UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Topics on Yorùbá nominal expressions Ajíbóyè, Ọládiípọ̀ Jacob 2005

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
831-ubc_2006-129751.pdf [ 13.05MB ]
Metadata
JSON: 831-1.0092961.json
JSON-LD: 831-1.0092961-ld.json
RDF/XML (Pretty): 831-1.0092961-rdf.xml
RDF/JSON: 831-1.0092961-rdf.json
Turtle: 831-1.0092961-turtle.txt
N-Triples: 831-1.0092961-rdf-ntriples.txt
Original Record: 831-1.0092961-source.json
Full Text
831-1.0092961-fulltext.txt
Citation
831-1.0092961.ris

Full Text

TOPICS ON Y O R U B A N O M I N A L EXPRESSIONS  by O L A D I I P O J A C O B AJIBOYE  B.A (1983), P.G.D.E (1988), M A (1992), University of florin  A THESIS SUBMITTED IN P A R T I A L F U L F I L M E N T OF T H E REQUIREMENTS FOR T H E D E G R E E OF DOCTOR OF PHILOSOPHY in  T H E F A C U L T Y OF G R A D U A T E STUDIES  (Linguistics)  T H E UNIVERISTY OF BRITISH C O L U M B I A November 2005  © Oladiipo Jacob Ajiboye, 2005  ABSTRACT  This thesis discusses four selected topics on Yoruba nominal expressions: the syntax of possessives, the construal of bare nouns, the marking of specificity and salience, and plural marking strategies. Regarding possessives, it is proposed that they have one base structure (a v P shell). The difference in surface linear order between verbal and nominal genitives is determined by which of the two arguments move. In nominal genitives, the possessum moves. In verbal genitives, it is the possessor that moves. Regarding the interpretation of Yoruba bare nouns, it is shown that they can be construed in one of three ways: as generics, as indefinites, or as definites. First, generics may be lexically conditioned (with permanent state predicates) or grammatically conditioned (with transitory predicates through the use of imperfective maa-n). Second, wherever a generic construal is illicit, an indefinite construal is licit. Third, definite construals are discourse-linked. Regarding specificity, it is shown that Yoruba overtly marks specificity on NPs with the element kan. Regarding salience, it is shown that definite DPs are morphologically marked as salient (by virtue of being unique, in an identity relation or additive) through the use of nda. Finally, regarding plural marking, it is shown that Yoruba uses three different strategies: contextually, semantically, or morphologically determined plurality. It is proposed that the deployment of the P L U R A L feature is determined by feature percolation or feature matching.  T A B L E OF CONTENTS Abstract  ii  Table of contents  iii  Acknowledgements...  x  Dedication  xv  Abbreviations  xvi  Chapter one: Introduction  1  1.1 Purpose and organization  1  1.1.1 Expression of possessive relation  1  1.1.2 Discourse related interpretation  3  1.1.2.1 Interpreting bare nouns  3  1.1.2.2 Marking nouns for specificity and salience  4  1.1.3 Plural strategies  4  1.2 Clause structure  6  1.2.1 Small clause  6  1.2.2 Relative clause  7  1.2.3 The Internal Structure of Yorilba DP  8  1.2.3.1 D P and the D position... 1.2.3.2 Genitive structure in DP  :  12  v  13  1.2.4 D(emonstrative) and NP  14  1.2.5 Modification  15  Chapter two: Yoruba Genitive constructions  21  2  Introduction  21  2.1 The semantics of the R relation  21  2.1.1 Three kinds of R-relation  22  2.1.2 "Possessive" versus "Genitive"  25  2.1.3 Possessor and Possessum  28  2.1.4 The three types of genitive constructions in Yoruba  ;  29  2.2 The syntax of the R relation  30  2.2.1 Possessor-Possessum as a co-argument relation  30  2.2.1.1 Base structure for Yoruba genitives  32  2.2.1.2 Other languages with overt genitive morpheme  37  2.2.2 Possessor-Possessum as a Head-Complement relation  38  2.2.3 Parallels between V-syntax and N-syntax  39  2.2.3.1 Parallels between the verbal possessive and the nominal genitive  39  2.2.3.2 Parallels between IP-syntax and DP-syntax  42  2.3 The Yoruba nominal genitive: syntax  43  2.3.1 Yoruba genitive constructions involve a co-argument relation  44  2.3.2 Possessum raising in Yoruba genitive constructions  44  2.3.3 What moves: N or NP?  48  2.3.3.1 Evidence from modifier placement..  49  2.3.3.2 Evidence from structural ambiguity  50  2.3.3.3 Evidence from selectional restrictions  :  51  2.3.4 Genitive pronouns  53  2.4 The Yoruba nominal genitive: syntax-phonology interaction 2.4.1 Spelling out the genitive: cross-linguistic evidence  56 .....51  2.4.2 The phonological distribution of the Mid Tone Syllable  61  2.4.3 Analysis I: the Mid Tone Syllable is an underlying mora \i  63  2.4.4 Analysis II: the Mid Tone Syllable as phonologically conditioned segment  68  2 AAA The Mid Tone Syllable is not a genitive morpheme  68  2.4.4.2 The Mid Tone Syllable as a prosthetic vowel  70  2.4.4.3 The Mid Tone Syllable as a mark of hesitation  74  2.4.5 Relating the Mid Tone Syllable to L-deletion  76  2.4.5.1 The necessary and sufficient conditions for L-deletion  83  2.4.5.2 MTS and L-deletion are in complementary distribution  84  2.5 Conclusion  Chapter 3:fr'-genitivesas reduced relative clauses  85  86  3  Introduction  86  3.1 The parallel between ti genitives and ti relative clauses  89  3.1.1 Two structures for relative clauses  '.  89  3.1.1.1 Relative clauses as complements to D  90  3.1.1.2 Relative clauses as adjuncts to N P  92  3.1.2 ti genitives are reduced relative clauses  94  3.1.2.1 M-tone ti and H-tone ti  100  3.1.2.2 MTS andHTS  103  3.1.2.3 M-tone ni and H-tone ni  104  3.1.2.4 The significance of H-tone/M-tone alternations  105  3.1.3 The ti N construction  106  3.1.3.1 ti as Complementizer analysis  107  3.1.3.2 ti as N analysis: Awobuluyi (2004)......  109  3.1.4 Assigning genitive case to the possessor  115  3.1.4.1 The co-argument analysis: exceptional case marking  116  3.1.4.2 The Head-Complement analysis: case under government  117  3.1.5 Assigning structural case to the raised possessum 3.1.5.1 The problem  118 '.  .....118  3.1.5.2 Assigning Structural case to the raised possessum I: Nominative.... 119 3.1.5.3 Assigning Structural case to the raised possessum II: Accusative.... 120 3.2 Genitive Case assignment and co-occurrence of the MTS with genitive ti  121  3.2.1 Case alternations: the Mid Tone Syllable or ti  121  3.2.2 Case stacking: the Mid Tone Syllable and ti  123  3.2.3 Interpretive effects of Case stacking  125  3.3 The co-occurrence of the copula je with genitive ti  127  3.4 Other syntactic contexts for the R-relation: prepositional uses of ti  130  3.4.1 lati as a locative P at the right edge of vP  131  3.4.2 ti as a locative P at the left edge of the vP  132  3.4.3 ti as a marker of adjunct extraction  134  3.5 Conclusion  135  Chapter four: Interpreting Yoruba bare nouns  136  4  Introduction  136  4.1 The determinants for interpreting Yoruba bare nouns  137  4.1.1 Verb classes  137  4.1.1.1 Permanent states  138  4.1.1.2 Temporary states  139  4.1.1.3 Events  140  4.1.1.4 Distinguishing "temporary-state-events" and "permanent states".... 141 4.1.2 Subject versus object position  143  4.1.3 Discourse-linking  144  4.2 What conditions the generic construal of bare nouns?  146  4.2.1 P(ermanent)-states have a lexical G E N operator  146  4.2.1.1 Bare noun object of transitive P-states  149  4.2.1.2 Bare noun subject of transitive P-states  150  4.2.1.3 Bare noun subject of intransitive Permanent-states  153  4.2.2 T(emporary)-states require a grammatical G E N operator  154  4.2.2.1 Bare noun object of transitive T-state  155  4.2.2.2 Bare noun subject of transitive T-state  155  4.2.2.3 On the absence of intransitive T-state in Yoruba  156  4.2.3 Events require a grammatical G E N operator  157  4.2.3.1 Bare noun object of transitive event  158  4.2.3.2 Bare noun subject of transitive event  158  4.2.3.3 Bare noun subject of intransitive event  159  4.2.4 Analysis of grammatically conditioned genericity  160  4.2.5 Extending grammatically conditioned genericity to Permanent State verbs.... 163 4.2.6 Contrast between English and Yoruba genericity: evidence for default aspectl66 4.3 What conditions the indefinite construal of Yoruba bare nouns? 4.3.1 Defining indefiniteness 4.3.2 Generic and indefinite construals are in complementary distribution  168 ••• 168 169  4.3.3 Elsewhere case: indefinite bare nouns are bound by existential operator 4.4 What conditions the definite construal of Yoruba bare nouns  171 174  4.4.1 Defining Definiteness  174  4.4.2 Definite construal is unavailable in out-of-the-blue contexts  175  4.4.3 Definite construal arises from Discourse linking  176  4.5 Consequences of the analysis  179  4.5.1 Bare nouns are structurally ambiguous  179  4.5.1.1 Bare nouns can be NP or DP  179  4.5.1.2 Critique of previous analysis (Ajiboye 2001)  180  4.5.2 Implication for the analysis of genitive construction 4.6 Conclusion  182 183  Chapter five: Marking specificity and salience in nominal expressions  184  5  Introduction  184  5.1 Specificity in Yoruba: kan.'.  186  5.1.1 Defining specificity  186  5.1.2 Marking specificity  187  5.1.2.1 Yoruba  187  5.1.2.2 English  189  5.1.2.3 Turkish  190  5.1.2.4 Gungbe  191  5.1.3 Testing for specificity I: Speaker knowledge Ludlow and Neale (1991)  193  5.1.4 Testing for specificity II: the subset relation En? (1991)  195  5.1.4.1 Yoruba  195  5.1.4.2 English  197  5.1.4.3 Turkish  199  5.2 Salience in Yoruba: nda  200  5.2.1 Defining Salience  201  5.2.1.1 Uniqueness  201  5.2.1.2Additivity  202  5.2.1.3 Identity  :  202  5.2.2 Marking salience in Yoruba  203  5.2.2.1 Focused naa-arguments: uniqueness or identity function  204  5.2.2.2 Nad marks additivity or identity with (non-focused) arguments  207  5.2.3 Testing for salience  210  5.2.3.1 Yoruba salient nominals can't be indefinite  210  5.2.3.2 Yoruba salient nominals can be "free choice"  212  5.2.3.3 No dedicated marking for salience in English  213  5.2.4 nda combines with other functional elements within the D P  214  5.3 The syntax of kan and nda  216  5.4 The number interpretation of nda  218  5.5 Conclusion  219  Chapter six: Plural strategies in Yoruba  220  6  Introduction  220  6.1 Contextually determined plurality  221  6.1.1 The General number analysis  222  6.1.2 Unspecified for number resulting in ambiguity  223  6.1.3 Unspecified for number with obligatory singular interpretation  225  6.2 Semantically determined plurality  226  6.2.1 The feature percolation analysis  227  6.2.2 Inherently plural Quantifiers  231  6.2.3 Inherently plural numerals  233  6.2.3.1 Accounting for the floating tone of the m-numerals  236  6.2.3.2 Other languages that pattern with Yoruba  238  6.3 Morphologically determined plurality 6.3.1 awgn marks plural on nouns  239 '  241  6.3.1.1 awgn as a pronoun  242  6.3.1.2 Appositive analysis of PL-NP  243  6.3.1.3 Internal structure of awgn  245  6.3.2 won- marks plural on demonstratives  248  6.3.3 Reduplication marks plural on modifiers  251  6.3.3.1 What prevents N copy: blocking?  258  6.3.3.2 The emergence of the unmarked  260  6.3.4. Multiple plural marking  •  6.3.5 Feature percolation versus feature matching 6.4 Comparing analyses  261 265 268  6.4.1 The plural parameter (Deprez 2004)  268  6.4.2 Plural marking as a functional head or as a modifier (Wiltschko 2004)  269  6.4.2.1 Plural marking as a functional head  270  6.4.2.2 Plural marking as a modifier  271  6.5 Conclusion  273  Chapter 7: Conclusion  274  7 Findings and open questions  274  7.1 Findings 7.2 Open questions  References  :  274 276  279  ACKNOWLEDGEMENT  BI a kd bd tete ku, a 6 jeran to terin The longer we live, the more we achieve There are a lot of people that I have to acknowledge for the role they play either severally or jointly in the course of pursuing the studies that produce this dissertation. I am lucky to have Rose-Marie Dechaine as my supervisor for so many reasons. First, she was instrumental to my coming to U B C . Second, she paid my flight ticket to U B C and provided me accommodation for the first two weeks of my arrival. Third, her deep knowledge of Yoruba language and my cultural background has been very helpful in putting me on the right track throughout my years of study at U B C . This dissertation stems from the first conference paper that I had in 2001 under her auspices. She has given me every support that I need to do thorough research: time, insightful and provocative thoughts and ideas, books and journals, computers for use at home and on campus. She also sponsored me to conferences where parts of this dissertation were presented and a linguistic field trip to Nigeria in 2003. Above all, she knows how to make me feel less insecure about my ideas while challenging me to make them better even if it is going to take extra time and pain. I am also grateful for benefiting from the following research grants from Social Sciences and Humanities Research Council (SSHRC) as her Research Assistant from 1999-2005: (i) #412-03-1003 (Principal Investigator: A . M . DiSciullo; Co-investigator: R - M . Dechaine (ii) #412-97-0016 (Principal Investigator: A . M . DiSciullo; Co-investigator: R . M . Dechaine) (iii) #410-2000-1234 (Principal Investigator: R - M . Dechaine); (iv) #12R57105. I am equally grateful to Douglas Pulleyblank who is both my teacher and a member of my committee. Since the day I met him, he has been very helpful emotionally and academically. His many years of study of the Yoruba language are valuable assets to me. I benefited immensely from his criticisms, suggestions, and comments on every part of this dissertation, most especially the phonological aspects. I acknowledge the financial support from his research grant (#410-2002-0041). Thanks to Anne-Marie, his wife, who provided me with some cooking utensils and a sleeping blanket when I first came. Finally, I also feel honoured by their occasional visits to me. The fact that Rose-Marie and Douglas are icons in the field of the Yoruba language who are to play significant roles in this study was anticipated in the following extracts from Awobuliiyi's referee report to the department of Linguistics U B C in 1999. " M r . Ajiboye holds a Master's degree in Yoruba language, which to us here is proof that he is capable of  successfully undertaking studies and original doctoral degree. We had hoped that he would do that here in this department. Now, this is not to say that we begrudge him his desire to do it in your department. Indeed, he has our very best wishes, as we know that he will be in good hands there." Truly, I have been in good hands as I have had a soft-landing into Doug and Rose-Marie's hands. I thank Martina Wiltschko for her valuable contributions as a member of my dissertation committee. She honoured all meetings that I requested; read every draft of the dissertation in the shortest possible time and gave me feedback both in written form and oral discussion. I also thank her for sponsoring me to the 2005 B L S conference where I presented chapter four of this dissertation. I am short of adequate words to convey my deepest appreciation to Victor Manfredi a.k.a Femuwe, because of the many roles he played in my doctoral program. First, he was also instrumental to my coming to Canada after his visit to Ilorin in 1997 on a Fulbright Scholarship. On this significant role, I say: Bi onko 6 ko, ki longbin 6 gbin. Bi ongbin o gbin, ki lonka 6 ka a. 'No planting without harrowing, and no harvest without planting'. Second, I thank him for moral and financial support. Third, I am grateful for the supply of a lot of relevant academic materials even at short notice and for making his library in Boston available to me each time I visited him. Finally, I thank him for his immense contribution in form of comments, criticisms and suggestions to the upcoming of this dissertation. Though some of his suggestions could not make it into this final draft, they form a strong basis for future research.  ,  I say thanks to oga Yiwola Awoyale, for recommending me to Rose-Marie to come and study at U B C . I am indebted to my instructors: Pat Shaw, Uyi Stewart, Laura Downing, Douglas Pulleyblank, Henry Davis, Rose-Marie Dechaine, Felicia Lee, Bryan Gick, Lisa Matthewson who all helped to shape my linguistic perspective and other faculty members: Joe Stemberger, Eric Baetson, Gunnar Hannson, Guy Carden, Hotze Rullmann, who have shared their knowledge with me either on personal grounds or by their comments at various presentations of my work in the department of linguistics seminar series. Uyi deserves a special thank you. His charge to me that I should always do one thing at a time is a guiding principle that helps me whenever I am overwhelmed by many deadlines at the same time. I also thank Bryan for allowing me unofficially to use his lab as my working place throughout the entire period of writing this dissertation. I would like to thank my Nigeria friends at U B C : Gloria and her husband Robert, Pius  Adesanmi, Chidi Oguamanam, Rashidat Ayinla, Olumide Ibikiinle, Lanre Ankeebi, Bayo Banjo, Bayo Aregbesola, and Chinedu. Outside the U B C community my thanks go to Kehinde and Victoria Ajibikee, Lukman and Mary Bamigbade, Victor and Fatima Enisanlayeye, Nikee and Josiah Akfnsanmf, Wale and Bukola Eniade, Folasade Daramola, Brando Adesuyan, Babatunde and Funnke Balogun and Mosope Fagbohungbe for making me less homesick. I owe my wife, Modupe special thanks for her moral and emotional support while both of us were here. I am grateful to my children Ola6nib6ji, Jenrola, Omole and pdunitan for their good behaviour during my six-year-absence from home.  I am also grateful to  Motunrayo for being an excellent foster mother for those years. I extend my appreciation to Mojirola and Titilayo, my college son Idowu Shittu, my junior sister, Iyabode, my senior sister Oyelaran, my aunt, madam Ebe Atolagbe, my dear brother Amos Fabiyi, my good friends Samson Olajide, jibola Abiodun, Idowu Awosejo, and Ladipo Odeyemi, my mother in-law madam Lucia 6ni for their moral support and constant visits to the children that I left behind. I was never bored by mails from brother Wale Jolayemi and calls from Anifowose Atolagbe asking of my welfare and progress in my studies. For this, I say thanks to the duo. I thank the ejire: Tayewo Aaniioluwa Paul and pmokehindegbegbon Ifeoluwa Peter here in Canada for being patient with me especially when pressure from work sometime make me react negatively to their rightful demands. They are an additional feather to my cap. A jagun, a keru, a keru, Oga ogo Io ba wa se. 'We fought a war, we won and we had booties because the Almighty God was on our side'. I would like to express special thanks to my classmates for their support, Leora Barel, Carrie Gillon, Masaru Kiyota, Gessiane Picanco, Sungunya Ruangjaroon, Naomi Sawai, Scott Shank, Kayono Shiobara, Linda Watt, Ian Wilson, and Rachel Wojdak and the following fellow graduate students for their friendship, Solveiga Armoskaite, Leszek Barzek, Clare Cook, Suzanne Gessner, Susie Jones, Eun-shook Kim, Karstin Koch, Jeff Muhlbauer, Sunyoung Oh, Jeremy Perkins, I owe Jeff Muhlbauer and Clare Cook special thanks for proofreading part of the early drafts of this dissertation and for sharing with me some of the data on Algonquian languages (Plains Cree and Blackfoot) from their fieldwork. Thanks to Sahffiq for technical support at the Interdisciplinary Speech Laboratory. Apart from being a native speaker of Yoruba, I conducted some fieldwork, therefore my deepest thanks go out to Olanikee Qla-Orie, Oluseye Adesola, Rashidat Ayinla, Lanre Arikeebi Oyekanmi Nash and Modupe Ajfboye who shared their knowledge of Yoruba  language with me by acting as consultants on some of the data that I use in this dissertation. I gratefully acknowledge the following sources of financial supports: (i) International Partial Tuition Scholarship (1999-2005), (ii) Alma Mater Society Student A i d Bursary (2000), (iii) Student A i d Bursary for Graduate Students (2005), (iv) Emergency fund for developing country students (financial aid from Canadian Bureau for  International  Education), (v) Stipends for Teaching Assistantship in the department of Linguistics (20012005), (vi) Salary as a Yoruba language Consultant to the 2003/2004 U B C linguistics field methods class. Finally, I give the Almighty God all the glory. Without his approval, it would have been impossible for me to come to U B C . Without his support and assistance, it would have been impossible for me to sail through the storms that I journeyed during the course of this study. The totality of what God has done for me is found in the following Yoruba saying: - A soro 6 se b'ohun t'Oluwa 6 fe A dun unse b'ohun t'Oluwa tise tan Aye e ma tu wo IseOluwani  (Sunny Ade 1983)  When God says yes, nobody can say no When God says no, nobody can say yes You dare not inquire It is God's work  Dedication  This dissertation is dedicated to my late parents Pa James Ajfboye and Mrs. Abigail Ayanponmile Ajfboye  ABBREVIATIONS  ACC  accusative  Agr  agreement  ARG  argument  Asp  aspect  AspP  aspect phrase  BN  bare noun  C  complementizer  CP  complementizer phrase  DEF  definite  Dem  demonstrative  DET  determiner  DP  determiner phrase  Emph  emphasis  ERG  ergative  Fut  future  GEN  generic  Gen  genitive  HTS  high tone syllable  I(NFL)  inflection  IMP  imperfective  Indef  indefinite  IP  inflectional phrase  Loc  locative  MB  Moba  MTS  mid tone syllable  Neg  negation  NOM  nominative  Nom  nominalizer  Num  numeral  0  object  OP  operator  PL  plural  Poss'r  possessor  Poss'm  possessum  Prog  progressive  P-state  permanent state  Q  quantifier  RP  resumptive pronoun  SAL  Salience  SG  singular  SLP  Stage Level Predicate  Spec  specifier  SPF  specificity  S  subject  SY  Standard Yoruba  T-state  Temporary state  V-event  event verb  CHAPTER ONE:  1.1  INTRODUCTION  Purpose and organization This dissertation investigates selected topics on Yoruba nominal expressions and  makes a number of proposals that enable us understand the internal and the external structure of the Yoruba DP. Although the internal structure of Yoruba nominals has been extensively described (Awoyale 1974; Awobuluyi .1978; Bamgbose 1966, 1967, 1990; Yusuf 1995), there remains the question of whether more recent theories of syntax such as the ' D P hypothesis' (Abney 1987; Aboh 1999, 2004; Cinque 1994, 2004, 2005; Kayne 1994; Longobardi 1994, 2000, 2003, 2004) can provide insight into the syntax, pragmatics-and semantics of Yoruba nominals. The dissertation deals with four related topics on nominal expressions, which are as follows: the expression of the possessive relation (chapters 2-3), how bare nouns are construed (chapters 4), how specificity and salience are marked (chapter 5), and plural marking strategies (chapter 6).  1.1.1  The expression of the possessive relation  Chapters 2 and 3 examine how the possessive relation (henceforth R) is expressed in nominal and verbal environments. The first part of chapter 2 discusses the semantics of the R relation, which forms the basis for possessive constructions. The second part deals with the syntax of the R relation, proposing that the Possessor-Possessum relation is best analyzed as a coargument relation. I also show that there is a parallel between verbal possessor constructions such as (la) and their nominal counterparts in (lb). I propose that both (la) and (lb) derive from the same base structure.  (1)  a.  Tiinde ni ile T. have house 'Tunde has a house'  Possessor  verb  Possession  b.  ile e Tunde house M T S T. 'Tunde's house'  Possessum  MPS  Possessor  Observe that the linear order of the possessor and the possessum in genitive constructions is the mirror image of what is obtained in verbal possessives. The other concern of chapter 2 is to account for the phonological content of the mid tone genitive morpheme, which shows up between arguments that are in a genitive relation as in (2a), but is sometimes absent as shown in (2b).  (2)  a.  M o de [ile e Tunde] lsg reach house MTS T. 'I got to Tunde's house.'  possessum MTS possessor  b.  G o m i n a pase k i won ri possessum MTS possessor governor give-order C 3 pi erect [ere Awolowo] si orita Basorun statue A . Loc junction B. 'The Governor ordered that the Awolowo's statue be erected at Basorun junction.'  An extensive account that aims at establishing the phonological content of the M-tone syllable as well as its syntactic and semantic functions is undertaken. Chapter 3 continues the analysis of genitive constructions with more complex structures, looking at the occurrence of possessum NPs with ri-phrases. The example in (3a) indicates that it is possible for the M tone //-element to occur between the possessum and the possessor. The example in (3b) shows the co-occurrence of the MTS and the ri-element. (3)  a.  Tolu fe . [ere ti A w o l o w o ] T. want statue C A . 'Tolu wants the Awolowo's statue.'  possessum C possessor  b.  Mo gba [owo o ti Bunmi] lsg get money MTS C B. 'I got Bunmi's money.'  possessum MTS C possessor  I propose that nominal genitives of the type shown in (2) are to be analyzed as plain genitive DPs and that D, which is occupied by the MTS, takes a small clause (vP) as its complement. In the same way, I propose that the genitive plus //-construction in (3) is to be analyzed as a  DP with its overt D taking a reduced relative clause (CP) as its complement.  1.1.2  Discourse related interpretation  Chapters 4 and 5 examine some discourse related interpretive properties of nominals in Yoruba based on whether they are new, familiar, specific or salient in the discourse.  1.1.2.1 Interpreting bare nouns  Chapter 4 establishes three different ways by which bare nouns can be construed: generic (4a), indefinite (4b), and definite (4c).  (4)  a.  Taye feran aja T. like dog 'Taye likes dogs.'  generic  b.  Taye rf aja T. see dog 'Taye saw a dog.'  indefinite  c.  Taye gbe aja T. carry dog 'Taye carried the dog.'  definite  (in proper discourse context)  I demonstrate that these construals are determined by the syntactic position of the bare noun (subject versus object), the predicate type (permanent state versus temporary state/event) as well as discourse linking. For a generic construal, this work shows that there are two possible ways by which this can be obtained: (i) lexically conditioned generics that are bound by a null generic (GEN) operator, and (ii), grammatically conditioned genericity that is introduced by the imperfective aspectual marker maa-n Lexically conditioned generic construals are found with permanent state predicates whereas grammatically conditioned generic construals are found with transitory (temporary  state and event) predicates. As for indefinite construals, I show that they are possible wherever a generic construal is not otherwise available. Thus, I demonstrate that a bare noun is generic if there is a G E N operator, and that elsewhere it is indefinite. Lastly, I show that in Yoruba, any bare noun can be definite in the proper discourse context.  1.1.2.2 Marking nouns for specificity and salience  Chapter 5 focuses on another discourse related interpretation. I show that nouns in Yoruba can be morphologically marked for specificity and salience. When a noun is unfamiliar and new (i.e. indefinite) it can be overtly marked for specificity with kan (5a). On the other hand, when a noun is familiar (i.e. definite) it can be overtly marked for salience with nda (5b).  (5)  a.  Taye ri aja kan T. see dog specific 'Taye saw a C E R T A I N dog.'  specific  b.  Taye ri aja naa T. see dog salient 'Taye saw the V E R Y dog.'  salience  I argue that the facts reported above for Yoruba contrast in interesting ways with a language like English. For example, in environments where English would require overt marking of definite or indefinite, Yoruba has null (0) marking; i.e. bare nouns. In environments where English does not have obligatory marking for specificity and saliency, Yoruba does, and so we get kan (specificity) and nda (salience).  1.1.3  Plural strategies  Chapter 6 discusses various strategies that Yoruba adopts to mark plural. The first thing that this work shows is that plural marking is not obligatory. However, I demonstrate that there  are three different ways by which plural marking is carried out. One way is through contextually determined plurality. These are the cases where there is no overt plural marking as such and a noun can be interpreted as singular or plural: (6)  Taye nf aja T. have dog = 'Taye has a dog.' = 'Taye has dogs  Another way is a semantically determined plurality. These are the cases where nouns take quantifiers and numerals that have an abstract [ P L U R A L ] feature. Such nouns are unambiguously interpreted as plural. In (7a), there is the quantifier die 'few' co-occurring with aja 'dog' and in (7b) the numeral meje 'seven' co-occurs with the same noun and in both cases, the noun is interpreted as plural.  (7)  a.  Taye ra [aja die] T. buy dog few 'Taye bought few dogs.'  b.  Taye ra [aja meje] T. buy dog seven 'Taye bought seven dogs.'  The third category is the morphologically determined plurality. These are the cases when nouns are overtly marked by plural words. I show that there are two syntactic positions for morphologically determined plurality: the noun and a modifying element in a nominal expression. A noun is marked for plural by the use of the plural word awgn as shown in (8).  (8)  Mo ra [awoniwe] lsg buy PL book 'I bought books in Canada.'  nf Loc  Kanada Place  For plural marking on modifiers, this is realized through the COPY of a modifier, giga giga 'tall', as in (9). (9)  [lie giga giga] wa nf Fankufa house COPY tall be L O C Vancouver 'There are high-rises in Vancouver.'  The example in (9) contradicts claims that have been made in the literature to the effect that Yoruba has only one plural word: awgn (cf. Dryer 1989, Rowlands 1969). I moreover demonstrate that the multiple strategies of plural marking attested in Yoruba are semantically driven.  1.2  Clause structure  This section introduces the type of clause structure that is assumed for the analysis of Yoruba genitive constructions: the small clause structure (§1.2.1) and the relative clause structure (§1.2.2). I then introduce Yoruba D P structure (§1.2.3), and consider its relation to demonstratives (§1.2.4) and modification (§1.2.5).  1.2.1  Small clause  1  I propose a small clause (vP) along the lines of Stowell (1981) and subsequent works to account for "plain" nominal and verbal genitives. In the proposed structure, the possessor argument occupies the specifier of this functional .projection, vP taking the possessum argument as its complement (cf. Bowers 1993, Svenonius 1994), as in (10). Within this approach, I further propose that v is covert in the nominal genitive constructions and it is represented as [0]. However, in verbal possessives, v is overt and realized as a possessive verbm"have'. (10)  vP Possesor v  —---^ Possessum  As I show, this v P introduces a D P layer, and the possessum NP moves to Spec DP where D is overtly realized as the genitive morpheme (the M-tone syllable) that relates the possessum Some of the ideas of small clause developed here are taken from "Introduction to Layers in DP" (Zamparelli 1995).  1  and the possessor together in the surface syntax.  (11)  2  DP  -Poss'm _ |  Syntactically, I argue that the relation (R) between the possessor and the possessum arguments must be mediated by a functional element. When this functional element is covert its complement semantically has to move to a "stronger" layer (Zamparelli 1995), i.e. the DP, where it is pronounced as a genitive DP. But when the functional element is overt the complement need not move.  1.2.2  Relative clause  The relative clause analysis adopted in this dissertation accounts for genitive plus ti constructions. Ti constructions found within genitive DPs are analyzed as reduced relative clauses that lack an IP layer and where C is headed by a M-tone ti. This reduced C P is introduced above the small clause, as in (12).  (12)  a.  b.  ere ti Kunle statue C K. 'statue of Kunle' CP  •Poss'm  2  See Stowell (1981) for the adjunction analysis of arguments within a small clause.  In a genitive D P where the M T S and ti are both overt, there are three layers of phrases comprising the DP, the CP and the vP.  (13)  a.  owo o ti money MTS C 'Kunle's money  Kunle K.  •Poss'm  1.2.3  The Internal Structure of Yoruba DP  The first thing to observe is that Yoruba surface ordering is usually head-initial, (14).  (14)  a.  V  DP  je 'eat  isu yam'  b.  P  DP  si 'to  Fankufa Vancouver'  c.  I  VP  yoo 'will  je eat  isu yam'  d.  C  IP  pe that  mo I  je eat  isu yam  However, nominal expressions show mixed properties. Some nominals have head-initial ordering as in (15a), while others have a head-final ordering as in (15b).  (15)  a.  Q  NP  b.  N P Dem  gbogbo omo all child 'all of the children' pmo y i i child this 'this child'  I propose that despite appearances, Yoruba is consistently head-initial. This accords with Kayne's (1994) proposal that all languages have a Specifier-Head-Complement order (S-HC). I propose that in Yoruba, any departure from the Specifier-Head-Complement order arises from movement to a higher specifier. I give the full structure in (16). In (16), Sal refers to salience marker and is analyzed as an adjunct to D(em)P. See chapter 5.  (16)  D(em)P D(em)P^^^^SAL (Spec)  NP  Mod  Poss'm In the proposed structure, the highest nominal projection is D(em)P, which dominates DP. Thus, when we have a demonstrative within a nominal expression, the D P is the complement of Dem. The D P in turn dominates a small clause (vP), which contain the possessive phrase. As for modifiers, they are right adjoined to NP. What follows is a break down of this D(em)P structure. In (17a), the genitive DP contains only the Possessor and the Possessum with the intervening genitive marker. I propose that the Possessum moves to Spec DP in order to obtain the surface linear order of Possessum-Possessor, as in (17b).  (17)  a.  aja a Tisa dog M T S teacher 'the teacher's dog'  b.  DP  D  vP  MTS Poss'r Poss'm Tisa aja  (18a) is a genitive D P whose Possessum takes a modifier. This modifier follows the Possessor in the surface syntax. But the modifier actually modifies the Possessum. The occurrence of the modifier after the Possessor is due to movement of the Possessum to Spec DP, as in (18b). (18)  a.  aja a Tisa giga dog MTS teachertall 'the teacher's tall dog'  b.  DP  The genitive DP in (19a) has the specificity marker kan which modifies the possessum. In this case, the Possessum first moves to the Spec of its DP and further moves to Spec of the higher DP to derive the surface linear order where the specificity marker directly follows the Possessor, as in (19b).  (19)  a.  aja a Tisa kan dog M T S teacher SPF 'the teacher's certain dog (a certain dog of the teacher)' DP:  D MTS  vP Poss'r DP Tisa D  NP  kan  aja J  The genitive DP in (20) has a modifier giga 'tall' and the salience marker naa. I treat both as adjuncts. The modifier modifies the Possessum. Assuming that the salience marker is in Ccommanding relation to both the possessor and the possessum, I argue that it may mark either DP as salient, hence (20a) is ambiguous between 'the V E R Y dog' or 'the V E R Y teacher'.  (20)  a.  [aja dog (i) = (ii) =  a Tisa giga naa] M T S teachertall SAL ' the V E R Y tall dog of the teacher' 'the tall dog of the V E R Y teacher' DP  b.  SAL  DP  D MTS  vP  naa  Poss'r  giga  Finally, in (21) is a DemP that contains a genitive DP. I treat the demonstrative as a functional head and propose that the whole genitive DP moves to Spec DemP. Observe that there is also a movement that is internal to the genitive DP, namely movement of the Possessum from the complement position of v to Spec DP.  (21)  a.  aja a Tisa yii dog M T S teacher Dem 'this teacher's dog'  b.  DemP  1.2.3.1 DP and the D position  Yoruba nouns can be bare in the sense that they need not have overt determiners. Yoruba bare nouns may be construed as generic, indefinite, or definite. I propose that bare nouns that are construed as definite have a (null) discourse-linked determiner as in (22), in which case they may have the structure of a DP with a null D, which takes the N P as its complement (cf. Abney 1997, Longobardi 1994, and subsequent works). (22)  Structure for definite construal of bare N DP  D  NP  0  aja  1.2.3.2 Genitive structure in DP  Yoruba nominal genitive constructions are post-nominal in the sense that the possessor follows the possessum. In this case the possessors are marked for Genitive case by a mid tone vowel, henceforth referred to as MTS, which I analyze as the genitive marker.  (23)  a.  ile e Tiinde house MTS T. 'Tunde's house'  b.  ile e re house MTS 3sg 'his house'  I claim that a genitive phrase has the structure of a DP whose D takes the small clause as its complement. The MTS occupies this D position.  (24)  Yoruba Nominal Genitive DP DP  D  vP  Poss'm  Observe that in the proposed structure, the possessor precedes the possessum in the v P . However in the surface syntax, the possessum precedes the possessor. To account for that linear order, I propose a raising analysis, which raises the possessum to Spec DP.  (25)  1.2.4  Yoruba Nominal Genitive D P DP  D(emonstrative) and NP  Two types of elements fill the demonstrative position in Yoruba. The first type, which I call basic demonstratives are yii'this'  andyen 'that'. They can be marked for number by the  prefix plural morpheme won (see chapter 6).  (26)  a.  omo yen child Dem 'that child'  b.  omo wonyen child Dem 'those children'  The other type of element that occupies this position is the deictic element kan which marks nouns for specificity.  (27)  ile kan house certain 'a certain house'  I analyze kan as a deictic D glosses as SPF in (28). This proposal has support from the distribution of these elements: demonstratives and kan are in complementary distribution. (28)  a.  *ile y if kan house Dem SPF  b.  * ile kan house SPF  yif Dem  Assuming Yoruba is Head initial for all phrases, I propose that movement of the N P to Spec DemP derives the surface ordering of demonstratives relative to the NP. (29)  a.  DemP  omo 'child'  b.  J yen 'that'  DP  omo | 'child' kan ^ 'certain'  1.2.5  Modification  Here, I gave a fairly detailed account of modifiers. This is because in the body of the dissertation, there is no separate discussion of this category other than in chapter 6 where they are treated as a class of plural word. Yoruba modifiers appear after the noun that they 3  modify: (30)  a.  [Aja dudu.] je egungun dog black eat bone 'The black dog ate the bone.'  [NPMod]  b.  [Igi nla] wa m igbo oro tree big be Loc forest sacred 'There are big trees in the sacred forest.'  [NPMod]  Modifiers as used in this dissertation are the adjectives (cf. Bamgbose 1967; Avvobuluyi 1978; Hawkins 1983; Aboh 1999, 2000; Cinque 2004; Dixon 2004; Lindsey and Scancarelli 1985). On the division of this category into open class versus closed class see Cinque (2004, 2005).  3  These modifiers fall into four distinct semantic classes (31). (31)  a. b. c. d.  Colour Modifiers: {dudu 'black', fun fun 'white', pupa 'red' } Dimension Modifiers: {kekere, 'small' kukuru, 'short', nla, 'big'} Quality Modifiers: [buruku 'bad', dara-dara 'good', lile 'hard', rere 'good' pataki 'important', tuntun 'new'} Quantity Modifiers: {die 'few', p u p ? 'many', + N U M E R A L S }  They show the ordering restrictions in (32).  (32)  4  [Colour > Dimension > Quality > Quantity]  The following examples illustrate these co-occurrence restrictions. When there is more than one modifier, in an NP such that colour is among them, the colour modifier must precede any of the other modifiers, with dimension in (33a), quality (34a), and quantity, (35a).  (33)  (34)  (35)  a.  Colour > Dimension Mo ra aja [dudu kekere] lsg buy dog black small T bought a small black dog.'  b.  ^Dimension > Colour *aja [kekere dudu]  a.  Colour > Quality Olu k6 fe omobinrin [dudu buruku] PN neg marry girl black bad 'Olu did not marry that nasty dark-in-complexion girl.'  yen Dem  b.  * Quality > Colour *aja [buruku dodo]  a.  Colour > Quantity Iya oniso-eran ni aja [dudu pupo] mother animal-trader have dog black many 'The woman who trades in domestic animals has many black dogs.'  b.  * Quantity > Colour *aja [pupo dudu]  However there is some variation in these restrictions as some of the starred orders are acceptable to some speakers of Yoruba.  4  With dimension, quality, and quantity modifiers, dimension must precede both quality (36a) and quantity (37a).  (36)  (37)  a.  Dimension > Quality N k6 fe aja [kekere lsg neg want dog small T do not want this ugly small dog.'  b.  * Quality > Dimension *aja [buruku kekere]  a.  Dimension > Quantity Mo ra isu [nla lsg buy yam big 'I bought few big yams.'  b.  buruku] bad  yii Dem  die] few  Quantity > Dimension *isu [die nla]  Further, when quality and quantity both modify an NP, the quality modifier must precede the quantity modifier.  (38)  a.  b.  Quality > Quantity Aja [buruku meje] wa nf dog bad seven be Loc 'There are seven terrible dogs in this city.' *aja  meje  ilu yif town Dem  buruku  Finally, a combination of colour, dimension and quality with an N P shows the co-occurrence order in (39a); colour must precede dimension, which in turn must precede quality. Finally, when all the modifiers combine with an NP, the ordering restriction that must be followed is illustrated in (39b).  (39)  a.  Colour > Dimension > Quality Q w o te oluko [dudu kekereburuku] yen hand reach teacher black small bad Dem 'That nasty small dark-in-complexion teacher is in trouble.'  b.  Colour > Dimension > Quality > Quantity Qba a fun Gomina ni king HTS give governor P esin [funfun nla daradara rnejo] horse white big nice eight 'The king gave the governor eight nice big white horses.'  I put forward two possible analyses of modifiers. They can either be analyzed as basegenerated adjuncts or as functional heads. If modifiers are adjuncts, then this gives the structure in (40) assuming right adjunction (cf. Cinque 1993). The entire D P raises to Spec DemP.)  (40)  hypothesis 1: modifiers as base-generated adjuncts (cf. Dechaine 1993) DemP (Spec) Dem  DP D ^ ^ ^ ^ ^ N P NP Mod esin [funfun nla daradara mejo]  The implications of modifier stacking for adjunct analysis is that one needs to stipulate a Yoruba-specific modifier sequence: Colour > Dimension > Quality > Quantity; which is the mirror image of English modifier sequences.  (41)  a.  Yoruba modifier sequence: [Colour > Dimension > Quality > Quantity]  b.  English modifier sequence: [Quantity > Quality > Dimension > Colour]  The examples in (42) and (43) illustrate the sequences in Yoruba and English.  (42)  Yoruba Qba king fun P  ra esin [funfun buy horse white aare Obasanjo president O.  nla big  daradara nice  rnejo] eight  (43)  English The king bought [eight nice big white] horses for President Obasanjo. The mirror image ordering could be taken as evidence that syntactic movement has  applied in Yoruba. And if that is the case, it is a good case to make for an analysis that treats modifiers as functional heads, which introduce a DP complement. The analysis proceeds as 5  follows. First the DP moves to Spec ModP and second, the ModP raises to Spec DemP. This corresponds to "snowball" movement (Aboh 1999). Snowballing as developed in Aboh (1999, 2004) is a syntactic mechanism that allows pied-piping of a maximal projection. In particular, it involves leftward movement of an NP to the specifier position of a higher phrase, call it YP, and the whole Y P which now harbors the NP and its complement moves to the specifier position of yet another higher phrase, call it XP. Depending on the number of intervening functional heads, snowballing continues until it gets to the Spec of the highest phrase. This contrasts with cyclic movement, which involves movement of a given phrase to higher specifier positions in a successive manner without having to be accompanied by the intervening host at the intermediate landing sites. For details, see Aboh (1999, 2004). The tree structure in (44) illustrates how the snowball movement derives the Yoruba data above.  Consistent with the idea that modifiers are functional heads is the fact they form a closed class in Yoruba (Abraham 1958; Bamgbose 1967; Welmers 1973; Awobukiyi 1978). 5  (44)  a.  DemP  b.  esin funfun nla daradara horse white big good 'eight nice, big white horses.'  mejo eight  ,  I retain the adjunct-based analysis of modifiers because it is able to account for why plural words/morphemes must adjoin to the lexical items they pluralize (chapter 6). In summary, this chapter has introduced us to the four related topics on nominal expressions in Yoruba that I cover in the entire dissertation as well the analyses that account for each of them. In the following chapters, I provide both empirical and theoretical evidence for each proposal.  CHAPTER TWO:  2  Y O R U B A G E N I T V E CONSTRUCTIONS  Introduction  This chapter discusses genitive constructions in Yoruba. It explores four themes. §2.1 accounts for the semantics of the relation between the two arguments that are in a genitive construction, i.e., the R-relation. In §2.2, I present the syntax of the R relation, while the focus of §2.3 is the syntax of Yoruba nominal genitives. The concern of §2.4 is how the syntax of Yoruba nominal genitives interacts with phonology. I conclude my findings in §2.5.  2.1  The semantics of the R relation  The terms "genitive" and "possessive" as used in this dissertation refer to constructions where two simple nouns enter into a relation with one another (Storto 2003). I first show in §2.1.1 that the R-relation is either pragmatically determined (via discourse-linking) as in (1) or lexically determined via the inherent meaning of relational nouns and inalienable bodypart nouns, as in (2). In this case it is the inherent meaning of the possessum that determines the kind of relation that holds between it and the possessor. Conflation or incorporation  (1)  Iwe e Tiinde book M T S T. 'Tunde's book'  discourse-linking  (2)  a.  baba a Tiinde father MTS T. 'Tunde's father'  relational noun  b.  apa a Tiinde arm M T S T. 'Tunde's arm'  inalienable body-part noun  I then address the distinction between the terms "possessive" and "genitive": while the former describes the semantics of the R-relation (§2.1.1), the latter describes the morphosyntax of the R-relation (§2.1.2). I introduce the terms "Possessor" and "Possessum" and show how they map onto the R-relation (§2.1.3). The section closes with a discussion of the morphosyntactic realization of the R-relation in Yoruba (§2.1.4).  2.1.1  Three kinds of R-relation  Nouns enter in a relation with each other in one of three ways. First is the kind of R relation that is Discourse-linked. According to Higginbotham (1983), an example such as (3a) is interpreted as in (3b).  (3)  a.  iwe e Tunde book- MTS T. 'Tunde's book'  b.  3x [iwe (x) A R(T,x)| = there exists x, x is a book and x stands in some relation R to Tunde  The R relation has its value supplied by discourse. This is the sense in which it is said to be Discourse linked (D-linked).  (4)  Simple N Possession R iwe  To say that the R-relation is D-linked means that iwe e Tunde 'Tunde's book' needs a discourse context for the relation to be understood. For example, depending on the context, the book in question could be 'the book that Tunde read', 'the book that Tunde drew a picture o f , 'the book that Tunde sang a song about', 'the book that Tunde owns', or even 'the book that Tunde threw into the ocean when he was fishing'.  With the pragmatically determined R relation, genitive constructions fall into three subtypes: genitives of possession, of depiction and of modification. While the genitive of possession in (5a) corresponds to the semantic notion of possession, the genitive of depiction in (5b) is ambiguous between possession and depiction. As for the genitive of modification in (5c), it is never construed as a possessive relation (cf. Partee and Borschev 1999: 174).  (5)  a.  ile e Tiinde house M T S T. 'Tunde's house'  genitive of possession  b.  aworan an Tiinde picture MTS T. 'John's picture'  genitive of depiction  c.  ogbontagi onfrdyin gem journalist 'a gem of a Journalist'  genitive of modification  These three subtypes of genitive construction are all instances of the relation R. I concentrate mainly on the genitive of possession and the genitive of depiction in this dissertation. In addition to these pragmatically determined R-relations, another R relation is found with relational nouns like babd 'father', where the relation is supplied by the meaning of the noun itself. Thus, R is lexically determined. In (5), under its most salient reading, babd means 'someone who stands in the father-of relation'. 'Tunde's father' doesn't need a discourse context for the relation to be understood: the person in question is unambiguously understood to be the one who stands in the father-of relation to Tunde (cf. Dechaine 1993: 127).  (6)  a.  baba a Tiinde father M T S T. 'Tunde's father'  b.  3x [baba (x) A baba (x,T) = There exists x, x is baba 'a father' such that x is the father of Tunde  When we say that R is lexically determined, the idea is that for relational nouns such as baba 'father', it is the noun itself that supplies the relation R, as in (7).  (7)  1  Relational Nouns baba R  The third type of R relation is found with inalienable nouns, which include body-part nouns such as apa 'arm' (8) (cf. Vergnaud and Zubizarreta 1992; Miihibauer 2004).  (8)  a.  apa a Tunde arm M T S T. 'Tunde's arm'  b.  3x [apa (x) A apa (x,T) = There exists x, x is apa 'an arm' such that x is the arm of Tunde  Such inalienable nouns have a special part-of relation (Dechaine 1993: 127-133; Miihibauer 2004a,b; 2005; Partee and Borschev 1999). Observe that the inalienable R relation is not discourse linked. Thus, as with relational nouns, the inalienable R relation is also lexically determined as in (9). In apa a Tunde 'Tunde's arm', this means the arm that is a part of Tunde. It's a component of his body, under its most salient reading.  (9)  2  Inalienable Nouns R is lexically determined apa R To summarize, simple N possession has an open semantic relation R whose content is  provided by the context (i.e. pragmatically determined). For the other two types of genitives, 'However, discourse linking is also possible with relational nouns in an appropriate context. Imagine the following scenario. "Every member of a community centre's 'Father's Group' was assigned a counselor. Counselor Tunde's father was very enthusiastic about the centre's program but Counselor Titi's father had no real interest in the group at all." In this context, discourse linking would take precedence over the kinship interpretation. However, such forced contexts are not considered here. See Burton (1995) for a discussion of the less salient readings of inalienable nouns. 2  namely relational nouns and inalienable body part nouns, the R relation is lexically determined.  2.1.2  "Possessive" versus "Genitive"  Attempts have been made to draw a distinction between "possessive" and "genitive" in the literature. For example, in semantics (Borer 2004), this kind of phrase is referred to as possessive because the focus is on meaning. Once possessive is mentioned, semantic consideration necessarily comes in, whereas "genitive" denotes a morphosyntactic (not semantic) relationship between a nominal and some other item (be it noun or verb). Following standard practice, I analyze genitive in terms of Case (Lindauer 1998). 3  Following from what is said above, one might conclude that the terms "genitive" and "possessive" are different names for the same kind of nominal expression. Observe that just as the nominal R relation is compatible with a wide range of pragmatically conditioned interpretations, as in (10a), so too is the verbal R relation, as in (10b).  (10)  a.  nominal R-relation (possessor/genitive) iwe e Tunde book MTS T. (i) 'the book that Tunde owns' (ii) 'the book that Tunde wrote' (iii) 'the book about Tunde'  b.  verbal R-relation (possessor/genitive) Tunde ni iwe T. have book (i) 'Tunde owns a book' (ii) 'Tunde wrote a book' (iii) 'Someone wrote a book about Tunde'  There is even a trend to decompose possession syntactically into location, BE and HAVE. See Benveniste (1971); Freeze (1992); Harley (1995); Partee and Borschev (2002). 3  Furthermore, just as lexical properties of nouns determine the R-relation, so too do lexical properties of arguments determine the relation expressed by a light verb. For example, in 4  (11a) the R-relation is pragmatically determined by discourse-linking while in (llb-c) the Rrelation is determined by the lexical properties of the relation and inalienable noun respectively. (11)  a.  bata a Tiinde shoe MTS T. 'Tunde's shoes'  discourse-linking  b.  egbon on Tiinde elder MTS T. 'Tunde's brother'  relational  c.  apa a Tiinde arm M T S T. 'Tunde's arm'  inalienable  Similarly, in (12) it is the inherent lexical properties of the object argument that determine how light verb je 'eat, ingest' is interpreted. In (12a), je isu translates to 'eat yam'; in (12b) je gbese translates literally as 'eat debt', i.e. to be in debt; in (12c) je egba 'eat cane' translates as 'to be caned'. Thus, just as the co-argument relation between [ A R G l v ARG2] is determined by the lexical properties of [ARG2] in the Possessive/genitive constructions of (11) , so too is the co-argument relation between [ARGl je ARG2] determined by the lexical properties of [ARG2] in (12).  (12)  a.  Bolii je isu B. eat yam 'Bolu ate yams.'  b.  Bolii je gbese B. eat debt 'Bolu is in debt.'  agent-theme  experiencer-theme  I define light verbs as verbs whose meaning and valence are determined exclusively by nouns with which they combine. 4  c.  Bolii je egba experiencer-theme B. eat cane 'Bolu received some strokes of cane.'  In Yoruba, it is possible to have a genitive relation without possession. Conversely all cases of nominal possession show some kind of genitive relation. In (13a), apb 'a bag' cannot be possessed by bata 'shoes' in any obvious way, although the shoes stand in a relation to the bag, e.g. by occupying the bag in spatial terms. But even though possession as such is not found in (13a) it is no less genitive than (13b) which can have a clearly possessive construal along the lines of 'the money that Tunde has'.  (13)  a.  [apo o bata] bag M T S shoe 'a bag of shoes'  b.  [owo o Tunde] Money MTS T. 'Tunde's money'  Based on (13), the conclusion must be that the notion of possession comes is slightly different from that of genitive. Another case to consider is shown in (14). In (14a), both nouns are animate. In that case Tunde does not possess his boss, but instead the genitive simply allows for the satisfaction of the R relation, since if you are a boss, you must be somebody's boss. Further, there are cases like (14b) where apd 'arm' contextually means 'sleeve'; metaphorically the gown can be thought to possess the sleeve just as I possess my own arm via the part-whole relation (cf. Miihibauer 2004; Partee and Borschev 1999).  (14)  a.  [oga a Tunde] master MTS T. 'Tunde's master'  b.  [apa (a) agbada] arm MTS garment 'a garment's sleeve'  What is reported of the genitive and possessive parallels the case of nominative Case and the external theta-role of the verb is. While many arguments with nominative Case are also agents, and many agents bear nominative Case, the correlation is not on one-to-one basis. For example, in (15a), Jenrola is agent, in (15b) the same Jenrola is the experiencer, whereas in (15c) Jenrola  is the possessor. In all three examples, Jenrola  is assigned  Nominative Case.  (15)  a  Jenrola je akara J. eat bean cake 'Jenrola ate the bean cake' 0: Agent Case: Nominative  b.  Jenrola feran akara Jenrola likes bean cake 'Jenrola likes bean cakes.' 0: Experiencer Case: Nominative  c.  Jenrola ni akara J. have bean cake 'Jenrola has a bean cake.' 0: Possessor Case: Nominative  Delsing (1998: 93-94) observes that though "the relation between possessor and possessum is semantically different from the relation between the subject and the verb, it seems to be morphosyntactically the same relation." On that view, genitive may be defined in terms of structural Case.  2.1.3  Possessor and Possessum  The other two terms that I will frequently use are Possessor and Possessum. A possessum NP does not necessarily mean an item possessed by X . Neither is it the case that a possessor NP always refers to the one who possesses X . Rather, they simply represent the two arguments that are in a genitive relation.  In the structure that I propose, which for now I call X P , the Possessor occupies the Specifier position in the base structure whereas the Possessum is in the complement position.  (16)  XP Possessor (Specifier) R  Possessum (Complement)  I claim that this Possessor-Possessum structure is the basis of all genitive constructions in Yoruba.  2.1.4  The three types of genitive constructions in Yoruba  There are three kinds of genitive constructions in Yoruba: (17)  a.  Nominal genitive  b. c.  Verbal genitive Nominal plus ti construction  The examples in (18) illustrate these three types. In (18a) the possessum and the possessor are separated by a mid tone mora (ji) herein referred to as the genitive marker (henceforth mid tone syllable (MTS)). The example in (18b) is the sentential counterpart of (18a). It is the kind of possessive that is assigned by the verb ni 'have'. Similarly, the example in (18c) parallels the example in (18a) in the sense that the possessum and the possessor are also separated by a mid tone element, here ti. (18)  a.  b.  Nominal genitive ile e Tiinde house M T S T. 'Tunde's house' Verbal genitive Tunde ni ile T. have house 'Tunde has a house.'  possessum  MTS  possessor  possessor  verb possessum  c.  Nominal plus ti construction eran ti Tunde animal of T. 'Tunde's animal'  possessum  C  possessor  In what follows, I sketch out a syntactic analysis of the R relation. After that, I proceed to account for the first two types of genitive constructions shown in (18a-b), namely nominal genitives and verbal genitives. The third type of genitive construction, namely the nominal plus ti construction, is treated in chapter 3.  i 2.2  The syntax of the R relation  A genitive construction establishes a relation R between two nouns: the possessor and the possessum. The syntax of the Possessor-Possessum relation is treated in one of two ways in the literature. On one view, the Possessor-Possessum relation corresponds to a co-argument structure in the syntax (Kayne 1993, 1994; Cinque 2003 among others). On another view, the Possessor-Possessum relation corresponds to a Head-Complement structure in the syntax (Jakendoff 1977). I discuss both proposals and adopt the co-argument analysis for Yoruba.  2.2.1  Possessor-Possessum as a co-argument relation  The co-argument relation analysis that I adopt is found in Kayne (1994). In this approach an abstract head mediates the relationship between the two arguments. Motivating a unified analysis for genitive constructions, Kayne (1994) argues that both pre-nominal and postnominal genitives derive from one source. The examples in (19) with the corresponding 5  structures in (20) illustrate this.  In the Linear Correspondence Axiom (LCA), Kayne (1994) claims that "Heads must always precede their associated complement position...specifier positions must invariably appear to the left of their associated head never to the right" (Kayne 1994: xiii). This proposal makes the claim that the universal ordering between a head and its dependents is Specifier-Head-Complement (S-H-C), as in (i). 5  (19)  a. b.  John's book: [ ...D° [John ['s book]]] book of John: [^..book, [ of [John ['s tj]]] DP  D  The structure in (20a) base generates the prenominal genitive that is marked by the Saxon genitive '5. To derive the post-nominal genitive, Kayne assumes that a D/P layer is introduced and the possessum moves to Spec DP/PP while ' o f occupies D/P; as in (20b). (20)  a.  DP/PP  D/P^^^^FP  book  b.  DP/PP book, D/P of  FP John F  's  *i (adapted from Kayne 1994)  In this dissertation, following Kayne's proposal, I assume that all Yoruba genitives originate from the same base phrase structure. This phrase structure consists of the possessor and the possessum, which stand in a co-argument relation.  (i)  Kayne's 1994 LCA XP (Specifier)  XP (Head)  (Complement)  X° YP YP, the complement of X" is a sister to this Head and is dominated by the XP. This implies that the kind of linear order that one would expect assuming this structure will be such that a complement must always follow its associated head and that specifier always precedes the phrase it is sister to thus giving S-H-C linear order.  2.2.1.1 Base structure for Yoruba genitives I propose a small clause vP as the base structure for all genitive phrases in Yoruba along the lines of Stowell (1981, 1983), Chomsky (1981), Haegeman (1991), Dechaine (1993), and Radford (1997) among others. First, I define a 'Small Clause' as a defective nominal or verbal clause containing two arguments and a relational head that may but need not be pronounced (cf. Stowell 1981, 1983, Chomsky 1981). On this assumption, the possessor, 6  which is one of the arguments, occupies Spec v P and the possessum, which is the other argument, is the complement of this relational Head (v), as in (21). Observe that this structure parallels the relational structure above. Thus v is the syntactic counterpart of the semantic R relation.  (21)  Base Structure for genitive constructions vP Possessor  — v  Possessum  There are two things that need to be pointed out. First, the same structure can account for both nominal and verbal possessives. Second, the relational head can be an abstract v in which case it is covert, i.e. a null relational verb (cf. Radford 1997: 201). It can also be a lexical head in that case it is overt (cf. Freeze 1992, Kayne 1993 and Harley 1995). Precisely, when the vP is a defective nominal clause, I propose that v is null, as in (22).  7  In the cited works, the phrases in brackets in (i) are arguably small clauses, with the subject argument in boldface. (i) a. I consider [John a friend] b. I expected [Serena to win the tournament] The other type of small clause according to Stowell (1978), Couquaux (1982), and Burzio (1986) consists of copular sentences, which have their base as small-clauses under "be" in IP. The example in (ii) illustrates this. The idea is that (ii-a) originates as a small clause in (ii-b), and that the argument raises to [Spec IP] to satisfy the Extended Projection Principle and get Case ii-c). (ii) a. John is the best man in town. b. [IP is [SC [John] [the best man.in town]]] c. [[John]_i is [SC t_i [the best man in town]]] (Zamparelli 1995) • 'Assuming that the R-relation corresponds to a head, we expect that in compounding of the type bookstore such a head is present as well. Possible evidence for such a head might come from German where in many of such compound we find an interfix Hahn-en-fuss (rooster leg) (Wiltschko personal communication). 6  vP  (22) Possessor  v  Possessum  0 On the other hand when v P is a verbal clause, v is pronounced as a possessive verb, as in (23). (23)  Possessum ni The split between overt v (ni) and covert v (0) in Yoruba has empirical support from language-internal evidence. It reflects a general parallelism that occurs between N and V . Further, as I demonstrate, the fact that v is null in a defective nominal clause leads to the obligatory raising of the possessum from the complement position to a higher position from where it can receive Case. T h e " v P " label taken at surface value may seem to be misleading since not all genitives can be described as verbal predicates. However, in this dissertation I use it as a cover term that refers to the base structure for nominal or verbal possessives. I could as well propose a different label like nP for the defective nominal clause, but I choose not to for the sake of capturing the generalization. With one base structure I have been able to show that there is the same relation (R) between two arguments in both nominal and verbal possessives. Thus, the head v is the syntactic equivalent of the R that relates the possessor to the possessum in semantics (cf. Hertz 1997: 513). Note that the specific label used for genitive constructions is not of any serious concern or problem to linguists, as reflected by the fact that there is no uniform label in the literature. For example, in many of the articles that appear in Alexiadou and Wilder (1998), which treats possessive constructions across languages, this phrase is mostly referred to as "Possessive", with "Poss" as the head, (24a). In Abney (1987), and Siloni (1997), the same construction is called a DP, and the Saxon genitive's is treated as D, a functional element,  which introduces another argument, (24b). In Hertz (1997) it is simply called an NP, (24c). Yet it is possible to label a genitive phrase simply an FP where " F " is an unspecified functional morpheme, (24d) (cf. Fukui 1986). Lastly, the same phrase has been referred to as an AgrP (Kayne (1994) where the possessor occupies the Spec AgrP and the possessum serves as the complement of the Agr, (24e). The aim of labeling a possessive phrase an AgrP is to show that there is some kind of agreement between the functional element and the NP in its Spec. Such proposal is unquestionable in languages that show agreement. The present proposal identifies R with a null v in nominal contexts, (24f).  [Poss'm]  Alexiadou & Wilder 1998  ID'S]  [Poss'm]]  Abney 1987, Siloni 1997  [NP Poss'r  [ ]  [Poss'm]]  Hertz 1997  d.  [FP Poss'r  [F'S]  [Poss'm]]  Fukui1986  e.  [  Agr  Ugr 'S ]  [Poss'm]]  Kayne 1994  f.  [  vP  [v0]  [Poss'm]]  present proposal  a.  [POSSP  b.  [DP Poss'r  c.  p  Poss'r  Poss'r  Poss'r  [POSS  ]  The other question that may arise is why can't we have a simple structure such as (25) where the possessor and possessum are simply sisters to each other? (25)  ^ Poss'r  Poss'm  First, the small clause analysis proposed here accounts for both nominal and verbal genitives with the same structure. The structure in (25) has no place for verbs. I have argued that a genitive phrase involves a relation (R) between two arguments (possessum and possessor) mediated by an intervening head, which may be overt or covert. The binary structure in (25) does not show such a relation. In syntax, the relationship between two nonidentical entities such as the one under review is best expressed through the use of an intervening head, which projects its own phrase. In my view, the structure in (25) is tenable when the two nouns refer to one entity such that one is saturating the other, as in the case of  appositives in the English examples in (26); or with identificational statements such as (27) and (28) from Blackfoot and Cree respectively.  (26)  a. b.  (27)  Anna Jeff this J. 'That is Jeff.'  (28)  [John the tailor] is here, I saw [my friend Jeff]. Blackfoot (Algonquian) (Miihibauer, personal communication)  Minos awa cat this  Cree (Algonquian)  'This is a cat.'  (Miihibauer, personal communication)  The corresponding structure for the English example in (26a) is given as (29). (29)  Appositive DPI  DP2  John  the tailor  DPI = DP2  However, genitive constructions are quite different from appositives, as there is a non-identity condition in the sense that the nouns that stand in the R relation are two distinct entities: one the possessor, the other the possessum. Thus in Jeff's dog, "Jeff" can be represented as "y" while "dog" can be represented as "x" in which case, "y * x". In nominal genitives, Yoruba has a genitive marker in the form of a M i d Tone Syllable (MTS) that shows up between the two arguments. The presence of the M T S 8  correlates with the fact that an R-relation holds between the nouns even though the MTS is  Note that it is not in every language that the genitive marker shows up between the possessor and the possessum. In Hungarian the genitive -ja marker comes after the possessum. (i) a. a te kalap-ja-i-d the you hat-POSS-PL-eSG b. (a) Mari kalap-ja-i-0 the Mary hat-POSS-PL-3SG 'Mary's hats' (Alexiadou and Wilder 1998: 3, ex.3)  not pronounced in "v", the syntactic head for the two arguments. Evidence for this claim comes from "v" which is pronounced as w i n verbal genitives.  (30)  a.  iwe e Tiinde book MTS T. 'Tunde's book'  b.  baba a Tunde father MTS T. 'Tunde's father'  c.  apa a Tiinde arm M T S T. 'Tunde's arm'  There is language-internal evidence in support of the type of structure that I propose. In verbal possessives, the two arguments are related by the intervening verb, which I claim is the spell-out of the relational v.  (31)  a.  Tunde ni iwe T. have book 'Tunde has a book.'  b.  Tiinde ni baba T. have father 'Tunde has a father.'  c.  Tiinde ni apa T. have arm 'Tunde has an arm.'  The fact that in verbal possessives v is pronounced is evidence that the possessor and possessum are in a co-argument relation. In the corresponding nominal possessives there is an abstract head linking the two nouns. The reason that a simple binary structure is inadequate as an analysis of genitive constructions therefore is that it consistently involves the presence of two referentially dependent nouns.  2.2.1.2 Other languages with overt genitive morphemes There is also cross-linguistic evidence in support of the small clause structure as there are a lot of languages that have overt morphemes for the genitive marker, e.g. Hebrew, Turkish, Japanese, Blackfoot and Cree among others. In (32a), the Hebrew Construct state is marked by sel. In (32b), the Japanese genitive marker is no. The same genitive marker is realized as -m and -im in Blackfoot (32c) and Cree (32d) respectively.  (32)  a.  b.  c.  d.  ha-bayit sel ha-'isa the-house of the-woman 'the woman's house'  Hebrew (Siloni 1997: 24 ex. 10c)  masaru-no hon masaru-Gen book 'Masaru's book'  Japanese  nisapikitsox my 'my ring'  Blackfoot  tsaatsi-m ring-poss  o-minos-im-a 3sg-cat-poss-obv 'his/her cat'  (Kiyota, personal communication)  (Muhlbauer, personal communication) Cree (Muhlbauer, personal communication)  In all of the languages above, each genitive phrase consists of two nominal arguments that are in relation as shown by the presence of the genitive marker. It follows then that an adequate structure must reflect such a relation and represent all the constituents (in this case, the two arguments and the genitive marker/verbal element). This is readily available in the small clause analysis that I propose. Observe that in the present analysis, the relational head v may be null. Regarding the examples in (32), there remains the question of whether the genitive marker (Hebrew sel, Japanese -no, Blackfoot -m, Cree -im) spells out the v position, or a higher D position. As we shall see for Yoruba, there are reasons to think that these genitive morphemes spell out an F-head, i.e., D. Relevant to the co-argument analysis is the fact that the relation between the possessor and possessum is transparently represented in the syntax.  2.2.2  Possessor-Possessum as a Head-Complement relation  Under X-bar theory (cf. Jakendoff 1977) the possessum and the possessor are standardly analyzed in terms of the Head-Complement relation. In this analysis, the possessum, which is the head, introduces the possessor as its complement. In this account, there is no separate node for the genitive marker even when the NP is preposed. In (33a), the complement NP (John) remains in its base position yielding a post-nominal genitive marked by ' o f . In (33b), the complement N P raises to Spec NP, yielding a prenominal genitive phrase marked by the Saxon genitive s. ,  (33)  a.  NP N' N ^ | book  ^ N P (Complement) (of) John  b.  NP N P f ' V (Spec) ...... /X N John's | house  '  (Compl) tj (cf. Jakendoff 1977)  Both the preposition 'of and the Saxon's in English are treated as part of the possessor NP. Applying this to Yoruba would mean treating both the MTS and ti as part of the possessor, as in (34). (34)  a.  NP N' N ^ | iwe 'book'  ^ N P (Complement) / \ (e) Tunde (ti) Tunde (e ti) Tiinde  Second, it would represent the possessor as the complement, and third it would mean merging the head, in this case D/C (in the form of the MTS and/or ti) with the complement. ,  2.2.3  Parallels between V-syntax and N-syntax  This section discusses parallels that exist between V-syntax and N-syntax as regards the R relation. In (35a), there is a verb that separates the possessor and the possessum. In (35b), there is a M-tone mora (pi), which I analyze as the genitive marker henceforth (MTS). This marker separates the possessum and the possessor.  (35)  a.  Lalupon nf onfbode L. have gate-man 'Lalupon has a gateman.'  Possessor  verb  Possessum  b.  onfbode e Lalupon house MTS L . 'Laiupon's gateman'  Possessum  MTS  Possessor  (35a) is a verbal genitive, while (35b) is a nominal genitive. On conceptual grounds we expect these parallels between N and V structures inasmuch as both express the co-argument relation that holds between a Possessor and a Possessum. I argue that the difference in linear order in the surface syntax between the V-syntax (where the possessor precedes the possessum) and N-syntax (where the possessum precedes the possessor) can be accounted for via a raising analysis. See chapter 3 for an account of Case assignment.  2.2.3.1 Parallels between the verbal possessive and the nominal genitive Languages have certain sentences that express possession. Such expressions of possession are in the sense of owner HAVE ownee relation (Kayne 1993, Harley 1995: 107, Manfredi 1994). This translates to possessor R possessum in the present analysis. In this way, nominal 9  possessives can be viewed as equivalent to verbal possessives. Thus, John's book is parallel  to 'John has a book' (Freeze 1992, Partee and Borschev 1999). The parallel between V-possessive and N-genitive is also contained in the works of Szabolcsi's (1983) on Hungarian and Freeze (1992), Kayne's (1993, 1994) and Alexiadou and Wilder's (1998) work on English. In their analyses, the claim is that the possessive DP (36a) and the possessive clause (36b) are related as in (37). In the possessive DP, the possessor moves to Spec AgrP (37a). In the verbal possessive, the possessor first moves to Spec AgrP and then moves to Spec IP, (37b).  (36)  a. b.  John's three books (are on the table). (A & W 1998: 6, 8a) John has three books. (A & W 1998: 6, 8b)  (37)  a. b.  [ D° [John Agr° [ three books]]] (A & W 1998: 6, 9a) [ Johnj has [ D° [tj Agr° [ three book]]]] (A& W 1998: 6, 9b) DP  IP  QP/NP  DP  QP/NP  In my analysis, I propose that the possessor raises to Spec, D P rather than Spec AgrP. Observe that Yoruba verbal possessives are of two types: the canonical ni H A V E construction in (38a), and the possessive dative je ti ('be to') in (38b). (38)  a.  Tiinde ni ile T. H A V E house 'Tunde has a house.'  b.  lie je ti Tiinde" house be of T. 'The house belongs to Tunde.'  With nominal possessives, we observe a Possessum Possessor order, (39).  (39)  ile e Tiinde house M T S T. 'Tunde's house'  The predicate element HAVE is interpreted in various ways in the literature: as auxiliary or verb (Kayne 1993), existential (Partee 2004, Szabolcsi 1994), locative or possessive (Freeze 1992) There is another verbal possessive ni. It is used in restricted context of 'his body' shown in (i). (i) Ebun un ni ara a re E. HTS own body MTS 3sg 'Ebun owns her body.' " This example is not discussed in this chapter because of its internal structure, especially the presence of the tielement. See chapter 3 for an account of f/-genitives. 9  10  If the parallelism between verbal and nominal genitives holds, one expects that verbal H A V E possessives will have the same structure as their nominal counterparts, as in (40).  (40)  a. Poss'r Tiinde  v ni  b.  ^ Poss'm ile  vP Poss'r Tiinde  v  Poss'm  MTS  ile  While the structure in (40a) is what is actually obtained in the surface syntax, (40b) is not attested, i.e. the M T S does not spell-out the relational v. As I show in the next section, although I still claim that both verbal and nominal genitives make use of the same base structure, v is not pronounced in nominal genitives in Yoruba, and the M T S spells out a higher functional head (namely D) but I am assuming that whenever we have a genitive DP the v is there even though it is not pronounced.  12  For verbal possessives, I adopt the base structure u.P in (40a). I argue, based on the data presented, that the verb ni occupies u and takes the possessum ile as its complement. On the other hand, Tiinde, the possessor, occupies Spec uP (cf. Szabolsci 1983, Alexiadou and Wilder 1998). One final point that I want to make about the account of verbal possessives is that the possessor NP can raise out of Spec uP. In that case, there will be an IP layer above the base structure. In the nominal genitives, it is the possessum NP that raises out of the uP. This is a core difference between nominal and verbal genitives.  The relation between D, v and R therefore is that at the semantic level, R is the head of the possession phrase. At the syntactic level, R is pronounced in v in verbal genitives whereas in nominal genitives, R is pronounced inD. 12  2.2.3.2 Parallels between IP-syntax and DP-syntax In this section, I present language-internal evidence in support of the proposed base structure based on the parallelism between IP-syntax and DP-syntax. This evidence comes from the grammatical particles that mark the two constructions. Recall that both the IP and the DP share the same base structure. In the IP domain, the head I(nfl) which is spelt out as HTS takes this vP base structure as its complement. Similarly, in the D P the head D, which is spelt out as M T S element takes the same vP base structure as its complement. One point that I want to emphasize about the account of verbal possessives is that the possessor raises out of the Spec vP to Spec IP to receive the N O M case, (41a). In nominal genitives, it is the possessum that raises to Spec D P to receive N O M / A C C Case depending on its syntactic position (41b). See chapter 3 for details. (41)  42  a.  IP-syntactic structure IP  b.  Gbada a nf G. HTS have 'Gbada has a house.'  a.  DP-syntactic structure DP  ile house.  b.  ile e Gbada house MTS G. 'Gbada's house'  I claim that the MTS is a syntactically conditioned phonological spell out of an F-head within the nominal phrase, (DP). On the other hand, the HTS in "I" is a syntactically conditioned phonological spell out of an F-head within the sentential phrase (IP) (cf. Dechaine 1993 among others).  13  I have established the existence of parallels between the verbal possessive and the nominal genitive. Languages have certain sentences that express possession in the sense of owner HAVE ownee relation, and this translates to possessor R possessum. I have equally established another parallel between IP-syntax and DP-syntax in support of the proposed base structure with evidence from the grammatical particles that mark the two constructions i.e., just as the head I(nfl) is spelt out as HTS and it takes this u P base structure as its complement, the head D, which is spelt out as M T S element, takes the same vP base structure as its complement.  2.3  The Yoruba nominal genitive: syntax  In this section, I present a detailed syntactic account of Yoruba nominal genitive constructions. First, I show that genitive constructions in this language involve a co-argument relation, §2.3.1. The analysis that I propose shows that the possessum NP must raise; §2.3.2. §2.3.3 gives various accounts of what determines movement of N or NP. The focus of §2.3.4  When the HTS occurs after the subject DP, it marks a non-future tense, (i). This contrasts with the presence of yoo, which marks the future tense, (ii). See Barczak (2004) for detailed discussion. (i) Ebun un sun E. HTS sleep = (a) 'Ebun is sleeping (+present).' = (b) 'Ebun slept.' (ii) . Ebun yoo sun E. FUT sleep 'Ebun will sleep.' This is the more reason why the two grammatical particles cannot co-occur, (iii). (iii) a. *Ebun un yoo sun b. *Ebun yoo un sun 13  is to show the parallels between N-syntax and V-syntax. The discussion in this section comes to ah end in §2.3.5 with the account of genitive pronouns.  2.3.1  Yoruba genitive constructions involve a co-argument relation  My goal is to extend Kayne's co-argument analysis to Yoruba genitive constructions involving the M tone mora, the verbal element, the M tone ti, and relate the last to the H tone ti. I argue that these morphemes are all functional heads, each of which takes a complement. I further show that Yoruba supports an additional step that is not found in Kayne's analysis of English: movement of the possessum N P from Spec C P to Spec DP in a reduced relative clause. The latter is discussed in chapter 3.  2.3.2  Possessum raising in Yoruba genitive constructions  The base structure adopted for the account of genitive constructions would imply that the possessor NP precedes the possessum NP. However in the surface syntax (43), the possessum NP actually precedes the possessor NP. (43)  a.  dje e Tiinde trick M T S T. 'Tunde's tricks'  b.  pata a Bunmi pants MTS B. 'Bunmi's pants.'  c.  oko o Tona vehicle MTS T. 'Tona's vehicle'  d.  apa a Kunle arm MTS K. 'Kiinle's arm'  I take this to indicate that the possessum N P raises to the specifier of another phrase higher. Thus we require the presence of a DP layer above the base vP structure as in (44). I further  propose that the Genitive marker in the form of the MTS occupies D. The structure illustrates that the possessor NP and the possessum NP are located below D. (44)  Nominal genitive structure DP  D  uP  MTS Poss'r / \ Kunle  v  Poss'm Z\ ile  The idea of having an additional layer above the base structure is consistent with what is observed in other languages. For example, in Hungarian, (Szabolsci 1983, 1994, Kayne 1994: 85-86, Delsing 1998) an additional functional head between D and N is suggested. Thus the Possessor morpheme is located below D in the extended projection. In that proposal there is a structural subject position below D in Spec PossP, as in (45a). This makes D parallel to C rather than I. The structure in (45a) accounts for the example in (45b): the determiner az sits in D, the possessum pronoun en sits in Poss and the possessum is contained within the NP.  (45)  a.  DP D' D  PossP  Possessive pronoun  head ~noun vendeg-e-m  genitival DP (Delsing: 1998:93)  b.  az en vendeg-e-m the I guest-POSS-lSG 'my guest' (Delsing 1998: 93) -  The structure in (44) predicts two possible outputs: it allows either the possessor or the possessum to raise to Spec, DP. However in Yoruba, with nominal possessives, only one of the two possibilities is attested. Precisely, only the possessum raises to Spec DP, thus giving the possessum-possessor linear order in the surface syntax. (46b) where the Possessor precedes the Possessum is ungrammatical. (46)  a.  ile e Kunle house MTS K . 'Kunle's house'  b.  *Kunle K.  e ile MTS house  The ill-formed surface string will be compatible with two analyses. The M T S is not in v because then we get the wrong word order. I illustrate this in (47). (47)  uP  Another way of deriving the ill-formed (46b) would be to raise the possessor to Spec DP, as in (48b). But such possessor raising is not possible in nominal context in Yoruba. The question is: why? I return to this question in chapter 3.  (48)  a.  *Kunleeile  b.  ^Nominal Possessor Raising *DP  For a language like English, observe that it is possible for either the possessum (49a) or the possessor (49b) to raise to Spec D P . (49)  14  a.  Possessum raising in English DP  b.  Possessor raising in English DP Poss'Tj Z\ John  D  uP  v  Poss'm  's  house  The implication of this analysis is that the Relational v position is pronounced in English as 's while the D position is realized as of. Thus, these two functional elements can co-occur in English. In Yoruba only the D position is pronounced (as the MTS). Indeed, the possessor need not raise. The proposed structure also predicts that we should be able to have John of's house, since both D and v are pronounced in English. 14  On the question of why ' o f is analyzed as D in English, this follows from the structure. If there are only two functional positions that must be filled and it is the case that ' o f precedes the Saxon 's in the surface syntax, the only position that ' o f can occupy is D. See Kayne (1994) who proposes that the relevant head can be D° or p  015  .  16  The table in (50) summarizes the discussion on the functional elements within the genitive DP in English and Yoruba. v and D are both pronounced in English, whereas in Yoruba only D is pronounced.  (50)  2.3.3  2 Internal structure of Genitive construction v Yoruba 0 English 's .  D MTS of  What moves: N or NP?  Although the structure proposed predicts phrasal movement, so far, we have only seen the movement of simple phrases that contain only the possessum noun. There is nothing that shows that what actually moves is the possessum noun rather than the whole phrase. In the next three subsections, I present three pieces of evidence that show that what moves is NP (rather than N).  This proposal contrasts with Alexiadou and Wilder's (1998). In their proposal, the Possessor Phrase is base generated. Assuming their proposal the possessor will remain in Spec,uP. While that will account for English examples such as 'John's house', such approach yields a wrong output at least for the Yoruba data where the Possessum NP always precedes the Possessor NP. Another question that arises is this: why can only the possessum move out of vP in Yoruba and why is it possible to move both the possessor and the possessum out of vP in a language like English? A comparison of a house of John's where the possessum moves with John's house shows that the former has a kind of "mild" topic reading (Muhlbauer, personal communication). The Yoruba counterpart ile e ti Tunde has the same reading. This claim is left to further research. 1  16  2.3.3.1 Evidence from modifier placement  The first piece of evidence comes from NPs that contain a noun and a modifier. The following examples show that whenever an N P raises, the modifier may raise with it: In (51a), the modifier kekere 'small' modifies the possessum noun apb 'bag' In (51b), the modifier giga 'tall' modifies the possessor noun obinrin 'woman'. Finally, in (51c), while the possessum noun apb 'bag' is modified by kekere 'small', the possessor noun obinrin 'woman' is modified by giga 'tall'.  (51)  a.  [Apo kekere e obinrin] bag small MTS woman 'I admire the woman's small bag.'  ^  wu mi please me  b.  [Ap6 o obinrin giga] w u m i bag M T S womantall please me 'I admire the tall woman's bag.'  c.  [Apo kekere e obinrin giga] w u mi bag small M T S womantall please me 'I admire the tall woman's small bag.'  I use (51c) as an illustration. Whenever the possessum NP raises, the modifier may raise with it. The possessum NP consists of the noun apb 'bag' and the modifier kekere 'small', both of which move to Spec DP. (52)  a.  Possessum raising with modifier DP Poss'mi D N apo  Mod kekere  ^—\ v N Mod |  I  obinrin  I  giga  ^  In this account, modifiers are treated as adjuncts.  What this shows is that whenever an  entity is adjoined to a phrase, and there is movement of that phrase, the adjunct can move with the phrase.  2.3.3.2 Evidence from structural ambiguity So far, we have seen cases of modifiers moving along with the nouns that they modify. There is, however, a situation that is yet to be reported, namely cases of structural ambiguity. When there is only one modifier within a genitive construction in certain contexts, the modifier can modify either the Possessor or the Possessum. For example, in (53) kekere 'small' may modify the Possessor Dupe, (53a) or the Possessum apd 'bag' (53b).  (53)  [Apo bag (a) (b)  d Dupe kekere] wu m i M T S D. small please lsg T admire the junior Dupe's bag.' 'I admire Dupe's small bag.'  The claim I make is that it is possible to leave behind an adjoined modifier when the NP that it is adjoined to moves. Thus, the surface string in (53) arises in one of two ways. First the Possessum" can be posited to occur by itself with the modifier attached to the Possessor, as in (54). (54)  a.  ,ap6]i o [NP bag  b.  DP  P0SS  . Dupe kekere] [tj Poss'r small R  'Poss'm  N  Mod  Dupe kekere 17  See chapter 6.  Second, the modifier can attach to the Possessum, but not raise along with it, as in (55). Even though the modifier is adjacent to the possessor in the surface syntax, it is construed as a Possessum modifier. (55)  a.  [apo] . o [Dupe] [ bag M T S D. 'Dupe's small bag' Poss  b.  m  NP  t  Poss  .  m  kekere]] small  DP  Based on these data, one can conclude that Yoruba permits modifier stranding. And when this happens, the stranded modifier still modifies the moved Possessum argument.  18  2.3.3.3 Evidence from selectional restrictions  I showed in the preceding section that a modifier occurring next to a possessor N P could modify the possessor or the possessum. However, there are certain modifiers occurring in the same position, which could only be construed as modifying only the possessum. For instance the modifier nla 'big' in (56a), and pelebe 'flat' in (56b), can only modify the Possessum  An alternative approach is to adjoin the modifier outside the small clause as in (i). (i) [kekere [vP [ . Dupe v [ . apo]]]] In this account there is going to be a successive raising of dpd around Dupe (attracted to the Spec of o) and raising (with pied piping) of [apo o Dupe] around kekere, (ii)-(iii). (ii) [kekere [DP . apo o [vP [ . Dupe v [tj]]] (iii) [XP[DP , apo o[vP[ . Dupe v ft kekere ft]]]] For details on pied piping of adjectives, see Cinque (2005a&b). 18  Poss  Poss  Poss  m  r  Poss  m  m  Poss  Poss  r  r  rather than the Possessor.  (56)  a.  [Apotii Dupe nla] w a nibi box M T S D. big be here (i) 'Dupe's big box(es) is here.' (ii) '*Big Dupe's box is here'  b.  Apo box (i) (ii)  o Dupe pelebesonu M T S D. flat loss 'Dupe's flat bag is missing.' '*Flat Dupe's bag is missing.'  The fact that the modifiers nla 'big' and pelebe  'flat' cannot modify the possessor is  accounted for via selectional restrictions. These modifiers can only modify [-human] rather than [+human] nouns. Thus, the examples in (57) are informed 19  (57)  a.  *Dupe D. '  nla big  •  b.  *Dupe D.  P?l?be big'  This contrasts with modifiers such as kekere, which means both 'small' and 'junior', or giga 'tall', which can modify both human and non-human nouns. (58)  a.  [apotii Dupe kekere] box M T S D. small (i) 'Dupe's small box(es)' (ii) 'Dupe (junior)'s box'  There is an exception to this rule. Consider (i) where there are two possible interpretations. In (b), the use of omo nla is idiomatic meaning 'a terrible child' rather than 'a big boy', (i) [apoti i omo nla] box MTS child large (a) 'the child's big box' (b) 'the terrible child's box' The apparent exception to the rule could be due to the fact that in (ib) nld is actually a pejorative use of the adjective (similar to the pejorative suffixes of other languages), in which case one could expect the cooccurrence of nld with 'small' (box of the child terrible small, or box of the small terrible child, meaning the terrible small child's box). 19  b.  [apoti box (i) (ii)  i Dupe giga] M T S D. ' tall 'Dupe's tall box(es)' 'tall Dupe's box' (the box of Dupe who is tall)  In this latter case, the modifier can modify either the possessum or the possessor.  2.3.4  Genitive pronouns  Looking back at all of the data so far presented, discussed, and analyzed, the possessor has always been a personal name. Yet given the right context, pronouns can replace nouns as possessors. In (59b), the clitic pronoun possessor takes the position of the personal name Kunle. (59)  a.  ile e Kunle house M T S K . 'Kunle's house'  b.  ile e re house M T S 3sg-Gen 'his/her house'  There are two facts that I bring to the attention of readers on pronouns: they divide into three classes weak (clitic), strong and reflexive pronouns, (cf. Cardinaletti 1998 on Italian and 20  Spanish, and Schoorlemmer 1998 on Dutch), and they can feature in genitive constructions. As background information, I provide a data set to show the pronoun paradigm when they enter into genitive constructions.  21  The terms weak and strong as used in the Yoruba literature is based on certain common properties that members of each group share which others do not share. Two of such properties is that the strong- pronoun behaves like proper noun in the sense that it can be modified and it can be focused whereas the weak pronoun can neither be modified nor focused. For details, see Bamgbose (1967, 1990) Awobuluyl (1978) Pulleyblank (1986), Dechaine (1993) among others. The tonal alternation on the mid tone genitive marker in (60) is not discussed in this dissertation due to space. Readers are referred to Akinlabi and Liberman (2000). 2 0  21  (60)  Weak Possessorpronouns a. Tiff fe [iwe e mi] T. want book MTS lsg 'Titi wants my book.' b  Titi fe [iwe e re] T. want book MTS 2sg 'Titi wants your book.'  c.  Titi fe [iwe e re] T. want book M T S 3sg 'Titi wants her book.'  d.  Tiff fe [iwe e wa] T. want book M T S l p l 'Titi wants our book.'  e.  Titi fe [iwe e yin] T. want book MTS 2pl 'Titi wants your book.'  f.  Titi fe [iwe e won] T. want book MTS 3pi 'Titi wants their book.'  Observe that in Yoruba, as in many other languages, whenever a pronoun functions as a possessor, it takes a different form from when it functions as subject or object NP. To illustrate this consider the 3sg clitic pronoun. In the subject position (61a), it is 161, in the object position (61b), it is / i / .  2 2  But when it functions as a possessor (61c) it takes the form  /re/.  (61)  a.  [O] ri 3sg.NOM see 'S/hesawme.'  mi lsg.Acc  Note that in Yoruba, the object pronouns take different forms depending on their syntactic position. For example, the 3sg clitic is invariantly "6" when in subject position whereas its shape in the object position is determined by the shape of the final vowel of the verb, (i) a. Ade gb[e e] Ade carried it b. Ade t[a a] Ade sold it 22  b.  Mo ri lsg.NOM see 'I saw him/her.'  [i] 3sg.ACC  c.  Mo ri [iwe lsg.NOM see book 'I saw his/her book.'  e [re] MTS 3sg.Gen  Comparing the personal name-as-possessor with pronoun-as-possessor raises one crucial question. Is the form a pronoun takes due to Case assignment? M y answer is in the affirmative. Observe that like most languages , Yoruba pronouns have different forms 23  depending on their syntactic position and function. I therefore propose that when a pronoun functions as a possessor, it must take the genitive form. This is why only (62a), and neither (62b) nor (62c), is well formed.  (62)  a.  iwe e re book M T S 3sg.Gen 'his/her book'  b.  *iwe book  e 6 M T S 3sg.NOM  c.  *iwe book  e e M T S 3sg.ACC  Just as possessors appear in the genitive Case, so too do subjects and objects take specific Case forms. Take for instance, the difference in form between the 3sg in subject and object position in (61). In the subject position where it receives N O M Case, its form is 6 (61a), whereas when it is in the object position where it receives A C C Case, the form it takes is i, (61b). The base structure in (63) illustrates the account of possessive pronoun, which is not in any way different from the one I already showed for possessors that are personal names.  ^ake for example the English 3sg masculine pronoun. When in the subject position, it takes the form he which is synonymous with NOM Case. When in object position, the same pronoun is realized as him. This is also synonymous with ACC Case. Finally, when the same pronoun functions as a genitive pronoun it is realized as his. This again is synonymous to Gen.  (63)  DP  D  vP Poss'r [Gen]  v  Poss'm  re  0  iwe  Again, the possessum NP moves to Spec DP as shown in (64). 24  (64)  DP Poss'm D  vP  iwe MTS Poss'r [Gen] re  v  ti  1  (Z)  I now turn to the phonological account of the genitive marker that shows up in Yoruba genitive construction, namely the mid tone syllable.  2.4  The Yoruba nominal genitive: syntax-phonology interaction  The discussion has so far concentrated on the semantic and syntactic basis of the Yoruba genitive construction. In this section, I turn to the question of what conditions the phonological shape of the genitive construction, focusing on the status of the M T S that appears between the possessor and the possessum which for now is represented as a mora \i.  For details on Yoruba pronouns, see Adesola (2001); Ajiboye (2005a); Akinlabi and Liberman (2001); Avvobuluyi (1978, 1995); Bamgbose (1966, 1967, 1990); Dechaine (1993); Dechaine and Wiltschko (2002a,b); Manfredi (1987, 1995) and Pulleyblank (1983). 2 4  (65)  a.  Poss'r VI]  Poss'm [Y2  ii  ile e 6j6 house M T S O. 'Ojo's house'  b.  After considering how the syntactic structure of genitive constructions maps onto phonological forms cross-linguistically (§2.4.1), I review the phonological distribution of the MTS (§2.4.2). Following my analysis of the MTS as a functional element, and the claim that it occupies a D position, I argue that its presence is regulated by syntax but its overt realization is phonologically determined. This means though it is found in the environment of a nominal phrase its pronunciation is determined by certain phonological rules. It also implies that while syntax determines where we get the M T S , phonology determines when and how it is to be pronounced. 1 consider two different accounts of the MTS. One analysis 25  claims that the morphosyntax inserts a genitive formative in the appropriate place. This is the analysis that I pursue here (§2.4.3). Another analysis treats the M T S as part of the word/lexical item (§2.4.4). In that account, the MTS is analyzed as a prosthetic vowel that is used to restructure non-canonical nouns in Yoruba (Awobuliiyi 2004). I argue that the Standard Yoruba (SY) and M o b a data are consistent with the first analysis though there is some variation. The second analysis is rejected as it accounts for neither the SY data nor the M o b a data . This section closes with a discussion of the relationship between MTS and L 26  deletion (§2.4.5).  2.4.1  Spelling out the genitive: cross-linguistic evidence  Most languages have morphemes that are used to indicate the genitive relation between two  This is in line with the view that phonological content implies the presence of a Functional head, but the absence of such content does not imply the absence of a Functional head (Dechaine 1993: 84). Another alternative analysis, which I do not consider in this dissertation, treats it as a syntactically conditioned copy that marks the R relation between the possessor and the entity that is possessed. This view has been implicitly expressed in Bamgbose (1990), Akinlabi and Liberman (2000) and explicitly in the work of Dechaine (2001) and Ajiboye (2004b). For details on copying analysis, see also Postma (1995), Mortenson 2003 Dechaine & Ajiboye (2004). 2 6  arguments, although the phonological representation of these morphemes may vary. Observe that in the genitive base structure (66), there is only one position available for this morpheme in the base structure namely, the v. (66)  Poss'm  However it may be pronounced as D in the surface syntax. In particular, English pronounces both D (of) and v ('s), (67). The Saxon morpheme has three allomorphic variants (Chomsky and Halle 1968, Russell 1997, Sotiloye 1999): [z], [s] and [az]. After a voiceless consonant, it is realized as [s]. After a voiced consonant, it is [z]. After a fricative or an affricate, it is [3Z].  (67)  a.  book of John  b.  (i) (ii)  /d3an + z b u k / /pop + s buk/  'John's book' 'Pope's book'  (iii)  /d3Ad3 + a z b u k /  'Judge's book'  c.  book of John's  The relevant structures are given in (68). (68)  (a)  NofN N'sN  N of N's  book  V  c.  book D of  vP John v  In Hebrew there is a distinction between the 'free state' and the 'construct state' genitive.  (69)  a.  ha-bayit sel the-house of 'the man's house'  b.  bayt ha-'is house the-man 'the man's house'  ha-'is the-man  free state  construct state (Siloni 1997: 21)  Phonologically, Hebrew marks its construct genitive on the head noun. When the head noun is stressed, it bears a full vowel. But when there is a shift of stress from the head noun to the genitive noun phrase, other phonological processes such as vowel reduction usually follow (Siloni 1997). Thus comparing (69a) and (69b), one can see that the head noun bayit 'house' in (68a) takes the full vowel because it is stressed in free state genitives. On the other hand, when the stress is lost on the possessum (69b), the vowel of the head noun reduces to a schwa. Siloni (1997) concludes that the shift in stress on the possessum is an indication that an abstract genitive Case is assigned by the head noun in construct state NPs. In my own account it appears that only the D position is pronounced in Hebrew. The relevant structures are given in (70).  ha-bayitj D  vP  v 0 In (70a), the possessum (ha-bayit 'the house') raises to Spec DP, and D is occupied by sel ' o f . In (70b), the possessum (bgyt 'house') again raises to Spec DP; this time, however, D has no phonological content. This empty D correlates with de-stressing of the possessum noun in the so-called construct state. Turning now to Yoruba, we observe that there are two grammatical elements that mark the genitive. There is MTS (71a), and ti ' o f (71b). Each of these occurs between the possessum and the possessor.  (71)  a.  epo o Dotun oil MTS T.' 'Dotun's oil'  b.  epo ti Dotun oil of T." 'oil of Dotun'  Though I treat both examples as genitive DPs, I propose different structures for each of them. I analyze the MTS as D, which takes a small clause as its complement, as in (72a). I analyze the ti element as C, which takes a small clause as its complement, as in (72b).  (72)  a.  DP  We have already seen the motivation for (72a). Readers are referred to chapter 3 for the motivation and discussion of the structure in (72b).  2.4.2  The phonological distribution of the Mid Tone Syllable  I propose that though the M T S has a syntactic function, its spell-out is phonologically conditioned. This is consistent with the claim that the MTS is a functional head that is overtly represented unless a phonological rule prevents it. The phonological rule that determines when the MTS is to be pronounced is the focus of this section. The examples in (73) show that the MTS is obligatory before consonant-initial nouns. Moreover, observe that the vocalic content of the MTS is always identical to the vowel that precedes it. In particular, the examples given in (73) show that V i can instantiate all of the possible oral vowels in Yoruba: {i, u, e, o, e, o, a}.  2 7  27  This is also true of the nasal vowels: {in, un, on, an} as the examples in (i) illustrate.  (73)  obligatory presence of MTS a. on i Kunle H] M [H 'Kunle's head'  *ori Kunle  b.  ojii u Kunle H] M [H ' Kunle's eye(s)'  *oju Kiinle  c.  ete e Kunle L] M [H 'Kiinle's lips'  *ete Kiinle  d.  owo o Kunle H] M [H 'Kunle's money'  *owo Kiinle  e.  ise e Kunle H] M [H 'Kiinle's work'  *ise Kiinle  f.  oko o Kunle LJ M [H 'Kiinle's vehicle'  *oko Kiinle  g.  aja a Kiinle H] M [H 'Kiinle's dog(s)  *aja Kiinle  Before vowel-initial nouns, MTS is optional, as in (74). Again, observe that when the MTS is present its vocalic content is identical to the vowel that precedes it.  28  orfn in Tiinde chewing stick MTS T. 'Tunde's tooth brush' ogiin un Tiinde MTS T. inheritance 'Tunde's inheritance' opon on Tunde tray MTS T. 'Tunde's tray' adan an Tunde T. bat MTS 'Tunde's bat' Some Yoruba linguists (e.g. Awobuliiyi 2004) report that the MTS is never present in vowel-initial nouns. There are two reasons for this opposing view. One is dialectal variation. As pointed out in Awobuliiyi (2004), there are some dialects of Yoruba where this segment is not pronounced before vowel-initial words. However, the fact that this segment is pronounced in Standard Yoruba calls for further research into other dialects. If it turns out that there are more dialects that support its absence, then we can conclude that it is an innovation in Standard Yoruba. The other reason is some phonological rule, which might rule out the sequence of three vowels in an NP. Awobuliiyi however admits that this mora can be realized when one hesitates. This is why he concludes that the presence of this mora neither has any syntactic nor semantic function. The data reported here reflect what has been established in the literature (Bamgbose 1990; Akinlabi and Liberman 2000) as well as the writer's judgment, namely that at normal speech rate, the MTS is usually absent before V-initial nouns, but at slower rates, the MTS is present.  (74)  optional presence of MTS a. on (i) Ade 'Ade's head'  H] M [ M  b.  oju (u) Ebun 'Ebun's eye(s)'  H] M [L  c.  ete (e) 'Olu'slips'  L] M [ M .  d.  owo (o) Ojo 'Ojo's money'  e.  ise (e) Eleyinmi 'Eleyinmi's work'  H] M [ M  f.  oko (o) oga 'master's vehicle'  L] M [L  g.  apa (a) isikalu Tsikalu's arm(s)  H] M [L  Olu  (-  H]M[L  ;  In what follows, I account for how the MTS comes to be realized. I consider two analyses: one claims MTS is an underlying mora \i (§2.4.3); the other claims MTS is a phonologically conditioned segment (§2.4.4).  2.4.3  Analysis I: the Mid Tone Syllable is an underlying mora ]i  The genitive marker always takes the exact shape pf the final vowel of the possessum, (75a), or that of the modifier if present, (75b).  (75)  a.  ile e TM house M T S T. 'Titi's house'  b.  ile nla a Titi house big MTS T. 'Titi's big house'  The question that I address is how the genitive marker comes to assume the segmental shape  of the preceding vowel without inheriting its tone. For example, in (75a), [e] is the segmental component of the H tone [e]. This excludes the H-tone. In accounting for this, I compare data from Standard Yoruba with M o b a and claim that the difference between the two varieties is due to the way they represent this genitive marker in the phonology. I propose that it is the morphosyntax that inserts a genitive formative in the possessum-possessor boundary. The underlying phonological form of this formative is stated as (76).  (76)  Phonological representation: The genitive marker is a mid tone mora pi with no vocalic specification.  This proposal states that the genitive marker is underlyingly a mid tone mora that has no segmental features of its own and occurs in the structure in (77)  (ll).  29  D  I  vP  ...  M On this view, in SY, this mora undergoes assimilation of the segmental features of the preceding vowel, (78). This is Contrary to Akinlabi (1985); Akinlabi and Liberman (2000); Pulleyblank (1986, 2004) and Ajiboye et al (2004) which claim that such moral is underlyingly without a tone of its own and therefore must be assigned the default M tone.  30  Dechaine (2001:101) claims that 'Genitive K shows up in the form of a M-tone mora or timing unit' (cf. Bittner and Hale (1995). 2 9  30  There is an alternative analysis that treats MTS as an underlying l\l in Standard Yoruba. On that view, the claim is that assimilation processes that involves this vowel shows that it is the weakest in Yoruba and as such it is susceptible to assimilating features of other adjacent segment. Thus, in (i-a), where I'll is the first of the two contiguous vowels at word boundary, it assimilates the segmental features of the second vowel, lol. In (i-b) where lil is the second vowel, it still assimilates the feature of the preceding vowel, /a/ (cf. Awobuliiyi 1982). (i) • a. ni ooriin > loooriln HAVE incense 'have incense' b. ara imode > araamode person 'an Imode person' Note that such assimilation only targets segmental features. This is consistent with the observation that in genitive constructions, assimilation does not target the tone of the segment that is copied.  (78)  a.  ile pi Tunde > house T. 'Tunde's house'  ile  e MTS  Tunde  b.  owo pi Tunde > money T. 'Tunde's money'  owo  o MTS  Tunde  The mora pi undergoes regressive assimilation:  (79)  pi  e M Observe that there are cases when the MTS may be unpronounced, as in (80).  (80)  a.  fila pi oga > cap 0 master 'master's cap'  fila  oga  b.  ewa pi Ojo money 0 O. 'Ojo's beans  ewa  Ojo  >  For speakers who do not pronounce the MTS in this context, situation is that although the mora is there, the assimilation rule does not apply. In order to capture the case where it is pronounced as in (78) and the case where it is not pronounced, as in (80), I argue that the rules of the phonology of the language cause deletion in appropriate contexts and On independent grounds, we also know that the relation between an unspecified mora in Standard Yoruba and l\l is close (Pulleyblank 1988 and Awobuluyl 1982, 1997). It would therefore not be surprising to find such a correspondence here. If there was an underlying l\l in SY and if assimilation was applicable or obligatory, then the surface effect would be exactly as we observe it and undistinguishable from having an underlying /mora/ that is not specified for any feature. However the current proposal is much simpler since it does not require a stipulation of a segmental representation of this morpheme underlyingly.  assimilation in appropriate contexts. In (81a), before Consonant initial possessors, it is obligatory for the rule of assimilation to apply to the genitive mora. In (81b), before Vowel initial possessors, for speakers who pronounce the M T S , the same rules of phonology applicable to Consonant initial possessors also apply. Whereas for speakers who do not pronounce the MTS before Vowel initial possessors, the rule of regressive assimilation fails to apply.  (81)  a.  fila y. Kunle > cap K. 'Kiinle's cap'  fila  a MTS  Kiinle  b.  fila n cap 'Ojo's cap  fila  (a) MTS  Ojo  Ojo O.  A comparison of Standard Yoruba with the M o b a dialect shows that the phonetic representation of the MTS in the two speech forms is not the same. I claim that this is due to the way the rules of phonology apply in the two varieties. First, consider the following data.  31  In Moba, before consonant-initial nouns the vocalic content of the MTS is the vowel III. Moba  (82)  Standard Yoruba  Gloss  a.  erf  i  Kiinle  on  i  Kiinle  'Kiinle's head'  b.  ojii  i  Kiinle  ojii  u  Kiinle  'Kiinle's eye(s)'  c.  ete  i  Kiinle  ete  e  Kiinle  'Kiinle's lips'  d.  eo  i  Kunle  owo  0  Kiinle  'Kiinle's money'  e.  use  i  Kunle  ise  e  Kiinle  'Kiinle's work'  t  f.  oko i  Kiinle  oko 0  Kiinle  'Kiinle's vehicle'  g-  apa i  Kiinle  apa a  Kiinle  'Kunle's arm(s)  The difference between the two dialects in lexical items such as 'head', 'money' and 'work' is not crucial to the discussion here. 31  I propose that despite the differences in the surface form, the M T S in M o b a is also underlyingly a mid tone mora just as in SY. When the possessor is consonant initial, I claim that this mora undergoes default feature assignment and surfaces as  (cf. Pulleyblank 1983,  1988 and Akinlabi 1985). (83)  The pi undergoes default feature assignment  However, when the possessor is vowel initial, the mora is not assigned any default feature, so it has to delete since it is unpronounced in the surface syntax.  (84)  a.  fila pi Kiinle > cap 0 K. 'Kiinle's cap'  fila  i  b.  fila pi cap 0 'Ojo's cap'  fila  Ojo  Ojo O.  >  Kunle  This is why we do not have examples such as (85) in Moba. (85)  a.  *fila cap  i oga MTS master  b.  *ewa i money MTS  Ojo O.  There is one vital point to note with respect to the M T S in M o b a and Standard Yoruba. The point is that while the MTS is optionally pronounced in Standard Yoruba when the possessor is vowel initial, in Moba, it must not be pronounced.  To recap, I have presented the facts of both the Standard Yoruba and M o b a data. The morphosyntax makes the genitive formative available in the appropriate place (i.e. in the Possessum-Possessor domain). The underlying phonological form of this formative is a mora  \i without vocalic or tonal specifications. I have also shown that it is the rules of the phonology that cause the different surface forms in Standard Yoruba and M o b a . In Standard Yoruba, the rules of assimilation and deletion apply in appropriate cases, i.e. obligatory regressive assimilation whenever the possessor is Consonant-initial and optional deletion whenever the possessor is Vowel-initial. In M o b a the rules of default feature assignment and deletion apply in appropriate contexts, i.e. default feature assignment whenever the possessor is Consonant-initial and obligatory deletion whenever the possessor is Vowel-initial. This account derives all the observed facts in a simple fashion.  2.4.4  32  Analysis II: the Mid Tone Syllable as phonologically conditioned segment  This section examines an alternative proposal by Awobuluyi (2004). There are three points that are relevant here. First, Awobuluyi challenges the claim that the M T S is a genitive marker, §2.4.4.1. Second, he claims that the MTS is a prosthetic vowel, §2.4.4.2. Third, he treats the MTS as a hesitation marker, §2.4A3.1 examine each point briefly.  2.4.4.1 The Mid Tone Syllable is not a genitive morpheme  In reaction to Ajiboye (2004b), which is the same view that I pursue further here, Awobuluyi (2004) claims that there is no genitive morpheme that exists between the Possessor and the Possessum in Yoruba. According to him, the MTS that shows up in genitive constructions has neither syntactic nor semantic content. On this view, he contends that the: While evidence that comes from different dialects is desirable and should be pursued in helping us to know more about what operates in SY (Bamgbose 1986; Awobuluyi 1988), such evidence (especially from a few dialects) is not absolute. There is a need for study of other dialects to see whether the story is the same. Observe that there are cases where the standard form of a language does something that is completely different from what its dialects do. One case in point is the occurrence of /u/ and nasal vowels word initial of certain nouns in some dialects; in particular Moba but not in Standard Yoruba (Ajiboye 1991). 3 2  possession is expressed in Standard Yoruba with noun phrases formed by simply juxtaposing either two noun phrases or a noun phrase and an appropriate possessive pronoun in that order. (Awobuluyi 2004:1) Consider the examples in (86) taken from the cited work.  (86)  a.  Owo Ojo money (of) Ojo 'Ojo's money'  b.  Owo re money his/her/its 'his/her/ money; money for it' (Awobuluyi 2004:2)  Under Awobluliiyi's account, the examples in (86) will have the structure in (87) as the genitive phrase.  33  (87) Poss'm owo  .  Poss'r Ojo  According to that analysis there is nothing like a genitive marker in Yoruba. The first thing one observes about the examples in (86) and the representation in (87) is that they do not show any relation between the possessor and the possessum. I have demonstrated that there exists a relation between the possessor and the possessum in Yoruba as in many other languages. I have also argued that the appearance of the MTS between these two arguments is symptomatic of this relation. The claim is that when it is covert, this is due to phonological rules that prevent the MTS from being pronounced. Thus, the absence of  The example in (86b) is not attested in Standard Yoruba as the correct form obligatorily takes the MTS. owo o re money MTS 3sg 'his/her money' The form reported in (86b) is only allowed in the orthography by convention. Nevertheless, in some dialects of Yoruba especially in Moba, there is also a parallel form to (i), which shows the absence of the MTS, (ii). (ii) eo 0 rin (Moba dialect) money 3sg (i)  phonological content is still interpreted as being associated with relational meaning. In other words, the phonologically empty category is licensed syntactically. More revealing are the numerous examples that have been previously cited which show contexts where the presence of the MTS is obligatory. As a reminder, I give one such example below.  (88)  a.  ile e Kunle house MTS K . 'Kunle's house'  b  *ile house  Kiinle K.  The ill-formedness of (88b) is due to the absence of the M T S . Awobuluyi's proposal represents the possessor and the possessum as sisters. On the justification for why the genitive construction in Yoruba does not have a genitive marker, he has this to say:  whether a given NP1 NP2 construction signifies possession or mere apposition in the language is actually a function either of the context or of the meanings or semantic classes of the individual lexical items involved (Awobuluyi 2004: 6). Recall from above that the R relation may be contextually or lexically determined. In this respect the present analysis accords with Awobuluyi's general claim. However, while Awobuluyi treats the Possessum-Possessor relation as a special case of apposition, the present analysis distinguishes the two as discussed above.  2.4.4.2 The Mid Tone Syllable as a prosthetic vowel  Awobuluyi argues that appearance of the M T S in genitive constructions is a purely phonological phenomenon. Taking evidence from many dialects of Yoruba such as Moba, E k i t i , Ikare and O w o , he asserts that the M T S is a prosthetic vowel that is used to  restructure non-canonical nouns in Yoruba. Canonical nouns are defined as follows: (i) they 34  are non-derived; (ii) they are not borrowed; and (iii) they have V - C V structure. Some representative examples are given in (89).  (89)  a.  ike 'hunch back'  b.  omi 'water'  c.  epo 'oil'  d.  efp 'vegetable' pja 'market'  Now consider the examples in (90), which are either borrowed as in the case of (a-c), or derived through a deletion process as in (d-e). f  Moba  35  Standard Yoruba  Gloss  a.  i-gomina  gomina  'governor'  b.  i-telp  telp  'tailor'  c.  i-bhiki  biriki  'brick'  d.  i-dele  Dele  Yoruba name  e.  i-kiinle  Kunle  Yoruba name (Awobuluyi 2004: 9-10)  In earlier work (Ajiboye 1991), it has been established that most nouns that begin with a consonant in Standard Yoruba are pronounced with vowel I'll preceding such consonants in Moba. In fact, in the cited work, I mentioned that cases where there is deletion of a syllable or two in personal names before the output produce Cinitial nouns, such nouns must obey the epenthetic rule that enforces III to be inserted. Thus, idele could be a reduced form of Awodele, Ogiindele, Oludele etc. Readers are referred to that work for more data and full explanation. In the original work, Awobuluyi simply puts 'Dialects' in the place where I put Moba. I make this change to be more precise. Moba is one of the dialects that has the so-called prosthetic vowel. 3 5  For examples such as those in (90), Awobuluyi's proposal will yield the structure in (91), where the so-called prosthetic vowel is part of the NP. (91)  DP  D  NP i-gomina  According to Awobuluyi, whenever consonant initial nouns enter into a genitive construction, it is the same prosthetic vowel that shows up. Some illustrative examples are given in (92).  (92)  Moba. a. ule i- gomina house governor 'Governor's house' b.  aso .i-telo cloth tailor 'the tailor's dress'  c.  eo i-bfrikila money bricklayer 'bricklayer's money'  d.  ata i-dele pepper D. 'Dele's pepper'  e.  owu i-kunle cotton K . 'Kunle's cotton'  In Ajiboye (2004a,b) and also in this present work, I treat the so-called prosthetic vowel as a functional element that occupies the D position, as in (93).  (93)  DP  D"  NP  i  gomina  One other argument that Awobuluyi puts forward in support of the claim that the MTS is a prosthetic element is with respect to the tonal melody it bears. On why the M T S bears M tone, he asserts that:  M T S cannot but bear mid tone, as it is a mere surface manifestation of a prosthetic vowel i with mid tone that is required to restructure otherwise phonologically deviant or noncanonic nouns when used at least NP-internally in all varieties of Yoruba (p. 17). As already discussed, Awobuluyi's account of MTS is at variance with mine. The first problem is the constraint that restricts the prosthetic segment to NP internal positions. The immediate question that follows is why the NP internal condition? If indeed this is a prosthetic vowel then we should be able to have it in word initial position in Standard Yoruba. This is not the case as shown by the unattested form in (93b).  (94)  a.  M o r i on i-gomina lsg see head MTS-governor T saw the governor's head.  b.  *Ig6mina governor  ri see  Standard Yoruba  mi. lsg  Even in M o b a , which is famous for the use of vowel l\l in consonant initial nouns, the occurrence of III in sentence initial position is optional.  (95)  a.  I-kunle  ri mi see lsg Tkunle saw me.'  b.  Kunle ri mi see lsg 'kunle saw me.'  Moba  (96)  a.  I-gbeinga ya a Question-tag come 'Did Igbenga come?'  b.  Gbeinga  ya  Moba  a  Question-tag come 'Did Gbenga come?'  Moreover, the optionality of Moba / i / is completely general. While the examples in (95) and (96) show the optionality of III for nouns in subject position, (97) show that III continues to be optional when the noun is in object position.  (97)  a.  Mi in pe lsg HTS call 'I called Ikunle.'  Ikiinle  b.  Mi in pe lsg HTS call 'I called Kunle.'  Kunle  Moba  However, in genitive constructions mid tone / is obligatory before consonant initial nouns in Moba.  (98)  a.  Iwe i-kunle re e book MTS-T. C O P Dem 'This is kunle's book'  b.  *Iwe  Kiinle  re  Moba  e  In conclusion, in M o b a M-tone / is optional with ordinary nouns but it is obligatory in genitive constructions (except if the possessor is V.-initial). This syntactic contrast is not captured by Awobuluyi's analysis.  2.4.4.3 The Mid Tone Syllable as a mark of hesitation  The last claim contained in Awobuluyi (2004) is that whenever the MTS is found in genitive  constructions involving vowel initial possessors, it signals hesitation. As earlier mentioned about the speech of this researcher and as also found in the literature (Bamgbose 1990, Akinlabi and Liberman 2000), the presence of the MTS in genitive constructions is not a way of expressing hesitation. Akinlabi and Liberman's observation, which supports my analysis, is contained in the following excerpt:  A M i d tone vowel occurs pervasively (though usually optionally) in the middle of such genitive constructions. It assimilates in quality to the vowel that precedes. This vowel is obligatory only when the possessor (the noun in second position) is consonant-initial. When the second noun is vowel-initial (the normal situation), then the vowel is optional. In the case of a vowel-initial possessor, it is natural to think of this extra vowel as an optional possessive morpheme. In the case of a consonant-initial possessor, we are tempted to think of it as an empty prefix. (Akinlabi and Liberman 2000: 17) As mentioned before however, the presence of the M T S is optional before vowel-initial nouns and this optionality is sensitive to speech rate. Thus, in (99) at fast speech rate the MTS is not pronounced, at slow speech, it is.  (99)  a.  ile (e) ojo house M T S O. 'Ojo's house'  b.  omo (o) akin child M T S A . 'Akin's child'  c.  oko (o) ojo vehicle MTS O. 'Ojo's vehicle'  "  (Akinlabi and Liberman 2000: 18)  In conclusion, I have discussed two different analyses of the M T S : (i) as a syntactically conditioned mora pi; (ii) as a phonologically conditioned segment. I have established that the data from both Standard Yoruba and M o b a is consistent with analysis (i). In view of the above, I reject analysis (ii) as it accounts for neither the Standard Yoruba data nor the M o b a data.  2.4.5 Relating the Mid Tone Syllable to L-deletion There is yet another indirect piece of evidence that supports the presence of a genitive marker in genitive constructions. This evidence comes from the loss of a low tone on the final segment of the possessum phrase. Although this is a phonological process, it nevertheless has syntactic implications. Bamgbose (1967, 1990) and Dechaine (2001) have discussed the deletion of the final low tone of a possessum noun. My goal is to find answers to the following questions: (i) Is it sufficient that the possessum end with a segment bearing an L tone or do we need to look beyond the last segment of the possessum for L deletion to apply? (ii) Just as the final L tone of the possessum NP can be deleted, is it also possible to delete the final H tone of the possessum NP? (iii) Is tone deletion of the possessum N P sensitive to syntactic structure; in particular, does a [Possessum-Possessor] sequence behave the same as a [Posessum-Modifier] sequence? Addressing the above questions not only corroborates previous analyses, it also accomplishes three additional things. First, in the course of attempting to answer these questions, I consider a broader range of data than has been previously examined. Second, I demonstrate that in order for L-deletion to apply, all the segments within the possessum must bear L-tone. Third and most important, the deletion of L tone and its replacement with M tone is shown to correlate with genitive marking. Consider the following data: 36  (100) a.  b.  (101) a.  asa a Dudu custom MTS African 'African customs' asa custom  Dudu African  ewu u Dele dress MTS D. 'Dele's dress'  Again, this claim is subject to an instrumental phonetic study examining the timing and duration of the midtone bearing genitive mora and the option of contracting it with the final syllable of the head noun. The relevance of such an experiment is to show that the contracted form is significantly shorter than the full form. It will also enable us to map the relative timing of the tonal and segmental events (i.e. whether the L to M transition happens differently if the M is a result of elision), or if it is a 'lexical' M (Ajiboye et al 2004).  3 6  (102) a.  b.  (103) a.  b.  (104) a.  ewu dress  Dele D.  iru u tail MTS 'lion's tail'  kinihun lion  iru tail  kinihun lion  bia igo o bottle M T S beer 'bottle of beer' igo bottle  bia beer  apo o ti bag MTS C 'our own bag' wa apo ti our bag C 'our own bag'  iw a (our  (cf. ( Bamgbose 1990)  The examples in (100)-(104) split into two: the presence of the MTS in all the (a) examples and its absence in all the (b) examples. There is one other vital point to make with respect to the (b) examples: the replacement of the L tone on the final syllable of the first noun with M tone. The surface forms in the (b) examples in (100)-(104) are accounted for as follows. As a starting point, since the environment for L deletion is where all the vowels of the possessum bear L tone, I assume that the vowels in this environment are associated with one L underlyingly (cf. Leben 1973 among others).  When the L tone of the final vowel in the possessum N P deletes, we are left with two morae that have no segmental features. The mora of the final vowel of the possessum also deletes and the vowel is associated with the mora of the genitive marker. Finally this vowel is assigned the default M tone.  (106) a  p  pi  o  0  L  +  pi  M  My goal here is to account for why there is L-tone deletion of the final vowel of the possessum. I also account for why the L-tone that is deleted is being replaced with M-tone. Lastly, I show the reason for the absence of M-tone genitive marker when the final L-tone of the possessum N P deletes. Dechaine's (2001) observation is that the genitive L-deletion is optional before a Consonant initial possessor. If the L tone does not delete then the genitive marker must be pronounced, otherwise it yields a wrong output.  (107) a.  ilu bata drum 'the bata type of drum'  b.  ilu u bata drum MTS b. 'the bata type of drum'  c.  *ilii  bata  Before a V-initial possessor, Dechaine contends that genitive L-deletion is obligatory, as long as the initial vowel of the second noun (the Possessor) does not bear L, (108). (108) c.  pataki importance  emu palm wine  b.  *pataki emu importance palmwine 'the importance of palm wine'  c.  *pataki importance  i emu MTS palmwine  (adapted from Dechaine 2001: 102)  Dechaine further observes that if an initial vowel of the possessor bears L tone, then Genitive L-deletion doesn't apply (109a), nor does the M T S surface, (109b). In this case, the possessum surfaces with its lexical L tone throughout, (109c). (109) a.  *ilu  oga  , b.  *ilu  u  c.  oga  ilu oga 'master's drum  Observe that all the nouns discussed so far that undergo L-deletion have L tone throughout. Now consider nouns, which have only a final L-tone.  (110) a.  pako HL 'chewing stick'  b.  oko 'motor'  ML  c.  orulu 'drummer'  MHL  d.  iwakiwa LLML 'bad character'  e.  alapata 'butcher'  MHML  When such nouns are put into genitive constructions where they function as possessums, L deletion is not possible, (111). In (112), when L-deletion is enforced on the possessum, the output is ungrammatical.  (Ill)  a.  pak6 o Kunle ' Kunle's chewing stick'  b.  oko o Kunle 'Kunle's motor'  c.  onilu u Kunle 'Kunle's drummer'  d.  iwakiwa a Kunle 'Kunle's bad character'  e.  alapata a 'Kunle's butcher'  Kunle  a.  *pako Kunle  *HL  >  HM  b.  *oko Kunle  *ML  >  MM  c.  *onilu Kunle  *MHL >  MHM  d.  *i w a k i w a Kunle  *LLML  >LLMM  e.  *alapata Kunle  *MHML  >MHMM  The fact the final L in all the possessums cannot delete indicates the L-tone of the possessum does not freely delete. There is one more thing to test even with a possessum that allows L deletion: this is whether the size of the possessum N P is in any way a crucial factor in determining L deletion. This I test with examples in (113-115) where the possessum consists of two, three and five syllables respectively. As the output reflects, each of them triggers L deletion. This suggests that the size of the possessum is not a conditioning factor for L deletion.  (113) a.  b.  asa a Kunle custom MTS K . 'Kunle's habit' asa Kunle custom K.  (114) a.  b.  (115) a.  b.  ogedee Kunle banana MTS K . 'Kiinle's banana' ogede Kunle banana K. agbaagba a Beyiose elder^ MTS 'Beyiose's elders' agbaagba elder  Beyiose B.  I now turn to the syntactic contexts of L deletion. Following Dechaine (1993, 1997, 2001), I propose that just as L deletion of a monosyllabic verb is syntactically determined, L deletion in genitive constructions is also syntactically conditioned. Nominal L deletion is restricted to genitive constructions; thus, if the second entity is not a possessor there cannot be L deletion. One context that we test is when possessum nouns take a modifier. As the following examples show, the final L tone of the nouns is obligatorily retained.  (116) NP + Modifier a. asa buruku custom bad 'bad customs'  *asa buruku  b.  e w u funfun garment white 'white garment'  *ewu funfun  c.  iru gigun tail long long tail'  *"iru gigun  d.  igo kekere bottle small 'small bottle'  *igo kekere  e.  apo pelebe bag flat 'flat bag'  *apo pelebe  This contrasts with (117) where the modifier is replaced with a possessor. Note that both the modifiers in (116) and the possessors in (117) begin with a consonant. Thus while the L tone of the base is retained in (116), it deletes in (117) and consequently a default M tone is assigned.  (117) Possessum NP + Possessor NP Input a. asa ]i Kunle 'Kunle's habits'  Output asa Kunle  b.  ewu JA Kunle 'Kunle's garment'  ewu Kunle  c.  iru n Kunle 'Kunle's tail'  iru Kunle  d.  igo y. Kunle 'Kunle's bottle'  igo Kunle  e.  ap6 \i Kunle 'Kunle's bag'  apo Kunle  If any of these modifiers changes its category from modifier (118a) to noun (118b), then, L deletion becomes possible.  (118) a  b.  igo dudu bottle black 'black bottle'  = dudu is a modifier  igo Diidii bottle D. 'the bottle of Mr. Dudu  = dudu is a noun  4  Finally, L deletion within the Possessum N P is not restricted to the Possessum. If the Possessum takes a modifier that bears L tone throughout as in (119), we observe that the L deletion rule still applies to the final vowel of the modifier, (120).  42  See Awobuluyi (2004) for relevant discussion.  (119) a.  [Apo gudugbe e Kunle] bag bulge ' MTS K . 'Kunle's bulging bag fell off"  jabo fall-off  [igba aimo on wa] ni Olorun time ignorant MTS l p l FOC God 'Our time of ignorance has been overlooked by God' c.  (120) a.  A k o palaba a Gbada sheath flat MTS K . 'Gbada's flat sheath is missing.' [Apo gudugbe Kunle] bag bulge K. 'Kunle's bulging bag fell off"  fojufoda overlook  sonu miss  jabo fall-off  b.  Olorun [Igba aimo wa] ni fojufoda overlook time ignorant lpl FOC God 'Our time of ignorance has been overlooked by God'  c.  A k o palaba Gbada sheath flat K. 'Gbada's flat sheath is missing.'  sonu miss  I have established that the L deletion that takes place in genitive constructions corresponds to genitive marking in Yoruba. In such cases, the segment whose inherent L tone deletes is assigned default M-tone. This is another piece of evidence in support of Dechaine's (2001) claim that there is a high degree of interdependency between phonology and syntax in Yoruba such that the knowledge of one greatly helps the understanding of what goes on in the other. L-deletion is a morphosyntactic phenomenon that is phonologically conditioned. This is because deletion itself is, by its very nature, phonological. But the process is conditioned by syntax; it takes place in a genitive construction. The problem that I address next shows that L-deletion and MTS are in fact two sides of the same coin.  2.4.5.1 The necessary and sufficient conditions for L-deletion  Recall the three questions that I set out to address regarding L-deletion. First is to find out if the possessum ending with a segment bearing L tone suffices for L deletion or do we look  beyond the last segment of the possessum. Second is to test if it is possible to delete the final H tone of the possessum NP. Third is to establish how crucial is the element that follows the possessum (possessor versus modifier) as a determining factor for L-deletion to take place. My finding shows that before L deletion can take place, the possessum NP or the modifier to the possessum NP must bear L tone throughout. I also established that when a possessum or its modifier ends in H tone, it couldn't delete for purposes of marking the genitive construction. And finally we saw that for L deletion to take place in a Yoruba DP, the targeted element must be (part of) a possessum NP in a genitive construction.  2.4.5.2 MTS and L-deletion are in complementary distribution  I have argued that the MTS is a genitive marker that occurs between the possessum and the possessor. The final issue to discuss is the relation between the occurrence of the MTS and this L-deletion. M y claim is that the substitution of the default M-tone for the deleted L tone is formally equivalent to the MTS. In particular, I will show that L tone deletion occurs whenever the MTS is not pronounced.  (121) a.  If MTS is pronounced, then no L deletion  b.  If MTS is not pronounced, then L deletes  First, I present data that show that whenever the MTS is pronounced, L-deletion cannot take place (122)-(124).  (122) a.  b. (123) a.  b.  Kunle asa a 'Kunle's habit' *asa  a  Kunle  ogede e 'Kunle's banana' e  Kunle  Kunle  (124) a.  b.  agbaagba a 'elders of Beyiose'  Beyiose  *agbaagba  Beyiose  a  Second, I present data that show that whenever the MTS is not pronounced, L deletion must take place (125)-(127). This is irrespective of the number of the syllables of the possessum NP. Cases where the MTS is absent and we fail to delete the final L tone as shown in the (b) examples are unattested.  (125) a.  b. (126) a.  asa 'Kunle's habit'  Kunle  *asa  Kunle  9g?de 'Kunle's banana'  Kunle  6gede  Kunle  f  (127) a.  b.  agbaagba 'Kunle's banana'  Beyiose  *agbaagba  Beyiose  In summary, I have argued that the overt realization of MTS and L-deletion of the final vowel of the possessum NP are two sides of the same coin.  2.5  Conclusion  I have discussed four major themes relating to genitive constructions in Yoruba. The first two themes dealt with the semantics and the syntax of genitive constructions. First, I proposed that at the semantic level, there is an R relation between the two arguments in a genitive construction. I made the same claim at the syntactic level by claiming that v is the (morphosyntactic) realization of the (semantic) R relation. I argued that the relational head v  may be null (in nominal genitives) or it may be pronounced (in verbal genitives). Still on the syntax of Yoruba genitives, I demonstrated that both nominal and verbal possessive constructions derive from the same base structure: a vP shell. The surface structure of both involves movement: while the possessum moves to Spec D P in nominal genitive constructions, it is the possessor that moves to Spec IP in verbal possessive constructions. I also claimed that the M T S that appears between the possessum and the possessor in the nominal genitive occupies D. Evidence for the raising analysis is found in the account of the phonological content of the genitive marker, which takes the form of the mid tone syllable. My account revealed that the MTS can be analyzed as a syntactically conditioned mora. This implies that while syntax determines where we get the morpheme, phonology determines its form and when it is to be pronounced.  C H A P T E R T H R E E : ^'-GENITIVES A S R E D U C E D R E L A T I V E C L A U S E S  3  Introduction  Chapter 2 established that there is a parallel between the semantic and syntactic structures of the two arguments that are in a genitive relation (R). Thus in (1), the possessor and the possessum are linked together by v (which instantiates the R relation). The possessor occupies the Specifier position of the vP shell whereas the possessum is the complement of v.  (1)  vP Poss'r  —— v  Poss'm  In the surface syntax of Yoruba genitives I argued that the nominal genitive and the verbal possessive both derive from the base structure in (1). In the nominal genitive there is a D P superstructure, which hosts the MTS in D as in (2a). The example in (2b) illustrates this.  (2)  a.  DP  D  vP  MTS Poss'r Poss'm  b.  ile e Kunle house MTS K . 'Kunle's house'  Similarly, in the verbal genitive, there is an IP superstructure that hosts the HTS in I(NFL) as in (3a) with the example in (3b) as an illustration.  (3)  a.  IP  Poss'm ni b.  6ke e ni O. HTS have 'Oke has a house.'  ile house  With nominal genitives the possessum raises to Spec DP whereas with verbal possessives, it is the possessor that raises to Spec IP. And while the v position is null in genitive DPs, in the verbal possessives, the verb ni 'have' occupies v. In this chapter, I extend the analysis to genitive constructions that contain the element ti. Four such constructions will be discussed. First, I establish a relation between the genitive M-tone ti construction (4a) and the H-tone ti relative clauses (4b).  (4)  a.  ere ti Kiinle statue C K. 'statue of Kunle'  N  ti  N  b.  ere ti Kiinle ni statue C K. own 'The statue that Kunle owns'  N  ti  N  VP  Second, I discuss the relationship between the f/-genitive and the M T S . In particular, in addition to the MTS and ti occurring by themselves, as in (5a-b), they may also co-occur as in (5c).  (5)  a.  ere e Kiinle statue M T S K . 'Kunle's statue'  b.  ere ti Kiinle statue C K. 'statue of Kunle'  .  N  MTS N  N  ti  N  c.  ere e ti statue M T S C 'statue of Kunle'  Kunle K.  N  M T S ti  N  Third, I consider the co-occurrence of ti with the verb je in verbal possessor constructions.  (6)  Owo je ti Kiinle money is C K. 'The money is to Kunle.'  N  je  ti  N  Fourth, I compare the M-tone ti found in genitive environments and the M-tone ti found in non-genitive environments e.g. the M-tone ti that functions as a P with preverbal locatives (7a) and the M-tone ti that functions as a P in post-verbal locatives (7b).  (7)  a.  Mo [ti ibe] de n i aaro lsg P place arrive Pmorning T got back from there in the morning.'  ti  N  V  b.  Mo de [lati Kanada] lsg arrive P Canada T arrived from Canada.'  V  ti  N  My proposal is two-fold. I argue that there are instances where M-tone ti behaves like a complementizer and instances where it behaves like a preposition. I propose that the ti in genitive constructions is to be analyzed as C, the head of a reduced relative clause. Second, the ti in non-genitive constructions is to be analyzed as P, the head of a prepositional phrase. What all these constructions have in common is the presence of the ti element. The chapter is organized as follows. In §3.1, I discuss the parallel between ti genitives and ti relative clauses. §3.2 discusses the co-occurrence of the Mid Tone Syllable with genitive ti in relation to Case assignment. In §3.3, I give an account of the co-occurrence of the copula je with genitive ti. §3.4 deals with the prepositional use of ti. §3.5 concludes.  3.1  The parallel between ti genitives and ti relative clauses  This section addresses the parallel that exists between M-tone ti that occurs in genitive construction (8a) and the H-tone ti that occurs in relative clause (8b).  (8)  a.  ere ti Kunle statue C K. 'statue of Kunle'  N  ti  N  b.  ere ti Kunle ni statue C K. own 'The statue that Kunle owns'  N  ti  N  VP  I argue that M-tone ti genitives are reduced relative clauses while H-tone ti constructions are full relative clauses. The first step is to provide theoretical motivation for the claim that ti genitives are a kind of relative clause. To this end, I briefly present the two kinds of structures used for relative clauses.  3.1.1  Two structures for relative clauses  Observe that English relative clauses can have overt Wh-pronouns, as in (9), or overt complementizers as in (10).  (9)  a. b.  [the book [which John read _]] [the man [who you saw _]]  (10)  a. b.  [The book [that John bought _ ]] [The house [that I saw _ ]]  I briefly review two accounts of relative clauses: namely the complementation analysis and the adjunction analysis. I adopt the complementation analysis to account for the Yoruba data.  3.1.1.1 Relative clauses as complements to D  In the complementation analysis of Vergnaud (1976) and Kayne (1994), a relative clause (CP) is treated as a complement to D and the nominal head (NP) is generated inside the relative clause itself. In English, for example, the relativized argument moves from a vPinternal position to Spec CP, in as in (11). (11)  a.  DP D the  CP bookj •  IP that  John; VP  Applying the same analysis to Yoruba let us examine cases involving movement of object NPs. Consider (12a). As shown in (12b), the possessum ere 'statue' raises to Spec CP. (12)  a.  ere ti Kunle ni _ statue C K. own 'the statue that Kunle owns'  b.  CP ere  IP  C ti  Kiinle VP  V ni  It is equally possible for the subject NP to raise as in (13). When this happens in Yoruba, there is a resumptive pronoun at the extraction site (cf. Carstens 1986, 2003, Takahashi and Yuksek 2004, Aboh 2005).  (13)  a.  obinrinj ti 6; ni woman C RP have 'the woman that has a statue'  ere statue  b.  okunrirti ti 6j ni man C RP own 'the man that owns a house'  ile house  c.  omo; ti 6 jogunope child C RP inherit palm tree 'the child that inherited a palm tree'  d.  iyawo ti o ni wife C RP have 'the wife that is stubborn'  i  {  agidi stubborn  I illustrate this with example (13a) as (14). The raised subject N P moves from Spec v P to Spec CP. The resumptive pronoun is spelled out in Spec, IP.  I turn to the second type of analysis that has been proposed in the literature, namely relative clauses as adjuncts to NP.  3.1.1.2 Relative clauses as adjuncts to NP  In what has become the standard account (cf. Carnie 2002 among others), relative clauses are treated as adjuncts. In that approach, a relative clause is adjoined to the head N P and the Head of the relative clause is generated outside the relative clause. The WH-pronoun that appears in these relative clauses is treated as an operator (OP). The operator is basegenerated in the object position and raises to Spec CP; this movement is internal to CP.  DP  (15) D the  NP NPi book  CP OP which t  C  0  IP John read  t  Observe also that the Wh-operator, which is co-indexed with the head of the relative clause book. The co-indexation is to show that "book" and "which" are the same. In the case under review, movement of the operator is overt. Even, when there is no overt operator, it is still assumed that there is covert movement of the operator from the object position to CP, (16).  DP  (16) D the  NP NP book  CP OPi C  IP  that  John read tj  i i i  Using the same approach for Yoruba yields (17): the head noun ere 'statue' is generated outside the CP, and this C P is adjoined to the NP. The null operator covertly moves to the Spec CP from its object position and is co-indexed with the head noun.  (17)  DP D  NP CP  NP ere  Op A  c  IP  tl  Kiinle ni t; i i i  Analyzing Yoruba data using the adjunct clause analysis makes no distinction between a relative clause that has a complementizer and one that has a relative pronoun because there is no overt OP in this context. In Yoruba, only the C position is pronounced. The kind of Wh-elemerits that would function as operators in Yoruba do not appear in relative clauses. This is what I show in (18) and (19) where only the (a) examples are grammatical in the sense that the presence of ewo 'which', (18a) and ta 'who', (19a) is allowed in Wh-constructions and not in relative clauses as the ungrammaticality of the (b) examples show.  (18)  a.  Iwe ewo ni o book which Foe 2sg Which book did you buy? *Iwe book  (19)  ewo o which 2sg  wh-question  ra buy  relative clause  ra buy  a.  Aja a ta ni dog M T S whose Foe 'Whose dog did you see?'  o 2sg  b.  *Aja a ta o ri dog M T S who 2sg see [the dog of who that you saw]  n see  wh-question  relative clause  This tells us that unlike English, Yoruba has no relative pronouns, as such the Spec C P will always be null and movement of OP will always be covert. Next, I turn to the account that establishes that genitive ti and complementizer ft'spell out C.  3.1.2 ti genitives are reduced relative clauses  I argue that ti genitives are reduced relative clauses and the head of this reduced relative clause takes a vP clause as its complement. First, observe that the ft'-genitives shown in (20) have the surface structure of " N ti N " . A s with any other type of genitive DPs, N-ft'-N expressions occur in the range of argument positions e.g. in subject position (20a) and object position (20b).  (20)  a.  [6ro ti Taye] dun mango of T. sweet 'The mango of Taye is sweet.'  b.  M o fo [agbon ti lsg break coconut of 'I broke the coconut of Jenrola.'  subject  Jenrola] J.  object  How are the N-fi-N expressions analyzed as reduced relative clauses? Starting with the proposal of a single base-structure for all genitives in Yoruba, a C P layer is introduced above the base structure as in (21). The ti element, like a true complementizer, occupies C and it takes the v P as its complement. The possessum N P (dro, 'mango')* moves from the vPinternal complement position to Spec CP.  (21)  DP  D  CP poss rrij  One consequence of this proposal is that it brings to light the relationship between the M tone ti genitive marker and the H-tone ti relative marker. In particular, in the proposed analysis, they both spell out C. The contrast between the two is further illustrated in (22) where H-tone ti heads a full relative clause and in (23) where M-tone ti heads a reduced relative clause.  (22)  a.  , ere ti Kunle ni statue C K. own 'the statue that Kunle owns' full relative clause: H-tone tf CP  IP  vP Kiinle ni ere (23)  a.  ere ti Kiinle 0 statue C K. 'statue of Kunle  reduced relative clause: M-tone ti CP  vP Kiinle 0 ere Further, while the M-tone ti found in a reduced C P can only take a small v P shell as its complement, the H-tone ti found in a full C P can only take an IP as its complement. Compared to a full relative clause, the reduced relative clause is deficient in two ways. First, a full CP has an IP (22a), which is lacking in the reduced relative clause, (23b). The reduced relative clause can only take a small vP as its complement. The second deficiency of a reduced C P is that M-tone  contrasts with the H-tone ti  of a full relative clause with respect to the tone. Observe that it is a general property of Yoruba complementizers that they require H-tone (Dechaine 1993). I show this with H-tone ki'm (24a), H-tone pe (24b) and a combination of pe and ki, both of which bear H tone (24c). As already shown, the C of a reduced relative clause bears M tone. (24)  a.  Mo gba ki Ade wa lsg admit C A. come 'I admitted that Ade shouldcame.'  b.  Mo gba pe Ade lsg admit C A. T admitted that Ade came.'  c.  Mo gba pe ki Ade lsg accept C C A. 'I accepted that Ade should come.'  wa came  wa came  Despite these deficiencies, the claim that M-tone ti in a reduced relative clause is a C still holds in the sense that it takes a clause (vP) as its complement. Apart from the fact that H-tone ti and M-tone ti are analyzed as complementizers, the issue of the landing site for the moved NP is also very significant. In the full relative clause, the raised N P lands at Spec, CP. However, in a ti genitive construction, it is possible for the NP to raise to Spec C P as in (21) above or out of Spec CP to Spec D P when D is pronounced.  As shown in (25), the position is pronounced as M-tone (here [e]). It is the possibility of raising from Spec C P to Spec DP which accounts for the co-occurrence of the MTS (in D) with M-tone ti (in C).  (25)  a.  [  iwe [D e book M T S 'Tunde's book' DP  b.  [  CP  [C ti  [ Tunde^v0[t]]]] T.. vP  DP iwe.  D MTS  CP t IP Tunde 0 tj  Yoruba is not the only language whose NP can raise past Spec CP. This is also attested in Romanian. In fact it is obligatory for the raised NP to move as far as Spec DP. The example in (26a) and the structure in (26b) illustrate this: carte 'book' moves from the complement of the verb citit 'read' first to Spec C P and finally to Spec DP.  (26)  a.  carte-a book-the  pe care pe which  am citit-o I.have read-it (Kayne 1994: 88)  DP carteD  CP  IP pe care  am citit- Oj  The summary of the similarities and differences among Yoruba, English and Romanian is given in (27). It shows that the three languages can have their object N P raised to Spec CP.  While English NP must not raise past Spec CP, Romanian obligatorily raises to Spec DP. Yoruba N P on the other hand can either raise to Spec C P and stop, or it can raise to Spec D P if D is pronounced. (27)  DP Rom/Y or CP Eng/Yor IP  vP Poss'r v  t J  The factor that determines how far an NP can raise in Yoruba therefore is the presence or the absence of the MTS. The generalization is this. One, when D is not pronounced, Yoruba N P remains in Spec CP, as in (28a). This is the case with full relative clauses and reduced relative clause that lack overt D. Two, when D is pronounced, Yoruba behaves like Romanian in the sense that the object NP must raise to Spec DP, as in (28b). This accounts for the co-occurrence of MTS with M-tone ti.  (28)  a.  [ iwej [Cti ] [ Tunde book that T. 'the book that Tunde owns' CP  IP  [ owo [Do money MTS 'money of Tunde' DP  ni own  tcpti [ C  ti of  Tiinde 0 T.  til  The table in (29) below captures the case of raising in the three languages discussed here.  (29)  Raising within the genitive construction in Yoruba, Romanian and English Yoruba movement to Spec C P movement to Spec DP  V V  Romanian  English  V V  Finally, there is one major difference between a relative clause and a rf-genitive that I need to point out. Apart from the possibility of raising the possessum/object, (30a) it is also possible for the possessor/subject to raise in a full relative clause, (30b).  (30)  full relative clause a. [ iwe\ [C ti [ omo ra tj]] book that child buy 'the book that the child bought' CP  b.  objlposs'm raising  ]P  [ p9 9i [C titip °i child that RP 'the child that bought the book' m  C  v  ra buy  iwe]]] subjlposs V raising book  However, possessor raising does not take place in a n'-genitive, as seen by the contrast between (31a) and (31b).  1  The ungrammaticality of (31b) is because it is the  subject/possessor that is raised. The failure of possessor raising in nominal genitive is due to the linear order restriction.  (31)  genitive DP with ti-construction a. [cpdrOi [C ti [ Kunle mango of K. 'the mango of Kunle'  [ 0ti]]]  vP  ^fcpKunle; [Cti K. of  v  [yp tj [ 0 6ro]]] v  objl poss'm raising  subj/poss'r raising  The disparity in raising might relate to another diagnostic measure for reduced and full relative clauses. Reduced relative clauses are found only in genitive constructions whereas full relative clauses are not restricted; as such, they can be found in genitive or non-genitive constructions. To allow the subject to raise in reduced relative clauses is to derive an unattested output. These differences are summarized in (32).  A subject NP in a full relative clause = a possessor NP.in a reduced relative clause. Similarly, an object NP in a full relative clause = a possessum NP in a reduced relative clause. 1  (32)  Distinction between full CP and reduced CP H-tone ti = full relative clause Object/possessum raising . V Subject/possessor raising V  M-tone ti = reduced relative clause  V X  The main claim so far is that the M-tone ti constructions involve a reduced relative clause whereas the H-tone ti constructions involve a full relative clause. The next task is to show that the proposal is independently motivated. I will demonstrate that H-tone ti is consistent with Spec-Head agreement in C, and that this occurs in the extended verbal domain. 2  Similarly, I will show that M-tone ti is consistent with Spec-Head agreement in C and that this occurs in the extended nominal domain. I will support this parallel with other parallels between HTS in I(nfl) and MTS in D, and H-tone ni and M tone ni. These parallels between H-tone and M-tone elements are summarized in (33).  (33)  Parallels between H-tone and M-tone elements in Yoruba M-tone ni M-tone ti H-tone ti H-tone ni HTS V V Nominal v' V Verbal V  MTS V X  The discussion in the four subsections is organized as follows. In §3.1.2.1, I establish the parallel between M-tone ti and H-tone ti m terms of agreement. The discussion in §3.1.2.2 focuses on the similarity between the MTS that occupies D and HTS that occupies I(nfl). In §3.1.2.3 I show that the same parallel exists between the M-tone ni that introduces nominal predicates and the H-tone ni that introduces verbal predicates. §3.1.2.4 summarizes the significance of H-tone/M-tone alternations.  3.1.2.1 M-tone ti and H-tone ti  There are two considerations that support the claims that the M-tone ti construction involves a reduced relative clause whereas the H-tone ti construction involves a full relative clause: theory internal and language internal (i.e., empirical) evidence. On the latter, I argue that the :  cf. Dechaine 1993's "default agreement".  H-tone ti is an Agreeing Complementizer marking Spec-Head agreement in the extended verbal domain, whereas the M-tone ti is an Agreeing Complementizer that shows up in the extended nominal domain . 3  There is some indication that agreement as the head of the subject relative clause is associated with a resumptive pronoun, which may or may not agree with its antecedent. The 4  fact that there is some form of agreement is an indirect evidence in support of my claim that H-tone ti shows Spec-Head agreement in Yoruba. For example, a 1 or 2 person antecedent st  nd  may be associated with a resumptive pronoun, which agrees in person and number, or the resumptive pronoun may surface in the 3 person form. This is shown for 1 person (singular rd  st  and plural) as well as 2 person (singular and plural) in examples (34)-(37). nd  (34)  (35)  (36)  (37)  3  4  a.  Emi ti mo nf lsg C lsg have T that own a statue sold it.'  ere ta statue sell  a 3sg  b.  Emi ti 6 nf lsg C 3sg have T that own a statue sold it.'  ere ta statue, sell  a 3sg  a.  Iwo ti o nf ere ta 2sg C 2sg have statue sell 'You that own a statue sold it.'  a 3sg  b.  Iwo ti 6 hi ere ta 2sg C 3sg have statue sell 'You that own a statue sold it.'  a 3sg  a.  Awa tf a nf ere ta lpl C 3pl have statue sell 'We that own a statue sold it.'  a 3sg  b.  Awa ti 6 nf ere ta lpl C 3sg have statue sell 'We that own a statue sold it.'  a 3sg  a.  E y i n tf e nf ere ta 2pl C 2pl have statue sell 'You that own a statue sold it.'  a 3sg  Cf. que/qui alternation as a form of agreement in French. See Dechaine (1993) for a treatment of the 3rd person singular form in Yoruba as default agreement.  b.  E y i n tf 6 nf ere ta 2pl C 3sg have statue sell 'You that own a statue sold it.'  a 3sg  As for the 3 person forms, with 3sg there is no alternation between the agreeing form and rd  the non-agreeing form of the resumptive pronoun since the resumptive pronoun is itself 3sg, (38).  (38)  a.  6un tf 6 nf ere ta 3sg C 3sg have statue sell 'S/he that owns a statue sold it.'  a 3sg  b.  Okunrin tf 6 nf ere ta man C 3sg have statue sell 'The man that owns a statue sold it.'  a 3sg  Finally, a 3 person plural pronoun (39) or common noun (40) antecedent may be associated rd  with a resumptive pronoun, which agrees in person and number, as in the (a) examples or the resumptive pronoun may surface as 3sg, as in the (b) examples.  (39)  a.  A w o n tf won nf ere ta lpl C 3pl have statue sell 'They that own a statue sold it.'  a 3sg  b.  A w o n tf  a  6  nf  ere  ta  i  (40)  lpl C 3sg have statue sell 'They that own a statue sold it.'  3sg  a.  Okunrin meji ti won ni man two C 3pl have 'The two men that own a statue sold it.'  ere ta statue sell  a 3sg  b.  Okunrin meji tf 6 nf man two C 3sg have 'The two men that own a statue sold it.'  ere ta statue sell  a 3sg  I conclude that the agreement that optionally shows up between the head of subject relative clause and its resumptive supports the assumption that H tone ti is an indication of agreement in a relative clause.  3.1.2.2 MTS and HTS Another parallel between H-tone ti and M-tone ti relates to the HTS found in I(infl) in the verbal domain and the MTS found in D in the nominal domain. In Yoruba, HTS is obligatory between a full DP subject and a bare verb (Dechaine 1993: 457). While the HTS instantiates Spec, Head agreement in IP (41), MTS spells out Spec, Head agreement in DP, (42).  (41)  a.  Ebun u n j tete E. HTS eat lottery 'Ebun won a lottery.' e  ) b.  (42)  a  IP  iwe e Tunde book M T S T. 'Tunde's book' DP Poss'mi Zx  vP  D  iwe  MTS  Poss'r Tunde 0  I suggest that just as HTS in the verbal domain parallels M T S in the nominal domain, so too does the H-tone ti of full relative clauses parallel the M-tone ti of the reduced relative clauses. Moreover, just as HTS occurs in the verbal domain so does H-tone ti. And just as MTS occurs in the nominal domain, so does M-tone ti.  3.1.2.3 H-tone ni and M-tone ni  I will now show that the contrast between H-tone ni and M-tone ni provides another indirect support for the claim that H-tone ti spells out Spec-Head agreement when C P dominates IP, while M-tone ti spells out Spec-Head agreement when C P dominates a defective nominal clause. The occurrence of the H-tone m in a clause shows a verbal predicate 'have'. That this ni is a verb is confirmed by the presence of the HTS, (43a). The ungrammaticality of (43b) shows that whenever verbal H tone ni is present, it is obligatory for the HTS to be present.  (43)  a.  Agbe e ni owo farmer HTS have money ' A farmer has money.'  b.  *Agbe0 farmer  ni have  Tiinde T  The behaviour of H-tone ni contrasts with M-tone focus/copular ni, which is never preceded by HTS. This is because, as observed in Abraham (1958:435), M-tone ni is not a true verb as shown by the ungrammaticality of (44b). The absence of HTS with copula ni may be symptomatic of the fact that this copula only introduces nominal predicates. Thus, when we make it behave like a true verb, which requires a HTS as in (44b) then the output is un grammatical.  (44)  a.  Agbe ni farmer be 'Tunde is a farmer.'  Tiinde T.  b.  *Agbe farmer  ni be  e HTS  Tiinde T.  The contrast between H-tone m and M-tone ni is illustrated with the structures in (45).  (45)  a.  verbal predicate IP  HTS  b.  ni owo  nominal predicate IP Agbe AGR  NP  ni  Tunde  In summary, I have shown that H-tone ti and M-tone ti have certain things in common that makes it appropriate to analyze them as C: they both take a clause as their complement. I have shown supporting language internal evidence that M-tone ti is to be analyzed as the head of a reduced relative clause in nominal domain. Such evidence includes the MTS that occupies D and M tone ni in nominal predicates. The final thing that I do is to show the significance of these findings.  3.1.2.4 The significance of H-tone/M-tone alternations  To summarize, this section has established that both ti and ti are Agreeing complementizers in Yoruba. The generalization that emerges is that every occurrence of H-tone mora shows agreement in the verbal or extended verbal domain whereas every occurrence of M-tone mora shows agreement in the nominal domain. This holds of H-tone ti versus M-tone ti, of HTS versus MTS, and of H-tone ni versus M-tone ni.  3.1.3  The ti N construction  I extend the account of genitive DPs as reduced relative clauses to another type of construction in which ti appears but which is at least in one respect different from the ones we have so far seen. It contains just the r/-element and a possessor N P as opposed to the genitive DPs in reduced relative clauses, which contain Possessum ti Possessor. This type of construction has been referred to as 'Possessive noun phrases with no Possessee' (Hertz (1997: 517), i.e. a genitive D P that lacks an overt possessum N P since the object that is possessed is phonologically null. Because such genitives lack an overt possessum, the latter must be discourse linked.  (46)  a.  Context: Imagine books belonging to Kunle, Olu and Ade are missing and somebody asked if I see any of the books, then I replied with (46b).  b.  Mo ri [ti lsg see of 'I saw Kunle's. '  Kunle] K.  5  The idea of interpreting (46) as a genitive construction raises a number of questions. First is the question of whether there is any semantic and or syntactic relationship between this construction and the type that has an overt possessum NP, compare (47a) that has an overt possessum NP with (47b), which lacks an overt possessum NP. (47)  a.  Mo ri [oko ti lsg see farm C 'I saw the farm of Ojo.'  Ojo] O.  b.  Mo ri {pro lsg see 'I saw Tunde's.'  Tunde] T.  ti C  The similarity in meaning between (47a) and (47b) is not in doubt. Again, the only difference is that (47b) cannot be rendered in the out-of-the-blue but it can follow (47a) in discourse. The same is applicable in (48). In an appropriate context, the possessum of ti Tunde is  5  Note that though  ti tiinde translates to English Saxon's: Kunle's, it has a different structure.  recoverable. Since in the context of (48), it has the same interpretation as aso ti Kunle, we expect-that the syntax will be the same on the basis of the parity in meaning. Note that the occurrence of aso 'cloth' in the first genitive DP in (48b) makes its occurrence in the second DP "awkward". (48)  a.  Mo ri [aso ti Ojo], lsg see cloth of O. , sugbon n ko ri [pro ti Tunde] but lsg N E G see of T. 'I saw the dress of Ojo but I didn't see Tunde's own.'  b.  Mo ri [aso ti Ojo], lsg see cloth of O. sugbon n ko ri #[aso ti Tunde] but lsg N E G see cloth of T. 'I saw the dress of Ojo but I didn't see the dress of Tunde.'  There are two accounts for ti N constructions. One claims that ti is a complementizer (herein) while the other claims it is a noun (Awobuluyi 2004).  3.1.3.1 ti as Complementizer analysis We can account for n'-constructions that lack a possessum N P the same way we account for fi-constructions with an overt possessum NP. As such, I propose that this ti element is the same as the one we find in rf-genitives, so it must be analyzed as a complementizer. I propose in line with the current analysis that the ti-N construction should be analyzed as a full DP and its possessum NP be represented as a null pronominal (pro). In this case, it is the pro possessum that covertly moves to Spec CP.  (49)  a.  DP  D  CP  0 C ti  vP Poss'r Tunde  L  v  pro  0  This movement is independently motivated. In cases with an overt possessum, it is obligatory that the possessum raises to Spec CP. As we saw in the earlier account with overt possessums, the pro cannot raise beyond Spec C P if the D position is null, otherwise we derive a wrong output, at least in M o b a (50), where the genitive MTS is' invariant I'll.  (50)  *i MTS  ti C  Ojo O.  There is a sharp contrast when we try to compare Yoruba 'ti N ' and English ' o / N ' . For example, the phrases in (51) are not well formed since those PPs cannot exist independently in English. (51)  a. b.  *ofJohn *of Mary  In English, the possessum NP must be pronounced in ^/--constructions.  (52)  a. b.  book of house of  John Mary  What accounts for this difference is not presently known. However, the examples in (51) actually have counterparts in so-called Saxon genitive constructions that are well formed (Fukui 1986, Siloni 1997, Hertz 1997).  (53)  a. b.  John's Mary's  Observe that Saxon (Y) genitives can take 'own' as gap filler for the missing possessum, (54) whereas o/-genitives cannot, (55). (54)  a. b.  John's own Mary'sown  (55)  a. b.  *own of *own of  John Mary  Finally, the difference between the Saxon 's genitive and the 'of genitive is the reason why it is possible to drop 'own' in the former but not in the latter. As for the Yoruba MTS genitives, they do not allow the elision of the possessum NP as shown by the ungrammaticality of (56b). (56)  a.  b.  ile e Kunle house M T S K . *e  Kunle  To summarize, under the present account the ti in ' N ti N ' and the ti in 'ti N ' are both analyzed as a functional element that occupies C. The data poses a challenge to an alternative analysis, which treats ti as an NP. I discuss this other analysis in what follows.  3.1.3.2 ti as N analysis: Awobuluyi (2004) Awobuluyi (2004) argues that the //-element found in nominal environments as discussed in the previous sections is itself a noun, which by implication must have the distribution of a DP. According to Awobuluyi:  the element ti must be some kind of nominal...and, therefore, have the feature [+N], since it requires to have the prosthetic vowel i attached to its left margin within noun phrases... Pointing in this same direction is the further fact that the element readily enters into construction with the possessive qualifier re 'his/her/its' and, together with the latter, functions as subject or object (Awobuluyi 2004: 14).  The examples below are used to support this claim. In particular is (57b) where the MTS, which undergoes assimilation attaches to ti.  (57)  (58)  a.  aja ti Kunle dog that-of Kunle 'Kunle's dog'  b.  aja a-ti dog MTS-that-of 'Kunle's dog'  a.  Ti re pupa that-of his/her/its be-red/brown 'His/Hers?its is red or brown.'  b.  Mo ri ti re I saw that-of his/her/its 'I saw his/hers.' (Awobuluyi 2004: 14-15)  Kiinle K.  However, the claim made by Awobuluyi that ti is a kind of noun makes incorrect predictions regarding morphology, semantics, and syntax of ti. First, the morphological shape of all canonical nouns in Yoruba is minimally V C V . Some representative examples are given in (59).  (59)  a.  owo 'money'  b.  ilii 'town'  c.  ede 'shrimps'  d.  apo 'bag'  e.  eja 'fish'  As for ti, it is just a C V . Based on this morphological criterion ti does not qualify as a noun. It requires at least a prosthetic vowel to make it conform to the shape expected o f most  nouns. Though adding I'll to ti as Awobuluyi's work suggests, satisfies the morphological requirement, it creates another problem namely, that no such word as *iti exists in any of the Yoruba dialects (not even in M o b a which is famous for its use of vowel / i / in consonant initial nouns) as the following examples illustrate.  (60)  6  Moba a. i-pako 'plank' b.  i-korfko 'grass'  c.  i-fere 'whistle/horn'  d.  i-dada 'personal name'  e.  i-bata 'a kind of drum'  <  And for consonant initial nouns that allow the i-prefix, such nouns must be at least two syllables: C V - C V not one C V as in the case of ti. In (61), each example has two forms. The first begins with a consonant the second begins with the prosthetic vowel. They all consist of two syllables (CV-CV)  (61)  Standard Yoruba/ M o b a a. yanrin i-yanrin 'sand' b.  busun i-busun 'sleeping place'  c.  buso 'mileage'  i-buso  d.  lekun 'door'  i-lekun  Observe that in Standard Yoruba, theses examples are without this prosthetic vowel. See Ajfboye (1991) for relevant data and discussion.  6  The second problem is semantics. Either in citation or in context, the meaning of ti as a noun is not transparent. While the whole sentence is interpretable in (62a) because each word has its meaning, the sentence in (62b), is not. I trace this to ti, which lacks its own meaning because it is a functional element.  (62)  a.  Ade ri iwe A. see book 'Ade saw a book.'  b.  *Ade ri  ti  Another piece of evidence is the contrast between ti and bi. The latter has been treated as a defective noun because it can assign genitive Case (Dechaine 1993, 2001). According to 7  those accounts, this lexical item bi can be analyzed as the nominal head of a relative clause, roughly meaning ' [the] way [in/by which]'. Dechaine gave five pieces of evidence to show that bi is a noun. One that is relevant to our discussion assuming that truly bi is a noun is that ' a s N , ^'licenses Genitive Case' (Dechaine (1993: 95). (63)  a.  ...ki a sise [DP bi i that lpl work way Gen '.. .that we work in the manner of a slave'  eru] slave  b.  jimo 6 wi ejo bi i awewa like Agr say case way Gen grumbler 'Jimo complains [in] the way of a grumbler' (Dechaine 1993: 95)  In contrast, ti cannot occur with MTS.  a.  *Mo lsg  ri see  ti of  i MTS  re 3sg  b.  *Mo lsg  ri see  ti of  i MTS  wa lpl  8  In my own view, bi is defective as noun for the same morphological reason as ti is: they both consist of CV only, nouns must be minimally be V-CV. For this reason, I do not consider bia noun as well. For more on the treatment of bi as a noun, see also Abraham (1958), Awobuluyi (1978), Oyelaran (1982) and Dechaine (2001). 7  8  Another piece of evidence that supports the claim that ti is not a noun is the use of the MTS as a diagnostic tool. Only when a true possessum N P appears with the possessor must we have the MTS:  (65)  (66)  a.  iwe *(e) [Tunde] book M T S T. 'Tunde's book is red.'  pupa red  b.  ti of  pupa red  a.  iwe *(e) [re book M T S his/her/its 'His/Her book is red.'  pupa red  b.  ti of  pupa red  (*e) [Tunde] M T S T.  (*e) [re M T S his/her/its  Further, if 'the element ti must be some kind of nominal' as Awobuluyi claims, then, we must be able to replace ti with a possessum noun. The example in (67b) shows that we cannot replace ti ' o f with iwe 'book', as the output is not well formed. (67)  a.  Ti re pupa that-of his/her/its be-red/brown 'His/Hers/Its is red.' (Awobuluyi 2004: 14-15)  b.  *Iwe re book his/her/its  c.  Iwe e re book M T S his/her/its 'His/Her book is red.'  pupa be-red/brown pupa be-red/brown  Observe that in (67a) where ti is present, the M T S is absent simply because ti is not a possessum. The pro does not trigger MTS because it has no phonological content. However, in (67c) where an overt possessum is present, it requires the presence of the MTS. The absence of the MTS yields a wrong output. This is another evidence that shows that ti is not a possessum noun. Further evidence that supports the claim that ti is not a possessum noun comes from  the fact that it can co-occur with a possessum like iwe, as in (68). Thus, iwe is neither sitting in the same position as ti nor is it referring to the same entity iwe. (68)  Iwe e ti book M T S of 'His/Her book is red.'  re his/her/its  pupa be-red/brown  The final piece of evidence comes from what constitutes a genitive phrase. In a possessive phrase, there is usually one possessor and one possessum. In (68), it is indisputable that iwe and re are the only two nominals. There iwe is the possessum and re is the possessor. For this reason ti and iwe cannot both be possessums. In this case, ti is not a constituent independent of the noun it follows. Further, the complement of ti i.e., the possessor cannot be stranded. This is illustrated with the example in (69c).  (69)  a.  Mo ri [ti lsg see C T saw Ade's.'  b.  [Ti AdeJ ni mo of A. FOC lsg 'It is Ade's that I saw.'  ri see  [tj  c.  *Ade ni  mo  ri  [ti  _]  A.  lsg  see  of  FOC  Ade] A.  Since Yoruba permits movement of the possessor, if this possessor has to move, it must satisfy the stranding condition, (70).  (70)  Stranding Condition When a Possessor moves, a resumptive pronoun that agrees in number and person is obligatorily inserted at the base position of the moved possessor DP  I show this with the example in (71) and (72). The violation of this stranding condition, as in the (b) examples, leads to ungrammaticality.  (71)  (72)  a.  [Ti Adei] ni mo of A. FOC lsg 'It is Ade own that I saw.'  b.  [Adej] ni mo ri [ti rej A. FOC lsg see of 3sg 'It is Ade, whose "x" I saw.'  c.  *Adei ni mo A. FOC lsg  a.  [Ti Ade ati Oliij] ni mo of A. and O. FOC lsg 'It is Ade and Olu's own that I saw.'  ri see  b.  [Ade ati Oliij] ni mo ri [ti A. and O. FOC lsg see of 'It is Ade and Olu, whose "x" I saw.'  wonj 3pl  c.  *[Ade ati A. and  [ti of  Oliij] O.  ri see  ri [t,] see of  [ti of  ni mo FOC lsg  ti]  ri see  [tj  t]  With all the evidence and arguments above, I conclude that ti is not a noun in Yoruba. Rather, ti is a functional head (C). And whenever the possessor moves, ti must also move or else there is a resumptive pronoun in place of the moved possessor (cf. Awoyale 1985).  3.1.4  Assigning genitive case to the possessor  There are two things to remind readers about the discussion on the genitive DPs so far. First, in the surface syntax, the two arguments within this phrase are related by the M T S , (73a) or by ti in the genitive plus ^'-construction, (73b).  (73)  a.  owo o Kunle money MTS K . 'Kunle's money'  b.  ere ti Kiinle statue of K. 'the statue of Kunle'  Second, both the MTS and the ri-element can co-occur, (74).  (74)  a.  aja a ti Bunmi dog M T S of B. 'dog of Bunmi (as opposed to someone else's dog)'  b.  ile e ti Dada house MTS of D. 'house of Dada' (as opposed to someone else's house)'  Having established the status of M T S and M-tone ti as functional heads (D & C respectively), in what follows, I discuss the function that they perform. I argue that these two elements are genitive Case assigners. In the next two subsections, I present two analyses of genitive Case assignment that is internal to genitive DPs, namely the raising analysis (Kayne 1994, Borsley 1997, 2004 and Bianchi 2000) and the traditional complementation analysis (Dechaine 2001).  3.1.4.1 The co-argument analysis: exceptional Case marking  I propose that the assignment of genitive Case to the possessor in Yoruba involves a kind of Exceptional Case Marking (cf. Haegemann 1994, Lee 1995) where the M T S in D assigns Genitive case to the possessor argument in Spec vP, as in (75). (75)  Exceptional Case marking  r  DP  This configuration is an instance of Exceptional Case Marking since it is exceptional for an argument to be assigned Case by an element outside of the projection that contains the argument (here vP). Observe that in (75) and elsewhere, within the genitive DP, the possessum N P must raise to either Spec C P or Spec DP. There is a question of what motivates movement of the possessum out of the complement position to Spec DP/CP. One reason for possessum raising is to receive non-genitive Case.  3.1.4.2 The Head-Complement analysis: Case under government In the complementation analysis, where the possessum is the head noun and the possessor is the complement, genitive case is locally assigned to the DP. In this instance, the genitive Case assigner is P or K(ase) (Dechaine 2001). This is what (76) illustrates.  (76)  a.  b.  ile e Jimo house MTS J. 'Jimo's house' DP  D  NP N ile  KP K ^ " ^ D P e  G  e  n  jimo  (adapted from Dechaine 2001)  In the current analysis genitive Case is not assigned locally since D , which is the Case assigner occupies D and D in my proposal is introduce in the layer above the vP shell, the Spec of which the Possessor occupies.  3.1.5  Assigning structural case to the raised possessum  I address three things in the following subsections: the problem of raising in relation to Case assignment (§3.1.5.1); how the raised possessum is assigned the N O M Case if in the subject position, (§3.1.5.2); and A C C Case if in object position, (§3.1.5.3).  3.1.5.1 The problem  The problem of Case assignment in the raising analysis is not peculiar to Yoruba. Similar problems have been observed for English (Kayne 1994; Borsley 1997) and Polish (Borsley 1997). Let's consider the English example in (77), where 'picture' has moved from its object position to Spec CP.  (77)  a.  [  DP  the [  CP  A [ Bill liked which pictureTil ip  I b.  [ pthe [CP  . [  D  DP  which picture] [ Bill liked  I c.  [DP the [  picture  t ]]]  IP  : NP  I DP  I [CP  [DP which t ] [ Bill liked]]] NP  IP  The explanation offered by Kayne as to why the NP 'picture' moves is faulted by Borsley, who notes that the problem of raising has nothing to do with Case assignment. Borsley notes that picture has been assigned Case by which that occupies D position ever before it raises. Note that pronouns are regarded as D in Kayne's analysis. This is because, ever before movement, which a D element, governs the NP: picture. The other problem noted by Borsley is that while Kayne claims that the NP receives Case from a higher D, this same NP receives Case from its trace. This, according to Borsley, leads to Case conflict since this amounts to duplication of Case assignment. What Borsley 9  considers a problem is not so in the current analysis as such duplication will be regarded as an instance of Case stacking. Cf. free relatives in German (Wiltschko personal communication).  A similar problem arises for Polish, where an N P and its relative pronoun are assigned different Case because of movement of the NP. In (78), the NP pana 'man' receives Accusative Case while the relative pronoun kidry receives Nominative Case. The problem is why should the NP and the relative pronoun receive a different Case.  (78)  Widzialem tego pana, kiory zbil ci szybe7 saw-lSG the-ACC man-ACC who-NOM broke your-SG glass-ACC T saw the man who broke your glass.' (Borsley 1997: 638).  Again, in the current proposal, Polish case is another instance of Case stacking. The following two subsections show how both nominative Case and accusative Case are assigned to nominals within genitive constructions in Yoruba.  3.1.5.2 Assigning Structural case to the raised possessum I: Nominative  The problem I aim at addressing here is which of the two arguments within a given genitive DP receives which Case. I propose that when a D P is in the subject position, Nominative Case is assigned to the Possessum. On this view, Possessum raising is necessitated by the need to receive N O M Case.  (79)  Assigning nominative Case to Genitive DP  [+NOM] D  vP Poss'r ^  tp '  OSS m  [0 Case] There is one instance of grammatical Case assignment in Yoruba. In subject position the entire D P is qualified for nominative Case assignment. Observe that Nominative Case is assigned by the HTS in I(nfl). In (80), this Nominative case assigner shows up as a H-tone  syllable, here ' a ' (whose vocalic content is determined by assimilating the vocalic features of the immediately preceding vowel). Note the similarity between the M T S that assigns Gen Case and the HTS that assigns Nominative Case: they both take the segmental copy of the preceding vowel. The difference lies in tone. M T S is found in nominal environments, whereas HTS is found in verbal environments.  (80)  [lie e Gbada] house MTS.Gen G. 'Gbada's house got burnt.'  a HTS.NOM  jona burn  The N O M Case assigned to possessums is another instance of Exceptional Case Marking. The same is true of assigning A C C Case to the possessum, which I consider next.  3.1.5.3 Assigning Structural case to the raised possessum II: Accusative  If in the object position, the entire DP is disposed to receive accusative Case. The A C C Case assigned to the possessum NP within the genitive DP is another instance of Exceptional Case Marking. I show this in (81), where even though the whole DP is assigned the A C C Case, the possessum is the recipient as shown by the arrow. (81)  Genitive D P in object position  Poss'm [+ACC]  ~ D  vP Poss'r [+Gen]  -— v  t  Possm  In (82) the L-tone verb ra 'buy' assigns Accusative Case to gkg g Tafa 'Tafa's vehicle'. There is a syntactically conditioned phonological processes described as Yoruba L-raising (Ajfboye et al 2004) which raises the L-tone of the verb to M .  1 0  This case has been analyzed as L tone drops in DSchaine (2001). According to that approach, this 'L-drop' is a kind of Accusative Case marking. 10  (82)  Accusative Case Jimo ra [oko 9 J. buy+ACC yam MTS+Gen 'Jimo bought Tafa's vehicle.'  Tafa] T.  This ends the discussion on the external Case assignment. I turn to Case assignment that is internal to the genitive construction.  3.2  Genitive Case assignment and co-occurrence of the MTS with genitive  I have shown that it is possible for the MTS and ti to co-occur in a genitive D P in which case both D and C positions are pronounced. What I intend to discuss here is the function they perform when the MTS and ti occur separately as well as when they co-occur. I argue that each of them can assign Genitive Case. When each of them independently assigns Genitive 11  Case, this is an instance of Case alternation (§3.2.1). I also show that they can jointly assign Genitive Case, this is what I refer to as Case stacking (§3.2.2).  3.2.1  Case alternation: Mid Tone Syllable or ti r  The assignment of genitive Case can be carried out either by the M T S or the M-tone tielement. The factor that determines which of them will assign Case depends on the landing site of the raised possessum. When the possessum NP raises to Spec DP, it is this MTS that assigns Genitive Case. This is illustrated in (83b) where the MTS assigns Genitive Case.  (83)  a.  iwe e Kunle book MTS K . 'Tunde's book'  " There is an alternative analysis that will treat ti and MTS as inherently morphologized Case elements rather than Case assigners. Such alternative is not considered in this study.  DP  POSS'M  The claim made for the MTS with respect to Case assignment is also applicable to ti, namely when it is the only functional element present in genitive phrase, it assigns Gen Case to the possessor. This implies that when the possessum raises to Spec CP, the ^-element, which occupies C is the only functional element that is available and as such, it assigns Gen Case to the possessor. This is what the structure in (84b) illustrates.  (84)  a.  12  obe ti Ebun knife C E. 'knife of Ebun' DP  D  CP Poss'm vP ti  Poss'r V  ^Poss'm  0 In summary, I have shown that each of the two functional heads can independently assign genitive Case to the possessor NP. The next thing I consider is how genitive Case is  One consequence of this analysis is that we expect the H-tone ti to be able to assign Case in relative clauses. For example, in English, (i),/or is assumed to be a complementizer which assigns Case. (i) I want for Serena to leave. This is going to be one focus for future research. 12  assigned when both elements co-occur.  3.2.2  Case stacking: Mid Tone Syllable and ti  We have seen that either the MTS or M-tone ti may intervene between the possessum and the possessor. In addition, it is possible for the MTS and ti to co-occur as in (85). (85)  a.  Mo ri [ile e ti lsg see house M T S C 'I saw the house of Tunde.'  Tunde] T.  b.  [He e ti Tiinde] house M T S C T. 'The house of Tunde is tall.'  ga be-tall  I propose that the co-occurrence of these two functional heads (MTS as D, and ti as C) is an instance of Case stacking. I show that Yoruba Case stacking has an interpretive effect and results in emphasis on the possessor. First, I define Case assigner stacking in Yoruba as the phenomenon whereby a DP is associated with two Case assigners. Rather than claiming that each argument is assigned genitive Case by each Case assigner, I propose that both the MTS and the /i-element multiply assign genitive Case to the possessor. (86)  DP  The Yoruba type of Case stacking contrasts with the kind of Case stacking reported in  the literature where one NP is assigned two different Cases. In Guugu Yalanji, it is possible for a Possessor to be assigned Genitive Case as well as Ergative Case, as in (87) where ndamun marks Genitive Case and du marks the Ergative Case on the noun Dicki.  (87)  Dicki-ndamun-du Dick-GEN-ERG 'Dick's dog'  kaya-ngka dog-ERG (Sadler and Nordlinger 2001: 4)  Similarly, in Korean, Case stacking involves co-occurrence of eykey and -ka, which assign Dative/Locative and Nominative Cases respectively to Cheli, as illustrated in (88).  (88)  Cheli- eykey-ka ton-i manh-ta CD A T - N O M money-NOM a.lot-DECL 'It is Cheli who has a lot of money.' (Yoon 2004: 4)  The same phenomenon has been reported in many Australian languages with non-nominative subjects (Nordlinger 1998; Sadler and Nordlinger 2004) and in Korean (Yoon 2004) and references cited therein.  13  In the case of Yoruba reported herein, there are two distinct  elements that assign the Case. There remains an unresolved problem with respect to the Case stacking analysis proposed for Yoruba. This is in relation to the position of the M T S to the possessor. Recall that on independent grounds, I have shown that each of the M T S and M-tone ti assigns genitive Case to the Possessor under Exceptional Case Marking. The proposal that the MTS co-assigns genitive Case with M-tone ti is unusual since the M T S is not in the right configuration for Case assignment. The Exceptional Case Marking as it is discussed in the literature does not usually extend to cases where the Case assigner and the Case assignee are not in a local relation.  One characteristic feature of Case stacking is that it is sensitive to syntactic function (Sadler and Nordlinger 2004: 1). In Korean it is in non-nominative DPs, whereas in Yoruba as reported herein, it is in genitive DPs. 13  3.2.3  Interpretive effects of Case stacking  Yoruba Genitive Case stacking has a semantic effect: a possessor that is multiply assigned Genitive Case has an emphatic interpretation. Following Bamgbose (1966), an emphatic genitive DP involves emphasis on the possessor. This kind of genitive is marked by both the MTS and ti. To test this claim, consider the examples in (89) and (90). (90b) is more emphatic than (89b) as it indicates that the statue being referred to is only that of Awolowo. (89)  (90)  a.  Context: I went to Bodija and saw Awolowo's statue. When I got back home, I say (89b) to my wife.  b.  Mo ri [ere e Awolowo] lsg see statue MTS A. T saw the statue of Awolowo at Bodija.'  a.  Context: I was told that there are many statues of Nigeria leaders erected at different locations in Bodja but I only saw Awolowo's. When my wife asked whose statue did I see, I say (91b). ,  b.  Mo ri [ere e ti Awolowo] lsg see statue MTS C A. T saw the statue of Awolowo at Bodija.' (as opposed to say statue of Murtala Mohammed)  ni Loc  Bodija B.  ni Loc  Bodija B.  This emphatic force of multiply Case-marked genitives is further exemplified in (91), where there is a comparison between Kiinle and Tunde's statues. The presence of ti in ere e ti Tunde makes it more emphatic and explains the reason why it is Ade's preference. (91)  14  (i)  14  [Ere e Kiinle] dara sugbon [ere e ti Tiinde] ni Ade fe statue M T S K . nice but status MTS of T. F O C A . want 'Kiinle's statue is fine but it is the statue of T U N D E that Ade wants.'  Recall that MTS is optional when the possessum is M-tone final. a. igd o ti Tiinde bottle MTS C T. 'the bottle of Tunde' b. ere (e) ti Tiinde M statue MTS C T. 'the statue of Tunde' c. owo o ti Tiinde H statue MTS C T. 'the money of Tunde'  MTS  (MTS)  MTS  The claim being made for Yoruba has support in the literature. In Korean, Yoon (2004: 6) claims that Case stacking also gives rise to a focus-like interpretation in nonnominative subjects.  15  In (92), a Dative-marked nominal that is an underlying Object  'advances' to become a surface Subject which in turn means it can be marked with N O M Case. This leads Yoon to conclude that Cheli has a focus interpretation. (92)  a.  Cheli-eykey-ka ton-i manh-ta Cheli-DAT-NOM money-NOM a.lot-DECL 'Cheli has a lot of money.'  b. ^  IP — ^  Exp[Dat-Nom] [NOM]  Case ('Cheli') = D A T (in Sp V ) = N O M (in Sp V or Sp I) F  = DAT-NOM  t  VP  I[+nom]  •L  ...Focus interpretation (due to position)  Exp[Dat] [Norn] ThmV [Nomf  V :<Exp[Dat], Thm> or <Exp, Thm> (Yoon 2004: 6)  Thus the structural position of Cheli as a non-nominative subject earns it a focus interpretation. Notice that the kind of Korean Case stacking is applicable to different Cases, whereas the one discussed for Yoruba involves multiple assignment of the same Case by two different Case assigners. Recall that when there is one Case assignment, there is one movement. This is witnessed if either the M T S (83), or M-tone ti (84) is the Case assigner; the Posssessum moves once. However when there is Case stacking there is multiple movement of the Possessum, (86). In other words, Yoruba Case-stacking is a diagnostic for multiple movement of NP. Similarly, in Korean, such multiple movement of N P has the same effect.  See Schiitze (1996, 2001) for arguments against the view that Case Stacking involves focus.  Relating multiple movements to Case, the claim is that multiple movements of the Possessum induce a focus effect. To conclude, I have shown that Case assignment can be carried out by the MTS or ti. I treat this as an instance of Case alternation. And when the two jointly assign genitive Case to the possessor, I call this an instance of Case stacking. In this latter case, in addition to being Case marked, the possessor is made prominent by way of emphasis.  3.3  The co-occurrence of the copula je with genitive ti  I extend my analysis to yet another kind of construction where the ti element is found, namely in a verbal construction involving the verb je. Some examples are given in (93). (93)  a.  lie yif je ti Kiinle house Dem is C K. 'This house is to Kiinle = This house is Kunle's.'  b.  Owo money ='That = 'The  c.  lie yen je ti baba a mi land Dem is C father MTS lsg 'That land is to my father.' = That land belongs to my father.'  naa je ti Bunmi spef is C B. very money is to Bunmi.' very money is Bunmi's.'  Like the //-construction, the je ti construction involves a Possessum-Possessor co-argument relation. I propose that in both contexts, ti heads a defective CP, as in (94). (94)  a.  DP  D  CP  (MTS) C ti  vP  b. V  L je  DP D  CP  C  vP  ti The verb je is also found in equative constructions where two nominals refer to the same person or object. In (95a), agbe 'farmer' and gba 'king' are equated with each other. Similarly, in (95b) Bgse and gmg rere 'nice girl' refer to the same person.  (95)  a.  Agbe je oba farmer is king 'Farmers are king.'  b.  Rose je omo B. is child 'Bose is a nice girl.'  16  Subject Predicate  rere nice  In addition to the je copula, Yoruba has another copula, ni. The je copula contrasts with the ni copula in the following way: while (95) has a Subject-Predicate order (96) has a PredicateSubject order. Thus, in (95a) with je, the first nominal, agbe 'farmer' is the Subject, and the second nominal, gba 'king' is the Predicate. However, in (96a), with ni, the first nominal, gba 'king' is the Predicate, and the second nominal, agbe 'farmer' is the Subject.  (96)  a.  Qba ni agbe (je) king FOC farmer is 'Farmers are king.'  Predicate Subject  b.  O m o rere ni Bose (je) child nice FOC B. is 'Bose is a nice girl.'  cf. predicate inversion (Moro 1995) For discussion of the [Subject-Predicate] versus [Predicate-Subject] order, see Manfredi (1994).  17  This much establishes that je introduces two nominals, a Subject and a Predicate, as in (97).  (97) DP V  DP  Now consider what happens when ye combines with ti. On independent grounds, I have proposed that ti heads a reduced relative clause. Its appearance with equative je follows automatically: The je introduces a DP complement, which, as before, hosts a rz'-clause. This is illustrated in (98). VP  (98) DP  v  DP  I D  CP  vP ti  Poss'r Kunle  v  Poss'm  0 In order to derive the surface linear order, there are three steps in the movement of the possessum: first, it moves to Spec C P and from there to Spec DP; finally, it moves to Spec VP.  Note that only the possessum but not the possessor can move to Spec, IP in this type of construction.  3.4  Other syntactic contexts for the R-relation: prepositional uses of ti  So far, I have shown two environments where we see reduced relatives, namely with genitive DPs and the je ti construction. In those cases M-tone ti is the head of a reduced relative clause. In this section, I consider yet other contexts where we see the ti element, namely nongenitive environments. I show that in these constructions ti has the status of a preposition. There are three types of ft'-elements that qualify to be analyzed as prepositions: the prepositional preverb //', the post-verbal preposition, lati and the ti that is a marker of locative adjunct extraction. The relation between C and P is discussed by Emonds (1985: 291ff) and Schulte (2001); certain prepositions behave sometimes also behave like complementizers. English for is an example of such a preposition. Compare (99) where for behaves like a true preposition that takes an NP as its complement with (100) where for has the status of a complementizer.  (99)  a.  Bake a cake [for Mary]  b.  I bought this shirt [for Peter and Paul]  c.  The musician performed [for the guests]  (100) a.  They arranged [for a friend to deliver the package]  b.  We bought some books [for Sam to read]  c.  He was anxious [for the tools to be bought.] (Emonds 1985:291)  The structure in (101a) illustrates the prepositional status of for. The structure in (101b) illustrates the complementizer status of for. (101) a.  ...PP  Mary  b.  CP  I analyze Yoruba ti in the following section as an element that functions as a preposition.  3.4.1  lati as a locative P at the right edge of vP  This section establishes that prepositional ti occurs in post-verbal PP in the form of lati' as 9  illustrated in (102) and (103). The examples (102b-c) and (103b-c) show that lati cannot occur pre-verbally.  (102)  a.  Mo de [lati Kanada] ni lsg arrive P Canada P T arrived from Canada in the morning.'  aaro morning  b.  *Mo lsg  [lati P  Kanada] Canada  de ni arrive P  aaro morning  c.  *Mo lsg  ni P  aaro morning  [lati P  a.  lati Eko fo ni aaro Mo fly P Lagos P morning lsg 'I airlifted from Lagos in the morning.'  b.  *Mo lsg  lati P  Eko fo Lagos fly  ni P  c.  *Mo lsg  lati P  Eko ni Lagos P  aaro fo morning fly  Kanada] Canada  de. arrive  aaro morning  It appears lati is derived. It may be possible to decompose it to ni 'locative' and ati 'and' but there is no semantic relationship between these two components and the whole to support such a suggestion. As to this, I treat this element as a compound P. 19  I propose that lati PPs are introduced as complement to v, as in (104). (104)  vP Mo" v.' de  PP P  DP  lati  Kanada  The structure in (104) predicts that the PP cannot be an adjunct to vP. This is why (105a) is bad; as such, we cannot have the structure in (105b) where the PP is adjoined to the vP. (105) a.  *Mo lsg  de Ibadan lati arrive I. from  Kanada Canada  This section has established two points about lati phrase: it occurs post-verbally and it can only be a complement of v.  3.4.2  ti as a locative P at the left edge of the vP  As a preposition, ti also introduces expressions of location (Awobuluyi 1978: 98). One other thing to add to this is the structural position of the ri-phrase relative to the verb: the PP that ti heads precedes the verb. This is shown by the (a) examples in (106)-(109). In the (b) examples of (106)-(109), the ri-phrase immediately follows the verb, and in the (c) examples, the ri-phrase is sentence final; such outputs are ungrammatical. This establishes that the PP headed by locative ti is pre-verbal and must also follow the subject DP.  (106)  (107)  a.  Mo [ti ibe] de ni aaro lsg P place arrive P morning 'I got back from there in the morning.' *Mo lsg  de [ti arrive P  ibe] place  *Mo lsg  de ni arrive P  aaro  [ti P  ibe] place  6 [ti ile] bere isekuse ni aaro 3sg P house start bad-beahviour P morning 'He began behaving badly from home in the morning.'  b.  *6  c.  *o 3sg  (109)  morning  aaro morning  a.  3sg  (108)  ni P  aa.  bere start  isekuse [ti bad-beahviour P  ile] house  ni P  aaro morning  bere start  isekuse ni bad-beahviour P  aaro morning  [ti P  ile] morning  ni P  aaro morning  [ti Kanada Mo Eko] folo si P Canada lsg Lagos fly go P 'I took a morning flight from Lagos to Canada.'  *  b.  *Mo lsg  f6lo si fly go P  Kanada Canada  [ti P  Eko] n i Lagos P  aaro morning  c.  *Mo lsg  folo si fly go P  Kanada Canada  ni P  aaro morning  [ti P  aa.  Ibadan n i [ti Ejigbo] de Olii P E. arrive I. P O. 'Olu arrived in Ibadan via Ejigbo in the morning.'  aaro morning  b.  *01ii O.  de Ibadan [ti arrive I. P  Ejigbo] E.  ni P  aaro morning  c.  *01u O.  de Ibadan n i P arrive I.  aaro morning  [ti P  Ejigbo] E.  Eko] Lagos  I propose that the preverbal r/-phrase is a left adjoined PP as in (110). There is one movement that is involved in the derivation, namely the subject DP moves from Spec vP to Spec IP.  IP  (110) DP  vP 014  pp  vP DP t  ti 'from'  DP  v  DP  de 'arrive'  Ibadan  Ejigbo  One important thing to note is the contrast between post-verbal lati whose PP is treated as a complement to v and the pre-verbal ti whose PP is analyzed as an adjunct.  3.4.3  ti as a marker of adjunct extraction  There is another kind of ti that I want to talk about. This is the one that is found at an adjunct extraction site. Cook (2004) claims that the ri-element found in W H constructions in Yoruba is a mark of locative extraction and proposes that ti be treated as a locative head. The (a) examples in (111)-(112) show constructions that depict location of certain events. When a locative expression is questioned, a locative Wh-word appears in sentence-initial position, and locative ti appears in preverbal position, as in the (b) example of (111-112). The absence of ti in those contexts leads to ungrammaticality, as in the (c) examples of (111-112). (Ill)  (112)  a.  Ade je 6ro ni oja A. eat mango P market 'Ade ate mangoes at market.'  b.  Nibo ni Ade ti joro where F O C A Loc eat-mango 'Where did Ade eat mangoes?'  c.  *Nibo ni  Ade  joro  Ade ta ilu ni oja A. sell drum P market 'Ade sold drums at market.'  Nibo ni Ade ti talu where FOC A Loc see-drum 'Where did Ade sell drums?' c.  *Nfbo ni  Ade  talu  Cook proposes a LocP for the ft'-phrases found in WH-constructions where the Loc takes a uP as its complement, as in (113). She argues that the overt realization of this Loc is in complementary distribution with the OP in Spec LocP. Thus if the Spec is pronounced Loc is null and vice versa. (113)  Yoruba locative P in Wh- extraction LocP .  OP^^^^Loc' Loc  uP  ti  talu (Cook 2004: 34)  To recap, this section has examined the kind of ti that is found in locative expressions claiming that its presence is an indication that extraction has taken place.  3.5  Conclusion  This chapter has discussed two types of ri-constructions, namely the type found in genitives and the one found in non-nominal environments. I have established that ^'-constructions in genitive environments are reduced relative clauses and the M-tone ri'-element was analyzed as a C. I proposed that, as a functional element, ti assigns genitive Case to the possessor when it is the only Case assigner available, but when it co-occurs with the MTS they both coassign genitive Case. I treated this as an instance of Case stacking. I also argued that ticonstructions in non-genitive environments are prepositional phrases, some of which are preverbal while others are post-verbal.  C H A P T E R F O U R : INTERPRETING Y O R U B A B A R E NOUNS  4  Introduction  Yoruba lacks obligatory determiners (contra B a m g b o s e 1967, 1990, 2001; Awobuluyi 1978), and so makes extensive use of bare nouns (BNs). These bare nouns can be construed 1  in at least one of these three ways : generic (la), indefinite (lb), or definite (lc). 2  (1)  a.  Oyinbo gbadun siga. European enjoy cigarette 'Europeans enjoy cigarettes.'  generic  b.  Mo ri aja. lsg see dog 'I saw a dog.'  indefinite  c.  Aja gbo mi. dog bark lsg 'The dog barked at me.'  definite (in discourse context)  English does not permit the occurrence of bare nouns in the context of Yoruba examples in (1) . The parallel examples in (2) illustrate this.  (2)  a. b.  *I saw dog. *Dog barked at me.  For the latter, nouns require overt determiners as shown in (3) . 3  (3)  a. b.  I saw a dog. A dog barked at me.  'The bare nouns discussed in this dissertation focus mainly on count nouns partly because mass nouns are not clearly distinguished from count nouns in Yoruba. See chapter 6 for more discussion. It is worth noting one obvious distinction between an indefinite and a definite noun namely that when an indefinite reading is obtained there is usually a question like which dog? or 'where?' In (lb) for instance, there may arise the need for either of the questions to be asked by the listener. By contrast, when a definite reading is obtained, it does not attract such questions. This is because in the latter case, there is usually a prior knowledge about the object that is being referred to. The example in (i) shows a context where English allows the absence of an overt determiner, namely with socalled'bare plural nouns'. (i) Birds lay eggs. 2  3  As discussed in the literature, Yoruba is not the only language whose nouns lack overt determiners. The examples in (4) also show that Japanese nouns can be bare. (4)  a.  John  -ga hon -Nom book 'John read a book.'  -o yonda -Acc read  b.  John  kai-ta wrote  c.  inu -ga heya-ni haitte-kita dog Nom room-to in came 'The dog entered the room.'  -ga ronbun-o -Nom article 'John wrote an article.'  4  (adapted from Fukui 1995: 105)  This chapter examines the distribution and interpretational variability of bare nouns in Yoruba. There are four basic questions that I address. What are the determinants for interpreting Yoruba bare nouns? (§4.1) What conditions the generic construal of bare nouns? (§4.2) What conditions the indefinite construal of Yoruba bare nouns? (§4.3) What conditions the definite construal of Yoruba bare nouns? (§4.4) The focus of §4.5 is the consequences of this analysis. §4.6 concludes.  4.1  The determinants for interpreting Yoruba bare nouns  Three factors determine how bare nouns are construed in Yoruba: (i) the verb type; (ii) syntactic position; and (iii) discourse linking.  4.1.1  Verb classes  The kind of verbs that a bare noun occurs with plays a crucial role in determining how these nouns are to be construed (Carlson 1977; Dechaine 1993; Chierchia 1995; Deprez 2004  4  The suffix on bare nouns in Japanese is an instance of case marking.  among others). In this section, I examine three types of verbs in relation to the distribution and construal of Yoruba bare nouns: (i) permanent states, (ii) temporary states, and (iii) dynamic events. I conclude that the construal of Yoruba bare nouns is sensitive to these distinctions (cf. Dechaine 1993: 437).  4.1.1.1 Permanent states (P-state)  Permanent states are verbs that express a permanent state of the individual. They do not denote actions that take place, as they involve no activity (van Voorst 1992:81). Permanent state verbs usually refer to a state or condition which is not changing or likely to change over time; they only have atemporal and abstract qualities (Bach 1981: 71; Carlson, 1979). They denote the kinds of states that people or things can be "in" for a long time. This can be a state of love, hate, wisdom, tallness etc. In fact, Dowty (1979: 126) claims that these verbs are somehow simpler or more limited in their interpretations than other kinds of verbs. I f transitive, permanent state verbs encode a relation between an experiencer (subject) and the theme (object).  (5)  a.  M o feran Titi lsg love T. 'Hove Titi.'  b.  Ade kdrira mi A. hate lsg 'Ade hates me.'  transitive P-state  If intransitive, permanent state verbs attribute a property to their subject: (6)  a.  Mo go lsg be-stupid T am stupid.'  b.  Dupe gbon D. be-wise 'Dupe is wise.'  intransitive P-state  Sehun u n ga S. HTS be-tall 'Sehun is tall.'  4.1.1.2 Temporary states (T-state)  I define temporary states as verbs that express behaviours that last for a short time. These verbs are transitory since they express the state that does not last for a long time and they can be transitive, as in (7).  (7)  a.  Titf bfnu T. be-annoy 'Titi was annoyed.'  (bi provoke  b.  Mo -yari lsg refuse T refused.'  (ya ?  d.  Mo pebi sun (pa lsg be-hungry sleep beat 'I went to bed without food.' re it be-tired T am tired.' 6  inu) transitive T-state stomach  ori) head  ebi) hunger  mi lsg  One thing about the transitivity of temporary state verbs is the problem of transparency. Consider, hinu 'be annoy' in (7a), which can be split into bi inu, 'provoke stomach' where the verb is the monosyllabic word bi (Abraham 1958:103) and the object inu, the meaning of each lexical item has little in common with the compound form. In other words, the meaning of the verb cannot be fully realized in isolation. This is why such verbs are said to have an idiomatic use (Oyelaran 1982).  4.1.1.3 Events  I define events as verbs that involve a dynamic situation which changes over time, and whose duration is limited. If transitive, they encode a relation between an agent (subject) and the theme (object). Thus, each of the verbs in (8) can take two arguments: a subject and an object.  (8)  a.  Motiin ri mi M. see lsg 'Motun saw me.'  b.  Okete gbe iho rodent dig hole ' A rodent dug a hole.'  c.  Okete hu ebu rodent uproot yam-seed ' A rodent uprooted a yam-seed.'  d.  Ode e pa ogidan hunter HTS kill leopard 'The hunter killed a leopard.'  e.  Olu je isu O. eat yam 'Olu ate yams.'  transitive event  Intransitive events include but are not limited to, examples in (9).  (9)  a.  Mo rin lsg walk 'I walked.'  b.  Eye e ba bird HTS perch 'The bird perched.'  c.  Olu joko O. sit-down 'Olu sat down.'  intransitive event  d.  Taye dide T. stand-up 'Taye stood up.'  These verbs do not encode any relation since there is usually one argument. These verbs however have one thing in common they depict activities that can only last for a short time. Note also that some intransitive events are monosyllabic (9a-b) while some are disyllabic (9c-d).  4.1.1.4 Distinguishing "events/temporary-states" and "permanent-states" There is one property that both events and temporary-states have in common, which distinguish them from permanent-states, namely they are transitory. The discussion in this section explores this distinction. I briefly examine three ways of distinguishing between "event/temporary-states" from "permanent-states". One is that the distinction between permanent-states and events/temporary-states involves a [±dynamic] contrast. A dynamic verb is one that can be used in the progressive (continuous) aspect, indicating an unfinished action. Going by this definition, event/temporary state verbs are dynamic. In Yoruba, the aspectual particle n meaning progressive is only compatible with event/temporary-state verbs (10). (10)  a.  Mo n je lsg prog eat T am eating yam.'  isu yam  Mo n bfnii lsg prog angry 'I am annoyed.'  n  V-event  n  T-state  O  In contrast, permanent-state verbs cannot take the progressive marker, (11).  (11)  a.  *Mo lsg  n prog  feran isu like yam  b.  *Mo lsg  n prog  go stupid  *S  n  P-state  n  P-state  O  Another way by which event/temporary state and permanent-state verbs is that event/temporary-state verbs can either be construed as past or present perfect, (12). (12)  Mo 1 sg (a) (b)  je isu eat yam 'I ate yams.' 'I have eaten yams.'  past present perfect  When the example in (12) serves as an answer to the question in (13) then, the tense in relation to the time of utterance can be that of simple past or past perfect.  (13)  Nje o je Q-tag 2sg eat 'Did you eat yam?'  isu yam  In contrast, permanent-state verbs in Yoruba usually convey a non-past meaning, (14).  (14)  a  Mo feran ijo lsg like dance 'I like dancing.'  non-past  b.  Mo go lsg stupid 'I am stupid.'  non-past  Lastly, the interpretation of bare nouns is sensitive to the permanent-state and event /temporary-state distinction. The contrast between the interpretation of oldgbd as 'a cat' (15a), and 'cats in general' (15b), is attributable to the difference in the verb it combines with. The verb pa ' k i l l ' in (15a) is eventive whereas korira 'hate' in (15b) is stative.  (15)  a.  Mo pa ologbo lsg kill cat 'I killed a cat'  b.  Mo korira ologbo lsg hate cat 'I hate cats.'  This establishes that the construal of bare nouns is sensitive to the particular verbs they occur with. This is treated in greater detail in §4.2. I turn to another factor that is crucial to the construal of bare nouns, namely the syntactic position of bare nouns. The two positions controlled for are subject versus object.  4.1.2  Subject versus object position  Another crucial factor in understanding how a given bare noun is to be construed is its syntactic position (cf. Rullmann 2003: 3). A t first, it appears Yoruba manifests an unrestricted distribution of bare nouns in argument positions in the sense that they can occur in subject as well as object position.  (16)  a.  b.  5  Aja ri mi dog see lsg ' A dog saw me.'  -  Mo ri aja lsg see dog 'I saw a dog.'  Thus, there is sometimes a symmetrical relation between the subject and object in the sense that they can be interpreted the same way as shown in (16). In some other cases there is an asymmetrical relation. For example, while it is possible to interpret a bare noun as generic when in object position, such interpretation is not available i f the same bare noun is in subject position as in (17). I discuss this in §4.2.  (17)  a.  Aja feran mi dog like lsg *'Dogs like me.'  b.  Mo feran aja lsg like dog 'I like dogs.'  ^he fact reported here agrees with English but contrasts with Romance. In the latter, "bare nouns are confined to complement positions and excluded from pre-verbal subject positions" (Longobardi 2004: 582).  4.1.3  Discourse-linking  The term discourse linking refers to a situation where the construal of a bare noun is supplied by the discourse. I define a discourse span as a group of sentences that focus on a topic (Wolfart 1972). Discourse-linked nominals are nouns that are either new or familiar referents in such a string of sentences. When a noun is mentioned at the beginning of a discourse, it is said to be a new referent. When it comes up in subsequent discussion, it becomes a familiar referent. The impact of discourse context on how Yoruba bare nouns are to be construed is 6  enormous. In Yoruba, as I show later in the chapter, a bare noun can only be construed as definite in a proper discourse context. Since Yoruba lacks overt determiners, it is not easy to identify discourse-linked utterances. In contrast, in English, which has overt determiners, the definite determiner links the noun it modifies to a noun that occurs earlier in the discourse, as illustrated in (18).  (18)  7  Yesterday, a snake entered my room. I was afraid to kill the snake.  In (18), the snake in the second sentence is linked to a snake in the first sentence. The Yoruba parallel example to (18) is given in (19). It is not possible to have a definite construal of ejd 'snake' in (19b) as we do in English example in (18).  (19)  a.  Ni ana ejd 6 w o - inu ile mi. P yesteray snake HTS enter inside house lsg 'Yesterday, a snaked entered into my house.'  b.  Eru u ba mi lati fear HTS catch lsg to #'I was afraid to kill the snake.'  pa kill  ejd snake  For reasons not quite understood, Yoruba equivalent of (18) shown in (19) is odd. Pronoun is preferred instead, (20). For details concerning discourse linking, see Stalnaker (1978); Pesetsky (1987); Enc (1991); Dechaine (1993); Lopez (2000); Doherty (2005); Kiss (2005) among others.  6  (20)  a.  Nf ana ejo 6 wo inu ile mi. P yesterday snake HTS enter inside house lsg 'Yesterday, a snaked entered into my house.'  b.  Eru u ba mi fear HTS catch lsg 'I was afraid to kill it.'  lati to  pa kill  a it  As mentioned above, the reason for this is that Yoruba lacks overt determiners like English 'a/an' and 'the' that can link the first and the second ejo 'snake' in (19). This supports Kiss's (2005: 132) claim that discourse linking is much easier to obtain in a language with overt determiners. The other alternative to the use of pronoun is to mark ejo 'snake' as salient by adding the salient marker naa to it, as in (21), otherwise a proper discourse context must be set.  8  (21)  a.  Ni ana ejo 6 wo inu ile mi. P yesteray snake HTS enter inside house lsg 'Yesterday, a snaked entered into my house.'  b.  Eru u ba mi lati pa fear HTS catch lsg to kill T was afraid to kill the very snake.'  ejo naa snake Sal  The account of how salience is marked in Yoruba is given in chapter 5.  9  However, this is not always the case. For example in 'The sun shines.' or 'The world is round', the noun that is being modified by the determiner the is unique and cannot be linked to a noun that occurs earlier in the discourse. (i) a. I told my friend that a snake and a spider entered my house yesterday. He asked me what I did to them and I said I killed the spider. He went further to ask of the snake then I said (i-b). b. Eru u ba mi lati pa ejo fear HTS catch lsg to kill snake 'I was afraid to kill the snake.' See §4.4 for details on discourse linking. This is not to say that object nouns cannot be construed as definite without being marked for salience. As we shall see, definiteness of objects is obtained in contrastive construction or when the entity is physically present. 7  9  4.2  What conditions the generic construal of bare nouns?  Following Chierchia (1995) and Greenberg (1998), I define generic sentences as those with bare nouns, which express "non-accidental, law-like generalizations". According to Carlson (1977, 1980) and Cohen (2004), the semantics of generic nouns show that they have a "quasi-universal flavor". As I show herein, Yoruba generic bare nouns entail kind referring expressions (cf. Dechaine & Manfredi 1998; Chierchia 1995, 1998; Longobardi 2003 and Miiller 2003: 72). I propose that Yoruba bare nouns can be construed as generic in one of two sets of contexts: in one set of contexts, a bare noun is construed as generic based on verb type. This 10  relates to the distinction between permanent-state and transitory (event/temporary-state). This is discussed as an instance of lexically conditioned genericity (§4.2.1). In the other set of contexts a bare noun can be generic in the presence of a characterizing (habitual) aspect (cf. Delfitto 2002; Longobardi 2002). I call this grammatically conditioned genericity, which in Yoruba requires the presence of imperfective maa-n (§4.2.2).  4.2.1  P(ermanent)-states have a lexical GEN operator  Yoruba bare nouns can be construed as generic if one of two conditions holds, as in (22).  (22)  Lexically conditioned G E N requires permanent-state verbs. Grammatically conditioned G E N requires imperfective maa-n.  Lexically conditioned genericity is when the construal of a bare noun as generic is dependent on the verb type. A lexically conditioned generic construal does not require any special marking as genericity appears to be encoded in the verb (Chierchia 1995: 219). Descriptively, there are three contexts for the lexically conditioned generic construal of Yoruba bare nouns:  '"Chierchia (1995) and Carslon and Pelletier (1995) claim that nouns occurring as subject of Individual Level Predicates (ILP) can be construed as generic in English.  (23)  First generalization about genericity of Yoruba bare nouns (lexically conditioned genericity): A Yoruba bare noun can be construed as generic when it occurs as: (i) the object of a transitive permanent-state, (ii) the subject of an intransitive permanent-state or (iii) the subject of a transitive permanent-state verbs, but only if the object is also a bare noun. In view of the above, I propose that lexically conditioned genericity is encoded in  permanent-state verbs in Yoruba and as such, in the appropriate syntactic position, a bare noun occurring with a permanent-state verb can be construed as generic. However, as observed by Chierchia (1995:202), the idea that certain verbs "are somehow inherently generic cannot be straightforwardly implemented in strict lexicalist terms. They need to be operated on by G E N " . Along this line, I further propose that Yoruba permanent-state verbs are associated with a null generic operator (henceforth GEN) that binds any bare noun in its scope (Dechaine 1993: 495). It is this G E N operator that licenses bare nouns as generic 11  subject to certain other conditions that are discussed below. I adopt the structure in (24) following Chierchia (1995). Observe that G E N is base generated as an adjunct to V P .  (24)  v? argument v  VP GEN  VP V  argument  (adapted from Chierchia 1995: 213)  " Dechaine (1993) claims that operators divide into two classes according to whether they take scope over a predicate or an argument. The relevant part of the construal condition is shown in (i). (i) construal condition For X to be construed with an operator, X must be in the domain of the operator.  Given the proposal that G E N is adjoined to V P , it means that G E N can only scope over object bare nouns.  12  As we shall see, this has consequences for how subject BNs are to be  construed as generic. A n operator scopes over an argument that it c-commands. In the case under discussion, only the object bare noun, and not the subject bare noun, is in c-command relation with G E N : (25)  Absence of scope over subject of permanent-states a. b.  If bare N is OBJ, then G E N If bare N is SUBJ, then *GEN  However, we shall see that the Yoruba data show that it is sometimes possible for G E N to move to a position where it can scope over the subject B N . I propose that G E N in certain syntactic contexts, undergoes Q-raising from where it can scope over both subject and object BNs (cf. Diesing 1992; Szabolcsi 2000).  (26)  XP  argument v  VP t EN G  v  V [P-state]  VP  argument  I split my discussion to three parts. In §4.2.1.1,1 discuss genericity of object bare nouns of transitive permanent-states. In §4.2.1.2, I account for genericity of subject bare nouns of transitive permanent-states where I claim that for a subject bare noun of transitive  The scope of an operator "is the domain within which it has the ability to affect the interpretation of other expressions." (Szabolcsi 2000:607)  permanent-states to be construed as generic, it must obey certain constraints. The focus of §4.2.1.3 is the account of genericity of subject bare nouns of intransitive permanent-states.  4.2.1.1 Bare noun object of transitive P-states When a bare noun functions as an object of a permanent-state verb, it can be construed as generic. (27a) describes a situation when I like dogs in general. Similarly, in (27b), it is the case that Moji hates cats in general.  (27)  a.  Mo feran aja lsg like dog T like dogs.'  object of P-state  b.  Moji korira ologbo M. hate cat 'Moji hates cats.'  object of P-state  By contrast, objects of permanent-state verbs are construed as indefinite when marked for specificity, as in (28), or as definite (in proper discourse contexts), as in (29b). (28)  (29)  13  a.  Mo feran aja kan lsg like dog Spf T like a certain dog.'  indefinite construal of object of P-state  b.  Moji korira ologbo kan M. hate cat Spf 'Moji hates a particular cat.'  a.  Felicitous Context: M y wife is shocked to hear that I like Tunde's dog even though he is not friendly with us. When she asked if what she heard is true; I said (29b).  b.  66to n i , mo feran aja trute copula lsg like dog 'It is true, I like the dog.'  13  definite construal of object of P-state  See chapter 5 for an account of how kan marks indefinite nouns as specific.  Consequently, I propose that an object bare noun of a permanent-state verb is generic when it is in the domain of a lexical generic operator, which licenses it. This is illustrated in (30). By hypothesis, permanent-state verbs are associated with a G E N OP in the lower V P ; this G E N OP c-commands the object position.  (30)  Lexical GEN with permanent-state verbs: BN in O position DP^^^^^vP Taye  v  VP GEN^  ^ V P  V ^ ^ ^ ^ B N [P-state] Vgeneric  ,L feran  ^ aja  This much establishes the claim that when a bare noun functions as an object of a permanent state verb, it can be construed as generic.  4.2.1.2 Bare noun subject of transitive P-states  The focus of this section is what conditions the construal of subject bare nouns as generic. First, consider the examples in (31). Aja 'dog' (31a) and ologbo 'cat' (31b) are construed as generic in the sense that it is the case that the former, i.e., aja 'dogs' generally like bones and the latter, i.e., ologbo 'cat' generally hate akara 'bean-cake'.  (31)  a.  Aja feran egungun dog like bone 'Dogs like bones.'  b.  Ologbo korira akara Cats hate bean-cake 'Cats hate bean-cakes.'  subject oj'transitive F'-state (1)  In contrast to the examples in (31) are those in (32) where the same bare nouns occurring as subject of the same permanent-state verbs cannot be construed as generic. The 14  example in (32a) cannot mean that it is generally the case that dogs like me. Similarly, in (32b) too, it is not generally the case that cats hate Moji.  (32)  a.  Aja feran Taye dog like T. *'Dogs like Taye.'  b.  Ologbo korira Moji cat hate M . *'Cats hate Moji.'  subject of transitive P-state (2)  This raises the question of what brings about this distinction. I first account for the case where both the subject and object are bare nouns. Here we observe object dependency, i.e., the genericity of the subject depends on the genericity of the object. In the two examples in (31), both subject and object are bare nouns. It follows that like object bare nouns, it appears that a subject bare noun can also be generic. Based on this fact, I propose the structure in (33) , where G E N moves to a position where it can scope over both the subject and the object bare nouns.  (33)  Lexical GEN with permanent-state verbs: BN in S & O position vP GEN  vP  Note also that an indefinite construal of these bare nouns requires the presence of the specificity marker while their construal as definite requires a proper discourse context. 14  As the structure in (33) shows, two crucial conditions are needed for subject bare nouns to be construed as generic. First, the object must be a bare noun. Second, there is abstract movement of G E N from its adjoined V P position to a position where it can scope over the subject. We are left with the account of the examples in (32) where subject bare nouns occur with transitive permanent-state verbs whose objects are not bare (i.e., Taye, a lsg object pronoun (32a) and Moji, a personal name, (32b)). I illustrate this in (34). Observe that abstract movement of G E N to a position where it could c-command the subject bare noun should make a generic construal possible, yet this is-precisely the interpretation that is unavailable. (34)  Non-genericity with permanent-state verbs: BN in S position GEN  vP  BN * generic t zx Aja 'dog'  VP  GEN  - V [P-state|  DP / \ Taye  feran 'like' The question that this analysis raises is why is it not possible for any subject bare noun of a permanent state verb to be construed as generic? The genericity of a subject bare noun seems to depend on the genericity of an object noun. This means in order for a subject bare noun of a P-state to be construed as generic, the object must also be bare; this is an instance of object dependency. Even when G E N raises to a position where it can scope over the bare noun subject, it is still not possible for the bare noun to be construed as generic. This establishes another point namely, though the issue of scope is very crucial, it is not the only determining factor. As we shall see, for examples such as (32), a generic construal is possible only if imperfective maa-n is present (see §4.2.3). I conclude that genericity is made available to bare nouns in both subject and object positions but with a constraint on those in subject position. Thus, generic construal of Yoruba bare nouns is both structurally constrained and lexically determined.  r 4.2.1.3 Bare noun subject of intransitive Permanent-states  A bare noun of an intransitive permanent-state is construed as generic. In (35a), it is the case that elephants are generally considered big animals. Similarly, tortoises are generally 15  considered wise, (35b).  (35)  a.  Erin tobi elephant be-big 'Elephants are big.'  subject of intransitive P-state  b.  Ijapa gbon tortoise be-wise 'Tortoises are wise.'  subject of intransitive P-state  The question is how do we account for the genericity of a subject bare noun of an intransitive permanent-state? The account is as follows. I propose that intransitive permanent-states are unaccusative (Chomsky 1991; Lasnik 2000; MacDonald 2004). As such, the internal argument receives its generic construal in the (underlying) object position and raises to the (surface) subject position for Case. This claim is in line with the observation that objects of permanent-states can be generic. In the current analysis, the only way to explain why a subject bare noun of an intransitive permanent-state is generic is to assume that it starts out as an internal argument, as in (36).  In Yoruba mythology, elephants are held in high esteem. They are described as splendid such as contained in this orikl'praise song'. (i) a. Kfku erin a da bf ile o GER-die elephant it be like house emph 'When an elephant lies down dead it has the size of a house.' 15  b.  A-j-ku  erin  a  da  bf  oke  Nom-Neg-die elephant it be like mountain 'When an elephant is alive, it has the size of a mountain.'  o...  emph  (36)  IP  I  VP GEN  VP V  ARG  gbon  This implies that intransitive P-state verbs are unaccusative and that their object must raise in the overt syntax to receive Case.  4.2.2  16  T(emporary)-states require a grammatical GEN operator  This section accounts for the grammatically conditioned generic construal of bare nouns. I claim that this occurs with transitory verbs (temporary-states and events). I start my discussion with the second generalization about genericity of Yoruba bare nouns:  (37)  Second generalization of genericity of Yoruba BNs (grammatically conditioned genericity) A Yoruba B N can be construed as generic if: it is an argument of a Temporary)-state verb marked by imperfective maa-n  17  Note that the corresponding permanent state predicates in Italian are not unaccusative (Cinque 1990). However, the sentences corresponding to (35) are also bad in Italian with bare nouns. For details see the cited work. Barczak (2004) treats maa-n as grammatical particles that can jointly mark a habitual sentence. In this analysis, I treat them as imperfective. There remains the question of whether compositionality plays any role in the fusion of the two particles for the purpose of marking the imperfective. The specific questions that one may ask are the following: (a) how is a bare noun interpreted when each of the particles occurs by itself? (b) Does [maa-n] ever occur in a sentence, which is imperfective without being habitual or vice versa? To answer the first question, consider the following sentences, (i) a. Aja maa ri mi. 'A dog will see me.' b. Aja n_ ri mi. 'A dog is seeing me/Dogs see me.' c. Aja maa feran mi. 'A dog will like me.' d. Aja n feran mi. '*A dog is liking me/#Dogs like me.' The examples in (i- a & c) show that when maa occurs by itself, a bare noun can only be interpreted as indefinite and the sentence can only reflect future tense. On the other hand, when n occurs by itself, the interpretation of a bare noun depends on the kind of verb. With an event verb, a bare noun can be interpreted as 16  17  The discussion here focuses on three different contexts where bare nouns can occur with temporary-states. In §4.2.2.1, I discuss the genericity of bare noun objects of transitive Tstates. In §4.2.2.2, I discuss the genericity of bare noun subjects of transitive T-states. Finally, in §4.2.2.3,1 discuss the genericity of bare noun subjects of intransitive T-states.  4.2.2.1 Bare noun object of transitive T-state  Observe that Yoruba transitive temporary-state verbs translate into English as intransitive adjectives, which are accompanied by copula verbs. The example in (38a) is an occurrence of aja 'dog' in object position without the imperfective and it can only be interpreted as indefinite. However, with the presence of imperfective maa-n in (38b), the interpretation of ajd 'dog' is generic, thus it is the case that in circumstances, dogs are always hungry.  (38)  a.  Ebi pa aja hunger-HTS seizes dog ' A dog slept without food.'  sun sleep  b.  Ebi maa-n pa aja sun hunger-IMP seizes dog sleep 'Dogs go to sleep without food.'  object of transitive T-state  4.2.2.2 Bare noun subject of transitive Temporary-state The discussion in this section is about the occurrence of bare noun subjects of transitive Tstate verbs. Again, such bare nouns continue to be non-generic without the presence of imperfective maa-n. This is shown by the contrast between (39a), where imperfective is absence and (39b) where it is present. Only in the latter is the generic construal possible.  indefinite when the action is progressive or generic. But with a state verb (i- d), the indefinite reading is not obtained because the sentence cannot be interpreted as progressive. Similarly, it,is not felicitous to interpret the bare noun as generic. As shown in all the examples so far, we do not have the problem of interpretation of bare nouns as generic with the combination of the two particles. This explains why I decide to treat maa-n as imperfective in this dissertation.  (39)  a.  Aja pa ebi sun dog seizes hunger sleep ' A dog slept without food.'  b.  Aja maa-n pa ebi sun dog IMP seizes hunger sleep 'Dogs sleep without food.'  subject of transitive T-state  Similarly, when the bare noun ekute 'mouse' occurs with the existential verb wa 'be' in (40), it only shows a temporary presence of ekute 'mouse' in that particular location, and not that mice are always in the house. Again, only when imperfective maa-n appears as in (40b), is the genericity of this bare noun is obtained.  (40)  a.  Ekute mouse (i) (ii)  wa nihil ile subject of transitive T-state be inside house *'Mice are in the house.' (*generic) ='There is a mouse in the house/There are mice in the house.'  b.  Ekute maa-n wa ninu ile mouse IMP be inside house 'Mice are found in the house.'  4.2.2.3 On the absence of intransitive T-state in Yoruba  This section shows that there are no intransitive temporary state verbs in Yoruba. This is a lexical gap when Yoruba is compared to other languages. In order to establish this claim I compare Yoruba with English. The examples in (41) are all adjectival intransitive verbs. As such, they do not take object nouns.  (41)  a.  John is tired.  b.  John is angry.  c.  John is happy.  d.  John is sad.  However, the Yoruba parallel examples in (42) show that these verbs are transitive and therefore require object nouns.  18  Also worthy of note is the fact that the subject can be an  expletive pronoun as in (42a).  (42).  a.  O re Tunde it sick T. 'Tunde is tired.'  b.  Tunde bi T. provoke 'Tunde was angry.'  c.  Tunde ni ayo T. have joy 'Tunde is happy.'  d.  Tunde ba inu je T. spoil stomach 'Tunde is sad.'  inu stomach  19  In conclusion, the data presented in §4.2.2.1-§4.2.2.3 establish the point that all temporarystate verbs in Yoruba are transitive. I now turn to a presentation of the data of event verbs.  4.2.3  Events require a grammatical GEN operator  I will show here that like temporary-states, event verbs also require a grammatical G E N operator in order a for bare noun to be construed as generic: (43)  Third generalization of genericity of Yoruba BNs (grammatically conditioned genericity) A Yoruba B N can be construed as generic if: it is an argument of an event verb marked by imperfective maa-n;  The following three subsections look at three different contexts where bare nouns can occur with eventive verbs. §4.2.2 discusses the genericity of bare noun objects of transitive events. 18  In most cases, the object forms part of their meaning.  §4.2.2.2 discusses the genericity of bare noun subjects of transitive events. Finally, in §4.2.2.3,1 discuss the genericity of bare noun subjects of intransitive events.  4.2.3.1 Bare noun object of transitive event  When a bare noun occurs as the object of a transitive event verb, it cannot be construed as y  generic except i f imperfective maa-n is present. In (44a) where egungun 'bone' is an object, it is not the case that generally, I eat bones. Once imperfective maa-n is introduced in (44b), this bare noun is interpretable as generic: each time I eat, I eat bones. (44)  a.  Mo lsg (i) (ii)  je egungun eat bone * T ate bones.' =T ate a bone.'  b.  Mo maa-n je lsg I M P eat 'I eat bones.'  object of transitive event  egungun bone  4.2.3.2 Bare noun subject of transitive event The same can be said of bare noun subjects of transitive events. (45a) shows aja 'dog' as the subject of the verb ri 'see'; it is construed as indefinite. On the other hand, in (45b), where imperfective maa-n is introduced, it can be that each time I jump over the fence of a building a dog will see me. (45)  19  a.  Aja dog (i) (ii)  ri mi see lsg *'Dogs see me.' =A dog saw me.'  b.  Aja maa-n ri dog IMP see 'Dogs see me.'  subject of transitive event  mi lsg  The verb baje 'spoil' is one of the splitting verbs whose object occurs in between them (Awobuluyi 1978).  Lastly, in (46), we see aja 'dog' and egungun 'bone' as subject and object of ri 'see' respectively. None of them has a generic construal. With the introduction of imperfective maa-n, as in (46b), a generic construal of these bare nouns is obtained. Thus, in (46b) dogs in general eat any kind of bone.  (46)  a.  Aja dog (i) (ii)  je egungun eat bone *'Dogs eat bones.' ='A dog ate a bone.'  b.  Aja maa-n je dog IMP eat 'Dogs eat bones.'  subject & object of event  egungun bone  4.2.3.3 Bare noun subject of intransitive event  This section briefly examines cases relating to the occurrence of bare nouns subjects of intransitive event verbs and establishes that like other transitory verbs, these verbs require the presence of imperfective maa-n before such bare nouns can be construed as generic. The example in (47a) contrasts with (47b) in one respect the absence of imperfective in (47a) and its presence in (47b). This contrast in turn brings about a difference in interpretation. Thus, (47a) can only mean that there is a dog that is asleep whereas (47b) indicates that it is the case that dogs sleep under certain conditions.  (47)  a.  Aja sun dog sleep 'A dog is asleep.'  b.  Aja maa-n sun dog IMP sleep 'Dogs sleep.'  subject of intransitive event  In summary, I have examined different contexts where grammatically conditioned genericity is realized. Contrasts are made between contexts that permit generic construal with contexts that do not. The conclusion that I reach is that genericity of bare nouns in all of these  contexts is possible with T-states and events only if imperfective maa-n is introduced. What remains to be resolved is how to capture this generalization. Based on this fact, I propose that " this imperfective marker carries a quantificational feature that forces the presence of G E N in its local environment, as suggested by Chierchia (1995: 202).  4.2.4  Analysis of grammatically conditioned genericity  The analysis of grammatically conditioned genericity of bare nouns that I propose draws largely from the work of Kamp and Reyle (1993: 569); Chierchia (1995) and Kiss (1998). In their aspectual operator-based analysis, Kamp and Reyle treat "progressive" and "perfect" as aspectual operators (Asp OP). According to them, these operators transform the meaning of the underlying non-progressive or non-perfect verb, verb phrase or sentence into that of its progressive or perfect counterpart respectively. Adapting their operator-based analysis, I propose that imperfectivemaa-n is associated with a vP-external aspectual operator, as in (48). This makes available the generic construal of bare nouns in either subject or object position. (48)  Grammatically conditioned genericity with T-state/event AspP  BN  Vgeneric V  BN ^. Vgeneric [T-state/event]  Specifically, I propose that imperfective maa-n is associated with a G E N operator, which is introduced in Spec AspP. The location of the imperfective above V P is supported by the fact that in Yoruba, the position of Aux is higher than V P (Dechaine 1993; Oyelaran 1989; Barczak 2004). The structures in (49-51), which are representative samples of the data  presented above, illustrate how the combination of imperfective marker with G E N derives the generic construals of bare nouns in all argument positions with event verbs.  (49)  Grammatically conditioned genericity with events: BN in S position AspP  v  >  VP  Vgeneric V Z\ Aja  (50)  | [event]  DP Z\ mi  Grammatically conditioned genericity with events: BN in O position AspP GEN Asp maa-n  vP v  VP DP^ Z\ V Mo. [event] | |  BN  p. Vgeneric ^ egungun  (51)  Grammatically conditioned genericity with events: BN in S &0 AspP GEN uP  Asp maa-n  VP  v  BN Vgeneric V Aja  BN Vgeneric  [event] egungun  jt  The introduction of the imperfective aspectual marker maa-n in conjunction with the G E N operator accounts for the generic construal of bare nouns with transitory verb constructions. I extend this analysis to cases involving permanent state verbs. Finally, note that in the surface syntax, the subject N P precedes the imperfective. Once the subject bare noun receives its generic reading, it moves to Spec IP where it receives nominative case from I(nfl). IP  (52) BN Vgeneric Aja  AspP  I GEN  vP  Asp maa-n  VP  v l  BN  DP [event] ri  mi  4.2.5  Extending grammatically conditioned genericity to Permanent State verbs  Recall that a bare noun cannot be construed as generic when it occurs as the subject of a transitive permanent-state verb whose object is not bare. Here, the presence of imperfective maa-n is required for a generic construal. In (53a), it is the case that generally anywhere I go, and dogs see me, they always play with me. In (54a), it is the case that for certain behaviour of Ade, generally, teachers hate him. In the (b) examples in (53)-(54) where the imperfective marker is absent, a generic construal is not possible.' (53)  (54)  a.  Aja maa-n feran mi dog IMP like me 'Dogs like me.'  b.  Aja dog (i) (ii)  a.  Oluko maa-n korira Ade teacher IMP hate A . 'Teachers hate Ade.'  s  b.  feran mi like lsg *'Dogs like me.' ='A dog likes me.'  Oluko korira Ade teacher hate A . (i) ^'Teachers hate Ade.' (ii) = ' A teacher hates Ade.'  The structure in (55) represents all the cases discussed above since there is no constraint other than the presence of imperfective marker and the positioning of G E N in a place where it can scope over these bare nouns.  (55)  Grammatically conditioned genericity with P-state AspP GEN Asp maa-n  vP VP  v  BN Vgeneric V ' [P-state]  BN Vgeneric  Note that in the surface syntax, the subject NP precedes imperfective. By hypothesis, this is because the subject moves to Spec IP where it receives nominative Case from I(nfl). IP  (56)  BN Vgeneric I Oluko  AspP GEN Asp maa-n  vP v  VP  DP [P-state]  Ade  korira The last environment to consider is where bare nouns are construed as generic without the presence of the imperfective. Predictably, this environment optionally permits the presence of imperfective maa-n to mark genericity. In (57a), the bare noun ajd 'dog' occurs as the object of the permanent-state verb feran 'like' with an optional imperfective maa-n. With or without the imperfective marker, the bare noun is construed as generic. Similarly in (57b) where both subject and object bare nouns of the same verb occur with the imperfective marker, which is optional, both bare nouns are construed as generic.  (57)  a.  Mo (maa-n) lsg IMP 'I like dogs.'  feran aja like dog  object of state  Aja (maa-n) dog IMP 'Dogs like bones.'  feran like  subject and object of state  egungun bone  I illustrate this with the structure in (58). The optionality of the imperfective marker is reflected by the parentheses.  (58)  IP BN  Vgeneric Aja  I  AspP GEN Asp (maa-n)  vP VP  v l  BN  v'  BN  Vgeneric [P-state] egungun feran I summarize the findings of genericity of bare nouns in Yoruba as (59).  (59)  generic construal of Yoruba bare nouns. PERMANENT STATE V E R B S O s S&O  Bare noun Imperfective: maa-n  \  (V)  TEMPORARY-STATE & EVENT VERBS S&O S o  X  X  V  V  X  -  9  X  *.V •  In conclusion, I have provided a unified,analysis of generic construal of bare nouns claiming that any bare noun can be construed as generic i f grammatically conditioned. There  remains the question of whether the account of Yoruba genericity can be applied to other languages.  4.2.6  Contrast between English and Yoruba genericity: evidence for default aspect  Extending the analysis of Yoruba bare nouns to English, I propose that imperfective maa-n always introduces a G E N operator. While Yoruba has an overt imperfective, English has a covert imperfective. First, observe that English bare nouns (i.e. nouns without overt determiners) can be construed as generic with permanent-state verbs, (60a). The same interpretive effect holds of Yoruba, (60b). (60)  20  a.  Dogs like bones GENx GENy [dog(x), bone(y), like(x.y)  P-state  b.  Aja feran egungun dog like bone 'Dogs like bones.'  P-state  However, with events, Yoruba contrasts with English in how bare nouns are construed. While English bare nouns can be construed as generic, Yoruba bare nouns cannot be.  (61)  a.  Dogs eat bones GENx GENy [dog(x), bone(y), like(x,y)  Event  b.  Aja dog (i) (ii)  Event  je egungun eat bone *'Dogs eat bones.' =A dog ate a bone.'  Generic construal of a bare plural noun (subject and object) is possible with both permanentstate and event verbs in English. In Yoruba, only permanent-state verbs permit generic bare nouns.  Further, in English, when the object of an event verb is not a bare noun, genericity still holds: (i) Dogs like me. 2 0  (62)  Contrast between Yoruba and English generic construals Yoruba Event/T-state & G E N X P-State & G E N V  English V V  The generalization that emerges when English is compared to Yoruba is the following: in English generic construal is available with events/temporary whereas in Yoruba, it is not. The question that arises is why should this be so. First, recall that Yoruba obtains generic readings with event verbs via grammatically conditioned genericity. (63)  Aja dog  maa-n je IMP eat  . egungun bone  generic  In the corresponding English example in (64), a generic reading is possible without any overt marking of imperfectivity. (64)  Dogs eat bones GENx GENy [dog(x), bone(y), like(x,y)  generic  I would like to suggest that the imperfective is the source of genericity with event verbs in both languages. Note however, that in the English examples there is no overt imperfective marker present. The question that arises is how then do we harmonize the facts of Yoruba (which has an overt imperfective) and the facts of English (where there is no overt imperfective). Observe that in Yoruba, the default aspect/tense value is the perfective/past, (65). In English, the default aspect/tense value is the imperfective/present, (66). (65)  (66)  j i m o je isu a. Jimo ate yam b. Jimo has eaten yam c. *Jimo is eating yam d. *Jimo eats yams  PAST (Present) PERFECTIVE *PROG *IMP  Jimmy eats yam a. habitual and imperfective b. historical present 21  21  This is in the context of story telling when a series of successive events is encoded.  The generalization that emerges is the following. The unmarked verb form in English is the Imperfective whereas the unmarked verb form in Yoruba is the Perfective. Assuming that the generic construal is closely linked to imperfective, this means that the generic construal is conditioned by the imperfective in both Yoruba and English. The imperfective in Yoruba is overtly realized through the use of maa-n, whereas in English the imperfective is covert.  (67)  a. b.  Jimo [ John [  IMP  IMP  maa-n] je isu 0 ] eats yam  Yoruba genericity English genericity  What remains for further study is whether or not this claim holds in other languages.  4.3  What conditions the indefinite construal of Yoruba bare nouns  I have shown that there are two ways that a bare noun can be construed as generic, via lexically conditioned genericity or via grammatically conditioned genericity. In this section, I account for the indefinite construals of these bare nouns.  4.3.1  Defining indefiniteness  My interpretation of Yoruba bare nouns as indefinites follows Heim (1982), Enc (1991) and Matthewson (1998). According to Enc/s Familiarity and Heim's Novelty Conditions: A l l indefinites in a sentence must be novel, in the sense that they must introduce into the domain of discourse referents that were not previously in the discourse. A l l definites must be familiar, in the sense that the discourse referents they are mapped onto must have been previously introduced into the discourse. In other words, indefinites cannot have antecedents in the discourse, whereas, definites must. (Enc 1991:7) We can infer from Enc/s idea that a bare noun is indefinite if it is novel at the current stage.of the conversation (cf. Heim's 1982, 1988, 2002 Novelty Condition). In this case, a bare noun is considered indefinite if it refers to a new discourse referent since it has not been previously  mentioned. The notion of novel is illustrated in (68). In particular, it will be wrong to have (68b) at the beginning of a conversation.  (68)  Novel context a. I met [a man] today. b. *I met [the man] today.  (Matthewson 1998: 32)  Replicating this in Yoruba, we see that the bare noun oblnrin 'woman' occurs in both indefinite and definite contexts: (69)  4.3.2  a.  Nf ojii ala mi mo In face dream lsg lsg 'In my dream, I saw a woman.'  ri saw  obinrin. woman  b.  Obinrin ti mo n je iya a woman C lsg saw C O P mother G E N 'The woman that I saw was my mother.'  mi lsg  Generic and indefinite construals are in complementary distribution  There are two generalizations that hold for the interpretation of bare nouns as indefinite or generic in Yoruba:  (70)  a. b.  Whenever indefinite is possible, generic is illicit Whenever indefinite is illicit, generic is required  I already gave examples that illustrate the contexts in which the generic construal of bare nouns is possible. What follows is the data that show contexts where indefinite construal of bare nouns is obtained. The examples in (71) show that when a bare noun occurs either as the subject of a transitive event verb (71a), or as the subject of an intransitive event verb (71b), it can be construed as indefinite. (71)  a.-  Aja dog = (i) *(ii)  ri mi see me ' A dog saw me.' 'Dogs see me.'  subject of transitive event  b.  Aja dog = (i) *(ii)  sun sleep 'A dog is asleep.' 'Dogs are asleep.'  subject of intransitive event  Similarly, a bare noun in the object position of a transitive event verb (72a) is also obligatorily construed as indefinite. Further, (72b) shows that when bare nouns occur as both the subject and the object of a transitive verb, they are also obligatorily construed as indefinite. (72)  a.  Mo lsg = (i) "(ii)  ri aja see dog 'I saw a dog.' 'I see dogs.'  object of transitive event  b.  Aja dog  ri egungun see bone 'A dog saw a bone.' 'Dogs see bones.'  subject & object of transitive event  = 0) "(ii)  On the other hand, the examples in (73) and (74) illustrate the occurrence of bare nouns with state verbs. First, when a bare noun occurs in the subject position of a transitive permanentstate (73a), or the subject of an intransitive temporary-state (73b), it can be construed as indefinite.  (73)  a.  Aja dog = (i) (ii)  feran mi like lsg 'A dog likes me.' 'Dogs like me.'  Subject of transitive P-state  Subject of intransitive T-state Erin in bfnii elephant HTS be-angry = (i) ' A n elephant is angry.' (ii) 'Elephants are (always) angry.' By contrast, in (74a) where a bare noun occurs in the object position of a transitive permanent-state an indefinite construal is not possible. are bare nouns, an indefinite is not possible.  22  And when both subject and object  (74)  a.  = (ii)  feran like 'I like 'I like  Aja dog *(i) = (ii)  feran egungun like bone ' A dog likes a bone.' 'Dogs like bones.'  Mo lsg  *0) b.  aja dog a dog. dogs.'  Object of transitive P-state  Subject & object of transitive P-state  Note that the examples in (74), which do not have indefinite construal are cases where bare nouns are obligatorily construed as generic. Thus, generic and indefinite construals are in complementary distribution.  4.3.3  Elsewhere case: indefinite bare nouns are bound by existential operator  I argue that the construal of bare nouns as indefinite is a last resort i.e. whenever the generic construal is not available then the indefinite construal is obtained as a last resort. In this case, such bare nouns are treated as existentials. The theory of existentiality that I adopt dates back to Heim (1982) and subsequent works. This theory claims that an indefinite noun may be existentially bound. Thus in (75), 'a dog' is considered indefinite in the semantic realm.  (75)  a. b.  John sees a dog 3x dog(x) A see (J, x)  Heim's idea is that a sentence such as shown in (76a) is interpretable only if the indefinite noun is existentially bound (76b). This is what is referred to as existential closure.  See chapter 5 for the account of how indefinite construal of (74) can be obtained. There I argue that an object bare noun of a state verb has to be marked for specificity.  2 2  = existential closure dog(x) (adapted from Heim 1982) With this background, I propose that in the absence of the abstract lexical G E N operator, a default existential operator (3) is introduced. I claim that the indefiniteness of bare nouns in Yoruba induces the presence of an existential operator (3). Bare nouns in Yoruba can therefore be analyzed as containing (free) variables, assuming that the operator binding bare nouns (variables) saturates the expression (cf. Heim 1982, 1988; Borer 2004: 8). s  Linking this proposal with the earlier proposal for genericity predicts that 3 will be in  complementary distribution with the generic construal. This in fact is the case. In (77), where the bare noun aja 'dog' occurs as the subject of n'see', an event verb, the generic construal is not available.  (77) Default existential with event verbs: BN in S position 3  vP  BN vlndef Aja 'dog'  v  VP V [event] | ri 'see'  DP y\ mi 'me'  The case of an object bare noun with a event verb is shown in (78). A generic construal is ruled out because the lexical G E N is not compatible with temporary-state and event verbs; the existential operator is introduced as the last resort. Consequently, the object B N noun is interpreted as indefinite.  (78) Default existential with event verbs: BN in O position 3 DP z \ Mo T  VP  v V [event]  BN vlndef Zx aja 'dog'  ri 'see'  The last case illustrated in (79) involves the occurrence of bare nouns as both the subject and the object of event verbs. They continue to be construed as indefinite because of the same reason: event verbs are not compatible with the lexical G E N operator. (79) Default existential with event verbs: BN in S & O position vP  3  BN Indef Zx Mo  'Aja'  VP  v V [event] gbe 'pick'  BN Indef Zx egungun 'bone'  On the other hand, whenever a generic construal is required, an indefinite construal is not available. This fact is illustrated in (80), where an indefinite reading is unavailable for the bare nouns since the lexically conditioned generic operator is present. In particular, the introduction of an existential operator is not an option since lexical G E N operator is obligatory.  (80)  vP GEN  vP BN VP  *Indef /"/generic Aja 'dog'  V [P-state] feran 'like'  BN *Indef /-/generic egungun 'bone'  The summary of these findings is given in (81). In particular, note the complementarity between indefinite and generic construals of bare nouns.  (81)  complementarity of indefinite and generic construal of Yoruba bare nouns P E R M A N E N T STATE VERBS o S S&O  Indef Bare noun  V  G E N Bare noun  4.4  J  TEMPORARY STATE & EVENT VERBS o s&o S  V  V  V  X  X  X  \  V  V  What conditions the definite construal of Yoruba bare nouns  This section is concerned with identifying the factors that determine whether bare nouns can be construed as definite. In §4.4.1,1 define the term "definiteness". §4.4.2 discusses contexts where a definite construal is not possible. In §4.4.3,1 show that a definite construal of bare nouns arises from Discourse linking.  4.4.1 Defining Definiteness  In fashioning out a definition of "definiteness", I combine Heim's (1982, 1988) and Matthewson's (1998) definitions. The issue of a definite noun being familiar is expressed in  the Familiarity Condition. According to Heim and Matthewson, a noun is definite if it is familiar at the current stage of the conversation. In order to obtain a definite reading the discourse referent that the noun is mapped onto must have been introduced earlier into the discourse. The examples in (82) are used to illustrate this concept. r (82)  a. b.  I bought [a car] yesterday. [The car] is a Toyota brand.  23  Context =0 Context =82a  'Car' in (82a) is unfamiliar since the hearer does not know about it prior to the time it was first mentioned. If (82b) follows (82a), then it means the hearer knows about the car already, giving rise to a definite construal. So the difference in the construal of 'car' as definite or indefinite hinges on the principle of familiarity. Further, there is an additional term that links familiarity and discourse together, namely the notion of common ground (Stalnaker 1974, Strawson 1952 and Matthewson 1998). A definite noun is familiar to the common ground of the discourse, which usually holds between the speaker arid the hearer.  4.4.2  24  Definite construal is unavailable in out-of-the-blue contexts  With the right verb, a Yoruba bare noun can be construed as indefinite in out-of-the-blue contexts. However, in out-of-the-blue contexts, no bare noun can be construed as definite. (83)  a.  Mo lsg = (i) *(ii)  ri see 'I saw T saw  aja dog a dog.' the dog.'  b.  Aja ri mi dog see lsg = (i) A dog saw me.' * (ii) 'The dog saw me.'  See Matthewson (1998: 32) for exceptions to the claim that definite descriptions must always be familiar to the common ground of the discourse.  2 3  This much establishes the desirability of discourse context for a definite construal of Yoriiba bare nouns.  4.4.3  Definite construal arises from Discourse linking  In appropriate discourse contexts, any bare noun can be construed as definite. The examples in (84)-(86) reflect definite bare nouns occurring with event verbs. Each of the boldfaced bare nouns is linked to a discourse context. The mention of aja 'dog' when in subject position (84c), or in object position (85c), shows that it is a familiar referent having being mentioned in the discourse contexts, (84a) and (85a) respectively. As such, it is construed as definite. However, in out of the blue contexts, it will be infelicitous to say either of (84c) or (85c).  (84)  (85)  BN a.  V O (BN in S position) Felicitous Context: M y wife asked if the security guard saw me as I was jumping over the fence to pluck some mangoes in Kunle's orchard to which I answered no. She then asked about his dog and I replied with (84c).  b.  Infelicitous Context: I walked into my house and said (84c) to my wife.  c.  Aja ri dog see 'The dog saw me.'  S  V  a.  Felicitous Context: M y wife asked if I saw my friend at the park yesterday to which I answered no. She then asked about my friend's dog and I replied with (85c).  b.  Infelicitous Context: I walked into my house and said (85c) to my wife.  c.  M o ri aja lsg see dog T saw the dog.'  BN  mi lsg  (BN in O position)  For more on definitions of definiteness see Enc (1991); Borer (2004) and the references therein.  Similarly, aja 'dog' and egungun 'bone' in (86c) are properly linked to the discourse context in (86a). This is why they are both interpreted as definite. On the other hand saying (86c) in out-of-the-blue context is infelicitous.  (86)  BN a.  V BN (BN in S and O position) Felicitous Context: M y brother's dog was not with him when he was eating, so he dropped a piece of bone at one corner of the room for the dog and went to the kitchen to wash his hands. By the time he came back, the bone had disappeared. Even though he did not see the dog, he said (86c) to his friend.  b.  Infelicitous Context: I walk into my house and say (86c) to my friend:  c.  Aja je egungun dog eat bone' 'The dog ate the bone.'  The next set of examples in (87)-(88) shows bare nouns occurring with permanent state verbs. Again, the B N in the (c) examples can be construed as definite in an appropriate context, as in the (a) examples. But in out-of-the-blue contexts, a definite construal of these bare nouns is not felicitous.  (87)  (88)  BN a.  V O (BN in S position) Felicitous Context: M y wife wants to know why Taye always gives Tunde's dog. In reply I say (87c).  b.  Infelicitous Context: I just entered the room and said (87c) to my brother.  c.  Aja feran Taye dog like T. 'The dog likes Taye.'  S a.  V BN (BN in O position) Felicitous Context: Taye has a dog and a cat. Each time he feeds them he spends more time with the dog than the cat so his friend wants to know why and the father says (26c).  b.  Infelicitous Context: I just walked in to my wife and said (88c).  c.  Taye feran aja T. like dog 'Taye likes the dog.'  (89)  BN a.  V BN (BN in S and O position) Felicitous Context: M y dog has just finished eating the bones that I served it. It stretched its legs, wagged its tail and rubbed its body against mine. I said (89c) to my friend who was watching with interest:  b.  Infelicitous Context: Without any bone around, I said (89c) to my friend who was playing with my dog.  c.  Aja gbadun egungun dog enjoy bone 'The dog enjoyed the bones.'  The generalization that emerges for the definite construal of a bare noun is straightforward: Yoruba bare nouns can be interpreted as definite i f they are discourselinked. On the other hand, in an out-of-the-blue context, no bare noun is construed as definite in Yoruba. These findings are summarized as (90). (90) Definite construal of Yoruba bare nouns T E M P O R A R Y STATE & EVENT VERBS Obj s S&O  PERMANENT STATE VERBS Obj S&O S  Out-of-the-blue  X  X  X  X  X  X  Discourse-linked  V  V  V  V  V  V  Finally, the table in (91) brings to light the three construals of bare nouns discussed in the preceding sections.  (91) GENERIC, INDEFINITE AND DEFINITE construals of Yoruba bare nouns PERN 1ANENT STAT!E V E R B S O S S&O GENERIC  V  V  X  INDEFINITE  X  X  V  DEFINITE  V  V  V  TEMPORARY STATE & EVENT VERBS 0 s S&O x--  V V  In what follows, I present some consequences of my analysis.  V V  V V  4.5  Consequences of the analysis  Yoruba bare nouns can be construed in one of three ways: generic, indefinite or definite. What remains to be addressed are the consequences of my analysis. In §4.5.1, I discuss the structural ambiguity of bare nouns, while §4.5.2 looks at how the B N analysis plays out in Yoruba genitive constructions.  4.5.1 Bare nouns are structurally ambiguous  Observe that in some languages whose bare nouns have been analyzed, the interpretation of bare nouns is structurally constrained (Lpngobardi 1994, Dechaine & Manfredi 1998). This section looks at the internal structure of Yoruba bare nouns. Based on the facts established in the preceding sections, I claim that there are two structural representations for all bare nouns in the language. On the one hand are bare nouns that do not require contexts for their interpretation. These are the bare nouns that are interpreted as generic or existential. On the other hand are the bare nouns whose interpretation is contextually determined. These are the bare nouns that are interpreted as definite. In principle, there are two possible syntactic structures for bare nouns; they are either analyzed as NPs or as DPs (with a null D). This raises the question of which of the two structures we should adopt for Yoruba. There are three hypotheses:  (92)  a. b.  H I . Bare nouns are always "bare NPs" H 2. Bare nouns are always "DPs"  c.  H 3. Bare nouns are structurally ambiguous between NP and DP.  I argue in the following section that Yoruba supports the third hypothesis.  4.5.1.1 Bare nouns can be NP or DP  I propose that Yoruba bare nouns that are construed as definite have a (null) discourse-linked determiner, in which case they may have the structure of a DP, as in (93). On this view, not  all bare nouns are bare in structure at the abstract level (cf. Cheng & Sybesma's 1999: 518 discussion of Chinese bare nouns).  (93)  Structure for definite construal of bare N DP D ^ ^ ^ ^ N P 1 0  ^ aja  On the other hand, I propose that both indefinite and generic bare nouns, which have been shown to be in complementary distribution, have an NP structure (94).  (94)  25  Structure for indefinite/generic construal of bare N NP aja  There seems to be no motivation for claiming that indefinite bare nouns have a different structure (DP) from generic bare nouns (NP) in the current analysis. A s I have shown generic and existential construals of bare nouns in Yoruba are in complementary distribution. As such, they should have the same syntactic structure.  4.5.1.2 Critique of previous analysis (Ajiboye 2001)  Ajiboye (2001) claims that Yoruba bare nouns with (in)definite construals are DPs, (95); whereas bare Ns with generic construals are NPs, (94) (cf. Dechaine & Manfredi 1998). One major difference between this previous proposal and the current proposal Ajiboye (2001) is the issue of government.  This is contrary to Miiller (2003: 72) where it is claimed that 'genericity may be expressed by the Determiner Phrase (DP)'. 25  (95)  DP  (indefinite  (96)  NP  generic  N  (Ajiboye 2001:9)  Ajiboye (2001) further distinguishes definite and indefinite DPs, following Longobardi (1994), who claimed that null D is indefinite (i.e. existential) but only i f governed. In that analysis, null D was possible only if DP was governed by the verb.  (97) V  ^ _ — S t r u c t u r e DP  for indefinite construal of bare N  N  (Ajiboye 2001: 7)  So, for the structure of indefinite bare nouns in object position, the D P is base generated with the N P occurring as the complement of the null D. On the other hand, a subject DP is said to be ungoverned, so subject null D is (correctly) predicted to not be indefinite in Yoruba. On what accounts for the definite construal of bare noun subjects, following Longobardi (1994), Ajiboye (2001) proposed that bare Ns are interpreted as definite if NP moves to Spec DP (98). One thing to note is that this movement is not detectable, as D is always null. (98)  DP NP^^  structure for definite construal of bare N  ^ D ' D ^ 0  ^t  N P  The table in (99) below summarizes the findings in (Ajiboye 2001).  (99)  summary of bare noun construals in Yoruba (Ajiboye 2001)  construal of bare noun Subject * Indefinite V Definite Generic V  Object V V V  Structure DP with governed null D DP with movement of NP to Spec, DP NP  The analysis is faced with the open question: what permits movement of NP to Spec, DP for definite bare nouns? Why should both definite and indefinite bare nouns have one structure and yet allow generic bare nouns to have a different structure? In the present account, I have shown that both generic and indefinite bare nouns pattern the same way. In light of this I have proposed that they should have the same structure. On the other hand, since definite bare nouns are context sensitive, their interpretation is not sensitive to position. A l l that is required is that they be discourse-linked. Thus, there is the need to modify the proposal put forward in Ajiboye (2001).  4.5.2  Implication for the analysis of genitive constructions  In chapter 2, it was established that the D P inside the genitive construction has an overt D in the form of the MTS.  (100) Iwofun ni [oko o Tona] optional FOC vehicle MTS T. 'Traveling in Tona's vehicle is optional.'  definite  The present chapter proposed that bare nouns that are definite have the structure of a DP with a null D. Bringing these two ideas together predicts that Yoruba genitive DPs should be construed as definite since it has a D position that is occupied by the M T S , which is sometimes null. However, in certain context, a genitive construction may be construed as indefinite. One context is where the speaker has knowledge of the referent i.e., but it has to be marked  for specificity. This is the situation where by the speaker knows that Tona, a transporter has more than one vehicle and he only saw one of such vehicles, (101). (101) M o ri [oko o 'Tona kan] lsg see vehicle MTS T. certain T saw a certain vehicle of Tona's.'  26  indefinite  The point that is being established here is that the proposal that definite bare nouns have the structure of a D P has support from genitive constructions whose D has no independent morpheme and they are also interpreted as definite.  4.6  Conclusion  I have established that a bare noun that is analyzed as an N P can be construed as generic i f lexically conditioned with permanent state verbs or grammatically conditioned through the use of the imperfective maa-n in transitory verbs. Otherwise it is existential. On the other hand, a Yoruba bare noun that has the structure of a DP (in an appropriate discourse context) is construed as definite. This is summarized in (102).  (102) Summary: Construals of bare nouns in Yoruba Structure interpretation conditioned by: surface form i) lexical G E N operator (i.e. P-state Generic 'N U N ] (GEN) verb) ii) grammatical G E N operator (i.e. imperfective maa-n) (i.e. Event/T-state verb) elsewhere condition: N Indefinite U N ] if B N * G E N , then BN= 3 (3) definite contextually conditioned N U D NP] (Def) 0  I extend this proposal to English and make the generalization that a language can either make use of the imperfective to mark genericity overtly (as in Yoruba) or covertly (as in English). In chapter 5, I account for another dimension of the interpretation of nominal expressions, namely the conditions under which they are construed as specific or salient. 2 6  For account of specificity, see chapter 5.  C H A P T E R F I V E : SPECIFICITY A N D S A L I E N C E IN Y O R U B A  5  Introduction  The claim of chapter 4 was that when a bare noun is interpreted as generic or indefinite, it must have the structure of an NP, (1). On the other hand, when a bare noun is interpreted as definite, it must have the structure of a DP, (2). (1)  (2)  a.  GEN/3 =  [NP]  b.  Mo korira [aja] lsg hate dog 'I hate dogs.'  generic  c.  Mo ri [aja] lsg see dog 'I saw a dog.' ' ,  indefinite  a.  DEF =  b.  (in appropriate discourse) Mo ri [aja] lsg see dog 'I saw the dog.'  •  [D NP] 0  definite  The goal of this chapter is to discuss how specificity and salience are realized in Yoruba nominal expressions. First, when a noun takes kan (3a), it has the interpretation of 'a certain N ' , whereas if the same noun takes ndd, (3b) it can mean 'the particular N ' .  (3)  a.  Mo ri aja kan lsg see dog certain 'I saw a certain dog.'  specificity  b.  Mo ri aja naa lsg see dog very 'I saw the very dog.'  salience  I argue that while indefinite NPs are morphologically marked for.specificity with kan, 1  definite DPs are morphologically marked for salience with naa I demonstrate that a nominal expression is interpreted as specific if it refers to an entity the speaker has some knowledge about (cf. Aboh 2004: 77). In (3a), the dog must be known to the speaker whereas (3b) indicates that the dog in question must have not only been mentioned before, it must be known to both the speaker and the hearer. I claim that kan marks NPs as specific and as such must be a D, (4).  (4)  a. b.  [DpNPitD Aon] I,] DP  Recall from chapter 4, the null D marks bare nouns as definite. This is a contrast between empty D and overt D in Yoruba: a null D marks bare nouns as definite whereas an overt D marks bare nouns as specific. Further, I propose that naa as a marker of salience is a modifier and it is to be analyzed as an adjunct to DP with the structure in (5).  (5)  a. b.  [  NP[D0]t]«a a ;  D P  DP SAL  The chapter is organized as follows. §5.1 accounts for specificity in Yoruba. §5.2 discusses salience in Yoruba. The focus of §5.3 is the syntax of kan and naa while in §5.4,1 give an  'See Hellan (1981); Fodor and Sag (1982); En? (1991); Ludlow and Neale (1991); Ioup (1997); and von Heusinger 2002).  account of number interpretation of nominal expressions when combined with nda. §5.5 concludes.  5.1  Specificity in Yoruba: kan  This section surveys the way specificity is marked in Yoruba and three other unrelated languages, namely English, Turkish and Gungbe. I find out that there are different ways by which specificity is marked in these languages. I show that while Yoruba uses the element kan, English optionally marks specificity with adjectives, Turkish does so with accusative and genitive Case markers, and Gungbe with c[e. First, I start by defining the term "specificity".  5.1.1  2  Defining specificity  One important distinction at the discourse level relates to the speaker's knowledge of the referent. This is often discussed in terms of specificity, where [+specific] corresponds to a referent known to the speaker and [-specific] corresponds to a referent that is not known to the speaker (Ludlow and Neale 1991; Haspelmath 1997:108-109). Recall that the [±definite] distinction is based on the discourse status of the referent (new/familiar). These two properties interact as follows. First an unfamiliar /new discourse referent, i.e. an indefinite NP, may be known or unknown to the speaker: an indefinite N P may be [+specific] or [specific]. In contrast, a familiar discourse referent i.e., a definite DP is necessarily known to the speaker: a definite D P must be [+specific]. Thus an indefinite N P can be non-specific or specific, while a definite DP must be specific. The table in (6) summarizes this.  The notion of specificity as a key concept in the semantics of reference is not novel in the literature (Fodor 1971; Enc 1991; Ludlow and Neale 1991; Hasplemath 1997; Mathewson 1998; Ginnakidou 1998; von Heusinger 2002, Aboh 2004 among others).  (6)  specificity versus (in)definiteness (cf. Enc 1991) indefinite definite -i-def, +spf -def, -spf -def, +spf  The table in (6) divides nouns into three categories. One category consists of nouns that are indefinite and at the same time non-specific. Another category consists of nouns that are indefinite but they are specific. The third category consists of nouns that are definite and by implication must be specific. With this in mind, I consider how specificity is marked in Yoruba and three other languages.  5.1.2  Marking specificity  This section investigates ways by which nouns are marked for specificity in Yoruba, English, Turkish and Gungbe and establishes that the four languages can all overtly mark indefinite nouns for specificity.  5.1.2.1 Yoruba  In Yoruba, an indefinite bare noun that is new and unfamiliar by definition can only be interpreted as non-specific. In (7c), aga 'chair' is indefinite because Jibola is looking for just any chair to sit on; he has no particular chair in mind.  (7)  a.  Background: When Jibola is tired, he sits down on whatever chair is at hand.  b.  A a r e m u Jibola fatigue catch J. Jibola is tired  c.  O n w a aga ti y66 jokoo le 3sg Prog look chair C Fut sit on 'He was looking for a chair to sit on.'  indefinite non-specific  d.  6 ri awon aga ni igun yara 3sg see P L chair Loc corner room 'He saw some chairs at the corner of his room.'  e.  6 si jokoo lori okan ninu 3sg then sit on one among awon aga ti 6 ri PL chair C 3sg see 'He then sat on one of the chairs that he saw.'  indefinite non-specific  indefinite non-specific  In (8c), the first, aga 'chair' is specific because there is a particular chair Jibola looks for whenever he is tired and unless he finds that chair, he doesn't sit down. And in (8d), aga 'chair' has the status of a definite noun.  (8)  a.  Background: When Jibola is tired, there is a particular chair in his house that he sits on.  b.  Aare m u jibola fatigue catch J. Jibola is tired  c.  O n wa [aga kan] 3sg Prog search chair spec ti 6 maa n jokoo le C 3sg HAB Prog sit on 'He is looking for a certain chair that he sits on.'  d.  6 3sg ni Loc 'He  e.  6 si jokoo le 3sg then sit on 'He then sits on it.'  indefinite specific  ri [aga ti 6 n wa ] definite specific see chair C 3sg Prog search igun yara corner room saw the chair he is looking for at the corner of the room.' e 3sg  Because Jibola has a particular chair in mind, the chair needs to be marked for specificity. In Yoruba, to make such nouns specific requires the presence of kan , (8c). Going by the 3  The word kan is ambiguous between the numeral 'one' and the specific marker 'certain', (i) Mo ra ile kan lsg buy house one  examples in (7) where Jibola is looking for just any chair to sit on and (8) where he has a particular chair in mind, with the latter requiring the element kan, one can conclude that this element functions as a specificity marker. Observe that the bare N in (8d) is construed as definite specific because it is the specific chair that he is looking for that he found. This means that while indefinite bare nouns are marked for specificity, Yoruba definite bare nouns are not overtly marked for specificity.  5.1.2.2 English  In English, an indefinite NP can be interpreted as specific or non-specific. The examples in (9) show that indefinite NPs may be specific without being morphologically marked.  (9)  a. b.  I want a book. I would like to buy a coat.  specific/non-specific specific/non-specific  In English, it is also possible for adjectives like 'certain', 'particular', and 'specific' to mark indefinite NPs as specific.  91  In this case, specificity is overtly marked (Enc 1991, von  Heusinger 2002). ='I bought one house.' = 'I bought a certain house.' The numeral 'one' and the specificity marker are different morphemes underlyingly. The numeral has an initial vowel /o-kan/, which deletes when it functions as a modifier, whereas the specificity marker is consonant initial, kan. In (ii-a) the initial vowel of the numeral one is deleted. This is not unusual in cases where two vowels are juxtaposed across word boundary (Bamgbose 1967, Oyelaran 1971 among others). In (ii-b), gkan 'one' appears in sentence-initial position, in which case the initial vowel is retained. (ii) a. Mo ra ile kan lsg buy house one T bought one house.' b. Okan ni mo ra. one FOC lsg buy 'It is one (X) that I bought.' In (iii-a), kan is the specific marker when linked with (iii-b). (iii) a. Mo ri ile kan ni Ibadan lsg see house certain P T saw a certain house in Ibadan.' b. lie naa ga house very tall 'The very house is tall.' As observed by Enc, there are certain restrictions on how adjectives of specificity work for English. For example, there is a limitation on distribution. See Enc (1991: 4) for details.  91  (10)  a. b.  I want a certain book. I like to watch a particular movie,  specific specific  Comparing English to Yoruba, there is one striking similarity and difference between the two: marking specificity on indefinite NPs is not optional in Yoruba as it is in English. On the other hand, both behave the same way with respect to marking definite nouns as specific they require no overt marking.  5.1.2.3 Turkish  According to Enc (1991) and von Heusinger (2002), Turkish marks specificity with overt Case morphology. In this language, both object and subject nouns can be so marked for specificity. First, let us look at how specificity is marked on object NPs. A n N P that lacks accusative case is interpreted as non-specific (11a & 12a) whereas an object that bears the accusative morpheme such as -yu and -i, in ( l i b ) and (12b) respectively can only be interpreted as specific. (11)  (12)  Ali bir piyano Ali one piano ' A l i wants to rent a piano.'  b.  Ali bir piyano-yu kiralamak Ali one piano-Ace to rent ' A l i wants to rent a certain piano.'  a.  (ben) bir kitap oku-du-m I a book read-past-lsg 'I read a book.'  non-specific indefinite  (ben) bir kitab-i I a book-acc T read a certain book.'  specific indefinite  The suffixes -yu, ( l i b ) and  kiralamak to rent  istiyor non-specific indefinite wants  a.  oku-du-m read-past-lsg  istiyor specific indefinite wants (En? 1991:4-5)  (von Heusinger 2002: 255)  (12b) perform a dual role of marking accusative Case and  specificity on the object N P they attach to. The distinction between a specific and nonspecific N P is easily identified: if [+ACC], then [+specific].  As reported in von Heusinger (2002), the same contrast exists for subject NPs. Specificity is marked on the subject that bears a genitive Case. "An embedded subject with the genitive-case suffix is interpreted as specific and one without the genitive-case suffix as non-specific (p.256)." In (13b), the genitive case -un marks haydut 'robber'as specific.  (13)  a  [kby-ii haydut bas-ti-in]-i non-specific indefinite village-acc robber raid-Nom-poss.3sg]-acc duy-du-m. hear-Past-lsg 'I heard that robbers raided the village.  b  [koy-U bir haydut-un bas-ti-in]-i specific indefinite village-acc a robber-gen raid-Nom-poss.3sg]-acc duy-du-m. hear-Past-lsg 'I heard that a certain robber raided the village, (von Heusinger 2002: 256)  The contrast between Yoruba and Turkish is that while specificity is always marked by the same morpheme kan in Yoruba, this is not so in Turkish. The marking of specificity in Turkish depends on its syntactic position: in object position, Accusative -yu l-i marks specificity, in subject position, Genitive -un marks specificity.  5.1.2.4 Gungbe  Gungbe is part of the Gbe group of languages (a sub-group of the K w a family) spoken in parts of Benin and Togo in West Africa (Aboh 1999). As reported in Aboh (2004:76-77), when a noun occurs by itself as in (14a), it is construed as non-specific. But when a noun takes die as in (14b), that noun is construed as specific. (14)  a  KDku K. emi 3sg 'Koku  m5n tavo, ce bo dp see-perf table lsg-Poss and say-perf na XD tavo Fut buy table saw my table and then said he would buy a table.'  b.  K5ku K. emi 3sg 'Koku (Aboh  m5n tavo ce bo dp see-perf table lsg-Poss and say-perf na x5 tavo dje Fut buy table spf . saw my table and then said he would buy a specific table.' 2004: 77) t def|  In Gungbe, just as indefinite DPs are overtly marked for specificity, definite DPs are also marked for specificity. In (15a), Q[e is the specific marker for tavo 'table' to make it specific (i.e. an indefinite specific table). In (15b), when tavo 'table' is accompanied by 75 it is construed as definite and specific.  (15)  a.i  b.  Koku K. emi 3sg 'Koku  i r o n tavo ce bo dp specific indefinite see-perf table lsg-Poss and say-perf na XD [tavo c|,e] Fut buy table spf . saw my table and then said he would buy a specific table.'  K5ku K. emi 3sg 'Koku (Aboh  iron tavo see-perf table na x5 [tavo Fut buy table saw my table and then 2004: 76)  t def]  ce bo dp specific definite lsg-Poss and say-perf 15] spf said he would buy that specific table.' l+def|  Yoruba and Gungbe show similarity in the sense that they both have dedicated specificity morphemes, which follow the nouns they so mark. They differ in one respect: while Yoruba covertly marks definite nouns as specific, Gungbe does so with an overt morpheme. In (16) is the summary of how indefinite NPs are marked for specificity in these languages discussed here. While both Yoruba and Gungbe obligatorily mark nouns for specificity with dedicated words: kan and q\e respectively, Turkish indirectly marks specificity through Case. English on the other hand optionally marks specificity with a range of lexical items (usually drawn from the adjective class).  (16)  a. Marking specificity on indefinite NPs in Yoruba, English, Turkish, and Gungbe Yoruba English Turkish Gungbe BN non-specific indefinite BN aN BN specific indefinite a (certain) N N kan N-CASE N-cie a (specific) N a (particular) N  On a final note, observe that Gungbe goes a step further than other languages under comparison in the sense that it overtly marks definite DPs for specificity w i t h i j . The standard assumption in the literature is that definite DPs are inherently [+specific] and need not be morphologically marked for specificity. This is consistent with the data from Yoruba and English. The Gungbe data indicate that there is a need for further research on the status of ID to determine whether it does something else rather than being a specific marker as reported in Aboh (2004). I now turn to the question of how to test for specificity. In what follows, I carry out two of such tests: the Speaker knowledge test (§5.1.3) and the Subset relation test (§5.1.4).  5.1.3  Testing for specificity I: Speaker knowledge test (Ludlow and Neale 1991)  The focus in this section is the Speaker knowledge test of Ludlow and Neale (1991). Their test for specificity relies on whether or not the speaker has knowledge of the referent. First, I start with non-knowledge of a referent. I illustrate this in (17).  (17)  a.  Mo n lo lona oko lsg Prog go P-path farm 'I was walking along a path to my farm.'  b.  Mo ri oya lsg see grass-cutter 'I saw a grass-cutter. '  indefinite non-specific  5  In (17b), oya 'grass-cutter' is construed as indefinite non-specifics object i.e., it is a referent that is not known to speaker at this point in the discourse.  What happens when the speaker has knowledge of referent? To answer this question, I break my discussion into two parts. First, I present the situation where the speaker has someone in mind. This is the case with an indefinite specific object/person. Recall that such new referent is marked with kan, (18b). (18)  a.  Mo n lo lona oko lsg Prog go P-path farm I was walking along a path to my farm  b.  Mo ri oya kan lsg see grass-cutter 'I saw a certain grass-cutter.  indefinite specific  In (18b), oya 'grass-cutter is construed as an indefinite specific object since it is a referent that the speaker has in mind at this point in the discourse. The next case to consider is where the referent is familiar to the speaker. This is the situation where the referent has to be a definite specific object. The first time that oya is mentioned (19a), it is a new referent in the discourse and it is unfamiliar to the speaker. However, the mention of oya the second time in (19b) makes it has become a familiar referent to the speaker. This is what makes it as definite and specific. Observe that in Yoruba, a definite specific noun is not overtly marked for specificity.  (19)  5  a.  TakiiteOlu pa [oya] trap O. kill grass-cutter 'Olu's trap caught a grass-cutter.'  b.  O l u gbe [oya] sinu apo O. put grass-cutter inside bag 'Olu put the grass-cutter in the bag.'  c.  6 di ile it results home 'He left for home.'  This is a kind of animal of the rodent family.  definite specific  To summarize, I have shown the correlation that exists between specificity and speaker's knowledge of the referent: when the speaker has no knowledge, the referent is non-specific and when he does, the referent is specific.  5.1.4  Testing for specificity II: the subset relation (Enc, 1991) <  The second test that I consider in my analysis of Yoruba specificity is from Enc (1991). Enc, claims that specificity involves a weak link that of being a subset of or standing in some recoverable relation to a familiar object. In Enc/s analysis, if there are two sentences such that the NP in the second sentence is specific, the specific NP picks out a member of the previously introduced NP in the first sentence. On the other hand, a non-specific NP picks out a separate entity, not already under discussion. In that case, such an N P is in a disjoint relation to the first referent. I adopt this subset relation test first to Yoruba data and extend this to English and Turkish.  5.1.4.1 Yoruba  A Yoruba indefinite non-specific NP behaves as predicted by the Enc, test in the sense that it is usually in a disjoint relation with a previously mentioned NP. The only available interpretation of awgn gmgbinrin 'girls' in (21) in relation to awgn gmg 'children' in (20) is that of disjointness as there is no way to link the non-specific gmgbinrin 'girl' to awgn gmg 'children'.  (20)  [ A w o n omo] n sere ninu ogba PL child prog play inside garden 'Children are playing in the garden.'  (21)  [Awon omobinrin] gbadun ara-won PL girl enjoy themselves 'Some girls enjoyed themselves.'  non-specific indefinite  The disjoint relation is illustrated in (22).  (22)  Yoruba Disjoint relation omo  (^^^binrin^)  However, a subset relation can also hold between two NPs in one of two ways. First is where the second NP is marked for specificity and is a subset of a definite NP. The examples in (23) and the illustration in (24) show this type. Here gmgbinrin 'a girl' is marked for specificity by kan: gmgbinrin kan 'a certain girl' is interpreted as a subset of gmg ileiwe 'school children'.  (23)  a.  c.  e.  Context: Some students asked Sehun to give them some water. After they had gone, the following dialogue ensued between him and his mother. Mother:  A w o n ta lo bu o m i fun PL W H Foc-2sg take water for 'Who are those people you served with water?'  Sehun.  awon omo ileiwe PL child school 'the school children'  Mother:  Se o mo won Q-tag 2sg know 3pl 'Do you know them?'  Sehun  Mo mo omobirain kan indefinite specific lsg know girl Spf 'I know a certain girl (among them).'  This subset relation is shown in the diagram in (24).  (24)  Yoruba subset relation: specific indefinite  definite specific  The other case is where the second NP is a definite specific and is a subset of an indefinite non-specific. (25)  a.  Olu ni eran osin O. have animal domestic 'Olu has some domestic animals'  indefinite non-specific  b.  N i akoko ounje in time food When it is time to feed them  c.  [Adiye] ni 6 maa n koko toju definite specific chicken Foe 3sg H A B prog first take-care ' He feeds the chickens first.'  In (25c), adiye 'chicken' is one of the domestic animals that has been previously mentioned, hence it is definite and qualifies to be a subset of the other domestic animals. I illustrate this with the subset diagram in (26).  (26)  Yorubd subset relation: specific definite  I present the same test for English in the next section and show that English parallels Yoruba.  5.1.4.2 English  The examples that I present here are an adaptation of Enc's (1991) data. Consider the pair of sentences in (27) the first of which contains the N P "several children" (27a), and the second NP containing "some girls" (27b).  (27)  a.  [Several children] were playing in the yard,  b.  [Some girls] enjoyed themselves.  Enc's analysis predicts that girls in (27b) can only be specific if it is a member of children in (27a). If it is non-specific, then it means girls is not a subset of children. The latter is the case i.e., the NP girls is in disjoint relation to the NP children. This reading is shown in (28). (28)  Disjoint relation in English some girls  J  (several children  I turn to the case of an indefinite specific, which shows a subset relation. (29)  A lot of children are expected in the hall. Sentence 1: Meaning 1 Meaning 2  Many children came in. I know a girl. A girl is part of the set of children A girl is distinct from the set of children  = specific indefinite = non-specific indefinite  The subset relation holds, if girl is part of the set of children earlier mentioned. In this case, the Meaning 1 is assumed, that of a specific indefinite. This is illustrated in the diagram in (30) . (30)  Subset relation in English "childreT "girT  It is also possible for an NP to be overtly marked for specificity:  (31)  a. b. c.  A lot of children are being expected in the hall. Many children came in. I know a certain girl.  In (31) a certain girl is part of the set of children, if it a specific indefinite. I show the relevant subset relation in (32). (32) a certain girl children  To conclude, in English, the subset relation holds between two NPs if one of them is (overtly) marked for specificity.  5.1.4.3 Turkish  Recall that in Turkish, the signal for specificity of an object N P is the presence of the accusative Case marker. In this language, non-specificity also implies disjointness. I illustrate this in the following examples. (33)  a.  b.  Odam-a birka? ?ocuk girdi my-room-Dat several child entered 'Several children entered my room.'  non-specific indefinite  Iki k i z taniyordum two girl I-knew 'I knew two girls.'  non-specific indefinite  (Enc 1991: 6)  (En? 1991: 6)  Iki kiz 'two girls' in (33b) is expected to show a disjoint relation with birkag gocuk 'several children' in (33a) since it is not marked for accusative. I show the disjoint relation between the two NPs in (34).  (34)  Disjoint relation in Turkish ?ocuk  However, when an NP takes an Accusative Case, it is interpreted as specific. Thus, iki k±z-i 'two girls' in (35b) is a subset of birkag gocuk 'several children' in (35) since it is marked for the accusative Case.  (35)  a.  b.  Odam-a birkag ?ocuk girdi my-room-Dat several child entered 'Several children entered my room.' Iki k i z - i  (En? 1991: 6)  taniyordum  two girl-Ace I-knew 'I knew two girls.'  (En? 1991: 5)  The subset relation is illustrated in (36). (36)  Subset relation in Turkish  In conclusion, I have shown that though languages may use different mechanisms to mark specificity, they converge with respect to these two tests: speaker knowledge and the subset relation. A NP that a speaker has knowledge of is always specific. Similarly, a NP that is a subset of another NP previously mentioned in the discourse is necessarily specific.  5.2  Salience in Yoruba: nda  In this section, I demonstrate that the element nda marks nouns as salient in one of three ways: by picking out a unique referent (37a), by identifying a referent with a previously mentioned referent, (37b) or by adding a referent to a previously established set (37c).  (37)  a.  b.  c.  Aja naa ni 6 le gbo dog S A L FOC 3sg able bark 'It is the O N L Y dog that can bark at me.'  mi lsg =  Uniqueness  Mo ri aja naa lsg see dog S A L 'I saw the V E R Y dog.'  =  Identity  Aja naa ri mi dog S A L see lsg 'The dog A L S O saw me.'  =  Additive  Some linguists have claimed that nda is a determiner (Bamgbose 1967, 2000; Awobuliiyi 1978) while some linguists describe it as post-nominal modifier (Dechaine 1993: 84). In what follows, I am going to show that the function and distribution of nda leads one to conclude that it is not a determiner but a marker of salience.  5.2.1  Defining Salience  Simply put, saliency implies being significant and striking. A noun is made salient when it is particularly important to the speaker and he wants to draw the audience's attention to its striking feature. As pointed out above, salience marking with nda performs three functions in Yoruba: the "uniqueness" function, the "additive" function, and the "identity function. I define each of these terms in the following subsections.  5.2.1.1 Uniqueness  There are two different factors that people consider when they talk about definiteness: uniqueness and familiarity. The Yoruba salience marker nda, like definites, may also pick out a unique referent. A noun phrase that encodes uniqueness implies that the individual it refers to (say 'the dog') is distinct from all other referents in the context, it's a 'one of a kind' referent. For example, in a context where I have both a dog and a cat as pets, if I say, "I'm going to go feed the cat", I mean the unique, 'one of a kind' cat. There are no other cats in the context. This meaning shows up most clearly in English when the determiner gets focused - " T H E cat". Levinson (2005: 1) claims that the referent of a noun phrase is unique or uniquely identifiable if there is only one most salient candidate for being the referent. In order words, the entity is distinguished from other members of the set as unique in some way. In this case, A is unique since it is the only referent, which can satisfy the description. 6  This is schematized in (38). (38)  Uniqueness function = only N A satisfies a description that the other discourse referents: B , C, D don't satisfy.  ©  B  C  D  Another view is contained in Heim (1988) where there is some sense in which an element that denotes uniqueness relates to specificity.  6  5.2.1.2 Additivity  In addition to having a uniqueness function, the Yoruba salience marker nda also has an additive function. This is schematized in (39), when A is added to the set of relevant 7  discourse referents. (39)  Additive function = also N A satisfies a description that the other discourse referents B C D previously satisfied.  B  C  D  CD  B  C  D  5.2.1.3 Identity  A third function associated with the Yoruba salience marker nda is the identity function. Thus, a noun marked by nda is understood to be identical to a previously mentioned discourse referent. This is schematized in (40) where A ' is identical to A . (40)  Identity function = the very N A ' satisfies the same description as A i.e. A ' is identical to A A...  A'.  In what follows, I illustrate how salience is marked in Yoruba and English. I go into details by showing how Yoruba nda performs the three different functions of uniqueness, additivity and identity. I also show that while Yoruba marks salience with nda, English has no dedicated morphology for this function.  5.2.2 Marking salience in Yoruba  I establish here that the presence of nda in Yoruba nominal expressions marks salience rather than definiteness. I start by showing that the misconception of interpreting naa as a definite determiner can be attributed to the fact that naa is often translated as the in English (Manfredi 1992). Consider (41) where the (a) example has a bare noun with an indefinite 8  construal while in the (b) example the same noun marked with naa has a definite construal.  (41)  a.  Mo ra ile lsg buy house 'I bought a house.'  b.  lie naa nf yara meje house naa have room seven 'The house has seven rooms.'  In (42b), the intuition of most native speakers of Yoruba is that naa is the same as English the? But naa does not strictly speaking mark definiteness. To see this, consider the examples in (42). In (42a), the first mention of gya indicates that it is a new referent-that is neither familiar to the speaker nor the hearer. In this case, its interpretation is indefinite. In (42b) the mention of gya again shows that it is now a familiar referent that both to the speaker and the hearer and here the bare noun is interpreted as definite. Finally in (42c), where the noun occurs with ndd the interpretation that gya 'grass-cutter' receives is that of identity with the previously mentioned gya 'grass-cutter'. According to Cohen (2004) additives are special types of focus-sensitive particles. . Manfredi (1992) claims that Yoruba morphemes usually glossed as determiners, and (mis)translated as English definite articles, are both syntactically and semantically more like demonstrative adverbs or deictic nouns. (i) a. Iwe e (naa) Ayo book of at that very (yen) b. Iwe e Ayo book of that-one c. Iwe e Ayo (yii) book of this-one (Manfredi 1992: 207) During the course of writing this chapter I had the privilege of speaking with native speakers of Yoruba who are not linguists telling them how I got stuck interpreting nouns that take the element ndd as in ile nda. To my surprise, the first response I got from one of them is this: 'that is being specific'! This response is a confirmation that many speakers of Yoruba have the intuition that nda marks nouns for definiteness/specificity. But from what I show in this section, we see that nda does not also quite fit into the description of specificity as it is used in the literature, rather nda marks nouns for "salience" in some way. In particular, I show that this notion of salience overlaps with specificity, it is not reducible to specificity. 7  9  (42)  a.  Takute O l u m u [oya] trap P N hold grass-cutter 'Olu's trap caught a grass-cutter.'  indefinite  b.  Olu gbe [oya] lole PN carry grass-cutter go-house 'Olu carried the grass-cutter home.'  definite  c.  [Oya naa] ni w o n fi jeyan definite salience grass-cutter S A L Foe 3pl use eat-pounded yam 'The very grass cutter served as meat with which they ate pounded yam.'  More generally, with nominal expressions that are unambiguously definite (e.g. proper names and nominal expressions that have demonstratives), I show that nda may mark uniqueness, additivity or identity. Moreover, the availability of these construals is sensitive to the syntactic status of the argument as stated in (43).  (43)  The distribution of nda (i) nda marks uniqueness or identity with focused arguments (ii) nda marks additivity or identity with non-focused arguments  The next two subsections show how these distinctions are realized.  5.2.2.1 Nda marks uniqueness or identity with focused arguments Like most K w a languages such as Gungbe, Fongbe, Gengbe (Aboh 2004), focus constructions in Yoruba are "utterances that involve a preconstructed domain...it can also be presupposed due to specific situational knowledge or even to general, may be culturally determined, background knowledge" (Bisang and Sonaiya 2000). They require leftward movement of the focused entity to sentence initial position and such entity is marked by the particle ni (Yusuf 1995). Both nouns and proper names that are marked by nda can be focused irrespective of whether they are singular or plural. When this happens, the focused arguments are construed as unique ' O N L Y the X ' reading or as standing in an identity relation with a previously discourse referent 'the V E R Y X ' reading. This is illustrated in (44) and (45) respectively.  (44)  (45)  Singular NP in Focus position a.  Context for uniqueness interpretation Ojo has a goat, a sheep and a tiger. One day, the tiger killed Ojo's guard. Ojo's son, who for the first time got to know how dangerous tigers are, ran to his father and said (44b). In response, the father said (44c).  b.  Ekun ni 6 pa ode tiger Foe 3sg kill hunter ' It is the tiger that killed the guard!'  c.  Ekun naa ni 6 le pa tiger S A L Foe 3sg able kill 'ONLY the tiger can kill the guard.'  a.  Context for identity interpretation It is a known fact to Kunle's father that there is one particular tiger in the Ibadan zoo where Kunle works that is very dangerous. This morning there was a report that the tiger killed a guard. When Kunle's father asked if it is the tiger that killed the guard. Kunle replied with (45b).  b.  Ekun naa ni 6 pa ode. tiger S A L Foe 3sg kill guard ' It is the VERY tiger that killed the guard.'  ode. guard  I extend the searchlight to coordinated nouns. Again, we see that focused arguments marked with nda are construed either as unique (46), or as standing in an identity relation (47).  (46)  (47)  Plural NP in Focus position a. Context for uniqueness interpretation Imagine that the following animals are paired together in a hunting expedition: (i) aja 'dog' and ologbo 'cat' (ii) agutan 'sheep' and ewure 'goat'. A t the end of the day, group (i) killed some game while group (ii) killed nothing. Determining the potentiality of which group can hunt, Speaker A rendered (46b). b.  [Aja ati ologbo naa] ni 6 le dog and cat S A L Foe 3sg able ' It is ONLY the dog and the cat that can hunt.'  sode. do-hunting uniqueness  a.  Context for identity interpretation Imagine that the following animals were paired together in a hunting expedition: (i) aja 'dog' and ologbo 'cat' (ii) agutan 'sheep' and ewure 'goat'. A t the end of the day, group (i) killed a rabbit while group (ii) killed  nothing. A week later the same contest was repeated and only group (i) still succeeded in killing some game, then speaker A rendered (48b). b.  [Aja ati ologbonaa] ni 6 pa dog and cat S A L Foe 3sg kill 'It is the VERY dog and cat that killed a game.'  eran. animal  Since proper names are treated as definite DPs (Longbardi 1994), it is also predicted that they should pattern like other definite nouns when they are focused. This is what the examples in (48)-(51) show. The examples in (48) and (49) show individual proper names while in (50) and (51) show coordinated names with nda.  (48)  (49)  (50)  Individual proper name in Focus position a. Context for uniqueness Olu has threatened to kill Ojo's dog. One day, the dog was reported dead. Ojo's son who heard of Olu's threat ran to his father and said (48b). In response, the father who also knows Olu to be a wicked person said (48c). b.  Olu ni 6 pa aja. O. Foe 3sg kill dog 'It is Ohi that killed the dog.'  c.  Olu O. (i) (ii)  a.  Context for identity Olu hates dogs. Last week, he killed Ojo's dog under the pretext that it ate his meat. Yesterday, another dog was killed and it was alleged that the same Olu was responsible. Speaker A who witnessed the killing said (49b).  (b)  Olu O. (i) (ii)  naa ni 6 le pa aja. S A L Foe 3sg able kill dog = 'It is ONLY Olu that could kill the dog.' *'It is ALSO Olu that could kill the dog.'  naa ni 6 pa aja. S A L Foe 3sg kill dog ='It is the VERY Olu that killed the dog.' *'It is ALSO Olu that killed the dog.'  Coordinated Proper names in Focus position a. Context for uniqueness The following children are paired together for a dancing competition (i) Kiinle and Titi (ii) Tunde and Moji. At the end of the day, group (i) won the competition. Commenting on the result, Speaker A rendered (50b).  (51)  b.  Kunle K. (i) (ii)  ati Titi naa ni 6 le jo. and T. S A L Foe 3sg able dance ='It is ONLY Kunle and Titi that can dance.' ^'It is ALSO Kunle and Titi that can dance.'  a.  Context for identity A week later the same contest was repeated and the same group (i) won the competition, then speaker A rendered (51b).  b.  Kunle K. (i) (ii)  ati Titi naa ni 6 le jo. and T. S A L Foe 3sg able dance =' It i s the VERY Kunle and Titi that can dance.' *'It is ALSO Kunle and Titi that can dance.'  Thus, in focus position, nominal expressions that occur with nda are interpreted either as unique ( O N L Y X ) or as identical to a previously mentioned referent (the V E R Y X ) . Crucially, the additive function of nda (ALSO X) is not found with focused elements.  5.2.2.2 Nad marks additivity or identity with (non-focused) arguments  Nouns and proper names can also be marked by nda in non-focused positions. When this happens, the non-focused arguments can be construed as added to the set of relevant discourse referents (ALSO X ) or are construed as identical to a previously mentioned referent (the V E R Y X). I first consider the additive reading. I split the relevant data into two. In (52)-(53) are the examples that show how the occurrence of nda with individual nouns and personal names gives rise to an additive interpretation. (52)  Individual nouns in A-position (a) Out of fear that someone might want to steal his pets, my neighbour left her dog and cat in the living room and went shopping. When she returned she quickly looked for the cat and found it sleeping on the floor. Looking around, she discovered that the dog too was lying on a couch. After a sigh of relief she said (52b). b.  Aja naa wa nibe. dog S A L be Loc-there 'The dog ALSO is there.'  (53)  Individual Personal name in A-position a. M y wife left Taye and Kehinde with our neighbour and went shopping. On her return she called on Taye from the gate and he answered. Calling Kehinde, she received no answer. She asked me if he had left our neighbour's place and I said (53b) because I just heard his voice within an earshot. b.  Kehinde naa wa K. S A L be 'Kehinde ALSO is there.'  nibe! Loc-there  Similarly, in (54)-55) are the examples showing the co-occurrence of coordinated nouns and personal names with the same salient element giving us the additive interpretation.  (54)  Coordinated nouns in A-position a. M y domestic animals: a dog, a cat, a goat and a sheep are never found outside after 6pm. By 7pm this evening, my son came to inform me that my goat and sheep were in the backyard. To my disbelief, I followed him to the backyard. When we got there, truly the goat and the sheep were there. Looking further, he saw some other animals and said (54b). b.  (55)  , Aja ati ologbo naa wa nita dog and cat S A L be Loc-outside 'The dog and the cat ALSO are also outside.'  Coordinated personal names in A-position a. Bose ran into the house from the backyard and told me that some people were abusing her because she was playing with her brothers. I asked her to point at them. First, she pointed at Titi and Moji. When I asked of the other people, she said (55b). b.  Olu ati Kunle naa wa O. and K. S A L be 'Olu and Kunle ALSO are there.'  nibe. Loc-there  I discuss the identity interpretation next beginning with the example of an individual noun (56) and an individual personal name (57). The identity interpretation is obtained for both of them, which implies that such nominal expressions have been previously mentioned in the discourse.  (56)  Individual noun in A-position a. My son told me there was a dog in the backyard that stared at him as he was entering through the gate. When we got there the dog was still there and my son said (56b). b.  (57)  Baba, aja naa wa nibe. father, dog S A L be Loc-there ' Father, the VERY dog is there.'  Individual personal name in A-position a. M y son told me that one man named Olu who he has just pursued for stealing his money was in the backyard staring at him. By the time my son led us there, he was still there and my son said (57b). b.  E wo 6 Olu naa wa 2pl look 3sg O. S A L be 'Look, the VERY Olu is there.'  nibe. Loc-there  In (58)-(59) are the coordinated nouns and personal names. Like their individual counterparts, when they take the salient element naa, both the identity and additive interpretations can be obtained but I only illustrate the former with the example in (58).  (58)  Coordinated nouns in A-position a. Ladi told me that a dog and a cat were in the backyard starring at him as he entered. When we got there they were still there and he said (58b). b.  (59)  Aja ati ologbo naa wa nibe. dog and cat S A L be Loc-there 'The VERY dog and cat are there.'  Coordinated personal names in A-position a.  Olu and Kunle are naughty boys. Few minutes ago, Omole came and told me the two were in the backyard throwing pebbles at him. When we got there, they were still there and he said (59b).  b.  Olu ati Kunle naa wa nibe. O. and K. Sal be Loc-there • 'The VERY Olu and Kunle are there.'  The conclusion that I draw from the data presented here is that non-focus arguments marked with naa are construed either as additive (the A L S O X reading) or as standing in an  identity relation to a previously mentioned referent (the V E R Y X reading). Notably, nonfocused arguments are never construed as unique (the O N L Y X reading). The table below shows the overall summary of N naa in both focus and non-focus constructions.  (60) N nda in focus and non-focus domains N naa FOCUS POSITION UNIQUENESS FUNCTION V IDENTITY FUNCTION A D D I T I V E FUNCTION  5.2.3  NON-FOCUS POSITION V  Testing for salience  The goal of this section is to demonstrate that salient nominals can't be indefinite but they can be "free choice" (defined below). I use two diagnostics to establish this claim.  5.2.3.1 Yoruba salient nominals can't be indefinite  Relevant to the present discussion is the fact that Yoruba bare nouns can be either indefinite or definite. The salient marker nda predictably should not co-occur with indefinite NPs because they are not discourse-linked nouns. However, at the surface level, it is difficult to tell which bare nouns are definite and which ones are indefinite. The question that arises is the following: how do we know that truly, indefinite nouns cannot take ndal One way to test this is by finding unambiguous indefinite nouns to combine with nda. The readily available nouns are the negative polarity items (NPI) i.e. the " N - k i - N " derived nominals (Koch 2004). On independent grounds, Koch claims that Yoruba negative polarity items such as enikeni in (61) are existentials and as such are indefinites.  (61)  Ko daju pe enikeni N E G certain that person.kf.person 'I am not certain that anyone is here.'  wa be  nfbi in.here (Koch 2004: 3)  If negative polarity items in general, and enikeni in particular, are indefinites, then we expect that their co-occurrence with nda should be bad. This prediction is supported by the ungrammaticality of (62b) where we see that nda cannot occur with enikeni. (62)  Enikeni lo wobe focus le anyone Foc-3sg able enter-there yara ti k6 nilekun room that Neg have-door 'It is anyone that can enter a room without a door.'  a.  10  *Enikeni anyone yara ti room that  naa SAL ko Neg  lo Foc-3sg nilekun have-door  le able  wobe enter-place  I extend the search to three other syntactic environments, including negation (63), conditionals (64) and Yes-no questions, (65). In all of these contexts, the co-occurrence of naa with the indefinite enikeni  'anyone' yields ungrammaticality as shown in the (b)  examples of (63-65). (63)  (64)  a.  Enikeni  b.  * Enikeni naa ko anyone S A L neg  a.  T i enikeni ba pe so if anyone Aux call tell 'If anyone calls, tell me  /  (65)  nile. at-home  ko si anyone neg be 'No one is at home.'  negation  si nile. be at-home fun m i . for lsg  b.  *Ti enikeni naa ba pe so fun m i . if anyone S A L Aux call tell for lsg  a.  Nje o ri enikeni Q-tag 2sg see anyone 'Do you see anyone?'  conditional  Yes-no Q  This saying is very famous among the eegun aldrmjo (elewe) 'traveling dancers'. It is contained in one of the songs that are used to compare a lady who is not attached to any man to a room that has no door. As such, any man can ask for her hand in love.  b.  *Nje o r i enikeni naa Q-tag 2sg see . anyone S A L  The fact that salient marking is not possible with enikeni is consistent with Koch's claim that it is a polarity item. But as I show in the next subsection, when enikeni occurs in a free choice environment it can occur with nda.  5.2.3.2 Yoruba salient nominals can be "free choice"  Drawing from Shank (2004: 70-72), I define free choice items as indefinite NPs whose domain has been widened. When these items are in a focus construction, the contextually 11  supplied domain, which restricts a negative polarity indefinite NP, can be widened. In this section, one case involving domain widening is considered for enikeni, and I argue that the fact that nda can co-occur with this item is due to the widening of its domain. By so doing, a possible scalar implicature is canceled. Specifically, whenever nda co-occurs with enikeni, its interpretation is no longer limited to "anyone". Rather, it has been widened to include "anybody at all".  (66)  a.  Context: O k o y a , a business mogul is throwing a party to mark his 60 birthday. T i t i , who is a friend to one of his daughters to her disbelief, heard that the party is open. She went to her friend and asked (66b). Then, her friend replied with (66c).  b.  Se ooto ni pe enikeni naa ni Q true be C anyone S A L Foe 6 le wa si pati ojo-ibi dadi. 3sg able comeLoc party day-birth daddy 'Is it true that it is anybody at all that can attend daddy's birthday party?'  c.  Ooto ni, enikeni naa ni 6 true be anyone SAL Foe 'It is true, anybody at all can attend.'  th  le 3sg  wa able  come  "For details on definition of "free choice", see also among others Haspelmath (1997) and Giannakidou 1998).  In this context, enikeni has a free choice reading. With this I conclude my discussion reiterating the claim that the distribution of naa is semantically conditioned: it never cooccurs with indefinite non-specifics (e.g. someone/anyone), but when its domain is widened, it can have a free choice reading that enables it to be construed as "anybody at all".  5.2.3.3 No dedicated marking for salience in English  I have shown that Yoruba has a dedicated word that marks salience on nouns. This is the element naa. This element is capable of marking nouns for uniqueness, additivity, or identity. In this section, I show that just as English does not overtly mark specificity on definite DPs, so it is with salience: English does not have dedicated words to mark nouns as salient. In this case, a definite DP is also considered as salient (Heim 1982). This fact is illustrated in (66). The claim is that the man can either be specific (66b) or salient (66c). Although, there is no distinct morpheme used in English to mark salience, as mentioned above, salience can be marked prosodically by means of an intonational prominence of the determiner.  (67)  a.  [A man] came in.  b.  [The man] sat down.  definite specific  c.  [THE man] was John's father.  definite salience  As we have seen in the translation of the various Yoruba examples, English does have lexical items that perform the functions of uniqueness, additivity and identity, namely only, also and very. (68)  12  a.  Only the man sat down.  uniqueness  b.  The man also sat down.  additivity  c.  The very man sat down.  Identity  From the foregoing, it is clear that English marks nouns for salience optionally. Worthy of note is that English does not have dedicated word for salience. Those words identified as capable of marking salience can also be used for other functions. For example 'very' can also function as a degree adjective, (i) a. I like the lady b. I like the lady very much 12  In conclusion, while Yoruba marks uniqueness, identity, and additive with nda, In English, different items fulfill these functions.  5.2.4 Nda combines with other functional elements within the DP  The salient marker ndd can also co-occur with either the specificity marker kan or demonstratives. I also show that ndd, kan and demonstratives yli/yen can all co-occur with a definite DP in certain ways. First, when kan co-occurs with nda in focus construction, the definite construal of the DP is maintained and in addition such DP is marked as standing in an identity relation to a previously mentioned referent that is unique. Of particular interest is the fusion of identity and uniqueness interpretation. I present the full data set in (69)-(71).  (69)  [NP SPF SAL] in focus position Aja kan naa nimo ri / dog SPF S A L FOC lsg see (i) ='It is the VERY SAME dog that I saw.' (ii) 'It is the dog TOO that I saw.'  Identity/unique * Additive  The next set of examples show how demonstratives combine with the salient marker. Observe that the linear order is Demonstrative-Salience.  (70)  [NP Dem Sal] in focus position Aja yii naa ni mo ri dog D E M S A L FOC lsg see (i) = 'It is THIS VERY dog that I saw.' (ii) * 'It is this dog TOO that I saw.'  IdentityIunique * Additive  And when all the three elements: demonstrative, specificity and salient markers combine to mark a DP, the linear order is Demonstrative-Specificity-Salience and the DP is marked as standing in an identity relation to a previously mentioned referent that is also unique. This is shown in (71).  (71)  [NP D E M SPF SAL] in focus position [Aja yii kan naa] ni mo ri dog Dem certain sal FOC lsg see (i) = 'It is THIS VERY SAME dog that I saw.' (ii) * 'It is A L S O this certain dog that I saw.'  Identity/unique * Additive  The examples in (72)-(74) illustrate the occurrence of these same elements in the same combination with bare nouns in non-focus position. Like the focused construction, both identity and uniqueness readings are obtained. When nouns co-occur with the salient and specificity markers (72), or with the specificity marker, the salient marker, and a demonstrative (74), such nouns are marked as standing in identical relation to a previously mentioned referent that is unique. In contrast, when a noun co-occurs with the salient marker and a demonstrative (73), such noun is marked as standing in identical relation to a previously mentioned referent and at the same time marked for additivity. (72)  (73)  (74)  [NP SPF SAL] in non-focus position Mo ri aja kan naa lsg see dog SPF S A L (i) ='I saw the VERY SAME dog.' (ii) * T saw certain dog TOO.'  Identity/unique * Additive  [NP D E M SAL] in non-focus position Mo ri aja yii naa lsg see dog DEM SAL (ii) * T saw THIS V E R Y dog.' (i) = 'I saw this dog TOO.'  * Unique Identity/additive  [NP D E M SPF SAL] in non-focus position Mo ri aja yii kan naa lsg see dog D E M SPF S A L (i) = 'I saw THIS V E R Y S A M E dog.' (ii) * 'I saw this dog TOO.'  Identity/unique * Additive  The summary of the findings is shown in (75). While focused arguments of the string: [Nouns/Proper names (Dem) (SPF) naa] are construed as unique and identical to a previously mentioned referent, when the same arguments of the same string are in non-focused position, they are construed as additive and identical to a previously mentioned referent. In all cases, whether focused or not, they are always construed as identical to a previously mentioned referent.  (75)  Interaction of nda, kan and DEM with bare nouns in focus and non-focus positions  Proper name + naa Definite noun + naa NP Dem + naa NP Spf + naa NP Dem Spf + naa  Focus (A') position Non-focus (A) position unique additive identity unique additive identity V V V V x . --« x V V V V * •i' x # x^.-.-J •Jr'-'x .-} \ X V V V \ V V X V •BiJHHl \ \ V V V  In conclusion, this section has established the following. First, in focus constructions, a nominal expression (proper name or noun) that co-occurs with the salient marker (with or without a demonstrative or the specificity marker kan) is always marked as standing in an identity relation to a previously mentioned referent that is unique. Second, in non-focus constructions, a nominal expression (proper name or noun) is marked as standing in an identity relation to a previously mentioned referent and as additive if such a noun co-occurs with the salient marker nda (with or without a demonstrative). Third, in non-focus constructions, when a bare noun is marked for salience and specificity, such a noun is marked as standing in an identity relation to a previously mentioned referent that is unique Thus, the presence of the specificity marker kan precludes the additive reading, and forces the identity reading. The reason as to why this comes to be so is subject to further research. I now turn to the syntax of these two elements in §5.3.  5.3  The syntax of kan and nda  Having shown the distribution and the semantics of kan and nda, I now give a syntactic account. I have shown that indefinite NPs are morphologically marked for specificity with kan. I analyze kan as D and propose the structure in (76) where the N P starts as a complement of D and moves to Spec DP to derive the surface linear order of [NP D].  (76)  [ o p N P i \D kan]  a.  b.  DP NP  D  l  NP  I kan On the other hand, I propose that naa as a marker of salience modifies D(em)Ps and is to be analyzed as an adjunct to D(em)P with the structure in (77).  (77)  a-  [  D(em)P  NP[D(EM)yif]t ]naa i  b.  D(EM)P D(em)P^^^^SAL NP  naa  As the structure in (77) shows, it is possible for the salient marker naa to mark the entire DP as salient or the functional element within the DP in this case, yli 'this', since it is in ccommand relation to the entire DP. 1 assume the former, namely that naa marks the whole 13  DP.  l3  Viewing naa as a modifier of functional structure has cross-linguistic support. The intuition of native speakers  of English is that in the phrase [even this book], the nuclear stress (association with focus effect) is on [this] not on [book]. [Book] is anaphoric, so somehow it escapes from the scope of the focus modifier [even] as shown by the contrast between (i-b) and (i-c). (i)  a.  DP D  NP book  this DP  DP  even D  NP book  The number interpretation of DPs that are marked by nda is the focus of the last section.  5.4  The number interpretation of ndd  This section discusses the connection of the salient marker nda with number marking. I observe that nouns that are marked by nda have only a singular interpretation except if the noun is independently marked for plural. First consider the examples in (78) where a bare noun occurs by itself. The outputs show that bare nouns are ambiguous between singular and plural interpretation.  (78)  a.  b.  Mo ri aja lsg see dog = 'I saw a dog.' = 'I saw dogs.  singular plural  Taye ra oro T. buy mango = 'Taye bought a mango.' = 'Taye bought mangoes.'  singular plural  The examples in (79) show that when bare nouns occur with only the salient marker the interpretation of these bare nouns is singular.  (79)  a.  NP + nda = singular  Mo ri aja naa lsg see dog S A L 'I saw the very dog.'  *DP  c.  DP  even D this  NP :  .  [book ] FOC  b.  Taye ra oro naa T. buy mango S A L 'Taye bought the very mango.'  NP + ndd= singular  It follows then that if a bare noun that occurs by itself is ambiguous between singular and plural interpretation but the same noun is obligatorily interpreted as singular when it takes this salient marker, it must be the case that the salient marker is inherently singular. To test this, we can look at the behaviour of nda with nouns that are overtly marked as plural. If nda is inherently singular, such sequences will be ill-formed. If nda is pragmatically singular, such sequences will be well-formed.  (80)  a.  Mo ba [awon ijimere ] nfnu lsg meet P L red-monkey P T met some red monkeys in the farm.'  b.  Mo le [awon ijimere] lsg greet P L red-monkey T chased the very red monkeys.'  oko farm  naa]] SAL  (80b) indicates that the salient marker is pragmatically singular rather than inherently singular.  5.5  Conclusion  I have shown how specificity and salience readings are obtained in Yoruba nominal expressions. I argued that an indefinite N P that is morphologically marked for specificity with kan has the interpretation of 'a certain N ' , whereas a definite DP that is morphologically marked for salience with nda can mean 'the V E R Y N \ ' O N L Y N ' or ' A L S O N ' . I also showed that though salience is not compatible with indefiniteness, in free choice cases, the salient marker nda could combine with indefinite NPs whose domain had been widened, yielding an 'anybody at all' reading. Syntactically, I claimed that kan is a deictic D whereas nda can modify a functional head, such as D(em) and it is analyzed as an adjunct to the DP.  C H A P T E R SIX: P L U R A L STRATEGIES IN Y O R U B A  6  Introduction  The concern of this chapter is to account for how plural marking is carried out on Yoruba nominal expressions. I have identified three strategies (cf. Ajiboye 2005c). First is contextually determined plurality, as in (1). The term "contextually determined plurality" is intended to capture the fact that, in the absence of a P L U R A L feature, the interpretation of a nominal expression as singular or plural is contextually determined. This is the focus of §6.1.  (1)  Moji M. (i) (ii)  ri [ejo] lona oko see snake on-path farm 'Moji saw a snake on her way to farm.' 'Moji saw snakes on her way to farm.'  The second plural-marking strategy is semantically determined, as in (2). The term "semantically determined plurality" is intended to capture the fact that certain lexical items, by virtue of their inherent semantics are necessarily construed as plural. In Yoruba, this is the case for group-denoting quantifiers and for numerals whose cardinality is two or greater than two. I account for this in §6.2.  (2)  a.  Mo ra [iwe pupo] lsg buy book many 'I bought many books.'  b.  Mo ra iwe meji lsg buy book two 'I bought two books.'  The third strategy is morphologically determined plurality, as in (3). The term "morphologically determined plurality" is intended to capture the fact that certain morphemes have an exclusively plural function, i.e., they mark plurality and nothing else. In Yoruba, there are three kinds of plural morphemes: the plural marker awgn that occurs before nouns, the plural prefix -won that attaches to demonstratives, the plural copy  morpheme, that copies modifiers to indicate plurality on the NP. 1 account for this in §6.3. 1  (3)  a.  awon okunrin Mo kf lsg greet PL man T greeted men.' Mo ri [aja lsg see dog 'I saw these dogs.'  won-yii] PL-DEM  Yoruba ko ni [asa buruku Y. Neg have custom COPY 'Yoruba does not have bad customs.'  buruku] bad  I propose that plurality is associated with two syntactic positions. One is with nouns, the other with non-nouns. I also demonstrate that Yoruba plural words are left adjoined to the nouns/non-nouns that they mark for plural.  6.1  Contextually determined plurality  The first observation is that if a noun occurs by itself or with a modifier, there is no way to know if that noun is to be interpreted as singular or plural.  (4)  a.  Moji M. (i) (ii)  ri [ejo] lona oko see snake on-path farm 'Moji saw a snake on her way to farm.' 'Moji saw snakes on her way to farm.'  b.  Moji M. (i) (ii)  ri [ejo dudu] lona oko see snake black on-path farm 'Moji saw a black snake on her way to farm.' 'Moji saw black snakes on her way to farm.'  Strictly speaking, both semantically determined plurality and morphologically determined plurality involve the semantic feature [PLURAL]. The distinction lies in the fact that plural quantifiers introduce a [PLURAL] feature by virtue of their group denoting semantics, while plural morphemes introduce a [PLURAL] feature and nothing else. 1  I show that the interpretation of such nominal expressions as singular or plural in Yoruba is partly contextually determined. I claim that this is an instance of General number.  6.1.1 The General number analysis  The idea of General number adopted here follows Rullmann and You (2003) and Rullmann (2004). On the basis of the semantic and pragmatic properties of bare nouns, they conclude that bare nouns are neither singular nor plural; i.e., they are unspecified for number (cf. Deprez 2004: 10). In order to determine whether a singular or a plural interpretation will be applicable, one has to put such bare nouns in a sentence. Rullmann and You claim that the example in (5), which has general number, has the same truth conditions as (6) which has a semantically singular object (p.2).  (5)  Zuotian wo mai le shu yesterday I buy A S P book 'Yesterday, I bought one or more books.' (Rullmann and You 2003: 1)  (6)  Zuotian wo mai (yi) yesterday I buy one 'Yesterday, I bought a/one book.'  ben CL  shu book (Rullmann and You 2003: 2)  The next two subsections examine count and mass nouns in Yoruba and show that both categories behave the same way with respect to the general number property. There are two distinctive ways by which nouns that are unspecified for number in Yoruba can be interpreted. On the one hand, there are nouns that are not specified for number and as such they can be interpreted as singular or plural. I treat this type in §6.1.2. On the other hand, there are nouns that are not specified for number and they can only be interpreted as singular. I treat this in §6.1.3.  6.1.2  Unspecified for number resulting in ambiguity  When a count noun occurs by itself or when it takes a modifier, it can either be construed as singular or plural. In (7a) ejd 'snake' occurs independently and as shown in interpretation (i), it can refer to one snake. But it can also be the case that the snake is more than one, as in interpretation (ii). Similarly, in (7b) where ejd 'snake' takes the modifier dudu 'black', it can either be construed as singular or plural. (7)  a.  Moji M. (i) (ii)  ri [ejd] lona oko see snake on-path farm 'Moji saw a snake on her way to farm.' 'Moji saw snakes on her way to farm.'  b.  Moji M. (i) (ii)  ri [ejd dudu] lona oko see snake black on-path farm 'Moji saw a black snake on her way to farm.' 'Moji saw black snakes on her way to farm.'  Like count nouns, bare mass nouns can also be interpreted as singular or plural depending on context. In (8a), miliiki 'milk' can be interpreted as singular or plural. It can also be interpreted as a quantity. Similarly, when it takes a modifier as in (8b) the same two interpretations of singular and plural are obtained. (8)  a.  Mo ra [miliiki] lsg buy milk = 'I bought a tin/carton of milk.' = ' I bought some tins/cartons of milk.'  singular/plural  b.  Mo ra [miliiki gberefu] lsg buy milk powdered = 'I bought a tin/sachet of powdered milk.' = 'I bought some tins/sachets of powdered milk.'  singular/plural  Since Yoruba nouns can be unspecified for number, this suggests that number marking is underdetermined in Yoruba. Based on the data and the discussion above I propose the structure in (9) for Yoruba bare nouns and claim that if an N P is unmarked for plural, then the N P can be interpreted as singular or plural.  (9)  a.  NP ejo 'snake(s)'  b.  NP NP^^^^Mod ejo dudu 'black snake(s)'  As a result of being unspecified for number, the interpretation of such an N P as singular or plural is contextually determined. The example in (10) shows a context where aja 'dog' can be interpreted as singular. (10)  a.  Context for singular interpretation: I. was walking on a path to my house at night and a dog appeared from nowhere. When I got home, I say (10b) to my wife.  b.  Mo ri aja lojii ona lsg see dog on road T saw a dog on my way home.'  ile house  On the other hand, the example in (11) shows a context where aja 'dog' can be interpreted as plural. In Saworoide movie, Adebdmf was telling a story of a hunter and his dogs to his children.  (11)  2  a.  Context for plural interpretation: The hunter in the story uses songs to summon his dogs in times of danger. The dogs then would run quickly to his aid.  b.  Aja a mi da o dog M T S lsg Q-tag Emph 'Where are my dogs?'  c  Aja ode dog hunter 'my hunting dogs'  Saworoide is one of the most famous Nigeria home video movies written by Professor Akfnwunmi, produced in (1999) by the Mainframe Film Production under its director; Tunde Kelanf.  2  d.  okemokerewu..., osopakagbomomi..., ogbalegarawe  e.  Aja ode dog hunter 'my hunting dogs'  f.  E sare e mia b6 o 2pl run 2pl Prog come Emph 'You should all come immediately.'  The mention of okemgkerewu,  gsgpakagbgmgmi,  and ogbalegarawe in this song leaves  no one in doubt that ajd 'dog' can only be interpreted as plural. Crucially, Yoruba differs from some other languages where the expression of plurality is inflectional. In such languages, plural marking is obligatorily expressed via a dedicated plural morpheme. Consider the English examples in (12). Adding the plural suffix 's to 'dog' differentiates between the singular interpretation (12a) and the plural interpretation (12b). (12)  a. b.  I saw a dog on my way home.' I saw dog-s at the Vancouver city hall today.'  singular plural  I conclude along the lines of Corbett (2000) that the issue of ambiguity of number interpretation of nouns as singular or plural is one of the peculiarities of languages with no overt dedicated plural marking. The next section addresses the other case of the unmarked 3  nouns that are unambiguously interpreted as singular.  6.1.3  Unspecified for number with obligatory singular interpretation  The other context where bare nouns are not specified for number is when they co-occur with a demonstrative (13a) or the numeral "one" (13b). In this context, the noun is obligatorily interpreted as singular.  4  Matthevvson (personal communication) notes that Brazilian Portuguese has number-neutral bare nouns and has a real plural marker. Note that the Demonstrative follows N. I assume that a demonstrative is a functional head that takes an NP as its complement. Further, I assume that Yoruba is a Head initial language. So, in cases where the head follows  3  4  (13)  a.  Mo ri [aja lsg see dog 'I saw this dog.'  yii] DEM  N  b.  Mo ri [aja lsg see dog 'I saw one dog.'  kan] one  N  DEM  NUM=1  The obligatory interpretation of nouns as singular whenever they take demonstratives being reported here is a little surprising. Recall that when a noun occurs by itself, it can be interpreted as either singular or plural. One wonders while taking a demonstrative will make a noun to be obligatorily singular. In light of this, I assume that demonstratives in Yoruba are inherently singular and therefore force nouns they occur with to have an obligatory singular interpretation. This claim is further discussed in §6.3 where demonstratives are marked for plural. To recap, we have seen that a noun that is not specified for number can be interpreted as singular or plural if it occurs by itself or if it takes a modifier. On the other hand, if the noun occurs with an inherently singular numeral ("one") or a demonstrative, that noun is obligatorily interpreted as singular.  6.2  Semantically determined plurality  One means by which NPs are unambiguously expressed as plural in Yoruba is when they occur with group-denoting expression that is inherently plural. The lexical items that are inherently plural in Yoruba nominal expressions are quantifiers and numerals. This is in contrast with Chierchia's (2005:8) claim that quantifiers generally lack inherent number  the complement in the surface syntax, I account for this via movement. In this case, the NP moves to Spec Dem (cf. Kayne 1994). (i)  DemP NP Dem  feature. Rather they receive this through agreement. As I show in this section, Yoruba quantifiers are inherently plural and this is why whenever they combine with nouns which are unspecified for number such nouns are obligatorily interpreted as plural.  (14)  a.  Mo ra [iwe pupo] lsg buy book many 'I bought many books.'  b.  Mo ra [iwe meji] lsg buy book two 'I bought two books.'  These quantifiers and numerals have an abstract [PLURAL] feature, which can be realized on nouns that they co-occur with. I call this an instance of semantically determined plurality. I propose that the plural interpretation of such nouns is accounted for through a feature percolation mechanism. The formulation of this mechanism is the subject of the next section.  6.2.1  The feature percolation analysis  There are two mechanisms of plural marking that I consider: feature percolation and feature matching. I start by giving a general view of the feature percolation mechanism before coming to the plural feature percolation in particular. I formulate the notion of percolation in the sense of copying where the copied feature is a, as outlined in (15).  (15)  a.  Node X immediately dominates node Y X •Y  b.  Y is specified for the feature a, X is unmarked for the feature a X  c.  The feature a is copied on to X [a] X  I illustrate how plural feature percolation works with nominal expressions that are pluralized by awgn '3pl pronoun' or a demonstrative. Observe that the plural feature can be realized on the noun (16a), the demonstrative (16b) or on both (16c). (16)  a.  awon isu yen PL yam Dem 'those yams'  b.  isu won- yen yam P L Dem 'those yams'  c.  awon isu PL yam 'those yams'  PL  won- yen PL Dem  PL  N  Dem  N  PL-  Dem  N  PL-  Dem  The assumption is that plural feature percolation copies the plural feature of a node onto the node that immediately dominates it, as in (17).  Input  Output Plural percolation [PLURAL]  It is possible for percolation to come from more than one node within a nominal expression. Using (16) as an illustration, plural feature can percolate through awgn as in (18a), through the prefix -won as in (18b) or through both awgn and -won as in (18c). In the latter case, the DP ends up receiving the plural feature from both NP and Dem.  (18)  a.  Plural percolation through awgn D(em)P PL  ^NP,,r PL  NP  . I  I  awon b.  D(em) y  i{  isu  plural percolation through w o n D(em)P PL  plural-percolation through awoii and won D(em)P PL  PL  NP  awon  aja  I.  PL  Dem  L  won  The mechanism of feature percolation is not new in the literature. In its broad use as a well-formedness condition, Selkirk (1982) and Scalise (1984) define percolation as follows:  (19)  If a constituent a is the head of a constituent 6, a and (3 are associated with an identical set of features (syntactic or diacritic) (Selkirk 1982: 21)  In the same spirit, Owolabi (1995: 106) claims that percolation is a device which enables a complex word to inherit the syntactic properties (or features) of its head. This implies that feature copying is usually from the head. These percolation approaches differ from the present analysis in one respect. In my own proposal, what makes an N P plural does not rely on the head alone. In particular, a plural feature of an adjunct can percolate onto the NP if the head noun is not specified for plural. Further, the general formulation of feature percolation in (15) can apply to a wide  range of domains in the grammar of any language. This depends on the target of copying as well as what is copied. When a feature a is copied in Phonology (e.g. under Optimality theory), the process is called reduplication, as in (20a). This process can manifest itself as feature spreading or assimilation, where phonological features are spelled-out at the level of pronunciation (PF). A t the syntactic level, C O P Y a is a syntactic feature, (20b). This is carried out through movement. In the minimalist theory this copying process involves movement of a phrase. In this approach, the base is deleted after movement has taken place. This is called "copy" and "erase" (Chomsky 1993, 1995) and it manifests itself in the linear order of lexical items. Lastly in semantics, a semantic feature that is copied is manifested in meaning, (20c).  (20)  COPY in three domains of grammar  COPY a  Feature  Process  Target  Output  a  a = phonological feature  b  a = syntactic feature a = semantic feature  [x] [x] segments [x] - W phrases [x] [x] features  pronunciation (PF) linear order  c  "reduplication" local, no deletion "movement" non-local, deletion "percolation" local, no deletion  Meaning (LF)  Closely linked to feature percolation is feature matching which is more restrictive than feature percolation in the sense that all members of a nominal expression within the phrase must agree with respect to the plural feature. Feature matching contrasts with feature percolation; in the latter the requirement is that at least one member of the phrase has a plural feature. I discuss the formulation of feature matching mechanism in §6.3.5. Finally, in the account of the plural strategies across languages, I propose (21).  (21)  Plural marking Plural is resolved through feature-matching if possible, elsewhere through feature-percolation.  I am assuming that Yoruba always uses the elsewhere condition i.e., the feature percolation mechanism. Having laid out the mechanism that is involved in accounting for plural marking, I  now turn to the two types of semantically driven plural marking in Yoruba.  6.2.2  Inherently plural Quantifiers  There are three quantifiers that force a plural construal on NPs in Yoruba. These are pupo 'many',cY/e 'few' and gbogbo  'all'.  5  In (22), where the only element in the nominal  expression apart from the bare N is the quantifier, we observe that the whole NP is construed as plural.  (22)  6  a.  M o ra [we pupo] lsg buy book many 'I bought many books.'  NPQ  b.  M o ra pwe die] lsg buy book few 'I bought few books.'  NPQ  c.  M o ra gbogbo lsg buy all ,'I bought all (the) books'  iwe book  Q NP  The task here is to account for the obligatory plural construal of the noun with these quantifiers. The data in (22) suggest that there is an abstract [PLURAL] feature on quantifiers that forces the plural interpretation on NPs. Using the feature percolation mechanism, I propose that the abstract [PLURAL] feature of the quantifier percolates onto the N P as illustrated in (23) . Lawal (1986, 1989) and Adewole (1989) claim that pupo is derived from the verb po 'be plenty'. Two other words that are derived from the same word are opo and opolppo, both of which mean 'many/plenty'. ^he syntactic position of gbogbo in relation to pupo 'many', and die 'few' is not clear at the moment except to stipulate that while gbogbo, a universal quantifier precedes the NP, pupd 'many' and die 'few' follow the NP. Note also that the Plural morpheme gbogbo can co-occur with awon. When this happens, gbogbo precedes, awon. (i) a. Gbogbo awon 91110 on de all PL child HTS arrive 'All the children have arrived.' b. *Awon gbogbo omo on de PL all child HTS arrive 5  (23)  a. NP  &>L  rl  pupo [PLURAL] 'many'  iwe 'book'  NP,PL NP  QPL  N | iwe 'book'  die [PLURAL] 'few'  L  NP,PL {.PL  gbogbo [PLURAL] 'all'  NP N iwe 'book'  Evidence from French supports the claim that quantifiers are inherently plural. The word plusieurs 'many' combines only with a plural noun. Compare (24a) where the noun is 7  plural with (24b) where it combines with a singular noun. For the latter, it results in ungrammaticality.  (24)  a.  plusieurs chevaux many horse.PL 'many horses'  b.  *plusieurs cheval many horse.sg 'many horse'  So, the fact that I establish here is that plusieurs is like Yoruba pupo in the sense that it only  7  One should not mistake the's' at the end of this quantifier to mean a plural morpheme.  occurs with plural nouns. In English, the quantifiers 'many' and 'few' among other quantifiers are also inherently plural. Compare (25a) where 'many' combines with a plural noun with (25b) where it fails to combine with a singular noun.  (25)  a.  many orange-s  b.  *many orange  8  With this, I conclude that like French plusieurs, and English many, the Yoruba quantifiers pupo 'many', die 'few', and gbogbo 'all', are inherently plural. As such, they pluralize the NP they occur in through feature percolation. However, in a language like English, there must be an agreement between the noun and the quantifier with respect to the plural feature.  6.2.3  Inherently plural numerals  Another source of plural marking on Yoruba NPs comes from numerals. This should not come as a surprise, as any noun that takes a numeral denoting a set with cardinality l>\l should be able to have a plural interpretation in any language (Corbett 2000: 211). This is why Ionin and Matushansky (2004) claim that the semantics of numerals is the same crosslinguistically. According to them numerals always signify plural (cf. Chierchia 2005). First, I claim that numerals apart from 'one' all bear an abstract plural feature that enable us to interpret the noun they co-occur with as plural. I later show that English numerals predictably contrast with Yoruba in the sense that there must always be agreement between the noun and  Although, the quantifier 'many' is supposedly inherently plural, the phrase: [a man] may occur .with this quantifier in certain restricted contexts. (i) a. During the 1930s [many [a man]] sold his farm and moved west. b. During the 1930s [many men] sold their farms and moved west c. *many man Observe also the parallel situation in other English quantifiers: 'every' versus 'all'. Both denote groups. While 'all' takes a plural NP, 'every' takes the unmarked form. (ii) a. Every man b. All men c. *A11 man  8  its constituents in English. Such agreement is not required in Yoruba. I argue that this is because the two languages adopt different strategies for plural marking. Let us now go into details on how numerals acquire plural marking status. First, observe that Yoruba numerals have different forms (Abraham 1958, Bamgbose 1967, Awobuluyi 1978, Ajiboye & Dechaine 2004). Ajiboye & Dechaine discuss two forms that are crucial to my account of plural formation: the base form and the m-form, where the latter is derived from the base numeral by prefixing m- and H tone. The m-form contrasts with the base form phonologically because the initial tone of the base numeral is replaced with H tone. Some representative examples are given in (26). (26) a. b. c. d. e.  Base eni okan eji eta erin  m-form Output Gloss m + eni *meni 'one' 'one' m + okan *mokan m- + eji meji 'two' 'three' m- + eta meta merin 'four' m- +erin (adapted from Ajiboye & Dechaine 2004: 6)  Observe that the numeral 'one', which has two base forms (em'and okan), cannot take the m- prefix. In what follows, I present the syntactic distribution of the base and m-form. First, both types can occur by themselves as nouns: (27)  a.  M o ra eji lsg buy two 'I bought two.'  b.  M o ra meji lsg buy two 'I bought two.'  Second, only the m-form can occur as a modifier, and therefore only the m-form seems capable of marking plural (28). (28)  a.  *Mo ra lsg buy  [iwe eji] book two  b.  M o ra [iwe meji] lsg buy book two 'I bought two books.'  The m-form of numerals cannot occur with 6k an 'one' and eni 'one', (29) to modify nouns.  9  This suggests two things: (i) only a numeral that denotes a set with cardinality l>\l can be used to derive the 'm-numerals', (ii) the m-form has to do with the semantics of more than 'oneness'. I return to this later. *Mo lsg  ra buy  [iwe book  mokan] one  *Mo lsg  ra buy  [iwe book  meni] one  The analysis of numeral-as-plural-feature-bearer follows the one proposed for quantifiers. Adjoining the appropriate numeral form to an NP yields the desired result. The abstract [PLURAL] feature of the numeral percolates through the numeral to the NP, as illustrated in (30).  9  10  Note that in counting, eni can be in the m-form.  It appears only the m-form can co-Occur with a noun. But observe that the base form like eji 'two' also inherently contains an abstract [PLURAL] feature. As such it should be able to qualify to mark plural on nouns. But this is not the case. However there are certain instances where only the base form can modify nouns and consequently mark them for plural. There are a few things to note about such numerals. Firs't, they do not allow the m-form, (i-a). Second, they precede the noun they modify, (i & ii). Third, they are multiples of ten starting from ogun 'twenty'. See Abraham (1958:xxxii-xxxvi), (i-iii). (i) a. Sehun ra [ogun iwe] S. buy twenty book 'Sehun bought twenty books.' b. *Sehun ra [iwe mogun] (ii) a. Jenrola ta [pgbon isu] J. sell thirty yam 'Jenrola sold thirty yams.' b. *Jenrola ta [isu mogbon] When these numerals follow the nouns, they show ordinals and as such no longer mark nouns for plural. (iii) a. Sehun ra [iwe ogun] S. buy book twenty ' Sehun bought the 20 book.' b. Jenrola ta [isu 9gbon] J. sell thirty yam 'Jenrola sold the 30 yam.' There is more to say than claiming either that only the m-form or the base form of numerals qualifies as a plural marker. What determines which numeral must be in the m-form and which one must be in its base form to mark plural as well as the linear order betweeh the numeral and the noun require further research. 1 0  lh  lh  iwe 'book'  meji [PLURAL] 'two'  Taking my discussion of quantifiers and numerals in these last two sections together, I conclude that quantifiers and numerals converge in the sense that they both have an inherent abstract [PLURAL] feature. The remaining issue to address is how to account for the H tone of the m-numerals.  6.2.3.1 Accounting for the floating H tone of m-numerals Observe that each base numeral in (31) begins with L tone whereas each m-numeral consists of H tone that associates with the initial vowel of the base numeral thereby displacing the underlying L tone. I am going to argue that H tone of the m-forms comes from the morpheme mil 'take'. (31)  a.  eji 'two'  m  + eji  meji  eta 'three'  m  + eta  meta  enn 'four'  m +erin  merm  efa 'six'  m '+ efa  mefa  eje 'seven'  m  meje  + eje  There are two possible hypotheses for how to account for this H tone. One hypothesis is that this floating H tone is there underlying with the m-prefix as part of the plural morpheme.  (32)  a. b.  m' m'  + +  eji eta  > >  meji meta  'two' 'three'  Since floating tones are not allowed in Yoruba (Bamgbose 1966 among others) there are two options: the H tone either deletes or it associates with a tone-bearing unit adjacent to it. In the case under consideration, the H tone associates with the initial vowel of the base and the L tone deletes.  H m  e  j  1  —>  L  m e j i  Thus one way of resolving the H L contour tone is through deletion of the L-tone.  11  The other hypothesis is that the H tone is associated to a vowel underlyingly before the vowel deletes and the tone in turn associates with the nearest adjacent tone bearing unit (TBU). On the phonological content of this vowel Abraham (1958: xxxii) suggests that it is a prefix ma-. According to him, whenever the base form such as eji 'two' follows the noun, the prefix ma- links this numeral to the noun. In order words numerals such as meji 'two' and meje 'seven' are derived from ma + eji and ma + eje. This suggestion faces two problems. First is that of meaning. It is difficult to pick out the meaning of ma either as a lexical item or when it combines with numerals. The second is a phonological problem. Assuming that the prefix is M toned ma, one is faced with the question of how H tone surfaces when it combines with numerals whose initial tone is L . In the alternative, I suggest that the morpheme is mu 'take'. This suggestion comes from our knowledge of the counting system in Yoruba. Native speakers count, meni 'take one', meji 'take two', meta 'take three' etc. It is probably the case that this morpheme combines with the base form and undergoes vowel deletion as shown in (34). (34)  a.  mu  eji  >  m'  eji  >  meji  b.  mu  eje  >  m'  eje  >  meje  For details on resolution of contour tones in Yoruba, readers are referred to (Bamgbose 1966; Akinlabi 1985; Pulleyblank 1986; Laniran 1992). 11  Note that there is nothing unusual with vowel and tone deletion in the language. In particular, this kind of deletion is robust when two vowels are juxtaposed across a word boundary (Bamgbose 1966, 1967, 1989; O l a 1989; Oyelaran 1971; 1990 among others). Once a tonebearing unit deletes in this context, there is bound to be an extra tone. On which tone to delete, the claim in the literature is that when there is H and L concatenation, the L deletes and H survives because the tonal hierarchy in Yoruba is H>L (cf. Chumbow 1982; Pulleyblank2004).  12  6.2.3.2 Other languages that pattern with Yoruba  Yoruba is not the only language where numerals form plural words. There are other languages that require no further marking whenever a numeral that denotes a set with cardinality />1/ is used. Hungarian is one of such languages. In (35), lany 'girl' is marked as plural only by the presence of the numeral ket 'two'.  (35)  Ket lany beszelget two girl.SG chat.SG 'two girls are chatting'  Hungarian (Corbett 2000: 211).  However, there is a slight difference between Yoruba and Hungarian. In Yoruba it is possible to use other plural words to mark nouns for plural even when a numeral is present. This contrasts with Hungarian, which does not allow any other plural marking. The question is how a language like Hungarian would be treated in the present analysis. The explanation is that Hungarian permits only one instantiation of the P L U R A L feature. It could also be that there is a language specific rule that prohibits further plural marking once a numeral is introduced. Note that even for Yoruba all that is required is for at least one plural.marker to be present.  14  There is also one problem with the second hypothesis. This is the problem of why men'i{mu eni 'take one') cannot combine with nouns. Observe however that when we count, we always have a sense of plurality. We don't just count "one" thing. Therefore counting requires > 1 thing. This suggests that, although we count "1", we are counting part of a plural set of objects. See §6.3.4 for multiple plural marking. For Hungarian, one can also speculate that PLURAL takes precedence over SINGULAR, hence when an NP 12  13  1 4  Further, there are languages where the presence of a numeral is mutually exclusive to dedicated plural word i.e., where a numeral and a dedicated plural word do not co-occur. The reason for this is because numerals occupy the same position as the plural word. In Gurung, a Tibeto-Burman language of Nepal, the plural word occurs in the same syntactic slot as numerals. The examples in (36) taken from Dryer illustrate this.  (36)  a  b.  (37)  a  b.  ca that 'those ca that 'those  pxra-bae mxi person walk-adj walking people so-bra mxi person Numeral three hundred people'  jaga PL  (Dryer 1989: 872)  *ca that  pxra-bae walk-adj  mxi person  so-bra jaga Numeral  *ca that  pxra-bae walk-adj  mxi person  jaga PL  PL  so-bra Numeral  This means that plural word and numeral are in complementary distribution (Dryer 1989: 871). Unfortunately, there is not enough data from Dryer's work to ascertain why plural and numerals cannot co-occur in this language. This concludes my discussion of the semantically driven plural strategy.  6.3  Morphologically determined plurality  The third way by which Yoruba marks plural on nominal expressions is through the use of  contains a numeral that has an abstract PLURAL feature and a noun with a SINGULAR feature, the NP is interpreted as plural because of this precedence constraint. This is illustrated in (i). NP (i) P1  Num  N  [PLURAL]  [SINGULAR]  Ket  lany  certain dedicated words/morphemes whose only function is to mark plurality. I refer to this as a morphologically determined plurality. There are three plural morphemes in Yoruba: awon marks plural on nouns, the prefix won- marks plural on demonstratives, and a copy marks plural on modifiers. I call these morphemes 'plural words' along the lines of Dryer (1989). The term "Plural word" in this dissertation is used for a word or a morpheme that gives a noun or an entity it co-occurs with a plural interpretation. Observe that plural words 15  cross-linguistically do not belong to a natural syntactic class. This is consistent with Dryer's observation that the grammatical category of words that function as plural words varies from language to language. In Yoruba, awon is a pronoun. As for the won- prefix, although it is 16  clearly related to the 3pl weak pronoun won/won, it does not have the same distribution as weak pronouns. As for modifiers, to qualify as plural words, a modifier must undergo copying. Thus a modifier-plural word consists of ' C O P Y ' in Yoruba. In order to account for how these plural words mark nouns for plural, I also appeal to the feature percolation mechanism already motivated above.  Cf. Corbett's (2000: 135) definition of 'plural words' as special 'number words' that languages use to indicate number and Dryer's (1989: 865) definition which says 'a plural word is a morpheme whose meaning and function is similar to plural affixes in other languages.' 16 For example Hawaiian (i) and French (ii) use articles to mark plural. (i) a. ke keiki SG child 'the child' b. na keiki PL child 'the children' (Hawaiian Dryer 1989: 872) (ii) a. la pomme [lae pom] the, fern apple 'the apple' b. les pommes [le pom] the, plur apple 'the apples' (French Dryer 1989: 873) Pefioles Mixtec combines plural words with numerals e.g. (Dryer 1989). (iii) oko kwee tee twenty PL man 'twenty men' (Penoles Mixtec Dryer 1989: 872) 15  16  6.3.1  awon marks plural on nouns  I have already established in Yoruba that when a bare noun occurs by itself, there is no way by which we know whether it is singular or plural. This is why I claim that Yoruba bare nouns are not specified for plural. The only way by which plural is overtly marked on nouns is through the use of awgn '3pl strong pronoun' (Dryer 1989, Rowlands 1969). 17  18  In (38a)  awon marks gkunrin 'man' as plural, in (38b) it marks obinrin 'woman' as plural; in (38c) it marks both gkunrin 'man' and gmg Israeli 'child of Israel' as plural. However, in (38), in addition to awgn, there are other lexical items that relate to plural interpretation. As I demonstrated in §6.2, numerals like meta 'three' in (38a) have an abstract plural feature that can be realized on the noun in the absence of a plural word. Similarly, as I show in §6.3.2 below, the prefix won that attaches to the demonstrative yii in (38b) is a plural morpheme.  (38)  a.  [awon okunrin] PL man 'these three men'  meta y i i three this (Dryer 1989: 875)  b.  [Awon obinrin won-yi] w a tun p i n s i meji PL woman D E M come again divide to two 'These women again divide into two groups.' (Fagiinwa 1961: 15)  c.  [Awon okunrin kan] n i n u [awon omo Israeli] de ihin yii PL man Spef among P L child Israel reach place this 'Certain men among the children of Israel got here.' (Joshua 2: 2)  I propose the structure in (39) which shows that a Yoruba N P consists of a bare NP and a  Note that Yoruba is not the only language that uses 3pl to mark plurality. Others include Chamorro and Ngarinjin (Dryer 1989: 877), Angas and some Creoles (Corbett 2000: 135 fn.3). (i) a. mandjan biri stone PL 'stones' b. biri-ma-ra they-say-past 'They said.' (Dryer 1989: 87) See the chapter on genitive construction for the definition of strong pronoun. The universal quantifier gbogbo, patterns with awon in the sense that it also precedes the noun. 17  18  plural word that is left adjoined to the NP.  Applying the percolation mechanism, the plural  feature of awgn percolates onto the higher NP node. (39)  ^ ^ ^ N P  P  L  PL  NP  I  ^  awon  igi 'tree'  As we shall see, all plural morphemes are left adjoined to NP in Yoruba. For awgn, which is the plural morpheme associated with nouns, this means that it must precede the noun that it pluralizes, as shown in (40).  (40)  Plural marking on nouns a.  Mo bi [awon omo] lsg born P L child 'I have children.'  b.  Mo lsg  bi born  [*omoawon] child P L  This concludes the discussion on awgn as a plural word that marks plural on nouns. I now turn to awgn as a pronoun.  6.3.1.1  awgn  as a pronoun  Apart from functioning as a plural word, awgn is active in other domains in the syntax of Yoruba like any other Yoruba strong pronoun. Consider (41) and (42) where both the 3pl weak pronoun wgn and the 3pl strong pronoun awgn occur in argument position. (41)  a.  Jeff wanted to know if Solveiga and Mario were at my birthday party. In response, I said (41b).  This poses the problem of how to account for the disparity between Sabe, Idaiesa and Ketu dialects and Standard Yoruba. However there is not enough data that will reveal a wider range of the distribution of won in those dialects as such, I consider only the case of Standard Yoruba for this present study. 19  b.  Won wa 3pl come 'They came.'  Let us now consider the context where awgn is used. (42)  a.  Jeff was not sure if Sloveiga, her husband and my friends were at the party yesterday and when he asked of Solveiga I answered "no" because I did not see her. But when he asked of my friends, I said (42b) because I wanted him to know that truly they were present.  b.  Awon on 3 pi HTS 'They came.'  wa come  Two additional facts about awgn are revealed in the above examples. Compared to its clitic counterpart won, awgn is emphatic. The sentence in (41b) will be uttered in normal speech but when someone wants to be emphatic, (42b) is used. The use of awgn is also an expression of certainty. The speaker is very certain of the referents' presence in (42b) whereas (41b) does not necessarily reflect certainty or uncertainty of the referents' absence. These two properties are peculiar to all the so-called strong pronouns when they occur in argument position.  6.3.1.2 Appositive analysis of PL-NP A diachronic account contained in Dryer's work is that awgn in (43a) may be analyzed as a pronoun head, with gkunrin meta yii functioning as an appositive noun phrase. This predicts that awgn is also a NP. In the appositive analysis, the structure for (43a) would be (43b).  For details on the Yoruba pronominal system see among others Bamgbose (1967), Rowlands (1969), Pulleyblank (1986), Manfredi (1987), Dryer (1989), Adesola (2000) Dechaine & Wiltschko (2002a&b), and Ajiboye (2005a). 2 0  (43)  a.  Mo ri [awon ] [okunrin lsg see they man 'I saw them these three men.'  b.  meta yii]] three Dem  21  DP3 DPl"^ awon  ^DP2 okunrin meta y i i  In (43), awon and okunrin meta yii are said to refer to the same people, hence they are analyzed as separate DPs (DPI & DP2). Dryer's analysis correctly predicts that,strong pronoun, (44a) and weak pronoun (44b) can occur independently.  (44)  a.  Mo ri awon lsg see 3 pi. strong 'I saw them.'  b.  Mo ri won lsg see 3pl.weak 'I saw them.'  His analysis also incorrectly predicts that both strong pronoun (45a), and weak pronoun (45b), should be able to occur with nouns to mark them for plural. But this is only true of the strong pronoun. (45)  a.  Mo ri awon okunrin lsg see PL man 'I saw the men.'  This kind of nominal expression is actually found in some Southern dialects of English spoken in the United States as shown in (i). (i) a. I saw [them children] PL PL b. [Them dogs] are in the yard PL -PL (Wilson personal communication) The Standard form of (i) is rendered as (ii). (ii) a. I saw [the children] b. [The dogs] are in the yard Such dialects may be good for reconstruction on the 'plural agreement neutrality' that is witnessed today in Standard English whenever a plural noun takes the determiner the. (iii) a. the boy (singular) b. the boys (plural) z l  b.  *Mo lsg  ri see  won/won 3 pi. weak  okunrin man  The fact that w o n / w o n '3pl clitic pronoun' cannot occur with okunrin  'man' to mark it for  plural in (45b) casts doubt on Dryer's claim that awgn in (43) should be analyzed as an appositive. One way to tell whether (43) is a single DP or two DPs in apposition is the position of gbogbo. If (43) were to be two DPs in apposition, one would expect (46) to be possible such that gbogbo modifies the second DP.  (46)  *Mo ri lsg see  [awon] [3 PL]  [gbogbo [all  okunrin man  meta y i i ] three Dem]  What we have instead is (47), where gbogbo precedes the 3PL implying that gbogbo modifies the entire phrase.  (47)  M o ri [gbogbo lsg see [all T saw all these three men.'  awon okunrin PL man  meta y i i ] three Dem]  The fact that awgn cannot precede gbogbo is evidence that awgn is not a separate DP and as such, (43) cannot be analyzed as an appositive phrase in the language.  6.3.1.3 Internal structure of  awgn  There is one more question that I address, namely the morphology of 3PL awgn. Comparing the example from Sabee dialect with Standard Yoruba, it appears the 3pl pronoun has two forms awgn  versus wgn which might suggest that one is derived from the other  historically. 22  According to Mart/ (1992: 13), in Sabe, Idaiesa and Ketu dialects of Yoruba, the plural marker follows the noun it pluralizes. 2 2  (48)  a.  egi won tree 3 pi 'trees'  b.  awon i g i 3 pi tree 'tree'  Sabee  Standard Yoruba  The question is which one is the base and which one is derived? There are two hypotheses that I put forward.  (49)  HI:  awgn is the basic form and wgn is derived from it through a phonological process which deletes the first vowel.  H2:  wgn is the basic form and awgn is derived from it through a morphological process of prefixation that prefixes a- to wgn.  At first glance, both hypotheses look plausible and seem to have support from internal evidence. But I am going to argue in support of the second hypothesis. I consider the deletion option first. Deletion as a phonological process is robust and productive in Yoruba, and it occurs in many word formation processes. The examples in (50) show deletion involving two concatenated vowels at morpheme boundary where /a/ is the initial vowel of the second morpheme. What one observes is that it is the first vowel that deletes. (50)  a  Ije Nom eat 'enjoyment'  aye life  >  ijaye  b.  ,a-igbe Nom-NEG live 'act of not living'  aye > world  aigbaye  c.  Iko ade > N o m gather crown 'gathering crown.'  ikade  d.  Ini ade > Nom have crown 'having crown'  ilade  e.  I-  igbafe  gbo afe > Nom hear enjoyment 'loving socialization'  The fact that 'a' is not deleted in any of these examples provides a strong argument against the claim that won is derived from awgn through deletion of the initial a-. On the other hand, there is ample data to back up the claim that awgn is derived through a morphological process of prefixation. In fact there are many nouns that are derived by prefixing an affix to a verb including a-. Some examples are given in (51).  (51)  a.  ato Nom urinate 'semen'  >  ato  b.  alo Nom go ' departure/leaving'  >  alo  c.  abo Nom arrive 'arrival'  >  abo  d.  awa ye Nom drive survive 'driving safely'  awaye  e.  aro pin Nom think end 'consider hopeless'  aropin  In view of the fact that word formation with the a- prefix is possible and plausible, I adopt the second hypothesis in accounting for the morphology of awgn, claiming that it is derived from wgn through prefixation of a-. The alternative hypothesis, namely that wgn is derived from awgn through deletion of the initial vowel, is rejected because it posits a word 23  formation process (deletion of the initial vowel /a/) that is otherwise unattested in Yoruba. Finally, I also need to point out that the plural word awgn shares some morphological structure and process with demonstratives in the sense that that they can take "e -/I"  ' This is not say that V2 can never be deleted in Yoruba.  nominalizer prefix just as won takes "a-". This process derives demonstrative nouns. I show this in (52). (52)  a.  eyii Nom Dem 'this one' lyen Nom Dem 'that one'  Note that each of the demonstrative nouns in (52) consists of the base: yii and yen.  6.3.2  won- marks plural on demonstratives  This section discusses how demonstratives are used as plural words. First, I show that the base form of demonstratives is unmarked for plural. That is why the nouns they combine with is obligatorily interpreted as singular. (53)  a.  Mo ra ile yii lsg buy house Dem Thought this house.'  b.  Mo ta ile yen lsg sell house Dem 'I sold that house.'  Demonstratives are pluralized by prefixing the morpheme won- as in (54).24  Note also that whenever won combines with demonstratives to form plurals the derived word can in turn undergo another nominalization process by prefixing i-. This further supports the claim that awon is derived from won by prefixation of a(i) a. iwonyi 'Nom PL Dem 'these ones' b. iwonyen Nom PL Dem 'those ones' One can also claim that awon is derived from the plural prefix won by prefixing /a/ to the former. However such argument is not tenable considering the fact that the output *awgn is not attested. 2 4  n  (54)  a.  won-yn PL-Dem 'these'  b.  won yen PL Dem 'those'  Observe that when a demonstrative is unmarked and it takes a noun, the N P is obligatorily interpreted as singular (55a). To get a plural interpretation, the demonstrative (55b) must be marked for plural.  (55)  a.  Mo ri [aja lsg see dog ='I saw this dog.' saw these dogs.'  yii] DEM  NP  Mo ri [aja lsg see dog T saw these dogs.' *'I saw this dog.'  won-yii] PL-DEM  NP  DEM  PL-DEM  Recall that in the present analysis, demonstratives are heads and they take NPs as their complement. The N P moves to Spec D(em)P, which derives the surface linear order. The plural feature on the demonstrative percolates to the D(em)P and assigns the whole phrase its plural feature.  (56)  Plural percolation via demonstrative D(em)P NP / \ iwe  PL  ^D(em)  P L  Dem won  yii  l  NP  Finally, there is a common factor shared by the plural word a-wgn on nouns and the plural prefix won on demonstratives, namely the presence of L-tone on them:  (57)  a.  a-  won  b.  won- y i i  It is desirable to inquire if there is anything that L tone can tell us about plural marking in Yoruba. The questions that arise are the following. Is there something that this L tone can tell us about plural marking in Yoruba or is its presence a mere accident? Is there any difference between the L tone won that marks plural on demonstratives and the H toned pronouns won and the M toned won that are used as weak pronouns?  '  One possible analysis is to claim that like its M-tone and H-tone counterparts, won L tone is a pronoun. Then, the claim is that the tonal alternation is syntactically motivated: the pronoun bears H tone when in subject position (58a) and M when it functions as an object (58b). In contrast, when this pronoun functions as a plural morpheme it bears L tone (58c).  (58)  a.  Won ri mi 3pl see lsg 'They saw me.'  b.  Mo ri won lsg see 3pl 'I saw them.'  c.  Mo lsg  ri see  won- y i i PL- Dem  T saw these Xs.'  On the other hand, the L-tone w o n neither appears as a subject nor as an object NP:  (59)  a.  *Won  b.  PL *Mo lsg  ri see ri see  mi lsg won PL  In the same vein, the M / H toned form does not function as a plural morpheme: (60)  a.  *won - y i i  b.  *won y i i .  One argument against this claim is that the plural morpheme on demonstratives always bears L tone as opposed to H or M . Observe that, in contrast to won which may surface with H, M or L tone the 3pl strong pronoun awgn bears initial L tone irrespective of whether it functions as an independent pronoun (61a-b) or as a plural word (61c). (61)  a.  Awonri mi 3pl see lsg 'They saw me.'  b.  Mo ri awon lsg see 3pl 'I saw them.'  c.  Mo ri awon omo lsg see PL child 'I saw children.'  This is an area for further research.  6.3.3  Reduplication marks plural on modifiers  25  Another way by which Yoruba expresses plurality on its nouns is through the use of modifiers. A modifier-as-plural-word consists of a 'COPY-modifier'. Bare nouns with plain 26  Cross-linguistically, reduplication is extensively used to form plural forms. There are several variations on how reduplication operates, depending on the base form of the noun or the entity that is reduplicated for the purpose of marking plural. In any case, all of the reduplication processes create a reduplicant, a prefix whose phonetic content is from the segments that make up the base. I should say the use of "COPY" as a mechanism of marking plural is famous among the Yoruba people, as demonstrated in the early novels. The example in (i) is taken from one of the works of Fagunwa, a famous Yoruba novelist. The copied modifier nla in citation means 'big' but when used as a plural word as in (i), it means 'great Xs'. 2 6  modifiers display an ambiguity between a singular and plural interpretation. This is illustrated in all the (a) examples in (62)-(66). However, this ambiguity disappears once the modifier is copied. This is reflected in the (b) examples. (62)  (63)  a.  II u town = (i) = (ii)  b.  Ilu yii ni [asa buruku town D E M have custom COPY 'This town has bad customs.'  a.  Peju P.  = 0) = 0)  yii ni [asa buruku] D E M have custom bad 'This town has a bad custom.' 'This town has bad customs.'  [boolu pupa] ta ball' red sell 'Peju sold a red ball.' 'Peju sold red balls.'  Peju ta [boolu P. sell ball' 'Peju sold red balls.' (64)  (65)  (i)  buruku] bad  pupa pupa] COPY red  a.  Omole fo [igo palaba] O. wash bottle flat = (i) 'Omole washed a flat bottle.' = (ii) 'Omole washed flat bottles.'  b.  Qmole fo [igo palaba palaba] O. wash bottle COPY flat 'Omole washed flat bottles.'  a.  Mo lsg = (i) = (ii)  ra [ogede nla] buy banana big 'I bought a big banana.' T bought big bananas.'  Iwo ni O ni [agbara nla] 2sg FOC 2sg own power big ti O fi da gbogbo [nkan nla iila] inu aye... that 2sg use create all thing COPY big inside world 'You are the one that has a strong power that you used in creating all the great things in this world.' (Fagunwa 1961: 146) See Bamgbose (1967: 112-113); Pulleyblank (2002) and Ajiboye and Dechaine (2004) for various accounts of Yoruba reduplication.  (66)  b.  Mo ra [ogede lsg buy banana T bought big bananas.'  nla nla] COPY big  a.  Abiola ni [ile giga] n i Eko A. have house tall P Lagos = (i) 'Abiola has/owns a tall building in Lagos.' = (ii) 'Abiola has/owns tall buildings in Lagos.'  b.  Abiola ni [ile giga giga] n i A. have house COPY tall P 'Abiola has/owns high-rise buildings in Lagos.'  Eko Lagos  Following Orie and Pulleyblank (2002) and Ajiboye and Dechaine (2004), I assume that the copied entity is at the left edge of the base. Consequently, I adopt the structure in (67) for 27  the Yoruba copy-modifier. The plural marker is left adjoined to the modifier to form a plural modifier (cf. Kayne 1994). (67)  Mod PL COPY [nla  Mod nla] big  The analysis of modifiers is the same as that of demonstratives reported above. For completeness, I show how the mechanism of feature percolation derives plural NPs with modifiers using (65a) as an illustration. (68)  plural percolation through modifier NP PL  Mod  While this claim is not transparent in full reduplication as it is difficult to know which one is the base and which one is the COPY, evidence comes from partial copy, which derives gerunds. The fact that only (i-a), which has the COPY to its left is grammatical attests to this claim, (i) a. de di-de 'arrive' 'arrival' b. de *de-di  2 7  Yoruba speakers do not have the same judgments on which sub-classes of modifiers can be copied to mark plural. While B a m g b o s e (1967: 112-113) reports that quality modifiers (e.g. buruku 'bad') and quantity modifiers (e.g. nla 'big') can be copied to 'indicate more-fhan-one-ness', some other native speakers cannot process copying that involves colours (such as dudu 'black') or qualities (such as buruku 'bad') as plural words. For the purpose of this dissertation, I consider quality, dimension and colour as modifiers 28  that can undergo copying for plural marking. Some representative examples are given in (69). (69)  a.  aja buruku buruku dog bad bad 'dangerous dogs'  quality  b. .  aja nla dog big 'big dogs'  dimension  c.  aja dudu dudu dog > black black 'black dogs'  nla big  colour  Putting aside speaker variation, it remains that in Yoruba, not all modifiers are eligible for copying to mark plural. In particular the class of attributives (70a) , most ideophones (70b), 29  30  and locatives (72-73) cannot be copied to form plurals.  Observe that quantity modifiers (quantifiers and numerals) are treated as a kind of plural word with an abstract [PLURAL] feature. They therefore require no copying to function as plural words. When they are copied they modify verbs, (i-b). (i) a. Mo ra iwe pupo lsg buy book many 'I bought many books.' b. Mo ra pupo pupo lsg buy COPY many 'I bought in large quantity (i.e., the buying was done in large quantity).' Attributive as used here is a kind of modifier that describes or characterizes the mental state of the noun it modifies. This contrasts with the standard use of the term as any adjective, which appears directly beside the noun. These modifiers are attributives because they assign some kind of quality to the noun they modify. According to Doke in Awoyale (1974:139), an ideophone is a word, often onomatopoetic, which describes a qualificative, predicative, or an adverb with respect to sound, colour, smell, manner, state, action or intensity. Actually, there is a category of ideophones that can be copied to mark plural (cf. Beck 2005). 28  29  3 0  (70)  (71)  (72)  a.  *aja dog  olonbuniku COPY  olorfburuku unlucky  b.  *aja dog  jatijati jatijati COPY feckless  a.  eyin tooth 'left  osi left tooth'  b.  *eyfn osi  a.  apa otun arm right 'right arm'  b.  *apa  31  otun  osi •  otun  The reason why locative modifiers cannot undergo copying might be because 'left' and 'right' are unique. The next thing that I would like to discuss is the size of the copied item. Whenever a modifier is copied for the purpose of marking plural, it is the whole word that is copied. In most cases, full copying is subject to certain phonological constraints. In particular, O l a (1995), and Orie and Pulleyblank (2002) claim that what determines the size of what is to be copied is the principle of "foot binarity". I also observe that whenever a modifier is copied, it does not undergo any phonological process of either deletion or assimilation either at the segmental or the tonal  The category of ideophones that can be copied to mark plural in Yoruba is shown by the examples in (i), (cf. Beck 2005). (i) a. Mo ra igo palaba lsg buy bottle flat = 'I bought a flat bottle.' = ' I bought flat bottles.' b. Mo ra igo palaba palaba lsg buy bottle COPY flat 'I bought flat bottles.' 31  level. This may be because the kind of copying under discussion is syntactic or because of its phonology: the modifiers in question are all consonant initial. a.  nla big  >  nla nla COPY big  b.  pupa red  >  pupa pupa COPY red  c.  kekere > small  33  kekere kekere COPY small  One final question that I address is what happens when two or more modifiers modify a noun. In chapter 1, I have shown that Yoruba modifiers fall into four distinct semantic classes of colour, dimension, quality, and quantity. Also, in my discussion of overt plural marking with modifiers above, I have shown that the C O P Y of the modifier functions as plural word. There is a restriction on the copying process when there is more than one modifier in a nominal expression. The observation is that only a modifier that is adjacent to a noun can undergo copying. This claim is supported by the examples in (74) where there are two modifiers: nla 'big' and tuntun 'new' that modify ile 'house'. A s it turns out, only the first of the two, namely nla can be copied.  (74)  a.  ile nla nla house P L big 'new big houses'  b.  *ile  nla  tuntun new  tuntun tuntun  See Inkelas and Zoll (2000) and Pulleyblank (2002) among others for a discussion of phonological versus morphological reduplication. Observe that Yoruba also makes use of the partial copying strategy to derive gerunds. If modifiers were to be partially copied for the purpose of plural marking, the process would involve copying the first consonant of the modifier and the insertion of a fixed segment (the H toned HI). a. kekere *ki-kekere COPY-small small *di'-dudu b. dudu black COPY-black *ni-nla c. nla COPY-big big Unfortunately, partial copying does not apply to modifiers not to talk of using it for plural marking. I conclude that copying that involves plural marking must be full and not partial.  32  33  The second restriction is that only one modifier can be copied. So, in (75b) despite the fact that the adjacent modifier is copied thus satisfying adjacency constraint, it is not possible to extend copying to the next modifier.  (75)  a.  ile nla nla house P L big 'new big houses'  tuntun new  *ile nla house P L  tuntun tuntun PL new  nla big  Even when the noun itself is marked for plural such ungrammaticality or infelicity (as witnessed in (74) remains. This is what I show in (76)-(77).  (76)  a.  awon ile kekere kekere PLhouse small PL'new small houses'  b.  a.  awon PL-  (77)  a.  awon aja dudu d u d u kekere PLdog PL black small 'small black dogs'  b.  *awonaja  d u d u [kekere  kekere]  PL-  black P L -  small  ile kekere house small  dog  tuntun new tuntun tuntun PLnew  'small black dogs'  It appears there is more to be done before we can make a definite conclusion on specific restrictions guiding modifier copying when they are stacked within NPs in Yoruba. I address what prevents N copy for the purpose of plural marking next.  6.3.3.1 What prevents N copy: blocking  Cross-linguistically, copying can be used to mark a number of grammatical functions including plurality, distributivity, perfective aspect, continuous/progressive aspect, diminutives, augmentatives, intensification (Moravcsik 1978, Ghomeshi et al. 2004). The issue of why nouns are not copied in Yoruba for the purpose of plural marking is the concern of this section. I have shown that Yoruba makes use of copying modifier for this purpose. On the other hand, Mandarin copies classifiers for plural marking.  (78)  (yi) zhang -zhang zhi Mandarin one CL C L paper 'pieces of paper (with an emphasis on the individual pieces)' (Yang 2004: 1)  If Yoruba were to be like Mandarin we expect (79) where the classifier is copied. But such examples are ill-formed in Yoruba.  (79)  * awe awe sheet sheet  iwe paper  Yoruba  The question that arises is this: why is it that nouns cannot be copied to mark plurality on nouns in Yoruba? Observe that there are many languages that copy nouns for the purpose of plural marking. For example, Amele forms its plural on nouns by copying the nouns.  (80)  a.  dana 'man'  b.  dana-dana 'men'  34  Amele  (Corbett 2000: 136-137)  Similarly, in Malay full copying of nouns creates the indefinite plural nouns. This kind of copying is also very common in New Guinea (Cook personal communication). (81)  a.  bunga 'flower'  Malay  Amele is a Madang sub-group within the Trans New Guinea languages (Corbett 2000).  b.  bunga-biinga 'flowers'  (McShane and Nirenburg 2003: 17)  Again, Yoruba is not like Amele and Malay as shown by the ungrammaticality of (82) where the noun fails to copy. (82)  a.  okunrin 'man/men'  b.  *okunrin okunrin [men]  Yoruba  From the foregoing, it is clear that Yoruba contrasts with Mandarin on the one hand and Amele and Malay on the other as it neither copies classifiers nor nouns for the purpose of plural marking. This is puzzling, because Yoruba makes use of extensive nominal copying in other parts of the grammar (Awoyale 1974; Awobuluyi 1995; Bamgbose 1990; Pulleyblank 2000; Ajiboye et al 2003; Ajiboye & Dechaine 2004).  35  What this boils down to is that variation with respect to copying might be a language specific rule. In Yoruba such a rule will pick out modifiers i.e. copying is only available for plural marking if the base is a modifier. I treat the Yoruba case of using COPY for modifiers but not for nouns as an instance of 'The Emergence of the Unmarked (TETU) as I will discuss in the next subsection.  Despite the claim that copying of nouns is very robust, I observe that there are a lot of nouns that cannot be copied on their own. (i) a. ile *ileele *iliile 'house' b. owo *ow66wo *ovvoowo 'money' For those nouns to be copied, it must be aimed at deriving negative polarity items or 'any bad X' in which case there is an infixation after copying (Koch 2004, 2005, Ajiboye 2004). (ii) a. ile ilekile 'house' 'any house/bad house' b. owo owokowo 'money' 'any money/ill-gotten money' Like modifiers, what determines the type of nouns that can be copied needs further research.  6.3.3.2 The emergence of the unmarked  There appear to be two cases of noun copy that we can talk about across languages. First, there is copying of nouns to mark plural, this I call the general case since many languages make use of this device. The other copying of nouns for more specified functions other than plural marking; this I call the restricted case. This is the category that Yoruba belongs.  (83)  a.  Case I: (General) e.g. Amele [a ] [a ] => pluralization N  b.  N  Case II (Restricted) e.g. Yoruba [a ] [a ] => quantification N  N  What we already saw is that Yoruba does not use the general case. On the restricted Case, whenever nouns are reduplicated in Yoruba, they have the specific function of distributive quantification with numerals (84) or temporal expressions (85).  (84)  (85)  a.  a.  eji => 'two'  eji two  two  eje 'seven'  eje eje seven seven  ale => 'night'  ale night  ale night  ose => 'week'  ose week  ose week  =>  ejeeji 'all two' ejeeje 'all seven'  => =>  alaale 'every night osoose 'every week >36  I claim that because N-copy has a dedicated function in Yoruba, it cannot be interpreted as pluralization. The generalization is that the more specified meaning (i.e. distributive quantification of numerals and temporal nouns) blocks the more general meaning (i.e. pluralization of N). As a last resort, Yoruba has to adopt the marked form of plural marking  For details, see (Ajiboye & Dechaine 2004).  i.e. the use of C O P Y modifier. This is what I call The Emergence of the Unmarked (TETU).  37  Observe that in many languages, whenever modifiers are reduplicated they have a degree reading. This is the general case. English is a language that adopts this device. The restricted case (i.e. Case II), which Yoruba adopts, is using the copy form to mark plural.  (86)  a.  Case I:  Degree reading e.g. L  tt  Mod] [ ° - Modi  a tall tall 'a very tall man' b.  English  man  Case II:  Plural marking e.g. Yoruba)  okunrin man  giga giga tall tall 'tall men'  =>  TETU  [ ° ; Mod] [ ° ; Mod]  Adapting this to plural marking makes the prediction that if a language fails to use COPY (i.e. reduplication) on N for the purpose of marking plural, then that language should use the marked form i.e. COPY on modifier if it has a reduplication process.  6.3.4  38  Multiple plural marking  As mentioned earlier, there are two syntactic positions available for plural marking within Yoruba nominal expressions the noun or the modifying elements (quantifiers/numerals, demonstratives and modifiers). What I have accounted for so far are cases involving one of the two positions at a time. In this section, attention is going to be on cases where both syntactic positions are marked for plural. This is what I call multiple plural marking. First, I present the data. In (88a-d) the noun aja 'dog' is marked for plural by awon. In addition the The idea of the emergence of the unmarked (TETU) is not novel in Phonology literature (McCarthy and Prince 1993, 1994 and in syntax too (Bermudez-Otero & Borjars 2002). In Optimality Theory (OT), a constraint can be active even if it is crucially dominated. A low-ranking markedness constraint, then, can decide between candidates, as long as they tie on all higher-ranking constraints. The linguistic structure that is unmarked with respect to this constraint can emerge in such circumstances (McCarthy and Prince 1994).  3 7  demonstrative is marked for plural with the prefix won- (87b), the modifier by its COPY (87c), and in (87d), each entity within the whole nominal expression is marked for plural by the appropriate plural morpheme.  (87)  a.  awon aja PL dog 'seven dogs'  meje seven  b.  awon aja PL dog 'these dogs'  won-yi Dem-PL  c.  awon aja PL' dog 'blackdogs'  dudu dudu COPY black  d.  awon aja dudu dudu meje won-yi PL dog COPY black seven Dem-PL 'these seven black dogs'  As with the account of cases where just one member of a nominal expression is marked for plural, these multiply marked NPs also involve percolation of the P L U R A L feature. I demonstrate this in (88). In (88a), both the plural feature of awgn and the abstract [ P L U R A L ] feature in meje percolate their plural feature onto the NP. In (88b), plural percolation is realized on the NP through the plural word awgn and the prefix won of the demonstrative. In (88c), plural percolation comes via awgn and the copied modifier dudu. Lastly, in (86d), plural percolation comes via awgn, the copied modifier dudu, the abstract [PLURAL] feature from meje and the prefix won of the demonstrative. The different arrows show the direction of the percolation.  Hovvever, this prediction is not true cross-linguistically as there are languages (e.g. Halkomelem), which marks plural on nouns with plural words that are not copied.  38  (88)  a.  plural percolation through noun & Num  Num NP meje / X [PLURAL' aja 'seven 'dog' 'seven dogs'  awon  plural-percolation through noun & Dem ^ 7 D(em)P PL  C^T^PL | awon  Dem NP 71 ^ PL Dem aja [ | 'dog' w o n yi 'this' 'these dogs' P  plural-percolation through noun & Mod NP PL  awon  aja dudu 'dog' 'black dogs'  dudu 'black'  d.  plural-percolation through noun, Mod, Num & Dem D(em)P ^ PL  Modp  ^ PL Num [ ^ | won aja P L Mod meje 'dog' | | [PL] dudu dudii 'seven' 'black' 'these seven black dogs' J p L  awon  L  Dem ^ yi 'this'  Plural construals on nouns in Yoruba divide into three interpretive classes as summarized in (89). Class 1 involves cases where an N P is not specified for number, in that context, the N P is interpreted as singular or plural. Class 2 involves cases whereby a noun can be accompanied by either an unmarked demonstrative or the numeral "one" and the nouns are obligatorily interpreted as singular. Finally, class 3 divides into two sub-groups. On the one hand are cases where bare nouns are either accompanied by a numeral that is greater than "one" or a quantifier. On the other hand are cases where a dedicated plural morpheme occurs with N . In all of these, the NPs are obligatorily interpreted as plural. (89)  A summary of the analysis C L A S S Syntactic context 1 N N + Mod N + Dem 2 ; N + Num 1 N + Num >1 3 (i) N +Q P L - N + Dem/Mod (ii) N + PL-Dem/PL-Mod P L - N + PL-Dem/PL-Mod  Interpretation S I N G U L A R OR P L U R A L SINGULAR PLURAL  This catalogues the various ways in which Yoruba nouns come to be interpreted as singular or plural. What remains to be discussed is how the feature percolation mechanism adopted in  the account for Yoruba contrasts with feature matching. After laying out feature matching mechanism, I use English as a case study.  6.3.5  Feature percolation versus feature matching  Feature matching is a mechanism whereby plurality is marked on all members of a nominal expression. In a language like English, which makes use of this mechanism, the way it works is as follows. Let 6 be the NP and a and |3 be constituents under 8. If 8 is plural, then both a and p must be plural (cf. Yang 2004, Wiltschko 2004). (90)  8  PLURAL  °-PLURAL .  ^PLURAL  There are four environments that can be examined in English to illustrate feature matching, as shown in (91).  (91)  a.  noun demonstrative  b.  noun quantifier  c.  noun numeral  d.  noun determiner  I look at only one of the cases, namely noun-demonstrative since it presents an interesting contrast with the facts of Yoruba already accounted for. If a noun in English takes a demonstrative, both must match with respect to number. When the noun is singular it can only take a singular demonstrative, (92a). On the other hand, a singular noun cannot take a plural demonstrative, (92b).  (92)  a. b.  this book *this book-s  Further, when the demonstrative is in the plural form, the noun and the demonstrative must  agree with respect to plural marking (93).  (93)  a. b.  these books * these book  The effect that we see in (93) is referred to as concord or agreement in the literature (Corbett 2000, Wiltschko 2004, 2005; Yang 2004 among others). In the current account I analyze English plural marking as an instance of feature matching. This means all members of a nominal expression must have the same number feature. The facts of (92) and (93) are illustrated in (94) and (95). In (94a), the demonstrative this and the noun book have the same number feature in that they both share the singular feature hence feature-matching is observed. However, in (94b) where the singular demonstrative this combines with the plural noun books, feature-matching is not observed and this leads to ungrammaticality. (94)  a.  DP  [ S G ]  D[SG]  NP  this  book  b.  *DP  [ S G ]  [ P L ]  D[SG]  NP  (PL  ]  I this  book-s  Similarly, in (95a), the demonstrative these and the noun books observe feature matching in that they both share the plural feature. On the other hand, in (95b) where the plural demonstrative combines with the singular noun, it violates feature-matching, hence the ungrammatical output. (95)  a.  DP,[PL] D  [PL]  these  NP  [ P L ]  book-s  [SG]  NP, [PL]  D(SG]  these  boy  Crucially, Yoruba differs from English in that nouns need not agree with demonstratives in terms of plural marking. As shown in (96b), only the noun is marked for plural by awon whereas the demonstrative remains unmarked and the whole N P is interpreted as plural. In the same way^ in (97b), only the demonstrative and not the noun is marked for plural by won and the output is well-formed. (96)  (97)  a.  aja yn dog DEM 'this dog.'  b.  awon aja PL dog 'these dogs.'  yu  awon aja PL dog 'these dogs.  won- y DEM PL  a.  NP  DEM  PL-NP  DEM  PL-NP  PL-DEM  NP  PL-DEM  DEM  (cf. *this dogs) 1  won- yi aja PL DEM dog 'these dogs. (cf. * these dog)  The disparity between English and Yoruba is straightforwardly accounted for by assuming that while Yoruba makes use of feature percolation, English adopts feature matching. The table below further illustrates the similarities and differences between the two languages in relation to how demonstratives and nouns interact. (98)  Plural marking on noun-demonstrative in Yoruba & English Yoruba a. DEM +N v• DEM +N b. v c. DEM + N L V d. DEMp + N V  English  V  P L  P L  P  V  L  In conclusion, I have shown that feature percolation and feature matching yield  different results with respect to the typology of number marking. When a language adopts the feature percolation mechanism, marking plural on all the entities within an N P is optional. This is what Yoruba does. By contrast when a language makes use of the feature matching mechanism, it becomes obligatory for all members of an N P to agree with respect to the plural feature. This is what English does. This analysis thus explains why Yoruba but not English, can have phrases such as 'this dogs' and 'these dog'.  6.4  Comparing analyses  I review two different types of previous analyses of plural marking and assess them with respect to the Yoruba data that I have analyzed. These are the plural parameter proposed in Deprez (2004) and inflectional versus modificational plural marking proposed in Wiltschko (2004, 2005).  6.4.1  The plural parameter (Deprez 2004)  Deprez's (2004) Plural Parameter, which is a development of an earlier proposal (Deprez 1999, 2001) using a morphosyntactic parameter, distinguishes two broad sets of languages: [+PL] languages and [-PL] languages. According to that approach, languages are grouped based on whether or not the structure of their nominal expressions obligatorily includes a functional projection for number. Deprez's approach treats all languages that do not have inflectional morphology for plural as [-PL]. On this view, in [+PL] languages, number must project whereas in [-PL] languages, "the projection of NumP is optional and when it occurs, NumP does not have to contain a counter" (Deprez 2004: 10). The consequence of this proposal according to Deprez is three-fold:  (99)  a.  Only [-PL] languages allow direct access to the basic kind denotation of nouns. For [+PL] languages, access to the kind denotation requires the presence of a relevant operator.  b.  Only [-PL] languages can have bare nominals that are under-specified for number and thus compatible with either a plural or a singular construal depending on contextual factors.  c.  In [+PL] languages, in the absence of relevant morphology, bare 'singular' nominals, i.e., NumPs that contain an unsaturated counter, are excluded. (Deprez 2004: 13).  Going by Deprez's classification, Yoruba is a [-PL] language. Her claim that bare nominals in such languages are under-specified for number and thus compatible with either a plural or a singular construal depending on contextual factors is in line with the facts of Yoruba presented in §6.1. Just to remind readers, consider (100) where the bare noun erm can be construed as either singular or plural.  (100) M o ri erin ni ogba eranko lsg see elephant garden animal ='I saw an elephant in the zoo.' = 'I saw elephants in the zoo.' However, despite the fact that Deprez claims that in a [-PL] language the projection of NumP is optional, her analysis makes no predictions regarding the actual deployment of P L U R A L features in a [-PL] language. As I demonstrated in §6.1-6.3, [-PL] languages like Yoruba use feature percolation to mark plural. This differentiates Yoruba from [+PL] languages such as English where P L U R A L must satisfy feature matching (i.e. agreement).  6.4.2  Plural marking as a functional head or as a modifier (Wiltschko 2004,2005)  The other account of plural marking that I examine is one that draws a distinction between plural marking as a functional head or via modification. M y particular focus is on Wiltschko (2004, 2005).  6.4.2.1 Plural marking as a functional head  Using English as an example of a language where plural is inflectional Wiltschko (2004, 2005) argues that in languages with plural marking as a functional head, the functional head takes a noun as its complement as in (101).  (101) [  numC1  [Pmral]  numCI  N  N]]  (Wiltschko 2005: 5)  In particular, she claims that plural marking is associated with the functional head num(eral) Cl(assifier). The other significant thing about languages of this type is that number marking requires that there be number agreement on all the elements within a nominal phrase.  39  The  following examples illustrate this idea. (102) Obligatory plural marking in English phrases a. the two men b. *fhe two man (103) Structure of English pluralized nouns numCl numCL  n n ^  ^root  (Wiltschko 2005: 12)  Arguing for (101), Wiltschko (2005:12) claims that plural marking in a language like English "instantiates a functional head (numCl) and is therefore obligatory due to a so called (Exceptional Projection Principle) EPP-effect." She goes further by claiming that the category of the plural marker (numCl) projects and is thus syntactically transparent. This is what accounts for the obligatory agreement of number.  Cf. Corbett (2000) who observes that there are two types of agreement that exist between the controller (noun) and the target modifiers, which plural form is determined by the noun. The two are syntactic and  3 9  6.4.2.2 Plural marking as a modifier  For Wiltschko, modificational plural marking arises when plural marking is optional. In Halkomelem, where plural marking is not inflectional, the plural marker modifies roots. Wiltschko proposes that plural, as a modifier, is adjoined to category neutral root (Wiltschko 2005: 11). This idea is illustrated in (104).  (104)  Structure of Halkomelem pluralized nouns: n n  root plural^^^v/root  (Wiltschko 2005: 12)  This structure accounts for the data in (105) and (106). First, observe that plural marking on the root is optional and a noun that is marked for plural is compatible with a preceding plural marked determiner (105a). However, the noun does not require the determiner to be marked for plural before it is interpreted as plural, (106b).  (105)  (106)  Optional plural marking in Halkomelem phrases (a)  t'flem ye si:wi:qe sing det.pl man-pl 'The men are singing.'  (b)  t'flem te si:wi:qe sing det man-pl 'The men are singing.'  (a)  t'flem ye swiyeqe sing det.pl man 'The men are singing.'  (b)  t'flem te swiyeqe sing det man 'The man is singing.'  (Wiltschko 2005: 3)  semantic agreement. Corbett's claim is that the form determines the syntactic agreement while semantic agreement is determined by the meaning (p. 178).  One significant thing about "plural marking as a modifier" is that it is not subject to the Exceptional Projection Principle (EPP) as it is with "plural marking as a functional head". The structural difference between plural marking as a functional head or as a modifier is given as (107). The structure in (107a) illustrates English type of language where X is the Number (the head) and Y is the noun (the complement). In contrast, (107b) illustrates Halkomelem type of language X as a modifier does not project and as such it only adjoins to Y (the noun).  (107) a.  XP  English type  X ^ ^ ^ ^ Y P PLURAL b.  YP X  Halkomelem type  ' ^ ^ ^ Y P  PLURAL  (adapted from Wiltschko 2005: 6)  Concluding, Wiltschko claims "the obligatoriness of Number-marking in English is consistent with its inflectional status whereas the optionality of plural marking in Halkomelem suggests that we are not dealing with an inflectional category" (Wiltschko 2004:3,2005:11). Of the two mechanisms discussed by Wiltschko, the plural marking as a modifier approach easily extends to Yoruba: Yoruba is not an inflectional language therefore its number marking cannot project to a functional phrase. So Wiltschko's proposal should be able to account for Yoruba data. This account is consistent with my account, which treats plural words as adjuncts. However, Wiltschko's account falls silent regarding the actual deployment of plural morphemes in nominal expressions. I have shown that there is a way by which both inflection-based (syntactic) and modification based (semantic) analyses converge using feature-based mechanisms. Under this approach, the inflection-based analysis is accounted for using the feature matching mechanism, which requires that there be feature agreement within all members of a nominal expression. The modification-based approach is  analyzed as feature-copying where the plural feature is copied from a lower node to a higher node (feature percolation). One advantage of my proposal is that it does not really matter whether or not the plural word is a functional head. What determines plural interpretation is the "copy" mechanism involved. When plural is morphologically marked, the feature matching mechanism as expressed by the notion of agreement is applicable. On the other hand, when plural words are used, feature percolation is enforced.  6.5  Conclusion  This chapter has accounted for the various strategies that Yoruba adopts in plural marking. The general picture that emerges is that plural can be marked in one of three ways: contextually, semantically, or morphologically. I have shown that when a noun occurs by itself or when it takes a modifier, its construal as singular or plural is sensitive to context. Secondly, I have analyzed quantifiers and numerals as lexical items with an abstract P L U R A L feature, which percolates onto the NP. This is what I describe as an instance of semantically conditioned plurality. The third claim is that plural is also marked by certain morphemes, which can be realized either on the noun (in the form of awgn) or on a modifying element (in the form of reduplication) or on demonstratives (in the form of wgn-). The analysis of plural marking proposed for Yoruba makes a prediction that there are two ways by which languages may mark their nouns for plural. Languages that show agreement will use plural feature matching while languages that do not show agreement will use plural feature percolation.  C H A P T E R S E V E N : CONCLUSION  7  Findings and open questions  In this concluding chapter, I summarize the results of the research and the issues that need further attention. This thesis has focused on three areas of Yoruba nominal expressions: possessives, different ways by which argument expressions acquire discourse properties, and different strategies of marking plural. The following findings and open questions emerge.  7.1  Findings  I have proposed that Yoruba verbal and nominal genitives derive from the same base structure (a vP shell) and that the surface structure of both involves some kind of movement. While the possessum moves to Spec DP/CP in nominal genitives it is the possessor that moves to Spec IP in case of verbal possessives. I have attributed this movement as Case related. I also established that the M tone syllable that occurs between the possessum and the possessor syntactically is a relational element that shows that the two arguments are in a genitive relation. I argued that this MTS heads the genitive DP. Comparing the account of nominal genitives to verbal possessives, I established that what the MTS is to DP is what the HTS is to IP. I also proposed that the genitive marker is a mora y, that acquires its segmental content through assimilation of the last vowel of the possessum NP. Extending this further to more complex structures involving ^'-constructions, I argued that there is a link between the " M tone ti" in reduced clauses within a genitive DP and " H tone tf in full relative clauses. In particular, I analyzed both M-tone ti and H-tone ti as complementizers. I proposed that there is a syntactic relationship between the MTS and ti claiming that both are functionalheads with MTS spelling out D and M-tone ti spelling out C. More importantly, when these two elements co-occur, they do not only doubly mark genitive Case on the possessor, they also perform an additional function of emphasizing the possessor. I finally showed that the two elements differ in the sense that MTS is supplied by  phonology whereas ti is supplied in the lexicon. In the second part of the dissertation, I showed that bare nouns can be construed in one of three ways: as generic, indefinite or definite. I established that the generic reading arises in one of two ways; it can be lexically conditioned (with permanent-state verbs) or grammatically conditioned (through the use of imperfective maa-n). In each case, there is a G E N operator that takes scope over these bare nouns. I extend this proposal to English and suggest that genericity cross-linguistically can be accounted for using the grammatically conditioned approach. Thus, either a language makes use of an aspectual marker to mark genericity overtly as in Yoruba, or covertly as in English. I also established that, in Yoruba, in all contexts where a generic interpretation is not possible, bare nouns are construed as indefinite. This is what I refer to as the elsewhere condition. I concluded that generic and indefinite construals are in complementary distribution. I also established that, with the right discourse contexts, bare nouns could be construed as definite irrespective of their syntactic position. The three possible interpretations of bare nouns attested in Yoruba have afforded us the opportunity to provide syntactic structures for these bare nouns. While generic and indefinite bare nouns are analyzed as NP, definite bare nouns are analyzed as DP with a null D since they are always discourse linked. In all, there are four types of Ds in Yoruba: These are the deictic D kan, the demonstrative D (both of which are in complementary distribution with one another), the null D that is definite, and the mora pi D found in genitive constructions. In my account of how nouns are marked for specificity and salience, I claimed that specificity is morphologically marked on NPs in Yoruba with kan while salience is morphologically marked on nouns with nda. The findings here shed some light on the relation between the definite/indefinite contrast and the marking of specificity and saliency across languages. Thus, in environments where we expect overt marking, i.e., with indefinites and definites as in English and Gungbe, we have bare nouns in Yoruba. And in environments where we do not expect overt marking i.e. specificity and saliency as in English, we get kan and nda in Yoruba and dje in Gungbe. Regarding pluralization, I demonstrated that plural marking is optional in Yoruba but that when nouns are marked for plural, there are three different ways in which this is carried  out. First is through contextually determined plurality. These are cases where there is no overt plural marking: as such, a noun can be interpreted as singular or plural. The second strategy is semantically determined plurality. These are the cases where nouns take quantifiers and numerals (both of which I claim have an abstract [ P L U R A L ] feature). The third strategy is morphologically determined plurality. This is the strategy where nouns are overtly marked by plural words. I also established that there are two syntactic positions for morphologically determined plurality either on nouns (using the plural word awon) or on non-nouns (copy-modifiers or the prefix won-). The feature percolation and feature matching mechanisms that I proposed shed light on the fact that plural marking converges cross-linguistically irrespective of whether the element that marks plural is a lexical or a functional head. By this analysis, I have shown that all languages are sensitive to feature matching. Where a language fails to utilize this, it results in feature percolation, which has the status of an elsewhere condition.  7.2  Open questions  There are a number of open questions that future research needs to address. One is the controversy surrounding the existence of the M T S as a genitive marker and its overt realization in the environment of a vowel-initial possessors as well as the L deletion on the final vowel of the possessum. It would be very helpful to carry out some instrumental study especially as concerns whether the forms with the MTS before a vowel-initial possessor are actually identical to a V V sequence or whether they are slightly longer. This would (in)validate the claim that I made here to the effect that the contracted form is significantly shorter than the non-contracted form.. Another area of further research will be a comparative study of the syntax and semantics of genitive and appositive constructions. This is necessary concerning the claim in this study that the M-tone mora pi found between the possessum and the possessor is a genitive marker. Most appositives are true appositives in the sense that they involve a juxtaposition of two nouns that refer to one object without any intervening segment.  (1)  a.  Mo mo Sulu Qba lsg know S. king 'I know Sulu, the king'  b.  Awa Yoruba korira i w o s i lpl Y. hate insult 'We Yoruba hate insults.'  But we also notice that there is some kind of appositives in which, this MTS mora shows up.  (2)  a.  Ade wa ni [ana an Jimo] A. come P yesterday M T S Friday 'Ade came yesterday, which was Friday'  b.  [Eni in Satide] ni ojo ibf i today M T S S. FOC day birth Gen 'Today, which is Saturday is my birth day.' (adapted from Awobuliiyl 2004: 12)  mi lsg  One of the problems confronting the raising analysis in genitive constructions that I discussed in this dissertation is with respect to its link with Case assignment. This is another area where we have not been able to find empirical evidence in support of the claim. It becomes interesting when one considers the fact this is not a problem peculiar to Yoruba. Another problem relating to Case assignment is with respect to the position of one of the Case-assigners that are stacked in genitive constructions. The current proposal that the MTS as a member of stacked Case-assigners is capable of assigning Case to the possessor is stipulative since the MTS is not in the right configuration for Case assignment. There is the need to pursue with vigor further research in this area in future. There also arises an open question on why interpreting subject bare nouns as generic with permanent-state verbs depends on the gehericity of the object bare noun. In the same way, there is a puzzle on the asymmetry that exists between the subject and object bare nouns of permanent states: why a subject bare noun is obligatorily interpreted as indefinite, an object bare noun is obligatorily interpreted as generic. We equally need a precise syntax and semantics of salience marking on definite nouns than is currently understood. In particular it is desirable to be able to know whether this marker modifies the functional head or the whole DP.  Finally, the parity that exists between count and mass nouns in Yoruba with respect to how plural is marked on them calls for more study, as this will have some cross-linguistic consequences for our understanding of the count/mass distinction (cf. Chierchia 2005).  REFERENCES  Abney, Stephen Paul. 1987. The English Noun Phrase in Its Sentential Aspect. Doctoral dissertation. MIT. Aboh, Enoch Olabode. 1999. From the Syntax of Gungbe to the Grammar of Gbe. Doctoral dissertation. University of Geneva. Aboh, Enoch Olabode. 2004. The Morphosyntax of Complement-Head Sequences: Clause Structure and Word Order Patterns in Kwa. New York: Oxford University Press. Aboh, Enoch olabode. 2005. Clause structure and word order in Kwa. L O T Winter school, University of Gronigen. Abraham, R.C. 1958. Dictionary of Modern Yoruba. University of London Press. Adesola, Oluseye. 2001. A-bar Dependency in Logophoric Context. Ms. Rutgers State University, New Jersey. Adewole, Lawrence  Olufemi.  1989. Some Yoruba quantifier words and semantic  interpretation: a critique. Studies in African Linguistics. 20:79-88. Ajiboye, Oladiipo. 1991. Atunyewo fonoloji Mpba. Moba phonology revisited. Master thesis. University of Ilorin, Ilorin, Nigeria. Ajiboye, Oladiipo. 2001. The internal structure of Yoruba DP. 32 Annual conference of nd  African Linguistics. University of California, Berkeley. Ajiboye, Oladiipo. 2004a. The syntax and semantics of Yoruba Bare Nouns. Department of Linguistics seminar series. University of British Columbia. Ajiboye, Oladiipo. 2004b. The syntax of Yoruba genitive construction. 35  th  Annual  Conference of African Linguistics. Harvard University, Boston. Ajiboye, Oladiipo. 2005a. The syntax of Yoruba reflexive epithets. Journal of West African Languages. 31:89-122. Ajiboye, Oladiipo. 2005b. Interpreting Bare nouns in Yoruba. Berkeley Linguistics Society 31. University of California, Berkeley. Ajiboye, Oladiipo. 2005c. Aspects of plural marking in Yoruba. 36 Annual Conference of th  African Linguistics. Georgia Southern University. Savannah . Ajiboye, Oladiipo, Rose-Marie Dechaine and Osamuyi Stewart. 2003. The syntax of  nominalization in Edo and Yoruba. In New Trends in African Linguistics 6:1-15. Ajiboye, Oladiipo and Rose-Marie Dechaine. 2004. The syntax and semantics of Yoruba duplicative constructions: SS^Annual conference of African Linguistics. Harvard University, Boston. Ajiboye, Oladiipo. Rose-Marie Dechaine, Bryan Gick and Douglas Pulleyblank. 2004. Disambiguating Yoruba tones: the interface between syntax, morphology, phonology and phonetics. Ms. University of British Columbia. Akinlabi, Akinbiyi. 1985. Tonal underspecification and Yoruba tone. Doctoral dissertation, University of Ibadan, Nigeria. Akinlabi. Akinbiyi and Mark Liberman. 2000. The tonal phonology of Yoruba clitics. In B. Gerlach and J. Grijzenhout, eds. Clitics  in Phonology, Morphology  and Syntax.  Amsterdam: John Benjamins. Alexiadou, Artemis and Chris Wilder. 1998. Adjectival Modification and Multiple Determiners. In Alexiadou, Artemis and Chris Wilder, eds. Possessors, predicates and movement in the Determiner Phrase. Alexiadou, Artemis and Chris Wilder, eds. Amsterdam: John Benjamins Publishing Company, 303-332. Alexiadou, Artemis and Chris Wilder, eds. 1998. Possessors, predicates and movement in the Determiner Phrase. Amsterdam: John Benjamins Publishing Company. Awobuluyi, Oladele. 1978. Essentials of Yoruba grammar. Ibadan: Oxford Press. Awobuluyi, Oladele. 1995. Mofoloji Ede Yoruba. Lecture notes, University of Ilorin, Ilorin, 1-25. Awobuluyi, Oladele. 2004. On the so-called Genitive Morpheme in Standard Yoruba, read at the 24th W A C L , University of Ibadan, Nigeria. August 1-6. Awoyale, Y i w o l a . 1974. Studies in the syntax and semantics of Yoruba nominalizations. Doctoral dissertation. University of Illinois, Urbana-Champaign. Awoyale, Yiwc-la. 1985. Focus as an unbounded movement rule in Yoruba. Journal of the Linguistic Association of Nigeria 3: 745-84. Bach, Emmon. 1981. On time, tense and aspect: A n essay in English metaphysics. In P. Cole, ed. Radical pragmatics. 63-81. New York: Academic Press Bamgbose, A y o . 1966. A Grammar of Yoriibd. London: Cambridge University Press. Bamgbose, A y o . 1967. A Short Yoriibd Grammar. London: Cambridge University Press.  Bamgbose, A y o . 1986. Yoruba: a language in transition. Lagos: J.F. Odunjo memorial lectures organizing committee. Bamgbose, A y o . 1989. When rules fail: the pragmatics of vowel elision in Yoruba. Linguistics Association of Nigeria conference, Jos. Bamgbose, A y o . 1990. Fonglgji ati Girama Yoruba. Ibadan: University Press. Bamgbose, A y o . 2000. Coreferentiality and focus in textual cohesion: evidence from Yoruba. Journal of West African Languages. 28(2):59-70. Barczak, Leszek. 2004. Syntax and semantics of the Yoruba auxiliary mda. 34 Annual th  Conference of African Linguistics. Harvard University, Boston. Beck, David. 2005. Ideophones and adverbs as predicate-qualifiers. W S C L A 10. Benveniste, Emil. 1971. The nominal sentence. In Problems in General Linguistics. 132-144 (translated from French by M . E . Meed). Coral Gables, Florida: University of Miami Press [originally in Bulletin de la Societe de Linguistique de Paris 46 (1950): 19-36. Bermudez-Otero Ricardo & Kersti Bbrjars 2002. Markedness in phonology and in syntax: the problem of grounding. Bernstein, Judy. 1991. DPs in Walloon: evidence for parametric variation in nominal head movement. Probus 3:101 -126. Bianchi, Valentina. 2000. The raising analysis of relative clauses: a reply to Borsley. Linguistic Inquiry 31:123-140. Bisang Walter and Remi Sonaiya. 2000. Information structuring in Yoruba. Linguistics 38: 169-197. Bittner Maria and Ken Hale. 1995. The structural determination of Case and agreement. Linguistic Inquiry 18:1-44 Borer, Hagit. 2004. In name only: an Exo-skeletal triology Book 1. Ms. University of Southern California Borer, Hagit. 2001. One is the loneliest number. Handout, University of Southern California. Bowers, J. 1993. The syntax and semantics of nominals. In Proceedings of SALT-III3A-30. Borsley, Roberts D. 1997. Relative clauses and the theory of phrase structure. Linguistic Inquiry 28:624-647'. Borsley, Roberts D. 2001. More on the raising analysis of relative clauses. Ms. University of Essex.  Burton, S. 1995. Six issues to consider when choosing a husband. Doctoral dissertation, Rutgers University. Burzio, L. 1986. Italian Syntax. A Government-Binding Approach. Hingham, M A : Kluwer. Cardinaletti, Anna. 1998. On the deficient/strong opposition in possessive systems. In Alexiadou, Artemis and Chris Wilder, eds. Possessors, predicates and movement in the Determiner Phrase. Alexiadou, Artemis and Chris Wilder, eds. Amsterdam: John Benjamins Publishing Company, 17-54. Carlson, Gregory N . 1977. A Unified analysis of the English bare plural, Linguistic and Philosophy 3:413-458. Carlson, Gregory N . 1979. Generic and atemporal when. Linguistic and Philosophy 3:49-98 Carlson, Gregory N . 1980. Reference to Kinds in English. Doctoral dissertation, University of Massachusetts, Arnhrest. New York: Garland Press. Carlson, Gregory. 1982. Generic terms and generic sentences. Journal of Philosophical Logic 11:145-181. Carlson, Gregory N and Francis Jeffry Pelletier 1995, eds. The generic book. Chicago: The University of Chicago Press. Carsten, Vicki. 1986. Proper Government in Yoruba. University of California, Los Angeles. M . A . thesis. Carsten, Vicki. 2003. Rethinking complementizer agreement: Agree with a Case-checked goal. Linguistic Inquiry 34:393-412. Cheng, Lisa Lai-Shen and Rint Sybesma 1999. Bare and not-so-bare nouns and the structure of NP. Linguistic Inquiry 30:509-542. Chierchia, Gennaro. 1995. Individual level predicates as inherent generics. In The Generic Book. G . Carlson and F. Pelletier (eds.), 176-223. Chierchia, Gennaro. 1998. Plurality of mass nouns and the notion of "semantic parameter". In Susan Rothstein (ed.).  Evenst and Grammar. Netherlands: Kluwer Academic  Publishers, 53-103. Chierchia, Gennaro. 2005. Substance, object and number. Harvard University, Boston. Chomsky Noam and Morris Halle. 1968. The Sound Pattern of English. Chomsky Noam, 1981. Lectures on Government and Binding. Dordrecht: Foris Publications. Chomsky, Noam. 1991. Some notes on economy of derivation and representation. In  Principles and parameters in comparative grammar, ed. Robert Freidin, 417-454. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1993. A Minimalist Program for Linguistic Theory. In K . Hale and S.J. Keyser, (eds.), The View from building 20: Essays in Linguistics in Honor of Sylvain Bromberger, 1-52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1995. The minimalist Program. Cambridge, Mass.: MIT Press. Chumbo, Beban Sammy. 1982. Contraction and tone Polarization in Ogori. Journal of West Africa Languages 12:89-103. Cinque, Guglielmo. 1990. Ergative Adjectives and the Lexicalist Hypothesis Natural Language and Linguistic Theory 8:1-39. Cinque, Guglielmo. 1994. On the evidence for partial N-movement in the Romance DP. In Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi and Raffaella Zanuttini, eds. Paths towards Universal Grammar. Georgetown: University Press. Cinque, Guglielmo. 2003. The pronominal origin of relative clauses. Ms. University of Venice. Cinque, Guglielmo. 2004. Introduction to Internal structure of DP. Ms. University of Venice. Cinque, Guglielmo. 2005a. The Internal Syntax of DPs (with special reference to adjectives and relative clauses) TR 1:00-2:40. Cinque, Guglielmo. 2005b. Deriving Greenberg's universal 20 and its exceptions. Linguistic Inquiry 36:315-332. Cohen, Ariel. 2004. Existential Generics. Linguistics  and Philosophy  27:137-168.  Netherlands: Kluwer Academic Publishers. Cook, Clare. 2004. Adjunct Extraction in Nehiyawewin and Yoruba: neither adjunction nor extraction. Ms. University of British Columbia. Corbett, Greville. 2000. Number. Cambridge: Cambridge University Press. Couquaux, Daniel. 1982. French predication and linguistic theory. In R. May and J. Koster. (eds). Levels of Syntactic Representation. 33-64. Dordrecht: Foris. Cysouw, Michael. In press. A typology of honorific uses of clusivity. In Elena Filimonova, ed. Clusivity. Amsterdam: Benjamins. Dechaine, Rose-Marie. 1993. Predicates across categories: Towards a category-neutral syntax. Doctoral Dissertation. University of Massachusetts, Amherst.  Dechaine, Rose-Marie. 1997. On the left edge of Yoruba complements. Paper presented at the University of Ibadan seminar series. Dechaine Rose-Marie. 1999. What Algonquian morphology is really like: Hockett Revisited, in Bar-el, Leora, Rose-Marie Dechaine, and C. Reinholtz, eds. Proceedings from the Workshop on the Structure and Constituency in Native American Languages, 25-72. MIT Occasional papers in Linguistics 17, Cambridge, Massachusetts. Dechaine, Rose-Marie. 2001. On the left edge of Yoruba complements. Lingua 111: 81-130. Dechaine Rose-Marie and Victor Manfredi. 1998. S V O Ergativity and abstract Ergativity. Recherches lingusitiques de Vincennes 27: 71-94. Dechaine, Rose-Marie: and Martina Wiltschko. 2002a. Decomposing pronouns. Linguistic Inquiry. 33: 409-442. Dechaine, Rose-Marie and Martina Wiltschko. 2002b. Deriving reflexives. WCCFL  21  Proceeding. Mikkelsen, L. and C. Potts, eds. 71-84. Somerville, M A : Cascadilla Press. Delfitto, Dennis 2002. Genericity  in Language. Issues of syntax, logical form and  interpretation. Edizioni dell'Orso, Alessandria. Delsing, lars-Olof. 1998. Possession in German. In Alexiadou, Artemis and Chris Wilder, eds. Possessors, predicates and movement in the Determiner Phrase. Alexiadou, Artemis and Chris Wilder, eds. Amsterdam: John Benjamins Publishing Company, 93-108. Deprez Viviane. 1999. De la nature semantique des nominaux sans determinant en creole haitien. Langue 2:289-300. John Libbey Eurotext, Agence Universitaire de L a Fracophonie. Deprez Viviane. 2001. On the syntactic and semantic nature of Haitian Bare NPs. In: Cresti, Tortoa and C. Saterfield, T. (eds.), Current issues in Romance linguistics: Selected Papers from the 29 Linguistics Symposium on Romance Languages. Amsterdam: John th  Benjamins. Deprez Viviane. 2004. Morphological number, semantic number and bare nouns. Lingua. 30: 1-27. j Diesing, Molly. 1992. Indefinites. Linguistic Inquiry Monograph Twenty, Cambridge: MIT Press. Doherty Monika. 2005 Language-specific appropriateness. Linguistics 43:1 -15.  conditions for discourse  linking and  Dowry, David. 1979. Word meaning and Montague Grammar. Dordrecht: Kluwer. Dryer, Mathew. 1989. Plural Words. Linguistics 27:865-895. Dusterhoff, Kurt. 1997. Stage/Individual Level Predicates. Tue Jul 15 09:44:31 BST 1997. En?, Miirvet. 1991. The Semantics of Specificity. Linguistic Inquiry 22:1-25. Faguwa, Daniel. 1961. Adiitu Olodumare. London: Nelson. Fodor, Janet Dean. 1970. The linguistic description of opaque contexts. New York: Garland. Fodor, Dean Janet and Ivan Sag 1982. Referential and quantificational indefinites. Freeze, Ray. 1992. Existentials and other locatives. Language 68:553-596. Fukui, Naoki. 1995. The principles-and-parameters approach: a comparative syntax of English and Japanese. In M . Shibatani and T. Bynon ed. Approaches  to Language  typology. 327-72. Oxford: Oxford University Press. Fukui, Naoki. 1986. A theory of category projection and its applications. Doctoral dissertation, MIT. Ghomeshi Jila, Ray Jackendoff, Nicole Rosen and Kevin Russell. 2004. Contrastive focus reduplication in English (The Salad-salad paper). Natural Language and Linguistic Theory 22: 307-357. Giannakidou, Anastasia. 1998. Polarity  Sensitivity as (NonJVerdical Dependency.  Amsterdam. John Benjamins Publishing Company. Greenberg, Yael. 1998. A n overt syntactic marker for genericity in Hebrew. In Rothestein, Susan, ed. Events and Grammar. Netherlands: Kluwer Academic Publishers. Haspelmath, Martin. 1997. Indefinite Pronouns. Oxford. Clarendon Press. Haegeman, L. 1991. Introduction to Government & Binding Theory. New York: Blackwell. Haegeman, L. 1994. Introduction to Government and Binding Theory 2nd ed. Blackwell. Harley, Heidi Britton. 1995. Subjects, events and licensing. Doctoral dissertation. MIT Press. Hay ward, Richard J. 1979. Bayso revisited: some preliminary linguistic observations II. Bulletin of the School of Oriental and African Studies. University of London 42:101-132. Heim, Irene. 1982. The semantics of Definite and Indefinite Noun Phrases. Doctoral dissertation. University of Massachusetts. Amherst. Heim, Irene. 1988. The semantics of Definite and Indefinite Noun Phrases. New York: Garland Publishing. Heim, Irene. 2002. File change semantics and the familiarity theory of definiteness. In  Portner Paul and Barbara Partee, eds. Formal Semantics: The essential readings New York: Blackwell Publishing Hellan, Lars. 1981. A n argument for a transformational derivation of passives. In A . Belletti, L. Brandi, and L. Rizzi eds. Theory of Markedness in Generative Grammar: Proceedings of the 1979 GLOW conference. Pisa: Scuola Normale Superiore, 217-285. Hertz, Anna. 1997. On the dual nature of the 'possessive' marker in Modern English. Journal of Linguistics 33:511-537. von Heussinger, Klaus. 2002. Specificity and definiteness in sentence and discourse structure. Journal of Semantics 19:245-274. Higginbotham, James. 1985. On semantics. Linguistic Inquiry 16:547-93. Hoekstra, Teun. 1994. ' H A V E ' as B E plus or minus'. In Cinque Guglielmo, Jan Koster JeanYves Pollock, Luigi Rizzi and Raffaella Zanuttini eds. Paths Towards Universal Grammar. Georgetown: University Press. Ionin, Tania and Ora Matushansky. 2004. A singular plural to appear in B . Schmeiser, V . Chand, A . Kelleher and A . Rodriguez (eds.), Proceedings ofWCCFL  23, U C Davis.  Ioup Georgette. 1997. Setting Versus Resetting a U G Parameter: Evidence from a Chinese Learner of English. International Speakers Series, Department of Linguistics, University of Durham, U K , April 1997. Jersperson Otto. 1924. The philosophy of Grammar. London: Allen and Unwin. Kamp, Hans. 1981. A Theory of truth and semantic representation. In Jeroen Groenedijk, Theo Janssen and Martin Stokhof, eds. Formal methods in the study of Language: Proceedings of the Third Amsterdam Colloquium, Part I, Amsterdam: Mathematical Centre, 277-321. Kamp, Hans. 1988. The semantics of Definite and Indefinite Noun Phrases. New York: Garland Publishing, Inc. Kamp Hans and Uwe Reyle. 1993. From Discourse to Logic: Introduction to Modeltheoretic Semantics of Natural Languages, Formal Logic and Discourse Representation Theory '  Part 2. Netherlands: Kluwer Academic Publishers.  Kayne, Richard. 1994. The Antisymmetry of syntax. Cambridge: MIT Press. Kayne, Richard. 1993. Toward a modular theory of auxiliary selection. Studia Linguistica 47:3-31.  Keenan, E. L . 1985. Relative clauses. Complex Constructions. In T. Shopen, ed. Language typology and syntactic description. 2:141-170. Cambridge: University Press. Kiss, E. Katalin. 2005. Event types and discourse linking in Hungarian. Linguistics 43:131154. Kiss, E. Katalin. 1998. On generic and existential bare plurals and the classification of predicates. In Susan Rothstein, ed. Studies in Linguistics and Philosophy: Events and Grammar, 45-162. Koch, Karsten A . 2004. Any bad noun: N-ki-N in Yoruba. 35th Annual Conference on African Linguistics. Harvard University, Boston. Koch, Karsten A . 2005. N-kf-N in Yoruba and the semantics of "any". Berkeley Linguistics Society 31, University of California, Berkeley. Koopman, Hilda. 2000. The spec head configuration. In The syntax of specifiers and heads: collected essays of Hilda J. Koopman Routledge. Kratzer, Angelika. 1995. Stage level and Individual Level Predicates. In G . Carslon and J.F.. Pelletier, eds. The Generic Book. Chicago: University of Chicago Press. Landman, Fred. 2000. Events and Plurality: The Jerusalem. Lectures. Amsterdam: Kluwer Academic Publishers. Landman, Fred. 2004. Indefinites and the type of sets. Blackwell Publishing. Laniran, Yetunde. 1992. Intonation in tone languages: the phonetic implementation of tones in Yoruba. Doctoral dissertation. Cornell University, Ithaca. Lasnik, Howard. 2000 Subjects, Objects and the Extended Projection Principle, wwwlinguistics.stanford.edu/colloq/2000/2000feb04.html Lawal, Nike. 1986. Some Yoruba quantifier words and semantic interpretation. Studies in African Linguistics 17:95-105. Lawal, Nike. 1989. Some Yoruba quantifier words and semantic interpretation: a reply to a critique. Studies in African Linguistics 20:89-92. Leben, W. 1973. Suprasegmental phonology. New York. Garland. Lee, Jeong-Shik. 1995. The role of Case in Exceptional Case marking. Studies in Generative Grammmar 5:265-317. Levinson, Dmitry. 2005. Gradual uniqueness effect. 6 annual SemFest. th  Longobardi, Giuseppe 1994. Reference and Proper Names: A theory of N-Movement in  Syntax and Logical Form. Linguistic Inquiry 25:609-665. Lindauer, Thomas. 1998. Attributive genitive constructions in German. In Alexiadou, Artemis and Chris Wilder, eds. Possessors, predicates and movement in the Determiner Phrase.  Alexiadou, Artemis and Chris Wilder, eds. Amsterdam: John Benjamins  Publishing Company, 109-140. Longobardi, Giuseppe 2000a, The Structure of DPs: principles, parameters and problems. In Mark Baltin and Chris Collins, eds. Handbook of Syntactic Theory, Cambridge, M A : Blackwell. Longobardi, Giuseppe. 2000b, Posteverbal subjects and the mapping hypothesis. Linguistic Inquiry 31:691-702. Longobardi, Giuseppe. 2003. Toward a Unified Grammar of reference. To appear. In Zeitschrift fur Sprachwissenschaft. Longobardi, Giuseppe. 2004. The Structure of DPs: Some Principles, Parameters, and Problems. In The Handbook of Contemporary Syntactic Theory. Mark, Baltin and Chris Collins, eds. 562-603. New York: Blackwell Publisher. Ludlow, Peter and Stephen Neale 1991. Indefinite descriptions: In defense of Russel. Linguistics and Philosophy 14:171 -202. MacDonald,, Jonathan E. 2004. Spanish Reflexive Pronouns: a Null Preposition Hypothesis. WCCFL  23 Proceedings, ed. G. Garding and M . Tsujimura, Somerville, M A : Cascadilla  Press.http://www.sinc.sunysb.edu/Stu/jmacdona/pdfs/Wccfl%2023%20proceedings 20draft%20for%20webpage. McShane Marjorie and Sergei Nirenburg. 2003. Parameterizing and eliciting text elements across languages for use in natural Language Processing Systems. Ms. University of Maryland, Baltimore County. Manfredi, Victor. 1987. Anti-logophoricity as domain extension in Igbo and Yoruba, in Manfredi ed. Niger-Congo Syntax & Semantics 1:97-113. Manfredi, Victor. 1992. A typology of Yoruba nominalizations. In Collins, Chris and Victor Manfredi, eds. Proceedings of the Kwa Comparative Syntax Workshop. M I T Working Papers in Linguistics. Manfredi, Victor. 1994. Syntactic (de)composition of Yoruba 'be' and 'have', edit.es par L . Nash and G. Tsoulas, 237-52. Langues et Grammaire; Actes du Premier  Colloque.  Departement des Sciences du Langage, Universite Paris. Manfredi, Victor. 1995. Inflection by Default. Niger-Congo and Semantics 6:91-112. Marti, Palau. 1992. Note sur Langue et la Transcription. Matthewson, Lisa 1998. Determiner systems and quantificational strategies: evidence from Salish. The Hague Holland Academic Graphics. McCarthy, John J. and Alan S. Prince 1993. Optimality in prosodic morphology: the emergence of the unmarked. Handout from the talk presented at the N E L S 24. University of Massachusetts, Amherst. McCarthy, John J. and Alan S. Prince 1994. Emergence of the Unmarked. Unpublished Ms, University of Massachusetts and Brandeis University. Milsark, Gary. 1974. Existential sentences in English. Doctoral dissertation MIT. Moravcsik, Edith A . 1978. Reduplicative constructions. In Universal of human language 3:297-334. Greenberg, Joseph H . , Charles A . Ferguson and Edith A . Moravcsik, eds. Stanford: Stanford University Press. Moravcsik, Edith A . 1998. Classifiers and plural marking. Email document, 9 November 1998, Discussion List for The Association for Linguistic Typology. Moro, Andrea. 1995. The Raising of Predicates. Cambridge: University Press. Mortenson, D. 2003. Two types of variable elements in Among Anaphora. Workshop on the Form and Function of Pronouns. University of British Columbia. Miihibauer, Jeff.  2004. The semantics of possessor relations in Nehiyawewin. Ms.  University of British Columbia. Miihibauer, Jeff. 2004. Three kinds of nominals in Yoruba. University of British Columbia Working Papers in Linguistics. Miihibauer, Jeff. 2004. Indefinite relations and possession in Plains Cree. Ms. University of British Columbia Miiller, Ana. 2003. Generic sentences with indefinite and bare subjects in Brazilian Portuguese. In Jan Anderson, Paula Menendez-Benito and Adam Werle, eds. The Proceedings of SULA 2:71-86.' Nordlinger, Rachel. 1998. Constructive case: evidence from Australian Languaes. Stanford, C A , CSLI Publications. Ola Olanike. 1989. Yoruba vowel deletion revisited. Department of Lingusitics and Nigerian  Languages seminar series. University of Ilorin Ola planike. 1995. Benue-Congo Prosodic Phonology and Morphology in Optimality Theory. Doctoral dissertation. University of British Columbia. Owolabi, Kola. 1995. More on Yoruba prefixing morphology. In Owolabi, Kola, ed. Essays in honour of Ayo Bamgbose, 92-112. Oyelaran, Olasope Oyediji. 1970. Yoruba Phonology. Doctoral dissertation. Stanford: Stanford University. Oyelaran, Olasope Oyediji. 1982. On the scope of the serial verb construction in Yoruba. Studies in African Linguistics 13: 109-46. Oyelaran, Olasope Oyediji. 1989. Morphological and syntactic constraints on verbal auxiliaries in Yoruba. 4 Niger-Congo Syntax & Semantics Workshop, Universiteit van th  Tilburg. Oyelaran, Olasope Oyediji. 1990. Vowel eilison in Yoruba revisited. 21 Annual conference st  on African Linguistics. The University of Georgia. Partee, Barbara and Vladimir Borschev. 1999. Possessives, favorite and Coercion. In Anastasia Riehl and Rebecca Daly, eds. Proceedings of ESCOL99 173-190. Ithaca, New York. Partee, Barbara and Vladimir Borschev. 2002. Existential sentences, B E , and the genitive of Negation in Russian. Conference on Existence: semantics and syntax. Nancy, Sept. 26 Postma, G. 1995. Zero Semantics: A Study of the Syntactic Conception of Quantificational Meaning. Holland Institute of Generative Linguistics. Pulleyblank, Douglas. 1986a. Tone in Lexical Phonology. Dordrecht: Reidel. Pulleyblank, Douglas. 1986b. Clitics in Yoruba. Syntax and Semantics 19. The Syntax of Pronominal Clitics. Pulleyblank, Douglas. 1988. Vocalic underspecification in Yoruba. Linguistic Inquiry 19:233-270. Pulleyblank, Douglas. 2002. Yoruba patterns of reduplication. Ms. University of British Columbia. Pulleybllank, Douglas. 2004. A note on tonal markedness in Yoruba. Phonology.  21:409-  425. Cambridge: University Press. Pulleyblank, Douglas (in press) Patterns of reduplication in Yoruba, in Kristin Hanson &  Sharon Inkelas, eds. The Nature of the Word: Essays in Honor of Paul Kiparsky. The MIT Press. Pylkkanen, Liina. 2000. Representing Causatives. In Jackson, B . and T. Matthews, eds. Proceedings. C L C Publications. Cornell University, Ithaca, N Y . Radford, Andrew. 1997. Syntax: a minimalist introduction. Cambridge: University Press. Ritter, Elizabeth. 1995. On the syntactic category of pronouns and agreement. Natural Language and Linguistic Theory 13:405-443 Roberts, Ian 1997. Comparative Syntax. London: Ardnold. Rowlands, E.C. 1969. Teach Yourself Yoruba. London: English University Press. Rullmann, Hotze. 2003. General Number and the semantics and Pragmatics of Indefinite Bare Nouns in Mandarin Chinese. Ms. University of Calgary. Rullmann, Hotze. 2004. Bare nominals and general number. Paper presented at Department of Linguistics, U B C Job talks series. Russell, Kevin. 1997.0ptimality Theory and Morphology. In Optimality theory: an overview. Archangeli Diana and D. Terence Langendoen, eds. Massachusetts. Blackwells. 102-133. Sadler, Louisa and Rachel Nordlinger. 2004. Case stacking in Realizational Morphology. To appear. In Linguistics. Scalise, Sergio. 1984. Generative morphology. Dordrecht: Foris Publications. Schoorlemmer, Maaike. 1998. Possessors, articles and definiteness. Alexiadou, Artemis and Chris Wilder, eds. Possessors, predicates and movement in the Determiner Phrase. Alexiadou, Artemis and Chris Wilder, eds. Amsterdam: John Benjamins Publishing Company, 55-86. Schulte, K i m . 2001. Pragmatic relevance as cause for syntactic change: the emergence of prepositional complementizers in Romance. In, Blake, Barry J. & Kate Burridge eds. Historical Linguistics 377-389. Schutze, C. 1996. Korean 'case stacking' isn't: unifying noncase uses of case particles. In Proceedings of the North East Linguistic Society 26, K . Kusumoto (ed). 351-365. Amherst: G L S A Publications. Schutze, C. 2001. On Korean 'case stacking': the varied functions of the particles ka and lul. The Linguistic Review 18: 193-232. Selkirk, Elizabeth O. 1982. The syntax of words. Cambridge. MIT Press.  Shank, Scott. 2004. Domain widening. Doctoral dissertation. University of British Columbia. Siloni, Tal. 1997. Noun phrases and nominalizations: The syntax of DPs. Dordrecht: Kluwer. Sotiloye Bose. 1999. Phonology. In. Introduction to Linguistics. Yusuf, Qreed. ijebu-Ode: Shebiotimo Publications. 33-46. Stalnaker, R. 1974. Pragmatic presuppositions. In M . Munitz and P. Unger eds. Semantics and Philosophy 197-213. New York: University Press. Haunz, Christine. 2004. Perceived similarity: universal or language-dependent? 9  th  Conference on Laboratory Phonology. Storto, Gianluca. 2003. Possessives in context: issues in the semantics of possessive constructions. Doctoral dissertation. University of California, Los Angeles. Stowell, Timothy. 1978. What was there before there was there? In Proceedings of the Fourteenth Annual Meeting of the Chicago Linguistic Society. Donka Farkas ed. Chicago: Chicago Linguistic Society 14/458-471. Stowell, Timothy. 1981. Origins of phrase structure. Doctoral dissertation, MIT, Cambridge, Mass. Stowell, Timothy. 1983. Subjects across categories. The Linguistic Review 2: 285-312. Strawson, P. 1952. Introduction to logical theory. London: Methuen. Svenonius, P. 1994. Dependent Nexus: Subordinate predication structures in English and the Scandinavian Languages. Doctoral thesis, UCSC. Szabolcsi, Anna. 1983. The possessor that ran away from home. In The Linguistic Review 3: 89-102. Dordrecht: Foris Publications. Szabolcsi Anna. 1994. The noun phrase. In. F. Kiefer and E. Kiss, eds. Syntax and semantics 27: 179-274. The syntactic structure of Hungarian. San Diego: Academic Press. Szabolcsi Anna. 2000. The syntax of scope. Handbook of Contemporary Syntactic Theory, 607-634. Baltin andCollins, eds. New York: Blackwell. Takahashi Shoichi and Martina Gracanin Yuksek. 2004. The syntax of Ki in Haitian Creole. MITWPL. Vergnaud, J.R., and M . L . Zubizarreta. 1992. The definite determiner and the inalienable constructions in French and English. In Linguistic Inquiry 23:595-652. Welmers, William E. 1973. African Language Structures. Berkeley: University of California Press.  Wilhelm Andrea. 2005. Bare nouns in Dene S u l i n e . University of British Columbia Seminar. Williams, Edwin. 2000. Adjunct modification. Rivista di Linguistica 12(1):129-154. Wiltschko, Martina. 2004. On the absence of inflectional plural marking in Halkomelem Salish: a formal account. Ms. University of British Columbia. Wiltschko, Martina. 2005. Many things are not [+plural]. M s . University of British Columbia. Wolfart, H . Christoph. 1972. Plains Cree: A Grammatical Study. Transactions of the American Philosophical Society. New Series, 63.5. Yang, Henrietta. 2004. Classifiers and plurality in Mandarin NPs. 27th G L O W . Thessaloniki, Greece. Yoon, James H . 2004. Nonnominative (major) subjects and case stacking in Korean. In P. Bhaskararao and K . V . Subbarao eds. Non-nominative  subjects. Amsterdam: John  Benjamins, 1-47. Yusuf, Ore. 1995. Girdma YorubdAkotun-.Lecturenotes series 3. Ilorin. Obven Publishers. Zamparelli, Roberto. 1995. Layers in the Determiner Phrase. Doctoral dissertation. University of Rochester.  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0092961/manifest

Comment

Related Items