{"@context":{"@language":"en","Affiliation":"http:\/\/vivoweb.org\/ontology\/core#departmentOrSchool","AggregatedSourceRepository":"http:\/\/www.europeana.eu\/schemas\/edm\/dataProvider","Campus":"https:\/\/open.library.ubc.ca\/terms#degreeCampus","Creator":"http:\/\/purl.org\/dc\/terms\/creator","DateAvailable":"http:\/\/purl.org\/dc\/terms\/issued","DateIssued":"http:\/\/purl.org\/dc\/terms\/issued","Degree":"http:\/\/vivoweb.org\/ontology\/core#relatedDegree","DegreeGrantor":"https:\/\/open.library.ubc.ca\/terms#degreeGrantor","Description":"http:\/\/purl.org\/dc\/terms\/description","DigitalResourceOriginalRecord":"http:\/\/www.europeana.eu\/schemas\/edm\/aggregatedCHO","FullText":"http:\/\/www.w3.org\/2009\/08\/skos-reference\/skos.html#note","Genre":"http:\/\/www.europeana.eu\/schemas\/edm\/hasType","GraduationDate":"http:\/\/vivoweb.org\/ontology\/core#dateIssued","IsShownAt":"http:\/\/www.europeana.eu\/schemas\/edm\/isShownAt","Language":"http:\/\/purl.org\/dc\/terms\/language","Program":"https:\/\/open.library.ubc.ca\/terms#degreeDiscipline","Provider":"http:\/\/www.europeana.eu\/schemas\/edm\/provider","Publisher":"http:\/\/purl.org\/dc\/terms\/publisher","Rights":"http:\/\/purl.org\/dc\/terms\/rights","RightsURI":"https:\/\/open.library.ubc.ca\/terms#rightsURI","ScholarlyLevel":"https:\/\/open.library.ubc.ca\/terms#scholarLevel","Supervisor":"http:\/\/purl.org\/dc\/terms\/contributor","Title":"http:\/\/purl.org\/dc\/terms\/title","Type":"http:\/\/purl.org\/dc\/terms\/type","URI":"https:\/\/open.library.ubc.ca\/terms#identifierURI","SortDate":"http:\/\/purl.org\/dc\/terms\/date"},"Affiliation":[{"@value":"Arts, Faculty of","@language":"en"},{"@value":"Linguistics, Department of","@language":"en"}],"AggregatedSourceRepository":[{"@value":"DSpace","@language":"en"}],"Campus":[{"@value":"UBCV","@language":"en"}],"Creator":[{"@value":"Anghelescu, Andrei","@language":"en"}],"DateAvailable":[{"@value":"2022-03-18T15:45:22Z","@language":"en"}],"DateIssued":[{"@value":"2022","@language":"en"}],"Degree":[{"@value":"Doctor of Philosophy - PhD","@language":"en"}],"DegreeGrantor":[{"@value":"University of British Columbia","@language":"en"}],"Description":[{"@value":"This dissertation describes primary data illustrating phonological alternations in Nata, a Lacustrine Bantu language spoken in the Mara region of Tanzania. The data presented in this work was elicited from native speakers of Nata, both in Canada, and in Tanzania. The primary focus of the dissertation is in describing and analyzing patterns of tone and vowel harmony in nominal forms. These particular phenomena were chosen because they are bounded by prosodic domains. I propose that these types of phonological patterns can be analyzed in an emergent framework, Lexical Allomorphy (Archangeli & Pulleyblank, 2015, 2016, 2017). This framework allows for phonotactic restrictions that apply globally in the phonological grammar, but also morpheme specific conditions which characterize the distribution of particular sounds with respect to that morpheme. Throughout the dissertation I demonstrate that such a framework is sufficient to account for a variety of forms.","@language":"en"}],"DigitalResourceOriginalRecord":[{"@value":"https:\/\/circle.library.ubc.ca\/rest\/handle\/2429\/80971?expand=metadata","@language":"en"}],"FullText":[{"@value":"Prosodic phonology in NatabyAndrei AnghelescuB.A., Boston University, 2010M.A., The University of Toronto, 2011A 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)March 2022\u00a9 Andrei Anghelescu, 2022The following individuals certify that they have read, and recommend to the Faculty of Graduate andPostdoctoral Studies for acceptance, the dissertation entitled:Prosodic phonology in Natasubmitted by Andrei Anghelescu in partial fulfillment of the requirements for the degree of Doctor of Phi-losophy in Linguistics.Examining Committee:Douglas Pulleyblank, Professor, Linguistics, UBCSupervisorRose-Marie De\u00b4chaine, Professor, Linguistics, UBCSupervisory Committee MemberGunnar O\u00b4lafur Hansson, Associate Professor, Linguistics, UBCSupervisory Committee MemberAnne-Michelle Tessier, Associate Professor, Linguistics, UBCUniversity ExaminerPatrick Moore, Associate Professor, Anthropology, UBCUniversity ExamineriiAbstractThis dissertation describes primary data illustrating phonological alternations in Nata, a Lacustrine Bantulanguage spoken in the Mara region of Tanzania. The data presented in this work was elicited from nativespeakers of Nata, both in Canada, and in Tanzania. The primary focus of the dissertation is in describingand analyzing patterns of tone and vowel harmony in nominal forms. These particular phenomena werechosen because they are bounded by prosodic domains. I propose that these types of phonological patternscan be analyzed in an emergent framework, Lexical Allomorphy (Archangeli & Pulleyblank, 2015, 2016,2017). This framework allows for phonotactic restrictions that apply globally in the phonological grammar,but also morpheme specific conditions which characterize the distribution of particular sounds with respectto that morpheme. Throughout the dissertation I demonstrate that such a framework is sufficient to accountfor a variety of forms.iiiLay SummaryThis dissertation discusses sound patterns in Nata, a minority language of Tanzania. The data in this disser-tation is largely novel. The primary focus of this work is the restrictions on which syllable has high pitch(tone) within a word and which types of vowels can occur together within a word. The theoretical contri-bution of this dissertation is to demonstrate that an emergent approach to a phonological grammar yields asatisfactory analysis, while identifying puzzles in the system and challenging some notions of the depen-dency between parts of a word. The empirical contribution of this dissertation is an extensive description ofthe variety of ways a single unit of meaning (morpheme) can be pronounced in Nata.ivPrefacePortions of chapter 4 were published as Anghelescu, Gambarage, Lam, and Pulleyblank (2017). I was re-sponsible for primary data collection on nominal forms and assisted in writing sections relevant to that datain the aforementioned article. Gambarage was a linguistic consultant, assisted in the conceptual develop-ment, writing, and editing of the article. Lam was responsible for the primary data collection on verbal formsand assisted in the conceptual development of that section, as well as the writing and editing of the article.Pulleyblank assisted with the conceptual development, writing, and editing of the article. The fieldworkreported on in chapters 4 and 5 was covered by UBC Ethics Certificate # H16-01726.vTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvAbbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviSymbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .xviii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1.1 Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1.2 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.3 Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.3.1 Preliminary work and NaWoG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.3.2 Fieldwork . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3vi1.4 Nata Demographics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.1 Nata Phonology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.1.1 Segments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.1.2 Syllables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.1.3 Dahl\u2019s law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162.1.4 Prenasalized consonants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.2 Theoretical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.2.1 Phonological Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.2.2 Lexical Allomorphy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.2.3 Alignment Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272.2.4 Morph preference conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303 Syntax to Prosody Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.1 Syntactic Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.2 Prosodic Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.3 Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353.4 Internal nominal syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363.4.1 nP=\u03c9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.4.2 NumP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.4.3 DP=\u03c6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413.5 External nominal syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453.5.1 Associative Prefix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453.5.2 Locative Prefix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483.5.3 Comitative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493.6 Mapping Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494 Tone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 504.1 Tone in Phonology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50vii4.2 Background on tone in Nata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524.3 Prosodic and syntactic structure of nominals . . . . . . . . . . . . . . . . . . . . . . . . . . 574.3.1 Bare nominals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 574.3.2 Associative Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 574.4 Low tone class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594.4.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594.4.2 Classes 5, 9, 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654.4.3 Data summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 784.4.4 Analysis summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 794.4.5 Morph set relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 804.5 Final H tone class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 824.5.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 824.5.2 Classes 5, 9, 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 864.5.3 Data summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 934.5.4 Analysis summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 934.5.5 Morph set relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 944.6 Initial H tone class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 944.6.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 944.6.2 Classes 5, 9, 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1014.6.3 Data Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1144.6.4 Analysis summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1154.6.5 Morph set relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1164.7 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1185 Vowel Harmony . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1195.1 \u03c9 vowel harmony . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1205.2 \u03c6 vowel harmony . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1235.2.1 Analysis: \u03c6 vowel harmony . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1245.2.2 Prepositional prefixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1295.2.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131viii5.3 Retracting suffix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1335.4 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1426 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1446.1 Why emergence? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1446.2 Emergence contrasted . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1456.3 Nata and Jita . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1516.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156A Vowel hiatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162A.1 Prepositional prefixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162A.2 Locative prefixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165B Tone Evaluation Tableaux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168B.1 Low tone class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168B.1.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168B.1.2 Class 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170B.1.3 Class 9\/10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172B.2 Final H tone class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176B.2.1 Class 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178B.3 Initial H tone class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183B.3.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183B.3.2 Class 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187B.3.3 Class 9\/10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191ixList of TablesTable 2.1 Phonemic Vowel Inventory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9Table 2.2 Vowel minimal pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10Table 2.3 Vowel pairings with reference to examples in Table 2.2 . . . . . . . . . . . . . . . . . . 11Table 2.4 Surface Consonant Inventory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12Table 2.5 CV and CVV syllable shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14Table 2.6 Distribution of vowel length with respect to segmental context . . . . . . . . . . . . . . . 14Table 2.7 Distribution of long vowels with respect to prosodic position . . . . . . . . . . . . . . . 15Table 2.8 Onsetless syllables utterance initially, and medially (in rapid and careful speech) . . . . . 15Table 2.9 Utterance-initial onsetless syllables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16Table 2.10 Voiceless velars in prefixes before stems with initial voiced consonants . . . . . . . . . . 16Table 2.11 Voiced velars in prefixes before stems with initial voiceless consonants . . . . . . . . . . 17Table 2.12 Class 20 velar prefix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17Table 2.13 Sequences of velar prefixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18Table 2.14 Prenasalized obstruents within morphemes . . . . . . . . . . . . . . . . . . . . . . . . . 18Table 2.15 Stems alternate between plain and prenasalized initial consonants . . . . . . . . . . . . . 19Table 2.16 Noun class 3 prefixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23Table 3.1 Lexical entries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32Table 3.2 Key to prosodic constituents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34Table 3.3 The root Gaamb with different nominalizers . . . . . . . . . . . . . . . . . . . . . . . . 37Table 3.4 Stem suffix -i in the nominal domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38Table 3.5 Stem suffix -O in the nominal domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38xTable 3.6 Stem suffix -u in the nominal domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39Table 3.7 Stem suffix -a in the nominal domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39Table 3.8 Al(nP; \u03c9) and the \u03c9-word: C-initial nP . . . . . . . . . . . . . . . . . . . . . . . . . . . 40Table 3.9 Noun class prefixes (=Num) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41Table 3.10 Overt and phonologically null determiners . . . . . . . . . . . . . . . . . . . . . . . . . 42Table 3.11 ri-i-CV(CV. . . ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43Table 3.12 ri- is a determiner: ri-i-CVCV \u223c i-CVCV . . . . . . . . . . . . . . . . . . . . . . . . . 43Table 3.13 Al(DP; \u03c6 ) and the \u03c6 -phrase: C-initial DP (c5) . . . . . . . . . . . . . . . . . . . . . . . 43Table 3.14 c10: tSaa-[n, m, N, \u00f1, \/0]-CV(CV. . . ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44Table 3.15 tSaa- is a determiner: tSaa-(N-). . .\u223c (N-). . . . . . . . . . . . . . . . . . . . . . . . . . . . 44Table 3.16 Al(DP; \u03c6 ) and the \u03c6 -phrase: C-initial DP (C10) . . . . . . . . . . . . . . . . . . . . . . 44Table 3.17 Alignment of \u03c6 -phrase edge in c1 with non-overt determiner . . . . . . . . . . . . . . . 45Table 3.18 Associative construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Table 3.19 Consonant initial DPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47Table 3.20 Associative prefixes on c5 nouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47Table 3.21 Structure of a locative PP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48Table 3.22 Comitative PPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49Table 4.1 Tone properties (Yip, 2001 p. 65) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51Table 4.2 Tone by vowel length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52Table 4.3 Tuki forms with tone bearing nasal (Hyman & Biloa, 1992) . . . . . . . . . . . . . . . . 52Table 4.4 Realization of H tone on heavy syllables . . . . . . . . . . . . . . . . . . . . . . . . . . 54Table 4.5 Only falling H on penultimate long vowels . . . . . . . . . . . . . . . . . . . . . . . . . 54Table 4.6 Root allomorphs with penultimate long vowels . . . . . . . . . . . . . . . . . . . . . . . 54Table 4.7 c5 & c10 forms with alternating HL\u223cHH . . . . . . . . . . . . . . . . . . . . . . . . . . 55Table 4.8 Prefix allomorphs with long vowels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55Table 4.9 Prosodic structure of a simple nominal . . . . . . . . . . . . . . . . . . . . . . . . . . . 57Table 4.10 Complex nominals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58Table 4.11 Prosodic structure of a complex nominal . . . . . . . . . . . . . . . . . . . . . . . . . . 59Table 4.12 Underived L-tone root with overt determiner . . . . . . . . . . . . . . . . . . . . . . . . 59xiTable 4.13 Underived L-tone root with null determiner . . . . . . . . . . . . . . . . . . . . . . . . . 60Table 4.14 Derived L-tone root with overt determiner . . . . . . . . . . . . . . . . . . . . . . . . . 60Table 4.15 Derived L-tone root with null determiner . . . . . . . . . . . . . . . . . . . . . . . . . . 61Table 4.16 Associative L-tone root with overt determiner . . . . . . . . . . . . . . . . . . . . . . . 63Table 4.17 Associative L-tone root with null determiner . . . . . . . . . . . . . . . . . . . . . . . . 63Table 4.18 Class 5 structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65Table 4.19 L-tone c5\/6 roots with overt determiner . . . . . . . . . . . . . . . . . . . . . . . . . . . 66Table 4.20 L-tone c5\/6 roots with null determiner . . . . . . . . . . . . . . . . . . . . . . . . . . . 66Table 4.21 Associative L-tone c5\/6 roots with overt determiner . . . . . . . . . . . . . . . . . . . . 68Table 4.22 Associative L-tone c5\/6 roots with null determiner . . . . . . . . . . . . . . . . . . . . . 68Table 4.23 L-tone c9\/10 roots with overt determiner . . . . . . . . . . . . . . . . . . . . . . . . . . 70Table 4.24 L-tone c9\/10 roots with null determiner . . . . . . . . . . . . . . . . . . . . . . . . . . . 71Table 4.25 Associative L-tone c9\/10 roots with overt determiner . . . . . . . . . . . . . . . . . . . . 74Table 4.26 Associative L-tone c9\/10 roots with null determiner . . . . . . . . . . . . . . . . . . . . 75Table 4.27 Summary: Low tone root class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79Table 4.28 Underived Final H-tone roots with overt determiner . . . . . . . . . . . . . . . . . . . . 82Table 4.29 Underived Final H-tone roots with phonologically null determiner . . . . . . . . . . . . . 83Table 4.30 Associative final H-tone root with overt determiner . . . . . . . . . . . . . . . . . . . . . 84Table 4.31 Associative final H-tone root with null determiner . . . . . . . . . . . . . . . . . . . . . 84Table 4.32 Final H-tone c5\/6 roots with overt determiner . . . . . . . . . . . . . . . . . . . . . . . . 86Table 4.33 Final H-tone c5\/6 roots with null determiner . . . . . . . . . . . . . . . . . . . . . . . . 86Table 4.34 Associative final H-tone c5\/6 roots with overt determiner . . . . . . . . . . . . . . . . . 87Table 4.35 Associative final H-tone c5\/6 roots with null determiner . . . . . . . . . . . . . . . . . . 88Table 4.36 Final H-tone class 9\/10 roots with overt determiner . . . . . . . . . . . . . . . . . . . . . 89Table 4.37 Final H-tone class 9\/10 roots with null determiner . . . . . . . . . . . . . . . . . . . . . 89Table 4.38 Associative Final H-tone c9\/10 roots with overt determiner . . . . . . . . . . . . . . . . 91Table 4.39 Associative Final H-tone c9\/10 roots with null determiner . . . . . . . . . . . . . . . . . 91Table 4.40 Summary: Final H tone class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93Table 4.41 Underived Initial H-tone roots with overt determiner . . . . . . . . . . . . . . . . . . . . 94Table 4.42 Underived Initial H-tone roots with null determiner . . . . . . . . . . . . . . . . . . . . 95xiiTable 4.43 Derived Initial H-tone roots with overt determiner . . . . . . . . . . . . . . . . . . . . . 95Table 4.44 Derived Initial H-tone roots with null determiner . . . . . . . . . . . . . . . . . . . . . . 95Table 4.45 Associative Initial H-tone root with overt determiner . . . . . . . . . . . . . . . . . . . . 98Table 4.46 Associative Initial H-tone root with null determiner . . . . . . . . . . . . . . . . . . . . 99Table 4.47 Initial H-tone Class 5 roots with overt determiner . . . . . . . . . . . . . . . . . . . . . 102Table 4.48 Initial H-tone Class 5 roots with null determiner . . . . . . . . . . . . . . . . . . . . . . 102Table 4.49 Initial H-tone roots, c5, associative with overt determiner . . . . . . . . . . . . . . . . . 104Table 4.50 Initial H-tone roots, c5, associative with phonologically null determiner . . . . . . . . . . 105Table 4.51 Initial H-tone Class 9\/10 roots with overt determiner . . . . . . . . . . . . . . . . . . . . 108Table 4.52 Initial H-tone Class 9\/10 roots with null determiner . . . . . . . . . . . . . . . . . . . . 108Table 4.53 Initial H-tone roots, c9\/10, associative with overt determiner . . . . . . . . . . . . . . . . 111Table 4.54 Initial H-tone roots, c9\/10, associative with null determiner . . . . . . . . . . . . . . . . 111Table 4.55 Summary: Initial H tone class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115Table 5.1 Monomorphemic \u03c9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120Table 5.2 \u03c9 including [-O] nominalizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121Table 5.3 \u03c9 including [-a] nominalizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121Table 5.4 Evaluation of \u03c9 -HARMONY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122Table 5.5 Class prefix vowel quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123Table 5.6 Determiner vowel quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124Table 5.7 Evaluation of \u03c6 -HARMONY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125Table 5.8 Comitative prefix vowel quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130Table 5.9 Locative vowel quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130Table 5.10 Associative vowel quality (i) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131Table 5.11 Associative vowel quality (ii) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131Table 5.12 Retracting suffixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134Table 5.13 Retracted prepositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134Table 5.14 Non-retracted forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135Table 5.15 Root types: alternating and non-alternating . . . . . . . . . . . . . . . . . . . . . . . . . 136Table 5.16 \u03c6 Harmony summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137xiiiTable 5.17 Vowel Harmony Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142Table 6.1 Zahao verb stems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147Table 6.2 Yucunany 1sg subject\/possessor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148Table 6.3 Tone patterns of bisyllabic noun stems in Jita (Adapted from Downing, 1990 (69)) . . . . 151Table 6.4 Stem tone class in Nata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151Table 6.5 Tone patterns of trisllabic nouns in Jita . . . . . . . . . . . . . . . . . . . . . . . . . . . 152Table 6.6 Lexical entries for Jita stems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154Table A.1 Determiner vowel quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163Table A.2 CV- comitative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163Table A.3 Associative vowel quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164Table A.4 Comitative vowel quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164Table A.5 Unprefixed nouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166Table A.6 Locative vowel quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166xivList of FiguresFigure 1.1 Nata and NTK area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5Figure 2.1 Y model of grammar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20Figure 3.1 Output of MERGE(X ,Y ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32Figure 3.2 Output of MERGE(WIFE,CLASS 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . 33Figure 3.3 Prosodic hierarchy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34Figure 3.4 Syntactic structure of a prepositional phrase and default prosodic mapping . . . . . . . . 36Figure 3.5 Syntactic structure of a prepositional phrase . . . . . . . . . . . . . . . . . . . . . . . . 45Figure 3.6 Syntactic structure of Associative Construction . . . . . . . . . . . . . . . . . . . . . . 46Figure 3.7 Syntactic structure of a Locative Prepositional Phrase . . . . . . . . . . . . . . . . . . . 49Figure 4.1 Nata surface inventory of tones by syllable weight . . . . . . . . . . . . . . . . . . . . 53Figure 4.2 Jamsay inventory of tones by syllable weight . . . . . . . . . . . . . . . . . . . . . . . 53Figure 4.3 Syntactic structure of a simple nominal . . . . . . . . . . . . . . . . . . . . . . . . . . 57Figure 4.4 Syntactic structure of Associative Construction . . . . . . . . . . . . . . . . . . . . . . 58xvAbbreviationsAUG augmentativeASSOC associativeCOM comitativeCM class markerDET determinerDIM diminutiveEVAL evaluativeLOC locativePL pluralRT rootSG singularST stemTBU tone bearing unitxviSymbols- morpheme separator( ) optional* unattested or unacceptable! fatal constraint violationu optimal candidate\/ actual form but non-optimal\u03b9 intonational phrase\u03c6 phrase\u03c9 wordC consonantG glideH high toneL low toneN nasalV vowelxviiAcknowledgementsThis dissertation is a culmination of efforts on the part of a number of people. Foremost among them are mycommittee, Douglas Pulleyblank, Gunnar O\u00b4lafur Hansson, and Rose-Marie De\u00b4chaine, who tirelessly helpedguide me through this project, from fieldwork to conference presentations, to editing the final document.Besides them, my colleague and consultant, Joash Gambarage, provided endless hours of expertise andassistance. It has truly been a joy to contribute to the scholarship on Nata.Before I ever arrived in British Columbia, my fascination with linguistics and African languages wasfostered at Boston University by Professor Cathy O\u2019Connor. Her field methods course with Ariane Ngabeuon Medumba set me down a path focused on phonology and tone in particular. My experience at BU wasshaped for the better by working alongside Anna Belew, Nick Danis, and Katie Franich. Between UBCand BU, I had the pleasure to complete a MA at The University of Toronto. Thank you to my cohort, whoendured endless presentations on tone, and to my advisor Professor Alexei Kochetov.At UBC, I was lucky enough to work with many other talented students, within my own cohort, andmany who followed. I wish I could explain how each of you kept me enthused and energized to keepworking, but that would near certainly compose an entire volume; in lieu of that, I will say thank you toMichael Schwan, Blake Allen, Kevin McMullin, Stephanie Williams, James Crippen, Adriana Osa, andOksana Tkachman; I could not have sustained myself without your contagious excitement for linguistics.In addition to my peers and supervisors at UBC, I learned a tremendous amount about teaching andsharing information from being a TA for Kathleen Currie Hall. Eric Vatikiotis-Bateson was a good spiritedsource of criticism and wisdom who shaped much of my thought about language and academia early on inmy stay at UBC; he is missed.This work would be for naught without the many people who speak Nata and welcomed me to theirland. Winyanya, his wife, and their family, along with Wasato and her husband made my stay comfortablexviiiand entertaining, not to mention keeping me well fed. Baunsa, Sarota, Peter (Mtabiri), Sabiti, Mugesi, andWasato gave countless hours of their time to help myself and Joash Gambarage untangle patterns in Nata;for this I am eternally humbled and grateful.I have been fortunate enough to share much of my academic journey at UBC with Natalie Weber andElla Fund-Reznicek; they kept me sane for a good portion of it, and were constant listeners. My partnerNikki Martin has graciously shown patience without end for my work. Our dog Molu has contributed smallbreaks of sanity, as well as tangential frustrations. Besides them, my mother Doralina Anghelescu has beenmy most fervent cheerleader and supporter, not only in this project, but throughout life. Thanks mom!My deepest thanks to every one listed and unlisted who has supported me, commented on my work, orbeen excited to discuss it with me.xixChapter 1Introduction1.1 GoalsThis dissertation has three goals: documenting a variety of grammatical constructions in Nata, developingan analysis of tone and vowel harmony in Nata using lexical allomorphy, and suggesting further work.Since work on Nata is relatively sparse, this dissertation presents the findings of a number of recentstudies conducted with the same primary consultant (see Section 1.3 for more details on language consul-tants). The focus of this description is on phonology and morphology, particularly of nouns. Verbal toneand morphology is a notoriously complex issue in Eastern Bantu languages and while the data presented inthis thesis sheds light on many aspects, it is not exhaustive.1.1.1 DocumentationThis work represents a unique collection of data gathered through collaborative work with a speaker-linguist,a wide array of students, and an excellent group of native speakers. The body of this work includes repre-sentative data sets which illustrate some core properties of nominal tone. In addition, the introduction anddiscussion chapters present smaller datasets on a variety of phenomena. Appendixes include a large set ofelicitation based data points for reference.1.1.2 AnalysisI employ a surface oriented theory of morpho-phonology: Lexical Allomorphy (Archangeli & Pulleyblank,2015, 2016, 2017, 2021). This approach has at its core the idea that alternations can be described solelyin terms of the relationships between the morphemes\u2019 surface forms. I have chosen to use this model to1analyze Nata in order to offer a new approach to understanding tonal and vowel quality alternations, whichare relatively well studied in generative phonology.1.2 OrganizationThe remainder of this chapter covers the research methods employed and demographic information on Nata.I discuss where the language is spoken, the speakers of the language, and its endangerment.Chapter 2 provides background on the phonological and morphological properties of Nata as well astheoretical framing for the analysis found in later chapters. In Chapter 3 I describe the model of grammar Iadopt and how it relates syntax and phonology; I then examine how this is applied to relevant syntactic andmorpho-phonological structures in Nata. In Chapter 4 I examine a variety of contexts which condition tonalalternations. In Chapter 5 I review the basic properties of vowel harmony and extend the analysis proposedby Gambarage and Pulleyblank, 2017 to additional morpho-syntactic constructions. Finally, in Chapter 6 Idiscuss a number of issues regarding Lexical Allomorphy, tone, and the morpho-phonology of Nata.1.3 Methods and MaterialsThe majority of the data presented in this dissertation was collected during fieldwork in Nata Village, Tan-zania from July to August of 2016; this fieldwork was supported by a SSHRC grant to Douglas Pulleyblank.The success of this research would not have been possible without the immense hospitality and friendshipof the Winyanya family in Nata and the Gambarage family in Mugeta. Prior to this period, extensive elic-itation was conducted with a single speaker, Joash Gambarage. This period began in May 2012, and wassupported by graduate student grants from UBC as well as a SSHRC research grant to Douglas Pulleyblank.The following subsections describe the methods used in elicitation as well as the preparation and materials.1.3.1 Preliminary work and NaWoGThe first stage of research that would become part of this work was conducted at the Linguistics Departmentin the University of British Columbia. In this stage I was extremely fortunate to be working with a consultantwho was also a trained linguist.I began by investigating nouns in isolation. The goal of this stage was to collect vocabulary items andbegin to understand the basic phonotactics of Nata. Two structures were targeted: word shape and tone2location.1It was quickly determined that closed syllables (those with a coda consonant: CVC, CVVC) are non-existent in Nata.2. The possible position for each syllable in a word was determined by asking the consultantto generate forms that fit the structure. This process was repeated over the course of early elicitations untilboth the consultant and I felt certain that the syllable canon and restrictions on it were comprehensive overa large enough data set (200 nouns). In addition to generalizations about syllable shapes, it became clearthat nouns had restrictions on their minimum size; this restriction is influenced by morphology, cruciallyflagging noun class 5 and 9 as exceptional (Lafon, 1994; Maho, 1999).3Tone location was checked in parallel with word shape. Possible word shapes generated for the initialelicitation included high tones in all possible combinations across all syllable types. Within a short time itbecame clear that words contain at most one high in isolation; therefore we abandoned the search for wordswith multiple high tone syllables. Likewise, it became clear that short vowels never had contour tones andlong vowels with high tone only had falling high tones or level high. The possible tone shapes for nounswas pared down to three types after eliciting a large enough dataset.My research on Nata was greatly supplemented by a field methods course taught by Professor Rose-Marie De\u00b4chaine. The course began several months after the initial stage of research and filled in many gapsin my knowledge of the language. Research that began in the field methods course eventually transformedinto the Nata Working Group (NaWoG), whose members have created (and published) a number of materialswithout which this dissertation would not have been possible.1.3.2 FieldworkDuring the field methods and NaWoG stages, a great deal of data was collected. This included an extensiveamount of verbal forms and many recorded utterances. A database of 200 nouns elicited by the authorwas supplemented with a word list from the Tanzanian Language Survey (Nurse & Philippson, 1975) andan Ikoma word list drawn from Higgins (2012). A database of 236 frames was created to target nouns indifferent syntactic structures; these frames drew heavily from data gathered by NaWoG. The frames crossed1For a more extensive exploration of research methods on verbal tone, see Marlo (2013)2See 2.1.2 for discussion on exceptional syllable structure, notably nasal syllables as in Mma \u2019mother\u2019 and the reduction of thefirst person subject marker ni3See Section 3.4.2 for the forms of class 5 and 9 class markers. Further discussion and analysis of class 5 can be foundin Section 4.4.2, Section 4.5.2, and Section 4.6.2 with respect to tone; discussion of and analysis of class 9 can be found inSection 4.4.2, Section 4.5.2, and Section 4.6.2 with respect to tone.3with the nouns created a script with more than 100,000 data points; therefore, the databases were used as achecklist to ensure an extensive set of combinations were investigated.At some points the frames were altered in order to encompass novel data (adjectives, adverbs, propernames).4Participants for fieldwork were selected from a pool of Nata speakers identified by a trusted council ofelders as \u2018good speakers of traditional (without Swahili) Nata.\u2019 This was done partially in order to maintainan impartial status as the author was closely affiliated with the Winyanya family. As Nata is a highlymulti-lingual community (people travel from across Tanzania for work), it is important that speakers wereraised primarily speaking Nata as opposed to having married a Nata speaker and having some knowledgeof lexical items, but, for instance, poor control of the tone patterns (i.e. replacing them with Swahili stress.)The council served to select such individuals. As a result, the age range of consultants is limited, reflectingonly the speech of those 40 years and older. This is not to say that the Swahili-influenced speech of youngerspeakers is unworthy of investigation; however such an investigation goes beyond the scope of this project.Meetings were conducted at the Winyanya homestead, where the author resided from July to August2016. Due to the organization of Nata village, this location was known and easily accessible for participants.Elicitations lasted several hours in the morning and several hours in the evening, usually with a differentspeaker for each session. This schedule avoided speaker fatigue, researcher fatigue and did not requiremeetings to take place during mid-day heat.The script for elicitation was generated from the aforementioned databases which allowed for rapidtranscription of speakers\u2019 production. The script acted as a prediction to which annotations were made.Every elicitation session was recorded. Recording was done using an iPhone 6 with an Apple headset micclipped near the neck of the consultant (either to a lapel or other piece of clothing). As the primary purposeof this work is not detailed phonetic study, this recorder was more than sufficient.Elicitation was done in a number of ways: Nata-to-Nata (N-N) elicitation, where the author wouldpresent a sentence in Nata and ask for the speaker to produce a \u2018correct\u2019 pronunciation. English-to-Swahili-to-Nata (E-S-N), where the author would present a sentence or context to a translator (a native speakerof Nata, and Swahili who was also fluent in English); the translator would then present the material inSwahili and consultants would reply in Nata. Finally, English-to-Nata elicitation (E-N), where the author4This is largely due to the fact that elicitation was done in conjunction with a native speaker, which allowed for on the flymodification; for instance, if a speaker preferred a particular adjective as a gloss for Swahili kubwa \u2019big\u2019, it was straightforward toadmit variants on the expected Nata moko\u00b4ro \u2019big, best\u2019.4Figure 1.1: Nata and NTK areawould present material in English to a translator and the translator would then present that material in Nata.N-N elicitation is useful to checking paradigms. E-S-N is useful for contextual elicitation, as was oftenrequired for cases where the preprefix is omitted. E-N elicitation is useful for conversational elicitation,such as question-answer pairs. Beyond these three approaches to scripted elicitation, speakers often providednarratives or particular cultural knowledge.1.4 Nata DemographicsNata5 is a Lacustrine Bantu language (JE45); it is mainly spoken in Nata Mbiso, Nata Motukeri, and Makon-dose (Serengeti District, Mara Region, northwestern Tanzania), with some speakers in Mugeta and Kyan-dege (Bunda District). The map in Figure 1.1 indicates Nata Mbiso, as well as a yellow highlighted areaindicating where Nata, Ikoma, and Isenye, are spoken.5Here and throughout, I refer to Bantu languages without prefixes, as this text is rendered in English. In Nata, the language isekina\u00b4ata, and it\u2019s speakers aBa\u00b4naata. In this text, the language is referred to as Nata, its speakers as Nata (people\/speakers), and thename of the area in which primary fieldwork was conducted as Nata (village). Likewise (Ki)Swahili is referred to without a prefix,as are Ikoma and Isenye (among others.)5Nata is bordered to the east by the Ikoma language, to the west by the Isenye language, to the northwestby the Ngoreme language, and to the south by the Sukuma language. According to Muzale and Rugemalira(2008), there are 7,050 of speakers of Nata, based on 2002 population statistics from the National Bureau ofStatistics, and fieldwork done in 2006. Aunio et al., 2019 reports 11,500 speakers, based on the same censusdata and independent fieldwork. In either case, Nata has slightly less speakers than Ikoma.Nata is classified as JE45 by Maho (2009) in his update of Guthrie classification of Bantu languages.Nata is classified along with Ikoma and Isenye as a dialect continuum (with the ISO code NTK); however,recent research indicates that at least Nata and Ikoma differ significantly in their phonology (Anghelescuet al., 2017; Aunio, 2010, 2013, 2015; Gambarage & Pulleyblank, 2017; Higgins, 2012). Despite this, thereis a high level of lexical similarity between the three dialects (Hill et al., 2007). Shetler (2007) presentstwo possible accounts of the migration of people into the Mara region. One account posits a commonancestry for Haya (JE22), Kerewe (JE24), Jita JE(25), Loogori (JE41), Kuria (JE43), Nata (JE45), Ikoma(JE45), and Isenye (JE45); the subgroup containing the ancestors of the Kuria, Loogori, Nata, Ikoma andIsenye migrated eastward, splitting again into the Kuria and Loogori north of the river Mara and the Nata,Ikoma and Isenye south of it. The second account posits a common ancestor for the Ngoreme (JE401), Nata(JE45), Ikoma (JE45), Isenya (JE45), and Sonjo (E41). This account states that the ancestors of these groupstravelled westward to Lake Victoria.Nata children acquire Swahili as their L1 (Mekacha, 1993). This shift has been recent and rapid. Withinthe same family, siblings in their 30s have Nata as L1, while siblings in their late teens have Swahili asL1. When parents address children in Nata, children respond in Swahili. Hill et al., 2007 note that Nata\u201chas a certain amount of prestige attached to it by its speakers, but [. . . ] its functional value and use aredeclining, possibly signalling the death of the language variety in the coming generations.\u201d Children exposedto Nata commonly make mistakes in prosody, assigning prominence (length, increased pitch) to penultimatesyllables, presumably under influence from Swahili.Speakers identify at least four varieties of Nata, based on clan and geography.6 In the north, the Eki-masaabha dialect is spoken by the Abhamoriho clan. In the south, the Ekibhachuuta dialect is spoken by theAbhaghikwe clan. These varieties differ, for instance, in the amount of vowel devoicing, and morphologyof the locative prefix. The Abhangirate clan and Abhusai clan have their own dialects, which differ, for6Here, I use the Nata orthography to indicate the names of dialects\/clans. These can be rendered in IPA as follows: Ekimasaabha[ekimasa\u00b4aBa], Abhamoriho [aBamoriho\u00b4], Ekibhachuuta [ekiBatSu\u00b4uta], Abhaghikwe [aBaG\u00b4ikwe], Abhangirate [aBaNgirate\u00b4], and Ab-husai [aBusa\u00b4i].6instance, in vowel harmony of prefixes. Where it is relevant in this work, I will mention distinctions in formand offer analyses to capture such variation.7Chapter 2Background2.1 Nata PhonologyThis section describes the inventory of segments in Nata, as well as some phenomena which create alterna-tions that are either relevant to prosodic domains, or else prevalent enough in the data at hand to be worthexplaining. Throughout this section, I avoid delving too deeply into the prosodic domains themselves; in-stead, I use labels like prefix and stem, whose formal status I am not directly concerned with in this work.In Chapter 3 I discuss the syntactic constituents and their relation prosodic constituents, detailing the formsthat are presented in the current section.2.1.1 SegmentsVowelsNata has seven vowel qualities which can all be either long or short; the fourteen contrastive vowels of Nataare shown in the chart below. Minimal pairs, pairs of words with distinct meanings that differ in only onesegment, are provided as evidence of contrast between segments.11The notion of contrast is at least a century old. Sapir, 1925 describes the distinction between phonetic facts and phonologicalcontrast, as does Saussure in his Course in General Linguistics (deGorog et al., 1959). See Dresher, 2011 for a modern treatmentof contrast with a significant review of the concept of contrast in the history of phonology.8Table 2.1: Phonemic Vowel InventoryFront BackShort Long Short LongHigh ATR i ii u uuMid ATR e ee o ooRTR E EE O OOLow RTR a aaIn Table 2.1, the horizontal lines separate the vowels into three heights: the high vowels [i, u]; the midvowels [e, E, o, O] and the low vowel [a]. Note that the low vowel [a] does not share any patterned behaviourwith either front or back vowels; for the sake of the chart, it is shown in the same column as front vowels.Table 2.2 illustrates the contrastive relationship, based only on vowel quality (and not length), for eachpair of vowels; for seven vowels, there are 21 combinations to consider (given that ordering and identicalvowels are uninformative for this comparison).9Table 2.2: Vowel minimal pairsForm Gloss Form Glossa. k\u0131\u00b4r-a \u2018be quiet!\u2019 ke\u00b4r-a \u2018skip over s.o!\u2019b. s\u0131\u00b4r-a \u2018get finished!\u2019 sE\u00b4r-a \u2018despise!\u2019c. k\u0131\u00b4r-a \u2018keep quiet!\u2019 ku\u00b4r-a \u2018burrow!\u2019d. k\u0131\u00b4r-a \u2018keep quiet!\u2019 ko\u00b4r-a \u2018grow up!\u2019e. k\u0131\u00b4r-a \u2018keep quiet!\u2019 kO\u00b4r-a \u2018do!\u2019f. s\u0131\u00b4r-a \u2018finished!\u2019 sa\u00b4r-a \u2018snatch!\u2019g. re\u00b4r-a \u2018cry!\u2019 rE\u00b4r-a \u2018raise s.o.!\u2019h. ke\u00b4r-a \u2018skip over s.o!\u2019 ku\u00b4r-a \u2018burrow!\u2019i. re\u00b4r-a \u2018cry!\u2019 ro\u00b4r-a \u2018be bitter!\u2019j. re\u00b4r-a \u2018cry!\u2019 rO\u00b4r-a \u2018see!\u2019k. ke\u00b4r-a \u2018skip over s.o!\u2019 ka\u00b4r-a \u2018trick!\u2019l. hE\u00b4r-a \u2018multiply!\u2019 hu\u00b4r-a \u2018wash!\u2019m. rE\u00b4r-a \u2018raise s.o.!\u2019 ro\u00b4r-a \u2018be bitter!\u2019n. rE\u00b4r-a \u2018raise s.o.!\u2019 rO\u00b4r-a \u2018see!\u2019o. mE\u00b4r-a \u2018germinate!\u2019 ma\u00b4r-a \u2018finish!\u2019p. ku\u00b4r-a \u2018burrow!\u2019 ko\u00b4r-a \u2018grow up!\u2019q. ku\u00b4r-a \u2018burrow!\u2019 kO\u00b4r-a \u2018do!\u2019r. ku\u00b4r-a \u2018burrow!\u2019 ka\u00b4r-a \u2018trick s.o!\u2019s. Bo\u00b4r-a \u2018get lost!\u2019 BO\u00b4r-a \u2018deteriorate!\u2019t. Bo\u00b4r-a \u2018get lost!\u2019 Ba\u00b4r-a \u2018count!\u2019u. BO\u00b4r-a \u2018deteriorate!\u2019 Ba\u00b4r-a \u2018count\u201910Table 2.3: Vowel pairings with reference to examples in Table 2.2i e E u o O ai - 3 a. 3 b. 3 c. 3 d. 3 e. 3 f.e - - 3 g. 3 h. 3 i. 3 j. 3 k.E - - - 3 l. 3 m. 3 n. 3 o.u - - - - 3 p. 3 q. 3 r.o - - - - - 3 s. 3 t.O - - - - - - 3 u.a - - - - - - -All vowels can occur as short or long. A variety of morpho-phonological contexts involve long vowels(in generative terms: glide formation, prenasal lengthening, hiatus resolution).There is no evidence that the low vowel interacts with either front or back vowels in a particular way;this distinction does not appear to be relevant for the low vowel.2Ladefoged, 1964 illustrates that the position of the tongue root is relevant in making distinctions in thevowel systems of some African languages. The vowels [e] and [E] differ in tongue root, and the phonologicalspecification for tongue root; [e] is ATR, or advanced tongue root, while [E] is RTR, or retracted tongue root.The same distinction separates [o] and [O], respectively. The high vowels [i, u] pattern with advanced midvowels in harmony alternations. The low vowel [a] patterns with retracted mid vowels. The relationshipbetween these vowels, and the sense in which [i, u] pattern like advanced vowels, and [a] like a retractedvowel will be explored in Chapter 5.ConsonantsNata has the inventory of consonants shown in Table 2.4. Sequences of two characters in the chart below,such as <nd> or <rw>, represent a single sound; in these cases a voiced prenasalized alveolar plosive anda labialized approximant, respectively. I will overview some alternations found in Nata; however, this is notan exhaustive study of consonant alternations or the contexts which condition them.2The low vowel could be considered to lack any feature specification for front or back. In the terms of Dresher, 2009, theback\/front dimension is not contrastive for the low vowel.11Table 2.4: Surface Consonant InventoryLabial Alveolar Palatal Velar GlottalPlosiveVoiceless Plain t kVoiceless labialized tw kwVoiceless palatalized kjVoiced Pre-nasalized mb nd NgFricative B s S G hLabialized Bw sw GwPalatalized Bj GjAffricate tSPre-nasalized \u00f1tSNasal m n \u00f1 NLabialized mwPalatalized mjApproximant rLabialized rwGlide w jNata has a small set of voiceless plosive consonants, [t, k]. Plosives surface voiced when prenasalized.The voiceless [t, k] alternate with [nd] and [Ng], respectively. Likewise, the fricative [B] alternates with[mb], patterning like the plosives. See \u00a72.1.4 for a brief discussion of alternations involving prenasalizedsegments. Voiceless velar plosives alternate with the voiced velar fricatives; see section 2.1.3 for additionaldata regarding \u2018voicing\u2019 alternations between segments.The rhotic [r] is realized in a variety of ways, depending on speech rate and phonological context. Itmay be more trill like or more approximant like. This status is not uncommon in other Bantu languages, forinstance Downing, 1990 reports a similar situation in Jita.Two classes of conditions exist: speech rate and morphological context. An example of a speech rate12conditioned alternation occurs in sequences of r-vowel-r where the intermediate vowel is elided in fastspeech compared to careful speech and the resulting rhotic is a trill; take for instance the careful production[OmO\u00b4\u00f4O\u00f4O] \u2018fire\u2019 (c3) and the fast speech variant (with a trill) [OmO\u00b4rO] \u2018fire\u2019 (c3).An example of a morphologically conditioned alternation is found in at least one r-vowel-r stem: rOr\u2018see\u2019; when followed by the final vowel -a, the stem surfaces with an r-V-r shape: [aka\u00f4O\u00b4\u00f4a] \u2018they (sg.) saw\u2019;however, when followed by the extensional affix -ire, the stem surfaces with a palatal affricate: [taa\u00f4OtSE\u00b4]\u2018they (sg.) aren\u2019t seeing\u2019. Note that throughout this dissertation, these types of alternations will not bepresented, and therefore a single symbol will be used to indicate these sounds: <r>.The lateral approximant [l] is found in Swahili borrowings (eg. [omuwali] \u2019rice\u2019 (c3)) and ethnonyms(ex. [omulagoori] \u2019Ragoori person\u2019 (c1)). This issue is ripe for detailed acoustic and articulatory study, as itseems unlikely that simply ascribing free variation captures the nuances of the situation.In this work, I interpret consonants with a vocalic offglide (i.e. labialized and palatalized consonants)as a singular segment, similar to prenasalized consonants. As we shall see in the following section, thesetwo classes of segments also interact with syllable structure; when they occur in non-final position, they arefollowed by long vowels.2.1.2 SyllablesNata has two unambiguous syllable shapes: CV and CVV. This simple inventory relies on interpreting twotypes of potential sequences as complex segments; i) NC are treated as prenasalized consonants as opposedto a sequence of a nasal segment and consonant; ii) consonants followed by glides (CG) are treated asconsonants with either labial or palatal secondary articulation. An alternate analysis of these sequenceswould result in the following syllable shape inventory: CV, NCV, CGVV, CVV.13Table 2.5: CV and CVV syllable shapesShape Example GlossCV omote\u00b4 \u2018tree\u2019 (c3)oBu\u00b4saro \u2018beads\u2019 (c14)CVV umwaana\u00b4 \u2018child\u2019 (c1)angwi\u00b4ina \u2018crocodile\u2019 (c9)eGeteete\u00b4 \u2018spine\u2019 (c7)oGure\u00b4esa \u2018beard\u2019 (c7)Long vowels are found following consonants with secondary articulation (CGVV) and preceding pre-nasalized consonants (VVNC); short vowels are not found in these environments. This distribution is ex-pected given patterns of vowel length in these positions, as Choti, 2015 documents in a variety of Bantulanguages.Table 2.6: Distribution of vowel length with respect to segmental contextContext Example GlossCGVV atw\u0131\u00b4iga \u2018giraffe\u2019 (c9)ikjO\u00b4OndE \u2018honey badger\u2019 (c9)VVNC orotSuumbe\u00b4 \u2018cattle pen\u2019 (c11)orohe\u00b4ende \u2018lentils\u2019 (c11)CGV *atwigaVNC *orotSumbeIn utterance-final position, only short vowels, and not long vowels, are found. In all other positions,either long or short vowels are possible. If a syllable in utterance-final position has a CG onset, it will nothave a long vowel; in other words, the prohibition on long vowels in utterance-final position supersedes thegeneral pattern in Table 2.6. In the example below, brackets represent phrase edges and hash marks represent14utterance edges. The forms in Table 2.7 are single phrase utterances.Table 2.7: Distribution of long vowels with respect to prosodic positionPosition Example GlossPhrase Initial [tSa\u00b4aka]# \u2018lion\u2019 (c9)[riisima\u00b4]# \u2018pond\u2019 (c5)Phrase Medial [omuuk\u0131\u00b4]# \u2018girl\u2019 (c1)[omuna\u00b4ata]# \u2018Nata person\u2019 (c1)Phrase Final [a-ma-rwa\u00b4]# \u2018beer\u2019 (c5)Utterance-final *[tSa\u00b4akaa]#*[riisima\u00b4a]#*[amarwa\u00b4a]#In addition to these two syllable types, onsetless syllables occur phrase-initially in careful speech andutterance initially in both careful and rapid speech. In rapid speech vowel coalescence readjusts syllablesto be either CV or CVV. Consider the nominals in Table 2.8: when they are utterance initial, they beginwith an onsetless short vowel syllable; when they occur in rapid speech, this short vowel coalesces with thepreceding verb ni-.Table 2.8: Onsetless syllables utterance initially, and medially (in rapid and careful speech)Example Glossikjaara\u00b4 \u2018finger\u2019 (c7)niikjaara\u00b4 \u2018it\u2019s a\/the finger\u2019 (rapid speech)ni ikjaara\u00b4 \u2018it\u2019s a\/the finger\u2019 (careful speech)omoGo\u00b4ondo \u2018farm\u2019 (c3)noomoGo\u00b4ondo \u2018it\u2019s a\/the farm\u2019 (rapid speech)ni omoGo\u00b4ondo \u2018it\u2019s a\/the farm\u2019 (careful speech)amaka\u00b4ra \u2018charcoal\u2019 (c6)naamaka\u00b4ra \u2018it\u2019s a\/the charcoal\u2019 (rapid speech)ni amaka\u00b4ra \u2018it\u2019s a\/the charcoal\u2019 (careful speech)15There are two types of onsetless syllables: those with short vowels and those with long vowels, V andVV. Onsetless syllables with long vowels are highly restricted: in addition to the utterance-initial restrictioncommon to long and short onsetless syllables, onsetless syllables with long vowels only occur when theyare part of a prosodic phrase that is bisyllabic. In terms of morphology, onsetless syllables with long vowelsare strictly found as class 9 prefixes on nouns with monosyllabic roots.Table 2.9: Utterance-initial onsetless syllablesShape Example GlossV omuka\u00b4ri \u2018woman\u2019 (c1)V aBaka\u00b4ri \u2018women\u2019 (c2)VV aaswe\u00b4 \u2018fish\u2019 (c9)VV a\u00b4angwe \u2018leopard\u2019 (c9)2.1.3 Dahl\u2019s lawIn Nata, voiceless velar obstruents, [k, kw, kj], alternate with voiced velar fricatives, [G, Gw, Gj]; thoughonly the plain series are used for illustration. This type of alternation is commonly observed in EasternBantu languages and referred to as Dahl\u2019s Law (Davy & Nurse, 1982; Lombardi, 1995; Odden, 1994). Thisalternation can be observed by focusing on the form of class prefixes which contain a velar obstruent; amongthese are classes 7, 12, and 15. These prefixes can occur with different stems; some stems begin with voicedconsonants, while others with voiceless consonants. When occurring with a stem beginning with a voicedconsonant, the prefix velar is voiceless.Table 2.10: Voiceless velars in prefixes before stems with initial voiced consonantsForm Gloss(DET-CM-ST)e- ke\u00b4 -riBitSi \u2018girdle\u2019 (c7)a -ka\u00b4 -rOrO \u2018fire (DIM)\u2019 (c12)o- ko\u00b4 -Goro \u2018leg\u2019 (c15)16Table 2.11: Voiced velars in prefixes before stems with initial voiceless consonantsForm Gloss(DET-CM-ST)e -Ge -ku\u00b4Ba \u2018chest\u2019 (c7)a -Ga\u00b4 -tSoGu \u2018elephant (DIM)\u2019 (c12)o- Go -twe\u00b4 \u2018ear\u2019 (c15)The result of this alternation is that sequences of voiceless velars followed by voiceless obstruents arenot attested in Nata. This is not true of sequences of voiced velars followed by voiced obstruents, as the c20prefix illustrates.Table 2.12: Class 20 velar prefixForm Gloss(DET-CM-ST)O -GO\u00b4 -rOrO \u2018fire AUG\u2019 (c20)*O -kO\u00b4 -rOrO \u2018fire AUG\u2019 (c20)u -Gu -taaro\u00b4 \u2018river AUG\u2019 (c20)Stems can be preceded by more than one prefix. In cases where multiple velar prefixes precede a stem,they will alternate in voicing. If the last (right-most) prefix is voiced, the second-to-last will be voiceless.17Table 2.13: Sequences of velar prefixesForm Gloss(C15-(OM)-ST)ko -\u00f1w -a\u00b4 \u2018to drink\u2019Go -k\u0131\u00b4 -\u00f1w -a \u2018to drink it\u2019 (c7)Go -su\u00b4kur -a \u2018to dump out\u2019ko- Ge\u00b4 -sukur -a \u2018to dump it out\u2019 (c7)2.1.4 Prenasalized consonantsWhen nasals surface before other consonants they agree in place with the following obstruent consonant; thisproperty of nasals is common to many Eastern Bantu languages, as Choti, 2015 catalogues. Additionally,nasals never act as tone bearing units. For these reasons, I treat sequences of nasals and homorganic conso-nants as a single prenasalized segment (as opposed to a distinct syllable, or a coda consonant). Downing,2005b presents a coherent case for such a treatment for a number of Bantu languages.In this section we investigate the realization of two morphemes which include nasals: the class 9 prefix,and the 1SG person prefix. Nata has the following prenasalized consonants: [mb, Ng, nd], and [ntS], asillustrated in Table 2.14, focusing on noun stems.Table 2.14: Prenasalized obstruents within morphemesForm Gloss(DET-CM-ST)a-ma-\u00f1eembe\u00b4 \u2018mango\u2019u-Bw-ooNgo\u00b4 \u2018brain\u2019o-ku-Geenda \u2018traveling\u2019e-Ge-so\u00b4ontSo3 \u2018plate\u20193Prenasalization of the affricate [tS] is transcribed with an alveolar nasal : [ntS]. This transcription is at odds with a homorganic18Class 9When the nominal stem begins with a vowel, the class 9 prefix surfaces as a postalveolar nasal, [\u00f1]. Whenthe stem begins with a consonant, the stem consonant is prenasalized.Some stems begin with with [B, r, G] when preceded by a non-nasal class marker; these same stems beginwith a prenasalized stop [mb, nd, Ng] when they are used in the c9 context. In other words, the segmentalreflex of the c9 prefix is prenasalization of the initial stem consonant in such cases; this is indicated withmorpheme breaks.Table 2.15: Stems alternate between plain and prenasalized initial consonantsForm Gloss(DET-CM-ST)[B]\u223c[mb] a -ka -Bo\u00b4rete \u2018goat (DIM)\u2019 (c12)a -m -bore\u00b4te \u2018goat\u2019 (c9)[r]\u223c[nd] a -ka\u00b4 -roro \u2018pancreas (DIM)\u2019 (c12)a -n -do\u00b4ro \u2018pancreas\u2019 (c9)[G]\u223c[Ng] a-ka -Gu\u00b4Bo \u2018hippopotamus (DIM)\u2019 (c12)a -N -guBo\u00b4 \u2018hippopotamus\u2019 (c9)2.2 Theoretical BackgroundIn this section I will describe the general model of grammar which I assume for the remainder of this work.I will then look in some detail at the phonological framework adopted in this dissertation. In the followingchapter I will consider details of the syntactic framework.Grammatical knowledge can be organized into a number of distinct modules which are separate butcommunicating systems. There are three modules: the syntactic, phonological, and semantic; the second ofthese is considered in depth in this work. Roughly, the three modules are organized in a Y shape, such thatthe output of syntax is available to both the phonological and semantic modules (Chomsky, 1995) .4theory of prenasalization; however, it is not clear that this is an accurate rendering of the articulation. This topic requires furtherinstrumental investigation; in this work I use the symbol <n> for the sake of consistency with other work published on Nata19Figure 2.1: Y model of grammarThere is a restricted number of relations possible between the syntactic and phonological modules. Datathat is not accounted for by such relations would point toward a modification of this modular approach.The separation between syntax and phonology allows us to ask the question of what sort of interface thetwo modules have.5 Crucially, observe that the phonological module cannot influence the syntactic module;the relationship is unidirectional such that the output of syntax must be the input of the phonology module.The Lexicon exists in addition to the modules of grammar; it is involved with every module. I assumethat the lexicon contains a set of lexical items. These items are pairs of sets of allomorphs and syntacticinformation. This model of the lexicon shares some features with Distributed Morphology (Marantz, 1997);specifically that there is no prelexical component of the grammar that forms \u201cwords\u201d. Instead, structureslike words and phrases are composed by a singular syntactic mechanism. I return to this topic in the nextchapterIn the next section, we will investigate the machinery of the phonological module of grammar. Thoughwe will cover a range of phonological processes, the goal of this section is to understand the general mecha-nisms which will model both the mapping from syntax to prosody as well as the restrictions on how phono-(Anghelescu, 2012; Anghelescu et al., 2017; Gambarage, 2012; Gambarage & Pulleyblank, 2017; Johannes, 2007).4For an alternative approach to modelling language, consider (Jackendoff, 1997) proposed parallel model. Generally, the Y-model has less descriptive power, but is more concise in explanation (Irurtzun, 2009)5One possible result would be that, in fact, the two modules are not separate and require direct interaction. I do not find strongevidence for such a claim.20logical elements interact with prosodic structures. I begin by reviewing constraint-based phonological gram-mars with a focus on alignment and indexed constraints. I supplement this approach by introducing LexicalAllomorphy, a surface-oriented theory of morpheme representation. I then outline a theory of prosodicdomains and describe how it is instantiated in Nata.2.2.1 Phonological ModellingI adopt a constraint-based approach to modelling the phonological grammar of Nata. In this approach,phonological restrictions are encoded as constraints which are ranked and violable. Potential output formsare compared based on their violations of constraints, with the winning output being the one which violatesthe least highly ranked constraints.In this section I review the classical architecture, propose a modification to lexical inputs, and detailalignment constraints, which are crucial for relating prosodic domains to segmental material (and viceversa), and indexed constraints, which essentially define classes of morphemes which differ in phonologicalimplementation.Optimality TheoryOptimality Theory formalizes phonological grammars as four major components: LEXICON, which containsall of the underlying forms for a language; CONSTRAINTS, which are a universal set of phonotactics6; GEN-ERATOR, which creates the candidate output forms to be evaluated; and EVALUATOR, which assesses theviolations of the candidate set in order to determine the optimal form (McCarthy & Prince, 1993; Prince &Smolensky, 1993\/2004). These components function to choose a single optimal output form given a lexicalinput and a language specific ranking of constraints. For this dissertation, I make a different assumptionabout the nature of the lexicon and generator; therefore I will focus on the universal constraint set and theirevaluation.The universal constraint set represents the full range of statements (i.e. functions that evaluate the well-formedness of a phonological element) possible in a phonological grammar; the ranking of these constraintsrepresents the phonological grammar of a particular language.7 This follows the generative tradition of6Classical OT constraints are not only phonotactics; they encompass a much broader range of functions over inputs and outputs.For the purpose of this exposition, I will not delve into the distinction between MARKEDNESS and FAITHFULNESS, nor will Iconsider the properties of inputs and outputs. This is because I do not adopt these properties from the classical model (Prince &Smolensky, 1993\/2004).7Rankings between constraints are referred to as crucial when the domination relationship makes a difference in selecting theoptimal output over some other form. When the relationship is not crucial, the constraints are not crucially ranked; in tableaux this21parsimony and modularization in grammar; however, the specific form of these restrictions is a continuallyopen question.Evaluation functions as follows: for each candidate violations are assigned for each constraint. Begin-ning with the most highly ranked constraint, if a candidate has more violations of a constraint than someother candidate, it is eliminated from consideration. For each constraint this is repeated until a single formremains. This is the optimal output form.In this dissertation, I adopt many of the familiar notation conventions found in Optimality Theory. Aform is evaluated in a tableau, as shown in example (1). The anatomy of a tableau is as follows: the inputis in the top left cell; arranged along the top row in descending order of rank are (a subset of) constraints;dashed vertical lines indicate no crucial ranking, while solid vertical lines indicate that the left constraintcrucially dominates the right; in the left column, under the input, are the output candidates; the optimalcandidate is indicated with au(thumbs up) symbol; in each row to the right of the candidate form andseparated by a double vertical line is the evaluation for that candidate; violations are indicated with an *(asterisk); violations which remove a candidate from consideration are called \u2018fatal\u2019 and represented withan ! (exclamation point) following the violation mark.(1) Anatomy of a tableau[INPUT] CON1 CON2 CON3u a. WINNER *b. LOSER1 *!c. LOSER2 *!In the tableau in (1), CONSTRAINT1 (CON1) crucially dominates (>>) CON2, and CON2 crucially dom-inates CONSTRAINT3 (CON3). In this tableaux, (1a.) is the winning candidate but still violates CON-STRAINT3. Candidates in (1b. & c.) both violate more highly ranked constraints than CON3. The candidatein (1b.) fatally violates CON1, which is not violated by either of the other candidates; this removes (1b.)from consideration. The candidate in (2c.) fatally violates CON2, which is not violated by the candidate in(1a.); this eliminates the candidate in (1c.), leaving (1a.) as the optimal output.is indicated with a dashed line. Conceptually, non-crucially ranked constraints represent an analytical indeterminacy between twopotential grammars for some languages. In an actual learned OT grammar for some language, constraints are all ranked.222.2.2 Lexical AllomorphyInput forms are phonological elements to be evaluated by the grammar. In the theory of Lexical Allomor-phy (LA), input forms are represented by sets of observed (or inferred) surface forms for each morpheme.This approach contrasts with traditional approaches to generative phonology in which abstract underlyingrepresentations serve as the base for deriving surface forms for each morpheme (Archangeli & Pulleyblank,2015, 2016, 2017). Furthermore, inputs include syntactic structure containing allomorph sets. In this sectionI motivate a surface based approach and demonstrate how this model is applied to allomorphs in Nata.Allomorph sets are sets containing all of the surface forms of a particular morpheme and any morphsgenerated by redundancy (which we will turn to shortly). For example, in Nata, noun class prefixes withnon-low vowels alternate between high, mid advanced, mid retracted vowels, and glides; consider class 3:{mu, mo, mO, mw}. These allomorphs are shown in Table 2.16.Table 2.16: Noun class 3 prefixesForm Glosso-mo-te\u00b4 \u2018tree\u2019o-mu\u00b4-nwa \u2018mouth\u2019O-mO-tO\u00b4ONg-O \u2018string\u2019u-mw-aat\u00b4i \u2018medicine\u2019This set is not yet complete, as noun class prefixes may be either high or low toned. Combining thesepossibilities, the resulting allomorph set for class 3 is: {mu, mu\u00b4, mo, mo\u00b4, mO, mO\u00b4, mw}.(2) Noun class 3 prefix allomorph set:{mo, mo\u00b4, mu, mu\u00b4, mO, mO\u00b4, mw}One difference between Lexical Allomorphy in contrast to traditional Optimality Theory is that thenumber of candidates is strictly defined by the shape of the lexicon (and not by GEN). Because morphemesare simply sets of all surface forms, the total candidate set for any form is the product of all of the allomorphsets. For example, consider the form [o-mo-te\u00b4] \u2018tree\u2019 (c3); this form is composed of three morphemes. Theallomorph sets for these morphemes are given below.23(3) Allomorph sets for the morphemes in [o-mo-te\u00b4]DET: {o, oo, o\u00b4o, O, OO, O\u00b4O}C3: {mo, mo\u00b4, mu, mu\u00b4, mO, mO\u00b4, mw}TREE: {te\u00b4, te}In order to generate the candidate forms, we generate the product of these three sets. In terms of number,this is equal to 6\u22177\u22172 = 84 candidates. The full list of candidates is given below.(4) Candidate set for the form [{o, oo, o\u00b4o, O, OO, O\u00b4O}+{mo, mo\u00b4, mu, mu\u00b4, mO, mO\u00b4, mw}+{te\u00b4, te}]:o-mu-te\u00b4o-mu\u00b4-te\u00b4o-mo-te\u00b4o-mo\u00b4-te\u00b4o-mO-te\u00b4o-mO\u00b4-te\u00b4o-mw-te\u00b4o-mu-teo-mu\u00b4-teo-mo-teo-mo\u00b4-teo-mO-teo-mO\u00b4-teo-mw-teoo-mu-te\u00b4oo-mu\u00b4-te\u00b4oo-mo-te\u00b4oo-mo\u00b4-te\u00b4oo-mO-te\u00b4oo-mO\u00b4-te\u00b4oo-mw-te\u00b4oo-mu-teoo-mu\u00b4-teoo-mo-teoo-mo\u00b4-teoo-mO-teoo-mO\u00b4-teoo-mw-teo\u00b4o-mu-te\u00b4o\u00b4o-mu\u00b4-te\u00b4o\u00b4o-mo-te\u00b4o\u00b4o-mo\u00b4-te\u00b4o\u00b4o-mO-te\u00b4o\u00b4o-mO\u00b4-te\u00b4o\u00b4o-mw-te\u00b4o\u00b4o-mu-teo\u00b4o-mu\u00b4-teo\u00b4o-mo-teo\u00b4o-mo\u00b4-teo\u00b4o-mO-teo\u00b4o-mO\u00b4-teo\u00b4o-mw-teO-mu-te\u00b4O-mu\u00b4-te\u00b4O-mo-te\u00b4O-mo\u00b4-te\u00b4O-mO-te\u00b4O-mO\u00b4-te\u00b4O-mw-te\u00b4O-mu-teO-mu\u00b4-teO-mo-teO-mo\u00b4-teO-mO-teO-mO\u00b4-teO-mw-teOO-mu-te\u00b4OO-mu\u00b4-te\u00b4OO-mo-te\u00b4OO-mo\u00b4-te\u00b4OO-mO-te\u00b4OO-mO\u00b4-te\u00b4OO-mw-te\u00b424OO-mu-teOO-mu\u00b4-teOO-mo-teOO-mo\u00b4-teOO-mO-teOO-mO\u00b4-teOO-mw-teO\u00b4O-mu-te\u00b4O\u00b4O-mu\u00b4-te\u00b4O\u00b4O-mo-te\u00b4O\u00b4O-mo\u00b4-te\u00b4O\u00b4O-mO-te\u00b4O\u00b4O-mO\u00b4-te\u00b4O\u00b4O-mw-te\u00b4O\u00b4O-mu-teO\u00b4O-mu\u00b4-teO\u00b4O-mo-teO\u00b4O-mo\u00b4-teO\u00b4O-mO-teO\u00b4O-mO\u00b4-teO\u00b4O-mw-teAs an analyst, it is possible to rapidly limit the problem space just by reducing the number of relevantdeterminer allomorphs. It is clear that the forms beginning with long vowels are never optimal in phrase-initial position; this cuts the candidate set in half, removing the forms that begin with long vowels: DET:{o, O}. The long vowel allomorphs regularly occur when there is an additional nominal prefix precedingthe determiner. It is also clear that the retracted determiner only occurs in a special harmony context, againcutting the set in half: DET: {o}. The resulting fourteen candidates (1*7*2=14) represent a more manageableand informative problem space, particularly for the account of high tone placement and (general) vowelharmony.Practically, this procedure is undertaken regardless of the model of the lexicon; however, adopting anallomorphy based approach allows this procedure to be formalized in terms of a finite number of candidates.In other words, any formal analysis in OT shows only a relevant subset of forms for evaluation. No oneconsiders the infinite set; therefore everyone must use some partitioning algorithm, likely similar to the onedescribed above.Throughout the remainder of this dissertation I will use abbreviated candidate sets; however, I will spellout allomorph sets, allowing the full candidate set to be generated for the interested reader. Generally, I willfollow the same set of simplifying steps laid out above: removing candidates which have properties whichnever surface in the context under consideration.Selectional RestrictionsIn addition to phonotactics restricting the combinations of allomorphs, some morphemes require that anallomorph appear in a particular context. These restrictions generally require specific phonological content25to precede or follow an allomorph. From the perspective of a learner these patterns are the first type ofinference to be made; further learning then generalizes to account for phonotactics in place of morphemespecific restrictions. Nevertheless, grammars contain both the higher level generalizations, as well as mor-pheme specific restrictions.Selectional restrictions are modelled as a condition on each morpheme which is either satisfied or notsatisfied, following Archangeli and Pulleyblank, 2017. For example, the c9 prefix would be represented asin (5).(5) Lexical entry for class 9\/10 prefixSel: L: This allomorph must be followed by a syllable associated with a low tone.{N, \/0} L C9The satisfaction of these restrictions is represented by the constraint SELECT. If a morpheme\u2019s selectionalrestriction is satisfied, then SELECT is not violated; however, if the morpheme appears in a context that doesnot satisfy its selectional restriction, then SELECT is violated.In contrast to other theories, Lexical Allomorphy places all allomorphy inside of phonology in the sensethat there is no pre-processing of forms based on syntactic contexts. Embick, 2010 describes VocabularyInsertion as a process which inserts phonological structure at the syntactic level based on the local (syntactic)context. Consider the example of the English plural vocabulary items.(6) Vocabulary items for tense (English) (adapted from (4) in (Embick, 2010))T[past] \u2194 [-t]\/{LEAVE, BEND . . . }T[past] \u2194 [- \/0]\/{HIT, SING . . . }T[past] \u2194 [-d]In contrast, the three allomorphs of T[past] would be modelled as members of the set of allomorphs for theT[past] morpheme; each of these would either select or be selected for (or both) by particular types of roots.For instance, the root HIT would select for a following [- \/0] allomorph, as illustrated below. Note that thereare some details about how present tense morphology would work out that are being omitted here.(7) Lexical entries for English past tense construction{d, -t, - \/0} T[PAST]{hitSel: T[past] \/0} HIT26Another type of allomorphy exists, which is computed inside of the phonology. This type of allomorphywould be at play in the decision between English plural morphs like [-z, -s, and -@z]. The approach adoptedin this paper collapses all allomorphy into the same type, computed inside of the phonology. Lexical al-lomorphy is not the only approach which places the onus of allophonic variation inside the phonologicalgrammar. For a similar approach applied to Seenku tone sandhi, see McPherson, 2019.Redundancy RelationsThe lexicon is not entirely arbitrary. Learners are capable of forming generalizations over types of allo-morphs for a given morpheme such that they can predict allomorphs they have not encountered by general-izing from similar morphemes. Following Archangeli and Pulleyblank, 2016 I refer to these as morph setrelations (MSR) because they formalize which allomorphs can be predicted given a particular observation.These relationships capture patterned regularity found in a phonological grammar; in particular, similarmorphemes share similar allomorph sets.Let us consider a simple case of tone allomorphs. The prototypical Nata class marker prefix can beeither high or low tone. We will see in detail the cases in which these allomorphs surface in Chapter 4. Fora learner, this means that if they observe any class marker, they can predict at least one other allomorph thanthe observed one: if they observe a H allomorph, they may assume an L allomorph and vice versa.(8) Systematic class marker relations (tone)CV\u00b4 \u2194 CV`[mo\u00b4]C1 \u2194 [mo`]C1...While this relation is straightforward, other allomorphs have more complicated redundancy relationships.In particular, stems have a more elaborate system of tonal allomorphs, and in general vowel harmony allo-morphs are more nuanced. We will consider these cases in Chapter 4 and Chapter 5 respectively.2.2.3 Alignment ConstraintsThe material in this dissertation is largely concerned with prosodic domains and their edges. ALIGNMENTconstraints were developed to account for, among other things, relationships between morpheme edges(McCarthy & Prince, 1993); however, this mechanism has been extended to tone Cassimjee and Kisseberth,271998. In this section I will explain the function of alignment constraints and illustrate the set of mappingconstraints which reference syntactic structure.Although we are specifically discussing alignment, more generally all constraints are formally express-ible with the same schema in (9).(9) General constraint schema (Zoll, 1996)\u2200x(P(x)) where P is the phonological statement to be judged true or falseAlignment is a particular type of phonological statement that requires two elements to stand in the samerelative position (i.e. an edge). Alignment constraints have the general form in (10a.); the Edge function isdefined in (10b.)(10) General form of an alignment constraint McCarthy and Prince (1993)a. Align(Cat1,Edge1;Cat2,Edge2): \u2200 Cat1 \u2203 Cat2 such that Edge(Cat1, Edge1) andEdge(Cat2, Edge2) coincide.b. Edge(x,{L,R}): The left\/rightmost element in xZoll (1996) demonstrates that alignment should formally be composed of two functions: coincidence andprecedence. Informally, COINCIDENCE evaluates if two elements are standing in the same position relativeto some edge8; INTERVENTION returns violations for each element which occurs between two edges. Inthe definitions below, y > x is read as \u2018y occurs after x\u2019.(11) COINCIDENCE and INTERVENTION (Adapted from Zoll (1996)9)COINCIDENCE(x,y): x and y coincide iff Edge(x,{L,R})=Edge(y, {L,R})INTERVENTION(x,y): x (left-)intervenes between y and edge E iff y > x >E and x \u0338= \/0Following McCarthy (2003), I model alignment using just the coincidence relation. The general motivationis simply that constraints which only compute coincidence are more restrictive than admitting constraintswhich consider precedence. More importantly, the data presented in this dissertation does not cruciallyrequire it evaluating precedence.Consider the constraint below, a specific instance of the schema in (10a). Throughout this work, con-straints are presented in an example with a full name, followed by a short name (used in tableaux and other8The formulation of COINCIDENCE in terms of EDGE is essentially McCarthy and Prince, 1995\u2019s ANCHOR.9Zoll, 1996 further distinguishes between left and right intervention, of which only left intervention is relevant for the Nata data.28environments where space saving is necessary) and definition. Formal evaluation statements will be pro-vided for most constraints along with a prose description. When defining alignment constraints, C standsfor the coincide relation.(12) ALIGN(PHRASE , LEFT; HIGH , LEFT)AL(\u03c6 ,H): \u2200\u03c6\u2203 H C(Edge(\u03c6 ,L), Edge(H,L))For every phonological phrase, assign a violation mark if the left edge of that phrasedoes not coincide with the left edge of some high tone.2.2.4 Morph preference conditionsIn lexical allomorphy, phonological properties such as high tone, or vowel features are referenced in phono-tactics prohibiting configurations (globally or contextually, as we shall see in particular cases). In addition,some constraints can reference a particular morphological property.Archangeli and Pulleyblank, 2016 propose that for any allomorph of a morpheme that is labelled for aparticular morpho-syntactic feature, there exists a preference condition that favours the marked morph inthe context of the particular morpho-syntactic context. They claim that such a condition exists in the case ofEnglish irregular plurals.In English, the regular plural morphology fails to surface in cases where a noun has an allomorphspecified for the PLURAL feature, such as the form [fit] FOOT.PL (*[fUt-s]FOOT-PL). The morph set forthe noun FOOT is {fUt, fitPLURAL}. Because the morph [fit]PLURAL is marked for the PLURAL context, thereautomatically exists a preference condition, as shown below.(13) English plural preference conditionfit>fUt|PLURAL : The morph [fit] is preferred to the morph [fUt] in the plural morphological context.In English, another component of the irregular plural pattern has to do with the economy of morpho-syntacticfeature expression by the phonology. Specifically, morphs like [fit]PLURAL is marked for the PLURAL fea-ture, therefore a form which also includes an independent plural morpheme such as [fit-s] would violate acondition on morpho-syntactic features having a one-to-one correspondence with morph. In [fit-s], thereare two morphs that correspond to plural]. Archangeli and Pulleyblank, 2016 refer to this condition asMANIFEST.29In the case of Nata, morph preference is relevant because some root forms are marked to occur onlywith particular class feature contexts. Unlike the English case in which MANIFEST penalizes multiplephonological elements which corresponding to a single morpho-syntacic feature, these cases involve multi-ple phonological elements being marked for matching class features. In a sense, MANIFEST is orthogonalto the Nata cases, because each individual morph does correspond to a distinct bundle of class features; whatis crucial is that the distinct feature bundles of the root and pre-root morphemes have the same value. Theeffect of morph preference in the Nata cases is that morphs which are marked for a particular class contextare preferred in the presence of that class marker prefix over unmarked morphs.In general, MORPHPREF (MP) constraints evaluate a morpheme in a given morpho-syntactic contextand assign a violation if the properly specified morph is not selected. We shall see the function of theseconstraints when we examine nominal tone. Consider the example below, which requires that class markersand roots match in class feature.(14) MorphPref constraintMORPHPREF:RT-DET: For a root morpheme, if it has an allomorph specified for a classfeature, assign a violation mark if that allomorph is not used when ina \u03c6 preceded by a determiner with a matching class feature.Note that not all allomorphs have a particular indexing for a particular class feature. In the case of Nata,most roots do not have any class indexing; only a particular subset of roots have a particular allomorphmarked for a particular class, as we shall see in Section 4.4.2.3 SummaryIn this chapter we have looked at some crucial aspects of Nata: the sound inventory of the language, possiblesyllable shapes, and common alternations. In addition, we considered the theoretical underpinnings of thiswork: the complementary work of the syntactic and phonological modules, the basics of Optimality Theory,and of Lexical Allomorphy. In the following chapter I will consider in more detail the structure of Natawords and utterances, as mapped from syntactic structure to phonological structure.30Chapter 3Syntax to Prosody MappingIn this chapter I make a proposal for mapping syntactic structures to prosodic structures. I assume thatprosodic (i.e. phonological) processes apply to prosodic domains, not directly to syntactic ones. Neverthe-less, there is a correspondence between syntactic structure and prosodic structure. Studies with a similarperspective on how syntax and phonology interface include: Dobashi, 2003; Elfner, 2012; Hsu, 2016; Ito\u02c6and Mester, 2009, 2012, 2013; Marvin, 2003; Newell, 2008; Pak, 2008; Poletto, 1998; Ross, 2011; Sande,2017; Sandy, 2017; Varis, 2012; S\u02c7urkalovic\u00b4, 2015; Wang, 2014.I begin by describing some properties of the syntactic module of grammar: how syntactic structuresare built, what they are composed of, and how phonological forms are related to them. I then consider themapping of nominal expressions: what is their syntactic structure, what are the constraints that relate thatstructure to a prosodic one, and how are optimal prosodic outputs chosen.3.1 Syntactic StructuresThe syntactic module of grammar is, in generative terms, designed to predict all and only the grammaticalsentences of some language.1 I adopt a variant of syntactic theory known as Minimalism. In a minimalistapproach the function MERGE combines two elements and produces a labelled structure containing them(Adger, 2003; Chomsky, 2001).(15) MERGE(X ,Y )\u2192{X ,{X ,Y}}1See Boeckx and Hornstein, 2010 for a conceptual history of syntax.31The formal definition in (15) illustrates that MERGE applied to the elements X and Y, results in a set withlabel X (though it could have been Y) containing the pair (X,Y). This can be represented using a tree, as inFigure 3.1.XYXFigure 3.1: Output of MERGE(X ,Y )Syntactic elements are composed of features that encode information like category, person, number,gender, and tense. Features have values, for instance a number feature may be valued singular or plural, acategory feature may be valued noun, and so on.The lexicon contains sets of allomorphs and syntactic information, these are similar to \u2018vocabularyitems\u2019 in terms of Marantz, 1997. The lexicon functions as a sort of dictionary which transforms the terminalnodes in syntactic structures into corresponding phonological material. As we saw in the previous chapter,this phonological material is modelled as a set of surface forms, of which the phonological module selectsthe single optimal form for the particular context.Consider the categorized root ka\u00b4ri WIFE, and two class markers which could potentially merge with thisroot; these are shown in Table 3.1, with an explanation following.Table 3.1: Lexical entriesLabel P-Form2 S-FeaturesCLASS 1 mu {Num:Sg . . .}CLASS 2 Ba {Num:Pl . . .}WIFE ka\u00b4ri {uNum, Cat:n . . . }Before considering the specifics of the morphemes in (15), I will discuss some relevant aspects of syntac-tic features. Syntactic features drive the MERGE operation. Features come in two basic types: interpretableand uninterpretable. Uninterpretable features are notated with a \u2018u\u2019 preceding them. MERGE is motivated bythe checking of uninterpretable features (Adger, 2003; Svenonius, 2006). In order to fill the value of an un-2\u2018P-Form\u2019 is standing for the set of allomorphs for a particular morpheme.32interpretable feature, it must be adjacent to another feature bundle that has a value for that feature. Crucially,a syntactic derivation will fail if it contains uninterpretable features. Therefore, a successful derivation mustresult in the checking of all uninterpretable features.Returning to the morphemes in (15), the root WIFE has an uninterpretable Number feature \u2019uNum. Thisfeature can be valued by applying the MERGE operation to the root and a class marker. Consider Class 1,which has a value Singular (Sg) for Number. The merger of WIFE and CLASS 1 is shown in Figure 3.2. Iwill discuss the syntactic labelling of such structures later in this chapter.NumPka\u00b4riuNummuNum:SgFigure 3.2: Output of MERGE(WIFE,CLASS 1)3.2 Prosodic StructuresProsodic domains are hierarchically organized units which the phonological grammar can reference. Theconceptual development of these domains can be traced through several major works, beginning with Nesporand Vogel (1986) and Selkirk (1986).The lowest level of the prosodic hierarchy is the mora, which is dominated by the syllable, which is inturn dominated by the foot. This dissertation focuses primarily on domains larger than the syllable and foot.I adopt a relatively streamlined model which contains only three domains larger than the foot: word,phrase, and intonational phrase; this contrasts significantly with models which allow for sub-word domains,such as the prosodic macrostem and the prosodic stem.It is crucial at this point to disambiguate some technical terms. \u2018Words\u2019 are technical in the sense of theprosodic word, as well as informal, in the sense of the string of segments which occur between white spaces.I refer to the Prosodic word as a\u03c9-word, or simply\u03c9 . Informally, I sometimes refer to morphological words,prosodic words, and typographical words as \u2018words\u2019. When the distinction is important, I specify the typeof word; crucially, the prosodic unit is identified by the \u03c9 .Likewise, linguistic theory is rife with many types of phrases; this dissertation is concerned with phono-logical and syntactic phrases. Following the convention of using Greek letters to label prosodic structures, I33Intonational Phrase (\u03b9)Phrase (\u03c6 )Word (\u03c9)Foot (\u03a6)Syllable (\u03c3 )Mora (\u00b5)Figure 3.3: Prosodic hierarchyrefer to the Prosodic phrase as a \u03c6 -phrase, or simply \u03c6 . When referring to syntactic phrases, I will use thefamiliar \u2018XP\u2019 schema, where X stands for some type of phrase, for instance, DP, stands for a DeterminerPhrase, nP stands for nominalizer Phrase.Table 3.2: Key to prosodic constituentsName Short hand SymbolProsodic Word \u03c9-word \u03c9Prosodic Phrase \u03c6 -phrase \u03c6Intonational Phrase \u03b9-phrase \u03b9The literature on Bantu languages often includes additional prosodic domains, such as the Prosodic Stem(Downing, 1999; Downing & Kadenge, 2020; Hyman, 2009b; Mudzingwa, 2010). Such a domain could berelevant in Nata, for example in reduplication; however, for the study at hand, the distinction between aninner domain (\u03c9) and outer domain (\u03c6 ) is sufficient to characterize the behaviour of nominal tone and vowelharmony.343.3 MappingIn this section I describe the syntactic structure of nouns and how that structure is related to prosodic do-mains. I will review the syntactic structure of nominal phrases (nPs), determiner phrases (DPs) and prepo-sitional phrases (PPs); throughout this discussion I will provide a description of the forms involved in thestructures.I follow approaches like Selkirk, 2011 in assuming that the syntactic and prosodic hierarchies have adefault mapping which is modulated by phonological constraints. Categorized phrases, nP and vP, map toprosodic words; this domain has a unique minimal size restriction which has been a topic of much work inBantu languages (Downing, 1994, 1998, 1999, 2005a, 2006). Above the nP and vP, phrases such as NumP,DP, and PP, correspond to phonological phrase (\u03c6 -phrases).Constraints which map from syntax to prosody are given in the example below. Note that analysis ofmore languages may suggest that these constraints represent a more general constraint relating classes ofsyntactic structures to specific prosodic structures. I do not take up the issue of morpho-syntactic naturalclasses in this work.(16) Core Syntax to Prosody mapping constraintsAl(nP,\u03c9): \u2200 nP \u2203 \u03c9 C(Edge(nP,L), Edge(\u03c9 ,L))For every nP, assign a violation mark if the left edge of that nP does not coincidewith the left edge of some \u03c9-word.Al(NumP,\u03c6 ): \u2200 NumP \u2203 \u03c6 C(Edge(NumP,L), Edge(\u03c6 ,L))For every NumP, assign a violation mark if the left edge of that NumP does notcoincide with the left edge of some \u03c6 -phrase.Al(DP,\u03c6 ): \u2200 DP \u2203 \u03c6 C(Edge(DP,L), Edge(\u03c6 ,L))For every DP, assign a violation mark if the left edge of that DP does not coincidewith the left edge of some \u03c6 -phrase.Al(PP,\u03c6 ): \u2200 PP \u2203 \u03c6 C(Edge(PP,L), Edge(\u03c6 ,L))For every PP, assign a violation mark if the left edge of that PP does not coincidewith the left edge of some \u03c6 -phrase.The optimal mapping generated by these constraints is illustrated in Figure 3.4. These structures satisfyeach of the constraints given in (16) above.35PPDPNumPnPNumDPUtt\u03c6\u03c6\u03c6\u03c9nPNumDPSyntactic structure of a PP Prosodic structure of a PPFigure 3.4: Syntactic structure of a prepositional phrase and default prosodic mappingThe prosodic structure shown in Figure 3.4 involves the recursion of \u03c6 in the sense that there are two \u03c6sthat are dominated by other \u03c6s. In Nata, only the maximal \u03c6 is relevant; therefore, I assume a non-recursivestructure. In the remainder of this work, I will only indicate the maximal \u03c6 .To be explicit, there remains an open question of whether a constraint prohibiting the recursion of\u03c6 , *RECUR(\u03c6 ), is ranked above or below the constraints in (16). If *RECUR(\u03c6 ) is ranked above theseconstraints, then there will be no \u03c6 left edges corresponding to the DP or NumP in the cases when a PP exists,because there will only be one \u03c6 corresponding to the PP. If *RECUR(\u03c6 ) is ranked below the constraints in(16), then there will be nested \u03c6s as illustrated in (16). In Nata, there is no evidence to support a low ranked*RECUR(\u03c6 ); therefore, I rank it above the mapping constraints.In Section 3.4 I will consider the internal constituents of the nominal, starting from the inside (nominalphrase), and moving up to the determiner phrase. In Section 3.5 I will consider prepositional prefixes, whichare external to the nominal; I examine the associative, locative, and comitative.3.4 Internal nominal syntaxIn the following subsections I describe the nominal Phrase (nP), Number Phrase (NumP), Determiner Phrase(DP). I list the types and forms of morphemes in these syntactic domains and illustrate how these domainsare mapped to prosodic phrases.36The nP is either an a-categorical root merged with a categorizer (e.g. rem \u2018cultivate\u2019 + i \u2018agentivenominalizer\u2019), as in (17a.); or an inherently categorized root (e.g. kari \u2018woman\u2019), as in (17b.).3 The structurein (17a.) is referred to as a derived or complex nP, while the structure in (17b.) is referred to as a simplexnP.(17) Two types of nPsa. Complex nP b. Simplex nPnPni\u221aremnPnkariCategorizers merge with a-categorical roots in order to satisfy an uninterpretable category feature on theroots. For categorized roots, there is no uninterpretable category feature and thus a categorizer cannotmerge.3.4.1 nP=\u03c9I will now focus on the internal syntax of the nP. There are four categorizing stem suffixes in Nata: -i, -O,-u(\/-Bu), and -a. In the chart below I provide examples of stems in a variety of contexts.Table 3.3: The root Gaamb with different nominalizersForm Glosso-mu\u00b4-Gaamb-i \u2018speaker\u2019 (c1)e-k\u0131\u00b4-Gaamb-O \u2018saying\u2019 (c7)o-ku\u00b4-Gaamb-a \u2018to speak\u2019 (c15)De\u00b4chaine et al., 2017 identify two compatible usages for the -i suffixes. The first is stative nominalswith class 14 prefixes; these forms roughly have the meaning \u2018some state\/quality of being X\u2019, where X is the3There are several ways to model these two types of roots. The approach I suggest distinguishes a-categorical and categoricalroots based on the presence or absence of the n feature. Nominal categorizers are distinguished from roots by lacking the non-category (n) features which both types of roots presumably contain. In other words, I assume that categorized roots have a categoryfeature, plus some other features which define nominals that the categorizer lacks.37action denoted by the stem. These cases are all examples of nominals that do not have singular and pluralforms. The second case is compatible with agent nominals with class 1\/2 prefixes; these forms roughly havethe meaning \u2018some person who performs X\u2019, where X is the action denoted by the stem.Table 3.4: Stem suffix -i in the nominal domainForm Glosso-mo\u00b4-Gor-i \u2018buyer\u2019 (c1)a-Ba\u00b4-Gor-i \u2018buyers\u2019 (c2)o-Bo\u00b4-Gor-i \u2018job of buying\u2019 (c14)o-mu\u00b4-sOm-i \u2018educated person\u2019 (c1)a-Ba\u00b4-sOm-i \u2018educated people\u2019 (c2)o-Bu\u00b4-sOm-i \u2018literacy\u2019 (c14)The -O suffix is compatible with instruments with either class 3\/4 prefixes, or 7\/8 prefixes; additionally,some examples are found in c11. These forms roughly have the meaning \u2018some object used to perform X\u2019,where X is the action denoted by the stem.Table 3.5: Stem suffix -O in the nominal domainForm Glosso-mo\u00b4-tir-O \u2018ladder\u2019 (c3)e-me\u00b4-tir-O \u2018ladders\u2019 (c4)o-mo\u00b4-tum-O \u2018seam\u2019 (c3)e-me\u00b4-tum-O \u2018seams\u2019 (c4)e-B\u00b4i-Gaamb-O \u2018sayings\u2019 (c8)o-ru\u00b4-GEEnd-O \u2018journey\u2019 (c11)o-ro-huuNgur-O \u2018key\u2019 (c11)The -u suffix is compatible with patient nominals with either class 1\/2 prefixes, or 7\/8 prefixes; these38forms roughly have the meaning \u2018some entity which undergoes X\u2019, where X is the action denoted by thestem.Table 3.6: Stem suffix -u in the nominal domainForm Glosso-mu-sE\u00b4k-u \u2018person who is laughed at\u2019 (c1)a-Ba-sE\u00b4k-u \u2018people who are laughed at\u2019 (c2)e-ke-me\u00b4r-u \u2018thing which is swallowed\u2019 (c7)e-Be-me\u00b4r-u \u2018things which are swallowed\u2019 (c8)De\u00b4chaine et al., 2017 identify two compatible readings for the -a suffixes. The first is compatible withevent nominals with class 15 prefixes; these forms roughly have the meaning \u2018some X event\u2019, where X isthe action denoted by the stem. The second is compatible with manner nominals with class 5\/6 morphology;these forms roughly have the meaning \u2018the manner of performing X\u2019, where X is the action denoted by thestem.Table 3.7: Stem suffix -a in the nominal domainForm Glosso-ko-me\u00b4r-a \u2018act of swallowing\u2019 (c15)r-ii-Bin-a\u00b4 \u2018musical performance\u2019 (c5)Turning to the mapping between nP and prosodic domains, I propose that the left edge of the noun stem(nP) is mapped to the left edge of a prosodic word (\u03c9). The relevant constraint is reprinted from (16) above.(18) nP to \u03c9-word mappingAl(nP,\u03c9): \u2200 nP \u2203 \u03c9 C(Edge(nP,L), Edge(\u03c9 ,L))For every nP, assign a violation mark if the left edge of that nP does not coincide withthe left edge of some \u03c9-word.39This constraint is sufficient for the behavior of stems that begin with consonants and include two sylla-bles or more. Consider the chart in (19) which illustrates how various forms are evaluated against Al(nP;\u03c9). In the example below, the leftmost column shows syntactic bracketing, the second from the left showsprosodic bracketing and the third shows violation or satisfaction of AL(NP; \u03c9 )Table 3.8: Al(nP; \u03c9) and the \u03c9-word: C-initial nPSyntactic Form Prosodic Form Al(nP,\u03c9) Glossomo\u00b4[nPremi omo\u00b4[\u03c9 remi D \u2018farmer\u2019 (c1)aBa\u00b4[nPremi aBa\u00b4[\u03c9 remi D \u2018farmers\u2019 (c2)rii[nPBur\u0131\u00b4 rii[\u03c9Bur\u0131\u00b4 D \u2018feather\u2019 (c5)ama\u00b4[nPBuri ama\u00b4[\u03c9Buri D \u2018feathers\u2019 (c6)oro[nPs\u0131\u00b4ri oro[\u03c9s\u0131\u00b4ri D \u2018rope\u2019 (c11)tSaa[nPsir\u0131\u00b4 tSaa[\u03c9sir\u0131\u00b4 D \u2018ropes\u2019 (c10)3.4.2 NumPBantu languages, including Nata, are well known for their extensive class systems, manifested by concordialprefixes on nouns, adjectives, and verbs, among others. I follow De\u00b4chaine et al., 2014 in analysing class asa feature which can be realized on Num, among other heads. I follow Carstens, 2008; Kramer, 2015, andDe\u00b4chaine et al., 2014 in characterizing the class marker morphemes as the exponent of Num. Generally,Num merges with nP in order to satisfy an uninterpretable Number feature on the nP.4Table 3.9 presents canonical singular and plural pairings of the noun classes5, in Myers\u2019s (1987) terms,the genders.6 Pairs are separated by a horizontal line. Plural classes are shaded for each pair.7 The classprefix is in bold; for classes 9 and 10, where there are no bolded segments, the prefix is phonologically null.4Because I treat Class as a feature hosted on syntactic heads, it is also the case that class is associated with inner nominalaspect\/sort (to account for the mass reading of c14), Eval (to account for evaluative uses of the class prefixes, which notably neverstack in Nata, as they do in Shona (De\u00b4chaine et al., 2014)), or Det (to account for honorifics.)5I only consider noun class prefixes used in nominal classification. The locative prefixes in Nata occur before the determinerand class prefixes. I consider them separately in \u00a73.5.2.The locative classes have been numbered 16 (ha), 17 (ko), and 18 (mo). Thelatter two function as prepositional prefixes while the former acts as the noun class for locations. Gambarage, 2019 points out thatthe locatives do not have the same agreement pattern with verbs and adjectives that other nouns do, supporting their separation.6This subsection draws heavily from Johannes, 2007 with additional work from Gambarage and Pulleyblank, 2017 and my ownfield notes and elicitation with that author.7Class 14 does not have a plural counterpart as it largely consists of mass nouns (and characteristics.)40For each noun class prefix, I illustrate each of its vowel quality allomorphs (ignoring tonal differences.)Table 3.9: Noun class prefixes (=Num)Class Form Glossc1 o-mo-su\u00b4Be \u2018man\u2019o-mu-ka\u00b4ri \u2018woman\u2019o-mw-\u0131\u00b4ika \u2018killer\u2019c2 a-Ba-su\u00b4Be \u2018men\u2019a-B-\u0131\u00b4ika \u2018killers\u2019c3 o-mo-te\u00b4 \u2018tree\u2019o-mu\u00b4-nwa \u2018mouth\u2019O-mO-tO\u00b4ONg-O \u2018string\u2019u-mw-aat\u00b4i \u2018medicine\u2019c4 e-me-te\u00b4 \u2018trees\u2019e-mi-nwa\u00b4 \u2018mouths\u2019E-mE-tO\u00b4ONg-O \u2018strings\u2019i-mj-aat\u00b4i \u2018medicines\u2019c5 ri-i-Bur\u00b4i \u2018feather\u2019e-r-iinO\u00b4 \u2018tooth\u2019c6 a-ma\u00b4-Buri \u2018feathers\u2019a-m-inO\u00b4 \u2018teeth\u2019c7 e-Ge\u00b4-seku \u2018door\u2019e-Gi-sare\u00b4 \u2018twin\u2019E-kE\u00b4-mEr-O \u2018throat\u2019i-kj-O\u00b4OndE \u2018honey badger\u2019c8 e-Be\u00b4-seku \u2018doors\u2019e-Bi-sare\u00b4 \u2018twins\u2019E-BE\u00b4-mEr-O \u2018throats\u2019i-Bj-O\u00b4OndE \u2018honey badgers\u2019Class Form Glossc9 aa-swe\u00b4 \u2018fish\u2019a-sa\u00b4ato \u2018tilapia\u2019a-m-Ba\u00b4ata \u2018duck\u2019a-n-tSE\u00b4ra \u2018path\u2019a-N-goko\u00b4 \u2018chicken\u2019c10 tSa\u00b4a\u00b4-saato \u2018tilapia\u2019tSa\u00b4a\u00b4-m-Baata \u2018duck\u2019tSa\u00b4a\u00b4-n-tSEra \u2018path\u2019tSaa-N-goko\u00b4 \u2018chicken\u2019c11 o-ro-s\u00b4iri \u2018rope\u2019o-ru\u00b4-saro \u2018bead\u2019O-rO-hO\u00b4ONg-O\u00b4 \u2018sieve\u2019u-rw-aambe\u00b4 \u2018vapor\u2019c10 tSaa-sir\u00b4i \u2018ropes\u2019c14 o-Bu\u00b4-saro \u2018beads\u2019c14 o-Bo-k\u00b4ima \u2018ugali\u2019O-BO-tO\u00b4Or-O \u2018loads\u2019u-Bw-aana\u00b4 \u2018childhood\u2019c15 o-ko\u00b4-Goro \u2018leg\u2019o-ku-BOkO \u2018arm\u2019o-Gw-iitSoreru \u2018flood\u2019c6 a-ma\u00b4-Goro \u2018legs\u2019c16 a-ha-GE\u00b4rO \u2018location\u20193.4.3 DP=\u03c6We will now consider the Determiner Phrase. Determiners merge with NumP in order to satisfy a seman-tic requirement that entities have definiteness specified; to remain consistent, this can be modelled by anuninterpretable definiteness feature on nP being valued by merging nP and Det. Generally, the determineris a vowel preceding the class marker prefix, as we shall see shortly; however, classes 5 and 10 occur withdeterminers that are consonant initial. We shall discuss the consonant initial forms here.There are two types of determiners: i) a segmental allomorph (generally a vowel) preceding the class41marker prefix, as we have seen already, and ii) a phonologically null determiner. The phonologically nulldeterminer occurs in contexts where the speaker is not committing to existence of the nominal entity. Incontrast, a speaker commits to the existence of an entity referred to by a nominal with a phonologicallyovert determiner (Gambarage, 2019). These constitute separate determiner morphemes, the selection ofwhich is dependant on the intended meaning.Table 3.10: Overt and phonologically null determinersExample Gloss(DET-CM-STEM)o- mo-sirikare\u00b4 \u2018the\/a policeman\u2019 (c1)mo-sirikare\u00b4 \u2018no policeman\u2019 (c1)o- mo\u00b4-rem-i \u2018the\/a farmer\u2019 (c1)mo\u00b4-rem-i \u2018no farmer\u2019 (c1)o- mu-ka\u00b4ri \u2018the\/a woman\u2019 (c1)mu-ka\u00b4ri \u2018no woman\u2019 (c1)I propose that the left edge of the DP is mapped to the left edge of a prosodic phrase (\u03c6 ). The relevantconstraint, in (19), is reprinted from (16) above.(19) DP to \u03c6 -phrase mappingAl(DP,\u03c6 ): \u2200 DP \u2203 \u03c9 C(Edge(DP,L), Edge(\u03c6 ,L))For every DP, assign a violation mark if the left edge of that DP does not coincidewith the left edge of some \u03c6 -phrase.This constraint is sufficient for the behavior of DPs with determiners that begin with consonants: class 5 andclass 10.We will begin by considering class 5. Generally, class 5 nominals will have a CV-V-CV(CV. . . ) shape.The interpretation of [ri-] as a DET is supported by alternations between ri- and \/0- which are parallel withthe (semantic) alternations found with other determiners; specifically, the \/0-form determiner is used in non-existential contexts.42Table 3.11: ri-i-CV(CV. . . )Form Glossr\u0131\u00b4-\u0131\u00b4-Buri \u2018feather\u2019 (c5)ri-i-soso\u00b4 \u2018lung\u2019 (c5)r\u0131\u00b4-i-to \u2018leaf\u2019 (c5)ri-i-Bu\u00b4 \u2018ashes\u2019 (c5)The complementary distribution of overt and non-overt determiners acts as a diagnostic for identifyingdeterminers. The overt determiner can be identified by contrasting a nominal in a positive frame (\u2018I saw X\u2019)with that same nominal in a negative frame (\u2018I didn\u2019t see any X\u2019); the phonological material that occurs inthe positive, but not negative frame, is the determiner.Table 3.12: ri- is a determiner: ri-i-CVCV \u223c i-CVCVForm Glossomuka\u00b4ri aGatwa\u00b4 r\u0131\u00b4-i-to \u2018the\/a woman took the\/a leaf\u2019omuka\u00b4ri taagwiire\u00b4 i-to \u2018the\/a woman didn\u2019t take any leaf\u2019Class 5 nominals satisfy the DP to \u03c6 alignment phonotactic, as shown in Table 3.13.Table 3.13: Al(DP; \u03c6 ) and the \u03c6 -phrase: C-initial DP (c5)Syntactic Form Prosodic Form Al(DP,\u03c6 ) Gloss[DPr\u0131\u00b4\u0131\u00b4Buri [\u03c6 r\u0131\u00b4\u0131\u00b4[\u03c9Buri D \u2018feather\u2019 (c5)[DPriisoso\u00b4 [\u03c6 rii[\u03c9soso\u00b4 D \u2018lung\u2019 (c5)Class 10 forms have a determiner with the form tSaa-. The same diagnostic which identifies the class 5determiner can be applied to class 10. Compare the forms in Table 3.14 with those in Table 3.15.43Table 3.14: c10: tSaa-[n, m, N, \u00f1, \/0]-CV(CV. . . )Form GlosstSaa-swe\u00b4 \u2018fish\u2019 (pl.)tSa\u00b4a\u00b4-saato \u2018tilapias\u2019tSa\u00b4a\u00b4-n-duGi \u2018rats\u2019tSa\u00b4a\u00b4-m-Baata \u2018ducks\u2019tSaa-N-goko\u00b4 \u2018chickens\u2019tSaa-\u00f1-akwaaha\u00b4 \u2018armpits\u2019Table 3.15: tSaa- is a determiner: tSaa-(N-). . .\u223c (N-). . .Form Glossomuka\u00b4ri aGatwa\u00b4 tSaa-swe\u00b4 \u2018the\/a woman picked some\/the fish(pl)\u2019omuka\u00b4ri taa\u00f1\u00b4i swe\u00b4 \u2018the woman doesn\u2019t have any fish\u2019omuka\u00b4ri aGatwa\u00b4 tSaa-m-Bor\u00b4i \u2018the\/a woman picked some\/the goats\u2019omuka\u00b4ri taa\u00f1\u00b4i m-Bor\u00b4i \u2018the\/a woman doesn\u2019t have any goats\u2019Class 10 nominals satisfy the alignment phonotactic, as shown in Table 3.16.Table 3.16: Al(DP; \u03c6 ) and the \u03c6 -phrase: C-initial DP (C10)Syntactic Form Prosodic Form Al(DP,\u03c6 ) Gloss[DPtSa\u00b4a\u00b4nduGi [\u03c6 tSa\u00b4a\u00b4[\u03c9nduGi D \u2018rats\u2019 (c10)[DPtSa\u00b4a\u00b4mBaata [\u03c6 tSa\u00b4a\u00b4[\u03c9mBaata D \u2018ducks\u2019 (c10)[DPtSaaNgoko\u00b4 [\u03c6 tSaa[\u03c9Ngoko\u00b4 D \u2018chickens\u2019 (c10)44Table 3.17: Alignment of \u03c6 -phrase edge in c1 with non-overt determinerSyntactic Form Prosodic Form Al(DP,\u03c6 ) Al(NumP,\u03c6 ) Gloss[DP \/0 mo\u00b4remi [\u03c6mo\u00b4[\u03c9 remi D D \u2018no farmer\u2019 (c1)[DP \/0 muka\u00b4ri [\u03c6mu[\u03c9ka\u00b4ri D D \u2018no woman\u2019 (c1)[DP \/0 mosirikare\u00b4 [\u03c6mo[\u03c9sirikare\u00b4 D D \u2018no policeman\u2019 (c1)We will return to the vowel-initial determiners when we consider the distribution of tone across nomi-nals; as we shall see, the vowel-initial determiners are subject to a phonotactic which leads to misalignmentbetween syntactic domains and prosodic ones.3.5 External nominal syntaxIn this section I will consider a set of cases that involve structures built on top of the nominal DP which wehave seen thus far. The relevant morphemes are the associative, the comitative, and the locative. I treat thesethree as a single position class labelled Preposition, or Prep; these morphemes head preposition phrases(PP). The general structure for these phrases is given in Figure 3.5.PPDPNumPnPNumDPrepFigure 3.5: Syntactic structure of a prepositional phrase3.5.1 Associative PrefixThe associative prefix marks a relationship between two nouns, which can range in meaning from possessionto general association (Gregersen & Welmers, 1974; Van de Velde, 2013). The first of these nouns is thehead noun. The associative prefix occurs on the second (non-head) noun and reflects the noun class of thehead noun. The second noun is also referred to as the modifier. In this context, the head noun is the noun45which which the verb agrees. The associative marker can occur with and without the DET.Table 3.18: Associative constructionForm Gloss(DET-)CM1-NSTEM1 ASSOC1-(DET-)CM2-NSTEM2 \u2018N1 of N2\u2019Head noun Modifier nounThe tree representation of the full associative construction is given in Figure 3.6, with the Associativesubstituted for the generic preposition from the structure in Figure 3.5. Crucially, note that the head noun ismerged with the PP headed by the associative-marked modifier noun.DPNumPnPPPDPNumPnPNumDAssocnPNumDFigure 3.6: Syntactic structure of Associative ConstructionThe associative marker has two possible shapes for each class: CV- and C- shapes. The CV- forms forthe associative can be deduced from examining consonant initial DPs. Class 5 and 10 DPs are special in thatthey begin with a consonant, as shown in Table 3.19.46Table 3.19: Consonant initial DPsForm Glossr-\u0131\u00b4\u0131\u00b4-Buri \u2018feather\u2019 (c5)r-\u0131\u00b4\u0131\u00b4-to \u2018leaf\u2019 (c5)tSa\u00b4a\u00b4-n-tS ogu \u2018elephants\u2019 (c10)tSaa-N-goko\u00b4 \u2018chickens\u2019 (c10)The data in Table 3.20 illustrates associative prefix forms corresponding to classes 1-11 heads with aclass 5 modifier. The associative prefixes are displayed in bold.8Table 3.20: Associative prefixes on c5 nounsContext: DET-NUM-N ASSOC-rii-Buri\u2018N.c# of a feather.c5\u2019o-mu-ka\u00b4ri wa\u00b4-r-ii-Buri \u2018woman.c1 of feather.c5\u2019a-Ba-ka\u00b4ri Ba\u00b4-r-ii-Buri \u2018women.c2 of feather.c5\u2019o-mo-s\u00b4isi wo\u00b4-r-ii-Buri \u2018tamarind tree.c3 of feather.c5\u2019e-me-s\u00b4isi Ge\u00b4-r-ii-Buri \u2018tamarind trees.c4 of feather.c5\u2019r-\u00b4i\u00b4i-Buri re\u00b4-r-ii-Buri \u2018feather.c5 of feather.c5\u2019a-ma\u00b4-Buri Ga\u00b4-r-ii-Buri \u2018feathers.c6 feather.c5\u2019e-Ge-ku\u00b4Ba Ge\u00b4-r-ii-Buri \u2018chest.c7 of feather.c5\u2019e-Be-ku\u00b4Ba Be\u00b4-r-ii-Buri \u2018chests.c8 of feather.c5\u2019a-ku\u00b4ru je\u00b4-r-ii-Buri \u2018turtle.c9 of feather.c5\u2019o-ro-s\u00b4iri ro\u00b4-r-ii-Buri \u2018rope.c11 of feather.c5\u2019I propose that the left edge of the PP is mapped to the left edge of a prosodic phrase (\u03c6 ). The relevant8These forms are well formed utterances but are essentially word games insofar as there are not common or even specific itemsto which they refer. In other words, the concept of a \u2018tamarind tree of feather\u2019 is equally puzzling in English as the equivalent is inNata.47constraint in (20) is reprinted from (16) above.(20) PP to \u03c6 -phrase mappingAl(PP,\u03c6 ): \u2200 PP \u2203 \u03c9 C(Edge(PP,L), Edge(\u03c6 ,L))For every PP, assign a violation mark if the left edge of that DP does not coincide withthe left edge of some \u03c6 -phrase.This constraint accounts for the behaviour of all PPs, including the locative and nominative prepositions.(21) Alignment of \u03c6 -phrase edge in complex form[PP{w, wo}+[DP{o, oo, o\u00b4o\u00b4}+[NumP{mo\u00b4, mo}+[NP{rem}+{i}]]] AL(PP,\u03c6)AL(DP,\u03c6)u a. [\u03b9 [\u03c6w-o\u00b4o\u00b4-mo-[\u03c9 rem-i *b. [\u03b9w-o\u00b4o\u00b4-[\u03c6mo-[\u03c9 rem-i *!3.5.2 Locative PrefixThe locative prefixes \u2018ku-\u2019 and \u2018mu-\u2019 add the meaning \u2018on N\u2019, or \u2018in N\u2019, to a noun they precede, respec-tively. The locative prefixes, unlike the associative and comitative, are restricted to occurring with the overtdeterminer. Although the cooccurrence of a locative prefix with a non-overt determiner is theoretically co-herent, the context for such forms has yet to be discovered. See Gambarage, 2019 for more information ondeterminers and how they interact with locatives.Table 3.21: Structure of a locative PPForm GlossLOC-DET-NUM-N \u2018on, in N\u2019The tree representation of a locative prepositional phrase is shown in Figure 3.7.48PPDPNumPnPNumDLocFigure 3.7: Syntactic structure of a Locative Prepositional PhraseThe locative prefixes share the same mapping as the associative and comitative prefixes. As we shallsee in Chapter 4, all three of these prepositions behave the same with respect to tone assignment, preciselybecause they are form a single class for prosodic mapping purposes.3.5.3 ComitativeTwo nouns can be conjoined with the prefix \u2019na-\u2019. This prefix is slightly different from locatives and asso-caitves in that it results in different agreement with the conjoined nouns; for instance, a verb with a conjoinednominal subject may have a subject marker that does not match the class of either of the conjoined subjects.Below are some examples of nominals joined by the comitative prefix with nonovert and overt determiners.Table 3.22: Comitative PPsForm Glossmu-ka\u00b4ri na\u00b4-mO-rOrO \u2018no woman and no fire\u2019 (c1, c3)mo\u00b4-rem-i na-ki-\u00f1aata\u00b4 \u2018no farmer and no belt\u2019 (c1, c7)o-mu-ka\u00b4ri n-oo-mo-su\u00b4Be \u2018woman and man\u2019 (c1, c1)The comitative prefix, like the other prepositions, is mapped to a \u03c6 -phrase. Therefore, the comitativepatterns like the other prepositions for tone assignment.3.6 Mapping SummaryOnly maximal \u03c6 and \u03c9 are relevant. The highest phrase of PP, DP, NumP, will map to maximal \u03c6 . NP mapsto \u03c9 . These mappings are not violated.49Chapter 4ToneIn this chapter I will discuss the phonological representation of tone, the tonal configurations of citation formnouns, noun phrases, and two exceptional class prefixes. The analysis of tone is composed of two generalphonotactics requiring: i) that there is one and only one high tone per \u03c6 -phrase; and ii) that high tone isaligned as far to the left of the \u03c6 -phrase as possible. Some allomorphs impose additional requirements ontone: prepositional prefixes and the class 5 and class 9 prefixes. These aspects of the analysis are encodedin the lexical entries of these morphemes.I begin in Section 4.1 with a brief overview on tone in phonology, followed by general properties of tonein Nata in Section 4.2. In Section 4.3 I layout the relevant prosodic structures for investigating nominal tone.In Section 4.4 I focus on Low tone class. In Section 4.5 I focus on Final H tone class and in Section 4.6I focus on Initial H tone class. For each class I consider basic nominals (with overt and null determiners),nouns in classes 5, 9, and 10, and finally nominals with prepositional prefixes as well as their interactionwith determiners, and the aforementioned classes.I conclude the chapter in Section 4.7, summarizing the properties and analysis of nominal tone in Nata.4.1 Tone in PhonologyWelmers, 1959 is an early example of tone being considered as both a phonological and morphologicalobject. That is, tone can be a property of a particular syllable, but it may also be the realization of amorpheme. Hyman, 2006, 2011; Hyman and Leben, 2001 provides a modern definition in the spirit ofWelmers:50(22) \u201c[A tone language is one] in which an indication of pitch enters into the lexical realization of at leastsome morphemes.\u201d (Hyman 2011:207)\u201cAn indication of pitch\u201d is meant to mean \u201ctone features or any other analytical device whose only functionis to characterize pitch\u201d (Hyman, 2006). One of the interesting properties of this definition is that in orderto be a tone language, tone must be introduced at the lexical level. In Nata, tonal allomorphs are present inthe lexical entries of many morphemes, satisfying this description.Tone has been recognized to have surprising properties in comparison to segmental phonology (Hyman,2009a). For instance, tone melodies can be identified as stable elements, separate from their segmentalhosts (Leben, 1978). The independence of tone from its host segment lead to proposals such as Goldsmith\u2019s(1975, 1976) autosegmental model. Yip, 2001 identifies the list of tone properties shown in Table 4.1.Table 4.1: Tone properties (Yip, 2001 p. 65)Property DescriptionMobility Movement away from point of originStability Survival after loss of original host segmentOne-to-many A single tonal feature shared by two or more segmentsMany-to-one Multiple tonal features surfacing on a single host segmentToneless segments Potentially tone-bearing segments that never acquire phonological toneThese properties of tone are used to motivate an autosegmental account in which tone is autonomousfrom the segment (vowel, mora) that hosts it. Regarding Nata, mobility and stability are of interest as Htones are attracted to a particular edge, but only for some types of forms. At the syllable level, Nata permitsonly one-to-one association between syllables and high tones. In contrast, some Bantu languages allowa single high tone to link to many syllables (Downing, 1990; Kanerva, 1990). Nata permits many-to-oneassociation at the syllable level: a H and L tone to both be linked to a single syllable, forming a contour(though LH is not permitted in Nata). Nata also permits one-to-many association at the mora level: a tonemay be linked to multiple moras within a syllable as in the case of a long level H, or L. We shall examinethe distribution and analysis of tone in more detail in the following section.514.2 Background on tone in NataWe will now discuss the distribution of tone in Nata. Only vowels host tone, and there is a distinctionbetween the short and long vowels. There are five tone patterns for a syllable in Nata: two melodies for lightsyllables (V), and three for heavy syllables (VV). Note the absence of a rising contour on heavy syllables inTable 4.2.Table 4.2: Tone by vowel lengthType Form GlossLight, High omo\u00b4remi \u2018the\/a farmer\u2019 (c1)Light, Low omosu\u00b4Be \u2018the\/a man\u2019 (c1)Heavy, Falling riiBuru\u00b4uNga \u2018the\/a egg\u2019 (c5)Heavy, Level H tSa\u00b4a\u00b4ndoro \u2018the\/a pancreas\u2019 (c10)Heavy, Low riihuundu\u00b4ku \u2018the\/a corpse\u2019 (c5)Heavy, Rise *riihuu\u00b4nduku \u2014In Nata non-vowel segments are not hosts for tone, though this is not universally the case (Yip, 2001).Compare Nata to languages where a nasal consonant bears high tone, for instance, Tuki (Hyman & Biloa,1992). In Tuki, a verbal suffix -m\u00b4 can surface with high tone1Table 4.3: Tuki forms with tone bearing nasal (Hyman & Biloa, 1992)Form Glossa`-n-d\u0131`nga`-m\u00b4 \u2018he loves me\u2019a`-n-da\u00b4nga\u00b4-m\u00b4 \u2018he loses me\u2019Nata does not have nasal segments which are moraic; however there are restrictions on how tone interactswith different moraic structures. Moras (\u00b5) encode a contrast between heavy and light syllables (Hyman,1This morpheme is the only high tone prelinked to a consonant and surfaces with it. The first person object prefix N- is prelinkedto a low tone but does not surface with it.521985). A relevant diagnostic for how tone associates to moras is the behaviour of heavy (bimoraic) syllables;do such syllables behave as one tone bearing unit (TBU) or two? For our purposes, heavy syllables are justthose with long vowels.2I claim that tones are linked to moras in Nata, as supported by a distinction between level H and fallingH on long vowels. The assumed relationship between tones and moras for light (monomoraic) and heavy(bimoraic) syllables is given in Figure 4.1.Light syllables Heavy syllables\u03bcCVH\u03bcCVL \u03bcCVVH\u03bc \u03bcCVVL\u03bc \u03bcCVVH\u03bcL \u03bcCVVL\u03bcH * Figure 4.1: Nata surface inventory of tones by syllable weightNata lacks a rising contour; however, such structures are permitted in other languages (Yip, 2001). Forexample, Heath, 2008 identifies the following pattern for Jamsay, a Dogon (Niger-Congo) language of Mali:light syllables may have high tone or low tone; heavy syllables may be high, low, rising (LH) or falling(HL). Given freely combining high and low tones, we predict exactly the pattern described by Heath.Light syllables Heavy syllables\u03bcCVH\u03bcCVL \u03bcCVVH\u03bc \u03bcCVVL\u03bc \u03bcCVVH\u03bcL \u03bcCVVL\u03bcH Figure 4.2: Jamsay inventory of tones by syllable weightIn Nata, there is a distributional restriction on long vowels with an H tone. If the long vowel is inthe penultimate syllable of a \u03c6 -phrase, then it will have a falling contour. If the long vowel is not in thepenultimate syllable, it will have a level H.2Coda consonants are generally prohibited in Bantu languages of the Mara region (Aunio, 2017), and are not attested in Nata,as shown in Section 2.1.2.53Table 4.4: Realization of H tone on heavy syllablesType Form GlossFalling riiBuru\u00b4uNga \u2018the\/a egg\u2019 (c5)Level H tSa\u00b4a\u00b4ndoro \u2018the\/a pancreas\u2019 (c10)In conjunction with the complementary distribution of level H and falling H on long vowels, recall thatthere are no long vowels in \u03c6 -phrase final positions, as shown in Section 2.1.2. This means that the onlypossible location for falling contours on heavy syllables is in the penultimate syllable of a \u03c6 -phrase.Table 4.5: Only falling H on penultimate long vowelsPosition Form GlossPenultimate [\u03c6omuna\u00b4ata] \u2018the\/a Nata person\u2019 (c1)Final *[\u03c6 riisima\u00b4a] \u2014The distribution of falling high tones (and long vowels) in Nata is reflected in the allomorphs for mor-phemes that can occur in the penultimate position of a \u03c6 -phrase. These are generally root morphemes(possibly with a suffix), which are syntactically the nP, as we saw in Section 3.4.3 For any root that hasa falling H contour on its penultimate syllable, there is not a corresponding allomorph with a level H tone(*na\u00b4a\u00b4ta, *Buru\u00b4u\u00b4Nga).Table 4.6: Root allomorphs with penultimate long vowelsMorpheme Gloss{na\u00b4ata} NATA{Buru\u00b4uNga} EGG3In the verbal domain, extensions and final vowels sometimes fall within this final two syllable window of a \u03c6 -phrase.54However, for forms in class 10, which have long vowels and may occur either in a penultimate orprepenultimate position, both level H and falling H tone allomorphs are attested. As predicted by the distri-bution of HH and HL on long vowels, the HL long allomorphs occur in penultimate positions while the HHallomorphs occurs elsewhere.Table 4.7: c5 & c10 forms with alternating HL\u223cHHForm Gloss[\u03c6 r\u0131\u00b4-i-to] \u2018the\/a leaf\u2019 (c5)[\u03c6 r\u00b4i-\u00b4i-Buri] \u2018the\/a feather\u2019 (c5)[\u03c6 tSa\u00b4a\u00b4-n-tSogu] \u2018the\/a elephants\u2019 (c10)[\u03c6 tSa\u00b4a-ka] \u2018the\/a homes\u2019 (c10)This alternation is reflected in a different configuration of allomorphs for these prefix morphemes thanfor the root morphemes in Table 4.7.Table 4.8: Prefix allomorphs with long vowelsMorpheme Gloss{tSaa, tSa\u00b4a, tSa\u00b4a\u00b4} DET.C10To account for the distribution of contour tone on long vowels, I assume that a general dispreferencefor contours prevents the selection of LH allomorphs, and limits HL to syllables in \u03c6 -phrase penultimateposition.(23) No contour tones Yip, 2001*CONTOUR: Assign a violation mark to any syllable whose moras are associated to more than onetone.In the context of the penultimate syllable, a phonotactic restriction on level H results in the selection of HL55allomorphs.4(24) No level H in penultimate position*LONGH\/PENULT\u03c3 : For every \u03c6 -phrase, assign a violation mark to the penultimate syllable ofthe \u03c6 -phrase if it is bimoraic and both moras are associated to a high tone.This is illustrated in the tableaux below, where a HL is permitted because it satisfies the higher rankedcontextual phonotactic on HL. Note that the default low tone option is not optimal due to a higher rankedconstraint which requires that \u03c6 has an H tone within it, as we shall see later.(25) Evaluation of H on heavy penultimate syllable[DP{tSaa, tSa\u00b4a\u00b4, tSa\u00b4a}+[NumP{ \/0, N}+[nP{ka}]]] *LONGH\/PENULT\u03c3 *CONTOURa. [\u03c6 tSa\u00b4a\u00b4-ka] *!u b. [\u03c6 tSa\u00b4a-ka] *For the remainder of this chapter, I will condense the constraints on contours tones into a single con-straint VV-TONE, defined below.(26) Tone contour constraintVV-TONE: Assign a violation mark to bimoraic syllables associated with LH. Assign a violationmark to bimoraic syllables that are not in \u03c6 -penultimate position and are associatedwith HL. Assign a violation mark to bimoraic syllables that are in \u03c6 -penultimateposition and which have both moras associated to an H.As we shall see throughout this chapter, \u03c6 -phrases in Nata generally have a single high tone syllable(Anghelescu, 2012; Anghelescu et al., 2017). This observation is compatible with an interpretation of highpitch as \u201caccent\u201d, where accent is defined by OBLIGATORINESS and CULMINATIVITY (Hyman, 2006).Obligatoriness is the property which requires that every word (or other domain) have a prominent syllable;culminativity is the property requiring that every word (or other domain) have at most one most prominentsyllable. This situation is not unique to Nata among Eastern Bantu languages to Nata. Schadeberg, 1973identifies Kinga (G.65), also spoken in Tanzania, as a language where words bear at most one high tone.4Nata never allows rising contours on long vowels; I assume a high ranking and unviolated constraint penalizing rising contours.Crucially, the constraint penalizing rising contours outranks the constraint penalizing falling contours: *LH\u226b*HL.56Finally, it is worth noting that many of the properties of nominal tone that are described and analyzed inthis chapter are shared by the verbal tone system. Anghelescu et al., 2017 compare tone on nouns and verbs,demonstrating that all of the nominal tone patterns appear in the verbal domain. Relative to that analysis,the current approach relies on prosodic domains that are related to syntactic structure, as described in thepreceding chapter. One stark difference is that in the verbal domain, a \u03c6 may occur with more than one Htone.4.3 Prosodic and syntactic structure of nominals4.3.1 Bare nominalsThe data in this section are the citation forms of nouns, in other words, the DP, as described in Section 3.4.3.DPNumPnPNumDFigure 4.3: Syntactic structure of a simple nominalTable 4.9: Prosodic structure of a simple nominalForm Gloss[\u03b9 [\u03c6 e-Ge\u00b4-[\u03c9seku]]] \u2018the\/a door\u2019 (c7)4.3.2 Associative ConstructionThe associative construction relates two nouns. The first of these is the head, and the second is the modifier,as described in Section 3.5.1. The associative construction will serve as the exemplar for tone in complexforms. Consider the complex forms in Table 4.10, illustrating the associative (a.), comitative (b.), andlocative (c.); for the associative and comitative constructions, the noun bearing the relevant prefix is givenin bold.57Table 4.10: Complex nominalsForm Glossa. e-Ge\u00b4-seku kjo-o-mu-ka\u00b4ri \u2018the\/a door of the\/a woman\u2019 (c7, c1)b. o-mu-ka\u00b4ri ne\u00b4-e\u00b4-Ge-seku \u2018the\/a woman and the\/a door\u2019 (c1, c7)c. mwe\u00b4-e\u00b4-Ge-seku \u2018on the\/a door\u2019 (c7)The syntactic structure of the associative construction is given in Figure 4.4.DPPPDPNumPnPka\u00b4rimu-o-kjo-DPNumPnPsekuGe\u00b4-e-e-Ge\u00b4-seku kjo-o-mu-ka\u00b4ri \u2018the\/a door of the\/a woman\u2019 (c7, c1)Figure 4.4: Syntactic structure of Associative ConstructionThe associative morpheme is displayed in a box for ease of parsing. The structure also representscomitative and locative forms; however locatives do not occur with a second DP.Like the simple nominals from the previous section, the complex nominals are utterances (\u03b9); however,because the associative construction contains two DPs, the corresponding prosodic structure also containstwo \u03c6 -phrases. Each \u03c6 -phrase contains an \u03c9-word, mapped from the nP.58Table 4.11: Prosodic structure of a complex nominalForm Gloss[\u03b9 [\u03c6 e-Ge\u00b4-[\u03c9seku]][\u03c6kjo-o-mu-[\u03c9ka\u00b4ri]]] \u2018the\/a door of woman\u2019 (c7, c1)Although both DP and PPs are mapped to \u03c6 via the constraints Al(PP; \u03c6 ) and Al(DP; \u03c6 ), respectively,the presence of *RECUR(\u03c6 ) prohibits \u03c6 from containing other \u03c6 \u2019s. Therefore, only the highest ranked ofthe XP to \u03c6 constraints can be satisfied. As we saw in the previous chapter, Al(PP; \u03c6 ) outranks Al(DP; \u03c6 ).The mapping developed in Section 3.3 predicts heads and modifiers should be tonally independent giventhat they are in distinct \u03c6 -phrases and our observations of simplex cases. As we shall see, this prediction isborne out.4.4 Low tone class4.4.1 GeneralLow tone roots are characterized by their general lack of a high tone allomorph. Words in Nata generallyrequire exactly one syllable with high tone. In order to satisfy this restriction, words with low tone rootssurface with high tone prefix allomorphs. Consider the forms in Table 4.12; these forms represent the generalcases involving underived low tone roots.Table 4.12: Underived L-tone root with overt determinerForm GlossD- NUM- na. e- ke\u00b4- Gi \u2018the\/a wasp\u2019 (c7)b. e- ke\u00b4- ribitSi \u2018the\/a waist\u2019 (c7)c. e- Ge\u00b4- seku \u2018the\/a door\u2019 (c7)d. o- ru\u00b4- BErE \u2018the\/some millet\u2019 (c11)In Nata, phonologically null determiners encode non-existential entities, as described in Gambarage,592019. In the context of low tone roots, the alternation between phonologically overt and phonologically nulldeterminer does not influence which syllable will bear high tone.Table 4.13: Underived L-tone root with null determinerForm GlossD- NUM- na. ke\u00b4- Gi \u2018no wasp\u2019 (c7)b. ke\u00b4- ribitSi \u2018no waist\u2019 (c7)c. Ge\u00b4- seku \u2018no door\u2019 (c7)d. ru\u00b4- BErE \u2018no millet\u2019 (c11)Consider the set of forms in Table 4.14 and Table 4.15, which include derived roots. These roots areformed by the addition of a nominal suffix vowel to an a-categorical stem. There is no tonal distinctionbetween low tone roots which are derived and those which are underived.Table 4.14: Derived L-tone root with overt determinerForm GlossD- NUM- na. o- mo\u00b4- rem-i \u2018the\/a farmer\u2019 (c1)b. o- mo\u00b4- Gor-i \u2018the\/a buyer\u2019 (c1)c. o- mo\u00b4- tir-O \u2018the\/a ladder\u2019 (c3)d. o- mo\u00b4- tum-O \u2018the\/a seam\u2019 (c3)60Table 4.15: Derived L-tone root with null determinerForm GlossD- NUM- na. mo\u00b4- rem-i \u2018no farmer\u2019 (c1)b. mo\u00b4- Gor-i \u2018no buyer\u2019 (c1)c. mo\u00b4- tir-O \u2018no ladder\u2019 (c3)d. mo\u00b4- tum-O \u2018no seam\u2019 (c3)The tableaux in (29) and (30) illustrate two general properties of \u03c6 -phrases: they must have a hightone syllable, and they must not have more than one high tone syllable. This property of Nata words hasbeen reported by Anghelescu, 2012, Lam, 2013 (for verbal tone), and Anghelescu et al., 2017. Theseproperties are encoded in the constraints below. Unlike previous formulations of these constraints, thevariants presented below are relative to the \u03c6 -phrase.(27) MONOH and HIGH\u03c6 constraintsMONOH: Assign a violation mark to a \u03c6 that contains more than one H tone.HIGH\u03c6 : Assign a violation mark to a \u03c6 that does not contain at least one H tone.As we shall see later, determiners have low and high tone allomorphs, like class prefixes. However,the high tone allomorph of the determiner does not appear in the general cases. This represents a conflictbetween aligning tone to the left edge of a word, and restricting high tones on onsetless short vowels, asencoded by the constraints below.(28) *(\u03c3 V\u00b4 and Alignment constraints*(\u03c3 V\u00b4: Assign a violation mark to any syllable without an onset and that is also associated with Htone.ALIGN(H,\u03c6 ): Assign a violation mark to the left edge of any \u03c6 that is not aligned with an H tone.We shall see evidence for the crucial ranking of *(\u03c3 V\u00b4 below HIGH\u03c6 when we consider class 5 cases in Sec-tion 4.4.2. More generally regarding rankings, tableaux display crucial constraint rankings for the analysis61throughout the document; however, justification for the rankings may not be provided for a particular set ofconstraints until after the constraints have been introduced.As illustrated in the tableaux below, the positional restriction on high tones out ranks alignment; there-fore, the forms with high tone class prefixes are optimal, despite violating alignment. The candidate in(29d.) loses to (29b.) because the former violates *(\u03c3 V\u00b4.The candidate in (29 a.) loses to (29 b.) because the former violates HIGH\u03c6 . The candidate in (29 c.)loses to (29 b.) because the former violates MONOH.(29) Evaluation of L-tone roots with overt determiner[DP{e, e\u00b4}+[NumP{Ge, Ge\u00b4}+[n{seku}]]]MONOHHIGH\u03a6*(\u03c3V\u00b4AL(H,\u03c6)a. [\u03c6 e-Ge-[\u03c9seku]] *!u b. [\u03c6 e-Ge\u00b4-[\u03c9seku]] *c. [\u03c6 e\u00b4-Ge\u00b4-[\u03c9seku]] *! *d. [\u03c6 e\u00b4-Ge-[\u03c9seku]] *!(30) Evaluation of L-tone roots with null determiner[DP[NumP{Ge, Ge\u00b4}+[n{seku}]]]MONOHHIGH\u03a6*(\u03c3V\u00b4AL(H,\u03c6)a. [\u03c6Ge-[\u03c9seku]] *!u b. [\u03c6Ge\u00b4-[\u03c9seku]]Low tone roots with associative prefixes do not surface with high tone on the class prefix. Instead, a hightone allomorph of the associative prefix is used for such cases. Note that there is an interaction between theassociative prefix and determiner, which results in the associative prefix occurring with a glide and vowelidentical to the determiner vowel; this is explored in Appendix A, but crucially is independent of tone.62Table 4.16: Associative L-tone root with overt determinerForm GlossASSOC- D- NUM- na. e-Ge-s\u00b4ima kjo\u00b4- o\u00b4- mo- rem-i \u2018the\/a well of the\/a farmer\u2019 (c7, c1)b. e-ki-GE\u00b4rO ke\u00b4- e\u00b4- Ge- seku \u2018the\/a thing of the\/a door\u2019 (c7, c7)Associative forms with a null determiner also surface with a high tone allomorph of the associativeprefix; however, there is no vowel coalescence interaction between the associative prefix and the determiner.Table 4.17: Associative L-tone root with null determinerForm GlossASSOC- D- NUM- na. Ge-s\u00b4ima ke\u00b4- mo- rem-i \u2018no well of no farmer\u2019 (c7, c1)b. ki-GE\u00b4rO ke\u00b4- Ge- seku \u2018no thing of no door\u2019 (c7, c7)Forms with an associative prefix and determiner are subject to restrictions on high tone on long vowels,because these forms have a long vowel.(31) VV-TONE constraintVV-TONE: Assign a violation mark to bimoraic syllables associated with LH. Assign a violationmark to bimoraic syllables that are not in \u03c6 -penultimate position and are associatedwith HL. Assign a violation mark to bimoraic syllables that are in \u03c6 -penultimateposition and which have both moras associated to an H.Candidates like those in (32c.), (32d.), (32f.), and (32g.) violate VV-TONE (f. and g. fatally so). VV-TONEpenalizes rising contours in any position, and falling contours outside of the penultimate syllable.The winning candidate in (32h.) satisfies both alignment and *(\u03c3 V\u00b4, because the left edge of the \u03c6 isnot vowel initial. This stands in contrast with the comparable forms with the same root, but without an63associative prefix; recall that in those cases the optimal forms do violate alignment in order to satisfy themore highly ranked *(\u03c3 V\u00b4.(32) Evaluation of associative L-tone root with overt determiner[PP{kjo, kjo\u00b4}+[DP{o, o\u00b4}+[NumP{mo, mo\u00b4}+[nP{rem}+{i}}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONEAL(H,\u03c6)a. [\u03b9 [\u03c6kjo-o-mo-[\u03c9 remi]] *!b. [\u03b9 [\u03c6kjo-o-mo\u00b4-[\u03c9 remi]] *c. [\u03b9 [\u03c6kjo-o\u00b4-mo\u00b4-[\u03c9 remi]] *! * *d. [\u03b9 [\u03c6kjo\u00b4-o-mo\u00b4-[\u03c9 remi]] *! *e. [\u03b9 [\u03c6kjo\u00b4-o\u00b4-mo\u00b4-[\u03c9 remi]] *!f. [\u03b9 [\u03c6kjo-o\u00b4-mo-[\u03c9 remi]] *! *g. [\u03b9 [\u03c6kjo\u00b4-o-mo-[\u03c9 remi]] *!u h. [\u03b9 [\u03c6kjo\u00b4-o\u00b4-mo-[\u03c9 remi]]Forms with an associative prefix and no determiner do not interact with VV-TONE, as there are no longvowels. As illustrated in the tableau in (33), the winning candidate in (33a.) satisfies alignment, while otherpossible candidates fail to satisfy alignment (33b.), MONOH (33c.), or HIGH\u03c6 (33d.).(33) Evaluation of associative L-tone root with null determiner[PP{ke, ke\u00b4}+[DP[NumP{mo, mo\u00b4}+[nP{rem}+{i}}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONEAL(H,\u03c6)u a. [\u03b9 [\u03c6ke\u00b4-mo-[\u03c9 remi]]b. [\u03b9 [\u03c6ke-mo\u00b4-[\u03c9 remi]] *!c. [\u03b9 [\u03c6ke\u00b4-mo\u00b4-[\u03c9 remi]] *!d. [\u03b9 [\u03c6ke-mo-[\u03c9 remi]] *!644.4.2 Classes 5, 9, 10Classes 5, 9, and 10 have different phonological shapes for their prefix and determiner morphemes. Thisinteracts with where high tone is realized. For forms with low tone roots, this is due to the availability ofa consonant initial determiner in class 5, and a non-TBU class marker in class 9. These morpheme shapesinteract with the positional restriction on H tone penalizing H on onsetless vowels.Class 5Class 5 has the same morpho-syntactic structure as other nominals; these structures have a noun stem pre-ceded by a class prefix, which is preceded by a determiner, as described in Section 3.4.3. Class 5 differsfrom the other classes in the forms of its determiner and class prefix. The class 5 determiner is not a V,but rather a CV. The class 5 prefix is not a CV-, but rather a V-. Lafon, 1994; Maho, 1999, 2003, 2009identify similar differences between class 5 and other noun classes in a variety of Bantu languages; oftenthese differences include tonal effects.Table 4.18: Class 5 structuresSyntactic Form Prosodic Form Gloss[DPr\u0131\u00b4-[NumP \u0131\u00b4-[nPBuri [\u03c6 r\u0131\u00b4-\u0131\u00b4-[\u03c9Buri \u2018feather\u2019 (c5)[DPr\u0131\u00b4-[NumP \u0131\u00b4-[nPkonono [\u03c6 r\u0131\u00b4-\u0131\u00b4-[\u03c9konono \u2018a\/the elephant\u2019s trunk\u2019 (c5)[DPr\u0131\u00b4-[NumP \u0131\u00b4-[nPrem-i [\u03c6 r\u0131\u00b4-\u0131\u00b4-[\u03c9 rem-i \u2018a\/the farmer (eval.)\u2019 (c5)Many class 5 forms have corresponding class 6 plurals. Class 6 has the same morpheme shapes as theclasses we have seen in the preceding section; therefore it is useful to compare roots in class 5 with theirplural class 6 forms.Low tone roots in class 5 surface with a level high tone on their initial syllable. This initial syllable iscomposed of a determiner morpheme and a class prefix morpheme. Corresponding class 6 forms surfacewith a high tone on the class marker, as we have seen previously for the general low tone root forms.65Table 4.19: L-tone c5\/6 roots with overt determinerForm GlossD- NUM- na. r\u0131\u00b4- \u0131\u00b4- Buri \u2018a\/the feather\u2019 (c5)b. a- ma\u00b4- Buri \u2018some\/the feathers\u2019 (c6)c. r\u0131\u00b4- \u0131\u00b4- konono \u2018a\/the elephant\u2019s trunk\u2019 (c5)d. a- ma\u00b4- konono \u2018some\/the elephant\u2019s trunks\u2019 (c6)e. r\u0131\u00b4- \u0131\u00b4- rem-i \u2018a\/the farmer (eval.)\u2019 (c5)f. a- ma\u00b4- rem-i \u2018some\/the farmers (eval.)\u2019 (c6)When a low tone root in class 5 occurs with a phonologically null determiner, the class prefix stillsurfaces with a high tone allomorph. This is unexpected given that generally high tone is prenasalized whenit occurs on an onsetless vowel.Table 4.20: L-tone c5\/6 roots with null determinerForm GlossD- NUM- na. \u0131\u00b4- Buri \u2018no feather\u2019 (c5)b. ma\u00b4- Buri \u2018no feathers\u2019 (c6)c. \u0131\u00b4- konono \u2018no elephant\u2019s trunk\u2019 (c5)d. ma\u00b4- konono \u2018no elephant\u2019s trunks\u2019 (c6)e. \u0131\u00b4- rem-i \u2018a\/the farmer (eval.)\u2019 (c5)f. ma\u00b4- rem-i \u2018some\/the farmers (eval.)\u2019 (c6)The evaluation of low tone roots in class 5 is straightforward given what we have seen so far. Like othernouns, these forms must have exactly one H within the \u03c6 . Like the associative forms, the H tone must belevel (given that it is not in penultimate position unless the root is monosyllabic, which we will not discuss66in this work.)The forms in (34b.) and (d.) lose to (34c.) because they violate VV-TONE. The form in (34a.) violatesHIGH\u03c6 . There is no possible form with multiple high tones, given the morphemes and allomorphs involvedin structures with a low tone root in class 5.(34) Evaluation of L-tone c5 root with overt determiner[DP{ri, r\u0131\u00b4}+[NumP{i, \u0131\u00b4}+[nP{Buri}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONEAL(H,\u03c6)a. [\u03b9 [\u03c6 ri-i-[\u03c9Buri]] *!b. [\u03b9 [\u03c6 ri-\u0131\u00b4-[\u03c9Buri]] *! *u c. [\u03b9 [\u03c6 r\u0131\u00b4-\u0131\u00b4-[\u03c9Buri]]d. [\u03b9 [\u03c6 r\u0131\u00b4-i-[\u03c9Buri]] *!The unexpected form of low tone roots in class 5 with a null determiner is a result of HIGH\u03c6 outranking*(\u03c3 V\u00b4, and the available allomorphs with H tone being limited. Because low tone roots generally do not haveany allomorphs with high tone, only the class prefix and determiner can supply the H required by HIGH\u03c6 .While the resulting form does violate *(\u03c3 V\u00b4, this violation is not fatal.(35) Evaluation of L-tone c5 root with null determiner[DP[NumP{i, \u0131\u00b4}+[nP{Buri}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONEAL(H,\u03c6)a. [\u03b9 [\u03c6 i-[\u03c9Buri]] *!u b. [\u03b9 [\u03c6 \u0131\u00b4-[\u03c9Buri]] *When low tone roots in class 5 occur with an associative prefix, the associative prefix occurs with a hightone allomorph. This is predicted by alignment. Corresponding class 6 forms are given in the chart below,these differ in that they surface with a long vowel composed of the associative prefix and determiner.67Table 4.21: Associative L-tone c5\/6 roots with overt determinerForm GlossASSOC- D- NUM- na. o-mu-ka\u00b4ri wo\u00b4- ri- i- rem-i \u2018the\/a woman of a\/the farmer (eval.)\u2019(c1, c5)b. o-mu-ka\u00b4ri wa\u00b4- a\u00b4- ma- rem-i \u2018the\/a woman of some\/a farmers (eval.)\u2019(c1, c6)Low tone roots in class 5 with an associative prefix and null determiner are similar to the general lowtone roots with associative prefixes. This is because the left edge of the \u03c6 is a CV-V and therefore VV-TONE plays a role. Unlike the general low tone roots with associative prefixes where the CV-V is theassociative-determiner; for low tone roots in class 5 with an associative prefix and null determiner the CV-Vis the associative-class prefix.Table 4.22: Associative L-tone c5\/6 roots with null determinerForm GlossASSOC- D- NUM- na. mu-ka\u00b4ri w\u0131\u00b4- \u0131\u00b4- rem-i \u2018no woman of no farmer (eval.)\u2019 (c1, c5)b. mu-ka\u00b4ri wO\u00b4- ma- rem-i \u2018no woman of no farmers (eval.)\u2019 (c1, c6)The optimal form for the low tone roots in class 5 with an associative prefix satisfy alignment. Dueto the number of morphemes involved, there are a number of possible forms that fatally violate MONOH(36d., h., i.), or VV-TONE (36b. & e.). The candidate in (36a.) fatally violates HIGH\u03c6 . The candidate in(36g.) loses to (36c.) because the former violates alignment.68(36) Evaluation of associative L-tone c5 root with overt determiner[PP{wo, wo\u00b4}+[DP{ri, r\u0131\u00b4}+[NumP{i, \u0131\u00b4}+[nP{rem}+{i}}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONEAL(H,\u03c6)a. [\u03b9 [\u03c6wo-ri-i-[\u03c9 remi]] *!b. [\u03b9 [\u03c6wo-r\u0131\u00b4-i-[\u03c9 remi]] *! *u c. [\u03b9 [\u03c6wo\u00b4-ri-i-[\u03c9 remi]]d. [\u03b9 [\u03c6wo\u00b4-r\u0131\u00b4-i-[\u03c9 remi]] *! *e. [\u03b9 [\u03c6wo-ri-\u0131\u00b4-[\u03c9 remi]] *! *g. [\u03b9 [\u03c6wo-r\u0131\u00b4-\u0131\u00b4-[\u03c9 remi]] *!h. [\u03b9 [\u03c6wo\u00b4-ri-\u0131\u00b4-[\u03c9 remi]] *! *i. [\u03b9 [\u03c6wo\u00b4-r\u0131\u00b4-\u0131\u00b4-[\u03c9 remi]] *!The evaluation of a low tone root in class 5 with an associative and a null determiner is identical to theevaluation of a low tone root in class 5 with a determiner (but no associative prefix). This is because theassociative and class prefix form a CV-V sequence, just like the determiner and class prefix do.(37) Evaluation of associative L-tone c5 root with null determiner[PP{wi, w\u0131\u00b4}+[DP[NumP{i, \u0131\u00b4}+[nP{rem}+{i}}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONEAL(H,\u03c6)a. [\u03b9 [\u03c6wi-i-[\u03c9 remi]] *!b. [\u03b9 [\u03c6w\u0131\u00b4-i-[\u03c9 remi]] *!c. [\u03b9 [\u03c6wi-\u0131\u00b4-[\u03c9 remi]] *!u d. [\u03b9 [\u03c6w\u0131\u00b4-\u0131\u00b4-[\u03c9 remi]]Class 9\/10Classes 9 and 10 differ from other classes in that they do not have a tone bearing class marker. Class 9 hasa V- determiner, like we have seen in the general cases; however, class 10 has a CVV- determiner.69Low tone roots in class 9\/10 have an additional allomorph with an initial H tone, as observed from theforms below. This allomorph is morpho-syntactically restricted to occur in the class 9 or class 10 context,as it never surfaces elsewhere, even if it would be phonologically optimal. Crucially, low tone roots do notgenerally have an H initial allomorph; this is strictly a property of class 9\/10 low tone roots.Low tone roots in class 9 with overt determiners surface with a high tone on the root initial syllable. Thesame roots in class 10 surface with the more general all low allomorph, and the class 10 determiner occurswith a high tone allomorph.Table 4.23: L-tone c9\/10 roots with overt determinerForm GlossD- NUM- na. a- su\u00b4kuBi \u2018a\/the hump\u2019 (c9)b. tSa\u00b4a\u00b4- sukuBi \u2018some\/the humps\u2019 (c10)c. a- Nwi\u00b4ina \u2018a\/the crocodile\u2019 (c9)d. tSa\u00b4a\u00b4- Nwiina \u2018some\/the crocodiles\u2019 (c10)e. a- su\u00b4gura \u2018a\/the tradition\u2019 (c9)f. tSa\u00b4a\u00b4- sugura \u2018some\/the traditions\u2019 (c10)g. a- m- Ba\u00b4ata \u2018a\/the duck\u2019 (c9)h. tSa\u00b4a\u00b4- m- Baata \u2018some\/the ducks\u2019 (c10)i. a- n- do\u00b4ro \u2018a\/the pancreas\u2019 (c9)j. tSa\u00b4a\u00b4- n- doro \u2018some\/the pancreases\u2019 (c10)Low tone roots in class 9 with a null determiner occur with the initial H allomorph. Low tone roots inclass 10 with a null determiner occur with the all low allomorph, which is unexpected because the resultingform lacks a high tone.70Table 4.24: L-tone c9\/10 roots with null determinerForm GlossD- NUM- na. su\u00b4kuBi \u2018no hump\u2019 (c9)b. sukuBi \u2018no humps\u2019 (c10)c. Nwi\u00b4ina \u2018a\/the crocodile\u2019 (c9)d. Nwiina \u2018some\/the crocodiles\u2019 (c10)e. su\u00b4gura \u2018no tradition\u2019 (c9)f. sugura \u2018no traditions\u2019 (c10)g. m- Ba\u00b4ata \u2018no duck\u2019 (c9)h. m- Baata \u2018no ducks\u2019 (c10)i. n- do\u00b4ro \u2018no pancreas\u2019 (c9)j. n- doro \u2018no pancreases\u2019 (c10)The evaluation of class 9 forms crucially involves a type of constraint we have not considered thus far,SELECTION; specifically, SEL:L.(38) SELECT:L constraintSEL:L: Assign a violation mark to any allomorph that has a Sel:L restriction not followedby a syllable with L tone.SEL:L is relevant because the class 9\/10 prefix has a selectional restriction; specifically, the prefixselects for a following syllable with low tone5, if it is not followed by such a syllable, a violation is assignedto the output form. The selectional restriction is marked on the class 9\/10 prefix morpheme: {N, \/0}L. SeeSection 2.2.2 for more discussion on selectional restrictions.In addition, notice that there is an allomorph which is indexed to class 9. We will return to the relevanceof this indexation when we consider class 10 forms with a null determiner.5Keep in mind that the form of the c9\/10 prefix being represented is not a nasal consonant, but the c9\/10 morpheme itselfconditions prenasalization. For this reason, it is not itself a syllable, like a CV- prefix is. The syllable following the prenasalizationis interpreted as being the same as the syllable which has the prenasalized onset.71The candidate in (39b.) violates *(\u03c3 V\u00b4, whereas the winning candidate in (39c.) does not. This isessentially because low roots in class 9 and 10 have a H initial allomorph which can satisfy HIGH\u03c6 whilenot violating *(\u03c3 V\u00b4. The candidates in (39a. & d.) violate HIGH\u03c6 and MONOH, respectively.(39) Evaluation of L-tone c9 root with overt determiner[DP{a, a\u00b4}+[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:LAL(H,\u03c6)a. [\u03b9 [\u03c6 a-mL-[\u03c9Baata]] *!b. [\u03b9 [\u03c6 a\u00b4-mL-[\u03c9Baata]] *!u c. [\u03b9 [\u03c6 a-mL-[\u03c9Ba\u00b4atac9]] * *d. [\u03b9 [\u03c6 a\u00b4-mL-[\u03c9Ba\u00b4atac9]] *! * *The evaluation of low tone roots in class 9 with a null determiner is dependant on HIGH\u03c6 , as there areonly two possible surface forms: one with a H and one without.(40) Evaluation of L-tone c9 root with null determiner[DP+[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:LAL(H,\u03c6)a. [\u03b9 [\u03c6mL-[\u03c9Baata]] *!u b. [\u03b9 [\u03c6mL-[\u03c9Ba\u00b4atac9]] * *Low tone roots in class 10 surface with a high level tone allomorph of the determiner. Note that theclass 10 determiner does have a falling high contour ({tSa\u00b4a}), which surfaces with monosyllabic roots (notdiscussed in this work). The primary candidates of interest are the optimal form in (41b.), and the form withthe root initial H tone in (41d.) The loser in (41d.) violates the selectional restriction on the class 9\/10 prefixwhich requires a following low tone.72(41) Evaluation of L-tone c10 root with overt determiner[DP{tSaa, tSa\u00b4a\u00b4, tSa\u00b4a}+[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:LAL(H,\u03c6)a. [\u03b9 [\u03c6 tSaa-mL-[\u03c9Baata]] *!u b. [\u03b9 [\u03c6 tSa\u00b4a\u00b4-mL-[\u03c9Baata]]c. [\u03b9 [\u03c6 tSa\u00b4a-mL-[\u03c9Baata]] *!d. [\u03b9 [\u03c6 tSaa-mL-[\u03c9Ba\u00b4atac9]] *! *e. [\u03b9 [\u03c6 tSa\u00b4a\u00b4-mL-[\u03c9Ba\u00b4atac9]] *! *f. [\u03b9 [\u03c6 tSa\u00b4a-mL-[\u03c9Ba\u00b4atac9]] *! * *The surprising case of the low tone roots in class 10 with a null determiner is not correctly predictedby our current model of Nata phonology. This is because the observed surface form violates HIGH\u03c6 . Onerepair to this case is to invoke an undominated MORPHEME PREFERENCE constraint which penalizes anindexed form for occurring outside of the morphological context that it is indexed to; crucially, in this casethe constraint must penalize a root for not matching the class features of a class prefix, as by assumption thenull determiner lacks class features; consider the definition below in (42).(42) MORPHPREF:ROOT-CLASS MARKERMP:RT-CM: Assign a violation mark to any form that has a root marked for a class featureand that class feature is not the same as the class feature of the preceding classprefix within the \u03c6 .The constraint must be undominated because it outranks HIGH\u03c6 , which is the only constraint violated bythe observed surface form. As we shall see when we consider initial H roots, this constraint is independentlynecessary to decide between HL and LH allomorphs, as both are equally phonologically well formed. Un-fortunately, for the initial H cases, the ranking of MP:RT-CM is below HIGH\u03c6 and therefore, it will notsolve the misprediction in (43). We return to this constraint in for initial H roots in (70).73(43) Evaluation of L-tone c10 root with null determiner[DP[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:LAL(H,\u03c6)\/ a. [\u03b9 [\u03c6mL-[\u03c9Baata]] *!u b. [\u03b9 [\u03c6mL-[\u03c9Ba\u00b4atac9]] * *When low tone roots in class 9 and 10 occur with an associative prefix, the associative and determineroccur with high tone; this case is comparable to the general low tone root cases in so far as the associativeprefix and determiner constitute a long vowel, as discussed in Appendix A.Table 4.25: Associative L-tone c9\/10 roots with overt determinerForm GlossASSOC- D- NUM- na. e-r\u00b4i-ino rja\u00b4- a\u00b4- Nwiina \u2018the\/a tooth of a\/the crocodile\u2019(c5, c9)b. e-r\u00b4i-ino r\u0131\u00b4- tSaa Nwiina \u2018the\/a tooth of some\/a crocodiles\u2019(c5, c10)c. o-mo-r\u00b4iGo wa\u00b4- a\u00b4- m- Baata \u2018the\/a load of a\/the ducks\u2019(c3, c9)d. o-mo-r\u00b4iGo wO\u00b4- tSaa- m- Baata \u2018the\/a load of some\/a ducks\u2019(c3, c10)When low tone roots in class 9 and 10 occur with an associative prefix and a null determiner the asso-ciative prefix occurs with a high tone allomorph.74Table 4.26: Associative L-tone c9\/10 roots with null determinerForm GlossASSOC- D- NUM- na. r\u00b4i-ino re\u00b4- Nwiina \u2018no tooth of no crocodile\u2019(c5, c9)b. r\u00b4i-ino re\u00b4- Nwiina \u2018no tooth of no crocodiles\u2019(c5, c10)c. mo-r\u00b4iGo wO\u00b4- m- Baata \u2018no load of no ducks\u2019(c3, c9)d. mo-r\u00b4iGo wO\u00b4- m- Baata \u2018no load of no ducks\u2019(c3, c10)The optimal form satisfies SEL:L. In contrast, the losing candidate in (44e.) violates SEL:L. Moreover,the form in (44e.) fails to satisfy alignment, which the optimal candidate does satisfy.As we have seen previously for low tone roots, candidates which fail to satisfy H IGH\u03c6 (44a.) andMONOH (44g.-i.) in general lose to the optimal candidate. Again, as before, candidates which violateVV-TONE (44b. & c.) lose to the optimal candidate.75(44) Evaluation of associative L-tone c9 root with overt determiner[PP{wa, wa\u00b4}+[DP{a, a\u00b4}+[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:LAL(H,\u03c6)a. [\u03b9 [\u03c6wa-a-mL-[\u03c9Baata]] *!b. [\u03b9 [\u03c6wa\u00b4-a-mL-[\u03c9Baata]] *!c. [\u03b9 [\u03c6wa-a\u00b4-mL-[\u03c9Baata]] *! *u d. [\u03b9 [\u03c6wa\u00b4-a\u00b4-mL-[\u03c9Baata ]]e. [\u03b9 [\u03c6wa-a-mL-[\u03c9Ba\u00b4atac9]] *! *g. [\u03b9 [\u03c6wa\u00b4-a-mL-[\u03c9Ba\u00b4atac9 ]] *! * *h. [\u03b9 [\u03c6wa-a\u00b4-mL-[\u03c9Ba\u00b4atac9]] *! * * *i. [\u03b9 [\u03c6wa\u00b4-a\u00b4-mL-[\u03c9Ba\u00b4atac9 ]] *! *The evaluation of low tone roots in class 9 with a null determiner is essentially the same as with theevaluation for forms with an overt determiner, except VV-TONE plays no role, as there are no long vowelswhich could have high tone. As in other forms with the class 9\/10 prefix, using the H initial allomorph ofthe root results in violations of SEL:L, as in candidates (45c. & d.).(45) Evaluation of associative L-tone c9 root with null determiner[PP{wO, wO\u00b4}+[DP[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:LAL(H,\u03c6)a. [\u03b9 [\u03c6wO-mL-[\u03c9Baata]] *!u b. [\u03b9 [\u03c6wO\u00b4-mL-[\u03c9Baata]]c. [\u03b9 [\u03c6wO-mL-[\u03c9Ba\u00b4atac9]] *! *d. [\u03b9 [\u03c6wO\u00b4-mL-[\u03c9Ba\u00b4atac9]] *! *The evaluation of low tone roots in class 10 with an associative prefix is similar to cases we have alreadyseen in so far as the optimal candidate satisfies alignment. The candidate with a high tone allomorph of thec10 determiner (46c.) loses to the optimal candidate in (46b.) because the former violates alignment.76(46) Evaluation of associative L-tone c10 root with overt determiner[PP{wO, wO\u00b4}+[DP{tSaa, tSa\u00b4a\u00b4}+[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:LAL(H,\u03c6)a. [\u03b9 [\u03c6wO-tSaa-m-[\u03c9Baata]] *!u b. [\u03b9 [\u03c6wO\u00b4-tSaa-m-[\u03c9Baata]]c. [\u03b9 [\u03c6wO-tSa\u00b4a\u00b4-m-[\u03c9Baata]] *!d. [\u03b9 [\u03c6wO\u00b4-tSa\u00b4a\u00b4-m-[\u03c9Baata]] *!e. [\u03b9 [\u03c6wO-tSaa-m-[\u03c9Ba\u00b4atac9]] *! *g. [\u03b9 [\u03c6wO\u00b4-tSaa-m-[\u03c9Ba\u00b4atac9]] *! *h. [\u03b9 [\u03c6wO-tSa\u00b4a\u00b4-m-[\u03c9Ba\u00b4atac9]] *! * *i. [\u03b9 [\u03c6wO\u00b4-tSa\u00b4a\u00b4-m-[\u03c9Ba\u00b4atac9]] *! *The evaluation of low tone roots in class 10 with an associative prefix and a null determiner dependscrucially on the ranking of SEL:L over HIGH\u03c9 .(47) HIGH\u03c9 constraintHIGH\u03c9: Assign a violation mark to any \u03c9 that does not contain an H tone.The optimal form in (48b.) violates the lower ranked constraint, while the losing candidate in (48c.) satisfiesHIGH\u03c9 but violates the more highly ranked SEL:L. Additionally, the optimal candidate satisfies alignment.77(48) Evaluation of associative L-tone c10 root with null determiner[PP{wO, w\u00b4O}+[DP[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:LHIGH\u03c9AL(H,\u03c6)a. [\u03b9 [\u03c6wO-m-[\u03c9Baata]] *! *u b. [\u03b9 [\u03c6wO\u00b4-m-[\u03c9Baata]] *c. [\u03b9 [\u03c6wO-m-[\u03c9Ba\u00b4atac9]] *! *d. [\u03b9 [\u03c6wO\u00b4-m-[\u03c9Ba\u00b4atac9]] *!4.4.3 Data summaryIn this section we have observed the surface forms involving low tone roots will generally surface with a Has far left as possibly, in alignment with the \u03c6 , but also subject to the condition that onsetless short vowelsnot bear high tone.78Table 4.27: Summary: Low tone root classD-NUM-n Glossa) o-mo\u00b4-[rem-i] \u2018the\/a farmer\u2019 (c1)b) \/0-mo\u00b4-[rem-i] \u2018some farmer\u2019 (c1)c) r\u0131\u00b4-\u0131\u00b4-[rem-i] \u2018the\/a farmer (eval.)\u2019 (c5)d) \/0-\u0131\u00b4-[rem-i] \u2018some farmer (eval.)\u2019 (c5)e) a-m-Ba\u00b4ata \u2018the\/a duck\u2019 (c9)f) \/0-m-Ba\u00b4ata \u2018some duck\u2019 (c9)g) tSa\u00b4a\u00b4-m-Baata \u2018the\/a ducks\u2019 (c10)h) \/0-m-Baata \u2018some ducks\u2019 (c10)D-NUM-n ASSOC-D-NUM-n Glossi) e-Ge\u00b4-seku kjo\u00b4-o\u00b4-mo-[rem-i] \u2018the\/a door of farmer\u2019 (c7, c1)j) \/0-Ge\u00b4-seku ke\u00b4- \/0-mo-[rem-i] \u2018some woman of some farmer\u2019 (c7, c1)k) e-Ge\u00b4-seku ke\u00b4-ri-i-[rem-i] \u2018the\/a door of the\/a farmer (eval.)\u2019 (c7, c5)l) \/0-Ge\u00b4-seku k\u00b4i- \/0-\u00b4i-[rem-i] \u2018some door of some farmer (eval.)\u2019 (c7, c5)m) o-mo-r\u00b4iGo wa\u00b4-a\u00b4-m-Baata \u2018the\/a load of the\/a duck\u2019 (c7, c9)n) \/0-mo-r\u00b4iGo wO\u00b4- \/0-m-Baata \u2018some load of some duck\u2019 (c7, c9)o) o-mo-r\u00b4iGo wO\u00b4-tSaa-m-Baata \u2018the\/a load of the\/a ducks\u2019 (c7, c10)p) \/0-mo-r\u00b4iGo wO\u00b4- \/0-m-Baata \u2018some load of some ducks\u2019 (c7, c10)4.4.4 Analysis summaryThe analysis of low tone roots largely depends on the lexicon. Most low tone roots have no allomorphswith high tone; therefore, satisfaction of HIGH\u03c6 depends on the material to the left of the root. Alignmentwould prefer that the H tone be on the leftmost TBU in \u03c6 ; however, *(\u03c3 V\u00b4 prevents forms with high tone V-determiners from being optimal.79Class 5 cases rely on prohibitions on high tone over long syllables, which is encoded by VV-TONE.Class 9\/10 cases involve an additional root allomorph with high tone; however this allomorph is restrictedby morphological context, and crucially the phonotactic penalizing high tone on the syllable following thedeterminer.The puzzling case of class 10 forms with a null determiner remain unexplained by the analysis at hand.This case is puzzling because it violates H IGH\u03c6 , which is otherwise never violated by a surface form.Comparing the c9 and c10 surface forms with null determiners shows that they are only differentiated by theform of the root; in c9 the root occurs with the H initial allomorph, whereas in c10 the root occurs all low.Nevertheless, the mechanism that would differentiate between matched root\/prefix class is demonstrablyranked lower than the case of class 10 forms with a null determiner would require, as we shall see when weconsider initial H forms.4.4.5 Morph set relationsNot every possible distribution of tonal allomorphs are attested; for example, there is not a valid lexical entryfor a prefix with only a high tone allomorph. Following Archangeli and Pulleyblank, 2021, I will examinethe relations between morphs that are minimally different, as these are the types of pairs that a learner wouldconsider when forming lexical entries, and then later when expanding their lexicon.Prefixes generally occur with H and L tone allomorphs as we have seen for class prefixes, determiners,and prepositional prefixes. This is illustrated by the relation below:(49) Morph set relation: Tone (prefixes)MSRPfTone: Pairs of otherwise identical prefix allomorphs differ only with respect to an H tonein one being L in the other.{Mi,M j} Mi: [H]M j: [L]A learner can form the following condition, to ensure that they can produce novel forms when they encountera prefix that does not yet have a well established lexical entry. Essentially, this condition would state that aprefix has a H allomorph and has a L allomorph, or else is not a well formed prefix lexical entry.80(50) Morph set condition: Tone (prefixes)With respect to MSRPfTone, a nominal prefix morph set is ill-formed if M j is present and there is nocorresponding Mi, or if Mi is present and there is no corresponding M jMSCPfTone: *{M j,\u00acMi}*{Mi,\u00acM j}Roots occur in one of three types, as we will further explore in this chapter. They have different types oflexical entries than prefixes do, and a learner will need to discover this. For instance, if a learner posits thatthe condition in (50) above is totally general of the lexicon, they will incorrectly predict that roots shouldhave uniform tonal behaviour, and they observably do not.Low tone roots generally have only one allomorph, their namesake all low form. This is reflected in thelack of a relation, except in the case of class 9 roots, which have an allomorph with an initial high tone.(51) Morph set relation: Tone (Class 9 root)MSRC9Tone: Pairs of otherwise identical root allomorphs in class 9 differ only with respect to aninitial H tone in one being L in the other.{Mi,M j} Mi: [HL. . . ]M j: [L. . . ]If a root is in class 9 and has a low allomorph, it must have a initial H allomorph, and vice versa. Thisis true of low tone roots, but also of initial H roots; although in the case of initial H roots, they have bothallomorphs in all classes, as we shall see in Section 4.6.(52) Morph set condition: Tone (Low tone class 9 root)With respect to MSRC9Tone, a c9 nominal root morph set is ill-formed if M j is present and there is nocorresponding Mi, or if Mi is present and there is no corresponding M jMSCC9Tone: *{M j,\u00acMi}*{Mi,\u00acM j}814.5 Final H tone class4.5.1 GeneralForms with final H roots always occur with a high tone at the right edge of the \u03c6 (and \u03c9). They do not varyacross context (with overt\/null determiner, with\/without prepositional prefix, in class 9, 10, 5).Final H roots contrast with low roots in that they have a high tone allomorph. The existence of suchan allomorph will satisfy H IGH\u03c6 (and H IGH\u03c9); however, this morph is at odds with alignment, whichfavours forms with high tone at the left edge of the \u03c6 . Because final H roots only have one allomorph, wordsincluding these roots will always have a \u03c6 final H. Morphemes to the left of the root are prevented fromoccurring with H allomorphs by the highly ranked MONOH.In essence this class of roots is straightforward: the single root allomorph is invariant, all morphs to theleft of the root are low. Because prefix morphemes always have a low tone allomorph, there is no conflictthat could force violations of MONOH.Table 4.28: Underived Final H-tone roots with overt determinerForm GlossD- NUM- na. e- ke- hurEErO\u00b4 \u2018the\/a cooking pot\u2019 (c7)b. e- Gi- sEErO\u00b4 \u2018the\/a hide\u2019 (c7)c. e- Ge- teete\u00b4 \u2018the\/a husk\u2019 (c7)d. o- ro- tSuumbe\u00b4 \u2018the\/a cow shed\u2019 (c11)e. o- mo- sirikare\u00b4 \u2018the\/a policeman\u2019 (c1)82Table 4.29: Underived Final H-tone roots with phonologically null determinerForm GlossD- NUM- na. ke- hurEErO\u00b4 \u2018no cooking pot\u2019 (c7)b. Gi- sEErO\u00b4 \u2018no hide\u2019 (c7)c. Ge- teete\u00b4 \u2018no husk\u2019 (c7)d. ro- tSuumbe\u00b4 \u2018no cow shed\u2019 (c11)e. mo- sirikare\u00b4 \u2018no policeman\u2019 (c1)Final H roots are not derived from a-categorical stems and nominal suffixes. This property is not whollyunexpected, given that the nominal suffixes only have an allomorph with low tone.The evaluation of final H tone roots is essentially the same across contexts. The basic case belowillustrates this. The optimal candidate in (53a.) violates alignment, but unlike the other candidates in (53b.-d.), it satisfies MONOH.For completeness sake, the same set of cases we considered for the low tone root class, and which wewill consider for the initial H root class, is shown for the final H roots.(53) Evaluation of final H tone root with overt determiner[DP{e, e\u00b4}L+[NumP{ Gi, Ge\u00b4}+[n{teete\u00b4}]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03c6 e-Ge-[\u03c9 teete\u00b4]] *b. [\u03c6 eL-Ge\u00b4-[\u03c9 teete\u00b4]] *! * *c. [\u03c6 e\u00b4L-Ge\u00b4-[\u03c9 teete\u00b4]] *! * *d. [\u03c6 e\u00b4L-Ge-[\u03c9 teete\u00b4]] *! *83(54) Evaluation of final H tone root with null determiner[DP[NumP{ Gi, G\u0131\u00b4}+[n{teete\u00b4}]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03c6Ge-[\u03c9 teete\u00b4]] *b. [\u03c6Ge\u00b4-[\u03c9 teete\u00b4]] *!Table 4.30: Associative final H-tone root with overt determinerForm GlossASSOC- D- NUM- na. e-ki-GE\u00b4rO kjo- o- ro- tSuumbe\u00b4 \u2018the\/a thing of the\/a cow shed\u2019 (c7, c1)b. e-Ge-s\u00b4ima kjo- o- mo- sirikare\u00b4 \u2018the\/a thing of the\/a policeman\u2019 (c7, c1)Table 4.31: Associative final H-tone root with null determinerForm GlossASSOC- D- NUM- na. ki-GE\u00b4rO ke- ro- tSuumbe\u00b4 \u2018the\/a thing of the\/a cow shed\u2019 (c7, c1)b. Ge-s\u00b4ima ke- mo- sirikare\u00b4 \u2018the\/a thing of the\/a policeman\u2019 (c7, c1)84(55) Evaluation of associative final H-tone root with overt determiner[PP{kjoH, kjo\u00b4, }+[DP{o, o\u00b4}L+[NumP{mo, mo\u00b4}+[nP{sirikare\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6kjoH-oL-mo-[\u03c9sirikare\u00b4]] * *b. [\u03b9 [\u03c6kjoH-oL-mo\u00b4-[\u03c9sirikare\u00b4]] *! * *c. [\u03b9 [\u03c6kjoH-o\u00b4L-mo-[\u03c9sirikare\u00b4]] *! * *d. [\u03b9 [\u03c6kjo\u00b4-oL-mo-[\u03c9sirikare\u00b4]] *!e. [\u03b9 [\u03c6kjo\u00b4-o\u00b4L-mo-[\u03c9sirikare\u00b4]] *!f. [\u03b9 [\u03c6kjo\u00b4-o\u00b4L-mo\u00b4-[\u03c9sirikare\u00b4]] *! *g. [\u03b9 [\u03c6kjo\u00b4-oL-mo\u00b4-[\u03c9sirikare\u00b4]] *! *h. [\u03b9 [\u03c6kjoH-o\u00b4L-mo\u00b4-[\u03c9sirikare\u00b4]] *! * *(56) Evaluation of associative final H-tone root with null determiner[PP{keH, ke\u00b4, }+[DP[NumP{mo, mo\u00b4}+[nP{sirikare\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6keH-mo-[\u03c9sirikare\u00b4]] * *b. [\u03b9 [\u03c6keH-mo\u00b4-[\u03c9sirikare\u00b4]] *! *c. [\u03b9 [\u03c6ke\u00b4-mo-[\u03c9sirikare\u00b4]] *!d. [\u03b9 [\u03c6ke\u00b4-mo\u00b4-[\u03c9sirikare\u00b4]] *!854.5.2 Classes 5, 9, 10Class 5Table 4.32: Final H-tone c5\/6 roots with overt determinerForm GlossD- NUM- na. ri- i- tooka\u00b4 \u2018store\u2019 (c5)b. a- ma- tooka\u00b4 \u2018stores\u2019 (c6)c. ri- i- Birika\u00b4 \u2018kettle\u2019 (c5)d. a- ma- Birika\u00b4 \u2018kettles\u2019 (c6)Table 4.33: Final H-tone c5\/6 roots with null determinerForm GlossD- NUM- na. i- tooka\u00b4 \u2018store\u2019 (c5)b. ma- tooka\u00b4 \u2018stores\u2019 (c6)c. i- Birika\u00b4 \u2018kettle\u2019 (c5)d. ma- Birika\u00b4 \u2018kettles\u2019 (c6)86(57) Evaluation of Final H-tone c5 root with overt determiner[DP{ri, r\u0131\u00b4}L+[NumP{i, \u0131\u00b4}+[nP{Birika\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6 ri-i-[\u03c9Birika\u00b4]] *b. [\u03b9 [\u03c6 riL-\u0131\u00b4-[\u03c9Birika\u00b4]] *! *c. [\u03b9 [\u03c6 r\u0131\u00b4L-\u0131\u00b4-[\u03c9Birika\u00b4]] *!d. [\u03b9 [\u03c6 r\u0131\u00b4L-i-[\u03c9Birika\u00b4]] *! *(58) Evaluation of Final H-tone c5 root with null determiner[DP[NumP{i, \u0131\u00b4}+[nP{Birika\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6 i-[\u03c9Birika\u00b4]] *b. [\u03b9 [\u03c6 \u0131\u00b4-[\u03c9Birika\u00b4]] *! *Table 4.34: Associative final H-tone c5\/6 roots with overt determinerForm GlossASSOC- D- NUM- na. o-mu-ka\u00b4ri wo- ri- i- tooka\u00b4 \u2018the\/a woman of a\/the store\u2019(c1, c5)b. o-mu-ka\u00b4ri wa- a- ma- tooka\u00b4 \u2018the\/a woman of some\/a stores\u2019(c1, c6)87Table 4.35: Associative final H-tone c5\/6 roots with null determinerForm GlossASSOC- D- NUM- na. mu-ka\u00b4ri wi- i- tooka\u00b4 \u2018no woman of no store\u2019(c1, c5)b. mu-ka\u00b4ri wO- ma- tooka\u00b4 \u2018no woman of no stores\u2019(c1, c6)(59) Evaluation of associative final H-tone c5 root with overt determiner[PP{woH, wo\u00b4}+[DP{ri, r\u0131\u00b4}L+[NumP{i, \u0131\u00b4}+[nP{tooka\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6woH-riL-i-[\u03c9 tooka\u00b4]] * *b. [\u03b9 [\u03c6woH-r\u0131\u00b4L-i-[\u03c9 tooka\u00b4]] *! * *c. [\u03b9 [\u03c6woH-r\u0131\u00b4L-\u0131\u00b4-[\u03c9 tooka\u00b4]] *! *d. [\u03b9 [\u03c6woH-riL-\u0131\u00b4-[\u03c9 tooka\u00b4]] *! * *e. [\u03b9 [\u03c6wo\u00b4-riL-i-[\u03c9 tooka\u00b4]] *!f. [\u03b9 [\u03c6wo\u00b4-r\u0131\u00b4L-i-[\u03c9 tooka\u00b4]] *! *g. [\u03b9 [\u03c6wo\u00b4-r\u0131\u00b4L-\u0131\u00b4-[\u03c9 tooka\u00b4]] *!h. [\u03b9 [\u03c6wo\u00b4-riL-\u0131\u00b4-[\u03c9 tooka\u00b4]] *! *88(60) Evaluation of associative final H-tone c5 root with null determiner[PP{wiH, w\u0131\u00b4}+[DP[NumP{i, \u0131\u00b4}+[nP{tooka\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6wiH-i-[\u03c9 tooka\u00b4]] * *b. [\u03b9 [\u03c6wiH-\u0131\u00b4-[\u03c9 tooka\u00b4]] *! * * *c. [\u03b9 [\u03c6w\u0131\u00b4-i-[\u03c9 tooka\u00b4]] *! *d. [\u03b9 [\u03c6w\u0131\u00b4-\u0131\u00b4-[\u03c9 tooka\u00b4]] *!Class 9\/10Table 4.36: Final H-tone class 9\/10 roots with overt determinerForm GlossD- NUM- na. a- \u00f1- akwaaha\u00b4 \u2018a\/the armpit\u2019 (c9)b. tSaa- \u00f1- akwaaha\u00b4 \u2018some\/the armpits\u2019 (c10)Table 4.37: Final H-tone class 9\/10 roots with null determinerForm GlossD- NUM- na. \u00f1- akwaaha\u00b4 \u2018no armpit\u2019 (c9)b. \u00f1- akwaaha\u00b4 \u2018no armpits\u2019 (c10)89(61) Evaluation of Final H-tone c9 root with overt determiner[DP{a, a\u00b4}L+[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6 aL-\u00f1L-[\u03c9 tooka\u00b4]] *b. [\u03b9 [\u03c6 a\u00b4L-\u00f1L-[\u03c9 tooka\u00b4]] *! *(62) Evaluation of Final H-tone c9 root with null determiner[DP[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6\u00f1L-[\u03c9 tooka\u00b4]] *(63) Evaluation of Final H-tone c10 root with overt determiner[DP{tSaa, tSa\u00b4a\u00b4, tSa\u00b4a}+[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6 tSaa-\u00f1L-[\u03c9 tooka\u00b4]] *b. [\u03b9 [\u03c6 tSa\u00b4a\u00b4-\u00f1L-[\u03c9 tooka\u00b4]] *!c. [\u03b9 [\u03c6 tSa\u00b4a-\u00f1L-[\u03c9 tooka\u00b4]] *! *(64) Evaluation of Final H-tone c10 root with null determiner[DP[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6\u00f1L-[\u03c9 tooka\u00b4]] *90Table 4.38: Associative Final H-tone c9\/10 roots with overt determinerForm GlossASSOC- D- NUM- na. e-ki-GE\u00b4rO kja a- \u00f1- akwaaha\u00b4 \u2018the\/a thing of a\/the armpit\u2019(c7, c9)b. e-ki-GE\u00b4rO GE tSaa \u00f1- akwaaha\u00b4 \u2018the\/a thing of some\/a armpits\u2019(c7, c10)Table 4.39: Associative Final H-tone c9\/10 roots with null determinerForm GlossASSOC- D- NUM- na. ki-GE\u00b4rO kE \u00f1- akwaaha\u00b4 \u2018no thing of no armpit\u2019(c7, c9)b. ki-GE\u00b4rO kE \u00f1- akwaaha\u00b4 \u2018no thing of no armpits\u2019(c7, c10)(65) Evaluation of associative Final H-tone c9 root with overt determiner[PP{kjaH, kja\u00b4}+[DP{a, a\u00b4}L+[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6kjaH-aL-\u00f1L-[\u03c9akwaaha\u00b4]] * *b. [\u03b9 [\u03c6kja\u00b4-aL-\u00f1L-[\u03c9akwaaha\u00b4]] *! *c. [\u03b9 [\u03c6kjaH-a\u00b4L-\u00f1L-[\u03c9akwaaha\u00b4]] *! * *d. [\u03b9 [\u03c6kja\u00b4-a\u00b4L-\u00f1L-[\u03c9akwaaha\u00b4]] *!91(66) Evaluation of associative Final H-tone c9 root with null determiner[PP{kEH, kE\u00b4}+[DP[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6kEH-\u00f1L-[\u03c9akwaaha\u00b4]] * *b. [\u03b9 [\u03c6kE\u00b4-\u00f1L-[\u03c9akwaaha\u00b4]] *!(67) Evaluation of associative Final H-tone c10 root with overt determiner[PP{GEH, GE\u00b4}+[DP{tSaa, tSa\u00b4a\u00b4, tSa\u00b4a}L+[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6GEH-tSaaL-\u00f1L-[\u03c9akwaaha\u00b4]] * *b. [\u03b9 [\u03c6GEH-tSa\u00b4a\u00b4L-\u00f1L-[\u03c9akwaaha\u00b4]] *! *c. [\u03b9 [\u03c6GEH-tSa\u00b4aL-\u00f1L-[\u03c9akwaaha\u00b4]] *! * *d. [\u03b9 [\u03c6GE\u00b4-tSaaL-\u00f1L-[\u03c9akwaaha\u00b4]] *!e. [\u03b9 [\u03c6GE\u00b4-tSa\u00b4a\u00b4L-\u00f1L-[\u03c9akwaaha\u00b4]] *!f. [\u03b9 [\u03c6GE\u00b4-tSa\u00b4aL-\u00f1L-[\u03c9akwaaha\u00b4]] *! *(68) Evaluation of associative Final H-tone c10 root with null determiner[PP{kEH, kE\u00b4}+[DP[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6kEH-\u00f1-[\u03c9akwaaha\u00b4]] * *b. [\u03b9 [\u03c6kE\u00b4-\u00f1-[\u03c9akwaaha\u00b4]] *!924.5.3 Data summaryTable 4.40: Summary: Final H tone classD-NUM-n Glossa) o-mo-sirikare\u00b4 \u2018the\/a policeman\u2019 (c1)b) \/0-mo-sirikare\u00b4 \u2018some policeman\u2019 (c1)c) ri-i-tooka\u00b4 \u2018the\/a store\u2019 (c5)d) \/0-i-tooka\u00b4 \u2018some store\u2019 (c5)e) a-\u00f1-akwaaha\u00b4 \u2018the\/an armpit\u2019 (c9)f) \/0-\u00f1-akwaaha\u00b4 \u2018some armpit\u2019 (c9)g) tSaa-\u00f1-akwaaha\u00b4 \u2018the armpits\u2019 (c10)h) \/0-\u00f1-akwaaha\u00b4 \u2018some armpits\u2019 (c10)D-NUM-n P-D-NUM-n Glossi) e-Ge\u00b4-seku kjo-o-mo-sirikare\u00b4 \u2018the\/a door of the\/a policeman\u2019 (c7, c1)j) \/0-Ge\u00b4-seku ke- \/0-mo-sirikare\u00b4 \u2018some door of some policeman\u2019 (c7, c1)k) e-Ge\u00b4-seku ke-ri-i-tooka\u00b4 \u2018the\/a door of the\/a store\u2019 (c7, c5)l) \/0-Ge\u00b4-seku Gi- \/0-i-tooka\u00b4 \u2018some door of some store\u2019 (c7, c5)m) e-ki-GE\u00b4rO kja-a-\u00f1-akwaaha\u00b4 \u2018the\/a thing of the\/an armpit\u2019 (c7, c9)n) \/0-ki-GE\u00b4rO kE- \/0-\u00f1-akwaaha\u00b4 \u2018some thing of some armpit\u2019 (c7, c9)o) e-ki-GE\u00b4rO GE-tSaa-\u00f1-akwaaha\u00b4 \u2018the\/a thing of armpits\u2019 (c7, c10)p) \/0-ki-GE\u00b4rO kE- \/0-\u00f1-akwaaha\u00b4 \u2018some thing of some armpits\u2019 (c7, c10)4.5.4 Analysis summaryThe analysis of the final H tone class is essentially that no constraints particularly dictate the surface formbesides MONOH and HIGH\u03c6 . This is precisely because the roots in these cases have only one allomorph,and that allomorph has a high tone. Note that this also relies on nominal structure being minimally composed93of a root and class marker.4.5.5 Morph set relationsFinal H roots are simpler than low tone roots in that they never vary. If a root has an allomorph with a finalH tone, that set is well formed. The condition above does not apply to such roots, because that conditionholds of allomorphs with initial Hs, which final H roots do not have.4.6 Initial H tone class4.6.1 GeneralInitial H tone roots generally occur with a root allomorph that has a high tone on the initial syllable. Formsinvolving these roots also occur with high tone on the syllable preceding the root, or on the second syllableof the root. The roots involved have three relevant allomorphs: HL, LH, and L. The low allomorph is distinctfrom the low allomorph of low tone roots in that it also requires the preceding syllable have a high tone.Consider the forms shown in Table 4.41, which illustrate the basic cases involving initial H roots. Inthese forms, the root allomorph has a H on the initial syllable and there are no other high tones.Table 4.41: Underived Initial H-tone roots with overt determinerForm GlossD- NUM- na. o- mu- ka\u00b4ri \u2018the\/a woman\u2019 (c1)b. o- mu- Ga\u00b4ruka \u2018the\/a old man\u2019 (c1)c. e- Ge- s\u0131\u00b4ma \u2018the\/a well\u2019 (c7)d. o- ro- s\u0131\u00b4ri \u2018the\/a rope\u2019 (c11)Forms with an initial H and null determiners also surface with an initial H allomorph of the root.94Table 4.42: Underived Initial H-tone roots with null determinerForm GlossD- NUM- na. mu- ka\u00b4ri \u2018the\/a woman\u2019 (c1)b. mu- Ga\u00b4ruka \u2018the\/a old man\u2019 (c1)c. Ge- s\u0131\u00b4ma \u2018the\/a well\u2019 (c7)d. ro- s\u0131\u00b4ri \u2018the\/a rope\u2019 (c11)Derived roots can be of the initial H tone root class, as we saw with the low tone root class. Observe thatthese surface like their underived counterparts above.Table 4.43: Derived Initial H-tone roots with overt determinerForm GlossD- NUM- na. o- mu- sE\u00b4k-u \u2018person who is laughed at\u2019 (c1)b. e- ke- me\u00b4r-u \u2018thing which is swallowed\u2019 (c7)c. o- ko- me\u00b4r-a \u2018act of swallowing\u2019 (c15)Table 4.44: Derived Initial H-tone roots with null determinerForm GlossD- NUM- na. mu- [sE\u00b4k-u] \u2018person who is laughed at\u2019 (c1)b. ke- [me\u00b4r-u] \u2018thing which is swallowed\u2019 (c7)c. ko- [me\u00b4r-a] \u2018act of swallowing\u2019 (c15)95The evaluation of forms involving initial H roots is more complex than we have seen in the sense thatthere are more candidates. This is a function of more allomorphs for initial H roots than for low tone rootsand final H roots, which both generally have only one allomorph.Initial H root morphemes have two types of restrictions; the first is phonological selection, which weare already familiar with: the all low allomorph is phonologically restricted by the requirement that it ispreceded by a high tone, as governed by SELECTION. The second restriction is morphological; the LHallomorph is morphologically restricted to occur in a class 5 context (preceded by c5 prefix\/determiner), asgoverned by MORPHPREF. Recall that many roots can occur in class 5, as it is used to for an evaluativereading. In terms of analytical mechanisms, the low allomorph of the root has a SEL:H restriction, and theLH allomorph is indexed to c5; consider the corresponding constraints below.(69) SELECT:H and MORPHPREF:ROOT-DETERMINER definitionsSEL:H : Assign a violation mark to any allomorph that has a Sel:H restriction and is notpreceded by a syllable with H tone.MP:RT-DET: Assign a violation mark to any form that has a root marked for a class featureand that class feature is not the same as the class feature of thepreceding determiner within the \u03c6 .Note that because MP:RT-DET is ranked below SEL:L, the violations do not have any impact on theevaluations for low tone roots we saw in Section 4.4.2. Before we continue the analysis, we will pare downsome of the cases we consider for the sake of clarity. MONOH is undominated, therefore no candidate withhigh tone on more than one syllable will ever be optimal; from here on, I will not include this constraint northe forms that violate it. Consult Appendix B for full tableaux.The optimal form of the initial H root in the basic context (overt determiner, no prepositional prefixes)violates alignment; comparing the optimal form in (70c.) with the losing form in (70e.) that satisfiesalignment, observe that the latter violates *(\u03c3 V\u00b4 (like similar low tone root forms do), and additionallyviolates the selectional restriction on the low tone allomorph of the initial H root morph set (the syllablepreceding the root is not H). The losing candidate in (70b.) violates the morphological restriction on the LHallomorph of the initial H roots: it is not in a class 5 context; therefore both MP:RT-DET is violated. Thelosing candidate in (70d.) violates HIGH\u03c9 , which requires a high tone within the \u03c9; this is the counterpartto the familiar HIGH\u03c6 constraint. In contrast, the winner in (70c.) satisfies HIGH\u03c9 .96(70) Evaluation of Initial H-tone root with overt determiner[DP{e, e\u00b4}+[NumP{ Ge, Ge\u00b4}+[nP{s\u0131\u00b4ma, sima\u00b4c5, simaH }]]]HIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HMP:RT-DETHIGH\u03c9AL(H,\u03c6)a. [\u03c6 e-Ge-[\u03c9simaH ]] *! * * *b. [\u03c6 e-Ge-[\u03c9sima\u00b4c5]] *! *u c. [\u03c6 e-Ge-[\u03c9s\u0131\u00b4ma]] *d. [\u03c6 e-Ge\u00b4-[\u03c9simaH ]] *! *e. [\u03c6 e\u00b4-Ge-[\u03c9simaH ]] *! *The evaluation of initial H roots with a null determiner is similar to the evaluation with an overt deter-miner. Crucially this case involves a lower ranked MORPHPREF constraint which related root morphemeswith class marker morphemes, as opposed to the constraint related roots and determiners.(71) MORPHPREF:ROOT-CLASS MARKERMP:RT-CM: Assign a violation mark to any form that has a root marked for a class feature andthat class feature is not the same as the class feature of the preceding class prefixwithin the \u03c6 .This case is unique in illustrating that the features of these morphemes are directly related, as this is the onlycase in which the determiner has no class features, and there is no phonological reason to prefer the form in(72b.) to the winner in (72c.) aside from the mismatch in features between the root and class marker. Thelosing candidate in (72d.) fatally violates HIGH\u03c9 . The optimal candidate in (72c.) violates only alignment.97(72) Evaluation of Initial H-tone root with null determiner[DP[NumP{ Ge, Ge\u00b4}+[nP{s\u0131\u00b4ma, sima\u00b4c5, simaH }]]]HIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HMP:RT-DETHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03c6Ge-[\u03c9simaH ]] *! * * *b. [\u03c6Ge-[\u03c9sima\u00b4c5]] *! *u c. [\u03c6Ge-[\u03c9s\u0131\u00b4ma]] *d. [\u03c6Ge\u00b4-[\u03c9simaH ]] *!As was mentioned previously, this case also illustrates that MP:RT-CM is ranked under HIGH\u03c6 . This rank-ing means that the constraint cannot be used to differentiate between forms in (43), although it is sensitiveto exactly the relevant distinction.Forms with initial H roots and prepositional prefixes occur with the initial H allomorph of the root.Table 4.45: Associative Initial H-tone root with overt determinerForm GlossASSOC- D- NUM- na. e-ki-GE\u00b4rO kjo- o- mu- ka\u00b4ri \u2018the\/a door of the\/a woman\u2019 (c7, c1)b. e-Ge-s\u00b4ima kjo- o- mu- Ga\u00b4ruka \u2018the\/a door of the\/a old man\u2019 (c7, c1)Unlike forms we have seen thus far, forms with a prepositional prefix and null determiner involvinginitial H roots do not surface with the same relative H tone position. These forms surface with the all lowallomorph of the root and a H allomorph of the class prefix. This configuration is unexpected; however, aswe shall see, the selectional restriction of the low tone allomorph accounts for this surface form.98Table 4.46: Associative Initial H-tone root with null determinerForm GlossASSOC- D- NUM- na. ki-GE\u00b4rO ke- mu\u00b4- kari \u2018the\/a door of the\/a woman\u2019 (c7, c1)b. Ge-s\u00b4ima ke- mu\u00b4- Garuka \u2018the\/a door of the\/a old man\u2019 (c7, c1)Associative forms with initial H-tone roots and overt determiners illustrate a crucial interaction betweenthe selectional restrictions on the low tone preposition allomorph and the determiner. Up to this point, theserestrictions have not been relevant, though the morphemes which bear them have been discussed.The determiner has the same selectional restriction as the class 9\/10 prefix: it selects for a following lowtone.(73) Selectional restrictions determiners{V, V\u00b4}L DET{. . .}L: This morpheme (set) selects for a following syllable with low tone.SEL:L: Assign a violation mark to any allomorph that has a Sel:L restriction and is notfollowed by a syllable with L tone.The low tone allomorph of the prepositional prefixes (the associative in this case) selects for a followingH tone. This is distinct from the selection restriction of the all low allomorph of the initial H tone class ofroot morphemes.(74) Selectional restrictions for prepositions{CVH, CV\u00b4} ASSOCCVH: This morpheme selects for a following syllable with high tone.SEL:H: Assign a violation mark to any allomorph that has a Sel:H restriction and is notfollowed by a syllable with H tone.Before turning to how selectional restrictions interact in this case, let us consider the more usual inter-actions. The form in (75a.) predictably loses out to the optimal form because it lacks an H tone and violates99HIGH\u03c6 . The forms in (75e. & f.) violate VV-TONE because they include either a globally prohibited con-tour, or contextually prohibited counter. The form in (75g.) satisfies alignment and VV-TONE; however,it fatally violates SEL:H ; note that the comparable form involving a low tone root in (32) is not subject tothis restriction, because low tone roots do not have an all low allomorph which selects for preceding H.The forms in (75b.-d.) involve the interaction of selectional restrictions. The form in (75d.) uses the alllow allomorph of the root and satisfies the selectional restriction of that root morph, SEL:H , by occurringwith a H class prefix preceding the root; however, it violates SEL:L on the determiner, which the optimalform in (75c.) satisfies. The form in (75b.) loses to the optimal form in (75c.) because it violates MP:RT-DET by including the c5 morph of the root in a non-c5 context. The optimal form violates the selectionalrestriction of the low tone prefix morph, SEL:H.(75) Evaluation of associative Initial H-tone root with overt determiner[PP{kjoH, kjo\u00b4}+[DP{o, o\u00b4}L+[NumP{mu, mu\u00b4}+[nP{ka\u00b4ri, kariH , kar\u0131\u00b4c5}]]]]HIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6kjoH-oL-mu-[\u03c9kariH ]] *! * * *b. [\u03b9 [\u03c6kjoH-oL-mu-[\u03c9kar\u0131\u00b4c5]] *! * * *u c. [\u03b9 [\u03c6kjoH-oL-mu-[\u03c9ka\u00b4ri]] * *d. [\u03b9 [\u03c6kjoH-oL-mu\u00b4-[\u03c9kariH ]] *! * *e. [\u03b9 [\u03c6kjoH-o\u00b4L-mu-[\u03c9kariH ]] *! * *f. [\u03b9 [\u03c6kjo\u00b4\u2013oL-mu-[\u03c9kariH ]] *! * * *g. [\u03b9 [\u03c6kjo\u00b4\u2013o\u00b4L-mu-[\u03c9kariH ]] *! *The evaluation of initial H roots with an associative prefix and null determiner crucially relies on therestriction that low tone prepositional prefixes not be followed by a low tone. This is in contrast to theprevious cases with an overt determiner, where SEL:L crucially excluded forms with a H on the class prefixthat would satisfy SEL:H. Compare (75d.), a losing candidate in the previous tableaux, with (76d.), theoptimal form in the tableaux below. Because forms with a null determiner cannot violate SEL:L, the formin (76d.) wins over (76c.) which satisfies both SEL:H on the root, and SEL:H on the preposition.100(76) Evaluation of associative Initial H-tone root with null determiner[PP{keH, ke\u00b4}[DP[NumP{mu, mu\u00b4}+[nP{ka\u00b4ri, kariH , kar\u0131\u00b4c5}]]]]HIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6keH-mu-[\u03c9kariH ]] *! * * * *b. [\u03b9 [\u03c6keH-mu-[\u03c9kar\u0131\u00b4c5]] *! * *c. [\u03b9 [\u03c6keH-mu-[\u03c9ka\u00b4ri]] *! *u d. [\u03b9 [\u03c6keH-mu\u00b4-[\u03c9kariH ]] * *g. [\u03b9 [\u03c6ke\u00b4\u2013mu-[\u03c9kariH ]] *! *4.6.2 Classes 5, 9, 10Initial H roots in classes 5, 9 and 10 occur with a LH allomorph of the root in general. This contrasts with thecases we have just seen where forms with initial H roots occur with the initial H allomorph in general, and inthe case with a prepositional prefix and null determiner with the all low allomorph. In terms of analysis, theLH allomorph is indexed to morphological context. Beyond these basic cases, forms with a prepositionalprefix and null determiner surface with an initial H allomorph. This is comparable to the cases we have justseen where initial H roots in forms with a prepositional prefix and null determiner surface with an all lowroot allomorph; notice that in both situations H is one syllable right of the position it is realized in with anovert determiner.Class 5Initial H roots in class 5 with an overt determiner surface with high tone on the second root syllable. Con-sidering only two syllable roots, there is an ambiguity between final H and second syllable from the left H;however, trisyllabic cases illustrate that the more broad generalization is second syllable H.101Table 4.47: Initial H-tone Class 5 roots with overt determinerForm GlossD- NUM- na. ri- i- kara\u00b4 \u2018charcoal (piece)\u2019 (c5)b. a- ma- ka\u00b4ra \u2018charcoal (mass) (c6)c. ri- i- huundu\u00b4ku \u2018corpse\u2019 (c5)d. a- ma- hu\u00b4u\u00b4nduku \u2018corpses\u2019 (c6)e. ri- i- Buru\u00b4uNga \u2018egg\u2019 (c5)f. a- ma- Bu\u00b4ruuNga \u2018eggs\u2019 (c6)Initial H roots in class 5 with a null determiner also surface with a LH root allomorph.Table 4.48: Initial H-tone Class 5 roots with null determinerForm GlossD- NUM- na. i- kara\u00b4 \u2018charcoal (piece)\u2019 (c5)b. ma- ka\u00b4ra \u2018charcoal (mass) (c6)c. i- huundu\u00b4ku \u2018corpse\u2019 (c5)d. ma- hu\u00b4u\u00b4nduku \u2018corpses\u2019 (c6)e. i- Buru\u00b4uNga \u2018egg\u2019 (c5)f. ma- Bu\u00b4ruuNga \u2018eggs\u2019 (c6)The evaluation of initial H roots in the context of class 5, 9, and 10 is straightforward; in all of thesecontexts, the roots surface with the LH allomorph, which is lexically restricted to class 5. The form in (77a.)violates SEL:L. The forms in (77b.-e.) violate HIGH\u03c6 (among other constraints), which the optimal formin (77i.) satisfies.102(77) Evaluation of Initial H-tone c5 root with overt determiner[DP{ri, r\u0131\u00b4}L+[NumP{i, \u0131\u00b4}+[nP{Bu\u00b4ruuNga, BuruuNgaH , Buru\u00b4uNgac5}]]]]HIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DETSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 riL-i-[\u03c9Bu\u00b4ruuNga]] * *b. [\u03b9 [\u03c6 riL-i-[\u03c9BuruuNgaH ]] *! * * *c. [\u03b9 [\u03c6 riL-\u0131\u00b4[\u03c9BuruuNgaH ]] *! *d. [\u03b9 [\u03c6 r\u0131\u00b4L-\u0131\u00b4-[\u03c9BuruuNgaH ]] *!e. [\u03b9 [\u03c6 r\u0131\u00b4L-i-[\u03c9BuruuNgaH ]] *! *u i. [\u03b9 [\u03c6 riL-i-[\u03c9Buru\u00b4uNgac5]] *Forms in c5 with initial H roots and null determiners require a highly ranked constraint which penalizesroots without an index when a root with an index matching the prefix\/determiner is available. In other words,a constraint which would favour (78e.) over (78a.). This is similar to the constraint proposed as a solutionto the case in (43); however, that constraint assigns violations to indexed roots that do not match prefixmorphemes. The constraint at hand assigns violations to unindexed roots, as these do not have identicalclass features to their prefixes; furthermore, this constraint only needs to outrank SEL:H, as illustrated bythe case of associatives with initial H roots in c5 with an overt determiner (80). Note that it also must beranked under SEL:L, but has no relationship to MP:RT-DET.(78) *UNMARKEDROOTMORPH (* \/0RT)* \/0RT: Assign a violation to mark to forms with roots that are unspecified for a class feature.103(79) Evaluation of Initial H-tone c5 root with null determiner[DP[NumP{i, \u0131\u00b4}+[nP{Bu\u00b4ruuNga, BuruuNgaH , Buru\u00b4uNgac5}]]]]HIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 i-[\u03c9Bu\u00b4ruuNga]] *! *c. [\u03b9 [\u03c6 i-[\u03c9BuruuNgaH ]] *! * * *d. [\u03b9 [\u03c6 \u0131\u00b4-[\u03c9BuruuNgaH ]] *! * *u e. [\u03b9 [\u03c6 i-[\u03c9Buru\u00b4uNgac5]] *Initial H roots in class 5 with a prepositional prefix and overt determiner surface with the LH allomorphof the root. Compare the c6 forms, where the initial H allomorph is used. This demonstrates that the LHallomorph is generally favoured in the class 5 context.Table 4.49: Initial H-tone roots, c5, associative with overt determinerForm GlossASSOC- D- NUM- na. o-mu-ka\u00b4ri wo- ri- i- Buru\u00b4uNga \u2018the\/a woman of a\/the egg\u2019(c1, c5)b. o-mu-ka\u00b4ri wa- a- ma- Bu\u00b4ruuNga \u2018the\/a woman of some\/a eggs\u2019(c1, c6)Given the forms we have seen above involving initial H roots in class 5 contexts, the forms with apreposition and a null determiner are surprising, because they do not surface with the LH allomorph of theroot. Instead, such forms surface with the initial H allomorph of the root. This is also surprising consideringthe general cases of initial H roots with prepositional prefixes and a null determiner, where the all lowallomorph of the root is used. Considering the range of cases with initial H roots, the location of H tone isone syllable to the right in the context with a preposition and null determiner when compared to the contextwith a preposition and overt determiner.104Table 4.50: Initial H-tone roots, c5, associative with phonologically null determinerForm GlossASSOC- D- NUM- na. o-mu-ka\u00b4ri wi- i- Bu\u00b4ruuNga \u2018no woman of no egg\u2019(c1, c5)b. o-mu-ka\u00b4ri wO- ma- Bu\u00b4ruuNga \u2018no woman of no eggs\u2019(c1, c6)Because of the number of morphemes and selectional restrictions in the following tableaux, forms whichviolate the undominated constraint HIGH\u03c6 are omitted; similarly there are no candidates with vowel initialallomorphs, therefore *(\u03c3 V\u00b4 has been omitted.For the candidates in (80c.), the violation of * \/0RT illustrates the crucial ranking of * \/0RT over SEL:H;the latter constraint is violated by the winning candidate in (80e.), whereas the former constraint is violatedby the candidate in (80c.). The candidate in (80e.), which satisfies alignment, violates the selectional re-striction on the all low allomorph of the initial H root; were this an all low root, this would be the optimalcandidate, precisely because the all low allomorph of all low roots does not have a selectional restrictionrequiring a high tone on the preceding syllable. The losing candidates in (80b. & d.) violate VV-TONEbecause of their illicit contours on a long vowel: the candidate in (80b.) because rising contours are alwayspenalized, and the candidate in (80d.) because falling tone is penalized outside the penultimate syllable of a\u03c6 .105(80) Evaluation of associative initial H-tone c5 root with overt determiner[PP{woH, wo\u00b4}+[DP{ri, r\u0131\u00b4}L+[NumP{i, \u0131\u00b4}+[nP{Bu\u00b4ruuNga, BuruuNgaH , Buru\u00b4uNgac5}]]]]VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6woH-riL-i-[\u03c9Bu\u00b4ruuNga]] *! * * *b. [\u03b9 [\u03c6woH-riL-\u0131\u00b4-[\u03c9BuruuNgaH ]] *! * * *c. [\u03b9 [\u03c6woH-r\u0131\u00b4L-\u0131\u00b4-[\u03c9BuruuNgaH ]] *! * *d. [\u03b9 [\u03c6woH-r\u0131\u00b4L-i-[\u03c9BuruuNgaH ]] *! * * *u e. [\u03b9 [\u03c6woH-riL-i-[\u03c9Buru\u00b4uNgac5]] * *f. [\u03b9 [\u03c6wo\u00b4-riL-i-[\u03c9BuruuNgaH ]] *! *The evaluation of initial H roots in class 5 with a prepositional prefix and a null determiner are of interestbecause these are cases where the surface form does not include the morphologically specified LH allomorphof the root. Unfortunately, our analysis will incorrectly predict that the optimal form surfaces with the c5indexed form, precisely because of the ranking for * \/0RT established by (80). The observed surface form isindicated with a \u201c\/\u201d.If we ignore violations of * \/0RT the candidate in (81e.) that uses the LH allomorph, loses to the observedform (81a.) because the former violates SEL:H. However, because the candidates in (80a. & c.) violate* \/0RT, the predicted output is (81e.)106(81) Evaluation of associative initial H-tone c5 root with null determiner[PP{wiH, w\u0131\u00b4}+[DP[NumP{i, \u0131\u00b4}+[nP{Bu\u00b4ruuNga, BuruuNgaH , Buru\u00b4uNgac5}]]]]VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)\/ a. [\u03b9 [\u03c6wiH-i-[\u03c9Bu\u00b4ruuNga]] *! *b. [\u03b9 [\u03c6wiH-\u0131\u00b4-[\u03c9BuruuNgaH ]] *! * * *c. [\u03b9 [\u03c6w\u0131\u00b4-\u0131\u00b4-[\u03c9BuruuNgaH ]] *! *d. [\u03b9 [\u03c6w\u0131\u00b4-i-[\u03c9BuruuNgaH ]] *! * *u e. [\u03b9 [\u03c6wiH-i-[\u03c9Buru\u00b4uNgac5]] *! *Class 9\/10Initial H roots in class 9 and in class 10 with an overt determiner occur with the LH allomorph, as they do inclass 5. The observation that the LH allomorph occurs in all three of these contexts is crucial evidence thatthe form is morphologically conditioned, not phonologically conditioned: the class 5 pre-root phonologicalmaterial is a CVV, the class 9 pre-root material is a V, and the class 10 pre-root material is a CVV. The outlierof class 9 essentially supports the claim that phonological material is not conditioning the LH allomorph.Table 4.51 illustrates class 9\/10 plurals in contrast with class 12 evaluative forms. Nouns with initialH roots occur with the LH allomorph in classes 9 and 10, but those same roots occur with the initial Hallomorph in class 12 (which follows the general pattern shown in the preceding section).107Table 4.51: Initial H-tone Class 9\/10 roots with overt determinerForm GlossD- NUM- na. a- m- bara\u00b4he \u2018a\/the Thomson\u2019s Gazelle\u2019 (c9)b. tSaa- m- bara\u00b4he \u2018some\/the Thomson\u2019s Gazelles\u2019 (c10)c. a- ka- ba\u00b4rahe \u2018a\/the Thomson\u2019s Gazelle (eval.)\u2019 (c12)d. a- N- goko\u00b4 \u2018a\/the chicken\u2019 (c9)e. tSaa- N- goko\u00b4 \u2018some\/the chickens\u2019 (c10)f. a- ka go\u00b4ko \u2018a\/the chicken (eval.)\u2019 (c12)Initial H roots in classes 9 and 10 with null determiners also surface with the LH allomorph of the root.This is consistent with the general pattern that forms without prepositional prefixes demonstrate with overand null determiner (i.e. same H tone position, same root allomorph in the case of initial H roots).Table 4.52: Initial H-tone Class 9\/10 roots with null determinerForm GlossD- NUM- na. m- bara\u00b4he \u2018no Thomson\u2019s Gazelle\u2019 (c9)b. m- bara\u00b4he \u2018no Thomson\u2019s Gazelles\u2019 (c10)c. ka- ba\u00b4rahe \u2018no Thomson\u2019s Gazelle (eval.)\u2019 (c12)d. N- goko\u00b4 \u2018no chicken\u2019 (c9)e. N- goko\u00b4 \u2018no chickens\u2019 (c10)f. ka go\u00b4ko \u2018no chicken (eval.)\u2019 (c12)The winning candidate in (82b.) satisfies the selectional restrictions on the determiner and c9\/10 prefix;however, it violates alignment. The next best candidate in (82a.) fails to satisfy the selectional restrictionof both the determiner and class 9\/10 prefix, because the syllable following those morphs is not low. The108candidate in (82c.) violates *(\u03c3 V\u00b4; note that this type of form would win if it included a low tone root;however, because initial H roots have allomorphs that can provide an H tone in a position which does notviolate *(\u03c3 V\u00b4, the candidate at hand loses to them (82a.&b.).(82) Evaluation of Initial H-tone c9 root with overt determiner[DP{a, a\u00b4}L+[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 aL-mL-[\u03c9ba\u00b4rahe]] *! * *u b. [\u03b9 [\u03c6 aL-mL-[\u03c9bara\u00b4hec9\/10]] *c. [\u03b9 [\u03c6 a\u00b4L-mL-[\u03c9baraheH ]] *! * *The evaluation of initial H roots in class 10 with an overt determiner is similar to the evaluation we haveconsidered from comparable forms in class 9. The candidate in (83d.) loses to the winner (83b.) becauseit violates * \/0RT and H IGH\u03c9 . A similar candidate in (83c.) violates VV-TONE because it has a fallingcontour in a non-penultimate syllable. The candidate in (83a.) violates SEL:L, which is not violated by thewinner in (83b.). Note that the LH allomorph of the initial H root class is indexed to class 9 and 10, unlikethe HL allomorph of the low root, which is only indexed to class 9.(83) Evaluation of Initial H-tone c10 root with overt determiner[DP{tSaa, tSa\u00b4a\u00b4, tSa\u00b4a}+[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 tSaa-mL-[\u03c9ba\u00b4rahe]] *! * *u b. [\u03b9 [\u03c6 tSaa-mL-[\u03c9bara\u00b4hec9\/10]] *c. [\u03b9 [\u03c6 tSa\u00b4a-mL-[\u03c9baraheH ]] *! * *d. [\u03b9 [\u03c6 tSa\u00b4a\u00b4-mL-[\u03c9baraheH ]] *! *The evaluation of initial H tone roots in class 9 with a null determiner is very straightforward as theonly source of high tone is the root, which only has two high tone allomorphs: HL and LH. Of these two109allomorphs, the LH allomorph is also marked for c9, as we have already observed. The losing candidate in(84a.) violates SEL:L. The winning candidate in (84b.) only violates alignment.(84) Evaluation of Initial H-tone c9 root with null determiner[DP[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6mL-[\u03c9ba\u00b4rahe]] *! *u b. [\u03b9 [\u03c6mL-[\u03c9bara\u00b4hec9\/10]] *The evaluation of initial H roots in class 10 with a null determiner is identical to their class 9 counter-parts. This is because the only phonological difference between c9 and c10 forms is the determiner.(85) Evaluation of Initial H-tone c10 root with null determiner[DP[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6mL-[\u03c9ba\u00b4rahe]] *! *u b. [\u03b9 [\u03c6mL-[\u03c9bara\u00b4hec9\/10]] *Words with initial H tone roots in class 9 and 10 with overt determiners and prepositional prefixes occurwith the LH allomorph of the root.110Table 4.53: Initial H-tone roots, c9\/10, associative with overt determinerForm GlossASSOC- D- NUM- na. e-r\u00b4i-ino rja a- m- bara\u00b4he \u2018the\/a tooth of a\/the Thomson\u2019s Gazelle\u2019(c5, c9)b. e-r\u00b4i-ino rE tSaa m- bara\u00b4he \u2018the\/a tooth of some\/a Thomson\u2019s Gazelles\u2019(c5, c10)Like the class 5 counterparts, initial H roots in class 9 and 10 with a null determiner and prepositionalprefix differ from their counterparts with an overt determiner in that they occur with the HL allomorph ofthe root.Table 4.54: Initial H-tone roots, c9\/10, associative with null determinerForm GlossASSOC- D- NUM- na. e-r\u00b4i-ino rE m- ba\u00b4rahe \u2018no tooth of no Thomson\u2019s Gazelle\u2019(c5, c9)b. e-r\u00b4i-ino rE m- ba\u00b4rahe \u2018no tooth of no Thomson\u2019s Gazelles\u2019(c5, c10)The evaluation of initial H roots in class 9 with an overt determiner and prepositional prefix is essentiallythe same as for comparable class 5 cases in (80). The winning candidate in (86e.) violates SEL:H, butsatisfies all higher ranked constraints. The forms in (86c.) and (86d.) violate * \/0RT because they includeunmarked allomorphs of the root. Additionally, the candidate in (86d.) violates SEL:L. The candidates in(86a. & b.) fatally violate VV-TONE; the former because it has a rising contour which is penalized in allpositions, and the latter because it has a falling contour on a non-penultimate syllable.111(86) Evaluation of associative Initial H-tone c9 root with overt determiner[PP{rjaH, rja\u00b4}+[DP{a, a\u00b4}+[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 rja\u00b4-a-mL-[\u03c9baraheH ]] *! * *b. [\u03b9 [\u03c6 rjaH-a\u00b4-mL-[\u03c9baraheH ]] *! * *c. [\u03b9 [\u03c6 rja\u00b4-a\u00b4-mL-[\u03c9baraheH ]] *! * *d. [\u03b9 [\u03c6 rjaH-a-mL-[\u03c9ba\u00b4rahe]] *! * *u e. [\u03b9 [\u03c6 rjaH-a-mL-[\u03c9bara\u00b4hec9\/10]] * *As in the class 9 evaluation above, the optimal candidate in (87e.) violates SEL:H and alignment.The losing candidate in (87a.) satisfies SEL:H by virtue of not having a low tone prepositional prefix;however, it fails to satisfy the selectional restriction of the low tone allomorph of the root. The candidate in(87c.) satisfies both SEL:H and SEL:H , but fatally violates * \/0RT; the comparison between (87c.) and thewinning candidate in (87e.) illustrates that * \/0RT outranks SEL:H because if their ranking were reversedthe former would be the winning candidate in this evaluation. The candidate in (87d.) violates SEL:L, aswell as * \/0RT and SEL:H, in other words it is harmonically bounded by the winner since they both violateSEL:H and the latter also violates a set of other constraints which the winner does not. The candidate in(87b.) violates VV-TONE because it has a falling contour in non-penultimate position.112(87) Evaluation of associative Initial H-tone c10 root with overt determiner[PP{rEH, rE\u00b4}+[DP{tSaa, tSa\u00b4a\u00b4, tSa\u00b4a}+[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 rE\u00b4-tSaa-mL-[\u03c9baraheH ]] *! * *b. [\u03b9 [\u03c6 rEH-tSa\u00b4a-mL-[\u03c9baraheH ]] *! * * *c. [\u03b9 [\u03c6 rEH-tSa\u00b4a\u00b4-mL-[\u03c9baraheH ]] *! * *d. [\u03b9 [\u03c6 rEH-tSaa-mL-[\u03c9ba\u00b4rahe]] *! * * *u e. [\u03b9 [\u03c6 rEH-tSaa-mL-[\u03c9bara\u00b4hec9\/10]] * *The evaluations of initial H roots in class 9 and 10 with null determiners and prepositional prefixes failto produce the observed optimal candidate; as for the cases without a prepositional prefix, the c9 and c10evaluations are identical. The observed but non-optimal candidate is marked with a \u201c\/\u201d. We will return tothe mechanical problem with the analysis later in this chapter.Initial H roots in class 9\/10 with a null determiner and prepositional prefix actually occur with the initialH allomorph of the root. Our analysis predicts that such forms will surface with the all low allomorph of theroot and a H allomorph of the prepositional prefix. The optimal candidate in (88a.) only violates HIGH\u03c9and * \/0RT. The observed candidate in (88b.) violates SEL:L fatally, as well as alignment and * \/0RT. Notethat the candidate in (88c.) also fares better than the observed form in (88b.); the former violates SEL:H,which is crucially ranked lower than SEL:L, as we saw for previous cases.(88) Evaluation of associative Initial H-tone c9 root with null determiner[PP{rEH, rE\u00b4}+[DP+[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 rE\u00b4-mL-[\u03c9baraheH ]] *! *\/ b. [\u03b9 [\u03c6 rEH-mL-[\u03c9ba\u00b4rahe]] *! * *u c. [\u03b9 [\u03c6 rEH-mL-[\u03c9bara\u00b4hec9\/10]] *! *113The class 10 evaluation is provided below for completeness sake, but as mentioned, it is identical to thec9 case (because the form of the determiner is the only phonological distinction between c9 and c10).(89) Evaluation of associative Initial H-tone c10 root with null determiner[PP{rEH, rE\u00b4}+[DP+[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 rE\u00b4-mL-[\u03c9baraheH ]] *! *\/ b. [\u03b9 [\u03c6 rEH-mL-[\u03c9ba\u00b4rahe]] *! * *u c. [\u03b9 [\u03c6 rEH-mL-[\u03c9bara\u00b4hec9\/10]] *! *4.6.3 Data SummaryForms with initial H roots generally surface with an initial H root allomorph. In class 5, 9, and 10, the LHroot allomorph occurs. In contexts with a prepositional prefix, the initial H allomorph is used, except forclasses 5, 9, and 10, where the LH allomorph is used. In contexts with a prepositional prefix and a nulldeterminer, the H tone occurs one syllable to the right of the position it occurs in comparable contexts withan overt determiner.114Table 4.55: Summary: Initial H tone classD-NUM-n Glossa) o-mu-ka\u00b4ri \u2018the\/a woman\u2019 (c1)b) \/0-mu-ka\u00b4ri \u2018some woman\u2019 (c1)c) ri-i-Buru\u00b4uNga \u2018the\/an egg\u2019 (c5)d) \/0-i-Buru\u00b4uNga \u2018some egg\u2019 (c5)e) a-m-bar\u00b4ahe \u2018the\/a Thomson\u2019s Gazelle\u2019 (c9)f) \/0-m-bara\u00b4he \u2018some Thomson\u2019s Gazelle\u2019 (c9)g) tSaa-m-bara\u00b4he \u2018the Thomson\u2019s Gazelles\u2019 (c10)h) \/0-m-bara\u00b4he \u2018some Thomson\u2019s Gazelle\u2019 (c10)D-NUM-n P-D-NUM-n Glossi) e-Ge\u00b4-seku kjo-o-mu-ka\u00b4ri \u2018the\/a door of the\/a woman\u2019 (c7, c1)j) \/0-Ge\u00b4-seku ke- \/0-mu\u00b4-kari \u2018some door of some woman\u2019 (c7, c1)k) o-mu-ka\u00b4ri wo-ri-i-Buru\u00b4uNga \u2018the\/a woman of the\/an egg\u2019 (c7, c5)l) \/0-mu-ka\u00b4ri wi- \/0-i-Bu\u00b4ruuNga \u2018some woman of some egg\u2019 (c7, c5)m) e-r\u00b4i-ino rja-a-m-bara\u00b4he \u2018the\/a tooth of the\/an Thomson\u2019s Gazelle\u2019 (c7, c9)n) \/0-r\u00b4i-ino rE- \/0-m-ba\u00b4rahe \u2018some tooth of some Thomson\u2019s Gazelle\u2019 (c7, c9)o) e-r\u00b4i-ino rE-tSaa-m-bara\u00b4he \u2018the\/a tooth of Thomson\u2019s Gazelle\u2019 (c7, c10)p) \/0-r\u00b4i-ino rE- \/0-m-ba\u00b4rahe \u2018some tooth of some Thomson\u2019s Gazelle\u2019 (c7, c10)4.6.4 Analysis summaryThe analysis of initial H roots involves a number of morpheme specific phonotactics which are crucial.Roots have two allomorphs with special conditions. Class 9\/10 class markers, determiners, and prepositionalprefixes all have tonal phonotactics. MORPH PREFERENCE and * \/0RT are crucial for classes 5, 9, and 10;recall that MORPHPREF is also a potential solution to the case of class 10 forms with a null determiner,115except for the ranking required by the initial H forms we have considered.The analysis fails to predict the observed surface forms for initial H roots in class 9\/10 with a nulldeterminer and prepositional prefix. The source of the failure is both that the observed surface forms violateSEL:L and do not use the class 9\/10 indexed root allomorph.One relevant observation is that class 9 initial H roots do not regularly surface with an all L allomorph,the way other initial H roots do. Nevertheless, class 9 roots can be used in an evaluative class, which canbe used in an associative with a null determiner. In such cases, we would predict an all low allomorph. Ifthis is not the case then the lack of such an allomorph would change our analysis and we would predict thecorrect surface form.4.6.5 Morph set relationsInitial H roots have three allomorphs: HL. . . , LH(L). . . , and LL. . . . If a root morph set has a high toneallomorph, it has a corresponding low tone allomorph. This is true of all disyllabic and larger roots asidefrom the final H roots. The relation states that allomorphs with a non-final H tone have a corresponding Ltone allomorph.116(90) Morph set relation: Tone (Roots)MSRRtTone: Pairs of otherwise identical root allomorphs differ only with respect to a non-final Htone in one being L in the other.{Mi,M j} Mi: [. . . H. . . ]M j: [. . . L. . . ]Notice that this condition is not bidirectional; a H allomorph implies the existence of an L allomorph;however, a L allomorph does not imply the existence of a H allomorph. This is why low tone roots do notgenerally have any H tone allomorphs, but initial H roots have HL allomorphs and LL allomorphs. Theclass 9 low tone roots differ in that they are governed by bidirectional variants of this condition. Prefixes aregoverned by a version of this condition that globally applies to H tones (i.e. can relate final H tones to L),because every prefix we have considered is monosyllabic, the initial and final TBUs are identical.(91) Morph set condition: Tone (Roots)With respect to MSRRtTone, a root morph set is ill-formed if Mi is present and there is no correspond-ing M j.MSCRtTone: *{Mi,\u00acM j}The relation between HL. . . and LH(L). . . is unique to the initial H roots. This essentially means it does notrelate the low tone c9 root HL allomorphs with any other allomorph, but it does relate the initial H tone rootHL allomorph to an LH allomorph.(92) Morph set relation: Tone (Initial H roots)MSRInitHRtTone: Pairs of otherwise identical initial H root allomorphs differ only with respectto a [HL. . . sequence in one being [LH. . . in the other.{Mi,M j} Mi: [HL. . . ]M j: [LH. . . ]The condition governing this relation is bidirectional. An initial H root morph set is ill formed if it has aLH allomorph but not HL, and vice versa. If the relationship were applied to all roots, or to morphemes ingeneral, it would need to be unidirectional with {HL} being permissible, but {LH} being ill-formed; this isbecause cases like the low tone c9 roots exist, which have HL allomorphs, but no LH allomorphs.117(93) Morph set condition: Tone (Initial H roots)With respect to MSRInitHRtTone, an initial H root morph set is ill-formed if M j is present and there isno corresponding Mi, or if Mi is present and there is no corresponding M jMSCInitHRtTone: *{M j,\u00acMi}*{Mi,\u00acM j}4.7 Summary and ConclusionIn this chapter we have seen that nominal forms in Nata have several basic properties; additional morpho-logical complexity exists for classes 5, 9, and 10. There remain two unsolved issues, both involving theforms with a null determiner.Nominal forms in Nata have exactly one high tone. That high tone can be a contour if it is falling, andon the penultimate syllable; otherwise it must be a level high tone. The high tone cannot be realized on anonsetless vowel, in general. Regarding the placement of the high tone, we observed that this is largely afactor dependant on the type of root involved in the form. Final H roots do not vary and always occur withH on the final syllable of the form. Forms with low tone roots generally have H tone aligned as far to the leftas possible, given the restrictions on tone placement; a special case exists for low stems in c9, which havea morphologically marked allomorph with a H on the initial syllable. Forms with initial H roots generallyoccur with the initial H root allomorph; however, special morphological conditions for c5, 9, and 10 selectfor a LH allomorph of such roots. Finally, there is an interaction between initial H roots and conditions onprepositional prefixes which require those morphemes be followed by a L tone.The remaining issues are accounting for low tone roots in c10 with a null determiner, and initial H rootsin c9\/10 with a prepositional prefix and a null determiner. The first case is puzzling because the observedsurface form has no H tone. The problem for the analysis is that the requirement to have an H tone is veryhighly ranked, and otherwise unviolated by surface forms. The case involving initial H roots is complicatedbecause the observed forms violate a morpheme selection for c9\/10 which is otherwise active with bothovert and null determiners. In other words, the selectional restriction is crucial for some cases, but appearsto be ignored for the problematic cases. In order to better understand the analysis of these cases, we wouldneed more examples of cases where the SEL:L phonotactic could be violated, such as monosyllabic rootswith only H allomorphs.118Chapter 5Vowel HarmonyThe analysis of tone proposed in the previous chapter is predicated on the assumption of phonologicaldomains with particular structure. In this chapter we shall see that these domains are also functional inunderstanding vowel harmony.I will begin in Section 5.1 by considering the basic restriction on non-identical tongue root positionsfor vowels within the \u03c9 , following work by Gambarage and Pulleyblank, 2017. In Section 5.2 I examinehow vowels in class prefixes and other material outside of the \u03c9 alternate; this work expands on previousresearch and illustrates a generalized behavior of material within the \u03c6 , but outside \u03c9 . In Section 5.3, Iconsider a particular nominal suffix which conditions an alternation within the \u03c6 .Finally, I summarize the patterns of harmony across prosodic domains and generalizes over the types ofallomorphs relevant to vowel harmony in Section 5.4.Following many previous studies on tongue root harmony (Archangeli and Pulleyblank, 1994; Bakovic\u00b4,2000; Casali, 2003 among others) I adopt a binary phonological feature for tongue root: [\u00b1ATR]. Vowelswith advanced tongue root are [+ATR]; vowels with retracted tongue root are [-ATR].Ladefoged\u2019s (1964) X-ray analysis of Igbo vowels argues that tongue root position is the primary artic-ulatory distinction between two sets of vowels; vowels produced with an advanced tongue root have largerpharyngeal cavities than those with retracted tongue root. Later research demonstrated that the size of thepharyngeal cavity can be manipulated through other means in addition to tongue root manipulation (Hess,1992; Lindau, 1974, 1978; Tiede, 1996 on Akan). Further work has claimed that the advanced tongueposition can be articulatorily distinguished from a neutral tongue position (Fusheini, 2014 on Dagbani).The primary acoustic distinction between advanced and retracted vowels is that the latter have a lower119F1 in comparison to their advanced counterparts (Halle & Stevens, 1969). The retracted high vowels [I, U]share this raised F1, but also can have a slightly lowered F2 (compared to the advanced [i, u]) (Jacobson,1980). Further work is needed to place the tongue root system of Nata into this phonetic landscape.5.1 \u03c9 vowel harmonyTongue root harmony is a pattern in which vowels within a particular domain have the same value for tongueroot feature, [\u00b1ATR]. In Nata harmony operates slightly differently within the \u03c9 and outside \u03c9 but within\u03c6 . In this section I will present forms and analysis for patterns within \u03c9 .Within \u03c9 , mid vowels always agree for [\u00b1ATR] (Gambarage, 2013; Gambarage & Pulleyblank, 2017).This is illustrated by the examples in Table 5.1 a. & b., for retracted and advanced mid vowels, respectively.High and low vowels are unrestricted and may occur freely with retracted or advanced mid vowels, as wellas with other high and low vowels (c.-h.).Table 5.1: Monomorphemic \u03c9Form Glossa. e-ki-[\u03c9GE\u00b4rO \u2018thing\u2019 (c7)b. e-Be\u00b4-[\u03c9 tore \u2018cucumbers\u2019 (c8)c. a-ma\u00b4-[\u03c9saahE \u2018blood\u2019 (c6)d. o-Bu\u00b4-[\u03c9saro \u2018beads\u2019 (c14)e. o-Bo-[\u03c9 r\u00b4iBo \u2018wax\u2019 (c14)f. o-mo\u00b4-[\u03c9sukO \u2018pocket\u2019 (c3)g. o-mu-[\u03c9ka\u00b4ri \u2018woman\u2019 (c1)h. e-ke-[\u03c9 r\u00b4isa \u2018hammer\u2019 (c7)For forms with the nominalizer [-O] in \u03c9 , stems surface with retracted mid vowels, consistent with theobservations regarding the data in Table 5.1.120Table 5.2: \u03c9 including [-O] nominalizerForm GlossO-mO\u00b4-[\u03c9 tOONg-O \u2018string\u2019 (c3)E-kE\u00b4-[\u03c9mEr-O \u2018throat\u2019 (c7)O-rO-[\u03c9hO\u00b4ONg-O \u2018sieve\u2019 (c11)Compare these same roots with a retracted low final vowel, where roots occur with advanced vowels.Table 5.3: \u03c9 including [-a] nominalizerForm GlossGo-[\u03c9 to\u00b4oNg-a \u2018to pierce\u2019ko-[\u03c9me\u00b4r-a \u2018to swallow\u2019Go-[\u03c9ho\u00b4oNg-a \u2018to sieve\u2019This illustrates that within the \u03c9 retracted mid vowels are prohibited from occurring with advanced midvowels, and not a more general prohibition on advanced mid vowels followed by retracted vowels. We willreturn to forms with alternating roots in Section 5.3Analysis: \u03c9 vowel harmonyWe can account for the distribution of vowels in prosodic words via a phonotactic prohibiting mid vowelsfrom disagreeing in tongue root within that domain. Gambarage and Pulleyblank, 2017 formalize this viatwo co-occurrence restrictions relative to the stem domain. The analysis at hand is a modification of thisproposal that is relativized to the prosodic word (\u03c9).11Like Gambarage and Pulleyblank, 2017, I do not argue for any particular feature system with respect to vowels, specificallytheir height or tongue root configuration. Unlike Gambarage and Pulleyblank, 2017, I will use a binary tongue root and heightfeatures; although specific phrasing would need to be modified, this could be replaced with monovalent tongue root and heightfeatures.121(94) \u03c9-word Harmony Constraints adapted from Gambarage and Pulleyblank, 2017, p. 8*[ . . . +ATR\/-HI,-LO C0 -ATR\/-HI,-LO . . . ]*[A R]\u03c9 : For every advanced mid vowel within the \u03c9-word, assign a violation mark if that midvowel is followed by a retracted mid vowel.*[ . . . -ATR\/-HI,-LO C0 +ATR\/-HI,-LO . . . ]\u03c9*[R A]\u03c9 : For every retracted mid vowel within the \u03c9-word, assign a violation mark if that midvowel is followed by an advanced mid vowel.In (94) are two distinct constraints handling each of the possible orderings: i) advanced mid followed byretracted mid, and ii) retracted mid followed by advanced mid. In the remainder of this work I will representthese two constraints with a single constraint. To my knowledge there are no cases where the ranking of thetwo component constraints is relevant. The constraint is referred to as \u03c9 -HARMONY.(95) \u03c9-Harmony constraint\u03c9 -HARMONY: For every mid vowel within the \u03c9-word, assign a violation mark if the vowel inthe following syllable is a mid vowel that differs in the value for the feature[\u00b1ATR].Table 5.4 illustrates various types of \u03c9-words, both attested (a.-f.) and unattested (g. & h.), with theirevaluation for the harmony constraint given above.Table 5.4: Evaluation of \u03c9 -HARMONYForm \u03c9 -HARMONYa. a-ma-[\u03c9bO\u00b4kO Db. e-Gi-[\u03c9sEErO\u00b4 Dc. O-mO-[\u03c9 tOONg-O Dd. Go-[\u03c9 to\u00b4oNg-a De. E-kE\u00b4-[\u03c9mEr-O Df. o-mu[\u03c9ka\u00b4ri Dg. *. . . [\u03c9 tOONgo *h. *. . . [\u03c9merO *122Cases with inherently nominal stems (such as in Table 5.4 a. & b.) as well as those cases with anominalizer suffix (c.-e.) satisfy \u03c9 -HARMONY. Forms without mid vowels do not violate this constraint(f.) Furthermore, no attested \u03c9-words fail to satisfy this phonotactic; this is indicated by the negative datain (g. & h.)5.2 \u03c6 vowel harmonyWe will now consider material outside the \u03c9 , beginning with the class marker prefix and determiner. Gen-erally, prefix vowels are mid advanced when they occur before a stem beginning with an advanced vowel asin Table 5.5 a.; otherwise, the prefixes surface with a high vowel (b.).Table 5.5: Class prefix vowel qualityForm Glossa. [\u03c6o-mo\u00b4-[\u03c9sikera \u2018someone who enters\u2019 (c1)[\u03c6o-mo-[\u03c9su\u00b4kO \u2018pocket\u2019 (c3)[\u03c6 e-me-[\u03c9ke\u00b4ra \u2018tails\u2019 (c4)[\u03c6 e-Be\u00b4-[\u03c9 tore \u2018cucumbers\u2019 (c8)b. [\u03c6o-mu-[\u03c9ka\u00b4ri \u2018woman\u2019 (c1)[\u03c6o-ru\u00b4-[\u03c9saro \u2018bead\u2019 (c11)[\u03c6o-mu-[\u03c9 tE\u00b4rEbi \u2018wooden ladle\u2019 (c3)[\u03c6o-Bu-[\u03c9sO\u00b4Oh-u \u2018greediness\u2019 (c14)I follow Gambarage and Pulleyblank, 2017 in referring to stems with advanced initial vowels as ad-vanced stems (such as Table 5.5 a) and stems with retracted initial vowels as retracted stems (b).The examples in Table 5.5 illustrate the alternation in prefix height based on the initial vowel of the stem.When preceding an advanced stem, class prefixes surface with a mid advanced vowel; when preceding aretracted stem, class prefixes surface with a high vowel.In some forms, the determiner is a mid advanced vowel; however if the determiner is followed by aretracted vowel, it will surface with a high vowel instead of a mid vowel as in Table 5.6 b.123Table 5.6: Determiner vowel qualityForm Glossa. [\u03c6o-mo\u00b4-[\u03c9 rem-i \u2018farmer\u2019 (c1)[\u03c6o-Bu-[\u03c9sO\u00b4oh-u \u2018greediness\u2019 (c14)[\u03c6o-[\u03c9mw-iika\u00b4 \u2018pressure\/gas\u2019 (c3)[\u03c6o-[\u03c9mw-\u00b4iit-i \u2018killer\u2019 (c1)b. [\u03c6u-[\u03c9mw-a\u00b4ana \u2018child\u2019 (c1)[\u03c6u-[\u03c9mw-EEr\u00b4i \u2018moon\u2019 (c3)[\u03c6 i-[\u03c9kj-O\u00b4OndE \u2018honey badger\u2019 (c7)This indicates that the raising observed for class marker prefix vowels in Table 5.5 b. is a general strategyto avoid sequences of mid vowels that disagree in\u00b1[ATR] across the edge of \u03c9 , as opposed to an alternationspecific to that morpheme.5.2.1 Analysis: \u03c6 vowel harmonyExtending the \u03c9 -HARMONY constraint to \u03c6 is not sufficient to account for the alternation of class markerand determiner prefixes. Tongue root harmony in these domains is distinct in several ways.Inside the \u03c9 , only mid vowels are prohibited from disagreeing in [\u00b1ATR]; however, mixed values of[\u00b1ATR] are possible due to the low and [-ATR] vowel [a]. This can be observed in forms like [\u03c6ko-[\u03c9me\u00b4r-a]\u03c9 ]\u03c6 \u2018to swallow\u2019 and [\u03c6Go-[\u03c9 to\u00b4oNg-a]\u03c9 ]\u03c6 \u2018to pierce\u2019In contrast, within the \u03c6 -phrase, both mid and low [-ATR] vowels trigger an alternation in [+ATR]vowels to their left. The alternation of the class prefix illustrates this: [\u03c6o-mo\u00b4-[\u03c9 rem-i]\u03c9 ]\u03c6 \u2018farmer (c1)\u2019,[\u03c6o-mu-[\u03c9ka\u00b4ri]\u03c9 ]\u03c6 \u2018woman (c1)\u2019, [\u03c6o-mu-[\u03c9 tE\u00b4rEBi]\u03c9 ]\u03c6 \u2018ladle (c3)\u2019; see (5.5) for more.We will begin by focusing on cases where a nominal prefix has a high (advanced) vowel. In order toaccount for these, we require a phonotactic prohibiting advanced mid vowels followed by retracted vowels(i.e. a harmony phonotactic). The present analysis is based on the proposal by Gambarage and Pulleyblank,2017 but recast in terms of the prosodic system developed in this work.124(96) \u03c6 -Harmony constraint\u03c6 -HARMONY: For a mid vowel within the \u03c6 -phrase but outside the \u03c9-word, assign aviolation mark if the vowel in the following syllable differs in value for thefeature [\u00b1ATR].The restrictor clause \u201cbut outside the \u03c9-word\u201d is required to account for the distinct properties of \u03c9 and \u03c6 .As we shall see in Section 5.3, if the \u03c6 constraint applied to \u03c9 we would incorrectly predict roots wouldoccur with retracted mid vowels before both [-O] and [-a] nominal suffix vowels. This constraint varies inscope from the variant presented in Gambarage and Pulleyblank, 2017 because only prefixes immediatelyleft of the stem were considered, and as we shall see from the data presented below, all prefixes within the\u03c6 show harmonic behaviour.Table 5.7 illustrates several forms and how they are evaluated by \u03c6 -HARMONY. While \u03c6 -HARMONYis necessary to account for the observed alternations, it is not sufficient.Table 5.7: Evaluation of \u03c6 -HARMONYForm \u03c6 -HARMONYa. * [\u03c6 O-mO-[\u03c9ka\u00b4ri Db. [\u03c6o-mu-[\u03c9ka\u00b4ri Dc. * [\u03c6o-mo-[\u03c9ka\u00b4ri * (o. . . a)d. * [\u03c6 O-mO\u00b4-[\u03c9 remi * (O. . . e)e. * [\u03c6o-mu\u00b4-[\u03c9 remi Df. [\u03c6o-mo\u00b4-[\u03c9 remi DFor the form in Table 5.7 a., the retracted mid vowels within the \u03c6 -phrase but outside the \u03c9-word arefollowed by retracted vowels (one mid, one low). For the form in (b.), the advanced mid vowel within the\u03c6 -phrase but outside the \u03c9-word is followed an advanced high vowel, therefore not violating harmony; thecase is similar for the form in (e). The form in (f.), like the one in (e.) contains only advanced vowels.To complete our analysis, we consider the lexical entries of the class prefixes, specifically those withmid vowel allomorphs. For a class marker with mid vowel allomorphs both the [+ATR] and [-ATR] mid125variants will be present, as well as a high vowel allomorph.(97) Lexical entry for class prefixes{mo, mu, mO} C1{Ba} C2{Ge, Gi, GE} C7...Gambarage and Pulleyblank, 2017, following work such as Mascaro\u00b4, 2007, propose ranking allomorphssuch that the advanced mid vowel allomorphs are less marked than the retracted allomorphs. They observethat the vowel quality of prefix allomorphs varies in two dimensions: advanced vs. retracted and mid vs.high. Advanced is preferred to retracted, and mid is preferred to high. Because of the wider scope of theharmony phonotactic employed here, ranking \u03c6 -HARMONY over phonotactics will result in the observedoptimal forms.In cases with only high and mid vowel allomorphs, the harmony constraint rules out mismatches, asin the candidate in (98c.); a phonotactic preferring non-high vowels prevents forms with the high vowelallomorph from surfacing. The winning candidate in (98a.) uses the mid advanced prefix allomorphs(98) Mid vowel forms satisfy Harmony[{o, O, u. . .}+{ mo, mO, mu. . .}+{rem . . . }+{i}] \u03c6 -HARM [*-ATR] *[+HI]u a. [\u03c6o-mo\u00b4-[\u03c9 rem-ib. [\u03c6o-mu\u00b4-[\u03c9 rem-i *!c. [\u03c6 O-mO\u00b4-[\u03c9 rem-i *!(O\u00b4. . . e)In cases with retracted root initial vowels, the phonotactic preferring advanced vowels rules out retractedharmonic candidates, like the losing candidate in (99c.) Violations of *[+HI] and *[-ATR] are shown onlyfor prefix and determiner allomorphs.126(99) Evaluation of simple vowel harmony[{o, O, u. . .}+{ mo, mO, mu. . .}+{ka\u00b4ri . . .}] \u03c6 -HARM *[-ATR] *[+HI]u a. [\u03c6o-mu-[\u03c9ka\u00b4ri *b. [\u03c6o-mO-[\u03c9ka\u00b4ri *!(o. . . O) *c. [\u03c6 O-mO-[\u03c9ka\u00b4ri *! *d. [\u03c6o-mo-[\u03c9ka\u00b4ri *!(o. . . a)Before considering vowel harmony involving other morphemes, we will consider the relationship be-tween allomorphs of the class prefixes. The morphs of the class prefixes are shown below, repeated from(97).(100) Lexical entry for class prefixes{mo, mu, mO} C1{Ba} C2{Ge, Gi, GE} C7...Not every possible distribution of vowel harmony allomorphs are attested; for example, there is not a validlexical entry with only a high vowel allomorph. Following Archangeli and Pulleyblank, 2021, I will examinethe relations between morphs that are minimally different, as these are the types of pairs that a learner wouldconsider when forming lexical entries, and then later when expanding their lexicon.There are two relations between allomorphs of prefixes: one relating mid advanced and mid retractedvowels (mid vowels differing minimally in their tongue root configuration), as shown in (101); and anotherrelating mid advanced and high advanced vowels (advanced vowels differing minimally in their height), asshown in (102). Note that these relations apply to the class prefixes we have been looking at, but they alsomore generally apply to other nominal prefixes; we will consider other nominal prefixes in the followingsection.127(101) Morph set relation: tongue rootMSRTR: Pairs of otherwise identical nominal prefix allomorphs differ only with respect to amid advanced vowel in one being mid retracted in the other.{Mi,M j} Mi: [+ATR, -HI, -LO]M j: [-ATR, -HI, -LO](102) Morph set relation: heightMSRHi: Pairs of otherwise identical nominal class prefix allomorphs differ only with respect toa mid advanced vowel in one being mid high in the other.{Mi,M j} Mi: [+ATR, -HI, -LO]M j: [+ATR, +HI, -LO]These relations function to limit the forms of allomorphs that can be grouped within a class prefix lexicalentry. We will begin by considering tongue root.(103) Morph set condition: TR (nominal prefixes)With respect to MSRTR, a nominal prefix morph set is ill-formed if M j is present and there is nocorresponding Mi, or if Mi is present and there is no corresponding M jMSCTR: *{M j,\u00acMi}*{Mi,\u00acM j}A learner might acquire this sort of condition by observing data like which lead to the lexical entries in(100), where there are no lexical entries that have only advanced mid, or only retracted mid vowels (althoughthere are lexical entries that have only retracted low vowels). A learner might postulate that the conditionabove applies broadly to nominal prefixes (class prefixes, locative prefixes, locative prefixes, determiners);such a learner would find this condition would turn out to be correct given the data we shall see in theremainder of this chapter. However, a learner might also conservatively postulate that MSCTR applies onlyto nominal class prefixes. Such a learner would need to modify their grammar when encountering formswith prepositional prefixes, which also exhibit the same distribution of vowel harmony allomorphs withrespect to tongue root. A learner may also incorrectly generalize that all Nata morph sets are subject toMSCTR; as we shall see when we return to noun stems, this is not the case.128Now we will consider conditions between morphs based on vowel height. Unlike the tongue root MSC,this condition applies only to class prefixes, locative prefixes, and determiners. Associative prefixes are notsubject to this condition; we will consider locative and associative prefixes in the following section.(104) Morph set condition: HiWith respect to MSRHi, a class prefix or locative prefix morph set is ill-formed if M j is present andthere is no corresponding Mi, or if Mi is present and there is no corresponding M jMSCHi: *{M j,\u00acMi}*{Mi,\u00acM j}As with MSCTR, MSCHi can be acquired by considering class prefixes; a learner could then extend this con-dition to apply to locative prefixes. Alternatively, a learner might over generalize that MSCHi applies to allnominal prefixes, like MSCTR; in such a case, the learner would encounter data suggesting that hypothesisis incorrect, specifically from the associative prefix; we investigate this in the following section.Taken together, the conditions on tongue root and vowel height allomorph pairings will result in lexicalentries for class markers, such as those in (100). There will never be lexical entries for a class prefix whichhave an advanced mid vowel, but no retracted mid vowel, and vice versa; moreover, the allomorphs willvary just in tongue root (but not in height, or frontness\/backness). There will never be lexical entries for aclass prefix which have a mid vowel, but no high vowel, and vice versa. Note that morph sets with just lowvowels are permitted by virtue of not being restricted by a condition.In the following section, we will observe that the locative prefix has the same alternations and allomorphsas the class marker while the associative prefix has a different type of morph set and alternation (with respectto vowel harmony).5.2.2 Prepositional prefixesIn this section we will examine prefixes which occur to the left of the determiner. These morphemes arethe associative, locative, and comitative prefixes. Recall from Chapter 3 that these prefixes occur inside thesame \u03c6 as the determiner and class marking prefix, as indicated by the bracketing in the examples below.This dataset has numerous examples of vowel coalescence; however this is not the primary focus on thischapter; consult Appendix A for details about coalescence and how it can be analyzed.The comitative prefix is not of particular interest with respect to vowel harmony as it occurs with an129invariant low vowel. Because of coalescence, this morpheme only surfaces with an allomorph that has avowel when it is followed by a consonant initial morpheme. One such context is before the non-existentialdeterminer because this morpheme has a phonologically null allomorph, as we saw in Chapter 4. Considerthe examples in Table 5.8 illustrating the form of the comitative.Table 5.8: Comitative prefix vowel qualityBatarOOtSE\u00b4 muka\u00b4ri [\u03c6na\u00b4- \/0-mO-[\u03c9 tOONg-O \u2018they weren\u2019t seeing any woman.c1 and string.c3\u2019BatarOOtSE\u00b4 muka\u00b4ri [\u03c6na- \/0-ki-[\u03c9GE\u00b4rO \u2018they weren\u2019t seeing any woman.c1 and thing.c7\u2019BatarOOtSE\u00b4 muka\u00b4ri [\u03c6na\u00b4- \/0- \/0-[\u03c9ntSogu \u2018they weren\u2019t seeing any woman.c1 and elephant.c9\u2019The locative prefixes have a high vowel allomorph as well as both advanced and retracted mid vowelallomorphs, the latter of which we will see in the following section. These surface in the same contexts asthe analogous determiner and class prefix allomorphs, as illustrated by Table 5.9.Table 5.9: Locative vowel quality[\u03c6koo-mo-[\u03c9 s\u00b4isi (c3) \u2018on tamarind tree.c3\u2019[\u03c6moo-Ge-[\u03c9ku\u00b4Ba (c7) \u2018in chest.c7\u2019[\u03c6ku\u00b4u\u00b4-ma-[\u03c9Buri (c6) \u2018on feathers.c6\u2019[\u03c6muu-[\u03c9kj-a\u00b4arO (c7) \u2018in village.c7\u2019The associative prefix has a different set of allomorphs (with respect to vowel quality) than the classmarker prefixes we have focused on thus far; associative prefixes do not have allomorphs with high vowels.As with the comitative, we will consider a subset of cases where the morpheme following the associativeis consonant initial, prohibiting coalescence between the two morphemes (see Appendix A for informationregarding coalescence).130Table 5.10: Associative vowel quality (i)o-mo-s\u00b4isi [\u03c6wO\u00b4-tSaa- \/0-[\u03c9ntSogu \u2018tamarind tree.c3 of elephants.c10e-me-s\u00b4isi [\u03c6GE\u00b4-tSaa- \/0-[\u03c9ntSogu \u2018tamarind trees.c4 of elephants.c10\u2019e-Ge-ku\u00b4Ba [\u03c6GE\u00b4-tSaa- \/0-[\u03c9ntSogu \u2018chest.c7 of elephants.c10\u2019e-Be-ku\u00b4Ba [\u03c6BE\u00b4-tSaa- \/0-[\u03c9ntSogu \u2018chests.c8 of elephants.c10\u2019o-ro-s\u00b4iri [\u03c6 rO\u00b4-tSaa- \/0-[\u03c9ntSogu \u2018rope.c11 of elephants.c10\u2019When preceding a consonant initial DP, associative markers with non-low vowels alternate betweenadvanced and retracted vowel quality. The quality of the associative prefix vowel depends on the qualityof the vowel in the syllable to its right. If the vowel in the syllable to the right of the associative prefix isretracted, then the vowel of the associative prefix will be a retracted mid vowel.Table 5.11: Associative vowel quality (ii)o-mo-s\u00b4isi [\u03c6wo\u00b4-r-ii-[\u03c9Buri \u2018tamarind tree.c3 of feather.c5\u2019e-me-s\u00b4isi [\u03c6Ge\u00b4-r-ii-[\u03c9Buri \u2018tamarind trees.c4 of feather.c5\u2019e-Ge-ku\u00b4Ba [\u03c6Ge\u00b4-r-ii-[\u03c9Buri \u2018chest.c7 of feather.c5\u2019e-Be-ku\u00b4Ba [\u03c6Be\u00b4-r-ii-[\u03c9Buri \u2018chests.c8 of feather.c5\u2019o-ro-s\u00b4iri [\u03c6 ro\u00b4-r-ii-[\u03c9Buri \u2018rope.c11 of feather.c5\u2019If the vowel in the syllable to the right of the associative prefix is advanced, then the vowel of theassociative prefix will be an advanced mid vowel (see Table 3.20 in Section 3.3). Note that if the associativeprefix for a particular class has a low vowel, it will surface with a low vowel when it is followed by advancedvowels and retracted vowels (cf. c2 associative: wa\u00b4-tSaa-ntSogu, wa-mw-EEr\u00b4i, wa\u00b4-rii-Buri).5.2.3 AnalysisThe prefixes we have seen in this section come in three flavours: i) the comitative has a low vowel, andtherefore does not participate in vowel harmony alternations; ii) the locative, which functions like the classprefixes we have seen previously; and iii) the associative prefixes, which have different morphs than the131locatives and class prefixes. The comitative prefixes require no comment. Consider the lexical entries givenfor the locative prefixes; class prefixes are included for comparison. The retracted allomorph of the locativewill be discussed in the following section; for the time being recall that it is not preferred even when thelocative is followed by a retracted (low or mid) vowel, as shown in (5.9)(105) Lexical entry for locative prefixes{koo, kuu, kOO} LOC (ON){moo, muu, mOO} LOC (IN)Lexical entry for class prefixes{mo, mu, mO} C1{Ge, Gi, GE} C7...The ranking of *[-ATR] over *[+HI] is shown to be crucial by (106b. & c.); the surface form (106c.)satisfies *[-ATR], but not *[+HI], in contrast to (106c.)(106) Evaluation of locative vowel harmony[\u03c6{ko\u00b4o\u00b4, kO\u00b4O\u00b4, kuu. . .}+{ a, \/0. . . }+{ ma. . .}+[\u03c9{Buri . . .}] \u03c6 -HARM *[-ATR] *[+HI]a. [\u03c6ko\u00b4o\u00b4-ma[\u03c9Buri *!(o. . . a)b. [\u03c6kO\u00b4O\u00b4-ma[\u03c9Buri *(kO\u00b4O\u00b4)u c. [\u03c6ku\u00b4u\u00b4-ma[\u03c9Buri *(ku\u00b4u\u00b4)The analysis of the associative prefixes depends on the observation that this set of prefixes lacks highvowel allomorphs. The lexical entries for associative prefixes are given below; for comparison, the lexicalentries for class prefixes are shown in (105) above.(107) Lexical entry for associative prefixes{wo, wO} ASSOC .C1{Ge, GE} ASSOC .C7...132When there are no harmony violations, the vowel phonotactics select the surface form, as illustratedbelow.(108) Evaluation of associative vowel harmony (i)[\u03c6{wo\u00b4, wO\u00b4, . . .}+{ r, . . .}+{ ii. . . }+[\u03c9{Buri . . .}] \u03c6 -HARM *[-ATR] *[+HI]u a. [\u03c6wo\u00b4-r-ii-[\u03c9Burib. [\u03c6wO\u00b4-r-ii[\u03c9Buri *!(O. . . i) *(wO\u00b4)When there is a harmony violation, the limited set of vowel allomorphs for the associative prefix means thatviolations of *[-ATR] must be tolerated as there are no other allomorphs which satisfy this global constraintwhich do not result in a harmony violation; this is illustrated by the case below.(109) Evaluation of associative vowel harmony (ii)[\u03c6{wo\u00b4, wO\u00b4, . . .}+{ tSaa, . . . }+{ \/0. . . }+[\u03c9{ntSogu . . . }] \u03c6 -HARM *[-ATR] *[+HI]a. [\u03c6wo\u00b4-tSaa- \/0-[\u03c9ntSogu *!(o. . . a)u b. [\u03c6wO\u00b4-tSaa- \/0-[\u03c9ntSogu *(wO\u00b4)In terms of morph set conditions, the locative is subject to all the same conditions as class prefixes:MSCTR (103) and MSCHi(104). On the other hand, the associative is not subject to MSCHi (thought it issubject to MSCTR). The result of this is that associative prefix morph sets will never have a high vowelallomorph; therefore, they behave differently in the same harmonic contexts when compared to the locativeand class prefixes.5.3 Retracting suffixWe will now return to the stem and a special case of harmony involving a particular suffix. For some formsincluding the nominalizer suffix [-O], all mid vowels within \u03c6 have the same advanced\/retracted value.133Table 5.12: Retracting suffixesForm Gloss[\u03c6 O-mO\u00b4-[\u03c9 tOONg-O \u2018string\u2019 (c3)* [\u03c6o-mu\u00b4-[\u03c9 tOONg-O* [\u03c6u-mu\u00b4-[\u03c9 tOONg-O[\u03c6 E-kE\u00b4-[\u03c9mEr-O \u2018throat\u2019 (c7)* [\u03c6 e-k\u00b4i-[\u03c9mEr-O* [\u03c6 i-k\u00b4i-[\u03c9mEr-O[\u03c6 O-rO-[\u03c9hOONg-O\u00b4 \u2018sieve\u2019 (c11)* [\u03c6o-ru-[\u03c9hOONg-O\u00b4* [\u03c6u-ru-[\u03c9hOONg-O\u00b4The requirement that these forms have mid retracted vowels extends to the prepositional prefixes.Table 5.13: Retracted prepositionsForm Gloss[\u03c6kO\u00b4O\u00b4-mO-[\u03c9 tOONg-O \u2018on string\u2019 (c3)* [\u03c6ko\u00b4o\u00b4-mu-[\u03c9 tOONg-O* [\u03c6ku\u00b4u\u00b4-mu-[\u03c9 tOONg-O[\u03c6nO\u00b4-O\u00b4-mO-[\u03c9 tOONg-O \u2018and string\u2019 (c3)* [\u03c6no\u00b4-o\u00b4-mu-[\u03c9 tOONg-O* [\u03c6nu\u00b4-u\u00b4-mu-[\u03c9 tOONg-O[\u03c6mO\u00b4- \/0-mO-[\u03c9 tOONg-O \u2018of no string\u2019 (c3, c3)* [\u03c6mo\u00b4- \/0-mu-[\u03c9 tOONg-O* [\u03c6mu\u00b4- \/0-mu-[\u03c9 tOONg-OThis pattern has an additional wrinkle. Not every stem that includes a retracting suffix surfaces with134retracted prefixes. Consider data in Table 5.14, which includes the suffix \u2018-O\u2019 that we\u2019ve seen, but surfacewith high vowel class prefixes. Because the class prefixes are high and advanced, the determiner need notraise; likewise any other prepositional prefixes need not be mid retracted or raised.Table 5.14: Non-retracted formsForm Gloss[\u03c6o-mu\u00b4-[\u03c9 tEg-O \u2018trap\u2019 (c3)* [\u03c6O-mO\u00b4-[\u03c9 tEg-O[\u03c6mo\u00b4o\u00b4-mu-[\u03c9 tEg-O \u2018in trap\u2019 (c3)* [\u03c6mO\u00b4O\u00b4-mO-[\u03c9 tEg-OIt may appear that no phonological factor distinguishes between environments like [\u03c6 E-kE\u00b4-mEr-O] \u2018throat\u2019(c7) and [\u03c6o-mu-tEg-O] \u2018trap\u2019 (c3); however, Gambarage and Pulleyblank, 2017 observe that for all retractedstems that surface with retracted determiners and prefixes, those stems contain roots that have both advancedand retracted allomorphs. On the other hand, retracted stems that surface with advanced determiners andprefixes contain roots that only have retracted allomorphs.This is illustrated by comparing the same stems with different suffix vowels. In Table 5.15, roots arebolded; the first two pairs illustrate non-alternating roots (a.-d.), while the second pair illustrate alternatingroots (e.-h.).135Table 5.15: Root types: alternating and non-alternatingForm Glossa. [\u03c6o-mu\u00b4-[\u03c9 tEg-O \u2018trap\u2019 (c3)b. [\u03c6Gu-[\u03c9 tE\u00b4G-a \u2018to trap\u2019c. [\u03c6o-mu\u00b4-[\u03c9 tOOs-O \u2018an abuse\u2019 (c3)d. [\u03c6Gu-[\u03c9 tO\u00b4Os-a \u2018to abuse\u2019e. [\u03c6 O-mO\u00b4-[\u03c9 tOOng-O \u2018string\u2019 (c3)f. [\u03c6Go-[\u03c9 to\u00b4ong-a \u2018to pierce\u2019g. [\u03c6 E-kE\u00b4-[\u03c9mEr-O \u2018throat\u2019 (c7)h. [\u03c6ko-[\u03c9me\u00b4r-a \u2018to swallow\u2019The data presented in this section illustrate that a particular suffix in combination with a particular type ofroot result in a distinct repair to the general problem of advanced mid vowels followed by retracted vowels.Forms like the pair in (109g. & h.) illustrate that vowel harmony is not enforced in the same way within the\u03c9 as it is outside \u03c9 but within \u03c6 . Given that these roots have mid advanced and mid retracted allomorphs,but no high allomorphs, they are similar to the associative prefix. Recall from (5.10) that the associativeprefix occurs retracted when followed by low vowels (as well as when followed by mid retracted vowels,but those are all cases including the [-O] suffix.) I analyze this distinction in behaviour via the restrictorclause in the \u03c6 -HARMONY constraint, as alluded to previously.The chart below illustrates the types of surface allomorphs for the class prefix, determiner, and locative,given the contexts we have discussed in the preceding sections: the vowel preceding the vowel of interest(outside of the \u03c9 , but inside \u03c6 ), if the root is alternating or not, and if the form includes the retracting suffix[-O].136Table 5.16: \u03c6 Harmony summaryPrefix Following V Root Suffix Forma. [+ATR]\/mid [+ATR] non-alternating [+ATR] [\u03c6o-mo\u00b4-[\u03c9 rem-ib. [+ATR]\/high [-ATR] non-alternating - [\u03c6o-mu-[\u03c9ka\u00b4ric. [+ATR]\/high [-ATR] non-alternating [-ATR] [\u03c6o-mu\u00b4-[\u03c9 tOOs-Od. [-ATR]\/mid [-ATR] alternating [-ATR] [\u03c6 O-mO\u00b4-[\u03c9 tOOng-OIn summary, the retracting suffix [-O] imposes a distinct repair to the general prohibition on advancedmid vowels followed by retracted vowels. For forms with the retracting suffix, vowels to the left of the stemcontaining the suffix must be mid; this requirement in combination with the \u03c6 -HARMONY constraint resultsin the prefixes surfacing with mid retracted allomorphs. The retracting suffix is dependent on a particulartype of root, characterized in terms of allomorphs; when the suffix is paired with a root that has mid advancedand mid retracted allomorphs, the prefixes must have mid retracted vowels (5.16 d.). However, when thesuffix is paired with a root that has only the mid retracted allomorph, the prefixes are not subject to therequirement that they be mid (5.16 c.).AnalysisIn the previous section we observed that retracting suffixes are dependant on alternating roots to cause prefixmorphemes to surface with mid retracted allomorphs. There are two effects at play here: retracting suffixesmust be preceded by stems with retracted vowels, and alternating roots must be preceded by mid vowels.The former is already formalized by our \u03c9-word harmony constraint. We can encode the latter effect as aselectional restriction (Archangeli & Pulleyblank, 2017; Gambarage & Pulleyblank, 2017). It is importantthat all vowels within the relevant domain are mid, and not just the immediately preceding vowel; the caseevaluated in (113) illustrates this.Selectional restrictions are a general property of morphemes; on top of phonotactics (such as for har-mony), individual morphemes may have specific phonological requirements. See \u00a72.2.2 for more informa-tion about how restrictions are modelled in Lexical Allomorphy. The alternating roots have a selectionalrestriction that they must be preceded only by mid vowels within the \u03c6 , in other words, they have a non-137local selectional restriction for [-HI, -LO]. This restriction is crucial in capturing the behaviour of alternatingstems when it comes to vowel harmony, as we shall see shortly.(110) Lexical entry for alternating roots{mer, mEr}SEL: [-HI, -LO]\u03c6 THROAT{tooNg, tOONg}SEL: [-HI, -LO]\u03c6 STRING...{CeC, CEC}SEL: [-HI, -LO]\u03c6 RTSEL: [-HI, -LO]\u03c6 : Vowels within the same \u03c6 as this morpheme must be mid.Note that the lexical entry for the nominalizer suffix has no selectional restriction.(111) Lexical entry for nominalizer suffix{O} NOMAlternating stems have the special property of selection; like the individual allomorphs and the redun-dancy relations between them (see Section 2.2.2), the selection restriction is learned as a property of thesetypes of stems. Because the selectional restriction is a property of alternating stems, we correctly predictthat non alternating stems which have retracted mid vowels only cause raising in prefixes in order to satisfythe phonotactic on advanced mid vowels followed by retracted vowels.(112) Evaluation of a non-alternating stem with retracting suffix[{o, O, u}+{mO, mu, mo. . .}+{tEG}+{O}] SEL \u03c6-HARM\u03c9-HARM*[-ATR] *[+HI]a. [\u03c6o-mo\u00b4-[\u03c9 tEg-O *!b. [\u03c6o-mO\u00b4-[\u03c9 tEg-O *! *(mO\u00b4)c. [\u03c6 O-mO\u00b4-[\u03c9 tEg-O *!(m\u00b4O), *(O)d. [\u03c6u-mu\u00b4-[\u03c9 tEg-O *(mu\u00b4),*!(u)u e. [\u03c6o-mu\u00b4-[\u03c9 tEg-O *(mu\u00b4)The alternating stems limit the solutions to this prohibition by requiring mid vowels to their left. There-fore, the best surface forms in these cases will have retracted mid vowels.138Compare the candidate in (112e.) above with the analogous candidate with an alternating root in (113f.)below; the latter candidate fails exactly because it does not satisfy the selectional restriction of the stem{tooNg, tOONg}SEL: [mid]\u03c6 . The tableau below illustrates the activity of the \u03c9-word level harmony constraint:candidates in (113a.-c.) are all ruled out by this phonotactic. The only way to satisfy the \u03c6 -phrase harmonyconstraint while satisfying selection is the optimal candidate in (113h.).(113) Evaluation of a alternating root with retracting suffixBased on Gambarage and Pulleyblank, 2017 (20)[{o, O, u}+{mO, mu, mo. . .}+{tooNg, tOONg}SEL: [-HI, -LO]\u03c6+{O}] SEL \u03c6-HARMONY\u03c9-HARM*[-ATR] *[+HI]a. [\u03c6o-mo\u00b4-[\u03c9 tooNg-O *!b. [\u03c6o-mO\u00b4-[\u03c9 tooNg-O *!* * *(mO\u00b4)c. [\u03c6o-mu\u00b4-[\u03c9 tooNg-O *! * *(mu\u00b4)d. [\u03c6o-mo\u00b4-[\u03c9 tOONg-O *!e. [\u03c6o-mO\u00b4-[\u03c9 tOONg-O *! *(mO\u00b4)f. [\u03c6o-mu\u00b4-[\u03c9 tOONg-O *! *(mu\u00b4)g. [\u03c6 O-mo\u00b4-[\u03c9 tOONg-O *! *(O)u h. [\u03c6 O-mO\u00b4-[\u03c9 tOONg-O *(O), *(mO\u00b4)i.[\u03c6 O-mu\u00b4-[\u03c9 tOONg-O *! * *(O) *(mu\u00b4)j. [\u03c6u-mO\u00b4-[\u03c9 tOONg-O *! *(mO) *(u)Compare the candidates in (113i. and j.) to the optimal candidate in (113h.): both fail to satisfy the selec-tional restriction of the retracted stem allomorph, repeated below from (110).(114) SEL: [-HI, -LO]\u03c6 : Vowels within the same \u03c6 as this morpheme must be mid.The form in (114j.) illustrates that the distance between the allomorph which selects is not measured byimmediate precedence, but instead by precedence within the \u03c6 domain. If precedence was measured bysyllable adjacency, then the candidate in (114j.) would be optimal, as it wouldn\u2019t violate this version of theselectional restriction.2 This is because the vowel immediately precedence the retracting suffix is retracted.2This evaluation is consistent with a phase based interpretation of local allomorphy, such as Embick and Marantz, 2008, where139In this analysis, \u201cretracting\u201d suffixes formally have no special condition associated with them. The effectof a retracting suffix is a combination of the phonotactic prohibiting advanced mid vowels that are followedby retracted vowels. We can confirm this by considering the optimal forms if the selectional restriction in(114) were a property of the suffix and not the stem; lexical entries for this alternative are shown below.(115) Alternative lexical entries for alternating roots and nominalizer suffix{mer, mEr} THROAT{tooNg, tOONg} STRING...{CeC, CEC} RT{O}SEL: [-HI, -LO]\u03c6 NOMThe problem with this alternative becomes immediately clear if we consider cases with non-alternatingstems, like we saw in (112). In such cases, the observed surface form has an advanced high vowel in theprefix (116c.). The alternative analysis which attributes the selection of mid vowels to the nominalizer suffixincorrectly predicts that the prefix (and determiner) should be mid and retracted (116f.), as shown below.(116) Alternative evaluation of a non-alternating stem with retracting suffix[{o, O, u}+{mO, mu, mo. . .}+{tEG}+{O}SEL: [-HI, -LO]\u03c6 ] SEL \u03c6-HARM\u03c9-HARM*[-ATR] *[+HI]a. [\u03c6o-mo\u00b4-[\u03c9 tEg-O *!b. [\u03c6o-mO\u00b4-[\u03c9 tEg-O *! *(mO\u00b4)\/ c. [\u03c6o-mu\u00b4-[\u03c9 tEg-O *! *(mu\u00b4)d. [\u03c6u-mu\u00b4-[\u03c9 tEg-O *! *(mu\u00b4), *(u)e. [\u03c6u-mO\u00b4-[\u03c9 tEg-O *! *(mO\u00b4) *(u)u f. [\u03c6 O-mO\u00b4-[\u03c9 tEg-O *(O), *(mO\u00b4)A-categorical roots are interesting in the context of the morph set relations we have examined thus far.This group of morphemes do not pattern like class and locative prefixes or like associative prefixes withcyclic, category defining, heads (such as n), can influence allomorph selection for morphemes that are within their domain. In thispaper the same effect is derived by reference to phonological domains that are reflexes of syntactic domains (which are defined bysyntactic heads).140respect to their morph sets. Consider the types of morph sets that these morphemes can have.(117) Lexical entry for a-categorical roots{mer, mEr}SEL: [mid]\u03c6 THROAT{tooNg, tOONg}SEL: [-HI, -LO]\u03c6 STRING...{CeC, CEC}SEL: [-HI, -LO]\u03c6 RT{tEg, } TRAP{tOOs}SEL: [-HI, -LO]\u03c6 ABUSE...{CEC} RTThese two distinct morph sets share the same tongue root relation as nominal prefixes: there is a correspon-dence between mid advanced and mid retracted allomorphs.(118) Morph set relation: tongue root (a-categorical roots)MSRTR: Pairs of otherwise identical a-categorical root allomorphs differ only with respect to amid advanced vowel in one being mid retracted in the other.{Mi,M j} Mi: [+ATR, -HI, -LO]M j: [-ATR, -HI, -LO]Unlike nominal prefixes, which have a bidirectional condition on morphs, as shown in (103), a retractedmorph can occur without a corresponding advanced morph in the same set.(119) Morph set condition: TR (a-categorical roots)With respect to MSRTR, an a-categorical root morph set is ill-formed if Mi is present and there is nocorresponding M j.MSCTR: *{Mi,\u00acM j}This condition will prohibit morph sets which have an advanced morph but no retracted morph, but they donot penalize a morph set with an unpaired retracted allomorph. For this reason, a-categorical roots can eitherbe alternating, or invariably retracted, but not invariably advanced. Note that noun stems in general have no141such relation; for an underived stem, a learner would not acquire any particular relation between morphs norwould they learn a condition, precisely because these stems do not vary with respect to vowel quality.5.4 Summary and ConclusionIn this section I will list and review the vowel harmony generalizations we\u2019ve observed in this chapter, howthey are accounted for, and future directions for research.Table 5.17: Vowel Harmony Observationsi. Within stems, mid vowels are all advanced or all retracted.ii. Generally, vowels in the prefix, determiner, and locative morphemes are advanced mid vow-els unless they are followed by a retracted vowel in which case they will be advanced highvowels.iii. The nominalizing suffix, [-O], only occurs preceded by retracted mid vowels (i.e. no ad-vanced mid vowels precede this suffix).iv. When preceding a consonant initial DP, associative markers with non-low vowels alternatebetween advanced and retracted vowel quality.We observed that stems are restricted to either having advanced or retracted mid vowels. Consideringthe material inside of the \u03c6 we observed that the vowels outside of the stem are generally advanced (eithermid or high); mid vowels are more common, and the high variants occur preceding retracted vowels (eithermid or low). Observations (5.17 iii.) and (iv.) pertain to particular morphemes which condition alternationsfrom the general patterns observed in (i.-ii.)Considering the analysis of these observations, the first observation is true of \u03c9; the second observationis the general pattern of harmony in the \u03c6 . Both of these observations are the result of phonotactics prohibit-ing non-identical values of [\u00b1ATR] for a particular set of vowels within a domain. These phonotactics aremodelled using two co-occurrence constraints which penalize disagreeing features of [\u00b1ATR]. The \u03c9 levelconstraint only penalizes mid vowels that differ in [\u00b1ATR], while the \u03c6 level constraint penalizes vowelsthat differ in [\u00b1ATR] regardless of height; however, it does not consider vowels inside of the \u03c9 .The analysis of forms including stems with the [-O] suffix depends on one of two types of a-categorical142roots. Stems composed of an a-categorical root and suffix are distinct from non-alternating stem, whichhave no vowel harmony allomorphs, because the a-categorical roots have allomorphs with advanced midvowels and allomorphs with retracted mid vowels; in addition, the retracted allomorphs bear a selectionalrestriction which prohibits high vowels from preceding them. Not all a-categorical roots alternate, andthe non-alternating set do not bear the same selectional restriction; therefore the result is that the presenceof both an alternating a-categorical root and the [-O] suffix results in a unique harmonic configuration fornominal \u03c6 , where all mid vowels are retracted.The associative prefix is distinct from other prefixes (class marker, locative, determiner) in that it doesnot have a high vowel allomorph. This distinction in the type of allomorphs available for this prefix incontrast to other prefixes accounts for the associative prefix\u2019s distinct response to the harmony phonotactic.In terms of analysis, a learner would need to observe that the associative prefix is exceptionally not subjectto MSCHi (104).143Chapter 6DiscussionThe focus of this discussion is to evaluate what the emergent approach to morphology and phonology con-tribute to the study of Language and more specifically to the study of Nata. I will begin in 6.1 by answeringbroadly why I chose to use an emergent approach. In 6.2, I consider how emergence differs from a moretraditional auto-segmental phonology approach and the future directions that we can explore via emergence.Finally, in 6.3, I compare the tonal analysis of Nata with a simple analysis of Jita, as described by Downing,1990; this final case bolsters comparisons with other tonal examples, and provides a more similar languagecomparison.6.1 Why emergence?Tone and vowel harmony have been investigated for a relativity long time in the long span of generativelinguistics. Considering these phonological phenomenon in Nata from the perspective of emergent grammaris worthwhile because we already have an arsenal of tools to model them. Emergence offers to pare downall of these abstract mechanisms in place of surface forms, as modelled in lexical allomorphy.One central question is: how much abstract structure do we need to add back to account for tone andharmony? The conclusion is that the emergent approach is capable of explaining long distance movement,as seen in tonal alignment, as well as distinct phonological classes, as seen in tone classes. What is alsoclear is that for Nata abstract underlying forms, specifically floating tones and spreading harmonic vowelfeatures, are not required for an analysis of tone or vowel harmony.In the following section I will consider a number of approaches that are similar to lexical allomorphy, inso far as they are mechanisms to handle allomorphy. Setting aside underlying forms, the major distinction144between lexical allomorphy and other approaches to allomorphy is that in lexical allomorphy all forms arelisted allomorphs; in other approaches to allomorphy only some types of forms are listed as allomorphs,while others are generated via a phonological mechanism.6.2 Emergence contrastedBonet and Harbour, 2012 review a number of sources of allomorphy, concluding that there is a broad rangeof triggers, and relations between morphs, all of which could be described as allomorphy. Generally, theyobserve that allomorphy describes the relation between phonological forms of a morpheme that are notsystemically predicted via phonology.Regardless of the details, the relation to lexical allomorphy is simple: all surface forms are listed in thelexicon, as allomorphs would be treated. The choice of surface forms is still determined by phonotacticconsiderations (the ranking of phonotactic constraints and evaluation of potential surface forms.); however,the special selectional properties of allomorphs also are a regular feature of lexical allomorphy, not anexceptional property of only those allomorphs without sufficient phonological similarity to be reevaluatedvia a singular underlying form.Adopting lexical allomorphy ultimately removes the questions regarding where on the scale of predictable-to-listed a particular type of form is. Likewise, there is no real question as to what the nature of allomorphsis: they are surface forms. This removes the possibility of floating or other abstract features listed as allo-morphs. As we have seen, this limitation does not hinder the analysis of Nata. Bonet and Harbour, 2012point out that theories of allomorphy typically lack bounds on the number of allomorphs, and in this sense,lexical allomorphy is the same. In fact, being a Bantu language with an extensive class system, Nata gen-erally illustrates that a single morpheme may have upwards of ten allomorphs, as is the case for any of thenominal prefixes examined in this work.As an emergent approach to grammar, lexical allomorphy does not restrict the types of generalizationslearners can draw regarding locality and position of conditioning context. As we shall see, some approachesto allomorphy distinguish between stems\/affixes in their ability to trigger morph selection. As we haveseen in the case of the associative prefix, affixes can select allomorphs of stems (albeit under direct tonaladjacency). From the case of the retracting suffix and alternating stems, we can observe that stems canselect for properties of their prosodic domain, which is not a segmentally adjacent context (one morph ofthe alternating stem morpheme type selects for all mid vowels within \u03c6 ).145The emergent approach to grammar, as realized via lexical allomorphy, shares some similarities withsub-categorization frames proposed by Paster, 2009, 2015; Paster, 2006 to account for phonologically con-ditioned suppletive allomorphy (PCSA). Although this approach still retains underlying forms, the compar-ison is of interest because lexical allomorphy treats all alternations as suppletive allomorphy, in the sensethat individual morphs are chosen via the phonology; this is contrasted with an approach in which underly-ing forms are changed via the phonology in order to respect phonotactics resulting in the observed surfaceforms. In this section I will examine some of the ways in which Paster\u2019s (2006) approach to PCSA is similarto and different from an emergent approach.Paster, 2006 observes that PCSA is sensitive to underlying forms, and not to surface forms. This findingis of interest in the context of an emergent approach to allomorphy precisely because there is no distinctionbetween underlying and surface forms. To begin with, let us understand what PCSA refers to. The sense inwhich PCSA is phonologically conditioned is that allomorphs of affixes can be subcategorized by phono-logical properties of stems. The sense in which PCSA is suppletive is that particular allomorphs are deemedto be too different in structure for a phonological rule or constraint to relate them via a single underlyingform.Underlying forms are not available in an emergent approach. Two immediate consequences are that i) itis not possible to judge if two allomorphs are in a suppletive relationship; and ii) emergent allomorphy mustbe sensitive to surface forms.As we have seen, it is not the case that emergence has nothing to say about how allomorphs are related.In fact, relating allomorphs is a crucial portion of how a learner productively fills their lexicon, via morphset relations and morph set conditions. Nevertheless, morphs need not be related; some must simply beobserved by a learner because no systematic relation will produce them: they are arbitrary members of amorph set.In the emergent approach, the only types of phonological forms a learner can access are surface forms:snippets of spoken language they have perceived. It is not the case that there is no abstract informationstored along with these surface forms. As we have seen, selectional restrictions are required to modelarbitrary alternations; in other words, alternations which are not predictable from the general set of (ranked)phonological constraints.Given this way of aligning properties of PCSA with the emergent approach adopted in this work, it ispossible to compare the tonal cases in Nata to the cases Paster, 2006 considered. It is worth noting that of146137 examples of PCSA, only two are tonal. The first example is from Zahao, as described in Yip, 2004.Zahao verb stems have two forms, primary and secondary, the surface distribution of which is conditionedby syntactic or prosodic factors.Table 6.1: Zahao verb stemsPrimary Secondarya. hmaan L hmaan L \u2018be correct\u2019hreen L hren H \u2018lock up, close\u2019b. laam LH laam L \u2018dance\u2019hmaan L hmaan L \u2018be correct\u2019Naan H Naan L \u2018write\u2019The primary form of a stem is not predictive of the secondary form it will have; some primary L stemshave identical secondary forms, while others have H secondary forms. Secondary forms are not predictiveof primary forms; some secondary L stems have identical primary forms, while other have LH or H primaryforms. For an emergent approach, this would mean learners must encounter a stem in both forms, andwill never be able to predict for a novel form if it does or does not alternate. Where this type of splitbehaviour between stems is problematic for generative or optimizing approaches, it is neither problematicnor exceptional for an emergent approach.Yip, 2004 proposes each stem that alternates has two underlying forms; a constraint requiring all un-derlying forms to be realized in some environment then requires each of them surface. In addition to thesemechanisms, Yip proposes a scale of phonotactic constraints which limits how allomorphs of a stem canvary. Interestingly, the analysis proposed by Yip has many properties in common with how lexical allo-morphy analyses the predictable relationships between allomorphs via morph set relations and morph setconstraints. This case is significant to emergent grammar in that it poses a set of problems for generativeand optimizing models of phonology that just do not exist in the lexical allomorphy approach. This is pre-cisely because lexical allomorphy lists all allomorphs, and makes no distinction between those that are veryclearly phonologically related and those that are opaque to such an analysis. In addition, this approachshares with lexical allomorphy that learners make generalizations about how morphs are related.147The second example is from Yucunany dialect of Mixtepec Mixtec, as described in Paster, 2006. The 1sgsubject\/possessor morpheme alternates between two forms depending on the tone of the preceding stem.1Table 6.2: Yucunany 1sg subject\/possessorna`ma\u00b4 \u2018soap\u2019 na`ma\u00b4a` \u2018my soap\u2019v\u0131\u00b4lu\u00b4 \u2018cat\u2019 v\u0131\u00b4lu\u00b4u` \u2018my cat\u2019yu`u\u00b4ti \u2018sand\u2019 yu`u\u00b4ti\u0131` \u2018my sand\u2019so`ko` \u2018shoulder\u2019 so`ko` yu` \u2018my shoulder\u2019tutu` \u2018paper\u2019 tutu` yu` \u2018my paper\u2019cha\u00b4\u2019a` \u2018short\u2019 cha\u00b4\u2019a` yu` \u2018I am short\u2019ve\u2019e ncha\u00b4\u2019\u0131` \u2018black house\u2019 ve\u2019e ncha\u00b4\u2019\u0131` yu` \u2018my black house\u2019One form is analyzed as a floating L tone, the other form is [yu`]. The [yu`] form is used following lowtone final stems, while the floating L is used following H and M final stems.An area where emergence differs from an ASR approach is in handling tonal morphemes. If two mor-phemes or constructions are distinguished by only tone, then that tonal difference must be morphological.In a system where morphemes can be individual tones, a morpheme could correspond to an unassociatedtone. However, in a surface oriented system like emergence, a tone is never observed without a vowel tobe realized on (or other TBU). This does not mean such an alternation cannot be accounted for; learnerscan distinguish the two environments and generalize which allomorph corresponds to which morphologicalcontext.In terms of lexical allomorphy, the 1sg subject\/possessor has a number of allomorphs correspondingto each vowel hosting that low tone (in addition to the [yu`] morph): {a`, u`, \u0131`, yu`}. The low tone vowelmorphs are predictably related via an MSR, which may or may not be more generally active with respect toother morphemes. The choice of which low tone vowel is decided by a phonotactic preventing dissimilarsequences of vowels within the same syllable (see Appendix A) for a treatment of such sequences in Nata).The [yu`] morph is not predictable from a general morph set relation which would predict that every morphset could have such a morph instead of just the 1sg subject\/possessor. This morph must simply be observed1V=Mid tone, V\u00b4=H tone, V`=L tone148and learned.Regarding the [yu`] allomorph, Paster, 2006 proposes that it is subcategorized for a final L stem. Interms of lexical allomorphy, a phonotactic which penalizes long vowels with low tone (either globally,or in word\/phrase final position) would account for the distribution of 1sg subject\/possessor allomorphs. Incontexts where the possessor is preceded by a H or M tone, the V` form is used, with the appropriate matchingof vowel features. In contexts where the possessor is preceded by a L tone, the [yu`] allomorph is used. Inorder to prohibit over usage of the [yu`] form, a constraint penalizing structure would need to invoked; sucha constraint would prefer V forms to CV forms. This is similar to the notion of faithfulness penalizingepenthetic structure in classical Optimality Theory, as in Prince and Smolensky, 1993\/2004. However, itis clearly not directly analogous given the different assumptions about underlying forms being made in thecurrent work.Paster, 2006 relies on comparison with a more conservative dialect of Mixtepec Mixtec in which the twodistinct 1sg subject\/possessor morphs are not members of the same lexical entry, but are actually the politeand informal variants. Paster proposes that the polite\/informal register distinction was lost in the Yucunanydialect of Mixtepec Mixtec, leading to free variation between the forms. Because the tonal form is difficultto perceive after a low tone stem, the segmental variant was preferred there. At some point, the variantsceased to be in free variation, and the segmental variant was only found after low tone stems. This analysisis consistent with the proposal sketched above where the [yu`] allomorph is only tolerated when it avoids aparticular phonologically illicit sequence.Paster, 2006 predicts we should not find cases where an affix influences the surface form of a root;however, the analysis of prepositional prefixes includes a selectional restriction which impacts the optimalsurface form of the root, this was shown in Section 4.6. Note that both the stem and affix select for each other;it is nevertheless crucial that the affix select for the stem. Regardless, this property is not compatible withPaster\u2019s predictions. Moreover, it is not clear Paster would allow for long range effects like that proposedfor roots that alternate between mid advanced and mid retracted, as shown in Section 5.3. In some sensedomain membership could be considered adjacency, but it certainly is not identical to segmental, or syllabic,adjacency.Like PCSA, lexical allomorphy allows for a singular method of morph selection, as well as allowingfor non-phonologically optimizing outputs. The first property arises from considering all alternations to beallomorphy. The second property is due to the lack of restriction on what kind of selection constraints a149morph may have; additionally, there is no meta-condition on MSR\u2019s which limits what kind of morphs maybe related in a morph set for a particular lexical entry.Regarding the relationship between morphological and phonological constraints, in Nata we observedthat selection outranked phonotactics. There is nothing about emergent grammar, nor indeed about lexicalallomorphy, that predicts this to be the case for every language.With respect to prosodic domains, emergence does not provide any unique insights. What is crucial forthe analysis of Nata is that certain morphemes are parsed together by the phonology. Namely, the entire\u03c6 must be available, as shown by the effects that the associative prefix can have on stem morphs as wellas by the effect that alternating roots can have on prepositional prefixes with respect to vowel harmony.The weakest conclusion here is that phonology parses at least the \u03c6 as a \u201cchunk\u201d. There is no reason tobelieve that whole utterances are not parsed as a whole, with their internal prosodic structure simultaneouslyavailable for evaluation.Concerning Bantu languages, there is a general question regarding how right edges of syllables alignwith syntactic domains. In Nata this arises with respect to alignment of H tone on onsetless syllables. Inthis work, we have analyzed \u03c6 as being licit with an initial vowel; however, this vowel is not phonotacticallylicit to host a H tone. Another interpretation is that the initial vowel stands outside of the \u03c6 , which is whyit is not a valid host for H tone. One piece of evidence in support of this position is the observation thatwhen preceded by a verb, and in fast speech, the initial vowel of a noun (in other words, the determiner)will coalesce with the final vowel of the verb (see Appendix A for more on coalescence.) This suggeststhat the initial vowel can be parsed into the same \u03c6 as the verb, outside of the \u03c6 of the noun to which it(syntactically) belongs. This is not particularly problematic if we assume that the phonology has availableto it entire utterances (which would contain both the verbal and nominal \u03c6 \u2019s).As a general note on phonological learning, Lexical Allomorphy does not prescribe an order to whatlearners must encode first: constraints, or allomorphs. Specific events can trigger the learning of a newallomorph, such as observing a novel form at a high enough threshold; however, this can occur at any pointin phonotactic learning. Likewise, a phonotactic can be learned before the lexicon is (nearly) fully populated,or after a large amount of forms have been learned; and can be based on generalizations over large numbersof forms. In short, learning is not thought of as serial, but rather parallel. Moreover, it need not be the casethat a learner obtains a ranking for a constraint and never changes it; additional allomorphs or contexts mayshow that a particular constraint needs to be re-ranked. The model of phonological represented here is an150idealized final state.6.3 Nata and JitaOther Bantu languages have a similar inventory of nominal tone patterns; Jita (Downing, 1990) has the samethree surface stem classes as Nata when considering bisyllabic noun stems.Table 6.3: Tone patterns of bisyllabic noun stems in Jita (Adapted from Downing, 1990 (69))Class Example Gloss(D-NUM-n)Low (0 tone) o-mu-saani \u2018friend\u2019 (c1)Initial H o-mu-ga\u00b4si \u2018woman\u2019 (c1)Final H o-mu-tuuNga\u02c6 \u2018rich person\u2019 (c1)The comparable Nata forms are provided in Table 6.4:Table 6.4: Stem tone class in NataClass Example Gloss(D-NUM-n)Low o-mo\u00b4-rem-i \u2018farmer\u2019 (c1)Initial H o-mu-ka\u00b4ri \u2018woman\u2019 (c1)Final H e-Gi-sEErO\u00b4 \u2018the\/a hide\u2019 (c7)H tone on Jita noun stems is far more variable depending on phrasal position than the Nata analogues.Both the Initial H and Final H classes alternate between phrase final and phrase medial forms; note that theforms in the examples above are phrase final. In phrase medial position, the initial H class surfaces with Htone on the final syllable, as illustrated below.(120) Tone variation by phrase position in Jitao-mu-nye\u00b4e\u00b4mbe \u2018mango tree\u2019 (c3)151o-mu-nyeembe\u00b4 gwaate\u00b4mwa \u2018the mango tree was cut\u2019In phrase medial position, the final H class will surface without H tone on the stem.(121) Tone variation by phrase position in Jitao-mu-tuuNga\u02c6 \u2018rich person\u2019 (c1), mu\u00b4mu\u00b4ji \u2018in town\u2019o-mu-tuuNga mu\u00b4mu\u00b4ji \u2018rich person in town\u2019Trisyllabic stems have three types in phrase final position. These correspond to the three classes identifiedfor bisyllabic stems; however, instead of an Initial H class, trisyllabic stems have an Penultimate H class. Inthe context of a bisyllabic stem, initial and penultimate positions are identical.Table 6.5: Tone patterns of trisllabic nouns in JitaClass Example Gloss(D-NUM-n)Low (0 tone) o-mu-lamusi \u2018judge\u2019 (c1)Penultimate H e-Bi-tuuNgu\u00b4ru \u2018onions\u2019 (c8)o-mu-kwaaru\u00b4u\u00b4jo \u2018scalp scratcher\u2019 (c3)Final H li-darina\u02c6 \u2018tangerine\u2019 (c5)The penultimate H stem class splits into two subclasses when considered in a medial context. This isanalyzed as an underlying distinction in where H tone is linked. Downing, 1990 proposes a general shiftingrule that applies to underlying H tones, moving them one syllable to the right. The effect of this rule is clearin the medial contexts, as shown by the locative constructions below.(122) Antepenultimate stems in medial positions: shifting v. nonshiftingDerivation of shifting form iiBustan\u00b4i mumu\u00b4ji \u2018garden in town\u2019Underlying iiBusta\u00b4ni mumu\u00b4jiShift & Delink iiBustan\u0131\u00b4 mumu\u00b4(ji)Surface iiBustan\u00b4i mumu\u00b4ji152Derivation of nonshifting form eBituuNgu\u00b4ru mutSika\u00b4po \u2018onion in basket\u2019Underlying eBitu\u00b4uNguru mutSika\u00b4poShift & Delink eBituuNgu\u00b4ru mutSika\u00b4(po)Surface eBituuNgu\u00b4ru mutSika\u00b4poThe tone shifting rule is comprised of two independent rules: one which adds an association line one syllableto the right of a H tone, and a second rule, ordered after the first, which deletes all but the rightmostassociation line associated to a H tone; this is refereed to as shifting and delinking. Regarding the finalsyllable of the locative modifiers, these are considered to be extra tonal, and are not candidates for landingsites of shifted tones; Downing, 1990 considers all utterance final syllables to be extra tonal.The shifting rule applies to the head nominals in the locative construction; however, tone shift does notapply to both types of stems when the head nominals occur unmodified. This is because the final syllable ofthese forms is considered extra tonal, as it is utterance final. The shifting rule and extra tonality of phrasefinal syllables accounts for the asymmetrical behaviour of shifting and non shifting stems.(123) Antepenultimate H stems in final positionDerivation of iiBustan\u00b4i \u2018garden\u2019Underlying iiBusta\u00b4niShift & Delink iiBusta\u00b4(ni)Surface iiBusta\u00b4niDerivation of eBituuNgu\u00b4ru \u2018onion\u2019Underlying eBitu\u00b4uNguruShift & Delink eBituuNgu\u00b4(ru)Surface eBituuNgu\u00b4ruBecause our analysis of Nata is not formulated in terms of deriving surface forms from underlying forms,we will consider the Jita data in terms of Lexical Allomorphy. Jita has three types of nominal stems: Lowstems, Initial(2\u03c3 )\/Penultimate(3\u03c3 ) H, Antepenultimate H, and Final H. In Jita, there are two non-alternatingstems: Low, and Antepenultimate H. Penultimate H and Final H both alternate and are considered to havemore than one tonal allomorph, in terms of LA. Note that the stem labels are misleading from the perspective153of a surface oriented theory like LA; however, I will continue using them for the sake of continuity from ourpresentation of the data, and in following with Downing, 1990.Table 6.6: Lexical entries for Jita stemsTone class Lexical EntryLow {saani} FRIENDInitial H {ga\u00b4si, gas\u00b4i} WOMANPenultimate H {Busta\u00b4ni, Bustan\u00b4i} GARDENAntepenultimate H {tuuNgu\u00b4ru} ONIONFinal H {darina\u02c6, darina} TANGERINEIn Nata, the final H stem class is unique in that the only allomorph available to morphemes from thisclass is the final H allomorph. This allomorph is never optimal in any context except when it is the onlyallomorph available for a stem.Although the forms of the invariable stems in Jita and Nata are distinct, their relationship to the tonesystem is similar. In both Nata and Jita, there is a pressure to align H tone to an edge. In Nata, it is the leftedge, as we have seen in the previous chapter. In Jita, H tones shift rightward, albeit the shifting is limited tothe adjacent syllable. The Nata Final H and Jita Antepenultimate stems represent morphemes which alwaysviolate the basic rule of alignment in those languages, precisely because they lack any allomorphs whichwould satisfy the requirements of alignment.Of course, there are a number of other distinctions in the tone systems of Nata and Jita. One strikingdifference is that words in Jita do not require an H toned syllable. Therefore low stems do not interactwith c9 morphology in the same way that their Nata counterparts do; Downing, 1990 lists numerous formsin c9 which occur with a low stem and no H tone on the prefixal material. Likewise, forms with final Hstems in medial positions surface without an H tone. In conclusion, Jita does not support the same typeof morpheme-based tonal requirements that Nata does; c9 prefixes have no special effect on any stem, andneither do c5 prefixes. This is partially related to the lack of mandatory H on every word, and partiallyrelated to the rightward (i.e. away from the prefixal material) shifting in Jita.1546.4 ConclusionIn this work I have illustrated a number of tonal and vowel quality alternations in Nata, primarily focusingon the nominal system. I have extended previous analysis of the tone system to consider the effects ofprefixes, and likewise have considered these same domains in vowel harmony. I have developed an accountof tone and vowel harmony that foregoes underlying forms, and instead focuses of surface allomorphs andthe relationships between them. One significant conclusion of this study is that lexical allomorphy has thecapacity to account for aspects of prosodic phonology without access to abstract forms. Furthermore, withrespect to the study of allomorphy, this study illustrates that affixes can select for stem allomorphs, and thatlong distance effects are possible given reference to prosodic domains.155BibliographyAdger, D. (2003). Core syntax. Oxford University Press.Anghelescu, A. (2012). Morphophonology and tone in Nata. University of British Columbia WorkingPapers in Linguistics (UBCWPL), 34, 89\u2013103.Anghelescu, A., Gambarage, J., Lam, Z., & Pulleyblank, D. (2017). Nominal and verbal tone in Nata: Anallomorphy-based account. In J. Kandybowicz & H. Torrence (Eds.), Africa\u2019s endangeredlanguages: Documentary and theoretical approaches. Oxford University Press.Archangeli, D., & Pulleyblank, D. (1994). Grounded phonology. MIT Press.https:\/\/doi.org\/10.2307\/416805Archangeli, D., & Pulleyblank, D. (2015). Phonology without universal grammar. Frontiers in Psychology,6. https:\/\/doi.org\/10.3389\/fpsyg.2015.01229Archangeli, D., & Pulleyblank, D. (2016). Emergent Morphology. In D. Siddiqi & H. Harley (Eds.),Morphological metatheory (pp. 237\u2013270). John Benjamins Publishing.https:\/\/doi.org\/10.1075\/la.229.09arcArchangeli, D., & Pulleyblank, D. (2017). Phonology as an emergent system. In S. Hannahs & A. Bosch(Eds.), The routledge handbook of phonological theory (pp. 476\u2013503). Routledge.https:\/\/doi.org\/10.4324\/9781315675428-17Archangeli, D., & Pulleyblank, D. (2021). Emergent phonology. Language Science Press.Aunio, L. (2010). Ikoma nominal tone. Africana Linguistica, 16, 3\u201330.https:\/\/doi.org\/10.3406\/aflin.2010.986Aunio, L. (2013). Ikoma verbal tone. Nordic Journal of African Studies, 4, 274\u2013321.Aunio, L. (2015). A typological perspective on Bantu nominal tone: The case of Ikoma-Nata-Isenye inwestern Tanzania. Southern African Linguistics and Applied Language Studies, 33(3), 359\u2013371.https:\/\/doi.org\/10.2989\/16073614.2015.1105106Aunio, L. (2017). Syllable weight and tone in Mara Bantu languages. In P. Newman (Ed.), Syllable weightin African languages (pp. 191\u2013214). John Benjamins Publishing Company.Aunio, L., Robinson, H., Roth, T., Stegen, O., & Walker, J. B. (2019). The Mara languages JE40. In M. Vande Velde, K. Bostoen, D. Nurse, & G. Philippson (Eds.), The Bantu languages (pp. 501\u2013532).Routledge. https:\/\/doi.org\/10.4324\/9781315755946-17Bakovic\u00b4, E. (2000). Harmony, dominance, and control (Doctoral dissertation). Rutgers University.Boeckx, C., & Hornstein, N. (2010). The varying aims of linguistic theory. In J. Franck & J. Bricmont(Eds.), Chomsky notebook (pp. 115\u2013141). Columbia University Press.https:\/\/doi.org\/10.7312\/bric14474-006Bonet, E., & Harbour, D. (2012). Contextual allomorphy. In J. Trommer (Ed.), The morphology andphonology of exponence (pp. 195\u2013235). Oxford University Press.https:\/\/doi.org\/10.1093\/acprof:oso\/9780199573721.003.0007Carstens, V. (2008). DP in Bantu and Romance. In K. Demuth & C. De Cat (Eds.), The Bantu-Romanceconnection (pp. 131\u2013166). John Benjamins. https:\/\/doi.org\/10.1075\/la.131.10car156Casali, R. (2003). [ATR] value asymmetries and underlying vowel inventory structure in Niger-Congo andNilo-Saharan. Linguistic Typology, 7, 307\u2013382. https:\/\/doi.org\/10.1515\/lity.2003.018Cassimjee, F., & Kisseberth, C. W. (1998). Optimal domains theory and Bantu tonology: A case study fromIsixhosa and Shingazidja. In L. M. Hyman & C. Kisseberth (Eds.), Theoretical aspects of Bantutone (pp. 33\u2013132). CSLI.Chomsky, N. (1995). Bare phrase structure. In G. Webelhuth (Ed.), Government and binding theory and theminimalist program (pp. 385\u2013439). Blackwell.Chomsky, N. (2001). Beyond explanatory adequacy. MIT occasional papers in linguistics 20. MITWorking Papers in Linguistics.Choti, J. N. (2015). Phonological (a)symmetries of nasal prefixes in Bantu (Doctoral dissertation).Michigan State University.Davy, J., & Nurse, D. (1982). Synchronic versions of Dahl\u2019s law: The multiple applications of aphonological dissimilation rule. Journal of African Languages and Linguistics, 4(2), 157\u2013196.https:\/\/doi.org\/10.1515\/jall.1982.4.2.157De\u00b4chaine, R.-M., Mudzingwa, C., Girard, R., & Wiltschko, M. (2014). The internal syntax of Shona classprefixes. Language Sciences, 43(18). https:\/\/doi.org\/10.1016\/j.langsci.2013.10.008De\u00b4chaine, R.-M., Si, D., & Gambarage, J. (2017). Nata deverbal nominalizations. In J. Kandybowicz &H. Torrence (Eds.), Africa\u2019s endangered languages. Oxford University Press.https:\/\/doi.org\/10.1093\/oso\/9780190256340.003.0006deGorog, R. P., de Saussure, F., & Baskin, W. (1959). Course in general linguistics(A. R. Charles Bally Albert Sechehaye, Trans.; Vol. 45). Wiley. https:\/\/doi.org\/10.2307\/320587Dobashi, Y. (2003). Phonological phrasing & syntactic derivation (Doctoral dissertation). CornellUniversity.Downing, L. J. (1994). SiSwati verbal reduplication and the theory of generalized alignment. InM. Gonzalez (Ed.), Proceedings of the 24th meeting of the north-eastern linguistics society(pp. 81\u201395).Downing, L. J. (1998). On the prosodic misalignment of onsetless syllables. Natural Language &Linguistic Theory, 16(1), 1\u201352.Downing, L. J. (1999). Prosodic stem \u0338= prosodic word in Bantu. In A. Hall & U. Kleinhenz (Eds.), Studieson the Phonological Word (pp. 73\u201398). John Benamins. https:\/\/doi.org\/10.1075\/cilt.174.05dowDowning, L. J. (2005a). Morphological complexity and prosodic minimality. Catalan Journal ofLinguistics, 4, 83\u2013106. https:\/\/doi.org\/10.5565\/rev\/catjl.114Downing, L. J. (2006). Canonical forms in Prosodic Morphology. Oxford University Press.https:\/\/doi.org\/10.1093\/acprof:oso\/9780199286393.001.0001Downing, L. J., & Kadenge, M. (2020). Re-placing PStem in the prosodic hierarchy. The LinguisticReview, 37(3), 433\u2013461. https:\/\/doi.org\/https:\/\/doi.org\/10.1515\/tlr-2020-2050Downing, L. J. (1990). Problems in Jita tonology (Doctoral dissertation). University of Illinois atUrbana-Champaign.Downing, L. J. (2005b). On the ambiguous segmental status of nasals in homorganic NC sequences. InM. van Oostendorp & J. van de Weijer (Eds.), The internal organization of phonological segments(pp. 183\u2013216). de Gruyter.Dresher, B. E. (2009). The contrastive hierarchy in phonology. Cambridge University Press.https:\/\/doi.org\/10.1017\/cbo9780511642005Dresher, B. E. (2011). The phoneme. In M. van Oostendorp, C. J. Ewen, E. Hume, & K. Rice (Eds.), TheBlackwell companion to phonology (pp. 241\u2013266). Wiley-Blackwell.https:\/\/doi.org\/10.1002\/9781444335262.wbctp0011Elfner, E. (2012). Syntax-prosody interactions in Irish (Doctoral dissertation). University of Massachusets,Amherst.157Embick, D. (2010). Localism versus globalism in morphology and phonology. The MIT Press.https:\/\/doi.org\/10.7551\/mitpress\/9780262014229.001.0001Embick, D., & Marantz, A. (2008). Architecture and blocking. Linguistic Inquiry, 39(1), 1\u201353.https:\/\/doi.org\/10.1162\/ling.2008.39.1.1Fusheini, H. (2014). [ATR] feature involves a distinct tongue root articulation: Evidence from ultrasoundimaging. Lingua, 143, 36\u201351. https:\/\/doi.org\/10.1016\/j.lingua.2013.12.009Gambarage, J. (2012). Context-of-use of augmented and unaugmented nouns in Nata [UBC qualifyingpaper]. UBCWPL, 34, 45\u201359.Gambarage, J. (2013). Tongue root restriction and nominal morphological domains in Nata [UBCqualifying paper].Gambarage, J. (2019). Weak determiners in Bantu: Evidence from the syntax and semantics of pre-prefxesin Nata (Doctoral dissertation). University of British Columbia.Gambarage, J., & Pulleyblank, D. (2017). Tongue root harmony in Nata: An allomorphy-based account. InJ. Kandybowicz & H. Torrence (Eds.), Africa\u2019s endangered languages: Documentary andtheoretical approaches. Oxford University Press.Goldsmith, J. (1975). Tone melodies and the autosegment. OSU WPL 20: proceedings of the sixthconference on African linguistics (pp. 135\u2013147).Goldsmith, J. A. (1976). Autosegmental phonology (Doctoral dissertation). MIT.Gregersen, E. A., & Welmers, W. E. (1974). African language structures (Vol. 53). University of CalifroniaPress. https:\/\/doi.org\/10.2307\/413074Halle, M., & Stevens, K. (1969). On the feature \u201cadvanced tongue root\u201d. Quarterly progress report (MITresearch laboratory of electronics) 94 (pp. 209\u2013215).Heath, J. (2008). A grammar of Jamsay. Mouton de Gruyter. https:\/\/doi.org\/10.1515\/9783110207224Hess, S. (1992). Assimilatory effects in a vowel harmony system: An acoustic analysis of advanced tongueroot in Akan. Journal of Phonetics, 20, 475\u2013492. https:\/\/doi.org\/10.1016\/s0095-4470(19)30651-5Higgins, H. A. (2012). Ikoma vowel harmony: Phonetics and phonology. SIL E-books 43.Hill, D., Lindfors, A.-L., Nagler, L., Woodward, M., & Yalonde, R. (2007). A sociolinguistic survey of theBantu languages in Mara region. SIL.Hsu, B. (2016). Syntax-prosody interaction in the clausal domain: Head movement and coalescense(Doctoral dissertation). University of Southern California.Hyman, L., & Biloa, E. (1992). Transparent low tone in Tuki. Proceedings of the Eighteenth AnnualMeeting of the Berkeley Linguistics Society: Special Session on The Typology of Tone Languages,104\u2013127. https:\/\/doi.org\/10.3765\/bls.v18i2.1547Hyman, L. M. (1985). A theory of phonological weight. Foris. https:\/\/doi.org\/10.1515\/9783110854794Hyman, L. M. (2006). Word-prosodic typology. Phonology, 23, 225\u2013257.https:\/\/doi.org\/10.1017\/s0952675706000893Hyman, L. M. (2009a). How (not) to do phonological typology: The case of pitch-accent. LanguageSciences, 31, 213\u2013238. https:\/\/doi.org\/10.1016\/j.langsci.2008.12.007Hyman, L. M. (2009b). The natural history of verb-stem reduplication in Bantu. Morphology.https:\/\/doi.org\/10.1007\/s11525-009-9140-yHyman, L. M. (2011). Tone: Is it different? In J. A. Goldsmith, J. Riggle, & A. C. L. Yu (Eds.), Thehandbook of phonological theory (pp. 197\u2013239). Wiley-Blackwell.https:\/\/doi.org\/10.1002\/9781444343069.ch7Hyman, L. M., & Leben, W. R. (2001). Tone systems. In C. Gussenhoven & A. Chen (Eds.), Languagetypology and language universals: An international handbook. vol. 2. (pp. 44\u201365). OxfordUniversity Press. https:\/\/doi.org\/10.1093\/oxfordhb\/9780198832232.013.6Irurtzun, A. (2009). Why Y: On the centrality of syntax in the architecture of grammar. Catalan Journal ofLinguistics, 8, 141\u2013160.158Ito\u02c6, J., & Mester, A. (2009). The extended prosodic word. In J. Grijzenhout & B. Kabak (Eds.),Phonological domains: Universals and deviations. Mouton de Gruyter.https:\/\/doi.org\/10.1515\/9783110219234.2.135Ito\u02c6, J., & Mester, A. (2012). Recursive prosodic phrasing in Japanese. In T. Borowsky, S.Kawahara,T. Shinya, & M. Sugahara (Eds.), Prosody matters: Essays in honor of Elisabeth Selkirk. Equinox.Ito\u02c6, J., & Mester, A. (2013). Prosodic subcategories in Japanese. Lingua, 124, 20\u201340.https:\/\/doi.org\/10.1016\/j.lingua.2012.08.016Jackendoff, R. (1997). The architecture of the language faculty. MIT Press. https:\/\/doi.org\/10.2307\/417010Jacobson, L. C. (1980). Voice-quality harmony in Western Nilotic languages. In R. Vago (Ed.), Studies inlanguage companion series (p. 183). John Benjamins Publishing Company.https:\/\/doi.org\/10.1075\/slcs.6.10jacJohannes, J. G. (2007). The Ki-Nata noun structure (Master\u2019s thesis). University of Dar es Salaam.Kanerva, J. M. (1990). Focus and phrasing in Chichew\u02c6a phonology (Doctoral dissertation). StanfordUniversity.Kramer, R. (2015). The morphosyntax of gender. Oxford University Press.https:\/\/doi.org\/10.1093\/acprof:oso\/9780199679935.001.0001Ladefoged, P. (1964). A phonetic study of west african languages. University Press.Lafon, M. (1994). Shona cl. 5 revisited: A case against ri- as a class 5 nominal prefix. Journal of theUniversity of Zimbabwe, 51\u201380.Lam, Z. (2013). Nata verbal tonology [UBC QP].Leben, W. (1978). The representation of tone. In V. A. Fromkin (Ed.), Tone: A linguistic survey(pp. 177\u2013219). Academic Press. https:\/\/doi.org\/10.1016\/b978-0-12-267350-4.50011-xLindau, M. (1974). The feature advanced tongue root. In E. Voeltz (Ed.), Third annual conference onAfrican linguistics.Lindau, M. (1978). Vowel features. Language, 54(3), 541\u2013563. https:\/\/doi.org\/10.2307\/412786Lombardi, L. (1995). Dahl\u2019s law and privative [voice]. Linguistic Inquiry, 26, 365\u2013372.Maho, J. F. (1999). A comparative study of Bantu noun classes (Doctoral dissertation). Go\u00a8teborgUniversity.Maho, J. F. (2003). A classification of the Bantu languages: An update of Guthrie\u2019s referential system. InD. Nurse & G. Philippson (Eds.), The Bantu languages (pp. 639\u2013651). Routledge.Maho, J. F. (2009). The online version of the New Updated Guthrie List, a referential classification of theBantu languages [Acessed March, 27 2017]. http:\/\/goto.glocalnet.net\/mahopapers\/nuglonline.pdfMarantz, A. (1997). No escape from syntax: Don\u2019t try morphological analysis in the privacy of your ownlexicon. In A. Dimitriadis (Ed.), Proceedings of the 21st annual penn linguistics colloquium: Pennworking papers in linguistics (pp. 201\u2013225).Marlo, M. (2013). Verb tone in Bantu languages: Micro-typological patterns and research methods.Africana Linguistica, 19, 137\u2013234. https:\/\/doi.org\/10.3406\/aflin.2013.1020Marvin, T. (2003). Topics in the stress and syntax of words (Doctoral dissertation). Massachusetts Instituteof Technology.Mascaro\u00b4, J. (2007). External allomorphy and lexical representation. Linguistic Inquiry, 38(4), 715\u2013735.https:\/\/doi.org\/10.1162\/ling.2007.38.4.715McCarthy, J. J. (2003). OT constraints are categorical. Linguistics Department Faculty Publication Series,52. https:\/\/doi.org\/10.1017\/s0952675703004470McCarthy, J. J., & Prince, A. A. (1995). Faithfulness and reduplicative identity. University ofmassachusetts occasional papers 18: Papers in optimality theory (249\u2013384). GLSA.http:\/\/roa.rutgers.edu\/files\/60-0000\/60-0000-MCCARTHY-0-0.PDFMcCarthy, J. J., & Prince, A. S. (1993). Generalized alignment. In G. Booij & J. van Marle (Eds.), Yearbookof morphology (pp. 79\u2013153). Springer Netherlands. https:\/\/doi.org\/10.1007\/978-94-017-3712-8 4159McPherson, L. (2019). Seenku argument-head tone sandhi: Allomorph selection in a cyclic grammar.Glossa, 4(22), 1\u201341. https:\/\/doi.org\/10.5334\/gjgl.798Mekacha, R. K. (1993). The sociolinguistic impact of Swahili on ethnic community languages in Tanzania:A case study of Ekinata. [University of Bayreuth, Bayreuth].Mudzingwa, C. (2010). Shona morphophonemics: Repair strategies in karanga and zezuru(Doctoral dissertation). University of British Columbia.Muzale, H. R. T., & Rugemalira, J. M. (2008). Researching and documenting the languages of Tanzania.Language Documentation & Conservation, 1, 68\u2013108.Myers, S. P. (1987). Tone and the structure of words in shona (Doctoral dissertation). University ofMassachusetts Amherst.Nespor, M., & Vogel, I. (1986). Prosodic phonology. Foris. https:\/\/doi.org\/10.1515\/9783110977790Newell, H. (2008). Aspects of the morphology and phonology of phases (Doctoral dissertation). McGillUniversity.Nurse, D., & Philippson, G. (1975). Tanzania language survey (TLS). Retrieved August 26, 2019, fromhttp:\/\/www.cbold.ish-lyon.cnrs.fr\/Odden, D. (1994). Adjacency parameters in phonology. Language, 70, 289\u2013330.https:\/\/doi.org\/10.2307\/415830Pak, M. (2008). The postsyntactic derivation and its phonological reflexes (Doctoral dissertation).University of Pennsylvania.Paster, M. (2009). Explaining phonological conditions on affixation: Evidence from suppletive allomorphsand affix ordering. Word Structure. https:\/\/doi.org\/10.3366\/e1750124509000282Paster, M. (2015). Phonologically conditioned suppletive allomorphy: Cross-linguistic resultsandtheoretical consequences. In E. Bonet, M.-R. Lloret, & J. Mascaro\u00b4 (Eds.), Understandingallomorphy : Perspectives from optimality theory (pp. 218\u2013253). Equinox.Paster, M. E. (2006). Phonological conditions on affixation (Doctoral dissertation). University ofCalifornia, Berkeley.Poletto, R. E. (1998). Topics in Runyankore phonology (Doctoral dissertation). Ohio State University.Prince, A., & Smolensky, P. (1993\/2004). Optimality theory: Constraint interaction in generative grammar(tech. rep.) [Technical Report, Rutgers University and University of Colorado at Boulder, 1993.Revised version published by Blackwell, 2004.].Ross, B. B. (2011). Prosody and grammar in Dalabon and Kayardild (Doctoral dissertation). TheUniversity of Melbourne.Sande, H. L. (2017). Distributing morphologically conditioned phonology: Three case studies from Gue\u00b4bie(Doctoral dissertation). University of California, Berkeley.Sandy, C. S. (2017). Prosodic prominence in Karuk (Doctoral dissertation). University of California,Berkeley.Sapir, E. (1925). Sound patterns of language. In P. Swiggers (Ed.), The collected works of edward sapir(pp. 179\u2013194). De Gruyter Mouton. https:\/\/doi.org\/10.2307\/409004Schadeberg, T. C. (1973). Kinga: A restricted tone language. Studies in African Linguistics, 4, 23\u201348.Selkirk, E. (2011). The syntax-phonology interface. In J. Goldsmith, J. Riggle, & A. C. L. Yu (Eds.), Thehandbook of phonological theory (Second, pp. 435\u2013484). Blackwell Publishing.https:\/\/doi.org\/10.1002\/9781444343069.ch14Selkirk, E. (1986). On derived domains in sentence phonology. Phonology yearbook 3 (pp. 371\u2013405).https:\/\/doi.org\/10.1017\/s0952675700000695Shetler, J. B. (2007). Imagining Serengeti: A history of landscape memory in Tanzania from earliest timesto the present. Ohio University Press.Svenonius, P. (2006). Interpreting uninterpretable features [lingbuzz\/000333]. Linguistic Analysis.160Tiede, M. K. (1996). An MRI-based study of pharyngeal volume contrasts in akan and english. Journal ofPhonetics, 24, 3990421. https:\/\/doi.org\/10.1006\/jpho.1996.0022Van de Velde, M. (2013). The Bantu connective construction. In A. Carlier & J.-C. Verstraete (Eds.), Thegenitive (pp. 217\u2013252). John Benjamins. https:\/\/doi.org\/10.1075\/cagral.5.08velVaris, E. E. (2012). The Spanish femenine el at the syntax-phonology interface (Doctoral dissertation).University of Southern California.S\u02c7urkalovic\u00b4, D. (2015). The no-reference hypothesis (Doctoral dissertation). University of Troms\u00f8.Wang, C.-M. B. (2014). The prosody-syntax interaction in the \u201cyi-bu-qi-ba\u201d rule: A morphologicallyconditioned tone change in Mandarin Chinese (Doctoral dissertation). University of NorthCarolina at Chapel Hill.Welmers, W. E. (1959). Tonemics, morphotonemics, and tonal morphemes. General Lingustics, 4, 1\u20139.Yip, M. (2001). Tonal features, tonal inventories and phonetic targets (C. Iten & A. Neeleman, Eds.).University College London Working Papers in Linguistics, 13, 161\u2013188.Yip, M. (2004). Phonological markedness and allomorph selection in Zahao. Language and Linguistics,5(4), 969\u20131001.Zoll, C. (1996). Parsing below the segment in a constraint-based framework (Doctoral dissertation).University of California Berkeley.161Appendix AVowel hiatusIn this appendix I will examine hiatus between vowels within the nominal prefix domain. The effect ofvowel hiatus is to prohibit syllables from surfacing with a sequence of distinct vowel features, such as [iu]or [ei]. In Nata, there are no such observed sequences; however, morphology leaves the possibility open ifno intervening phonotactic penalizes non-identical vowel sequences.Vowel hiatus occurs when a morpheme with a vowel final allomorph ([. . . V]) is followed by a morphemewith a vowel initial allomorph ([V. . . ]). In such cases, the output would be a sequence of potentially non-identical vowels. In Nata, surface forms only exist long vowels of a single quality. In cases where V1and V2 are not identical, coalescence occurs, resulting in one set of vowel features taking over both timingslots. In Nata, both V1V2\u2192V1V1 and V1V2\u2192V2V2 are attested; however, V1V2\u2192V2V2 appears to be morecommon, as we shall see.This area of research is also known as hiatus resolution, or vowel elision, depending on which part ofthe metaphor is being focused on: the problematic sequence (hiatus), or the repair (coalescence, elision).A.1 Prepositional prefixesThe associative and comitative prefixes generally have a CV- shape, and occur preceding the determiner,which generally has a V- shape; thus, forms with a preposition and overt determiner could potentially sur-face as \u03c6 with a sequence of two adjacent but non-identical vowels. However, such sequences are whollyunattested in Nata.In order to examine coalesce, we will compare determiners on nouns without any prepositions to thosesame nouns with the class 2 associative prefix and with the comitative prefix. We begin with the forms162without prepositions.Table A.1: Determiner vowel qualitye-Be\u00b4-tore \u2018cucumbers.c8\u2019o-mo\u00b4-sikera \u2018someone who enters.c1\u2019 (derived)o-Bu-sO\u00b4oh-u \u2018greediness.c14\u2019 (derived)o-mu-tE\u00b4rEbi \u2018wooden ladle.c3\u2019o-mw-iika\u00b4 \u2018pressure\/gas.c3\u2019u-mw-a\u00b4ana \u2018child.c1\u2019u-mw-EEr\u00b4i \u2018moon.c3\u2019i-kj-O\u00b4OndE \u2018honey badger.c7\u2019In these forms, the determiner is either a mid advanced or high advanced vowel, depending on the[\u00b1ATR] value of the following vowel.Recall that the c2 associative surfaces as [Ba-] before a consonant. Similarly, the comitative surfaces as[na-] in such contexts, illustrated with a form with a c10 determiner, and a form with a phonologically nulldeterminer.Table A.2: CV- comitativena-tSaa- \/0-m-Bara\u00b4hE \u2018. . . and Thompson\u2019s Gazelles.c10\u2019na- \/0-ki-GE\u00b4rO \u2018. . . and not any thing.c7\u2019We will now consider the nouns above with prepositions.163Table A.3: Associative vowel qualitya-Ba-ka\u00b4ri Be-e-Be\u00b4-tore \u2018women.c2 of cucumbers.c8\u2019a-Ba-ka\u00b4ri Bo-o-mo\u00b4-sikera \u2018women.c2 of someone who enters.c1\u2019a-Ba-ka\u00b4ri Bo-o-Bu\u00b4-sO\u00b4Oh-u \u2018women.c2 of greediness.c14\u2019a-Ba-ka\u00b4ri Bo-o-mu-tE\u00b4rEbi \u2018women.c2 of wooden ladle.c3\u2019a-Ba-ka\u00b4ri Bo-o-mw-iika\u00b4 \u2018women.c2 of pressure\/gas.c3\u2019a-Ba-ka\u00b4ri Bu\u00b4-u\u00b4-mw-aana \u2018women.c2 of child.c1\u2019a-Ba-ka\u00b4ri Bu-u-mw-EEr\u00b4i \u2018women.c2 of moon.c3\u2019a-Ba-ka\u00b4ri Bi-i-kj-O\u00b4OndE \u2018women.c2 of honey badger.c7\u2019Table A.4: Comitative vowel qualitya-Ba-ka\u00b4ri ne-e-Be\u00b4-tore \u2018women.c2 and cucumbers.c8\u2019a-Ba-ka\u00b4ri no-o-mo\u00b4-sikera \u2018women.c2 and someone who enters.c1\u2019a-Ba-ka\u00b4ri no-o-Bu\u00b4-sO\u00b4Oh-u \u2018women.c2 and greediness.c14\u2019a-Ba-ka\u00b4ri no-o-mu-tE\u00b4rEbi \u2018women.c2 and wooden ladle.c3\u2019a-Ba-ka\u00b4ri no-o-mw-iika\u00b4 \u2018women.c2 and pressure\/gas.c3\u2019a-Ba-ka\u00b4ri nu\u00b4-u\u00b4-mw-aana \u2018women.c2 and child.c1\u2019a-Ba-ka\u00b4ri nu-u-mw-EEr\u00b4i \u2018women.c2 and moon.c3\u2019a-Ba-ka\u00b4ri ni-i-kj-O\u00b4OndE \u2018women.c2 and honey badger.c7\u2019The important observation regarding forms with an associative or comitative prefix is that the quality ofthe determiner does not alternate depending on the absence or presence of the prepositional prefix. In termsof the coalescence patterns mentioned earlier, this represents the V1V2\u2192V2V2 pattern.A phonotactic *HIATUS penalizes adjacent non-identical vowels. In addition, the same vowel qualityphonotactics we utilized in Chapter 5 account for which V-V sequence surfaces, much like the \u03c6 -HAR-MONY constraints. For the sake of simplicity, forms that violate harmony or tone phonotactics are omitted.164(124) Evaluation of coalescense[{na, no, nu, nO, ne, ni, nE, . . .}+{ o, u, O. . . }+{mo, mu, mO. . .}+{tE\u00b4rEbi . . . }] HIATUS *[-ATR] *[+HI]u a. [\u03c6no-o-mu-[\u03c9 tE\u00b4rEbib. [\u03c6na-o-mu-[\u03c9 tE\u00b4rEbi *!c. [\u03c6ne-o-mu-[\u03c9 tE\u00b4rEbi *!c. [\u03c6nu-u-mu-[\u03c9 tE\u00b4rEbi *!c. [\u03c6nO-O-mu-[\u03c9 tE\u00b4rEbi *!A.2 Locative prefixesThere are two additional prefixes that bear mentioning with respect to coalescence: the locatives [koo-] and[moo-]; \u2018on\u2019, and \u2018in\u2019 respectively. These forms differ from the associative and comitative in that the qualityof the vowel in the initial syllable is not determined by the determiner, but rather by the locative prefix. Interms of coalescence, they are the V1V2\u2192V1V1 pattern.We will begin by considering nouns without the locative prefixes.165Table A.5: Unprefixed nounso-mu-ka\u00b4ri \u2018woman.c1\u2019o-mo-s\u00b4isi \u2018tamarind tree.c3\u2019e-me-s\u00b4isi \u2018tamarind trees.c4\u2019e-Ge-ku\u00b4Ba \u2018chest.c7\u2019e-Be-ku\u00b4Ba \u2018chests.c8\u2019o-ro-s\u0131\u00b4ri \u2018on rope.c11\u2019a-Ba-ka\u00b4ri \u2018women.c2\u2019u-mw-EEr\u00b4i \u2018moon.c3\u2019a-ma-Buri \u2018feathers.c6\u2019i-kj-O\u00b4OndE \u2018village.c7\u2019a-tSaa-ntSoGu \u2018elephants.c10\u2019Compare the forms above to those with a locative prefix, below.Table A.6: Locative vowel qualitykoo-mu-ka\u00b4ri \u2018on woman.c1\u2019koo-mo-s\u00b4isi \u2018on tamarind tree.c3\u2019koo-me-s\u00b4isi \u2018on tamarind trees.c4\u2019koo-Ge-ku\u00b4Ba \u2018on chest.c7\u2019koo-Be-ku\u00b4Ba \u2018on chests.c8\u2019koo-ro-s\u0131\u00b4ri \u2018on rope.c11\u2019kuu-Ba-ka\u00b4ri \u2018on women.c2\u2019kuu-mw-EEr\u00b4i \u2018on moon.c3\u2019ku\u00b4u\u00b4-ma-Buri \u2018on feathers.c6\u2019kuu-kj-O\u00b4OndE \u2018on village.c7\u2019ku\u00b4-tSaa-ntSoGu \u2018on elephants.c10\u2019166Forms with the locative can never be interpreted as having a phonological null determiner. The deter-miner still encodes existence in the way determiners have been found to (Gambarage, 2019).The problem with analyzing such a case is that determiners for class 7, 8, and others, do not generallyhave back vowel allomorphs; they are distinct from c1, 3, 4, and others in that they occur with front vowels.Therefore, these allomorphs must be limited to just the post-locative position. The other option is that thereis a null allomorph of the determiner that still encodes existence, distinct from the null determiner whichsignifies a non-existential entity.167Appendix BTone Evaluation TableauxB.1 Low tone classB.1.1 General(125) Evaluation of L-tone roots with overt determiner[DP{e, e\u00b4}L+[NumP{Ge, Ge\u00b4}+[n{seku}]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03c6 eL-Ge-[\u03c9seku]] *! *u b. [\u03c6 eL-Ge\u00b4-[\u03c9seku]] * * * *c. [\u03c6 e\u00b4L-Ge\u00b4-[\u03c9seku]] *! * * * *d. [\u03c6 e\u00b4L-Ge-[\u03c9seku]] *! * *168(126) Evaluation of L-tone roots with null determiner[DP[NumP{Ge, Ge\u00b4}+[n{seku}]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03c6Ge-[\u03c9seku]] *! * *u b. [\u03c6Ge\u00b4-[\u03c9seku]] * *(127) Evaluation of associative L-tone root with overt determiner[PP{kjoH, kjo\u00b4}+[DP{o, o\u00b4}L+[NumP{mo, mo\u00b4}+[nP{rem}+{i}}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6kjoH-oL-mo-[\u03c9 remi]] *! * * *b. [\u03b9 [\u03c6kjoH-oL-mo\u00b4-[\u03c9 remi]] *! * * *c. [\u03b9 [\u03c6kjoH-o\u00b4L-mo\u00b4-[\u03c9 remi]] *! * * * * *d. [\u03b9 [\u03c6kjo\u00b4-oL-mo\u00b4-[\u03c9 remi]] *! * * * *e. [\u03b9 [\u03c6kjo\u00b4-o\u00b4L-mo\u00b4-[\u03c9 remi]] *! * * *f. [\u03b9 [\u03c6kjo-o\u00b4L-mo-[\u03c9 remi]] *! * * *g. [\u03b9 [\u03c6kjo\u00b4-oL-mo-[\u03c9 remi]] *! * *u h. [\u03b9 [\u03c6kjo\u00b4-o\u00b4-mo-[\u03c9 remi]] * *169(128) Evaluation of associative L-tone root with null determiner[PP{keH, ke\u00b4}+[DP[NumP{mo, mo\u00b4}+[nP{rem}+{i}}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6ke\u00b4-mo-[\u03c9 remi]] * *b. [\u03b9 [\u03c6keH-mo\u00b4-[\u03c9 remi]] * * *!c. [\u03b9 [\u03c6ke\u00b4-mo\u00b4-[\u03c9 remi]] *! * *d. [\u03b9 [\u03c6keH-mo-[\u03c9 remi ]] *! * * *B.1.2 Class 5(129) Evaluation of L-tone c5 root with overt determiner[DP{ri, r\u0131\u00b4}L+[NumP{i, \u0131\u00b4}+[nP{Buri}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 riL-i-[\u03c9Buri]] *! * *b. [\u03b9 [\u03c6 riL-\u0131\u00b4-[\u03c9Buri]] *! * * *u c. [\u03b9 [\u03c6 r\u0131\u00b4L-\u0131\u00b4-[\u03c9Buri ]] * *d. [\u03b9 [\u03c6 r\u0131\u00b4L-i-[\u03c9Buri]] *! * *(130) Evaluation of L-tone c5 root with null determiner[DP[NumP{i, \u0131\u00b4}+[nP{Buri}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 i-[\u03c9Buri]] *! * *u b. [\u03b9 [\u03c6 \u0131\u00b4-[\u03c9Buri]] * * *170(131) Evaluation of associative L-tone c5 root with overt determiner[PP{woH, wo\u00b4}+[DP{ri, r\u0131\u00b4}L+[NumP{i, \u0131\u00b4}+[nP{rem}+{i}}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6woH-riL-i-[\u03c9 remi]] *! * * *b. [\u03b9 [\u03c6woH-r\u0131\u00b4L-i-[\u03c9 remi]] *! * * *u c. [\u03b9 [\u03c6wo\u00b4-riL-i-[\u03c9 remi]] * *d. [\u03b9 [\u03c6wo\u00b4-r\u0131\u00b4L-i-[\u03c9 remi]] *! * * *e. [\u03b9 [\u03c6woH-riL-\u0131\u00b4-[\u03c9 remi]] *! * * *g. [\u03b9 [\u03c6woH-r\u0131\u00b4L-\u0131\u00b4-[\u03c9 remi]] * * *!h. [\u03b9 [\u03c6wo\u00b4-riL-\u0131\u00b4-[\u03c9 remi]] *! * * *i. [\u03b9 [\u03c6wo\u00b4-r\u0131\u00b4L-\u0131\u00b4-[\u03c9 remi]] *! * *(132) Evaluation of associative L-tone c5 root with null determiner[PP{wiH, w\u0131\u00b4}+[DP[NumP{i, \u0131\u00b4}+[nP{rem}+{i}}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6wiH-i-[\u03c9 remi]] *! * * *b. [\u03b9 [\u03c6w\u0131\u00b4-i-[\u03c9 remi]] *! * *c. [\u03b9 [\u03c6wiH-\u0131\u00b4-[\u03c9 remi]] *! * * * *u d. [\u03b9 [\u03c6w\u0131\u00b4-\u0131\u00b4-[\u03c9 remi]] * *171B.1.3 Class 9\/10(133) Evaluation of L-tone c9 root with overt determiner[DP{a, a\u00b4}L+[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 aL-mL-[\u03c9Baata]] *! * *b. [\u03b9 [\u03c6 a\u00b4L-mL-[\u03c9Baata]] *! * *u c. [\u03b9 [\u03c6 aL-mL-[\u03c9Ba\u00b4atac9]] * *d. [\u03b9 [\u03c6 a\u00b4L-mL-[\u03c9Ba\u00b4atac9]] *! * *(134) Evaluation of L-tone c9 root with null determiner[DP+[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6mL-[\u03c9Baata]] *! *u b. [\u03b9 [\u03c6mL-[\u03c9Ba\u00b4atac9]] * *172(135) Evaluation of L-tone c10 root with overt determiner[DP{tSaa, tSa\u00b4a\u00b4, tSa\u00b4a}L+[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 tSaaL-mL-[\u03c9Baata]] *! * *u b. [\u03b9 [\u03c6 tSa\u00b4a\u00b4L-mL-[\u03c9Baata]] * *c. [\u03b9 [\u03c6 tSa\u00b4aL-mL-[\u03c9Baata]] *! * *d. [\u03b9 [\u03c6 tSaaL-mL-[\u03c9Ba\u00b4atac9]] *! * * *e. [\u03b9 [\u03c6 tSa\u00b4a\u00b4L-mL-[\u03c9Ba\u00b4atac9]] *! * * *f. [\u03b9 [\u03c6 tSa\u00b4aL-mL-[\u03c9Ba\u00b4atac9]] *! * * * *(136) Evaluation of L-tone c10 root with null determiner[DP[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)\/ a. [\u03b9 [\u03c6mL-[\u03c9Baata]] *! * *u b. [\u03b9 [\u03c6mL-[\u03c9Ba\u00b4atac9]] * * *173(137) Evaluation of associative L-tone c9 root with overt determiner[PP{waH, wa\u00b4}+[DP{a, a\u00b4}L+[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6waH-aL-mL-[\u03c9Baata]] *! * * *b. [\u03b9 [\u03c6wa\u00b4-aL-mL-[\u03c9Baata]] *! * *c. [\u03b9 [\u03c6waH-a\u00b4L-mL-[\u03c9Baata]] *! * * * *u d. [\u03b9 [\u03c6wa\u00b4-a\u00b4L-mL-[\u03c9Baata]] * *e. [\u03b9 [\u03c6waH-aL-mL-[\u03c9Ba\u00b4atac9]] * *g. [\u03b9 [\u03c6wa\u00b4-aL-mL-[\u03c9Ba\u00b4atac9]] *! * *h. [\u03b9 [\u03c6waH-a\u00b4L-mL-[\u03c9Ba\u00b4atac9]] *! * * *i. [\u03b9 [\u03c6wa\u00b4-a\u00b4L-mL-[\u03c9Ba\u00b4atac9]] *! *(138) Evaluation of associative L-tone c9 root with null determiner[PP{wOH, wO\u00b4}+[DP[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6wOH-mL-[\u03c9Baata]] *! * * * *u b. [\u03b9 [\u03c6wO\u00b4-mL-[\u03c9Baata]] * * *c. [\u03b9 [\u03c6wOH-mL-[\u03c9Ba\u00b4atac9]] *! *d. [\u03b9 [\u03c6wO\u00b4-mL-[\u03c9Ba\u00b4atac9]] *! *174(139) Evaluation of associative L-tone c10 root with overt determiner[PP{wOH, wO\u00b4}+[DP{tSaa, tSa\u00b4a\u00b4}L+[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6wOH-tSaaL-mL-[\u03c9Baata]] *! * * *u b. [\u03b9 [\u03c6wO\u00b4-tSaaL-mL-[\u03c9Baata]] * *c. [\u03b9 [\u03c6wOH-tSa\u00b4a\u00b4L-mL-[\u03c9Baata]] * * *!d. [\u03b9 [\u03c6wO\u00b4-tSa\u00b4a\u00b4L-mL-[\u03c9Baata]] *! * *e. [\u03b9 [\u03c6wOH-tSaaL-mL-[\u03c9Ba\u00b4atac9]] *! * * * *g. [\u03b9 [\u03c6wO\u00b4-tSaaL-mL-[\u03c9Ba\u00b4atac9]] *! * * *h. [\u03b9 [\u03c6wOH-tSa\u00b4a\u00b4L-mL-[\u03c9Ba\u00b4atac9]] *! * * * *i. [\u03b9 [\u03c6wO\u00b4-tSa\u00b4a\u00b4L-mL-[\u03c9Ba\u00b4atac9]] *! * * *(140) Evaluation of associative L-tone c10 root with null determiner[PP{wOH, w\u00b4O}+[DP[NumP{N, \/0}L+[nP{Baata, Ba\u00b4atac9}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6wOH-mL-[\u03c9Baata]] *! * * *u b. [\u03b9 [\u03c6wO\u00b4-mL-[\u03c9Baata]] * *c. [\u03b9 [\u03c6wOH-mL-[\u03c9Ba\u00b4atac9]] *! * *d. [\u03b9 [\u03c6wO\u00b4-mL-[\u03c9Ba\u00b4atac9]] *! * *175B.2 Final H tone class(141) Evaluation of final H tone root with overt determiner[DP{e, e\u00b4}L+[NumP{ Gi, Ge\u00b4}+[n{teete\u00b4}]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03c6 eL-Gi-[\u03c9 teete\u00b4]] * *b. [\u03c6 eL-Ge\u00b4-[\u03c9 teete\u00b4]] *! * * *c. [\u03c6 e\u00b4L-Ge\u00b4-[\u03c9 teete\u00b4]] *! * * *d. [\u03c6 e\u00b4L-Ge-[\u03c9 teete\u00b4]] *! * *(142) Evaluation of final H tone root with null determiner[DP[NumP{ Gi, G\u0131\u00b4}+[n{teete\u00b4}]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03c6Ge-[\u03c9 teete\u00b4]] * *b. [\u03c6Ge\u00b4-[\u03c9 teete\u00b4]] *! *176(143) Evaluation of associative final H-tone root with overt determiner[PP{kjoH, kjo\u00b4, }+[DP{o, o\u00b4}L+[NumP{mo, mo\u00b4}+[nP{sirikare\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6kjoH-oL-mo-[\u03c9sirikare\u00b4]] * * *b. [\u03b9 [\u03c6kjoH-oL-mo\u00b4-[\u03c9sirikare\u00b4]] *! * * *c. [\u03b9 [\u03c6kjoH-o\u00b4L-mo-[\u03c9sirikare\u00b4]] *! * * *d. [\u03b9 [\u03c6kjo\u00b4-oL-mo-[\u03c9sirikare\u00b4]] *! *e. [\u03b9 [\u03c6kjo\u00b4-o\u00b4L-mo-[\u03c9sirikare\u00b4]] *! *f. [\u03b9 [\u03c6kjo\u00b4-o\u00b4L-mo\u00b4-[\u03c9sirikare\u00b4]] *! * *g. [\u03b9 [\u03c6kjo\u00b4-oL-mo\u00b4-[\u03c9sirikare\u00b4]] *! * *h. [\u03b9 [\u03c6kjoH-o\u00b4L-mo\u00b4-[\u03c9sirikare\u00b4]] *! * * *(144) Evaluation of associative final H-tone root with null determiner[PP{keH, ke\u00b4, }+[DP[NumP{mo, mo\u00b4}+[nP{sirikare\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6keH-mo-[\u03c9sirikare\u00b4]] * * *b. [\u03b9 [\u03c6keH-mo\u00b4-[\u03c9sirikare\u00b4]] *! * *c. [\u03b9 [\u03c6ke\u00b4-mo-[\u03c9sirikare\u00b4]] *! *d. [\u03b9 [\u03c6ke\u00b4-mo\u00b4-[\u03c9sirikare\u00b4]] *! *177B.2.1 Class 5(145) Evaluation of Final H-tone c5 root with overt determiner[DP{ri, r\u0131\u00b4}L+[NumP{i, \u0131\u00b4}+[nP{Birika\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6 ri-i-[\u03c9Birika\u00b4]] * *b. [\u03b9 [\u03c6 riL-\u0131\u00b4-[\u03c9Birika\u00b4]] *! * *c. [\u03b9 [\u03c6 r\u0131\u00b4L-\u0131\u00b4-[\u03c9Birika\u00b4]] *! *d. [\u03b9 [\u03c6 r\u0131\u00b4L-i-[\u03c9Birika\u00b4]] *! * *(146) Evaluation of Final H-tone c5 root with null determiner[DP[NumP{i, \u0131\u00b4}+[nP{Birika\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6 i-[\u03c9Birika\u00b4]] * *b. [\u03b9 [\u03c6 \u0131\u00b4-[\u03c9Birika\u00b4]] *! * *178(147) Evaluation of associative final H-tone c5 root with overt determiner[PP{woH, wo\u00b4}+[DP{ri, r\u0131\u00b4}L+[NumP{i, \u0131\u00b4}+[nP{tooka\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6woH-riL-i-[\u03c9 tooka\u00b4]] * * *b. [\u03b9 [\u03c6woH-r\u0131\u00b4L-i-[\u03c9 tooka\u00b4]] *! * * *c. [\u03b9 [\u03c6woH-r\u0131\u00b4L-\u0131\u00b4-[\u03c9 tooka\u00b4]] *! * *d. [\u03b9 [\u03c6woH-riL-\u0131\u00b4-[\u03c9 tooka\u00b4]] *! * * *e. [\u03b9 [\u03c6wo\u00b4-riL-i-[\u03c9 tooka\u00b4]] *! *f. [\u03b9 [\u03c6wo\u00b4-r\u0131\u00b4L-i-[\u03c9 tooka\u00b4]] *! * *g. [\u03b9 [\u03c6wo\u00b4-r\u0131\u00b4L-\u0131\u00b4-[\u03c9 tooka\u00b4]] *! *h. [\u03b9 [\u03c6wo\u00b4-riL-\u0131\u00b4-[\u03c9 tooka\u00b4]] *! * *(148) Evaluation of associative final H-tone c5 root with null determiner[PP{wiH, w\u0131\u00b4}+[DP[NumP{i, \u0131\u00b4}+[nP{tooka\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6wiH-i-[\u03c9 tooka\u00b4]] * * *b. [\u03b9 [\u03c6wiH-\u0131\u00b4-[\u03c9 tooka\u00b4]] *! * * * *c. [\u03b9 [\u03c6w\u0131\u00b4-i-[\u03c9 tooka\u00b4]] *! * *d. [\u03b9 [\u03c6w\u0131\u00b4-\u0131\u00b4-[\u03c9 tooka\u00b4]] *! *179Class 9\/10(149) Evaluation of Final H-tone c9 root with overt determiner[DP{a, a\u00b4}L+[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6 aL-\u00f1L-[\u03c9akwaaha\u00b4]] * *b. [\u03b9 [\u03c6 a\u00b4L-\u00f1L-[\u03c9akwaaha\u00b4]] *! * *(150) Evaluation of Final H-tone c9 root with null determiner[DP[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6\u00f1L-[\u03c9akwaaha\u00b4]] * *180(151) Evaluation of Final H-tone c10 root with overt determiner[DP{tSaa, tSa\u00b4a\u00b4, tSa\u00b4a}+[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6 tSaa-\u00f1L-[\u03c9akwaaha\u00b4]] * *b. [\u03b9 [\u03c6 tSa\u00b4a\u00b4-\u00f1L-[\u03c9akwaaha\u00b4]] *! *c. [\u03b9 [\u03c6 tSa\u00b4a-\u00f1L-[\u03c9akwaaha\u00b4]] *! * *(152) Evaluation of Final H-tone c10 root with null determiner[DP[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6\u00f1L-[\u03c9akwaaha\u00b4]] * *(153) Evaluation of associative Final H-tone c9 root with overt determiner[PP{kjaH, kja\u00b4}+[DP{a, a\u00b4}L+[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6kjaH-aL-\u00f1L-[\u03c9akwaaha\u00b4]] * * *b. [\u03b9 [\u03c6kja\u00b4-aL-\u00f1L-[\u03c9akwaaha\u00b4]] *! * *c. [\u03b9 [\u03c6kjaH-a\u00b4L-\u00f1L-[\u03c9akwaaha\u00b4]] *! * * *d. [\u03b9 [\u03c6kja\u00b4-a\u00b4L-\u00f1L-[\u03c9akwaaha\u00b4]] *! *181(154) Evaluation of associative Final H-tone c9 root with null determiner[PP{kEH, kE\u00b4}+[DP[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6kEH-\u00f1L-[\u03c9akwaaha\u00b4]] * * *b. [\u03b9 [\u03c6kE\u00b4-\u00f1L-[\u03c9akwaaha\u00b4]] *! *(155) Evaluation of associative Final H-tone c10 root with overt determiner[PP{GEH, GE\u00b4}+[DP{tSaa, tSa\u00b4a\u00b4, tSa\u00b4a}L+[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6GEH-tSaaL-\u00f1L-[\u03c9akwaaha\u00b4]] * * *b. [\u03b9 [\u03c6GEH-tSa\u00b4a\u00b4L-\u00f1L-[\u03c9akwaaha\u00b4]] *! * *c. [\u03b9 [\u03c6GEH-tSa\u00b4aL-\u00f1L-[\u03c9akwaaha\u00b4]] *! * * *d. [\u03b9 [\u03c6GE\u00b4-tSaaL-\u00f1L-[\u03c9akwaaha\u00b4]] *! *e. [\u03b9 [\u03c6GE\u00b4-tSa\u00b4a\u00b4L-\u00f1L-[\u03c9akwaaha\u00b4]] *! *f. [\u03b9 [\u03c6GE\u00b4-tSa\u00b4aL-\u00f1L-[\u03c9akwaaha\u00b4]] *! * *(156) Evaluation of associative Final H-tone c10 root with null determiner[PP{kEH, kE\u00b4}+[DP[NumP{N, \/0}L+[nP{akwaaha\u00b4}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6kEH-\u00f1-[\u03c9akwaaha\u00b4]] * * *b. [\u03b9 [\u03c6kE\u00b4-\u00f1-[\u03c9akwaaha\u00b4]] *! *182B.3 Initial H tone classB.3.1 General(157) Evaluation of Initial H-tone root with overt determiner[DP{e, e\u00b4}L+[NumP{ Ge, Ge\u00b4}+[nP{s\u0131\u00b4ma, sima\u00b4c5, simaH }]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03c6 eL-Ge-[\u03c9simaH ]] *! * * *b. [\u03c6 eL-Ge-[\u03c9sima\u00b4c5]] * *! *u c. [\u03c6 eL-Ge-[\u03c9s\u0131\u00b4ma]] * *d. [\u03c6 eL-Ge\u00b4-[\u03c9simaH ]] *! * * *e. [\u03c6 eL-Ge\u00b4-[\u03c9sima\u00b4c5]] *! * * * *f. [\u03c6 eL-Ge\u00b4-[\u03c9s\u0131\u00b4ma]] *! * * *g. [\u03c6 e\u00b4L-Ge-[\u03c9simaH ]] *! * * *h. [\u03c6 e\u00b4L-Ge-[\u03c9sima\u00b4c5]] *! * * *i. [\u03c6 e\u00b4L-Ge-[\u03c9s\u0131\u00b4ma]] *! * *j. [\u03c6 e\u00b4L-Ge\u00b4-[\u03c9simaH ]] *! * * * *k. [\u03c6 e\u00b4L-Ge\u00b4-[\u03c9sima\u00b4c5]] *! * * * *l. [\u03c6 e\u00b4L-Ge\u00b4-[\u03c9s\u0131\u00b4ma]] *! * * *183(158) Evaluation of Initial H-tone root with null determiner[DP[NumP{ Ge, Ge\u00b4}+[nP{s\u0131\u00b4ma, sima\u00b4c5, simaH }]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03c6Ge-[\u03c9simaH ]] *! * * * *b. [\u03c6Ge-[\u03c9sima\u00b4c5]] * *! *u c. [\u03c6Ge-[\u03c9s\u0131\u00b4ma]] * *d. [\u03c6Ge\u00b4-[\u03c9simaH ]] * *!e. [\u03c6Ge\u00b4-[\u03c9sima\u00b4c5]] *! * *f. [\u03c6Ge\u00b4-[\u03c9s\u0131\u00b4ma]] *! *184(159) Evaluation of associative Initial H-tone root with overt determiner[PP{kjoH, kjo\u00b4}+[DP{o, o\u00b4}L+[NumP{mu, mu\u00b4}+[nP{ka\u00b4ri, kariH , kar\u0131\u00b4c5}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6kjoH-oL-mu-[\u03c9kariH ]] *! * * * *b. [\u03b9 [\u03c6kjoH-oL-mu-[\u03c9kar\u0131\u00b4c5]] * * *! *u c. [\u03b9 [\u03c6kjoH-oL-mu-[\u03c9ka\u00b4ri]] * * *d. [\u03b9 [\u03c6kjoH-oL-mu\u00b4-[\u03c9kariH ]] * *! *e. [\u03b9 [\u03c6kjoH-oL-mu\u00b4-[\u03c9kar\u0131\u00b4c5]] *! * * *f. [\u03b9 [\u03c6kjoH-oL-mu\u00b4-[\u03c9ka\u00b4ri]] *! * *g. [\u03b9 [\u03c6kjoH-o\u00b4L-mu-[\u03c9kariH ]] *! * * * * *h. [\u03b9 [\u03c6kjoH-o\u00b4L-mu-[\u03c9kar\u0131\u00b4c5]] *! * * * * *i. [\u03b9 [\u03c6kjoH-o\u00b4L-mu-[\u03c9ka\u00b4ri]] *! * * * *j. [\u03b9 [\u03c6kjo\u00b4\u2013oL-mu-[\u03c9kariH ]] *! * * *k. [\u03b9 [\u03c6kjo\u00b4\u2013oL-mu-[\u03c9kar\u0131\u00b4c5]] *! * * *l. [\u03b9 [\u03c6kjo\u00b4\u2013oL-mu-[\u03c9ka\u00b4ri]] *! * *m. [\u03b9 [\u03c6kjoH-o\u00b4L-mu\u00b4-[\u03c9kariH ]] *! * * * *n. [\u03b9 [\u03c6kjoH-o\u00b4L-mu\u00b4-[\u03c9kar\u0131\u00b4c5]] *! * * * *o. [\u03b9 [\u03c6kjoH-o\u00b4L-mu\u00b4-[\u03c9ka\u00b4ri]] *! * * *p. [\u03b9 [\u03c6kjo\u00b4\u2013o\u00b4L-mu-[\u03c9kariH ]] *! * *q. [\u03b9 [\u03c6kjo\u00b4\u2013o\u00b4L-mu-[\u03c9kar\u0131\u00b4c5]] *! * *r. [\u03b9 [\u03c6kjo\u00b4\u2013o\u00b4L-mu-[\u03c9ka\u00b4ri]] *! *s. [\u03b9 [\u03c6kjo\u00b4\u2013o\u00b4L-mu\u00b4-[\u03c9kariH ]] *! * * *t. [\u03b9 [\u03c6kjo\u00b4\u2013o\u00b4L-mu\u00b4-[\u03c9kar\u0131\u00b4c5]] *! * *u. [\u03b9 [\u03c6kjo\u00b4\u2013o\u00b4L-mu\u00b4-[\u03c9ka\u00b4ri]] *! * *185(160) Evaluation of associative Initial H-tone root with null determiner[PP{keH, ke\u00b4}[DP[NumP{mu, mu\u00b4}+[nP{ka\u00b4ri, kariH , kar\u0131\u00b4c5}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6keH-mu-[\u03c9kariH ]] *! * * * * *b. [\u03b9 [\u03c6keH-mu-[\u03c9kar\u0131\u00b4c5]] *! * *c. [\u03b9 [\u03c6keH-mu-[\u03c9ka\u00b4ri]] * *! *u d. [\u03b9 [\u03c6keH-mu\u00b4-[\u03c9kariH ]] * * *e. [\u03b9 [\u03c6keH-mu\u00b4-[\u03c9kar\u0131\u00b4c5]] *! * *f. [\u03b9 [\u03c6keH-mu\u00b4-[\u03c9ka\u00b4ri]] *! * *g. [\u03b9 [\u03c6ke\u00b4\u2013mu-[\u03c9kariH ]] *! * *h. [\u03b9 [\u03c6ke\u00b4\u2013mu-[\u03c9kar\u0131\u00b4c5]] *! *i. [\u03b9 [\u03c6ke\u00b4\u2013mu-[\u03c9ka\u00b4ri]] *! *j. [\u03b9 [\u03c6ke\u00b4\u2013mu\u00b4-[\u03c9kariH ]] *! * *k. [\u03b9 [\u03c6ke\u00b4\u2013mu\u00b4-[\u03c9kar\u0131\u00b4c5]] *! *l. [\u03b9 [\u03c6ke\u00b4\u2013mu\u00b4-[\u03c9ka\u00b4ri]] *! *186B.3.2 Class 5(161) Evaluation of Initial H-tone c5 root with overt determiner[DP{ri, r\u0131\u00b4}L+[NumP{i, \u0131\u00b4}+[nP{Bu\u00b4ruuNga, BuruuNgaH , Buru\u00b4uNgac5}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 riL-i-[\u03c9Bu\u00b4ruuNga]] *! * *b. [\u03b9 [\u03c6 riL-\u0131\u00b4-[\u03c9Bu\u00b4ruuNga]] *! * * *c. [\u03b9 [\u03c6 r\u0131\u00b4L-\u0131\u00b4-[\u03c9Bu\u00b4ruuNga]] *! * *d. [\u03b9 [\u03c6 r\u0131\u00b4L-i-[\u03c9Bu\u00b4ruuNga]] *! * * *e. [\u03b9 [\u03c6 riL-i-[\u03c9BuruuNgaH ]] *! * * * *f. [\u03b9 [\u03c6 riL-\u0131\u00b4-[\u03c9BuruuNgaH ]] *! * *g. [\u03b9 [\u03c6 r\u0131\u00b4L-\u0131\u00b4-[\u03c9BuruuNgaH ]] *! *h. [\u03b9 [\u03c6 r\u0131\u00b4L-i-[\u03c9BuruuNgaH ]] *! * *u i. [\u03b9 [\u03c6 riL-i-[\u03c9Buru\u00b4uNgac5]] *j. [\u03b9 [\u03c6 riL-\u0131\u00b4-[\u03c9Buru\u00b4uNgac5]] *! *k. [\u03b9 [\u03c6 r\u0131\u00b4L-\u0131\u00b4-[\u03c9Buru\u00b4uNgac5]] *!l. [\u03b9 [\u03c6 r\u0131\u00b4L-i-[\u03c9Buru\u00b4uNgac5]] *! *187(162) Evaluation of Initial H-tone c5 root with null determiner[DP[NumP{i, \u0131\u00b4}+[nP{Bu\u00b4ruuNga, BuruuNgaH , Buru\u00b4uNgac5}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 i-[\u03c9Bu\u00b4ruuNga]] *! *b. [\u03b9 [\u03c6 \u0131\u00b4-[\u03c9Bu\u00b4ruuNga]] *! * *c. [\u03b9 [\u03c6 i-[\u03c9BuruuNgaH ]] *! * * *d. [\u03b9 [\u03c6 \u0131\u00b4-[\u03c9BuruuNgaH ]] *! * *u e. [\u03b9 [\u03c6 i-[\u03c9Buru\u00b4uNgac5]] *f. [\u03b9 [\u03c6 \u0131\u00b4-[\u03c9Buru\u00b4uNgac5]] *! *188(163) Evaluation of associative initial H-tone c5 root with overt determiner[PP{woH, wo\u00b4}+[DP{ri, r\u0131\u00b4}L+[NumP{i, \u0131\u00b4}+[nP{Bu\u00b4ruuNga, BuruuNgaH , Buru\u00b4uNgac5}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6woH-riL-i-[\u03c9Bu\u00b4ruuNga]] *! * * *b. [\u03b9 [\u03c6woH-riL-\u0131\u00b4-[\u03c9Bu\u00b4ruuNga]] *! * * * *c. [\u03b9 [\u03c6woH\u2013r\u0131\u00b4L-\u0131\u00b4-[\u03c9Bu\u00b4ruuNga]] *! * * *d. [\u03b9 [\u03c6woH\u2013r\u0131\u00b4L-i-[\u03c9Bu\u00b4ruuNga]] *! * * * *e. [\u03b9 [\u03c6woH-riL-i-[\u03c9BuruuNgaH ]] *! * * * * *f. [\u03b9 [\u03c6woH-riL-\u0131\u00b4-[\u03c9BuruuNgaH ]] *! * * *g. [\u03b9 [\u03c6woH\u2013r\u0131\u00b4L-\u0131\u00b4-[\u03c9BuruuNgaH ]] *! * *h. [\u03b9 [\u03c6woH\u2013r\u0131\u00b4L-i-[\u03c9BuruuNgaH ]] *! * * *u i. [\u03b9 [\u03c6woH-riL-i-[\u03c9Buru\u00b4uNgac5]] * *j. [\u03b9 [\u03c6woH-riL-\u0131\u00b4-[\u03c9Buru\u00b4uNgac5]] *! * *k. [\u03b9 [\u03c6woH\u2013r\u0131\u00b4L-\u0131\u00b4-[\u03c9Buru\u00b4uNgac5]] *! *l. [\u03b9 [\u03c6woH\u2013r\u0131\u00b4L-i-[\u03c9Buru\u00b4uNgac5]] *! * *m. [\u03b9 [\u03c6wo\u00b4-riL-i-[\u03c9Bu\u00b4ruuNga]] *! *n. [\u03b9 [\u03c6wo\u00b4-riL-\u0131\u00b4-[\u03c9Bu\u00b4ruuNga]] *! * *o. [\u03b9 [\u03c6wo\u00b4\u2013r\u0131\u00b4L-\u0131\u00b4-[\u03c9Bu\u00b4ruuNga]] *! *p. [\u03b9 [\u03c6wo\u00b4\u2013r\u0131\u00b4L-i-[\u03c9Bu\u00b4ruuNga]] *! * *q. [\u03b9 [\u03c6wo\u00b4-riL-i-[\u03c9BuruuNgaH ]] *! * *r. [\u03b9 [\u03c6wo\u00b4-riL-\u0131\u00b4-[\u03c9BuruuNgaH ]] *! * * *s. [\u03b9 [\u03c6wo\u00b4\u2013r\u0131\u00b4L-\u0131\u00b4-[\u03c9BuruuNgaH ]] *! * *t. [\u03b9 [\u03c6wo\u00b4\u2013r\u0131\u00b4L-i-[\u03c9BuruuNgaH ]] *! * * *u. [\u03b9 [\u03c6wo\u00b4-riL-i-[\u03c9Buru\u00b4uNgac5]] *!v. [\u03b9 [\u03c6wo\u00b4-riL-\u0131\u00b4-[\u03c9Buru\u00b4uNgac5]] *! *w. [\u03b9 [\u03c6wo\u00b4\u2013r\u0131\u00b4L-\u0131\u00b4-[\u03c9Buru\u00b4uNgac5]] *!x. [\u03b9 [\u03c6wo\u00b4\u2013r\u0131\u00b4L-i-[\u03c9Buru\u00b4uNgac5]] *! *189(164) Evaluation of associative initial H-tone c5 root with null determiner[PP{wiH, w\u0131\u00b4}+[DP[NumP{i, \u0131\u00b4}+[nP{Bu\u00b4ruuNga, BuruuNgaH , Buru\u00b4uNgac5}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)u a. [\u03b9 [\u03c6wiH-i-[\u03c9Bu\u00b4ruuNga]] * *b. [\u03b9 [\u03c6wiH-\u0131\u00b4-[\u03c9Bu\u00b4ruuNga]] *! * * *c. [\u03b9 [\u03c6w\u0131\u00b4-\u0131\u00b4-[\u03c9Bu\u00b4ruuNga]] *! *d. [\u03b9 [\u03c6w\u0131\u00b4-i-[\u03c9Bu\u00b4ruuNga]] *! * *e. [\u03b9 [\u03c6wiH-i-[\u03c9BuruuNgaH ]] *! * * * *f. [\u03b9 [\u03c6wiH-\u0131\u00b4-[\u03c9BuruuNgaH ]] *! * * * *g. [\u03b9 [\u03c6w\u0131\u00b4-\u0131\u00b4-[\u03c9BuruuNgaH ]] * *!h. [\u03b9 [\u03c6w\u0131\u00b4-i-[\u03c9BuruuNgaH ]] *! * *i. [\u03b9 [\u03c6wiH-i-[\u03c9Buru\u00b4uNgac5]] *! *j. [\u03b9 [\u03c6wiH-\u0131\u00b4-[\u03c9Buru\u00b4uNgac5]] *! * * *k. [\u03b9 [\u03c6w\u0131\u00b4-\u0131\u00b4-[\u03c9Buru\u00b4uNgac5]] *!l. [\u03b9 [\u03c6w\u0131\u00b4-i-[\u03c9Buru\u00b4uNgac5]] *! *190B.3.3 Class 9\/10(165) Evaluation of Initial H-tone c9 root with overt determiner[DP{a, a\u00b4}L+[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 aL-mL-[\u03c9baraheH ]] *! * * *b. [\u03b9 [\u03c6 aL-mL-[\u03c9ba\u00b4rahe]] *! * *u c. [\u03b9 [\u03c6 aL-mL-[\u03c9bara\u00b4hec9\/10]] *d. [\u03b9 [\u03c6 a\u00b4L-mL-[\u03c9baraheH ]] *! *e. [\u03b9 [\u03c6 a\u00b4L-mL-[\u03c9ba\u00b4rahe]] *! * * *f. [\u03b9 [\u03c6 a\u00b4L-mL-[\u03c9bara\u00b4hec9\/10]] *! *(166) Evaluation of Initial H-tone c10 root with overt determiner[DP{tSaa, tSa\u00b4a\u00b4, tSa\u00b4a}L+[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 tSaaL-mL-[\u03c9baraheH ]] *! * * *b. [\u03b9 [\u03c6 tSaaL-mL-[\u03c9ba\u00b4rahe]] *! * *u c. [\u03b9 [\u03c6 tSaaL-mL-[\u03c9bara\u00b4hec9\/10]] *d. [\u03b9 [\u03c6 tSa\u00b4aL-mL-[\u03c9baraheH ]] *! * *e. [\u03b9 [\u03c6 tSa\u00b4aL-mL-[\u03c9ba\u00b4rahe]] *! * * *f. [\u03b9 [\u03c6 tSa\u00b4aL-mL-[\u03c9bara\u00b4hec9\/10]] *! *g. [\u03b9 [\u03c6 tSa\u00b4a\u00b4L-mL-[\u03c9baraheH ]] *! *h. [\u03b9 [\u03c6 tSa\u00b4a\u00b4L-mL-[\u03c9ba\u00b4rahe]] *! * *i. [\u03b9 [\u03c6 tSa\u00b4a\u00b4L-mL-[\u03c9bara\u00b4hec9\/10]] *!191(167) Evaluation of Initial H-tone c9 root with null determiner[DP[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6mL-[\u03c9baraheH ]] *! * *b. [\u03b9 [\u03c6mL-[\u03c9ba\u00b4rahe]] *! *u c. [\u03b9 [\u03c6mL-[\u03c9bara\u00b4hec9\/10]] *(168) Evaluation of Initial H-tone c10 root with null determiner[DP[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6mL-[\u03c9baraheH ]] *! * *b. [\u03b9 [\u03c6mL-[\u03c9ba\u00b4rahe]] *! *u c. [\u03b9 [\u03c6mL-[\u03c9bara\u00b4hec9\/10]] *192(169) Evaluation of associative Initial H-tone c9 root with overt determiner[PP{rjaH, rja\u00b4}+[DP{a, a\u00b4}L+[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 rjaH-aL-mL-[\u03c9baraheH ]] *! * * *b. [\u03b9 [\u03c6 rja\u00b4-aL-mL-[\u03c9baraheH ]] *! * *c. [\u03b9 [\u03c6 rjaH-a\u00b4L-mL-[\u03c9baraheH ]] *! * *d. [\u03b9 [\u03c6 rja\u00b4-a\u00b4L-mL-[\u03c9baraheH ]] *! * *e. [\u03b9 [\u03c6 rjaH-aL-mL-[\u03c9ba\u00b4rahe]] *! *f. [\u03b9 [\u03c6 rja\u00b4-aL-mL-[\u03c9ba\u00b4rahe]] *! * * *g. [\u03b9 [\u03c6 rjaH-a\u00b4L-mL-[\u03c9ba\u00b4rahe]] *! * * *h. [\u03b9 [\u03c6 rja\u00b4-a\u00b4L-mL-[\u03c9ba\u00b4rahe]] *! * *u i. [\u03b9 [\u03c6 rjaH-aL-mL-[\u03c9bara\u00b4hec9\/10]] * *j. [\u03b9 [\u03c6 rja\u00b4-aL-mL-[\u03c9bara\u00b4hec9\/10]] *! *k. [\u03b9 [\u03c6 rjaH-a\u00b4L-mL-[\u03c9bara\u00b4hec9\/10]] *! *l. [\u03b9 [\u03c6 rja\u00b4-a\u00b4L-mL-[\u03c9bara\u00b4hec9\/10]] *!193(170) Evaluation of associative Initial H-tone c10 root with overt determiner[PP{rEH, rE\u00b4}+[DP{tSaa, tSa\u00b4a\u00b4, tSa\u00b4a}L+[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 rE\u00b4-tSaaL-mL-[\u03c9baraheH ]] *! * *b. [\u03b9 [\u03c6 rE\u00b4-tSa\u00b4aL-mL-[\u03c9baraheH ]] *! * * *c. [\u03b9 [\u03c6 rE\u00b4-tSa\u00b4a\u00b4L-mL-[\u03c9baraheH ]] *! * *d. [\u03b9 [\u03c6 rEH-tSaaL-mL-[\u03c9baraheH ]] *! * * *e. [\u03b9 [\u03c6 rEH-tSa\u00b4aL-mL-[\u03c9baraheH ]] *! * * *f. [\u03b9 [\u03c6 rEH-tSa\u00b4a\u00b4L-mL-[\u03c9baraheH ]] *! * *g. [\u03b9 [\u03c6 rE\u00b4-tSaaL-mL-[\u03c9ba\u00b4rahe]] *! * *h. [\u03b9 [\u03c6 rE\u00b4-tSa\u00b4aL-mL-[\u03c9ba\u00b4rahe]] *! * *i. [\u03b9 [\u03c6 rE\u00b4-tSa\u00b4a\u00b4L-mL-[\u03c9ba\u00b4rahe]] *! * *j. [\u03b9 [\u03c6 rEH-tSaaL-mL-[\u03c9ba\u00b4rahe]] *! * * *k. [\u03b9 [\u03c6 rEH-tSa\u00b4aL-mL-[\u03c9ba\u00b4rahe]] *! * * * *l. [\u03b9 [\u03c6 rEH-tSa\u00b4a\u00b4L-mL-[\u03c9ba\u00b4rahe]] *! * * *m. [\u03b9 [\u03c6 rE\u00b4-tSaaL-mL-[\u03c9bara\u00b4hec9\/10]] *!n. [\u03b9 [\u03c6 rE\u00b4-tSa\u00b4aL-mL-[\u03c9bara\u00b4hec9\/10]] *!o. [\u03b9 [\u03c6 rE\u00b4-tSa\u00b4a\u00b4L-mL-[\u03c9bara\u00b4hec9\/10]] *!u p. [\u03b9 [\u03c6 rEH-tSaaL-mL-[\u03c9bara\u00b4hec9\/10]] * *q. [\u03b9 [\u03c6 rEH-tSa\u00b4aL-mL-[\u03c9bara\u00b4hec9\/10]] *! * *r. [\u03b9 [\u03c6 rEH-tSa\u00b4a\u00b4L-mL-[\u03c9bara\u00b4hec9\/10]] *! *194(171) Evaluation of associative Initial H-tone c9 root with null determiner[PP{rEH, rE\u00b4}+[DP+[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 rEH-mL-[\u03c9baraheH ]] *! * * * *b. [\u03b9 [\u03c6 rE\u00b4-mL-[\u03c9baraheH ]] *! *\/ c. [\u03b9 [\u03c6 rEH-mL-[\u03c9ba\u00b4rahe]] *! * *d. [\u03b9 [\u03c6 rE\u00b4-mL-[\u03c9ba\u00b4rahe]] *! * *u e. [\u03b9 [\u03c6 rEH-mL-[\u03c9bara\u00b4hec9\/10]] * *f. [\u03b9 [\u03c6 rE\u00b4-mL-[\u03c9bara\u00b4hec9\/10]] *!(172) Evaluation of associative Initial H-tone c10 root with null determiner[PP{rEH, rE\u00b4}+[DP+[NumP{N, \/0}L+[nP{baraheH , ba\u00b4rahe, bara\u00b4hec9\/10}]]]]MONOHHIGH\u03a6*(\u03c3V\u00b4VV-TONESEL:HSEL:LMP:RT-DET*\/0RTSEL:HHIGH\u03c9MP:RT-CMAL(H,\u03c6)a. [\u03b9 [\u03c6 rEH-mL-[\u03c9baraheH ]] *! * * * *b. [\u03b9 [\u03c6 rE\u00b4-mL-[\u03c9baraheH ]] *! *\/ c. [\u03b9 [\u03c6 rEH-mL-[\u03c9ba\u00b4rahe]] *! * *d. [\u03b9 [\u03c6 rE\u00b4-mL-[\u03c9ba\u00b4rahe]] *! * *u e. [\u03b9 [\u03c6 rEH-mL-[\u03c9bara\u00b4hec9\/10]] * *f. [\u03b9 [\u03c6 rE\u00b4-mL-[\u03c9bara\u00b4hec9\/10]] *!195","@language":"en"}],"Genre":[{"@value":"Thesis\/Dissertation","@language":"en"}],"GraduationDate":[{"@value":"2022-05","@language":"en"}],"IsShownAt":[{"@value":"10.14288\/1.0407255","@language":"en"}],"Language":[{"@value":"eng","@language":"en"}],"Program":[{"@value":"Linguistics","@language":"en"}],"Provider":[{"@value":"Vancouver : University of British Columbia Library","@language":"en"}],"Publisher":[{"@value":"University of British Columbia","@language":"en"}],"Rights":[{"@value":"Attribution-NonCommercial-NoDerivatives 4.0 International","@language":"*"}],"RightsURI":[{"@value":"http:\/\/creativecommons.org\/licenses\/by-nc-nd\/4.0\/","@language":"*"}],"ScholarlyLevel":[{"@value":"Graduate","@language":"en"}],"Supervisor":[{"@value":"Pulleyblank, Douglas George","@language":"en"}],"Title":[{"@value":"Prosodic phonology in Nata","@language":"en"}],"Type":[{"@value":"Text","@language":"en"}],"URI":[{"@value":"http:\/\/hdl.handle.net\/2429\/80971","@language":"en"}],"SortDate":[{"@value":"2022-12-31 AD","@language":"en"}],"@id":"doi:10.14288\/1.0407255"}