z and projects x, (8a). At thispoint in the derivation, the projection which immediately dominates z and x is the maximal projection.In the second step of the derivation, the complex category x Merges with y and projects x, (8b). At thispoint in the derivation, the projection which immediately dominates z and x is no longer the maximalprojection. Nevertheless, at an earlier step in the derivation it was maximal, and this fact will be relevantfor root syntacticization, as I discuss in Section a. STEP 1: MERGE {z,x}xMAXz xMINb. STEP 2: MERGE {y,x}xMAXy xz xMINAlthough category labels like X0, X′, and XP are not formal elements, it is common practice touse these as informal labels in trees, and this thesis is no exception. I use X0 to designate minimalprojections, and XP for any higher projection of a complex category, as in (9) and (10). In (9) a headX0 merges with multiple a-categorical√ROOT adjuncts, represented by√ROOT1 and√ROOT2. (Seethe next section for an explanation of root syntacticization in my proposal.) In (10) a head X0 mergeswith two phrasal elements, represented by YP and ZP. Following Kayne (1994), I take the basic struc-tural contrast to be complement versus non-complement; consequently, there is no structural differencebetween specifiers and adjuncts, and the only difference is the order in which they Merge. This is alsoin conformity with the Inclusiveness Condition of Bare Phrase Structure (Chomsky 2007, 2008, 2013),which states there are no bar level distinctions in syntactic representations.7(9)√ROOT MERGES WITH HEAD X0XP√ROOT2 XP√ROOT1 X0(10) PHRASE MERGES WITH HEAD X0XPZP XPYP X01.2.2.2 Root syntacticizationRecent analyses treat roots as category-neutral formatives introduced uniformly as complements to acategorizing head. In one variant, the root is sister to a categorizing lexical head, such as a verbalizingv, (11a); this is found with Distributed Morphology (Marantz 1997; Siddiqi 2009) and AsymmetricMorphology (Di Sciullo 2005). In another variant, the root is sister to a functional head which serves tocategorize the root, as in (11b); this is found with Exoskeletal Syntax (Borer 2013).(11) a. [v [√ROOT ] ]b. [F [√ROOT ] ]The claim that roots are invariably complements is problematic, both theoretically and empirically.The theoretical problem is that if a√ROOT is a syntactic atom, it should be able to Merge as a head,complement, or adjunct. Despite this, the theories discussed above stipulate that a√ROOT must mergein one particular way—as sister to a syntactic head. There is no theory-internal reason for this to bethe case. The empirical problem is that roots in Blackfoot do not have a uniform morphosyntax, as ispredicted by the syntactic accounts in (11), where all√ROOTs merge in the same fashion. As I discussin Section 4.1, many Blackfoot roots have the distribution of phrasal adjuncts, while others are restrictedto verbal contexts and co-occur only with a verbalizing head.To account for these distinct distributions, I adopt a version of the proposal in Déchaine and Weber(2018). I propose that a-categorical roots are uniformly adjuncts, but that roots differ in the type ofsyntactic structure they adjoin to: some roots only adjoin to XMAX (where XMAX is calculated w.r.t. thesyntactic structure at the point in the derivation where the root merges), while other roots only adjointo XMIN. Following the informal labelling conventions I discussed in Section, a√ROOT either(i) adjoins to an XP phrase, (12a) or (ii) adjoins to an X0 head, (12b). I discuss the implications of thisproposal more fully in Section 4.1.8(12) ROOT SYNTACTICIZATION STRATEGIESa.√ROOT ADJOINS TO XPXP√ROOT XPb.√ROOT ADJOINS TO X0XP√ROOT X0This proposal solves the theoretical problem discussed above, because there is no external stipulationthat a√ROOT can only merge in one way. Although a√ROOT in this theory is always introduced viaadjunction, one hypothesis is that this restriction arises via theory-internal reasons. For example, DeBelder (2011) suggests that a√ROOT cannot project because it lacks syntactic features; this wouldexplain why a√ROOT only occurs as an adjunct. The proposal also solves the empirical problemdiscussed above, because the two different classes of Blackfoot√ROOTs emerge from the two differentMerge sites. This type of analysis of roots as a bare elements which does not need to be categorized onfirst Merge differs from most recent analyses, but is not without precedent. More recent developments inDistributed Morphology acknowledge the modificational nature of roots (Marantz 2013), and De Belder(2017) specifically argues that the non-head element of primary compounds in Dutch is a bare√ROOT.1.2.3 PhonologyResearch on the prosody-syntax interface over the past 40 years has shown that phonological domainsare independent from, but closely related to, syntactic structure. Even early observations indicatedthat the domains of phonological generalizations correspond systematically to syntactic constituents(Chomsky and Halle 1968; McCawley 1968; Selkirk 1974). However, these domains are often non-isomorphic to syntactic constituents and can be influenced by language-specific phonological properties,including rate of speech, weight/size of constituents, or stress or tonal properties (cf. Nespor and Vogel2007; Selkirk 1986, 2011). Because of this, I adopt a version of the Prosodic Hierarchy Theory (PHT),whereby phonological generalizations apply to a hierarchical prosodic structure which is independentfrom syntactic structure.In many instantiations of the Prosodic Hierarchy Theory, such as that in Nespor and Vogel (2007),the prosodic hierarchy is non-uniform. Itô and Mester (2012) call the higher levels “interface cate-gories”, which are extrinsically defined by their relation to syntax, while the lower levels are “rhythmiccategories”, which are intrinsically defined by sonority-related phonetic factors and rhythmic stress.Although the division is purely conceptual in most instantiations of the Prosodic Hierarchy Theory,Inkelas (1990, 1993) explicitly treated the Prosodic Hierarchy and the Metrical Hierarchy as two sepa-9rate hierarchies. This idea was taken up by much subsequent work (Branigan, Brittain and Dyck 2005;Downing 1999; Dyck 2009; Inkelas 1990; Itô and Mester 2012). I also adopt this approach. I discussmy assumptions about prosodic structure in Section and metrical structure in Section Prosodic structureProsodic Hierarchy Theory (PHT) was first developed in Selkirk (1978, 1980, 1981b) and further de-veloped by many researchers (a partial list: Beckman and Pierrehumbert 1986; Downing 1999; Hayes1989b, 1995; Hyman 1985; Inkelas 1990; Itô and Mester 2003, 2009a,b, 2012; Ladd 2008; Nespor andVogel 1982, 1983, 2007; Pierrehumbert and Beckman 1988; Selkirk 1984, 1996, 2009, 2011). There-fore, PHT itself has many different instantiations, but all of them assume there is a finite set of orderedprosodic categories, which are the domains for sets of phonological generalizations.The prosodic categories which I adopt are shown in (13). I follow the majority of researchers whoassume that this set of prosodic categories is universal. Not all researchers share this view; some arguethat not all levels of the prosodic hierarchy are instantiated in every language (Green 1997; Jun 2005;Labrune 2012), or that prosodic constituents are emergent (Schiering, Bickel and Hildebrandt 2010).(13) PROSODIC CATEGORIESIPhPPhPWdintonational phrasephonological phraseprosodic wordThe exact number of categories within the prosodic hierarchy differs among researchers. Manyworks also assume an Utterance (Utt) category above the IPh (Nespor and Vogel 2007; Selkirk 1978,1980, 1986; and subsequent work), but recent work has argued that the Utt can be viewed as the maximalIPh (Itô and Mester 2012; Kawahara and Shinya 2008, later incorporated into Match Theory; Selkirk2011). Still others have argued to expand the number of ‘phrasal’ categories (Nespor and Vogel 2007;Pierrehumbert and Beckman 1988; Poser 1984; Vogel 2009), or to add prosodic categories below theword level (Downing 1999; Inkelas 1990). However, these relatively minor differences do not affect theunderlying core assumption that there is a finite set of distinct prosodic categories.Phonological generalizations (e.g. phonotactic restrictions, segmental rules, suprasegmental rules)take a particular prosodic category as their domain of application. Said another way, phonologicalgeneralizations can diagnose prosodic constituents within a hierarchical structure structure. Nespor and10Vogel (2007: 58ff ) discuss four criteria which can be used to motivate phonological constituents. Astring is considered a constituent in phonology if:(14) PROPERTIES WHICH MOTIVATE PHONOLOGICAL CONSTITUENTSa. there are rules of the grammar that need to refer to it in their formulation, orb. there are rules that have precisely that string as their domain of application, orc. if the string is the domain of phonotactic restriction, ord. if the string must be posited as a domain which delimits the domains of stress patterns, inorder to explain the relative prominence relations (stress) within a given string.(Nespor and Vogel 2007: 58ff )In other words, if a phonological generalization cannot be stated without reference to a phonologicaldomain then the phonological generalization is evidence for a phonological constituent. The exact typeof phonological category cannot be determined by purely phonological factors, but can be extrinsicallydefined by its correspondence to syntax. I turn to these correspondence relationships in Section 1.2.4.Finally, there are inviolable constraints on hierarchical constituent structures which contain the or-dered categories in (13). Early proposals of PHT followed the Strict Layer Hypothesis (SLH; Beckmanand Pierrehumbert 1986; Nespor and Vogel 2007; Pierrehumbert and Beckman 1988; Selkirk 1981b,1984, 1986), which assumed that prosodic structures were strictly organized and had different propertiesthan syntactic structure. This formulation has since been split into a set of more primitive components(e.g. Inkelas 1989; Itô and Mester 2003; Nespor and Vogel 2007; Selkirk 1996). I take the dominancerelations in (13) to represent containment relations (see Pak 2008 for a good discussion of this point), asrequired by PROPER BRACKETING in (15). Proper bracketing follows from basic tree structure require-ments and I assume it is one of three inviolable constraints on constituent structure, along with PROPERHEADEDNESS, (16), and LAYEREDNESS, (17).1(15) PROPER BRACKETINGEvery C j (̸= CMAX) has one and only one mother node (i.e., a given constituent cannot simulta-neously be part of two or more higher prosodic constituents).(Itô and Mester 2003; Nespor and Vogel 2007)1A similar proposal in Peperkamp defines ‘proper nesting’ by using alignment constraints to require prosodic edges tocoincide (Peperkamp 1997: 37, (42)). Because alignment constraints are violable, this predicts that some languages tolerateviolations of proper nesting. In contrast, I take the containment relations to be inviolable and any apparent mismatches are dueto mismatches between prosodic and metrical categories as I discuss in Section below.11(16) PROPER HEADEDNESSEvery (nonterminal) prosodic category of level i must have a head; that is, it must immediatelydominate a category of level i−1.(Itô and Mester 2003; Nespor and Vogel 2007; Selkirk 1996)(17) LAYEREDNESSNo prosodic category of level i dominates a category of level j, j > i.(Nespor and Vogel 2007; Selkirk 1996)Other components of the original Strict Layer Hypothesis, such as NONRECURSIVITY and EX-HAUSTIVITY (Selkirk 1996) were later shown to be violable. For example, recursive prosodic structureswere first described by Ladd (1986, 1988), and his findings were later corroborated by Dobashi (2003),Elfner (2012), Féry (2011), Féry and Truckenbrodt (2005), Frota (2000), Gussenhoven (2004), Itô andMester (2003, 2012), Kabak and Revithiadou (2009), Kubozono (1987, 1989), van der Hulst (2010), &Wagner (2005, 2010), among others. I follow Selkirk (2011), who takes the strong position that con-straints like NONRECURSIVITY and EXHAUSTIVITY are not part of the universal constraint set CONin Optimality Theory (Prince and Smolensky 1993). Instead, any instance of prosodic recursion orlevel-skipping is the direct result of satisfying some syntax-prosody correspondence constraint. Metrical structureMetrical structure refers to a hierarchical representation of phonological units which are designated asrelatively “strong” or “weak” (Halle and Vergnaud 1987; Liberman 1975; Liberman and Prince 1977).Stress is seen as a relational property obtaining between strong and weak units. In line with manycommon metrical theories, I interpret metrical structure as a branching tree-like structure composed ofthe set of ordered metrical categories in (18).2(18) METRICAL CATEGORIESFtfootsyllablemora2The other widely-used representation in metrical theory is a grid-like structure of many levels, where prominence isdetermined mainly by edge alignment and rhythmic constraints (Gordon 2002; Prince 1983). Other models combine bothrepresentations.12The exact number of categories within the metrical hierarchy differs among researchers. For exam-ple, some do not include the mora as the lowest metrical category, and instead treat moras as a property ofsyllables rather than a prosodic unit (Itô and Mester 2003; Lunden 2006; McCarthy and Prince 1993a).Again, I take the dominance relations in (18) to represent containment relations. Hierarchical structurescontaining these categories must obey PROPER BRACKETING in (15), PROPER HEADEDNESS in (16),and LAYEREDNESS in (17).In addition, a relative prominence relation is defined for sister nodes within the tree: one nodeis assigned the value strong (s) and all the other nodes are assigned the value weak (w); the strongnode is referred to as the head of the constituent it is contained in.3 I assume that stress is simplythe property of being a foot head (Halle and Vergnaud 1987; Hayes 1995; Liberman and Prince 1977;Selkirk 1980), and that the abstract phonological property of stress may be expressed by some type ofphonetic prominence which may take different forms in different languages. However, not all foot headsmust be overtly realized with phonetic prominence. I analyze prominence in Blackfoot (realized as apitch peak on a single syllable) as the manifestation of stress on a head foot within the domain of stressin Section 3.2.2; the heads of secondary feet have no acoustic manifestation. For discussion on theexistence of ‘stressless’ feet see Bennett (2012), Buckley (2009), Crowhurst (1996), González (2007),& Halle and Vergnaud (1987).Following Inkelas (1990, 1993), the metrical categories exist in a hierarchy which is separate fromthe prosodic hierarchy entirely. Because prosodic and metrical structure form two distinct representa-tions, there must be rules of correspondence between them. I discuss these in Section CorrespondenceIn this section I discuss how the correspondences between syntactic, prosodic, and metrical constituentsis regulated. In Section I describe the constraints which require syntactic phases and cer-tain prosodic constituents to correspond exactly, as well as the constraints which contribute to non-isomorphism between the two representations. In Section I describe the constraints which re-quire edge alignment between prosodic and metrical constituents. Syntax-prosody correspondenceThe proposal I lay out in this section follows many recent theories of prosodic phonology which derivethe correspondence of syntactic and interface constituents via a set of ranked and violable constraintswithin an Optimality Theoretic (OT) framework (McCarthy and Prince 1993a,b; Prince and Smolensky1993), including Alignment Theory (Selkirk 1986, 1995), Wrap Theory (Truckenbrodt 1995, 1999),3The prominence relationship is often conflated with Proper Headedness. My definition of prominence follows Nesporand Vogel (2007), who distinguished Proper Headedness and prominence relations in their formulation of the Strict LayerHypothesis.13and Match Theory (Elfner 2012; Selkirk 2009, 2011).4 Within these models, syntactic structure (post-movement) forms the input to the phonological component of grammar, and theory-specific constraintscontrol the correspondence between syntactic constituents and prosodic constituents of particular types.I take Match Theory (Selkirk 2009, 2011) as my starting point, whose correspondence constraints(MATCH) require exact correspondence between syntactic constituents and prosodic constituents ofparticular types. Because phonological constraints in OT are evaluated in parallel, this predicts thatthe default correspondence between syntactic and prosodic structure can be violated in order to sat-isfy higher-ranked constraints.5 The MATCH constraints regulate correspondences between syntacticconstituents of certain types and prosodic constituents, represented informally in (19).(19) SYNTAX-PROSODY CORRESPONDENCES IN MATCH THEORY“syntactic clause” ([Comp, C] or [Comp, Force]) ←→ Ì (intonational phrase)“syntactic phrase” (XP) ←→ F (phonological phrase)“syntactic word” (X0) ←→ ř (prosodic word)As I demonstrate in later chapters, these correspondence relationships are too strict for Blackfoot.(Just as they are in other polysynthetic languages; see Miller 2018 for arguments.)• MATCH(ř) constraints extend too low:– The terminal X0 elements are bound suffixes and are never prosodified as a PWd.– The smallest constituent which has a prosodic correlate is the v*P phase.• MATCH(F) constraints extend too low:– Not all XPs are prosodified as PPhs. In particular, the v*P phase as well as each vP/VPprojection above this are prosodified as a PWd in Blackfoot.• MATCH(F) constraints extend too high:– Not all XPs are prosodified as PPhs. In particular, some functional projections between thefirst and second phase are prosodified outside the PWd but inside the PPh.4There are also earlier rule-based accounts in Nespor and Vogel (2007) and the Edge-Based model in Selkirk (1986).5Match Theory represents a sharp break from earlier theories of syntax-prosody correspondence such as Align-XP (Selkirk1986, 1995), Wrap Theory (Truckenbrodt 1995, 1999). Rather than assuming that prosodic structure has different propertiesthan syntactic structure, with algorithms designed to ‘flatten’ syntactic structure into a well-formed prosodic representation,Selkirk (2011) assumes that prosodic structure exactly matches syntactic structure by default. As a consequence, recursiveand weakly-layered structures are typical, and there are no markedness constraints against such structures in the theory. Othertypes of prosodic wellformedness constraints exist though, such as those requiring binarity.14I propose instead a Phase-Based Match Theory which is based on syntactic phases,6 as in (20). (SeeGuekguezian 2017 for a recent proposal along similar lines.)(20) PHASE-BASED CORRESPONDENCES FOR BLACKFOOTPhase II CP ←→ PPh (phonological phrase)Phase I vP/VP ←→ PWd (prosodic word)v*P ←→ PWdMIN (minimal prosodic word)A further problem is how to formulate the MATCH constraints themselves. As originally formu-lated in Selkirk (2009, 2011), the correspondence of syntactic and prosodic constituents is governed bya violable family of syntax-prosody correspondence (MATCH) constraints, which require exact corre-spondence between the left and right edges of syntactic constituents and prosodic constituents of par-ticular types. These constraints were later redefined in Elfner (2012) in terms of exhaustive dominance.However, Itô and Mester (2018) argue that the MATCH constraints as originally formulated in Selkirk(2011) are redundant in the sense that they not only (a) require the existence and correspondence ofparticular syntactic and prosodic constituents, but also (b) require an exact match of edges between thetwo. They suggest that we formally disentangle the two uses of MATCH constraints: MATCH should beredefined to only require the existence of particular constituents (which they call MATCH-∃ for clarity),and MATCH constraints regulate the details of exact correspondence. I adopt this idea here as well.The definitions below are modified from MATCH-∃ as defined in Itô and Mester (2018: 185, (43)).Specifically, I require that each v*P (the predicate of events) in the input representation must correspond(McCarthy and Prince 1993b) to a PWd constituent in the output representation, and vice versa. Simi-larly, each CP or DP phase must correspond to a PPh, and vice versa. There are two syntax-to-prosodycorrespondence constraints, (21), and two prosody-to-syntax constraints, (22).(21) SYNTAX-TO-PROSODYa. MATCH-∃(CP→PPH)Abbreviation: MATCH-∃ (CP) or M-∃ (CP)Given an input syntactic representation S and an output phonological representation P, suchthat SRP, assign a violation mark for every CP phase in S which does not have a correspon-dent PPh constituent in P.b. MATCH-∃(V*P→PWD)Abbreviation: MATCH-∃ (v*P) or M-∃ (v*P)Given an input syntactic representation S and an output phonological representation P, such6To my knowledge, no published work exists on Blackfoot intonational phrases, and nothing in this thesis bears on thestatus of IPh. I focus instead on the “word” level.15that SRP, assign a violation mark for every projection of a v/V in S which either (1) is a v*Pphase, or (2) dominates a v*P phase, which does not have a correspondent PWd constituentin P.(22) PROSODY-TO-SYNTAXa. MATCH-∃(PPH→CP)Abbreviation: MATCH-∃ (PPh) or M-∃ (PPh)Given an input syntactic representation S and an output phonological representation P, suchthat SRP, assign a violation mark for every PPh constituent in P which does not have acorrespondent CP phase in S.b. MATCH-∃(PWD→V*P)Abbreviation: MATCH-∃ (PWd) or M-∃ (PWd)Given an input syntactic representation S and an output phonological representation P, suchthat SRP, assign a violation mark for every PWd constituent in P which does not have acorrespondent constituent in S which either (1) is a v*P phase, or (2) dominates a v*P phase.For the MATCH constraints which require exact correspondence, I modify the definition of MATCH-PHRASE in Elfner (2012), which is based on the concept of exhaustive dominance.(23) EXHAUSTIVE DOMINANCEA syntactic node a exhaustively dominates a set of terminal nodes b iff a dominates all and onlythe terminal nodes in b . (Elfner 2012: 27, (17))(24) MATCH-PHRASE (to be revised)Suppose there is a syntactic phrase (XP) in the syntactic representation that exhaustively domi-nates a set of one or more terminal nodes a . Assign one violation mark if there is no phonologicalphrase (F) in the phonological representation that exhaustively dominates all and only the phono-logical exponents of the terminal nodes in a . (Elfner 2012: 28, (19))The advantage of this definition is that it is defined precisely, and that any phonologically nullmorphemes or traces are ignored. The disadvantage is that it is formulated too strictly and is violatedwhenever elements are epenthesized or deleted.To see this, suppose that x exhaustively dominates a set of terminal nodes, a = { x, y, z}, as in thefollowing tree.16(25) xx yy zSuppose further that the phonological exponents of the terminal nodes in a are as follows:(26) a. x↔ /bc/b. y↔ Øc. z↔ /def/Then a PPh which dominates all and only the phonological exponents of those terminal nodes, as incandidate (a) below, satisfies MATCHPHRASE. If a terminal node as a null phonological exponent, asfor y above, this does not affect the PPh. However, a PPh which dominates an epenthetic segment, like[a] in candidate (b), or a PPh which does not dominate a deleted segment, like [f] in candidate (c), willviolate this constraint. (PPh boundaries in the output are designated with curly braces { }.)(27)/bc-def/ MATCHPHRASEa. {bcdef}b. { a bcdef} ∗c. {bcde } ∗This behavior does not seem desirable. Because MATCH-∃ already requires a correspondence be-tween syntactic and prosodic constituents, I propose the following redefinition of MATCHPHRASE,which is localized to the strings of elements dominated by CP or PPh, and which ignores segments thatdo not have correspondents in both the input and the output. Following Guekguezian (2017), I splitMATCHPHRASE into two definitions, which penalize undermatches and overmatches, respectively, withviolations evaluated per segment.(28) MATCH(CP) (ALL)Abbreviation: MATCH(CP) (All) or MA(CP)Let S be an input syntactic representation and P be an output phonological representation. Sup-pose there is a CP constituent in S that exhaustively dominates a set of one or more terminal nodesa ∈ S, and there is PPh constituent in P and CP R PPh. Let S2 be the output string in P.17Assign a violation mark for every element that (1) is an exponent of a morpheme in a and (2) hasa correspondent in S2 which is not dominated by the PPh.(29) MATCH(CP)(ONLY)Abbreviation: MATCH(CP) (Only) or MO(CP)Let S be an input syntactic representation and P be an output phonological representation. Sup-pose there is a CP constituent in S that exhaustively dominates a set of one or more terminal nodesa ∈ S, and there is PPh constituent in P and CP R PPh. Let S2 be the output string in P.Assign a violation mark for every element that (1) is an exponent of a morpheme that is not in aand (2) has a correspondent in S2 which is dominated by the PPh.These definitions maintain the advantages from MATCHPHRASE but do not assign violations forepenthesized and deleted segments. To see this, suppose there is a CP in an input representation whichexhaustively dominates a set of terminal nodes, a = { x, y, z}, as in the following tree.(30)CPxy zwSuppose further that the phonological exponents of all the terminal nodes are as follows:(31) a. x↔ /bc/b. y↔ Øc. z↔ /def/d. w↔ /ghi/Now a PPh which dominates all and only the correspondents of the exponents of the terminal nodesin a , as in (a) below, satisfies both constraints. And a PPh which dominates an epenthetic segment,like [a] in candidate (b), or a PPh which does not dominate a deleted segment, like [f] in candidate (c),satisfies both constraints. Finally, candidates where one terminal element has either been moved outsidethe PPh (d) violate MATCH(CP) (All), and candidates where one terminal element has been moved18inside (e) violate MATCH(CP) (Only). Intuitively, these are the types of prosodifications that entail amismatch between syntactic and prosodic boundaries. (PPh boundaries in the output are designated withcurly braces { }.)(32)/bc-def-ghi/ MA(CP) MO(CP)a. {bcdef}ghib. { a bcdef}c. {bcde }d. {bcde}fghi ∗e. {bcdeg}hi ∗There are analogous MATCH(v*P) (All) and MATCH(v*P) (Only) constraints, as below.(33) MATCH(V*P) (ALL)Abbreviation: MATCH(v*P) (All) or MA(v*P)Let S be an input syntactic representation and P be an output phonological representation. Sup-pose there is a projection of a v/V in S which either (1) is a v*P phase, or (2) dominates a v*Pphase, which exhaustively dominates a set of one or more terminal nodes a ∈ S, and there is PWdconstituent in P and vP R PWd. Let S2 be the output string in P.Assign a violation mark for every element that (1) is an exponent of a morpheme in a and (2) hasa correspondent in S2 which is not dominated by the PWd.(34) MATCH(V*P)(ONLY)Abbreviation: MATCH(v*P) (Only) or MO(v*P)Let S be an input syntactic representation and P be an output phonological representation. Sup-pose there is a projection of a v/V in S which either (1) is a v*P phase, or (2) dominates a v*Pphase, which exhaustively dominates a set of one or more terminal nodes a ∈ S, and there is PWdconstituent in P and vP R PWd. Let S2 be the output string in P.Assign a violation mark for every element that (1) is an exponent of a morpheme that is not in aand (2) has a correspondent in S2 which is dominated by the PWd.By default then, MATCH constraints create a prosodic structure which is completely isomorphic tosyntactic structure. Subsequent works have redefined these constraints in various ways (Guekguezian192017; Itô and Mester 2018; Kalivoda 2018; Lee and Selkirk 2016), but the basic idea remains the same:syntactic and prosodic structural correspondences are regulated via ranked and violable constraints.Because phonological constraints in OT are typically evaluated in parallel, this predicts that the defaultcorrespondence between syntactic and prosodic structure can be violated in order to satisfy higher-ranked prosodic wellformedness constraints. Because MATCH is violated for every segment that ismisaligned, these definitions predict that misalignments will be minimal and at prosodic edges.To see how the constraints are assigned violations, see the tableau below, which uses the CP/PPhMATCH constraints as an illustration. The CP phase constituents in the input are marked with [ ]; thePPh prosodic constituents in the output are marked with ( ); corresponding constituents are markedby identical subscripts; x and y stand for strings of sounds. The MATCH-∃ constraints are assigned aviolation whenever the number of corresponding constituents do not match. The MATCH constraintsare violated if a prosodic constituent dominates not all or not only the exponents associated with thecorresponding XP.(35)[ x [ y ]1 ]2 M-∃ (CP) M-∃ (PPH) MA(CP) MO(CP)a. ( x ( y )1 )2b. ( x ( y )3 )2 ∗ ∗c. ( ( x )3 ( y )1 )2 ∗d. ( x )2 ( y )1 ∗ye. ( x y )2 ∗f. ( x y )1 ∗ ∗yWhen prosodic wellformedness constraints dominate MATCH constraints, mismatches occur. Thereare two constraints relevant for this thesis, defined below, although other prosodic wellformedness con-straints have been proposed.(36) EQUALSISTERS (EQSIS)Sister nodes in prosodic structure are instantiations of the same prosodic category. (Myrberg2013)(37) BINMIN (BIN)A PPh must consist of at least two prosodic words. (Inkelas and Zec 1995)The perfectly matched candidate (a) satisfies all the MATCH constraints but violates EQUALSISTERSand BINMIN. Because constraints are rankable, the MATCH constraints and prosodic wellformednessconstraints together predict a typology of languages.20(38)[ x [ y ]1 ]2 EQSIS BIN M-∃ (CP) M-∃ (PPH) MA(CP) MO(CP)a. ( x ( y )1 )2 ∗ ∗b. ( ( x )3 ( y )1 )2 ∗∗ ∗c. ( x )2 ( y )1 ∗∗ ∗yd. ( ( x )1 y )2 ∗ ∗ ∗y ∗xe. ( ( x )2 y )1 ∗ ∗ ∗y ∗xf. ( x y )2 ∗In the final section, I turn to the question of prosodic and metrical alignment. Prosody-metrical alignmentSince the prosodic and metrical hierarchies are separate, there must be some mechanism to force themto align. Recall that the dominance relations in (13) and (18) are essentially containment relations (seePak 2008 for a good discussion of this point). This is why categories within each hierarchy must obeyProper Bracketing (Itô and Mester 2003, discussed above). For example, PWds are contained entirelywithin PPhs, and syllables are contained entirely within Feet. There are no misaligned boundaries of thesort in (39) or (40).(39) MISALIGNED PROSODIC BOUNDARIES (DISALLOWED)* (xxxx)(xxxx) PPh. . . (xxxx). . . PWd(40) MISALIGNED METRICAL BOUNDARIES (DISALLOWED)*(xxxx)(xxxx) Foot. . . (xxxx). . .  (after Pak 2008: 47)The relationship between prosodic and metrical categories must be something other than a con-tainment relationship, because there are many types of misalignments between prosodic and metricalconstituents, including extrametricality of various kinds at edges and resyllabification across prosodicboundaries. Inkelas (1990) describes some of the mismatches between metrical and prosodic con-stituents, showing that they do not obey strict layering. Consequently, misalignments like those in (41)and (42) are tolerated.(41) MISALIGNED PROSODIC AND METRICAL BOUNDARIES (ALLOWED)(xxxx)(xxxx) PWd. . . (xxxx). . . Foot21(42) MISALIGNED PROSODIC AND METRICAL BOUNDARIES (ALLOWED)(xxxx)(xxxx) PWd. . . (xxxx). . . I proposed that edge-based Alignment constraints govern the relation between prosodic and metricalstructure (McCarthy and Prince 1993a; Selkirk 1986).7(43) GENERALIZED ALIGNMENT TEMPLATEAlign (Cat1, Edge1, Cat2, Edge2) ≡∀ Cat1 ∃ Cat2 such that Edge1 of Cat1 and Edge2 of Cat2 coincide.WhereCat1, Cat2 are one of the prosodic or metrical categories, andEdge1, Edge2∈ {Right, Left} (McCarthy and Prince 1993a)There are entire families of constraints, like the ones defined below which require the edges of feet toalign to different prosodic constituents. And indeed, languages make different rankings amongst theseconstraints. For example, in French stress is culminative at the phrasal level, but not necessarily at theword level, since rules of destressing may eliminate word stresses on the surface (Dell 1984). In Iwaija,the IP may be the domain of stress (Birch 2002). Later in Section 3.2.2 I argue that stress in Blackfootis culminative at the phrasal level as well.(44) a. ALIGN (IPh, L, Ft, L)b. ALIGN (PPh, L, Ft, L)c. ALIGN (PWd, L, Ft, L)(45) a. ALIGN (IPh, R, Ft, R)b. ALIGN (PPh, R, Ft, R)c. ALIGN (PWd, R, Ft, R)If the prosodic and metrical hierarchies are decoupled and instead aligned at edges, there is aninteresting prediction: there is the potential for a foot or a syllable to span a PWd boundary. I return tothis point in Section 5.2, where I argue that Blackfoot is such a language. To see how this would work,consider a case where a prefix attached to a v*P phase, such that the v*P phase will be MATCH-ed to aPWd in the tableaux below, I use ( ) for PWd, and [ ] for an input v*P or for a syllable in the output.Consider cases where the prefix ends in an open syllable, represented below by V-, and the PWdbegins in another vowel. Candidate (a) parses each vowel to a separate syllable, which satisfies bothMATCH constraints and ALIGN-IO(PWd,), but violates ONSET because the second vowel is parsed7Generalized Alignment Theory (McCarthy and Prince 1993a) allows morphological and prosodic/metrical categories tobe aligned directly. In my proposal the relationship between morphosyntactic and prosodic categories is regulated via MATCHconstraints, and the relationship between morphosyntactic and metrical categories is indirectly mediated via the prosodiccategories.22to an onsetless syllable. If this is a language with distinctive vowel length, then both vowels can beparsed to a single syllable, as in the remaining candidates. Candidates (b) and (c) satisfy ALIGN-IO(PWd,) by allowing imperfect correspondences between the input v*P and the output PWd, violatingMATCH(PWd) (All) and MATCH(PWd) (Only), respectively. The partial ranking ALIGN-IO(PWd,)would make (d) the optimal candidate, where both vowels are parsed to a single syllable, but the PWdboundary falls between the two moras in the middle of a syllable, satisying the two MATCH constraints.(46)V- [V  ] AL(PPh,) ONSET MA(PWD) MO(PWD) AL(PWd,)a. { [V] ( [V] [] ) } ∗b. { [VV] ( [] ) } ∗c. { ( [VV] [] ) } ∗d. { [V ( V] [] ) } ∗Now consider cases where the prefix is a single consonant, or ends in a single consonant, representedbelow by a C, and the PWd begins in a vocoid. Again, if the MATCH constraints outrank ALIGN-IOthen misalignments between prosodic and metrical structure are expected.(47)C- [V  ] AL(PPh,) ONSET MA(PWD) MA(PWD) AL(PWd,)a. { [CV] ( [] ) } ∗b. { ( [CV] [] ) } ∗c. { [C ( V] [] ) } ∗For languages where feet and syllables span PWd boundaries, the evidence for the PWd itself wouldbe entirely morphophonological. Since there is a prosodic boundary at the left edge of the PWd, wemight expect this boundary to condition other generalizations, such as phonological exponence of mor-phemes. We would also not expect any metrical generalizations to hold of the PWd. For example, ifthere are generalizations which hold of stress at the PPh level, such as obligatoriness, culminativity, oredge demarcation, then those properties should not hold of the PWd.1.3 Language overviewThe empirical focus of this dissertation is Blackfoot, a language of the Algonquian family,8 whichextends across North America from British Columbia to Labrador and as far south as North Carolina(Mithun 1999). Blackfoot is spoken in northern Montana, USA and southern Alberta, Canada, making8The Algonquian family is well-established by lexical reconstructions based on regular sound changes (cf. Aubin 1975;Bloomfield 1946; Hewson 1993; Pentland 1979; Siebert 1967).23it the westernmost language within Algonquian. The location of Blackfoot in relation to the rest of thefamily is shown in Figure 1.2. On this map, a dashed line outlines Wiyot and Yurok, two Californianlanguages which, together with Algonquian, form the Algic language family (Goddard 1975; Haas1958; Sapir 1913, 1922).In terms of similarities to other Algonquian languages, Blackfoot is one of the so-called “PlainsAlgonquian” languages, an areal grouping which includes Cheyenne and Arapaho-Atsina. In the mapin Figure 1.2, Cheyenne and Arapaho-Atsina are represented by the non-contiguous gray area to thesoutheast of Blackfoot. Any shared features within this group have been argued to be the result ofcontact instead of shared innovations (Goddard 1994; Mithun 1999). However, it is clear that Blackfootalso shares many features of other Algonquian languages. For example, the Plains Cree dialect of theCree-Montaignais-Naskapi dialect continuum is spoken to the north and east of Blackfoot, and the twolanguages share many morphosyntactic features (see Bliss, Déchaine and Hirose 2013; Déchaine andWeber 2015, 2018; Hirose, Déchaine and Bliss 2020).Blackfoot is spoken on three reserves in southern Alberta and a fourth reservation in northern Mon-tana, shown in gray in Figure 1.3. The Siksiká (Blackfoot) reserve is slightly southeast of Calgary nearGleichen. The Aapátohsipikani (Peigan, or Northern Peigan) reserve is located at Brocket, southwestof Fort Macleod, Alberta. The Káínai (Blood) reserve is southeast of the Piikáni, near Cardston andStand Off, Alberta. The Aamsskáápipikani (Blackfeet, or Southern Piegan)9 reservation is located innorthwest Montana in Glacier County. Each of the four reserves is associated with a different dialect.The four dialects are mutually intelligible, but contain lexical, morphosyntactic, and phonological dif-ferences (Frantz 2009; Frantz and Russell 2017; Peter 2014). There is also interspeaker variation withineach dialect, and the preface to the dictionary notes that ‘there is as much variation between speakersfrom the same reserve as there is between speakers from different reserves’ (Frantz and Russell 2017:xiii). The dialects are thus determined based on geographic location, but are probably not the only locusof variation.According to the 19th edition of the Ethnologue, the Blackfoot population is approximately 15,000in Canada, and another 1,600 in the United States (Lewis, Simons and Fennig 2016). There are roughly3,250 speakers of Blackfoot in Canada (Statistics Canada 2011), and 50 speakers in the U.S. (p.c. DarylKipp to Mizuki Miyashita in 2011; documented in Fish and Miyashita 2017). This number is similarto the estimate of 4,315 speakers given in Russell and Genee (2006). In the author’s estimation mostspeakers are age 50 or older, and the language is rarely, if ever, the first language of acquisition forchildren.In terms of genetic affiliation, Blackfoot has been called ‘the most divergent language within theAlgonquian family’ (Goddard 2015), containing archaic retentions (Goddard 1994, 2015) as well as9The spelling is usually ‘Peigan’ in Canada, but ‘Piegan’ in the States.241,000 kmBlackfootterritoryAlgonquianWiyot and YurokWiyotandYurokAlgic languagesFigure 1.2: Location of Blackfoot in relation to the Algonquian and Algic languages. Map by EricLeinberger.25Figure 1.3: Locations of Blackfoot reserves. Map by Kevin McManigal.substantial phonological innovation (Berman 2006, 2007). Goddard (2015) argues that Blackfoot wasthe first to diverge from a Proto-Blackfoot-Algonquian ancestor, and that Proto-Algonquian can bedefined as a subgroup by several innovations. Many cognates to other Algonquian languages havebeen claimed (Berman 2006, 2007; Goddard 2015; Michelson 1935; Proulx 1989, 2005; Taylor 1960;Thomson 1978; Weber 2017), but the many unresolved details of Blackfoot phonology and morphologymean that there is still no full comparative reconstruction of Blackfoot historical phonology. My hopeis that the contributions in this dissertation move us towards a better understanding of the relationshipof Blackfoot to the other Algonquian languages.261.4 Data and methods1.4.1 SourcesThe data in this dissertation come from two main sources. The first is original fieldwork by the author(2011–present), primarily with speakers of the Káínai (Blood) dialect. The person who first introducedme to Blackfoot and taught me is Beatrice Bullshields (29 May 1945–1 July 2015), who was calledTotsinámm (Resembles Her Family). She was from the Káínai (Blood) reserve, and was a monolingualBlackfoot speaker until age 7, when she began attending St. Paul’s residential school in Cardston. Atthe time that I met Beatrice, she had lived in Seattle and Vancouver for many years, but often visited herfamily for the Akóka’tssini (Sundance) in the summers. She was an experienced language consultant,and served as the language consultant for three field methods classes at UBC (1998–1999, 2004–2005,2011–2012). The data in Section 3.2.2 in particular is drawn from elicitation sessions with Beatrice, andI describe my methods for elicitation in that section. I am also indebted to Natalie Creighton from theKáínai reserve, who is called Ááhsaikamo’sáákii (Always Conquering Woman), and Rod Scout, whogrew up in Siksiká. Other people from the Blackfoot community have also shared their language withme, but their words are not in this dissertation.The second source of data is the Blackfoot dictionary (Frantz and Russell 2017), which includesforms from all four major dialects. If a lexical item is restricted to one or two dialects, that informationis often noted in the dictionary entry. All entry headers and examples are written using an orthographywhich maps transparently to a phonemic or broad phonetic transcription (Frantz 1978). When I includeexamples from the dictionary, I often include a phonetic transcription, which I derive from the ortho-graphic representation following the explanation of the orthographic system in Frantz (1978, 2009).This derivation is made explicit in Section PresentationI present data in a four- or five-line interlinear gloss. If the first line exists, it is a phonetic transcriptiongiven in square brackets, as in (48). Where data is taken from my own fieldwork, (48a), the first lineis a broad IPA transcription given in square [ ] brackets, and the second line is an orthographic repre-sentation. I mark examples from my own fieldnotes with an abbreviation of the speaker’s name andoften with the date it was elicited. Where data is taken from a written source, (48b), the second line isreproduced faithfully from the original source and the phonetic transcription in the first line is derivedusing the rules in Section 1.4.3 and given in double square J K brackets. If the written source is Frantzand Russell (2017), the example has no citation; all other sources are cited. If the morphemic analysisand/or gloss lines differ from the original source, the source will be preceded by ‘Adapted from’.27(48) EXAMPLES OF A FIVE-LINE INTERLINEAR GLOSSa. FROM FIELDWORK[içpU´m:a]iihpómmaawa[ii\ohpomm–aa]–Ø–wa[IC\buy–AI]–IND–3‘she bought’ (BB; 2013-14-01)b. FROM WRITTEN SOURCESJsE:p´i:sKsaipíís[sa–ip/i–:s]–Ø[out–bring/v–2SG:3.IMP]–CMD‘bring her out!’If the phonetic transcription line does not exist, then the top line is a faithful reproduction of theoriginal source’s orthographic transcription and is given in an italicized font, (49) and (50). In otherwords, all examples from written sources begin either with a broad phonetic transcription in doublesquare brackets, J K, or with an orthographic transcription in italics.(49) áakoohtoyiiwaaak–[yooht–o–yii]–Ø–waFUT–[listen–v–3SUB]–IND–3‘s/he will hear him/her’(50) áaksipapainoyiiwáyiaak–[ipapa–in/o–yii]–Ø–w=ayiFUT–[dream–by.sight/v–3SUB]–IND–3=OBV.SG‘she will see him in a dream’The orthographic line is followed by two lines for a morphemic analysis and gloss. I mark thefirst phase of events (v*P) in square brackets in these two lines. This constituent is equivalent to thetraditional Algonquian theme (Bloomfield 1946). I separate an X0-adjoined√ROOT from the head X0it adjoins to with a slash. For example, the v0 -o occurs alone in (49) but adjoined to the root√IN‘by sight’ in (50). The v0 in transitive stems and the V0 in intransitive stems are equivalent to thetraditional Algonquian final, which determines valency and the animacy of one of the arguments. A√ROOT/X0 complex is equivalent to a complex Algonquian final, which consists of a more concreteelement followed by an abstract element which only determines valency and animacy. The slash in thetranscriptions follows the practice in Denny (1984) of demarcating the concrete and abstract finals fromeach other with a slash.In general, I follow the Leipzig glossing rules (2015), such that any punctuation mark from the set{ ( ) – = * \ / } in the morphemic analysis line is repeated in the gloss line. The \ symbol is used to marknon-concatenative morphology, such as the vowel ablaut in (48a). When multiple words or features inthe gloss line correspond to a single unit of the morphemic analysis line, the glosses are separated by aperiod. Parentheses around a morpheme or phrase indicate that element’s optionality, in the sense thatthe speaker produced the utterance with and without the optional element and judged both as acceptablewithin the given context. Other judgement marks (* % ?) precede the entire interlinear gloss. In caseswhere the example was judged unacceptable, I will preface the translation with ‘Target’ or ‘Intended’.28The final line is a free translation as given by the speaker or reproduced faithfully from the originalwritten source. Some examples from Frantz and Russell (2017) are words which have been extractedfrom longer examples; in those cases I change the free translation accordingly (for example, to usepronouns instead of full DPs).1.4.3 TranscriptionThe orthography developed in Frantz (1978) matches a broad phonetic or phonemic transcription fairlyclosely (see the phonemic inventory in Chapter 2), except that <’> is used for [P], <y> for [j], and<h> for /x/. Long consonants and vowels use doubled letters; e.g. <pp> is a geminate [p:] and <aa>is a long [a:]. The long mid open vowels [E:] and [O:] are represented uniformly with the digraphs <ai>and <ao>. The diphthong [oi] is written straightforwardly as <oi>.I made the following changes when deriving phonetic transcriptions from orthography:• I replace with <VV> with [V:].• In Section 2.4 I argue that vowel length is distinctive in open syllables, but neutralized to shortin closed syllables. (See similar arguments in Elfner 2006b.) Therefore, I transcribe <ai> and<ao> as long vowels [E:] and [O:] in open syllables and short vowels [E] and [O] in closed syllables.• The only exception to this rule is before glottal stop. Frantz (2017) includes words with both<V> and <VV> before <’>, so I have maintained vowel length distinctions before glottal stop.Frantz (1978, 2009) describes <ai> and <ao> as diphthongs before glottal stop, so I transcribe[ejP] for <ai’> and [awP] for <ao’>.• In Section 2.2 I show that vowels are short and lax before geminate consonants or coda [s]. Frantz(1978, 2009) states that vowels are short and lax before doubled consonants; geminates and coda[s] are both written with doubled consonants. Therefore, I transcribe [I, U, E, O, 2] for<i, o, ai, ao,a> before doubled consonants.• The trigraph sequences <aii>, <aai>, <aoo>, and <aao> are not explicitly discussed in Frantz(1978, 2009). In my own fieldwork, I typically hear these sequences as a vowel sequence: [E:i]for <aii> or <aai>, and [O:o] for <aoo> and <aao>. The same sorts of transcriptions are usedby Taylor (1969), who worked with speakers of a different dialect in Montana. I have continuedto use these transcriptions for orthographic trigraphs.• I replace <’> with [P] and <y> with [j].• I replace intervocalic <CC> with [C:]. For syllabified transcriptions, I use two consonants torepresent the geminate, separated by a syllable boundary, as in [C.C].29• I replace the prevocalic sequences <ts> and <ks> with the assibilants [>ts] and [>ks]. I replaceprevocalic <tts> and <kks> with [>t:s] and [>k:s]. For syllabified transcriptions, I use [t.>ts] and[k.>ks].• I replace interconsonantal <s> with a syllabic [s"]. This is true even if the immediately precedingconsonant is <t> or <k>.• I replace a sequence like <Css> with [Cs:]. For syllabified transcriptions, if the long [s:] isfollowed by a consonant, I interpret the long [s:] as a bimoraic nuclear sibilant, and I transcribe itlike [Cs":.C]. If the long [s:] is followed by a vowel, I interpret the long [s:] as a nuclear sibilantwhich is simultaneously parsed to the onset of the following syllable, and I transcribe it like[Cs".sV].• The sequence <sss> only occurs between a consonant and an orthographic vowel: <CsssV>. Ireplace this sequence with [Cs::]. For syllabified transcriptions, I interpret the superlong [s::] asa bimoraic nucleic sibilant which is simultaneously parsed to the onset of the following syllable,and I transcribe it like [Cs":.sV].• I replace the post-vocalic sequence <st> with [st]. I replace post-vocalic <stt> with [st:]. Forsyllabified transcriptions, I use [st.t].• I replace the post-vocalic sequence <ssC> with [sC].• I transcribe <h> as [ç] if it follows orthographic <i>, [x] if it follows orthographic <a>, and[xw] if it follows orthographic <o>. If the vowel before the dorsal fricative was written with adoubled letter or digraph that represents a monophthong (e.g.<ii>, <oo>, <aa>, <ai>, <ao>)then I transcribe a short vowel before the dorsal fricative. If the vowel before the dorsal fricativewas written with a single letter (e.g. <i>, <o>, <a>) then I transcribe the dorsal fricative as asyllabic fricative with no vowel. Examples: <aah> becomes [ax], but <ah> becomes [x"].• The only exception to the preceding rule is at the beginning of an orthographic word. A shortvowel before <h> is always voiced in this environment, so I follow the same procedure as if thevowel were a digraph.• I transcribe verbs in indicative and unreal clauses with a final devoiced vowel. I transcribe suffixednouns with a final devoiced vowel.• Speakers of the Káínai dialect often have no acoustic or articulatory reflexes of the verbal suffix-wa ‘3’ or the nominal suffix -wa ‘PRX’. This is discussed in many sources, including Bliss(2013), Bliss and Gick (2009), Bliss and Glougie (2010), Gick et al. (2012), &Windsor (2017a,b).30I made the conservative decision to transcribe these everywhere that they occur in the orthography,even though this may be an inaccurate transcription for some dialects.1.5 Outline of the dissertationChapter 2 discusses aspects of Blackfoot phonology, with a focus on the phonemic inventory andsyllable structure. The analysis of syllable structure will be relevant to later analyses of vowel hiatusresolution in Section 3.2.1 and Section 4.2 and primary stress in Section 3.2.2.Chapter 3 lays out the correspondence between CP phrases and PPh constituents. I argue thatthe verbal complex in Blackfoot has the syntax of a CP with pro arguments in Section 3.1. Then inSection 3.2 I present two phonological generalizations which are delimited by the PPh constituent inthis prosodic structure. Finally, I argue in Section 3.3 that each DP and CP phase corresponds by defaultto a PPh. A high-ranked prosodic markedness constraint which requires prosodic sisters to be of thesame type forces the verbal complex in Blackfoot to correspond to a PPh constituent as well.Chapter 4 lays out the correspondence of transitive vP and intransitive VP phrases to PWd con-stituents. In Section 4.1 I discuss the internal syntax of the vP/VP phrase in Blackfoot, including thesyntacticization of√ROOTs. Then in Section 4.2 I present a process of epenthesis which is unique tothe PWd constituent. Finally, I argue in Section 4.3 that the first phase (the event predicate) correspondsby default to a PWd.Chapter 5 discusses some of the broader theoretical implications of my proposals of correspon-dence. I argue in Section 5.1 that the PPh and the PWd are distinct in Blackfoot, and point out that somephonological generalizations at both prosodic levels require the PWd to fall in the middle of a syllable.In Section 5.2 I discuss why this is expected, given my proposal that metrical constituents are dominateddirectly by the PPh in Blackfoot. I suggest that other languages with so-called “clausal words” are likeBlackfoot in that the PPh is the domain of metrical constituents.Chapter 6 concludes.31Chapter 2Blackfoot phonologyThis chapter discusses aspects of Blackfoot phonology which are relevant for the remainder of the thesis.Many examples throughout this dissertation include a phonetic transcription, so I begin by discussingthe contrastive units and allophones for consonants (Section 2.1) and vowels (Section 2.2). Some ofmy arguments go against the analyses from reference grammars like Frantz (2009), Taylor (1969), &Uhlenbeck (1938). In particular, I argue that the pre-assibilants /st/ and /st:/ contrast with plain /t/ and/t:/, and that the assibilant [>ks] contrasts with plain /k/. For the vowel inventory I argue that Blackfoothas three short vowels but five contrastive long vowels. I then introduce the phonetic characteristicsof suprasegmentals (Section 2.3), because the location of stress is a major diagnostic of the Phono-logical Phrase (PPh), which I discuss in Chapter 3. Finally, I discuss syllable structure in Section 2.4because syllable structure conditions several phonological processes at the Phonological Phrase (PPh)and Prosodic Word (PWd) levels, and because many of the phonetic transcriptions in this thesis includesyllabification.2.1 ConsonantsThis section lays out evidence for the phonemic consonant inventory in Table 2.1. I discuss regularallophony within the system and also point out ways in which my analysis differs from previous re-searchers. In particular, I argue that the (pre-)assibilants /st/, /st:/, and />ks/ are phonemic. Appendix Aincludes additional minimal pairs which support the claims in this section.2.1.1 Phonemic inventoryThe inventory of phonemic consonants is given in Table 2.1. The glide /w/ occurs in the labial columnand also in parentheses in the dorsal column because it has a complex (labio-velar) place of articulation.The coronal sounds are all dental/alveolar, except for the palatal glide /j/. For some speakers /s/32is retracted and distributed for the plain fricative and in the post-assibilants,1 approximately [ç], [>tç],and [>kç]. The alveolar post-assibilants [>ts] and [>t:s] are not included in Table 2.1, because they areregular phonological variants of /t/ and /t:/ (see Section 2.1.2). There is no laryngeal contrast inobstruents, and plosives are voiceless and unaspirated. As shown here, many consonants have short andlong counterparts. According to Derrick (2006), long consonants are approximately 100% longer thantheir short counterparts.Table 2.1: Blackfoot phonemic consonant inventoryLabial Coronal Dorsal GlottalStops p p: t t: k k: PAssibilants st st:>ksFricatives s s: xNasals m m: n n:Glides w j (w)Several of these consonants have an unusual distribution with respect to the other consonants, whichI discuss in Section 2.4 on syllable structure below. Specifically, the dorsal fricative /x/ occurs onlybefore obstruents, the glottal stop /P/ usually occurs before consonants (but can rarely occur betweenvowels), and the alveolar sibilant /s/ can occur next to either vowels or consonants.Blackfoot includes alveolar pre-assibilants ([st] and [st:], which are distinct from the cluster [st]),alveolar post-assibilants ([>ts] and [>t:s], and velar post-assibilants ([>ks] and [>k:s]). The status of theassibilants has been discussed frequently in the literature (Armoskaite 2006; Chávez Peón 2015; Denzer-King 2009, 2012; Derrick 2006, 2007; Goad and Shimada 2014a,b). I treat [st], [>ts] and [>ks] as single,complex consonants for the following reasons: (a) they have the same distribution as simplex consonantsin onset positions, (b) there are alternations between plosives and assibilants, which I discuss below, and(c) the duration of [s] in assibilants is shorter than the duration of [s] in other clusters (Derrick 2006,2007). Some of the assibilants have a predictable distribution and are not included in the inventoryabove. The alveolar post-assibilants [>ts] and [>t:s] are predictable variants of /t/ and /t:/ before [i]Section 2.1.2. As far as I know, the long (pre-)assibilants [st:], [>t:s], and [>k:s] always alternate withgeminate plosives [t:] and [k:]. I discuss why I take /st/, /st:/, and />ks/ to be phonemic further in thefollowing paragraphs.1Trask (1996) defines an assibilant as ‘[a]n affricate whose fricative element is a sibilant, such as the affricate [ts]’. While[>ts] is clearly an affricate since both places of articulation are alveolar and therefore homorganic, the complex segment [>ks]arguably is not an affricate since the two places of articulation are not homorganic. The Blackfoot literature frequently refersto [>ts] and [>ks] using the same term, whether that term is an ‘affricate’ (e.g. Denzer-King 2009; Elfner 2006a; Frantz 2017;Kaneko 1999; Taylor 1969) or ‘assibilant’ (Derrick 2006, 2007). Derrick used ‘assibilant’ to refer to morphologically derived[>ts] and [>ks] and ‘affricate’ to refer to morpheme-internal or non-derived [>ts] and [>ks]. I see no reason to distinguish betweenthem and use ‘assibilant’ for both cases.33I list />ks/ in the inventory above because it contrasts with /k/, as shown by the minimal pair in (51).2(51) CONTRAST BETWEEN [k] AND [>ks]a. JIsk´ItKisskít[issk–Ø–i]–t–Ø[by.body–v–TI1]–2SG.IMP–IMP‘break it!’b. JIs>ks´ItKissksít[issk–i]–t–Ø[urinate–AI]–2SG.IMP–IMP‘urinate!’The pre-assibilants /st/ and /st:/ only occur after a front vowel or diphthong, [i], [E], or [oi], but inthat position they can occur in the same phonological environments as the plain plosives /t/ and /t:/.For example, [st] and [t] can both occur in the environment [. . . a:ni a. . . ], (52)–(53). I take this tomean that /st/ and /t/ contrast.(52) [a´:.ni.sta]áánistaa[aan–ist–aa]–Ø–wa[say–v–3OBJ]–IND–3‘s/he is called/named’ (BB)(53) Jma:.ni.ta´.pi:.>ksiK˚maanitápiiksi[maan–itapi]–iksi[recent–person]–AN.PL‘young people’Similarly, [st:] and [t:] can both occur in the environment [#a:>ksi x". . . ], (54)–(55). I take this tomean that /st:/ and /t:/ contrast.(54) Jaˆ:.>ksIst.tx".ka.pi.waK˚áaksisttahkapiwaaak–[istta–hkap/i]–Ø–waFUT–[under–drag/AI]–IND–3‘the sun will set’(55) Jaˆ:.>ksIt.tx".si.waK˚áaksittahsiwaaak–[ittahs–i]–Ø–waFUT–[triumph–AI]–IND–3‘he will be triumphant’There are five reasons to consider /st/ to be a complex consonant rather than a consonant cluster.First, vowel length is distinctive before the pre-assibilant /st/, (56). As I argue in Section 2.4.2, vowellength is neutralized to short in closed syllables, so the fact that vowel length is distinctive before /st/means that the sibilant is not parsed to a syllable coda position.2Most works concentrate on the behavior of [k] and [>ks] before a following [i] (e.g. Armoskaite 2006). However, thedistribution of [>ks] is much broader than this. It can occur before any vowel quality, which I discuss in Appendix A.1.3.34(56) SHORT OR LONG VOWELS BEFORE [st] ASSIBILANTa. Jaˆ:.>ksi.sta.waP.si.waK˚áaksistawa’siwaaak–[istaw–a’si]–Ø–waFUT–[grow–AI]–IND–3‘he will grow’b. Jaˆ:.>ksi:.sta.pi.na.kaP.si.waK˚áaksiistapinaka’siwaaak–[iistap–inak/a’si]–Ø–waFUT–[away–roll/AI]–IND–3‘it will roll away’Second, the pre-assibilant /st/ is distinct from a heterosyllabic cluster /st/. Phonetically, the dura-tion of the sibilant [s] is much shorter in the pre-assibilant than in the cluster. Phonologically, vowelsare predictably short before the cluster, (57). Vowels are predictably lax before clusters and geminates,which I discuss below. I argue in Section that the [s] in [sC] clusters is parsed to a syllablecoda, which is why vowel length is neutralized to short.(57) ONLY SHORT VOWELS BEFORE [st] CLUSTERa. Jaˆ:.>ksIs.to.ji.waK˚áaksisstoyiwaaak–[issto–yi]–Ø–waFUT–[beard–AI]–IND–3‘he will have whiskers’b. * [i:st]Third, pre-assibilation also occurs on a geminate, [st:], as in (58). However, there are no clusters inBlackfoot where the second consonant is a geminate. That is, there are no sequences like *[st:], *[xt:],or *[Pt:]. This again shows that pre-assibilation patterns differently than a consonant cluster.(58) PRE-ASSIBILATION ON GEMINATE [t:]Jaˆ:.>ksIst.tx".ka.pi.waK˚áaksisttahkapiwaaak–[istta–hkap/i]–Ø–waFUT–[under–drag/AI]–IND–3‘the sun will set’Fourth, pre-assibilation only occurs on a homorganic plosive [t]. There are no other pre-assibilatedconsonants. However, a moraic coda [s] is not so limited in distribution and can precede any obstruent,as I discuss in Section This suggests that pre-assibilation is tightly bound to [t] specifically,and that they form a single unit together.35Fifth, morphemic alternations also support the idea that [st] is a unit. There is a process of gemina-tion where a stem-final [t] followed by one of the inverse morphemes which begin in -ok will geminateto [kk]. For example, the stem kot- ‘give’ ends in a /t/, which assibilates to [>ts] before [i] in (59a).When kot- ‘give’ is followed by -oki ‘2→1’, the sequence /t-ok/ is realized as [kk], (59b).(59) GEMINATION OF STEM-FINAL /t/a. [ko.>ts´is]kotsís[ko–t–:s]–Ø[give–v–2SG:3.IMP]–IMP‘give it to him!’ (BB)b. [kUk.k´it]kókkit[ko–t–oki]–t–Ø[give–v–INV.21]–2SG.IMP–IMP‘give it to me!’ (BB)The same process of gemination occurs when the stem ends in [st]. For example, the stem aanist-‘tell’ ends in /st/, (60a). When aanist- ‘tell’ is followed by -okoo ‘X→SG’, the sequence /st-ok/ isrealized as [kk], (60b). Crucially, gemination affects the entire [st] unit, suggesting that it is a singlecomplex consonant.(60) GEMINATION OF STEM-FINAL /st/a. [a´:.ni.sta]áánistaa[aan–ist–aa]–Ø–wa[say–v–3OBJ]–IND–3‘s/he is called/named’ (BB)b. [ni.ta´.nIk.ko]nitánikkoonit–[aan–ist–okoo]–(hp)1–[say–v–INV.X:SG]–(IND)‘I am called/named’ (BB)The inventories in Kaneko (1999, 2000) and Peterson (2004) also include a bilabial assibilant />ps/,presumably for reasons of symmetry with alveolar [>ts] and velar [>ks]. However, /p/ is never followedby a short sibilant the way that /t/ and /k/ are. The examples in (61a) and (61b) show that [t] and [k]can be followed by both short and long [s], where short [s] has been interpreted as one half of a complexassibilant consonant, and long [s] as a bimoraic syllable nucleus. (See Section 2.4.3 below for syllabicconsonants.) The examples in (61c) show that /p/ can be followed by a long [s:] but never a short [s].36(61) DISTRIBUTION OF LONG AND SHORT [s] AFTER OBSTRUENTSTranscription Orthography Glossa. >ts Jni>tsama´xkimaP>tsisiK˚ nitsamááhkima’tsisi ‘my broom’ts: JIsts":a´:kojiwaK˚ isstssáákoyiwa ‘it dripped’b.>ks Ji>ksa´PsiwaK˚ iksá’siwa ‘s/he hid’ks: Ji:tE´çks":akjoPpiK˚ iitáíhkssakio’pi ‘clothesline’c. * >ps — —ps: Ja´Pps":apo´pi:tK á’pssapópiit ‘ride around (in a vehicle)!’2.1.2 AllophonyI describe regular allophones in this section. In Section I argue that [>ts] is a positional variant of/t/, and in Section I describe the three realizations of the dorsal fricative /x/. Short alveolar assibilant [>ts]The assibilant [>ts] is the positional realization of underlying /t/ before [i].3 However, an opaque processof deletion often removes the conditioning vowel. First, consider the distribution of [t] and [>ts] beforedifferent vowel qualities as summarized below. (Vowel length is unspecified in the table.) Only [>ts]occurs before [i] or [i:]. Both [t] and [>ts] can occur before all the other vowels.Context: [i] [E] [a] [O] [o][t] 8 3 3 3 3[>ts] 3 3 3 3 3The following examples include [t] and [>ts] in overlapping phonological environments before eachof the vowels in the above table.3The alveolar assibilant />ts/ is transcribed in various fashions. It is included in phonemic inventories as [ts] in Derrick(2007) & Kaneko (1999), as an affricate [>ts] in Elfner (2006a), as an affricate [ts] in Denzer-King (2009), and as a ‘complexstop’ [>ts] in Goad and Shimada (2014a,b).37(62) [t] VS. [>ts] BEFORE [i]a. * [ti] b. Ja>ts´isaK˚atsísa[atsis]–a[pants]–PRX‘pants’(63) [t] VS. [>ts] BEFORE [o]a. J>ksi:stoj´iwaK˚ksiistoyíwa[ksiisto–yi]–Ø–wa[hot–II]–IND–3‘It is/was hot.’b. Jmi:s>tso´jisiK˚miistsóyisi[miistsis–oyis]–i[wood–lodge]–IN.SG‘house of wood’(64) [t] VS. [>ts] BEFORE [E]a. Ji:tE´:sokç"takjoPpiK˚iitáísokihtakio’pi[ii\it–a–[isok–iht/aki]–o’p]–i[IC\LOC–IPFV–[above–putv/AI]–21.IND]–IN.SG‘shelf’b. Jaˆ:>ksIs>t:si>tsE:sokç"kaPsiwaK˚áaksisttsitsaisokihka’siwaaak–isttsitsa–[isok–ihk/a’si]–Ø–waFUT–no.longer–[good–behave/AI]–IND–3‘In the beginning he will act good, butnot later.’(65) [t] VS. [>ts] BEFORE [O]a. Ji:tO´:matapapi>tsi>tsIs":kojiK˚iitáómatapapitsitsisskoyi5[ii\it–a–omatap–[apitsitsissko]–Ø]–yi[IC\LOC–IPFV–start–[change.color?.II]–IND]–IN.SG‘September’ (lit: ‘when the leaveschange color’)b. JIs>t:si>tsO´:matapaPpoPtakiwaK˚isttsitsáómatapa’po’takiwaisttsitsa–omatap–[a’p–o’t/Ø–aki]–Ø–wano.longer–start–[around–take/v–AI]–IND–3‘At the beginning she started to work butshe no longer does.’5Frantz and Russell (2017) also lists iitaomatapapitsitssko as an alternative, where the penultimate syllable [>tsIs] <tsiss>has been reduced to [ts":] <tss> with a long [s":] as the syllable nucleus peak.38(66) [t] VS. [>ts] BEFORE [a]a. JIsta´:kaPsiwaK˚isstááka’siwa[[issta–Ø–aki]–a’si]–Ø–wa[[stake–v–AI]–AI]–IND–3‘It lodged, landed on end.’b. JIs>tsa´:kaPsiwaK˚isstsááka’siwa[isstsaak–a’si]–Ø–wa[praise–AI]–IND–3‘He bragged (about himself).’The examples in (66) are a minimal pair, which establishes that the choice between [t] and [>ts]can distinguish lexical items. However, these two examples likely have different abstract phonologicalrepresentations as well, for the following reason. The suffix -a’tsis attaches to animate intransitive(AI) stems to form an instrumental nominalization (Bliss, Ritter and Wiltschko 2016). As you can see,an underlying [i] before a non-[i] vowel like the [a] in -a’tsis becomes a glide after a consonant like[p], (67a). But it deletes all together after an underlying /t/, even though it causes the /t/ to assibilateto [>ts], (67b). Therefore, the [ts] in (66) could be derived from an abstract underlyingly sequence /tiV/.I assume that all instances of [>ts] before non-[i] vowels are derived from sequences of /tV/.(67) a. JIs:apja´P>tsi:s>tsiK˚issapiá’tsiistsi[[issap–i]–a’tsis]–istsi[[watch–AI]–NMLZ]–IN.PL‘binoculars’b. Jkax>tsaP>tsi:s>tsiK˚kaahtsa’tsiistsi[[kaaht–i]–a’tsis]–istsi[[game–AI]–NMLZ]–IN.PL‘playing cards’There are also morphemic alternations which support the neutralization of /t/ and />ts/ before high,front vowels. All morphemes which end in [t] before non-high, non-front vowels end in [>ts] before highfront vowels. This pattern is exhibited in (68) using a wide-scoping negative prefix, which is [ma:t-]before [a], (68a), and [o], (68b), but [ma:>ts-] before [i], (68c). (For a description of the syntax andsemantics of maat- ‘NEG’, see Bliss 2013.) Because the alternant [ma:t-] occurs in a wider number ofenvironments than [ma:>ts-], I assume the underlying form of this morpheme is /ma:t-/ ‘NEG’. There areno prefixes which are realized with a final [>ts] in all positions, supporting the idea that [>ts] only exists asa positional realization of underlying /t/.(68) a. Jma´:.taP.si.mi.waK˚mááta’simiwamaat–[a’s–im/i]–Ø–waNEG–[young–body/AI]–IND–3‘he is not young’39b. Jma´:.to.ma.ni:.wa:.>tsi.>ksiK˚máátomaniiwaatsiksimaat–[om–an/ii]–Ø–w=atsiksiNEG–[truth–say/AI]–IND–3=NONAFF.3SG‘he did not tell the truth’c. Jma´:.>tsi.>tsi.no´.p´i:.wa:.>tsi.>ksiK˚máátsitsinópííwaatsiksimaat–itsin–[op–ii]–Ø–w=atsiksiNEG–among—[sit–AI]–IND–3=NONAFF.3SG‘he did not sit among them’ (Frantz and Russell 2017) Dorsal fricatives [ç], [x], and [xw]The dorsal fricative has several allophones depending on the quality of the preceding vowel: [ç] afterthe front vowels [i] and [E], and the diphthong [>oi], (69a), [x] after the low vowels [a], (69b), and [xw]after the round vowels [o] and [O], (69c). The phonetic realization of /x/ after low vowels is quite farback, and some researchers write it as a uvular [X] (Denzer-King 2009; Uhlenbeck 1938).(69) ALLOPHONES OF [x] AFTER VOWELSa. Jiçki>ts´ikawaK˚ iihkitsíkawa ‘seven’Ja´kEçki>tsiwaK˚ ákaihkitsiwa ‘it is dry’JE´:s>oiçta:waK˚6 áísoihtaawa ‘he is plating his food’b. J>ksa´xkojiK˚ ksááhkoyi ‘dirt’c. Jo´xwkotokiK˚ óóhkotoki ‘stone’JmO´xw>ksin2t>tsiwaK˚ máóhksinattsiwa ‘it looks red’All three allophones can also occur between consonants.(70) ALLOPHONES OF [x] BETWEEN CONSONANTSa. Jaˆ:kç"ki>tsikawaK˚ áakihkitsikawa ‘she will freeze her feet’b. JaPkx"ko´jiK˚ a’kahkóyi ‘curved geographical feature’c. Jn´itx"wkUkkaK˚ nítohkokka ‘she gave to me’I discuss the distribution of /x/ more in Section 2.4.6The vowel [>oi] resulted from /o+i/ coalescence. I have also heard something closer to [Y] as a variant of this dipthongbefore /x/ and /s:/. Miyashita (2018: 225) actually transcribes <aisoçtaa> [E:soçta:], where there is no reflex of the highvowel [i] in the vowel itself. The occurrence of [ç] in this example is opaque, since the front vowel that conditions theoccurrence of [ç] never appears in the surface pronunciation.402.2 VowelsThis section lays out evidence for the phonemic vowel inventory in Table 2.2. I discuss regular allophonywithin the system and also point out ways in which my analysis differs from previous researchers. Inparticular, I argue that the long mid vowels /E:/ and /O:/ are phonemic. Appendix A includes additionalminimal pairs which support the claims in this section.2.2.1 Phonemic inventoryThe inventory of phonemic vowels is given in Table 2.2. There are three short vowels /i o a/, andfive long vowels /i: o: E: O: a:/. The short vowels [E] and [O] are not contrastive, but do occur asallophones of long [E:] and [O:] before geminates and consonant clusters (Section 2.4.2). In Section 4.2.1I show that high front vowels come in two types, which I label [i1] and [i2], which have differentmorphophonological effects on neighboring consonants.Table 2.2: Blackfoot phonemic vowel inventoryfront central backhigh i i: o o:mid E: O:low a a:I have maintained [o] as the transcription for the high back rounded vowel, because [o] reflects themost common pronunciation, but I treat it as [+high]. The vowels [u] and [u:] are free variants of /o/and /o:/, respectively, with the closed vowel occurring more readily in accented syllables rather thanunaccented, but there is no contrast between these two vowel qualities. In Chapter 3 I discuss twoprocesses where [i] and [o] pattern together to the exclusion of the other vowels. The first process isvowel hiatus resolution when the second vowel (V2) is long. A [w] is epenthesized between the twovowels if V2 is [E], [O], or [a]; otherwise, vowel hiatus is resolved via coalescence. The second processis the calculation of stress in indicative clauses with no person prefixes. Stress falls on an initial heavysyllable if it contains [E], [O], or [a], and otherwise falls on a predictable syllable. In both cases, thevowels [i] and [o] pattern differently than [E], [O], and [a]. Because of this, I suggest that the relevantfeature is [±high], and that [i] and [o] are [+high] while [E], [O], and [a] are [-high].7The inventory I propose here is novel, as most researchers assume a symmetrical six-vowel inventorythat does not include /E:/ or /O:/ (e.g. Elfner 2006b; Frantz 2017). Others do include either short or7But see Elfner (2006b) for a different interpretation. She treats only [i] as [+high], and the relevant feature that splits [i]and [o] from the remaining vowels is [±ATR].41long versions of the mid vowels [E] and [O] in the inventory but also state that these two vowel qualitiesalways result from vowel coalescence (Kinsella 1972; Stacy 2004; Taylor 1969; Van Der Mark 2003).The argument for why the long mid vowels /E:/ and /O:/ are phonemic is that some instances of[E:] and [O:] occur infrequently in non-derived environments (e.g. morpheme-internally), where theycontrast with other long vowels. I have found one minimal pair, in (71), which shows that /O:/ and /a:/can distinguish lexical items.(71) MINIMAL PAIR: LONG [O:] VS. LONG [a:]a. JO:n´i:tKaoníít[ao–n/i–i]–t–Ø[hole–by.needle/v–TI1]–2SG.IMP–IMP‘pierce it!’b. Ja:n´i:tKaaníít[aan–ii]–t–Ø[say–AI]–2SG.IMP–IMP‘say (s.t.)!’I know of no minimal or near minimal pairs which show a lexical contrast between /E:/ and othervowels, but there are examples which demonstrate that long /E:/ and /O:/ can occur in overlappingcontexts with the other long vowels. For example, [E:], [a:], and [o:] can all occur in the environment[#s ta. . . ]; that is, after a PPh-initial [s] and before [ta. . . ], as shown in (72).(72) OVERLAPPING ENVIRONMENTS: LONG [E:] VS. LONG [a:] VS. LONG [o:]a.% JsE:tamitKsaitamit[sait–am/i]–t–Ø[breathe–be/AI]–2SG.IMP–IMP‘breathe!’ (Piikáni dialect)b. Jsa:ta´pi>ksItKsaatápiksit[saat–apik/i–Ø]–t–Ø[across–throw/v–TI3]–2SG.IMP–IMP‘throw it out!’c. Jso´:tamanista:waK˚sóótamanistaawasootam–[an–ist–aa]–Ø–wadespite–[tell–v–3OBJ]–IND–3‘I told her in spite of all else!’42Similarly, [O:], [a:], and [o:] can all occur in the environment [#m to. . . ]; that is, after a PPh-initial[m] and before [to. . . ], as shown in (73).(73) OVERLAPPING ENVIRONMENTS: LONG [O:] VS. LONG [a:] VS. LONG [o:]a. JmO´:toPki:>ksiK˚máóto’kiiksi[maoto’kii]–iksi[Buffalo.Women]–AN.PL‘Buffalo Women’s Society’b. Jma´:tokamoP>tsipoka´:wa:>tsiK˚máátokamo’tsipokááwaatsimaat–okamo’t–[ipok–aa]–Ø–w=aatsiNEG–normal–[small–AI]–IND–3=NONAFF.3SG‘he wasn’t born normal’c. Jmo´:tokiPpiK˚móótoki’pi[mootoki’p]–i[skull]–IN.SG‘skull’There are more examples of non-derived [E:] and [O:] in the appendix (Appendix A.2.3). Becausethe mid vowels in the previous examples are unpredictable in the sense that they are not the result ofvowel coalescence across a morpheme boundary, I consider them phonemic. Even so, these phonemesare marginal in the sense that there are relatively few instances of non-derived [E:]/[E] and [O:]/[O].Blackfoot is not the only Algonquian language with asymmetrical numbers of short and long vowels.For instance, Nishnaabemwin (Valentine 2001: 34–41), Oji-Cree (Slavin 2012), and Plains Cree (Russell2008; Wolfart 1973: 79) each have three short vowel qualities /i a o/ but four long vowels /i: e: a: o:/,where the long /e:/ has no phonemic short counterpart. However, Russell (1992) suggests that /e:/in Plains Cree is a heavy diphthong with the underlying representation /ai/, fused by a late, possiblyphonetic rule which is widely attested in external sandhi processes. If this is true, then Plains Cree has asymmetrical system at a more abstract level (three short vowels, and three long vowels, but with a fourthlong vowel quality on the surface). Northern Plains Cree apparently truly has only three long vowels /i:a: o:/ (Wolfart and Carroll 1973).A less clear example exists with Shawnee, which has four short ([i e a o]) and four long ([i: e: a: o:])phonetic vowels. Although Muzaffar (1997) states that all four long vowels are used in transcriptionsin Voegelin (1935, 1938), the long vowels in Shawnee were long assumed to be allophonic variantsof short vowels. For instance, Voegelin claims that ‘[q]uantitative differences in Shawnee vowels are43however not phonemic’ (Voegelin 1935: 36), even though he does not describe the conditioning factorsfor length.More recently, Andrews (1994: 23) has claimed that two of the long vowels, /i:/ and /a:/, are phone-mic. He bases this claim on the two minimal pairs in (74) and (75) below, which respectively show thatshort /i/ contrasts with long /i:/ and that short /a/ contrasts with long /a:/.(74) a. ho-wiiši’-ta 3-head-TI ‘he was in charge’ (of the Shawnee)b. wi’ši dog (Chrisley 1992: 48)(75) a. cˇaaki aama ‘all this’ (Voegelin 1953, episode 3a)b. cˇaki ‘small’ (Chrisley 1992: 8)Since he finds no such minimal pairs for [e:] and [o:], he assumes they are not contrastive. In otherwords, Andrews privileges the criteria of lexical distinction above all others for determining contrast,even though there often are few minimal pairs in Algonquian languages.He concludes – presumably speaking about all four vowel qualities – that ‘most vowel length al-ternations [. . . ] are phonetic and perhaps conditioned by following segments. More research is neededto clarify the nature of these alternations in Shawnee’ (Andrews 1994: 23). Without a description ofthose vowel length alternations, it is impossible to say whether all instances of [e:] and [o:] are derivedfrom short vowels or not. Perhaps, as is the case for Blackfoot /E:/ and /O:/ discussed below, there areunderived instances of [e:] and [o:] in Shawnee, even if no minimal pairs can be found. If so, that wouldbe a good argument to include long /e:/ and /o:/ as phonemes.2.2.2 AllophonyVowels have several predictable variants, which I describe in this section. In Section I describepositional variants of mid vowels before glottal stops, and in Section I show that lax vowelsoccur before some types of consonants. Mid vowelsBefore a glottal stop, [E:] is pronounced as a short diphthong [ej], (76a), and [O:] is pronounced as adiphthong [aw], (76b). For some speakers this diphthong begins with a mid vowel quality and is closerto [Ow] or [ow].44(76) MID VOWEL DIPHTHONGIZE BEFORE GLOTTAL STOPa. [>e´jPpoji]á-i’poyiwaa–[i’po–yi]–Ø–(wa)IPFV–[speak–AI]–IND–(3)‘he is speaking’ (BB)b. [a´kawPto]ákao’toowaakaa–[o’t–oo]–Ø–(wa)PRF–[here–go.AI]–IND–(3)‘he has arrived’ (BB) Lax vowelsShort, lax vowels [I], [E], [2], [O], [U], occur predictably before [sC] clusters, (77), and geminates, (78).There are no short or long non-lax vowels in these positions.(77) SHORT LAX VOWELS BEFORE [sC] CLUSTERS[Is>ts´i:jit] isstsííyit ‘listen!’ (BB)[nitE´spIn:iPp] nitáísspinnii’pa ‘I’m lifting it’ (BB)[p2ska:ni] passkaani ‘dance’ (BB)*[OsC] — —[mU´ski>tsipx"pi˚] mósskitsipahpi ‘heart’ (BB)(78) SHORT LAX VOWELS BEFORE GEMINATES[kIp:o˚xwkU´k:it] kippohkókkit ‘please give me it!’ (BB)[E´k:amoks"kaPsi] áíkkamokska’siwa ‘he runs fast’ (BB)[nit2´k:a] nitákkaawa ‘my friend’ (BB)[O´t:aki] áóttakiwa ‘bartender’ (BB)[mUt:o>ks´is] mottoksísa ‘knee’ (BB)[Ist:o2´n] isttoán ‘knife’ (BB)[E´s>t:si] áísttsiiwa ‘it hurts’ (BB)[In:>o´ist:o2ni] innóísttoaniwa ‘American (lit: has long knife)’ (BB)In the next section I turn to suprasegmental properties in Blackfoot.2.3 SuprasegmentalsThis section describes the phonetic characteristics of suprasegmental properties, because the location ofstress is one major diagnostic of the Phonological Phrase (PPh) constituent that sets it apart from the45Prosodic Word (PWd) constituent. I return to the phonological characteristics of stress in Section 3.2,where I argue that prominence is best analyzed in terms of metrical stress.Syllabic prominence in Blackfoot is signaled primarily with a higher F0, as well as greater intensityand duration (Van Der Mark 2003). The pitch of syllables without pitch accent is interpolated betweenthe edges of the PPh and the pitch peak on the most prominent syllable; that is, pitch rises to a pitch peakand falls steeply after the pitch peak (Miyashita and Weber 2020; Weber and Allen 2012). The mostprominent syllable in a PPh is marked in the orthography by an acute (´) accent (Frantz 2009; Taylor1969).Because pitch is the salient property, prominence in Blackfoot is commonly referred to as ‘pitchaccent’ (Frantz 2009: 3), and this term has been widely adopted (Kaneko 1999; Miyashita 2011; VanDer Mark 2003; Weber 2016a,b). This can be misleading, especially because Blackfoot displays differ-ent characteristics than other pitch accent languages; for example, Japanese allows unaccented wordswhile Blackfoot does not allow unaccented PPhs (see discussion in Stacy 2004). Moreover, Hyman(2006) argues convincingly that ‘pitch accent’ is not a distinct word prosodic type. Instead, pitch ac-cent languages display properties of both stress and tone languages. Under this view, a more accuratedescription is that Blackfoot has a privative tonal contrast (H vs. Ø). It is a ‘sparsely’ tonal languagebecause there is at most one H and at least one H within a PPh. The location of the pitch peak onnouns is not predictable (Weber and Allen 2012), but later in Section 3.2.2 I show that the location ofthe pitch peak on verbs is predictable, and displays rhythmic metrical properties. This leads me to ananalysis where prominence is the acoustic manifestation of primary stress in Blackfoot. One way tomodel this would be to associate the single obligatory H tone within the PPh to the head syllable of thePPh. These are pitch accents in the sense of Ladd (1986, 2008). (See Miyashita and Weber 2020 forone Autosegmental Metrical account of Blackfoot pitch contours.)Prominence in Blackfoot has different phonetic realizations depending on the location of the sylla-ble in the PPh (Frantz 2009; Stacy 2004; Taylor 1969). It is realized as a level high on any syllable butthe last, as in iihkíítaawa [iC.k´ı:.ta] ‘s/he baked’, and the pitch of the utterance as a whole gradually risesor falls to/from that peak. Before glottal stops, as for isî’katsiiwa [i.sˆıP.ka.tsi] ‘s/he kicked him/her’, theaccent is often realized as a falling tone. On the final syllable of a PPh, the pitch may be falling in care-ful speech, but more usually is signalled through greater intensity, as for inihkíwa [in.Pç".kˆı]∼ [in.Pç"."ki]‘s/he sang’. This variation is summarized in Table 2.3 using transcriptions of the person’s speech who in-formed my analysis of stress in Section 3.2.2. When creating phonetic transcriptions from orthographicsources, I use an acute accent in all of these environments.Although Frantz (2009) claims that pitch can also rise over a long vowel as well, as in [ma:.a´x.si˚]‘his/her grandparent’, these can be analyzed as two separate syllables with pitch accent on the secondsyllable only. To my ears there are always two distinct sonority or intensity peaks (one for each syllable).Additional evidence is the fact that long vowels in this position can be superlong (i.e. longer than a46Table 2.3: Positional realization of prominenceLevel high Falling IntensityPPh-medially [iç.k´ı:.ta] — —Before [P] — [i.sˆı.ka.>tsi] —PPh-finally — [in.Pç".kˆı] [in.Pç"."ki]normal long vowel), which is expected if there is a sequence of two syllables, and one or both of thesyllables contains a long vowel. And finally, the vowel quality can change between the two syllables, asin [mO:.o´.ji˚] ‘his/her mouth’. I know of no cases of a rising pitch across a short vowel. Because of this,I claim that prominence in Blackfoot is a property of entire syllables and not of individual moras. Thatis, the stress-bearing unit is the syllable.Only one pitch peak tends to be marked in the orthography per orthographic word, although there aresome exceptions. For example, certain DP enclitics add a pitch peak to the syllable immediately to theirleft (Bliss 2013; Glougie 2009), and this tends to be marked in the orthography. From my own field-work, I sometimes hear secondary pitch peaks, which do not have as high an F0 as the first, distributedrhythmically throughout the PPh, although these depend on rate of speech and person. Secondary pitchpeaks do not tend to be marked in the orthography.2.4 Syllable structureIn this section I address syllable structure in Blackfoot. This is relevant throughout the dissertation,because I transcribe syllable boundaries in many of the phonetic transcriptions, and those transcriptionsfollow the analyses I discuss here. Additionally, syllable structure conditions several phonological pro-cesses at the PPh level (e.g. [w]-epenthesis in the onset of heavy syllables and the location of primarystress) and the PWd level (e.g. epenthesis). I discuss these processes in Chapter 3 and Chapter 4.This section is organized as follows. In Section 2.4.1 I discuss open syllables in Blackfoot and showthat vowel length is distinctive in open syllables. Then in Section 2.4.2 I turn to closed syllables andshow that vowel length is neutralized to short in closed syllables. Finally, in sections Section 2.4.3 andSection 2.4.4 I turn to two unusual aspects of Blackfoot syllable structure. In Section 2.4.3 I discusssyllabic fricatives in Blackfoot. In Section 2.4.4 I discuss how the right edge of the PPh hosts an extraconsonant slot; I analyze these as onsets to degenerate syllables which contain no nucleus. Throughoutthe chapter I analyze syllabic structure using tenets of Moraic Theory (Hayes 1989a; Hyman 1985;Pulleyblank 1994) and Optimality Theory (McCarthy and Prince 1993a,b; Prince and Smolensky 1993).47Syllables in Blackfoot may be (C)V, (C)VV, or (C)VC.8 Crucially, there are no syllables like CVVC.Using Moraic Theory (Hayes 1989a; Hyman 1985; Pulleyblank 1994), I argue in this section that themaximal syllable in Blackfoot is bimoraic, (79), and that codas in Blackfoot are moraic. A constraintagainst trimoraic syllables explains why long vowels cannot occur before moraic codas.(79) MAXIMAL SYLLABLECV(){VC}In the next section I argue that vowel length is distinctive in open syllables. After that I show thatvowel length is neutralized to short in closed syllables. I take this as evidence that the coda is moraic.2.4.1 Open syllables: contrastive vowel length2.4.1.1 DataVowel length is distinctive in open syllables. As I will show later, vowel length is neutralized before codaconsonants, so this property provides crucial evidence that the following consonant is in a syllable onsetposition. Examples (80)–(83) include minimal pairs which show that short and long vowels contrastbefore short consonants. There are relatively few known minimal pairs in Blackfoot, although I collatemany more in Appendix A. In order to help catalog the existing minimal pairs, I give a citation aftereach pair to give credit to the researchers who first noticed that pair. The mid [-ATR] vowels are notshown here because they do not contrast for length.8Onsetless syllables are only tolerated at the left edge of the PPh. The right edge of the PPh allows an extra consonant,which I analyze Section 2.4.4 as a degenerate syllable.48(80) a. Jo:>tsis>ts´iniK˚ootsistsíni[ootsistsin]–i[3.palate]–OBV‘palate, his/her palate’b. Jo:>tsis>ts´i:niK˚ootsistsííni[ootsist–iin]–i[strawberry–berry]–IN.SG‘strawberry’(Denzer-King 2009: 18; Taylor 1969: 34)(81) a. Ja´jamo:waK˚áyamoowa[ayam–oo]–Ø–wa[aggrieved–go.AI]–IND–3SG‘he left with hurt feelings’b. Ja´ja:mo:waK˚áyaamoowa[ayaam–oo]–Ø–wa[misdirected–go.AI]–IND–3SG‘he went in a different direction’(Denzer-King 2009: 18)(82) a. Jaˆ:koPkowaK˚áako’kowaaak–[o’k–o]–Ø–waFUT–[fall–II]–IND–3SG‘it will be fall’b. Jaˆ:koPko:waK˚áako’ko:waaak–[o’k–oo]–Ø–waFUT–[raw–go.AI]–IND–3SG‘she will become exhausted’(Denzer-King 2009: 18)(83) a. Jaˆ:koka:waK˚áakokaawaaak–[ok–aa]–Ø–waFUT–[rope–AI]–IND–3SG‘he will rope’b. Jaˆ:ko:ka:waK˚áakookaawaaak–[ook–aa]–Ø–waFUT–[Sundance–AI]–IND–3SG‘she’ll sponsor a Sundance’(Frantz 2009: 2)Singleton consonants are parsed as the onsets to the syllable containing the following vowel ratherthan as a coda, (84). This follows from the cross-linguistic generalization that CV sequences are alwaysparsed into the same syllable (often termed the Onset Principle or Core CV syllable formation; seeClements and Keyser 1983; Itô 1986; Kahn 1976; Selkirk 1982; Steriade 1982).49(84) INTERVOCALIC ONSETSa. AFTER A SHORT VOWELVCVb. AFTER A LONG VOWELVCV2.4.1.2 AnalysisIn an Optimality Theory framework, the cross-linguistic typology of syllable shapes is regulated viasyllable markedness constraints ONSET and *CODA, which require onsets and prohibit codas, respec-tively (McCarthy and Prince 1993a; Prince and Smolensky 1993). I utilize a Moraic Theory for syllablestructure (Hayes 1989a; Hyman 1985; Pulleyblank 1994) which does not include structural positionslike “onset” or “coda” nodes. Therefore, I do not make direct reference to onsets and codas. Instead,these definitions are construed of as positional markedness constraints that prohibit particular structuresat the left and right edges of syllables.ONSET is defined in (85) as a categorical markedness constraint (McCarthy 2003). This definitionof ONSET penalizes any syllable whose left edge aligns with a mora, regardless of whether that moradominates a consonant or vowel.(85) ONSETAbbreviation: ONS* XFor each syllable, assign a violation mark if the leftmost segment is dominated by a mora.Onsetless syllables violate this constraint because the moraic syllable nucleus would be aligned withthe left edge of the syllable, but moraic onsets also violate this constraint because the mora associatedwith the onset consonant would be aligned to the left edge of the syllable. This behavior is favorable,since there are no moraic onsets in Blackfoot and moraic onsets are typologically rare (but see Topintzi2010 for evidence in favor of moraic onsets in some languages).50This formulation is superior to constraints that require the syllable to begin with a consonant (Mc-Carthy and Prince 1993a) or prohibit the syllable from beginning with a vowel (Kager 1999; McCarthy2003), both of which require the terms “consonant” and “vowel” to be defined. One natural definitionutilizes the features [+consonantal] for consonants and [-consonantal] for vowels. The problem is thatunder these definitions, intervocalic glides do not satisfy ONSET because glides are [-cons]. However, asI discuss in Section, vowel length is contrastive before glides just as it is for other onsets; glidestherefore seem to satisfy ONSET. The formulation in (85) avoids this problem by defining ONSET interms of moraic segments, rather than consonants or vowels.9The definition of *CODA in (86) below uses a positional markedness constraint to prohibit anysyllable that ends in a (non-glide) consonant, regardless of whether that consonant is the syllable nucleusor not. This makes sense for Blackfoot. As I discuss in Section, moraic [s] can be parsed as asyllable nucleus or coda consonant. It has the same distribution in both positions; namely, both types ofmoraic [s] occur only before obstruents. *CODA has the benefit of incurring a violation for any moraic[s], regardless of whether it occurs as a nucleus in an open syllable, or as a coda.(86) *CODAAbbreviation: *COD* X[+cons] For each syllable, assign a violation mark if the right edge of the syllable is aligned with theright edge of a segment associated with a [+cons] feature.The syllable markedness constraints, ONSET and *CODA, account for why short intervocalic con-sonants are preferentially parsed as onsets: this parse optimally satisfies both constraints. In the tableaubelow, the optimal candidate (a) satisfies both ONSET and *CODA by parsing the intervocalic consonantC as an onset. The competing candidate (b) violates *CODA, because it has a segment C that followsa tautosyllabic moraic segment (V), and ONSET, because the syllable begins with a moraic segment(V). Candidate (b) will always be less optimal than candidate (a), no matter where ONSET is ranked inrelation to faithfulness constraints.9See Smith 2002, 2004, 2012 for a discussion of different formulations of ONSET, and an alternative definition to minethat requires syllables to have a segment, a to the left of the syllable head, regardless of whether a is dominated by a mora ordirectly by the syllable node.51(87). . . VCV. . . ONS *COD a. V.CVb. VC.V ∗! ∗!Distinctive vowel length is represented in moraic theory as a difference in weight. Following Hayes(1989a) & Pulleyblank (1994), I assume that short vowels are associated with one mora in their under-lying representation, (88a), while long vowels are associated with two moras, (88b).(88) a. SHORT VOWELSVb. LONG VOWELSVLong vowels violate a general markedness constraint that prohibits segments from associating totwo moras. Consonants are unable to violate this constraint because consonants are never bimoraicin Blackfoot (and potentially universally), so I have named the constraint *V:, as in Benua (1995),Hammond (1997), Holt (1997), Keer (1999), Prince and Smolensky (1993), Rosenthall (1994), & Sherer(1994).10(89) *LONGVOWELAbbreviation: *V:*XAssign a violation mark for every segment which is associated to more than one mora.To derive the contrast between short and long vowels before simple onsets in Blackfoot, a faith-fulness constraint must rank above the markedness constraint *V:. Faithfulness will be measured usingCorrespondence Theory (McCarthy and Prince 1995, 1999). The correspondence relation can be definedas follows. For our purposes, S1 = input and S2 = output.10There is also a constraint which prohibits moraic consonants, *m/C, which I define and discuss Section 2.4.2. But seeMoren (1999), who explicitly argues against constraints like *LONGVOWEL and *GEMINATE and provides a unified theoryof moraicity across segment types (i.e. both consonants and vowels).52(90) CORRESPONDENCEGiven two strings S1 and S2, correspondence is a relation R from the elements of S1 to those ofS2. Elements a ∈ S1 and b ∈ S2 are referred to as correspondents of one another when aRb .(McCarthy and Prince 1995, 1999)The correspondence constraints that are relevant to long vowels are constraints that regulate corre-spondence between input and output moras: MAX-IO() in (91) and DEP-IO() in (92) below.(91) MAX-IO()Abbreviation: MAX()Given an input string S1 and and output string S2, such that S1 R S2, assign a violation mark forevery mora a ∈ S1 that does not have a correspondent b ∈ S2. (“No mora deletion.”)(McCarthy 2000)(92) DEP-IO()Abbreviation: DEP()Given an input string S1 and and output string S2, such that S1 R S2, assign a violation markfor every mora b ∈ S2 that does not have a correspondent a ∈ S1. (“No mora epenthesis.”)(McCarthy 2000)In a word like [aˆ:ko:ka:wa˚] ‘she’ll sponsor a Sundance’, (82b), the [o] in the second syllable is linkedto two moras in the input. For this vowel to surface faithfully, the faithfulness constraint MAX-IO()must be ranked above *V:. In the tableaux below, I have represented a long vowel (a vowel linked totwo moras) as [V:], and a short vowel (a vowel linked to one mora) as [V]. The optimal candidate (a)violates the constraint *V: because it has a long vowel in the output. Candidate (b) has a short vowelin the output, but incurs a violation of MAX-IO() because there is a mora in the input which does nothave a correspondent in the output.(93) [aˆ:.ko:.ka:.wa˚] áakookaawa ‘she’ll sponsor a Sundance’/a:k-o:ka:-wa/ MAX() *V: a. a:ko:ka:wa˚∗b. a:koka:wa˚∗!Crucial rankings: MAX-IO()≫*V:532.4.2 Closed syllables: vowel length neutralizationHaving shown that vowel length is distinctive in open syllables, I now show that vowel length is neu-tralized to short in closed syllables. I use this distribution to argue that codas are moraic in Blackfoot. DataSyllables can be closed by (a) a dorsal fricative, (b) an alveolar sibilant, (c) a moraically-licensed con-stricted glottis ([CG]) feature (which surfaces as a glottal stop), or (d) a geminate consonant. I describeeach of these in turn. Heterosyllabic clusters [çC] ∼ [xC] ∼ [xwC]Dorsal fricatives only occur immediately before a consonant. They can either occur after a vowel, whichI discuss here, or after a consonant, which I discuss in Section 2.4.3. In a post-vocalic position, the placeand rounding of the consonant is conditioned by the features of the immediately preceding vowel, asdescribed above in Section 2.1.2. The examples in (94) show that all three allophones of the codadorsal fricative can occur before obstruents, but not before nasals or glides. (A post-consonantal dorsalfricative has the same distribution: it can only occur before obstruents.)(94) CODA /x/ BEFORE OBSTRUENTSa. BEFORE [p]Jki>ts´i:kiçpaK kitsííkiihpa? ‘what did you do?’Jnita´ji:>tsImaxpIn:a:nK nitáyiitsimaahpinnaan ‘we are storing food’Joxwp´IniK˚ ohpíni ‘his/her lung’b. BEFORE [t]Jaˆ:>ksImiçtakiwaK˚ áaksimiihtakiwa ‘s/he will make a mess’Jsta´xtoxw>tsiK˚ stááhtoohtsi ‘(towards) underneath’Joxwto´:kIsiK˚ ohtóókisi ‘his/her ear’c. BEFORE [k]Ja>tsik´içka:tK atsikííhkaat ‘buy shoes!’Jna:ma´xkimaP>tsIsK naamááhkimaa’tsis ‘broom’Jo´xwkotokiK˚ óóhkotoki ‘rock’d. BEFORE [s]*JiçsKJaˆ:>ksi:taPpakaxsiwaK˚ áaksiita’pakaahsiwa ‘he will really get hurt’Joxwso´jIsiK˚ ohsóyisi ‘his/her tail’54e. BEFORE [>ts]JkiP>ts´i:ç>tsi:sK ki’tsííhtsiisa ‘put it aloft!’Jistax>tsIstotoxwsInK istaahtsistotoohsin ‘underwear’JIsko´xw>tsIkaK˚ isskóóhtsika ‘in the past’, ‘long ago’f. BEFORE [>ks]Jaˆ:>ksip´iç>ksini:wa´jiK˚ áaksipííhksiniiwáyi ‘she will slit it (e.g. a hide)’Ja´x>ksIk:amaPpamiwaK˚ ááhksikkama’pamiwa ‘it might be him’JIn:o´xw>ksisi:K innóóhksisii ‘elephant’g. BEFORE NASALS*[çm], *[xm], *[xwm]*[çn], *[xn], *[xwn]h. BEFORE GLIDES*[çw], *[xw], *[xww]*[çj], *[xj], *[xwj]Crucially, long vowels never occur before dorsal fricatives. The negative data in (95) uses the [xw]allophone, but the same generalization holds for the other two allophones as well.(95) NO LONG VOWELS BEFORE CODA [x]a. NON-SIBILANT OBSTRUENTS*[o:xwp´Ini˚] — —*[o:xwto´:kIsi˚] — —*[o´:xwkotoki˚] — —b. SIBILANT OBSTRUENTS*[o:xwso´jIsi˚] — —*[Isko´:xw>tsIka˚] — —*[In:o´:xw>ksisi:] — —I adopt a version of the analysis in Elfner (2006b), where /x/ is treated as an independent consonantwhich forms a moraic coda in syllables, as in (96). This analysis also explains why long vowels areprohibited before [xC] clusters: a syllable that contained a long vowel and a coda consonant would betrimoraic. Consequently, vowels are shortened in order to conform to a bimoraic syllable template.55(96) VELAR FRICATIVE AS CODAVxCThis moraic analysis of /x/ cannot account for the severely restricted distribution of [x] before shortobstruents. For this there is a diachronic explanation, since /x/ is the reflex of the first segment in a trueconsonant cluster in Proto-Algonquian (Berman 2006; Weber 2017). The second consonant of a trueconsonant cluster in Proto-Algonquian was always an obstruent. Therefore, the synchronic distributionof /x/ reflects the distribution of pre-consonantal consonants in Proto-Algonquian.Before moving on to discuss the distribution of moraic [s"], it is worth noting the differences be-tween the moraic analysis here and two previous analyses of for /x/. The first analysis does not treat/x/ as an independent segment at all, but only as the phonetic manifestation of pre-aspiration on the fol-lowing consonant (Reis Silva 2008; Weber 2016a,b). Under this proposal, Blackfoot obstruents have athree-way distinction: singleton unaspirated, geminate unaspirated, and preaspirated. This analysis suc-cessfully explains why [x] has such a limited distribution: it can only occur before voiceless obstruentsbecause it is the surface reflex of the preaspiration associated with the following obstruent. The surfacefricatives [ç] ∼ [x] ∼ [xw] under this analysis are a redundant secondary characteristic of pre-aspiration(Silverman 2003; Steriade 1999). However, this analysis fails to account for why pre-aspiration cannotoccur on geminate consonants, nor for why vowel length is neutralized before [x].The second analysis treats [x] as an allophone of /k/, because [x] and [k] occur in complementarydistribution (Kaneko 2000). The [x] occurs post-vocalically before another obstruent while [k] occurspre-vocalically or before an interconsonantal [s]. As I argued above, this means that [x] occurs inPPh-medial syllable codas while [k] occurs in PPh-medial syllable onsets. Additionally, [k] can occurPPh-finally but [x] cannot. She suggests that the velar stop /k/ spirantizes to [x] post-vocalically exceptPPh-finally.11 There are two issues with such an analysis. First, [x] is in complementary distributionwith all short obstruents, nasals, and glides, because all of these sounds have the same distribution as[k]; namely, they occur in syllable onsets. Thus, there is no reason to assume [x] is an allophone of/k/ in particular. Second, Blackfoot actively repairs consonant clusters which arise across morphemeboundaries via two types of epenthesis, and crucially not via spirantization. For example, in (97) the[k]-final prefix [pok-] ‘small’ combines with a [k]-initial noun [kax>tsaP>tsis] ‘(playing) cards’. Instead ofthe prefix-final [k] spirantizing to [x], a high front vowel [i] is epenthesized, which causes the preceding[k] to assibilate to [>ks].11Kaneko (2000) refers to the PPh as the ‘word’.56(97) a. Jkax>tsaP>tsi:s>tsiK˚kaahtsa’tsiistsi[[kaaht–i]–a’tsis]–istsi[[game–AI]–NMLZ]–IN.PL‘playing cards’b. Jpo´>ksikx">tsaP>tsi:s>tsiJ˚póksikahtsa’tsiistsipok–[[ikaht–i]–a’tsis]–istsismall–[[game–AI]–NMLZ]–IN.PL‘little playing cards’(not *Jpo´xwkx">tsaP>tsi:s>tsiK˚)In Section I argue that head-adjoined root /-p/ ‘tie’ begins with an underlying /p/. A highfront vowel [i] is epenthesized after the consonant-final root /aak-/12 ‘wrap with otter’ which causes thepreceding [k] to assibilate to [>ks], (98). Again, the root-final [k] in /aak-/ ‘wrap with otter’ does notspirantize to [x].(98) Jaˆ:ka>ksipista:waK˚áakaksipistaawaaak–[aak–p/ist–aa]–Ø–waFUT–[otter.wrap–tie/v–AI]–(IND)‘he will wrap the pole with the fur of anotter’(not *Jaˆ:kaxpista:waK˚) Heterosyllabic clusters [sC]The alveolar sibilant [s] in Blackfoot can occur in many positions of the word, including next to vowelsas well as next to other consonants, and has been widely discussed in the literature (Denzer-King 2009;Derrick 2006, 2007; Elfner 2006b; Goad and Shimada 2014a,b). However, pre-consonantal [s] hasthe same distribution as pre-consonantal /x/: both occur only before obstruents.13 Examples of pre-consonantal [s] are given in (99). Note that an /s/ before an /s/ is indistinguishable from a geminate,and I assume that this sequence is parsed as a single ambisyllabic geminate (Section (Moraic[s] also occurs between consonants, which I discuss in Section 2.4.3.)12This morpheme has a short vowel after consonants.13The term ‘pre-consonantal [s]’ does not include the pre-assibilant [st], for which see Section 2.1.1. Pre-consonantal [s] islonger in duration than the sibilant portion of [st] and occurs before a wider range of consonants other than just [t].57(99) [s] OCCURS BEFORE OBSTRUENTSa. NON-SIBILANT OBSTRUENTSJIsp´iks":kojiK˚ isspíksskoyi ‘high forest (of tall trees)’Jap2sta´:niK˚ apasstááni ‘bridge’JmU´ski>tsipx"piK˚ mósskitsipahpi ‘heart’b. SIBILANT OBSTRUENTS[kita´PpEs:2m:oka˚] kitá’paissammoka ‘he is looking around for you’[mU´s>tsi:wa˚] mósstsiiwa ‘vein’[otx"k>´ois>ksisi˚] otahkóíssksisi ‘brown-nosed horse’c. NASALS*[sm]*[sn]d. GLIDES*[sw]*[sj]Crucially, there are no examples with long vowels before pre-consonantal [s]. Following the samelogic as above, I will assume that coda [s] is moraic. Long vowels are prohibited before [sC] clustersbecause a syllable that contained a long vowel and a coda consonant would be trimoraic. Consequently,vowels are shortened in order to conform to a bimoraic syllable template. Again, this analysis does notexplain why coda /s/ occurs only before obstruents. For this there is a diachronic explanation, since/s/ is the reflex of the first segment in a true consonant cluster in Proto-Algonquian after PA * i or * i;(Berman 2006; Weber 2017).(100) ALVEOLAR SIBILANT AS CODAVsC582. Heterosyllabic clusters [PC]The glottal stop /P/ occurs before any of the permissible onset consonants. The examples in (101)illustrate [P] before short non-sibilant obstruents, sibilant obstruents, nasals, and glides.14(101) [P] OCCURS BEFORE ONSETSa. NON-SIBILANT OBSTRUENTSJa´PpoPtakiwaK˚ a’po’takiwa ‘he worked’JoPtak´i:waK˚ o’takííwa ‘it is round’JkoPk´ıjiK˚ ko’kíyi ‘(a) corner’b. SIBILANT OBSTRUENTSJsomo´PsItK somó’sit ‘fetch water!’JmaP>ts´ItK ma’tsit ‘take it!’JkaP>ksimo´jiK˚ ka’ksimóyi ‘sage’c. NASALSJapa´PmIsaK˚ apá’misa ‘lull her!’Ji:saPna:waK˚ iisa’naawa ‘he lost a fingernail’d. GLIDESJkik2´taPwa´:nojoPs":paK kikáta’wáánoyo’sspa ‘did you make archery equipment?’JIska´PjIs:iwaK˚ isská’yissiwa ‘he is exceptionally tough’The glottal stop has a variable realization as a full glottal stop, creaky voice, or length on the pre-ceding vowel, as discussed by Peterson (2004) and shown in (102). The exact realization depends onthe speaker, rate of speech, and morphophonological context.(102) SURFACE REALIZATIONS: [VPC] ∼ [VV˜C] ∼ [V:C]aikaiPni ‘He dies’a. [aIkaIPni]b. [aIkaII˜ni]c. [aIkajni] (= Peterson 2004, (3))Crucially, vowel length is not distinctive before a glottal stop. Following the same logic as above, Iwill assume that the glottal stop [P] is moraic. Long vowels are prohibited before [PC] clusters becausea syllable that contained a long vowel and a coda consonant would be trimoraic. Consequently, vowels14The glottal stop also occurs infrequently between vowels (Peterson 2004).59are shortened in order to conform to a bimoraic syllable template. Specifically, I follow the analysislaid out in Peterson (2004). To account for this variable phonetic realization, as well as some unusualmorphophonological processes that the glottal stop undergoes in Blackfoot, Peterson (2004) argues thatthe glottal stop in Blackfoot is a subsegmental constricted glottis ([CG]) feature which is moraicallylicensed. He accounts for the three surface realizations in (102) in the following way. If the [CG] featureis parsed to the final mora of the syllable, the vowel is short and the [CG] feature surfaces as a glottalstop, (103a). If it is parsed to the final mora of the syllable which is also linked to the vowel, then thevowel is long and the [CG] feature surfaces as creaky voicing on the vowel, (103b). If the [CG] featureis left unparsed to the syllable, then the vowel simply surfaces as long, (103c).(103) SUBSEGMENTAL MORAIC [CG]a. [VC]V[CG]Cb. CREAKY VOWELV[CG]Cc. LONG VOWELV[CG]C2. GeminatesLength is distinctive for obstruents and nasals. For example, the words [E´:pota:wa˚] ‘he is getting abeating’ and [E´:pUt:a:wa˚] ‘he is flying’ in (30b) only differ in the length of the medial consonant [t]or [t:] and the quality of the preceding vowel. (For further examples that short and long consonantscontrast, see Appendix A.1.2.)(104) a. JE´:pota:waK˚áípotaawaa–[ipot–aa]–Ø–waIPFV–[beat.v–3OBJ]–IND–3‘he is getting a beating’b. JE´:pUt:a:waK˚áípottaawaa–[ipott–aa]–Ø–waIPFV–[fly–AI]–IND–3‘he is flying’(Denzer-King 2009: 16, (7k,l))Geminates have the same distribution as heterosyllabic clusters; they occur intervocalically andcannot be followed or preceded by another consonant. As discussed in Elfner (2006b) & Frantz (2009),vowel length is neutralized before geminates, and only short, lax vowels can occur in this position.Some examples are shown in (105). In contrast, no long vowels occur before geminates, (106).60(105) SHORT V BEFORE GEMINATESa. NON-SIBILANT OBSTRUENTS[kIp:o˚xwkU´k:It] kippohkókkit ‘please give me it!’ (BB)JmUt:o>ks´isK mottoksís ‘knee’Jnit2´k:a:K nitákkaa ‘my friend’b. SIBILANT OBSTRUENTSJ2´s:aˆ:kIp:o´pi:tK ássáakippópiita ‘stop and sit for a moment!’JkU>t:s´i:sK kottsíísa ‘stuff it!’ (ANIM)J>ksI >k:sin2´>t:siwaK˚ ksikksináttsiwa ‘it is white’c. NASALSJIm:oja´:nK immoyáán ‘fur coat’Jij´In:aki:koanK iyínnakiikoan ‘policeman’d. GLIDES*[j:]*[w:](106) NO LONG V BEFORE GEMINATESa. NON-SIBILANT OBSTRUENTS*[ki:p:o˚xwkU´k:It] — —*[mo:t:o>ks´is] — —*[nita:k:a:] — —b. SIBILANT OBSTRUENTS*[a´:s:aˆ:kIp:o´pi:t] — —*[ko:>t:s´i:s] — —*[>ksi:>k:sin2´>t:siwa˚] — —c. NASALS*[i:m:oja´:n] — —*[ij´i:n:aki:koan] — —d. GLIDES*[j:]*[w:]This pattern suggests that geminates are, like heterosyllabic clusters, parsed to a moraic coda posi-tion. Following Hayes (1989a) I assume that an intervocalic geminate is parsed ambisyllabically to a61moraic coda position as well as to the following syllable node in order to satisfy ONSET, as in (107).Following the same logic as above, long vowels are prohibited before geminates because a syllable thatcontained a long vowel and a coda consonant would be trimoraic. Consequently, vowels are shortenedin order to conform to a bimoraic syllable template.(107) GEMINATE CONSONANTCVCVIn transcriptions with syllable boundaries (.), I write geminates as doubled sounds with the periodin the middle, to parallel the transcription of CC clusters, (109).(108) AMBISYLLABIC GEMINATE[ni.t2´k.ka:] ‘my friend’ =nit2ka(109) HETEROSYLLABIC CLUSTER[koP.ki] ‘corner’ =ko[CG]ki2.4.2.2 AnalysisThe preceding sections discussed four environments where vowel length is neutralized to short: before adorsal fricative, (110), an alveolar sibilant, (111), moraically-licensed [CG] feature, (112), or a geminateconsonant, (113). I argued that all four environments can be united under a moraic analysis: they allcreate a closed syllable with a moraic coda. This causes the preceding vowel to shorten in order toconform to a bimoraic syllable template.(110) VELAR FRICATIVE AS CODAVxC=(96)62(111) ALVEOLAR SIBILANT AS CODAVsC=(100)(112) SUBSEGMENTAL [CG] AS CODAa. [VPC]V[CG]Cb. CREAKY VOWELV[CG]C=(103a), (103b)(113) GEMINATE CONSONANTVC=(107)In the next sections, I address the analysis of vowel neutralization before heterosyllabic clusters andgeminates in turn. Heterosyllabic clustersIn this section I present an Optimality Theory analysis of vowel length neutralization in closed syllablesin Blackfoot. Because onsetless syllables15 and syllables with codas are tolerated, the two faithfulnessconstraints MAX-IO(Seg) and DEP-IO(Seg) must dominate the two markedness constraints ONSET and*CODA.15Onsetless syllables are tolerated only at the beginning of the PPh. PPh-internal onsetless syllables violate the morestringent constraint *HIATUS. See Elfner (2006b) for discussion.63(114) JoP.tak´i:.waK˚ o’takíwa ‘it is round’/o’takii-wa/ MAX(Seg) DEP(Seg) ONS *COD a. [oP.ta.k´i:.wi˚] ∗ ∗b. [toP.ta.k´i:.wi˚] ∗! ∗c. [o.ta.k´i:.wi˚] ∗! ∗Crucial rankings: {MAX(Seg), DEP(Seg)}≫ {ONS, *COD}Codas are predictably moraic, which is regulated by the constraint Weight-By-Position (WBP).Since there are no “coda” structural positions in Moraic Theory, I have formulated WBP as an alignmentconstraint (McCarthy and Prince 1993a). The constraint is satisfied by a mora at the right edge of asyllable regardless of whether that mora is linked to a consonant or a vowel. Therefore, open syllablessatisfy WBP. The constraint is formulated in such a way that degenerate syllables, which contain nomoras, vacuously satisfy this constraint. (Degenerate syllables are discussed in Section 2.4.4.)(115) WEIGHT-BY-POSITION = ALIGN(, R,, R)Abbreviation: WBPFor every syllable () which contains at least one mora, align the right edge of the syllable tothe right edge of a mora. (Based on Hayes 1989b)Any moraic consonant violates a general markedness constraint, *m/C.(116) * [+cons]Abbreviation: *m/CAssign a violation mark for any mora which dominates a [+cons] segment.(Broselow, Chen and Huffman 1997; Moren 1999)The weight of coda consonants is predictable, meaning that non-moraic consonants in the input willbecome moraic in the output if they are parsed to a coda position, thus incurring faithfulness violations.The following tableau for (101) [koPk´ıji˚] ‘(a) corner’ uses an input that contains a short vowel beforea non-moraic glottal stop. The faithful candidate (a) is not optimal, because it violates WBP, whichis undominated. The winning candidate (b) satisfies WBP by epenthesizing a mora. This violatesDEP-IO() because there is a mora in the output that does not have a correspondent in the input. Thiscandidate also violates the general markedness constraint, *m/C. Candidates (c) and (d) avoid violationsof WBP and *CODA by deleting or epenthesizing a segment, respectively.64(117) JkoP.k´i.jiK˚ ko’kíyi ‘(a) corner’/koPki-ji/ WBP MAX(Seg) DEP(Seg) MAX() DEP() *m /C *CODa. koP.k´i.ji˚∗! ∗ b. koPm .k´i.ji˚∗ ∗ ∗c. ko.k´i.ji˚∗!d. ko.Pi.k´i.ji˚∗! ∗Crucial rankings: WBP≫ {DEP-IO(), *m/C}; MAX-IO(Seg)≫ {DEP-IO(), *m/C, *CODA};DEP-IO(Seg)≫ {*m/C, *CODA}Under the Richness of the Base hypothesis (Prince and Smolensky 1993), there are no constraintson the input. The moraicity of codas is predictable, so an input which already has a mora should alsogenerate the same output. The coda in candidate (a) does not retain the mora, violating WBP and MAX-IO(). The optimal candidate retains the mora, confirming that WBP must be crucially ranked above*m/C. Note that since WBP ≫ *m/C, candidate (b) will be optimal no matter where MAX-IO() and*CODA are ranked with respect to WBP. Candidates (c) and (d) avoid violations of *m/C and *CODAvia deletion and epenthesis, respectively. Since MAX-IO(Seg)≫ {*m/C, *CODA}, MAX-IO() is notcrucially ranked; candidate (c) will be suboptimal no matter where MAX-IO() is ranked. And finally,the ranking DEP-IO(Seg)≫ {*m/C, *CODA} ensures that candidate (d) is suboptimal no matter whereMAX-IO() and DEP-IO() rank.16(118) JkoP.k´i.jiK˚ ko’kíyi ‘(a) corner’/koPki-yi/ WBP MAX(Seg) DEP(Seg) MAX() DEP() *m /C *CODa. koP.k´i.ji˚∗! ∗! ∗ b. koPm .k´i.ji˚∗ ∗c. ko.k´i.ji˚∗! ∗!d. ko.Pi.k´i.ji˚∗! ∗! ∗!I now turn to an analysis of vowel length neutralization before clusters. The driving force behindvowel shortening is a maximality constraint on syllables, *3, which is undominated in Blackfoot.16There is a fifth candidate which looks just like candidate (d) but which involves reassociating the mora linked to /P/ in theinput to the following (epenthesized) [i] in the output. This candidate would violate DEP-IO(Seg) as well as constraints whichprohibit the reassociation of moras. For one possible set of definitions against mora reassociation, see McCarthy (2000).65(119) *3μAssign a violation for every syllable which dominates more than two moras. (“No trimoraicsyllables.”) (Based on Prince 1990)The problem can be seen most clearly when the input contains a long vowel followed by a moraicconsonant, such as the input /ko:Pki/ for [koP.k´i] ‘corner’. In the tableau below, the faithful candi-date (a) violates *3. Candidates (b) and (c) satisfy *3 by incurring violations of MAX-IO() anddeleting a mora associated with the consonant [P] and the vowel [o], respectively. Deleting the moraassociated with the consonant in candidate (c) also incurs a violation of WBP, which is undominated.The optimal candidate (b) deletes one mora associated with [o] and satisfies *3 and WBP. This tableaucrucially shows that *3 dominates MAX-IO(). Candidates (d) and (e) satisfy *3 by deleting the [P]and the mora associated with it, and by epenthesizing a vowel, respectively.(120) JkoP.k´i.jiK˚ ko’kíyi ‘(a) corner’/ko:Pmki-yi/ *3μ WBP MAX DEP MAX() DEP() *V: *m /C *CODa. ko:Pm .k´i.ji˚∗! ∗ ∗ ∗ b. koPm .k´i.ji˚∗ ∗ ∗c. ko:P.k´i.ji˚∗! ∗! ∗ ∗d. ko:.k´i.ji˚∗! ∗! ∗e. ko:.Pi.k´i.ji˚∗! ∗! ∗ ∗Crucial rankings: *3≫ MAX-IO()To summarize, vowel length is neutralized before heterosyllabic clusters in order to avoid violationsof *3. Deleting a mora associated with a coda consonant would incur violations of WBP, but delet-ing a mora associated with a vowel avoids violations of *3 without incurring a fatal violation. Thetableaux above used examples which contain a coda [CG] feature, but the same ranking would also workfor syllables with coda /x/ or /s/. In the next section I turn to vowel neutralization before geminateconsonants. Geminate consonantsFollowing Davis (2003, 2011), Hayes (1989a), McCarthy and Prince (1996), Moren (1999), & Spaelti(1994) among others, contrastive consonant length is derived from an underlying moraic contrast. Underthis view, geminate consonants are underlyingly moraic while short consonants are not.66(121) MORAIC REPRESENTATION OF GEMINATESa. SHORT CONSONANTCb. GEMINATE CONSONANTCNon-moraic consonants are parsed preferentially to onset positions in order to satisfy ONSET, whichprohibits moras at the left edge of the syllable. The same constraint causes geminate consonants to beparsed to a coda position, because a moraic onset would violate ONSET. (For arguments that onsetscan contain moraic segments, see Hajek and Goedemans 2006; Hart 1991; Muller 2001; Topintzi 2006,2008.) Ambisyllabic ‘flopping’ ensures the following syllable has an onset (Hayes 1989a).In the tableau below, the solid lines represent the crucial rankings *3≫ MAX-IO() from (120)and MAX-IO() ≫ *LONGVOWEL from (93). Candidate (a) parses a moraic consonant to an onsetposition, violating ONSET but satisfying *CODA, while the optimal candidate (b) parses the moraicconsonant to a coda position, violating *CODA but satisfying ONSET. Since these two candidates other-wise incur the same violations, this shows that ONSET dominates *CODA. Candidate (c) does not havea moraic consonant at all, which avoids violations of ONSET, *m/C, and *CODA, but incurs a violationof MAX-IO(). This shows that MAX-IO≫ {*m/C, *CODA}.(122) JE´:pUt.ta:.waK˚ áípottaawa ‘s/he is flying, flies’/a-ipot-aa-wa/ *3μ ONS MAX() *V: *m /C *CODa. E´:.po.ta:.wa˚∗!∗ ∗∗ ∗ b. E´:.pUt.ta:.wa˚∗ ∗∗ ∗ ∗c. E´:.po.ta:.wa˚∗ ∗! ∗∗Crucial rankings: ONSET≫ *CODA; MAX-IO()≫ {*m/C, *CODA}Under the Richness of the Base hypothesis, there are also inputs that contain a long vowel followedby a moraic consonant. Candidates (a) and (b) are faithful to the moraic associations in the input, vio-lating *3 and ONSET, respectively. Candidates (c) and (d) satisfy *3 by removing a mora associatedwith the vowel or the consonant, respectively, violating MAX-IO() in both cases. Because the optimalcandidate is (c), *V: must dominate *CODA and *m/C.67(123) JE´:pUt.ta:.waK˚ áípottaawa ‘s/he is flying, flies’/a-ipo:t-aa-wa/ *3μ ONS MAX() *V: *m /C *CODa. E´:.po:t.ta:.wa˚∗! ∗ ∗∗∗ ∗ ∗b. E´:.po:.ta:.wa˚∗! ∗!∗ ∗∗∗ ∗ c. E´:.pUt.ta:.wa˚∗ ∗ ∗∗ ∗ ∗d. E´:.po:.ta:.wa˚∗ ∗ ∗∗∗!Crucial rankings: *V:≫ {*m/C, *CODA}Note that it is crucially the ranking of ONSET over MAX-IO() in the tableau above that makescandidate (b) suboptimal, because the only reason candidate (b) violates *3 is that the geminate con-sonant is an onset to a syllable with a long vowel. For words with a short vowel following the geminateconsonant, like [Im.mo.ja´:.n] ‘fur coat’, below, candidate (b) violates ONSET but not *3.(124) JIm.mo.ja´:.nK immoyáán ‘fur coat’/i:moji-aan/ *3μ ONS MAX() *V: *m /C *CODa. i:m.mo.ja´:.n ∗! ∗ ∗∗ ∗ ∗b. i:.mo.ja´:.n ∗!∗ ∗∗ ∗ c. Im.mo.ja´:.n ∗ ∗ ∗ ∗ ∗d. i:.mo.ja´:.n ∗ ∗ ∗∗!Crucial rankings: ONSET≫ MAX-IO(); *V:≫ {*m/C, *CODA}This ranking of general markedness constraints (*V: ≫ {*CODA, *m/C}) means that long vowelsin Blackfoot are more marked than moraic consonants. An alternative analysis would instead index thefaithfulness constraints MAX-IO() and DEP-IO() to be specific to vowels versus consonants. Undersuch an analysis, faithfulness to moraic consonants would rank higher than faithfulness to bimoraicvowels. Now that I have shown an analysis of vowel length neutralization in all types of closed syllables,I briefly discuss how to account for the restrictions on the first consonant in a heterosyllabic clusters.The current constraint ranking predicts that illicit consonant sequences will be repaired via epenthesis,which I argue in Section 4.2.1 is true. Restrictions on clustersThe analysis above accounts for vowel length neutralization in closed syllables but does not account forwhy the first consonant in a heterosyllabic cluster is restricted to /P/, /x/, or /s/. The only allowable68pre-consonantal consonants are the voiceless continuants (e.g. /x/ and /s/) or a subsegmental unit whichis unspecified for voice and continuancy (e.g. [CG]). However, all possible geminates are allowed.I hypothesize that some markedness constraints apply to specific clusters, and that geminates escapethe effects of these constraints because they are not clusters, but single consonants linked to a mora.There are two relevant constraints. The first is *[-cont][+cons], (125), which prohibits clusters wherethe first segment is a plosive or assibilant. The second is AGREE(voice), (126), which requires adjacentconsonants to have the same value for voicing.(125) *[-cont][+cons]Assign a violation for each sequence of consonants if the first consonant is [-cont].(126) AGREE(voice)Assign a violation for each sequence of consonants if the first consonant is [avoice] and thesecond consonant is [-avoice].A cluster like /sk/ or /xk/, (127), does not violate either markedness constraint. Such clusterssatisfy *[-cont][+cons] because [s] and [ç], [x], [xw] are not [-cont] segments. Such clusters satisfyAGREE(voice) because /s/ and /x/ are both [-voice] and only occur before obstruents, all of which arealso [-voice] in Blackfoot. The clusters surfaces faithfully, even though the optimal candidate incurs aviolation of *CODA. Note that (117) already determined that DEP-IO(Seg)≫ *CODA, so candidate (b)is suboptimal no matter where DEP-IO() ranks.(127). . . sk. . . *[-cont][+cons] AGR(voi) MAX DEP MAX() DEP() *COD a. s.k ∗b. si.k ∗! ∗!c. Ø.k ∗!Crucial rankings: MAX-IO(Seg)≫ *CODAA cluster like /Pk/, (128), also violates neither markedness constraint, because [P] is a subsegmental[CG] feature which has no [voice] or [cont] features (Peterson 2004). Both markedness constraints aresatisfied regardless of the features of the following segment. This predicts that /P/ should be able tooccur freely before any licit onset, which is true (see data in Section . . Pk. . . *[-cont][+cons] AGR(voi) MAX DEP MAX() DEP() *COD a. P.k ∗b. Pi.k ∗! ∗!c. Ø.k ∗!Crucial rankings: MAX-IO(Seg)≫ *CODAA cluster like /pk/, (129), violates *[-cont][+cons] because [p] is a [-cont] segment. Since (117)shows that DEP-IO(Seg) ≫ *CODA, the tableau below shows that *[-cont][+cons] crucially domi-nates DEP-IO(Seg), and that either *[-cont][+cons] or *CODA must dominate DEP-IO(). The optimalcandidate (b) violates DEP-IO(Seg) and DEP-IO(). An alternative candidate (c) avoids violations of*[-cont][+cons] by deleting one of the consonants, which shows that MAX-IO(Seg) outranks the DEPconstraints.(129). . . pk. . . *[-cont][+cons] AGR(voi) MAX DEP MAX() DEP() *CODa. p.k ∗! ∗ b. pi.k ∗ ∗c. Ø.k ∗!Crucial rankings: *[-cont][+cons] ≫ DEP-IO(Seg); MAX-IO(Seg) ≫ {DEP-IO(Seg), DEP-IO()}A cluster like /nk/, (130), violates AGREE(voice) because the nasal is [+voice] while the follow-ing plosive is [-voice].17 Since (117) shows that DEP-IO(Seg) ≫ *CODA, the tableau below showsthat AGREE(voice) crucially dominates DEP-IO(Seg), and that either AGREE(voice) or *CODA mustdominate DEP-IO(). The optimal candidate (b) violates DEP-IO(Seg) and DEP-IO(). An alternativecandidate (c) avoids violations of AGREE(voice) by deleting one of the consonants, which shows thatMAX-IO(Seg) outranks the DEP constraints. In Chapter 4 I present concrete evidence that illicit clustersare avoided via epenthesis rather than deletion.17Nasals in Blackfoot pattern as a class with [s] for at least one alternation: all three segments delete at the right edge ofnoun stems in particular morphophonological contexts (Frantz 2009; Weber 2016d). Nasals in many languages pattern with[+cont] segments (Mielke 2008), and so I have assumed that nasals in Blackfoot are [+cont]. No noun stems end in /x/, /w/,or /j/, which means that the only [+cont] segments which occur at the end of noun stems is exactly the class {m, n, s}.70(130). . . nk. . . *[-cont][+cons] AGR(voi) MAX DEP MAX() DEP() *CODa. n.k ∗! ∗ b. ni.k ∗ ∗c. Ø.k ∗!Crucial rankings: AGREE(voice)≫DEP-IO(Seg); MAX-IO(Seg)≫ {DEP-IO(Seg), DEP-IO()}Geminates are parsed to a coda position and the optimal, faithful candidate violates *CODA, (131).Candidate (b) violates DEP-IO(Seg) and DEP-IO(). This creates a marked vowel hiatus context, thusviolating a constraint like *HIATUS. (See discussion in Section for this constraint.) Candi-date (c) violates MAX-IO() in order to resyllabify the consonant as an onset.(131). . . VpV. . . *HIAT AGR(voi) MAX DEP MAX() DEP() *COD a. Vp.pV ∗b. Vp.pi.V ∗! ∗! ∗!c. V.pV ∗!Crucial rankings: MAX-IO()≫ *CODAIn the final two sections of this chapter I turn to two unusual aspects of syllable structure whichfeature in the transcriptions throughout the dissertation. In Section 2.4.3 I discuss the distribution ofinterconsonantal fricatives. This discussion augments previous research which argues that interconso-nantal fricatives in Blackfoot are parsed to a syllable nucleus on the basis of distribution and patterns ofalternation (Denzer-King 2009; Goad and Shimada 2014a,b; Miyashita 2018). In Section 2.4.4 I discussdegenerate syllables, which I define as a syllable without a nucleus. These syllables occur only at theright edge of the PPh in Blackfoot and so serve as evidence for that edge.2.4.3 Syllabic fricativesOnly two types of consonants can occur between consonants in Blackfoot: /x/ and /s/. The transcrip-tions throughout this dissertation treat interconsonantal /x/ and /s/ as syllabic consonants. This sectionincludes the data which supports this analysis. I discuss interconsonantal /x/ and interconsonantal /s/in turn.Dorsal fricatives can occur between two consonants. The examples in (132) show that all threeallophones of the dorsal fricative occur before obstruents but not before nasals or glides. This meansthat interconsonantal /x/ has the same distribution as post-vocalic (coda) /x/ (discussed above in Sec-tion NUCLEAR /x/ BEFORE OBSTRUENTSa. BEFORE [p]Jni>ts´ı:pç"pIn:a:nK nitsíípihpinnaan ‘we have/had bows and arrows’Jma´:ks":tx"pikoPto:waK˚ mááksstahpiko’toowa ‘why did she come here?’Jaˆ:kxw"pejPpi:jiwaK˚ áakohpai’piiyiwa ‘she will jump’b. BEFORE [t]Jaˆ:kxw"pç"ta:waK˚ áakohpihtaawa ‘she will use sugar’Jaˆ:>kspx"taniPsiwaK˚ áaksipahtani’siwa ‘she will accidentally cut her hand’JinPç"kxw"to´mo:sK inihkohtómoosa ‘sing for him!’c. BEFORE [k]Jaˆ:kç"ki>tsikawaK˚ áakihkitsikawa ‘she will freeze her feet’JaPkx"ko´jK a’kahkóyi ‘curved geographical feature’Jn´itxw"kUk:aK˚ nítohkokka ‘she gave to me’d. BEFORE [s]*[çs"]JIs:kx"s´ıP>tsitK isskahsí’tsit ‘forget about it!’Jaˆ:kxw"simaK˚ áákohsima ‘she boiled it’e. BEFORE [>ts]Jits":a´pç">tsiPpaK˚ itssápihtsii’pa ‘there is something inside of it’Jaˆ:>ksipx">tsaPps"siwaK˚ áaksipahtsa’pssiwa ‘she will make a mistake’Jaˆ:kxw">tsitoko:waK˚ áakohtsitokoowa ‘he will go through (e.g. a crowd)’f. BEFORE [ks]*[ç>ks"]Jaˆ:>ksipx">ksimIm:aK˚ áaksipahksimimma ‘he will smell like urine’Jaˆ:>ksipxw">ksi:mIn:akiwaK˚ áaksipohksiiminnakiwa ‘she will prune’g. BEFORE NASALS*[çm], *[çn]*[xm], *[xn]*[xwm], *[xwn]72h. BEFORE GLIDES*[çj], *[çw]*[xj], *[xw]*[xwj], *[xww]When the dorsal fricative occurs in a nuclear position, the place and rounding of the consonant isconditioned by the immediately preceding vowel in the underlying representation.18 Morphophonolog-ical alternations support such abstract representations. This can be illustrated with a minimal pair usingthe verb stems /opi:-/ ‘sit’ and /opi-/ ‘possess archery equipment’, as discussed in Frantz (1997: 6). Thestem-final vowel of /opi:-/ ‘sit’ is long before the glide [w] in (133a). This vowel surfaces as a shortvoiced vowel before the indicative clause-typing suffix -hp (/-xp/) and the dorsal fricative is parsed asa coda [ç], (133b). In other words, long vowels in open syllables alternate with short vowels before thedorsal fricative /x/: [V:] ∼ [V]. (This length alternation also occurs before other coda consonants andgeminates, as I discussed in Section 2.4.2.)(133) a. Jaˆ:kopi:waK˚áakopiiwaaak–[op–ii]–Ø–waFUT–[sit–AI]–IND–3‘she will sit’ (Frantz and Russell 2017)b. Jni>ts´i:piçpIn:a:nKnitsíípiihpinnaan20nit–[ii\op–ii]–hp–innaan1–[IC\sit–AI]–IND–1PL‘we sat/stayed’ (Frantz 1997: 6)In contrast, the stem-final vowel of /opi-/ ‘possess archery equipment’ is short before the glide [w]in (134a). When followed by the dorsal fricative in the indicative clause-typing suffix -hp (/-xp/), theunderlying sequence /ix/ surfaces as a nuclear [ç"], as shown in (134b). Alternations like this one lendcredence to the idea that a nuclear dorsal fricative arises from an underlying sequence of a short vowelplus a consonant.18This interaction is not opaque, even though the conditioning vowel is not present in the output. Miyashita (2018) arguesthat nuclear dorsal consonants could be viewed as the product of coalescence between an underlying vowel-consonant se-quence. The details of this proposal in OT remain to be worked out, but such a proposal would circumvent opacity, which isnot analyzable in parallel OT.20The first vowel of the stem in this example and also in (134b) exhibits a vowel ablaut to [i:] which occurs for stems whichbegin in a short vowel whenever they occur in past tense indicative clauses for third persons; see Appendix C.73(134) a. Jaˆ:kopiwaK˚áakopiwaaak–[op–i]–Ø–waFUT–[archery.gear–AI]–IND–3‘she will have archery equipment’(Frantz and Russell 2017)b. Jni>ts´i:pç"pIn:a:nKnitsíípihpinnaannit–[ii\op–i]–hp–innaan1–[IC\archery.gear–AI]–IND–1PL‘we (excl.) have/had bows and arrows’(Frantz and Russell 2017)I treat interconsonantal /x/ as a monomoraic fricative which forms the nucleus of the syllable.(135) MONOMORAIC NUCLEIC /x/CxThe alveolar sibilant [s] can also occur between two consonants (Goad and Shimada 2014a,b), whereit may be short, (136), or long, (137). Just like post-vocalic (coda) [s] (discussed in Section,interconsonantal [s] only occurs before obstruents.(136) INTERCONSONANTAL [s"] OCCURS BEFORE OBSTRUENTSa. NON-SIBILANT OBSTRUENTSJaˆ:>ksi>ks´ists"piPtakiwaK˚ áaksiksístspi’takiwa ‘she will be eager to begin’Jaˆ:koks"takiwaK˚ áakokstakiwa ‘she will read’JsE:ps"ka´pato:tK saipskápatoot ‘stretch it!’b. SIBILANT OBSTRUENTSJnis´ims"si:s>tsiK˚ nisímssiistsi ‘my drinks’JO:ks">ts´itK aokstsít ‘gnaw a hole in it!’Jaˆ:kits">ks´iPpojiwaK˚ áakitsksí’poyiwa ‘she will imply/insinuate’c. NASALS*[Csm]*[Csn]d. GLIDES*[Csj]*[Csw]74(137) INTERCONSONANTAL [s":] OCCURS BEFORE OBSTRUENTSa. NON-SIBILANT OBSTRUENTSJ´i:ks":pi:waK˚ ííksspiiwa ‘it (wooden) is high’JaPps":to´:sK a’psstóósa ‘beckon to her!’JIsts":ka´:nK isstsskáán ‘dust’b. NON-SIBILANT OBSTRUENTSJo´ts":si>tsima:niK˚ ótsssitsimaani ‘his/her baby’Ja:ps":>ts´i>tsikinitK aapsstsítsikinit ‘take off your shoes!’Jn´its":>ksiniPpaK˚ nítssksinii’pa ‘I know it’c. NASALS*[Cs:m]*[Cs:n]d. GLIDES*[Cs:j]*[Cs;w]I follow Goad and Shimada (2014a,b) in treating interconsonantal [s] as a mono- or bimoraic sibilantwhich forms the nucleus of the syllable.(138) a. MONOMORAIC NUCLEIC /s/Csb. BIMORAIC NUCLEIC /s/CsTo summarize, preconsonantal /x/ and /s/ have the same distributional restrictions, regardless ofwhether they occur after a vowel or a consonant. Under my analysis, this is because the constraints*[-cont][+cons] and AGREE(voice) (defined in Section are satisfied by a sequence of /x/ or/s/ followed by an obstruent. A moraic analysis of interconsonantal segments accounts for the evidenceabove from alternations and segmental distribution (see also Denzer-King 2009; Goad and Shimada2014a,b).A full analysis of Blackfoot moraic consonants goes well beyond the scope of this thesis. Forexample, why is it that the only consonants which can be parsed to as a syllable nucleus are [x] and [s],but more sonorous segments cannot? And why does