UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Wh-constructions in Nêhiyawêwin (Plains Cree) Blain, Eleanor M. 1997

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

Item Metadata


831-ubc_1997-250172.pdf [ 9.21MB ]
JSON: 831-1.0087953.json
JSON-LD: 831-1.0087953-ld.json
RDF/XML (Pretty): 831-1.0087953-rdf.xml
RDF/JSON: 831-1.0087953-rdf.json
Turtle: 831-1.0087953-turtle.txt
N-Triples: 831-1.0087953-rdf-ntriples.txt
Original Record: 831-1.0087953-source.json
Full Text

Full Text

WH-CONSTRUCTIONS IN N E H I Y A W E W E S (Plains Cree)  by  Eleanor Marie Blain  B.I.D. University of Manitoba, 1973 B.A. University of Manitoba, 1986 M . A . University of Manitoba, 1989  A THESIS SUBMITTED IN P A R T I A L F U L F I L L M E N T OF THE REQUIREMENTS FOR THE D E G R E E OF DOCTOR OF PHILOSOPHY in THE F A C U L T Y OF G R A D U A T E STUDIES (Department of Linguistics)  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH C O L U M B I A August 1997 © Eleanor Marie Blain, 1997  In  presenting  degree at the  this thesis  in  University of  partial  fulfilment  of  of  department  this thesis for or  by  his  or  requirements  British Columbia, I agree that the  freely available for reference and study. I further copying  the  representatives.  an advanced  Library shall make it v  agree that permission for extensive  scholarly purposes may be granted her  for  It  is  by the  understood  that  head of copying  my or  publication of this thesis for financial gain shall not be allowed without my written permission.  Department of The University of British Columbia Vancouver, Canada  Date  DE-6 (2/88)  ABSTRACT This thesis provides an analysis of wh-questions in Nehiyawewin (Plains Cree). The study is done within the Principles and Parameters framework (Chomsky 1981, 1986, 1995). I argue that Nehiyawewin wh-words like aw/na 'who' are not generated in argument position and do not undergo A-bar movement to Spec CP (Chapter 3). Rather, they are licensed as the predicate of a nominal clause, and respect the same syntactic constraints as other nominal clauses: they are strictly predicate-initial; obey a referentiality hierarchy; and display agreement for number, animacy and obviation (chapter 4). I analyze Nehiyawewin nominal clauses as IP with a null Infl head in which the predicate fronts to Spec CP. The clause-initial position of the wh-word is thus part of a more general process of predicatefronting. The nominal clause analysis of wh-words accounts for the absence of wh-movement per se in the language, as well as for the absence of wh in situ. However, based on their interpretive properties, wh-questions must contain an operator-variable chain. I argue that the operator-variable relation arises when the subject of the nominal clause links to an A position in a subordinate clause. This occurs in one of two ways: by means of the kdcomplementizer or the e- complementizer (Chapter 5). If the subordinate clause has kd-, the resulting structure is a relative clause which restricts the reference of the subject. This yields a cleft construction: Who is it, that Mary likes /, ? If the subordinate clause has e-, the clauses are conjoined, and null-operator movement in the subordinate clause forces an anaphoric relation between the wh-word and the A-position in the e- clause: Who is he & t  OPt Mary likes him . t  Having shown how Nehiyawewin wh-words are associated with an operator-variable chain, I then consider the consequences of the proposed analysis (Chapter 6). A defining property of wh-chains is their sensitivity to island effects. Consistent with this, there is an argument/adjunct asymmetry in Nehiyawewin, which in turn bears on the question of where overt arguments are positioned in a polysynthetic language. I argue that complement clauses are base-generated in an A-position, unlike overt DPs which are in an A'-position (adjoined ii  to IP). This explains why long-distance extraction is possible from complement clauses, while extraction from adjunct clauses is ungrammatical. Another property of wh-chains is their sensitivity to Weak Crossover (WCO). W C O effects are absent in Nehiyawewin wh-questions. I argue that WCO may be avoided because there is no movement of a truly quantificational operator in the sense of Lasnik and Stowell (1991), but rather movement of a null operator. I then propose a Weakest Crossover analysis for the absence of WCO, following Demirdache (1997).  iii  LIST OF ABBREVIATIONS  A A' agr / A G R conj CP DP dir fut GF GR IC Infl/I inv IP LF NP obv Op Pass perf pl pro prox Q rel sing s.t. Spec t th VP -vbVAI VII VTA VTI WCO  Argment (position) A-bar (non-argument) Agreement Conjunct prefix (complementizer) Clause Phrase (S') Determiner Phrase Direct] Future Grammatical Functions Grammatical Relations Initial Change Inflection Inverse Inflectional Phrase Logical Form Noun Phrase Obviative Operator Passive Perfective (aspect) Plural null Pronominal argument Proximate yes/no question marker relative clause marker (complementizer) singular something Specifier Trace theme Verb Phrase Verbalizing suffix (on nouns) Intransitive Verb with Animate subject Intransitive Verb with Inanimate subject Transitive Verb with Animate object Transitive Verb with Inanimate object Weak Crossover  iv  T A B L E OF CONTENTS Abstract  ii  Abbreviations  iv  Table of Contents  v  Acknowledgements  ix  Dedication  x  Chapter 1: 1.0  INTRODUCTION  Goals and Outline  1  1.1 Situating Nehiyawewin 1.1.1 The Consultants 1.1.2 The Literature 1.2 Situating the Theory 1.2.1 Lexical DPs are Adjoined Pro in Argument Position Optionality of NPs Free Word Order Discontinuous Constituency of NPs 1.2.2 Complementizer Selection 1.2.3 The Structure of Clauses 1.2.4 Levels of Representation: S-structure and LF Chapter 2: 2.0 2.1 2.1.1 2.2 2.2.1 2.2.2 2.2.3 2.2.4 2.3 2.4  6 7 8 9 10 11 13 14 15 16 18 21  HIERARCHIES AND ALIGNMENT CONDITIONS  Introduction Hierarchies and Alignment Conditions Hierarchies Alignment Domain is within the Clause Local (non-third) Participants Mixed Sets: Local and Non-local Participant Non-local Participants Only Overt Lexical NPs Non-local Participants: Animate and Inanimate Alignment: Inverse vs Passive Summary and Comments  v  22 24 28 34 35 37 39 41 45 47 52  Chapter 3:  WHERE IS THE WH-WORD?  3.0 Introduction 3.1 Nehiyawewin Wh-questions 3.1.1 The Position of the Wh-word 3.1.2 Wh-Agreement 3.1.3 Choice of Complementizer Complementizer kd3.1.3.2 Complementizers3.1.3.3 Independent mode 3.1.4 Summary 3.2 The Wh-/'« situ Hypothesis 3.2.1 The Wh-word is not in Argument Position 3.2.2 Yes/No Question Particles 3.2.3 Wh-Question Particles 3.2.4 Ambiguity of Wh-words 3.2.5 Evaluating the Wh-/>? situ Hypothesis 3.3 The Wh-movement Hypothesis 3.3.1 Wh-movement as A ' -binding 3.3.2 Multiple Wh-Questions 3.3.3 Strong Crossover Strong Crossover in Nehiyawewin 3.3.4 Weak Crossover Possessor DPs: Mohawk vs. Nehiyawewin Relative Clause DPs: Mohawk vs. Nehiyawewin 3.4 The Null-operator Movement Hypothesis 3.4.1 The Antecedent Condition on Null-operator Movement 3.5 Comments Chapter 4:  55 59 60 62 66 67 70 72 73 74 76 76 77 79 84 85 86 88 92 93 94 96 97 99 100 101  NOMINAL CLAUSES  4.0 Introduction 4.1 Nominal Clauses 4.1.1 English Nominal Clauses Predicative Nominal Clauses Equative Nominal Clauses 4.1.2 Nehiyawewin Nominal Clauses Classifying Nominal Constituents Subject and Predicate Ordering Three Paradigms Paradigm I: [+wh] Indefinite - Descriptors & Names  vi  103 106 106 107 112 114 115 119 121 122 4.1.3 4.1.4  Paradigm II: [+wh] Indefefinite - Deictics and pro Paradigm DJ: Descriptors & Names - Deictic DP Deictic DP (Non-overt Nominal) Deictic DP (Overt Nominal) Descriptor - Descriptor Summary Agreement and Structure in Nominal Clauses Nominal Clause Agreement Agreement for Number Agreement for Gender Agreement for Proximate/Obviative No Agreement for Person Verbal Agreement Three types of Agreement Nominal Clause Structures Analysis of Nominal Clauses Summary  Chapter 5:  CLEFTS A N D CONJUNCTS  5.0 Introduction 5.1 Agreement Across Clauses 5.2 Analysis of Wh-questions 5.2.1 Wh-questions with M-clauses Adjunction Sites for Aa-clauses Clefted Wh-phrase with kd- Complementizer 5.2.2 Wh-questions with e- clauses Null-Operator Movement Null-Operator Constructions Purpose Clause Parasitic Gaps Correlatives Conjunction with e- Complementizer 5.2.3 Summary Chapter 6:  124 125 125 127 128 130 132 134 134 134 135 136 138 13 8 141 142 148  149 150 154 156 157 15 9 165 167 169 169 170 170 173 183  ISLANDS AND W E A K CROSSOVER  6.0 Introduction 6.1 Complement Clauses vs. Adjunct Clauses 6.1.1 Long Distance Extractions 6.1.2 Island Extractions Adjunct Islands with osdm 'because'  vii  185 186 186 189 189 Wh-islands Wh-Islands with kispin 'if, whether' Escaping the Island Effects 6.2 Weak Crossover vs. Weakest Crossover 6.2.1 Weak Crossover Weak Crossover in Mohawk Parasitic Gaps in Mohawk 6.2.2 Weakest Crossover Demirdache (1997) 6.2.3 Evidence for Nehiyawewin Possessive DPs Relative Clauses Relative Clauses Associated with Subject 6.3 Comments Chapter 7: 7.0  CONCLUSIONS  Conclusions  226  Bibliography Appendix A: Appendix B: Appendix C: Appendix D:  191 196 197 201 201 203 204 208 210 215 215 217 219 224  231 Inflectional Paradigms Proximate/Obviative and Gender Independent and Conjunct Modes ka:- : Complementizer or Past Tense?  viii  242 246 251 253  ACKNOWLEDGEMENTS There are many people to whom I owe a great debt of thanks for the many ways they helped me accomplish this work. Of particular importance are the Nehiyawewin (Plains Cree) speakers over many years whose lessons in patience and good humour were as invaluable to me as their intelligence and proficiency in teaching me their language. These include Jane Tipewan, Bill Sewepagaham, Donna Paskemin and Mary Ann Palmer. It was through Donna's experience as a Cree teacher that I was able to complete my language reading requirement using Cree. Thank you also, Mary Ann, for inviting me to your sweatlodge to help me get my head together in times of stress — and for your prayers to the Grandfathers on my behalf. I also thank three speakers from Pelican Lake, Joyce Harris, Gladys Bill (who introduced me to her grandmother, Martha Bill), and Ron Chamakese. Also, thanks to Ida Bear who taught a course in Swampy Cree at the Univ. of Manitoba and to my friend, Barbara McLeod, whom I have also consulted in matters of Cree. The other critical component in the completion of this work has been my committee; Rose Marie Dechaine (Supervisor), Hamida Demirdache, and Henry Davis. Through their efforts, I have learned more about linguistics during the writing of this dissertation than I ever hoped to know. Each of them has contributed immeasurably to the final product. As a body they have prodded, threatened and cajolled me when I stalled; but they have inspired and encouraged me without fail. They have spent countless hours and much red ink on numerous revisions. It is due largely to their efforts that a formless mass of data and ideas has taken shape in a form which is coherent and which I hope will be of benefit to teachers and students of the language. Rose Marie, thanks God [sic] you arrived on the scene to rescue me in time to write this dissertation. I'm so glad another Prairie woman was interested in working on Cree. Working with you has helped me to make sense out of it all. Any clarity possessed by this work is due to your incredible ability to extract the structure which underlies the sprawl of my argumentation. Hamida, your enthusiasm and creativity have always been a source of delight and inspiration for me. And Henry of the many-colouredsocks, thank you for all the support and kindness you have shown me over these many years, and also for the final reading of this thesis. I am privileged to count you all among my friends. Other professors who have been my teachers and my support in a variety of ways include Dale Kinkade, Michael Rochemont, and Guy Carden (the trickster); Patricia Shaw, Doug Pulleyblank, and Bruce Bagemihl. Thank you for all you have done for me. Special thanks to Carmen de Silva, who is the centre around whom the department revolves. Thank you to the Algonquianists John Nichols, Charlotte Reinholtz, Alana Johns, David Pentland, Richard Rhodes, and especially to Chris Wolfart who has always believed in me. Special thanks to Debbie James who, as my external examiner, gave me so much feedback and many helpful comments. Thank you also to Strang Burton, Suzanne Urbanczyk, Zita McRobbie, and Leslie Saxon; and to fellow students (past and present) who have shared many good times: Yan Feng Qu, Nike Ola, Takeru Suzuki, Mandy Jimmy, Monica Sanchez, Myles Leitch, Susan Blake, Kimary Shahin, Leora Bar-el, Tomio Hirose — and especially to Nicole Horseherder, Ping Jiang-King, Helmi Braches, not to mention Elizabeth Curry and Lisa Matthewson with whom I shared some great movies and murder mysteries. Thanks to my friends on the outside: Kazuko Shimizu, Eileen O'Byme, Diane Fitzmaurice, Mike and Barb Angel, Ruth Swan and Joanne Phillips. Thanks to Carol Bullen, Kate Reynolds, Lois Marshall, Atsuko Sakaki, and especially for the laughter shared with Shiva Mojtabavi, whose crises always exceeded mine. Also to the special memory of Rick Coffield. Thank you to my family who supported me spiritually and emotionally through thick (my waistline) and thin (my wallet): My sister Irene and husband Herb Haynes, along with Andrew, Chris and Rosa, Heather, and Robert, not to mention Tony and Benny of a feline persuation. My brother, Roger, and his wife, Sharon (thank you, Sharon, for all those gifts of $$), Marguerite and Roger Luke. Thank you all. And last but not least, my parents, Roger (t May 7, 1997) and Gertie Blain. Recently I was looking through some old family souvenirs and came across my baby book where some of my earlier accomplishments were recorded. There was a space for "Baby's first word" (other than mama and dada). My mother had writ ~ to my utter amazement and delight — the word, 'Book'. Have you ever heard the like? This work has been financed in part by a fieldwork grant from the Phillips Fund, American Philosophical Society; and by monies from SSHRC #410-96-1445, a research grant to R.M. Dechaine (UBC). ix  Dedication To my parents, Gertie and Roger Blain  Chapter 1 INTRODUCTION 1.0 Goals and Outline This thesis has as its main goal to investigate the structure of wh-questions in Nehiyawewin (Plains Cree), an Algonquian language spoken in Saskatchewan and Alberta. The study is done within the framework of generative grammar, specifically within the Principles and Parameters framework of Chomsky 1981, 1986, 1995. A wh-question is here defined as an interrogative clause which makes use of elements such as awina 'who (prox)' or awini-wa 'who (obv)' and kikway 'what'. The examples in (1) illustrate two of the ways one can ask a wh-question: (l.a)  b)  awini-wa Mary ka-wapam-a-t who-obv Mary rel-see-DIR-3 Who did Mary see? awina Mary e-wapam-a-t who Mary conj-see-DIR-3 Who did Mary see?  There are three observations to be made about the surface properties of the wh-questions in (1). First, a wh-word construed with an obviative argument may show obviative agreement as in (l.a), or not, as in (l.b). Second, the wh-word is in sentence-initial position in both examples. Third, wh-questions may be marked with the kd- 'Relative' complementizer as in (l.a), or with e- complementizer as in (l.b).  1  These surface properties raise some theoretical questions. For example, how does the proximate/obviative contrast interact with the properties of the wh-construction? Also, given that Nehiyawewin simple NPs are freely ordered, what is the significance of the obligatory sentence-initial position of the wh-word? Finally, although wh-questions with kdand e- complementizers have the same interpretation, they have a different syntax. What are their respective syntactic structures and how does complementizer selection interact with the formation of wh-constructions? The following chapters address these issues as follows. In Chapter 2,1 discuss proximate/obviative along with the direct/inverse system, both of which are characteristic of Algonquian languages. This chapter provides background information that is essential to the arguments and analyses proposed and developed in subsequent chapters. While proximate/obviative and direct/inverse systems identify the prominence of an argument, whquestions establish a link between a wh-word and an argument position. Consequently, understanding the factors which determine argument prominence is a necessary preliminary step for any analysis of Nehiyawewin wh-questions. I propose that direct/inverse marking is determined by a system of hierarchies (Silverstein 1976, Siewierska 1993): alignment of these hierarchies is marked by direct morphology on the verbal complex, and non-alignment is marked by inverse morphology. This approach provides insight into the difference of inverse vs. passive, a distinction which has been much debated in the literature (cf. Wolfart 1973, Jolley 1982, Dahlstrom 1986, Thompson 1989, Klaiman 1992). I argue that inverse signals non-alignment of the Person Hierarchy (2 > 1 > 3-proximate > 3'-obviative > ...) with  2  the Grammatical Relations Hierachy (subject/agent > object/patient). In contrast to this, passive signals non-alignment of the Syntactic Hierarchy (subject > object) with the Thematic Role Hierarchy (agent > patient). Chapter 3 begins an in-depth look at the structure of wh-questions. There are two standard analyses of wh-constructions. Under the in situ analysis, the wh-word remains in an A-position at S-structure (the overt syntax), but undergoes abstract ("covert") movement to Spec CP at Logical Form (LF). This type of analysis is adopted by Cheng (1991) for Mandarin. The overt movement analysis claims that the wh-word is base-generated in an A position and undergoes subsequent overt movement to Spec CP. This type of analysis is adopted by Baker (1996) for Mohawk. I argue that Nehiyawewin wh-constructions do not involve wh-movement per se (either at LF or S-structure), so that neither the in situ nor the overt movement analysis apply. Rather, there is movement of a null operator (Op). Chapters 4 and 5 develop and motivate my analysis of Nehiyawewin wh-questions. Chapter 4 develops a nominal clause analysis, which claims that all wh-words are generated in an equational structure of the type 'who is x'. Because both the wh-word and the element with which it is equated are nominal, these clauses are referred to as nominal clauses. Nominal clauses with non-wh expressions are pervasive in Nehiyawewin: they may appear as independent clauses (Dechaine, to appear), and are also used to form focus constructions and wh-questions. In the course of motivating the nominal clause analysis, Nehiyawewin is contrasted with English, drawing on the work of Rapoport (1987), Heggie (1988) and Moro (1990).  3  Chapter 5 extends the nominal clause analysis to wh-constructions with kd- clauses and e- clauses. In general, wh-constructions must be licensed by operator movement ~ and the operator may be overt or null. English has both overt and null operators: in (3.a), the operator-variable chain involves an overt wh-operator who, while in (3.b) there is a (phonologically) null operator, represented as Op. In both cases, the gap (= ti) represents the trace of the moved operator. (2.a)  This is [ the man [ who; [ I talked to ti.]]]  b)  This is [ the man [ Op; [ I talked to ti.]]] I argue that operator movement in Nehiyawewin always involves movement of a null  operator, i.e., the operator never has phonological content. I further argue that both kdclauses and e- clauses host null-operator movement. Kd- clauses are relative clause structures as in (3.a) which entails a cleft structure with S-structure null-operator movement. On the other hand, e- clauses involve conjunction (see Blain 1995b, 1997) with LF nulloperator movement creating a kind of parasitic gap structure as shown in (3.b). (3.a) b)  Who is it that [ Op  ;  ... kd- ... tj ]  Who is he & [Op; ... e-... ti] The occurence of kd- complementizer in Nehiyawewin wh-questions and in other  operator-type constructions is documented in the Algonquian literature: Wolfart (1973) and Blain (1996b) for Plains Cree; Rogers (1978) and Johns (1982) for Ojibwa; Ellis (1983) and James (1991) for Moose Cree; Reinholtz and Russell (1995) for Swampy Cree. Whconstructions with kd- clauses parallel focus constructions in that both involve a defied  4  structure. The following English examples (with a copula and dummy subject it) illustrate this parallel: (4.a) b)  It is John [ Opi that [Mary kissed tj ]] Who is it [ Op; that [Mary kissed U ]]  In both the focus and wh examples, the clefted constituent is contained in a nominal clause which is generated sentence-initially: It is John... and Who is it... As for wh-questions with e- complementizer, this construction type has not, to my knowledge, been previously documented for Nehiyawewin (though H.C. Wolfart, p.a, acknowledges that it occurs). In this respect, the proposed analysis with e- clauses as coordinate clauses containing parasitic gaps (cf. Ross 1967, Williams 1988) makes an empirical contribution to the documention of Nehiyawewin question formation strategies. This is presented in Chapter 5. Chapter 6 closes the investigation of Nehiyawewin wh-questions by examining constraints on null-operator movement. I show evidence that long-distance operatorextraction out of complement clauses is grammatical, as opposed to operator-movement out of adjunct islands, which is ungrammatical. On the basis of this asymmetry, I conclude that complement clauses are base generated in argument position (while overt NPs are licenced as adjuncts). Extraction is also barred from wh-islands and complex NPs, as these constitute subjacency (Ross 1967, Chomsky 1973 and subsequent work) and C E D violations (Huang 1982). I also provide an account of the absence of Weak Crossover (cf. Chomsky 1976,  5  Koopman and Sportiche 1982, Reinhart 1983 for a discussion of Weak Crossover; Lasnik and Stowell 1991, and Demirdache 1997 for a discussion of Weakest Crossover).  1.1 Situating Nehiyawewin As noted above, Nehiyawewin (Plains Cree) is an Algonquian language — one of five major 1  dialects of Western Cree and often referred to as the 'Y-dialect'. This term is based on the pattern of sound correspondences which distinguishes between: Plains Cree:  Y-dialect  niya  T, me'  Swampy Cree:  N-dialect  nina  T, me'  Woods Cree:  TH-dialect  niSa  T, me'  Moose Cree:  L-dialect  nila  T, me'  R-dialect  nira  T, me'  Plains Cree (Nehiyawewin) is spoken in Alberta and across much of Saskatchewan. Swampy Cree is spoken in Northern Manitoba; Woods Cree occurs in a band stretching from LaRonge, Saskatchewan to Hudson's Bay (Churchill River system); Moose Cree is spoken around Moose Factory, Ontario, west of James Bay. R-dialect is spoken at Isle a la Crosse, Saskatchewan, and at the tip of James Bay in Ontario and Quebec (Tete-de-Boule Cree). (See the map in Wolfart 1973 for details.)  Other Algonqian languages include Ojibwa (spoken in various dialects throughout Ontario, and in Minnesota); Blackfoot in Alberta; the languages of the Innu peoples of Quebec/ Labrador (Eastern Cree); Micmac in New Brunswick; and many other language groups in the U S A , i.e., Fox, Mennomeni, Cheyenne, Malaseet and Passamaquoddy (in Maine). 1  6  1.1.1  The Consultants  The data for this thesis is from four main Nehiyawewin speakers: Jane Tipewan (J), Bill Sewepagaham (B), Donna Paskemin (D), and Mary Ann Palmer ( M A ) .  2  There were also a  few examples elicited from a group of three women in a single session: Theresia Boysis, Evelyn Enright, and Leona Martin ( who are represented as LET). Of my main consultants, Jane and Bill were the oldest — about 50 years. Mary Ann is in her mid-40s and Donna was the youngest in her early 30s. Bill is a school principal who was born in the Lesser Slave Lake region of Northern Alberta. Jane lives on the Wichikan Lake R. near Spiritwood, Saskatchewan. Donna is a Cree teacher who is originally from Sweetgrass R. near North Battleford, Saskatchwan; and Mary Ann is originally from Little Pine R., also near North Battleford. The data elicited represents subdialects of Plains Cree (Nehiyawewin) based on regional differences, both areal (Northern Alberta vs Western Saskatchewan), and age related. It should also be noted that my contact with these four speakers was sequential: (6)  Jane (50)  Spiritwood, Sask.  Summers of 1992 and 1993  Bill (5.0)  UBC  Nov. 1992 to July 1994  Donna (30s)  UBC  July 1994 to May 1995  Mary Ann (40s)  UBC  Jan. 1996 to April 1997  The letters shown in parentheses are used to identify them with respect to elicited examples; i.e., B.245 identifies sentence No. 245 elicited from Bill.  2  7  A s a result, it was not possible to verify, complete paradigms, and otherwise more closely investigate and contrast some o f the data gathered at much earlier periods o f my research. This invites further investigation in many areas. The data was collected primarily via translation o f sentences elicited in English and to a lesser degree by means o f Nehiyawewin sentences composed by the investigator. In the latter situation, the consultant was often asked to gloss my Nehiyawewin sentences into English; and this produced quite different results in some cases (see discussion in Chapter 6).  1.1.2 The Literature Although the descriptive work done on Algonquian languages in the Bloomfieldian and functionalist frameworks is vast, as attested by the extensive bibliography o f Pentland and Wolfart (1982), work in the generative framework is just beginning. The contributors in this framework include Blain (1995a, 1997) on Plains Cree; Brittain (1995) on Sheshashlt Montagnais (Eastern Cree); and Campana (1996) on the conjunct order in Algonquian languages. Dahlstrom (1986) gives an L F G account o f the role o f the lexicon in the syntax o f Plains Cree verbs; and Grafstein (1984, 1989) looks at argument structure and disjoint reference in Ojibwa. Dechaine (1996, 1997a, 1997b) is working on Plains Cree morphology and nominal predication, while Bar-el (1997) and Hirose (1997) examine binding conditions and inchoatives respectively. See also McGinnis (1996) on Ojibwa; Reinholtz and Russell (1995) on quantificational N P s in Swampy Cree, and also Russell and Reinholtz (1995),  8  (1996). Earlier work in the generative includes Frantz (1976, 1978, 1979) which deal primarily with Blackfoot. The generative work previously done on Nehiyawewin (Plains Cree) (see Dahlstrom 1986) has been based primarily on textual evidence, i.e., on the naturally-occurring forms o f the language which are found in published stories. The occurrence o f wh-questions in these texts is rare, and there has been no work done in this area at all. Grammars (i.e., Wolfart 1973) also provide a minimum o f descriptive information in this respect. In Ojibwa and the other Cree dialects, there are two people who have worked on relative clauses and, to some extent, on wh-questions. James (1991) looks at the use o f a set o f Moose Cree preverbs (i.e., the complementizers) in conjunction with relative clauses and wh-questions. Johns (1981, 1982) looks at relative clauses and (briefly) at wh-questions in Ojibwa; she claims (1982) that wh-words are defied in that language.  Reinholtz and Russell (1995) make the  same claim for Swampy Cree; however, in neither case do they consider the consequences o f such a proposal in any detail.  1.2 Situating the Theory In order to proceed with the investigation o f the properties o f wh-questions in Nehiyawewin, it is necessary to introduce some o f the backgrouond assumptions that I will be making about the clausal structure and the architecture o f the theory. There are four areas that are particularly relevant to the study o f wh-expressions. First, given that the wh-words that we will be looking at are wh-NPs, the first question that arises is how ordinary N P s link to  9  clauses (Section 1.2.1). Another issue involves the status o f kd- and e- as complementizers (Section 1.2.2). Third, given that wh-words generally co-occur with verbal clauses, I introduce the basic clause structure that I assume for Nehiyawewin (Section 1.2.3). A fourth point concerns the organization o f the theoretical model that I adopt, in particular the distinction between (overt) S-structure movement and (abstract) movement at Logical Form (Section 1.2.4).  1.2.1 Lexical DPs are Adjoined I am assuming that arguments o f a verb are never in argument position but are adjoined. Nehiyawewin sentences mark agreement on the head (i.e., the verb). This will explain why N P s are optional, and also the freedom o f word ordering. These properties are characteristic o f head-marking languages (Nichols 1986); that is, languages in which strong agreement morphology on the verb is used to express grammatical relationships. The prohibition against N P s in argument position o f a verb could be the result o f two quite different properties o f a language. This could be a result o f the Pronominal Argument Hypothesis which states that rich agreement on the verbs identifies and licenses pro in argument positions. This is the claim o f Reinholtz and Russell (1995) for Swampy Cree. Then the overt N P s are adjoined to IP (Baker 1996) and licensed by coindexation with a pro in argument position.  3  On the other hand, Nehiyawewin could be a language with obligatory  A'-scrambling out o f argument positions (cf. Mahajan 1990).  3  It is not within the scope o f  Baker's proposal is based on earlier work by Jelinek (1984) and Hale (1983).  10  this work to address this issue. In the meantime, I adopt the first o f these two proposals which has pro occurring in A-positions licensing (optional) adjoined N P s (i.e., the Pronominal Argument Hypothesis). The following subsections show that, in Nehiyawewin, the morphological properties o f the verbal clause and the syntactic properties o f overt N P s are at least consistent with this assumption. I first illustrate how rich agreement identifies a null pronominal (pro) in argument position (Section Then I show that Nehiyawewin has the three properties identified by Baker (1996) as being characteristic o f languages with pronominal arguments: optional use o f N P s (Secion, free ordering of N P s (Section, and the occurrence o f discontinuous nominal expressions (Section Pro in Argument Position Consider the following Nehiyawewin sentence, which involves a single word, a complex verb (as is typical in head-marking languages). In (7), the ni- prefix represents the first-person (the subject, in this case) and the suffixes tell us that the object is third-person; i.e., the -d 'direct' morpheme tells us that the subject is the higher (1st) person and -w represents 'third person' (object) (see discussion o f Person Hierarchy in Chapter 2). (7)  4  ni-wapam-a-w 1 -see -dir-3 I see him/her.  The ni- (or ki- ) prefix occurs whether that person is subject or object o f the verb (providing the other argument is 3rd-person). 4  11  For this single word with rich agreement, we can propose a sentence structure as shown (8). In (8.a), the verb has an argument structure with thematic roles for two argument positions, an Agent assigned to the Spec VP position, and a Theme assigned to the complement (sister) position of the verb. (8.a)  VP  < Agent, Theme > The agreement morphology on the verb identifies the arguments, i.e., first person subject and third person object. In the absence of overt NPs, the agreement morphology licences (empty) pronominal arguments in subject and object positions, as shown in (b). (8. b)  VP V  pro,  A pro  V  k  I  nij - wapam - aw  k  T see him/her'  < Agent;, Themek >  12  This establishes that overt N P s are not required in a sentence. What, then, are the properties characteristic o f N P s when they do occur? Optionality of NPs A s mentioned above, one o f the properties o f head-marking languages such as Nehiyawewin is that overt N P s are optional. This is illustrated in (9), where both arguments are thirdperson. Example (9.a) is a complete sentence as shown by the gloss; the inflectional morphology provides pronominal referents. (9. a)  5  wapam-a-w see-dir-3 He saw her (obv).  b)  wapam-a-w Mary-wa see-dir-3 Mary-obv He saw Mary (obv).  c)  John wapam-a-w John see-dir-3 John saw her.  d)  John wapam-a-w Mary-wa John see-dir-3 Mary-obv John saw Mary (obv).  Overt pronouns, when they occur, serve an emphatic function only. Dahlstrom (1995) claims for Fox that an emphatic pronoun introduces a new topic. In Nehiyawewin, this would be restricted to niya 'I, me' and kiya 'you'. Third-person wiya is further restricted. In Blain (1994, 1995a) I argue that wiya functions primarily as a topic-sensitive intensifies 5  13  Overt N P s may be used as shown in examples (9.b) to (d); however, they are optional and serve as referential antecedents for the pronominal arguments identified in the verbal inflectional morphology. The optionality o f N P s can be taken as evidence that the argument positions are saturated, i.e., the verb has discharged its thematic roles and its case features. Under this analysis, when N P s do occur, they cannot be occupying an argument position. According to Baker (1996), this is because an N P in argument position would need to be assigned Case, but the agreement markers on the verb have already absorbed the Case features. This forces lexical N P s to occupy non-argument positions, and specifically to be adjuncts o f the clause. Free Word Order In addition to being optional, when lexical N P s do appear, they are freely ordered (cf. Blain (1992, 1993); Dahlstrom (1986); Wolfart (1973, 1996)): (10.a) John John John John;  e-wapam-a-t o-mama-wa conj-see-dir-3 (3>3') 3-mother-obv he saw her his mother saw his; mother. B.1037  SVO  b)  John o-mama-wa e-wapam-a-t  SOV  c)  e-wapam-a-t o-mama-wa John  VOS  d)  e-wapam-a-t John o-mama-wa  VSO  e)  o-mama-wa e-wapam-a-t John  OVS  f)  o-mama-wa John e-wapam-a-t  OSV  14  A l l o f these word orders are also evidenced in texts (cf. Dahlstrom (1986) for Nehiyawewin/Plains Cree).  6  Inasmuch as adjuncts are more freely ordered than arguments, the free ordering o f N P s is consistent with the claim that they are licensed as adjuncts rather than as arguments. Discontinuous Constituency of NPs A third claim about overt N P s is that they involve discontinuous constituency (cf. for example, Reinholtz and Russell (1995)). Contrast the deictic D P s in (1 l.a) and (b). In (a), the two words are continuous and form a single constituent D P . In (b), the deictic anihi 'those' precedes the verbal constituent while the coreferent nominal awdsis-ak 'children' follows the verb. (11. a)  b)  e-wapam-a-t-ik [anihi awasis-ak] conj-see-dir-3-pl those child-pl H e saw those children. [ anihi] e-wapam-a-t-ik [ awasis-ak ] those conj-see-dir-3-pl child-pl H e saw those children.  Reinholtz and Russell (1995) present the same kind of evidence for Swampy Cree. I also refer the reader to Baker's Chapter 2 which shows this type o f evidence for M o h a w k .  7  N P ordering in Algonquian languages, although free, is sensitive to discourse-related factors, cf. Tomlin and Rhodes (1992) for Ojibwa, Starks (1993) for Woods Cree, and Dahlstrom (1995) for Fox.  6  Another possibility is that they form distinct nominal constituents — possibly occurring in separate clauses. For discussion of related issues, see Dechaine's (to appear) analysis of  7  15  Another example of discontinuous constituency involves a possessor NP, as in (12.a). Note that Nehiyawewin has a second-position Yes/No Q-particle ci. In (12.b), the Qmorpheme separates the possessor Bill from the possessee o-wicewdkan-a  'his friend'.  (12.a) Bill o-wicewakan-a e-wicih-iko-t Bill 3-friend-obv conj-help-inv-3 Bill's friend helped him. cf.J.252 b) Bill ci o-wicewakan-a e-wTcih-iko-t Bill Q 3-friend-obv conj-help-inv-3 Did Bill's friend help him? J.252 To summarize, I have established that Nehiyawewin has the hallmark properties of a head-marking language. Rich agreement morphology identifies null pronominals (pro's) in argument position. As for overt NPs, they display the following three properties: they are optional, they are freely ordered, and they may be discontinuous. These then are the baseline properties of ordinary NPs. As we shall see in subsequent chapters, the distribution of whNPs differs from that of lexical NPs.  1.2.2 Complementizer Selection Wh-words occur most commonly with kd- clauses or with e- clauses as shown in (13). (13.a) awini-wa Mary ksl-wapam-a-t who-obv Mary rel-see-dir-3 Who did Mary see?  predication in nominal clauses. Baker (1996) also discounts an analysis for discontinuous constituency in Mohawk NPs.  16  b)  awini-wa Mary e-wapam-a-t who-obv Mary conj-see-dir-3 Who did Mary see?  I analyze e- and kd- as complementizers. In Wolfart (1973:45 ff.) they are referred to as Conjunct Markers (following Bloomfield 1928, Ellis 1971). However, Wolfart notes that the "changed conjunct" forms in e- and kd- mark subordination and attributes to them meanings such as those shown in (14): (14.a) e -wapam-a -t conj-see -dir-3 '(that) he saw him'  8  b)  ka-wapam-a -t rel-see -dir-3 '(the one) that he saw' '(one one) that saw him'  The complex verb forms in (14) are derived by head movement in the syntax. The complementizer morphemes precede tense/aspect markers such as ki-, as shown in (15).  The term "changed conjunct" refers to Initial Change. The counterpart of Nehiyawewin e- in Ojibwa and other related languages (including Potawatomi discussed below) is Initial Change (IC) — a morphophonological constituent which consists of a pattern of ablaut of the first vowel of the stem (cf. Rogers 1978). Wolfart (1973) proposes that e- (underlyingly I'xl + IC) in Cree languages is just a "vehicle" for the IC process. The kd- complementizer in Nehiyawewin does not occur in ordinary declarative sentences but is restricted to operator environments, occuring in whquestions, focus constructions and relative clauses. In contrast, there are two kd- complementizers in the Moose Cree dialect (D. James 1991 and p.c; Clarke et al 1993). In relative clause (null) operator environments, kd-i is a frozen form historically derived from IC on Proto Algonquian *kiwi- 'go around doing X ' (Clarke et al 1993 — attributed to Ives Goddard). This kd-, complementizer occurs in all tenses together with the appropriate tense/aspect marker. In wh-questions, the operator movement appears to be contained in the synchronic IC process ~ which in past tense operates on underlying /ki-/ 'past tense' + IC to produce kd- (restricted to past tense). Nehiyawewin, on the other hand, has grammaticalized the IC process in the form of overt complementizers e- and kd- In other words, in my analysis, there is only one kd- in Nehiyawewin — which occurs in both relative clauses and wh-questions (see discussions in Appendix D). 8  2  17  Inasmuch as ki- is introduced at the IP level, this ordering is consistent with the hypothesis that e- and kd- occupy a Comp position above the IP projection — assuming (as in Baker 1985) that the order of morphemes reflects the hierarchical architecture of the clause (see next section). (15.a) e -ki-wapam-a -t conj-perf-see-dir-3 '(that) he saw him'  b)  ka-ki-wapam-a-t rel-perf-see -dir-3 '(the one) that he saw' '(one one) that saw him'  Theoretical considerations also support a complementizer analysis for this position. Halle and Marantz (1993) discuss the independent vs. conjunct order in Potawatomi (another Algonquian language). In that language, there is no overt equivalent of e- (see fn. 8) so that there is no overt COMP in either independent or conjunct modes. The evidence for the independent vs. conjunct modes consists of two patterns for negation and two paradigms of agreement morphology. In their analysis, Halle and Marantz (1993:147ff.) propose a (nonovert) C O M P in a functional projection higher than the tense/Infl projection which selects for the independent vs. conjunct paradigm.  1.2.3 The Structure of Clauses Following Kayne (1994), I propose that in Nehiyawewin all projections are head-initial and specifier-initial, giving a uniform Specifier-Head-Complement ordering as in (16):  18  (16)  XP  Spec  X'  Head  Complement  This structure is consistent with the relative ordering of morphemes in the verbal complex. For example, as just discussed in the previous subsection, the complementizers kd- and eappear as proclitics on the verbal complex. This accords with the claim that the CP projection is head-initial. Similarly, tense and aspect markers precede the verb stem, e.g., the future marker wt- in (17). This is also consistent with the idea that the IP projection is headinitial. (17.a) ni-wi-wapam-a-w 1 -flit-see -dir-3 'I will see him.' b) ni-wi-wapam-ik(o-w) 1 -fut-see -inv-(3) 'He will see me.' The verbs in (17) are in the Independent mode which involves both prefixal person agreement and suffixal number agreement. The person prefixes (ni- '1st person , ki- '2nd person, and 0 1  '3rd person') are positioned before tense/aspect prefixes, and can be analyzed as occupying the specifier position of IP. The person markers which occur in the verbal paradigm parallel the personal pronouns:  9  See also Halle and Marantz (1993:150) who analyze these person-markers in Potawatomi as clitics.  9  19  (18)  Pronominal Paradigm niya T, me' lsg kiya 2sg 'you' 's/he, him/her' 3sg wiya niya-ndn 'we (excl.)' lpl kiya-naw 2.1pl 'we (incl.)' kiya-wdw 2pl 'you all' wiya-wdw 3pl 'they'  Verbal Paradigm w'-nikamo-n ^7-nikamo-n 0-nikamo-w w'-nikamo-nan A7-nikamo-(na)naw A:z'-nikamo-(na)waw 0-nikamo-wak  The internal structure of the CP and IP projections are represented in the structure in (19): (19)  CP  C° ka-, e-  IP /' Spec /  S  s  x  s  r  Verb+agr  pro  As (19) illustrates, the ni- '1st person' prefix occupies Spec IP preceding tense/aspect markers (e.g. future wi-), and following Comp. Note, finally, that the arguments of the verb are base-generated in VP-internal positions as null pronominals (pro's) whose contents are identified by the agreement morphology on the verb.  10  This oversimplifies the matter. Note that prefixal person agreement is limited to the independent mode and does not occur with an overt Comp. Conversely, the complementizers kd- and e- occur only in the conjunct mode, which is associated with suffixal person agreement. 10  20  1.2.4 Levels of Representation: S-structure and Logical Form Within the model that I am assuming, there are two levels of representation in the syntax: Sstructure and Logical Form (LF). In principle, movement of a wh-word can take place at either of these levels. As discussed above, there are two types of Nehiyawewin whquestions, those with kd- complementizer and those with e- complementizer, and both involve movement of a null-operator. All things being equal, one might expect that movement of a null-operator can take place at S-structure or at LF. It is clear that complementizer kd- occurs only in operator environments. Complementizer e- occurs elsewhere — but including in wh-questions (an obligatory operator environment). (Note that e- never occurs in relative clauses or focussed NP constructions which require operator movement at S-structure.  11  I shall argue that, with e- clauses, null-operator movement takes  place at Logical Form. Having motivated these background assumptions for Nehiyawewin, I now turn to the question of how the verbal agreement morphology codes argument structure. This is necessary for two reasons: (i) to understand the significance of the morphology in whconstructions (which will be crucial in later chapters), and (ii) to illustrate how the proximate/obviative contrast interacts with other person agreement morphology.  In other words, there must be some form of overt (S-structure) evidence for relative clauses in a language where NPs are optional. 11  21  Chapter 2 HIERARCHIES AND ALIGNMENT CONDITIONS  2.0 Introduction In this chapter, we will look at the reference-tracking system which is characteristic of Algonquian languages. Two contrasts are coded in the inflectional morphology. These are: (i)  The proximate vs. obviative status of the third-person arguments, the contrast evidenced in the presence or absence of suffix -(w)a 'obviative' on NPs, including wh-words. The proximate third person is the more salient or topic-like third person. A third person discourse topic must be proximate, represented as [3]. On the other hand, obviative marks a less salient, nontopic third person, represented as [3 ]. See examples in the table in (3). 1  (ii)  1  The direct vs. inverse marking determines which of the two participants is the agent/actor; this is represented (in the examples below) by the suffixes -d/-i  'direct' and -ikoZ-iti 'inverse' on the verbs.  These systems are represented in the examples which follow. In all the examples, the agreement morphology indicates that there are two third persons involved. The d- 'direct' marker in (l.a) indicates that the 3-proximate person is subject and the 3'-obviative person is  The proximate/obviative distinction in the hierarchy is sometimes referred to as 3rd-person (3) vs. 4th-person (3') distinction. Some grammars even refer to a 5th person (3") (cf. Ellis (1983) and Wolfart (1973)). 1  22  object. In (b), the iko- 'inverse' suffix indicates that the 3'-obviative person is subject and the 3-proximate person is object. (l.a)  b)  John e-wapam-a-t Mary-wa John conj-see-dir-3 Mary-obv John (prox) saw Mary (obv). John-a e-wapam-iko-t Mary John-obv conj-see-inv-3 Mary John (obv) saw Mary (prox).  Wh-words are also marked for the proximate/obviative distinction, as seen in (2): (2.a)  awina e-wapam-a-t Mary-wa who conj-see-dir-3 Mary-obv Who (prox) saw Mary (obv).  b)  awini-wa Mary e-wapam-a-t who-obv Mary conj-see-dir-3 Who (obv) did Mary (prox) see. These two morphological systems — i.e., proximate/obviative and direct/inverse —  operate in conjunction with a third system, the person hierarchy. This hierarchy ranks persons and governs the use of the direct/inverse morphology in the clause. Generally, action by a higher-ranked person on or toward a lower-ranked person is direct. Actions in which the lower-ranked person is the agent are marked inverse. Given that wh-phrases and NPs are both marked for the proximate/obviative contrast and these morphological systems are an integral part of wh-questions as well as declarative sentences, we must understand how they function in the language. I propose that the interaction of the proximate/obviative distinction with direct/inverse marking is the byproduct of how certain hierarchies align (or fail to align) with each other.  23  It is necessary to provide a more thorough account of the systems which mark the binding relations within a clause — not only to accustom the reader to the various patterns involved within the clause, but also to observe how the systems function together to provide reference-tracking information in a larger sentential context. This information will be relevant to the investigation of wh-questions insofar as wh-words, as NPs, also bear morphology for obviation and participate in the systems described here.  2.1 Hierarchies and Alignment Conditions The reference-tracking morphology is sensitive to three sets of contrasts: (i)  the proximate/obviative distinction  (ii)  the direct/inverse distinction  (iii)  person distinctions: speaker (1st person), hearer (2nd person), other (3rd person)  In the first (i.e., proximate/obviative), it is the obviative which is the marked status with suffix -(w)a  2  (3.a)  on animate NPs.  Proximate 3-person (unmarked)  Obviative 3'-person (marked)  Mary napew sisip atim awina  Mary-wa napew-a sisip-a atimw-a awini-wa  'man' 'duck' 'dog' 'who'  'man (obv)' 'duck (obv)' 'dog (obv)' 'who (obv)'  The -wa form typically occurs with vowel-final stems while -a is used for consonant-final stems. There are some consonant-final words like atim 'dog' which have the plural form atimw-ak and the obviative form atimw-a which are analyzed as being underlyingly latimw-l (cf. Wolfart 1973).  2  24  3.b)  Possessor NPs: Proximate 3-person (unmarked) ni-mosom 'my grandfather' ni-mama 'my mother 1  Obviative 3'-person (marked) o-mosom-a 'his grandfather' o-mama-wa 'his mother'  3  In addition to the evidence with respect to NPs above, the proximate/obviative status of arguments is also indicated in the inflectional morphology of verbs and with possessor agreement on NPs. The second contrast relevant to reference-tracking morphology is direct/inverse marking. This marking occurs only with transitive verbs involving animate arguments. The direct/inverse system designates which of the two participants is the agent in the event denoted by the verb. These systems (proximate/obviative and direct/inverse) operate in conjunction with a third system, the Algonquian person hierarchy. Generally, action by a higher-ranked person 4  on or toward a lower-ranked person is direct. Actions in which the lower-ranked person is the subject are marked inverse. (4) Person Hierarchy: 2  >  1  >  3  <Subj/Agent> Action <Obj/Patient><  >  3'  >  3"  > <Obj/Patient> Action <Subj/Agent>  >  0  >  0'  = DIRECT = INVERSE  Information about the proximate/obviative status of inanimate NPs is restricted to the verbal morphology. Obviative status is not marked on inanimate NPs. See Appendix B for an outline of the proximate/obviative contrast in conjunction with gender and related issues in Nehiyawewin. 3  4  See Silverstein (1976) and Siewierska (1993).  25  The numbers representing third-persons in the hierarchy are based on conventions in Wolfart (1973), Wolfart and Carroll (1981), and Ellis (1983); for example: 5  2 = second person  3'-third person further obviative  1 = first person 3 = third person proximate (animate) 3 - third person obviative  0 = third person proximate inanimate 0 - third person obviative inanimate  For example, consider a situation with two animate participants, the hearer and the speaker (the so-called You & Me forms). If a transitive verb has a 2nd-person participant acting on a lst-person participant, then the morphology will show that the verb is Direct. If the person hierarchy is violated, i.e., if the lst-person participant is acting on the 2nd-person participant, then the verb will be morphologically-marked as Inverse. This system operates throughout the hierarchy including the relation between proximate and obviative third persons. As seen in (4), objects of inanimate gender are lowest on the hierarchy. Dechaine (1996 based on Hockett 1966) schematizes the person hierarchy asin (5).  6  The term local  (from Hockett (1966)) refers to the speech act participants, i.e., you and me. Non-local refers to all others, i.e., third-persons, which are further distinguished in the proximate/ obviative contrast. This distinction plays a role not only in the reference-tracking system but  The proximate and obviative terminology itself has a long tradition in the Algonquian literature, i.e., it is used in works like Bloomfield (1946), Hockett (1966), etc. 5  Note that an obviation contrast exists only with third-persons in Nehiyawewin. However, Dechaine observes that the independent personal pronoun system of Blackfoot (a neighbouring Algonquian language cf. Frantz 1991:74) has two singular forms for first and second persons — with suffixes which correspond to the proximate with -wa and obviative with -yi. Unexpectedly, there is only one form (with the -yi suffix) for the singular third-person pronoun. 6  26  also in discourse. First we will consider the characteristic features of proximate and obviative. GENDER  (5)  Inanimate  Animate  Local  Addressee [+2]  Non-local  Speaker [-2,-1]  [+1]  [3] [3'] [3"] prox. obv. further obv. <  [0] prox.  [0'] obv.  obviation contrasts  Proximate status is morphologically unmarked. In a possessor phrase (see (3.b)), a third-person possessor is proximate while the person/thing possessed is obviative. In a clause containing two third-person arguments, it is the subject which typically has proximate status.  7  In a given span of discourse with more than one third person, proximate status is assigned to only one of them — all the others being obviative. The proximate third person constitutes the discourse topic for that span of the discourse. In other words, the discourse topic (if third person) must be proximate. Proximate status can be reassigned to another third person to 8  reflect a change in topic status.  7  The subject is usually considered to be the unmarked sentential topic (cf. Erteschik-Shir 1993).  The discourse topic is the main participant in a section of a story/discourse which spans at least one sentence. This person is assigned proximate status ~ all other third persons being marked  8  27  Obviative third persons are morphologically marked with the suffix -(w)a. The obviative NP is typically the possessed NP in a phrase with a third-person possessor and the object of a verb in a clause with two third-person arguments. The discourse topic cannot be obviative.  2.1.1  Hierarchies  In Siewierska's (1993) Hierarchy of Hierarchies, there are three categories of hierarchies which she ranks as: Familiarity > Dominance > Formal. These larger categories may range over a sub range of hierarchies as follows: i)  Familiarity:  ii) Dominance:  topicality > givenness > definiteness/referentiality a) person:  1> 2 > 3  human > animals > other organisms b) semantic role: agent > patient > recipient.... iii) Formal:  a) structure:  simple > complex  b) length:  short > long  Very briefly, as noted in Siewierska (1993:831), these Linearization Hierarchies ("X > Y") may be understood as showing a preference for X to precede Y. Leftward placement is related to familiarity focus  ofinterest  = subjectivity  (Benveniste 1971), = empathy (Kuno 1976, 1987), =  (Zubin 1979), = viewpoint (DeLancey 1981), = point of view,  perspective.  Subjectivity is defined in Lyons (1982:101) thus: "The term 'subjectivity' refers to the way in  obviative ~ and this status is reflected in the proximate/obviative marking of verbs in a series of sentences until the proximate status is reassigned to some other third-person in the discourse.  28  which natural languages, in their structure and normal manner of operation provide for the locutionary agent's expression of himself and of his own attitudes and beliefs." The hierarchies relevant for the discussion at hand are those that concern familiarity and dominance. Hierarchies represent higher-order generalizations about the relative ranking or salience of a natural class of linguistically significant terms. For example, grammatical functions (GFs) such as "subject", "object", and "oblique" constitute a natural class. It is generally agreed that there is a sense in which "subject" is the most salient GF; and that, amongst the non-subject GFs, "object" is more salient than "oblique". This can be represented as a ranked list, with "X > Y" to be read as "X outranks Y" or equivalent "X is more salient than Y". Thus, a GF hierarchy would appear as in (6). (6) Grammatical Function Hierarchy: Subject > Object > Oblique ... As indicated by  whether these three terms exhaust the GF hierarchy remains an open  question. Another hierarchy that is often invoked is the Semantic Role Hierarchy (also called the Thematic Hierarchy or the 0-Hierarchy), which starts with the observation that roles such as "agent", "patient" and "goal" are not randomly assigned: agents are more likely to be subject, patients are more likely to be object, and so on. This can be understood as being the effect of a ranking of semantic roles, as in (7). Again, note that the actual number of terms which constitute this hierarchy are subject to debate.  29  (7) Semantic Role Hierarchy  Agent > Patient > Goal... This chapter looks at the properties of transitive verbs, i.e., verbs with two arguments, which project a structure as in (8). With respect to Grammatical Functions, the subject is structurally more prominent than the object inasmuch as it c-commands the object. (The subject in (8) is represented as being in Spec VP on the assumption that all arguments are contained within the lexical projection of V.) As for Semantic Roles, the agent role is more prominent than the patient role and therefore is linked to a more prominent argument position. (8)  VP Subject Verb  V Object  < Agent, Patient > Henceforth, I use the term Grammatical Relations (GR) Hierarchy as a cover term to refer jointly to the GF Hierarchy and the Semantic Role Hierarchy. Accordingly, I will often refer to subject/agent as outranking object/patient ~ representing an active sentence. In addition to the GR Hierarchy, Nehiyawewin (and Algonquian languages in general) also exploit hierarchies based on discourse prominence, gender and person in the verbal morphology. In the Participant Hierarchy in (9), the Discourse and Gender hierarchies represented in (a) and (b) are in conjunction with the general Person hierarchy in (c):  30  (9) Participant Hierarchies: (cf. Hockett (1966) for example) a)  Discourse Prominence:  proximate > obviative  b)  Gender:  animate  c)  Person:  2>1>3>3'>0>0'  > inanimate  In a clause with two third-person participants, proximate status is assigned to the subject/agent in the unmarked case, while the object/patient is obviative — as predicted by the Discourse Prominence hierarchy in (a). The Gender hierarchy in (b) is reflected in the morphological agreement between transitive verbs. A transitive verb with an animate object shows agreement for both arguments. If the object is inanimate, there is only subject agreement. (10)  a)  T A Verb (animate object):  S  agr  b)  TI Verb (inanimate object):  S  agr  O  agr  The Person Hierarchy in (9c) is a complex system with many sub-parts. The contents of (9.c) are unpacked in the configurations in (11), where " X > Y " indicates that the personvalue of X outranks the person-value of Y. (11)  Person:  person X > person Y  Local: Mixed: Non-local: Animate: Inanimate:  2 > 1 2/1 > 3 3 > 0 3 > 3' 0 > 0'  31  (hearer) > (speaker) (local) > (non-local) (animate) > (inanimate) (proximate) > (obviative)  If the subject/agent is higher on the Person hierarchy than the object/patient, then the verb is marked direct. If the object/patient is higher on the Person hierarchy than the subject/agent, then the verb is inverse. So, what precisely does the direct vs. inverse morphology do? Direct and inverse morphemes are referred to in the Algonquian literature as "theme markers". They have two functions: the first is to identify one of the arguments of the verb. For example, if an action involves local speech act participants, i.e., the speaker and the hearer, then the direct/inverse theme markers are chosen from the -i/-iti set. If the action involves a non-local person, i.e., a third person, then the direct/inverse theme markers are chosen from the -d/-ikw set. (12) Direct/Inverse markers a)  Local (2 & 1):  Direct = IAI Inverse = /—iti/  b)  Mixed (2/1 & 3):  Direct = /-a/ Inverse = /-ikw/  c)  Non-local (3 & 3'):  Direct = /-a/ (/-e/in Ind) Inverse = /-ikw/  Secondly, the theme marker locates the two arguments in relation to one another in the Person hierarchy and also locates the two arguments in relation to the GR hierarchy. If direct, then the higher person is the subject/agent. If inverse, then the higher person is the object/patient. This will be discussed in more detail in the following sections. In order to account for the direct/inverse contrast, I propose the following Alignment Conditions:  32  (13) Alignment Conditions a)  When the verb is marked DIRECT, the Person hierarchy (2>1>3>3'>0>0') and the Grammatical Relations hierarchy (subject/agent > object/patient) are aligned.  b)  When the verb is marked INVERSE, the Person hierarchy (2>1>3>3'>0>0') and the Grammatical Relations hierarchy (subject/agent > object/patient) are not aligned.  In other words, the direct/inverse contrast captures the alignment vs. non-alignment of two dominance hierarchies — the Person hierarchy and the GR hierarchy. Alignment and non9  alignment are defined as follows: (14.a) Alignment: Given two hierarchies a and (3, each associated with a ranked ordering of elements, then the two hierarchies are aligned if a given pair of elements X and Y has the same relative ranking on both hierarchies (i.e., either X>Y in both hierarchies or Y>X in both hierarchies). b) Non-alignment: Given two hierarchies a and P, each associated with a ranked ordering of elements then the two hierarchies are not aligned if a given pair of elements X and Y does not have the same relative ranking on both hierarchies (i.e., either X>Y in hierarchy a but Y>X in hierarchy P, or conversely). The effect of alignment and non-alignment is illustrated in (15). Note that for both the inverse and the direct, the Grammatical Relations hierarchy is constant; in particular, the Grammatical Functions hierarchy (subject>object) aligns with the Semantic Role hierarchy (agent>patient). We return to this below in Section  10  We will see below that the direct/inverse distinction is not involved with passive constructions which signal the non-alignment of the Semantic Role hierarchy with the Grammatical Function hierachy.  9  Similar observations are made in a chart in Dahlstrom (1986) ~ though the issue was not presented in terms of hierarchies and alignment of hierarchies. 10  33  (15) The (non)-alignment of the Person Hierarchy and the GR Hierarchy: Subject Agent  > >  Object Patient  a) Direct/Aligned  Person X >  Person Y *  b) Inverse/Non-aligned  Person Y <  Person X  * Where X outranks Y on the Person hierarchy. This schema illustrates the function of the direct vs. inverse theme markers on the transitive verb. Row (a) shows that when the Person hierarchy is aligned with the GR (subject/agent > object/patient) hierarchy, the morphology will be direct. In (b), the Person hierarchy is non-aligned and the theme marker will show an inverse verb construction. Ideally, all of the hierarchies in a clause will be aligned. This predicts that direct is the unmarked status and inverse is the marked status. This prediction is borne out inasmuch as the direct markers (-/', -d) are less complex than the inverse markers (-/'//', -ikw).  2.2 Alignment Domain is within the Clause In order to illustrate the role of the Alignment Conditions and the direct/inverse system, the data in the following discussion includes transitive verbs with an animate object. These are grouped according to three categories: (i)  Local (non-third) participants,  (ii)  Mixed sets (a local and a non-local),  (iii)  Non-local (third-person) participants only.  34  The inflectional morphology of each verb reflects the alignment or non-alignment of the Person with the GR hierarchies. Given that the GR hierarchies encode the argument structure properties of a verb, it follows that the direct/ inverse contrast is confined to the local clause; i.e., it only marks the alignment (direct) or non-alignment (inverse) of the coarguments of the same verb. I now illustrate the inverse/direct morphology for the three participant sets identified in (i) to (iii) above: local, mixed and non-local. This is illustrated for the two verbal paradigms, the Conjunct Mode and the Independent Mode, (see Appendix C for a description of these two paradigms.)  2.2.1 Local (Non-third) Participants  [2  >  <Subj/Agent> Action <Obj/Patient><  1  ]  >  3  >  3'  > <Obj/Patient> Action <Subj/Agent>  >  3"  >  0  >  0'  = DIRECT = INVERSE  The same set of theme markers, -/' 'direct' and -it(i) 'inverse', are used in the independent mode as are used in the conjunct mode when both participants are local (non-third person). The following set of examples illustrate the direct and inverse morphology in the conjunct mode: Local participants involve only first and second persons, i.e., the speaker and the hearer. 2 outranks 1 on the person hierarchy. The direct morpheme in (a) indicates alignment of the Person hierarchy with the GR hierarchy, i.e., a second person is acting on a  35  first person. The inverse morpheme in (b) indicates non-alignment of the Person hierarchy with the GR hierarchy, i.e., a first person is acting on a second person. (16.a) e -wapam-i -yan conj-see -dir-2sg You see me. b) e -wapam-it -an conj-see -inv-lsg I see you.  11  2 —> 1 direct = aligned  1 —> 2 inverse = non-aligned  In (a), the subject/agent is second-person and the object/patient is first-person. The verb morphology shows agreement for the higher person in the hierarchy, i.e., -(y)an '2nd person'. The morpheme -/' 'direct' is used only when arguments are restricted to local (nonthird) persons, and it signifies that the object/patient is lower on the Person hierarchy — i.e., that the Person hierarchy and the GR hierarchy are aligned. Since there can only be first- and second-person arguments involved, the object must be first-person.  12  In (b), the verb agreement -(y)cin 'first person' shows that one of the arguments is 1st person and the -/'/ 'inverse' marker (local participants only) signifies that the subject/agent is  Note that the person agreement markers -yan 'second person' in (16.a) and -an 'first-person' in (16.b) represent the subject/agent. However, it would be misleading to analyze them as subject agreement, as a more complete paradigm would show that they code person agreement rather than grammatical function agreement (i.e. subject/object agreement). Cf. Appendix A for a complete set showing the direct and inverse paradigms for a declarative sentence in both the independent and conjunct modes. Observe that the agreement morphology typically represents the argument which is higher on the hierarchy: i.e., the subject in the direct paradigms and the object in the inverse paradigms. However, in the mixed sets involving thirdpersons and non-third persons, many portmanteau morphemes are involved. 11  In this sense, the -/ 'direct' morpheme represents the lower person on the hierarchy, i.e., firstperson, and marks it as the object. With the (b) example, the -/'/ 'inverse' morpheme represents the higher person on the hierarchy, i.e., second-person, and marks it as the object. 12  36  lower on the hierarchy than the object/patient, i.e., that there is non-alignment of the person and G R hierarchies. Hence the object/patient must be second- and the subject/agent must be first-person. In the Independent paradigm, the argument which is higher on the hierarchy above is prefixed to the verb as shown in (17): 2 —> 1  (17.a) ki-wapam-i -n 2 -see -dir-sg You see me.  direct = aligned  b) ki-wapam-iti-n 2 -see -inv-sg I see you.  1 —>• 2 inverse = non-aligned  Note that it is the higher person on the hierarchy which is represented by the pronominal prefix in both the direct and inverse constructions of the Independent Mode.  2.2.2 Mixed Set: Local and Non-local Participant The direct/inverse system extends to third-person arguments. We look first at mixed sets with a third-person and a non-third-person argument. [ 2  >  1  >  <Subj/Agent> Action <Obj/Patient>  <  3  3'  >  >  3"  ]  o  >  >  <Obj/Patient>  = DIRECT  Action  <Subj/Agent>  = INVERSE  0'  The first examples involve the Conjunct Mode. The agreement morphology, including direct/inverse, occurs as portmanteau morphemes in this set:  37  (18.a) e -wapam-ak conj-see - 1>3 I see him. b) e -wapam-it conj-see - 3>1 He sees me.  1 -»3 direct = aligned (portmanteau) 3 -> 1 inverse = non-aligned (portmanteau)  As portmanteau morphemes, -ak 'dir 1>3' and -// 'inv 3>1' cannot be broken down into their constituent parts. They partially disguise the direct/inverse morphemes used when a thirdperson argument is involved. The vowel quality in -ak is the same as -a 'direct' though the vowel length is shorter. Similarly, the vowel quality and length in -/'/ is the same as the initial vowel in -ik(w) ~ -iko 'inverse'). In addition, [t] and [k] are both forms of third person inflection across the various paradigms (see Appendix A). To get a better sense of the Mixed forms, which involve 2/1 acting on 3 (direct) or 3 acting on 2/1 (inverse), let us consider the Independent Mode. Example shows the -d 'direct' morphology clearly: (19.a) ni-wapam-a -w 1 -see -dir-3 I see him. b) ni-wapam-ik 1 -see -inv He sees me.  1 —> 3 direct = aligned  3 -> 1 inverse = non-aligned  In the inverse example (b), the underlying l-ikw-w/ (-inv-3) is phonetically reduced to [-ik]. Once again, the higher ranking participant (i.e., lst-person) is indicated in the prefix regardless of its thematic role or grammatical function (subject/agent vs. object/patient). The -a  'direct' and -ik(w) ~ -iko 'inverse' morphemes indicate that at least one argument is third-  38  person. Example (b) shows the first-person prefix and the inverse marker indicating that there is a third-person argument and this third-person holds a higher argument position than the first-person. In other words, the third-person is subject/agent and the Person hierarchy is not aligned with the GR hierarchy. The verbal paradigms also include examples for 1/2 and 3' (3-obviative) arguments, both direct and inverse. These involve additional morphology to mark the obviative status of the non-1/2 participant. In the interest of simplicity, I do not discuss these forms. See Appendix A for a complete set.  2.2.3 Non-local Participants Only This section deals with verbs with non-local (third-person) participants only. These examples involve the proximate/obviative contrast between animate participants. Recall that the Algonquian tradition codes animate 3rd-persons as "3", and inanimates as "0". Obviation is marked with an apostrophe (3'), and further obviation with a double apostrophe(3"). 2  >  i  [  3  <Subj/Agent>  Action  <0bj'/Patient>  <  >  3'  >  > <Obj/Patient> Action <Subj/Agent>  3"  ]  o  >  o'  = DIRECT = INVERSE  In this set, the proximate/obviative contrast comes into play. If the verb is marked direct and the hierarchies are aligned, then the subject/agent is 3-proximate, i.e., unmarked third-person, while the object is marked 3'-obviative. Let us look at the independent mode forms involving third-persons:  39  (20.a) wapam-e -w see -dir-3 She sees him(obv) b)  3 —> 3' direct = aligned  wapam-ik(o-w) see -inv -3 He (obv) sees her.  3' -> 3 inverse = non-aligned  In Nehiyawewin, there is no prefix on the verb with only third-person arguments. In the inverse example (b), the underlying /-ikw-w/ (-inv-3) is (typically) phonetically reduced to [-ik]. The -8 'direct' and -iko 'inverse' morphemes, as noted above, indicate that at least one non-local third-person is involved and show whether or not the person hierarchy is aligned with the GR hierarchy. The person agreement marking in both (a) and (b) is for 3-proximate, the higher of the two non-local arguments; therefore, the other argument must be 3'obviative. Now we will look at the verb morphology for the conjunct mode: (2l.a) e -wapam-a -t  b)  3 —>• 3'  conj-see -dir-3 She (prox) sees him (obv).  direct = aligned  e -wapam-iko-t conj-see -inv-3 He (obv) sees her (prox).  3' - » 3 inverse = non-aligned  The -d 'direct' morpheme occurs in (a) where the 3 (prox) participant is the subject/agent and the 3' (obv) participant is the object/patient in accordance with the alignment conditions. The -iko 'inverse' morpheme in (b) indicates non-alignment: a higher-ranked person (3-proximate) is linked with a lower-ranked GR (object/patient), and a lower-ranked person (3'-obviative is linked with a higher-ranked GR (subject/agent).  40  Verbal morphology with direct/inverse marking occurs with all transitive verbs involving animate participants. The presence of direct/inverse marking does not depend on whether the arguments are non-overt pronominals in the form of pro, or whether they are overt NPs.  When the arguments are overt, they may be ordinary lexical NPs, wh-words, or  indefinites like awiyak 'somebody', and will themselves be marked for the proximate/ obviative distinction. Overt Lexical NPs In this section, we look at a wide range of data which illustrates the possibilities of the proximate/obviative system. The distinction between proximate and obviative third-persons is illustrated more clearly when there are overt lexical NPs which are marked to show the proximate/obviative distinction. The verbs in (22) are direct, indicating that the Person hierarchy, in this case (3>3'), and the GR hierarchy are aligned. The 3-proximate subject/agent in (a) involves a deictic DP and the object is marked obviative. In (a) and (c), the object/patient is animate but nonhuman. In (d), the same animal is subject/agent. Accordingly, the moose has been assigned proximate status as subject/agent of (d). (22.a) ana napew e-macitotaw-a-t mosw-a that man conj-hunt-dir-3 moose-obv That man hunted the moose. J.434 b) Bill e-pikiskwat-a-t Mary-wa Bill conj-speak -dir-3 Mary-obv Bill talked to Mary. J. 197  3 —> 3' direct = aligned  3 - » 3' direct = aligned  41  c) omacTw e-wapam-a-t mosw-a hunter conj-see-dir-3 moose-obv The hunter saw the moose. J.437  3 —> 3' direct = aligned  d) mos e-wapam-a-t omacTw-a moose conj-see-dir-3 hunter-obv The moose saw the hunter, cf. J.437  3 —» 3' direct = aligned  It should be noted, once again, that the verbal morphology identifies which NP plays which grammatical role (i.e., subject/agent or object/patient) in the sentence regardless of the word order. Dahlstrom (1986) provides examples from Plains Cree (Nehiyawewin) texts showing every possible word ordering combination of the arguments with respect to the verb and to each other. With non-local participants, if one wishes to elicit an inverse Nehiyawewin sentence as in (23.a), then a passive English sentence with two overt NPs may be used. Though the first gloss was used to elicit (a), the sentence is active and is better glossed as in the active English version. This will be discussed in Section In (b), the subject/agent is possessed by a third-person possessor. As noted above, the third-person possessor is proximate and the possessee must then be obviative. The inverse is used to specify that John's dog is chasing him (John) and not someone else. (23.a)  ana moswa e-macltota-iko-t anihi napew-a that moose conj-hunt-inv-3 that man-obv The moose was hunted by the man. J.43 5 OR: The man hunted the moose.  b) John o-tem-a e-nawaswat-iko-t John 3.poss-dog-obv conj-chase-inv-3 John's; dog is chasing him;. B.56  42  3'—> 3 inverse/non-aligned  3' —> 3 inverse/non-aligned  Because of the close connection between proximate status and topicality, the occurrence of an inverse verb form with third-person participants, i.e., a proximate object/patient, is often triggered by shifts in information structure. For example, in (24.a), the speaker assigns proximate status to the (contrastive) focussed NP which is object of the verb (contra the typical proximate subject = unmarked sentence topic). The sentence suggests that we are already talking about the woman (= discourse topic) that John saw and this sentence is correcting some mistaken claim as to her identity. In (24.b), the pronominal form of the object in the elicited sentence implies an antecedent in the discourse, i.e., old information, topicality, and the person that the conversation is about. Hence the proximate status of this pronominal object/patient. The overt NP,  thunder,  (which is syntactically animate) is assigned obviative status because it is  new information and the least topical even though it is subject. (24.a) eko Mary ka-wapam-iko-t John-a the very one Mary rel-see-inv-3 John-obv It was Mary that John saw. B.344  3'->3 inverse/non-aligned  b) e-sekih-iko-t peyisiw-a conj-frighten-inv-3 thunder(bird)-obv The thunder frightens him. B.8  3' -> 3 inverse/non-aligned  Just as ordinary NPs participate in the alignment of the Person hierarchy with the GR hierarchy, so too do wh-words. Wh-phrases can show the same proximate/obviative contrasts as a regular lexical NP, and the inflectional morphology for their corresponding arguments on the verb behaves in the same manner in both cases. (25) illustrates aligned direct structures in a wh-context. In (25a), the 3-proximate wh-phrase is subject/agent while :  43  the 3'-obviative NP is object/patient. The obviative-marked wh-phrase in (b) is object/patient of the verb. In both cases, the proximate NP is subject/agent and the hierarchies are aligned. (25.a) awma e-pakamahw-a-t John-a who conj-hit-dir-3 John-obv Who saw John? b)  3 —> 3' direct/aligned  awini-wa John e-pakamahw-a-t who-obv John conj-hit-dir-3 Who did John see?  3 -> 3' direct/aligned  (26) illustrates non-aligned inverse structures with wh-words. The (a) example has the most topical (proximate) pronominal argument as the object/patient and the subject/agent is the obviative wh-phrase. In (b), the object/patient is 2nd-person with a 3-proximate whphrase as subject/agent. In both examples the Person hierarchy is non-aligned with the GR hierarchies and the verb is marked as inverse. (26.a) awini-wa e-pakamahw-iko-t  b)  3' —> 3  who-obv conj-hit-inv-3 Who saw him/her?  inverse/non-aligned  awina e-wapam-isk who conj-see-3>2 Who saw you?  3 —> 2 inverse/non-aligned  2.2.4 Non-Local Participants: Animate and Inanimate In the previous section, we discussed transitive verbs in which both participants were animate third persons, and we saw that it is the alignment of the proximate/obviative contast with the GR hierarchy which determines whether a verb will be direct or inverse. What happens when inanimate participants are included? All things being equal, one might expect that any  44  animate third person acting on an inanimate (3 —> 0) would yield a direct verb form, while an inanimate acting on an animate (0 —> 3) would yield an inverse verb form (e.g. something falling on or coming into contact with someone). 2  >  i  >  [ 3  <Subj/Agent> Action  3'  >  > <Obj/Patient>  <Obj/Patient> < - - — Action  <Subj/Agent>  3"  >  0  >  0'  ]  = DIRECT = INVERSE  While direct verb forms of the (3 —» 0) type are unattested, it is possible to have inverse verb forms of the (0 —> 3) type. To see why there is this asymmetry, one must consider, in addition to the Person hierarchy, the effect of animacy on verbal subcategorization. In Nehiyawewin, and in Algonquian generally, verbs divide into two broad classes according to the animacy of the object they introduce. This distinction is reflected in the traditional Algonquian nomenclature: V T A verbs are transitive verbs with an animate object; VTI verbs are transitive verbs with an inanimate object. Only V T A verbs, are marked for direct/inverse morphology. Thus, if an object is inanimate, it will be introduced by a VTI verb, and so will not trigger direct marking of the (3 —> 0) type. However, it is possible for an inanimate subject to act on an animate object, i.e., inverse forms of the (0 —> 3) type are possible within the V T A paradigm. To see how this arises, let us look at the interaction of the hierarchies more closely.  13  As you will see in the paradigms in Appendix A, the VTI set does not have theme markers which contrast direct and inverse. The following theme markers correspond to Verb classes: Class I verb: -e 1/2 subject Independent Mode -am elsewhere (i.e., 3/3' subject Independent Mode and throughout Conjunct Mode) 13  45  In accordance with the distinction between V T A - VTI verbs noted above, it is impossible to get an inverse structure for VTI verbs in Nehiyawewin. The action of an inanimate subject on an animate object (0 —» 1) produces an inverse structure of a VTA-type verb. Let us look at this more closely. We have seen from the discussion above the following alignments for V T A verbs (Transitive with Animate object): (27)  a)  Animate  >  Inanimate  b)  Subject  >  Object  Agent  >  Patient  3'  <  3  c)  Consistent with the Alignment conditions, (27) is a context where the verb would be marked inverse since the Person hierarchy is not aligned with the GR hierarchies. With this in mind, consider the possibilities that arise with a VTI (Transitive with Inanimate object), as illustrated in (28). If the Person hierarchy in (28.c) is aligned with the GR hierarchies in (b) and the Animacy hierarchy in (a), then the verb could, in principle, be marked direct. However, because VTI verbs do not bear direct/inverse marking, this alignment is not coded in the verb morphology. (28)  a)  Animate  >  Inanimate  b)  Subject Agent 3  > > >  Object Patient 0  c)  Class II verb: -d Class III verb: -0  46  (29) illustrates the possibility of the Person/Animacy hierarchies not aligning with the G R hierarchy. In principle, this should yield an inverse verb form. But since VTIs do not inflect for the direct/inverse distinction, this non-alignment cannot be marked on a VTI verb. (29)  a)  Inanimate  <  Animate  b)  Subject  >  Object  Agent  >  Patient  0  <  3  c)  However, the resulting structure does surface as an inverse verb form, but it does so within the context of a V T A , as in (30) where the inanimate noun maskihkiy  'medicine' is acting on  an animate entity. (30)  awina maskihkiy e-nanatawih-iko-t  0 —> 3  who medicine.NI conj-heal.VTA-inv-3 Who did the medicine heal? MA.401  Inverse/non-aligned  The fact that it is possible to switch from a VTI to a V T A in order to express the inverse relation of an inanimate acting on an animate (0 -> 3) further confirms that the direct/inverse alternation does not affect GRs per se, but reflects the alignment of the Person hierarchy with GRs.  2.3 Alignment: Inverse vs. Passive We have seen that the inverse form is a marked form in that the alignment of the person hierarchy relative to the GR hierarchies is violated. Inverse structures are often discussed in relation to passive constructions (cf, for example, Wolfart 1973, 1991; Thompson 1989;  47  Dahlstrom 1986; Dryer 1992). In Jolley's (1982) Relational Grammar account, Inverse is analyzed as a passive. As suggested with respect to (23) above, there would seem to be a link between inverse structures and passive; for example, one way to elicit an inverse sentence with an obviative subject/agent is to elicit an English passive construction with two overt third-person arguments.  14  The examples in (31) illustrate this point. Although the sentence in (a) is  elicited via an English passive (in order to suggest the topicality, hence, the proximate status, of the object argument), the inverse structure is not morphologically passive. A Nehiyawewin passive is agentless, as in (b); the 3-passive morpheme -iht signifies that the third-person patient is the subject of the detransitivized verb. The oblique agent is not marked on the verb. (31.a) Jim e-pakamahw-iko-t J o e - w a Jim conj-hit s.o.-inv-3 Joe-obv 3' > 3 Joe (obv) hit Jim (prox). J.348 inverse/non-aligned ELICITED AS: Jim was hit by Joe. b) Jim e-pakamahw-iht Jim conj-hit-PASS.3 Jim was hit. cf.B.927  Passive  On the other hand, consider an example with only local (non-third) participants. If you elicit the passive sentence, I-was hit by you , the result would be a morphologically direct verb form: 14  (i)  ki-pakamahw-i-n  2-hit-dir-(l)sg 'You hit me.'  Since the 2nd-person agent is higher on the Person hierarchy than the 1 st-person patient, the structure is direct and the Person and GR hierarchies are aligned.  48  One might argue that if inverse and passive were the same, they should not involve different verbal paradigms. In the discussion below, the Alignment Conditions are used to provide an account of the difference between inverse and passive. The subject of a sentence is typically considered to be the unmarked sentential topic (cf. Erteschik-Shir 1993, for example).  It was noted (in relation to (15) above) that direct  15  is the unmarked (i.e., aligned) form insofar as the higher person on the Person hierarchy is the subject/agent (topic) of the sentence. Inverse is the marked (non-aligned) form of an active sentence in that the lower person on the Person hierarchy occupies the subject/agent (topic) position. Accordingly, we have seen that in direct constructions involving two third persons, the third-person subject is proximate.  16  The inverse construction, in which the 3'-obviative  argument is subject/agent, topicalizes (in the sentence context) the less "topic-like" argument according to the Person hierarchy.  17  Nehiyawewin active clauses function in accordance with the two components (subject/object and agent/patient) of the GR hierarchy to provide both direct and inverse alignments. In inverse structures, the misalignment is between the Person hierarchy and the two GR hierarchies. Thus, in the inverse example in (31 .a) above, we have the following structure:  15  Klaiman (1993) uses the term ontological  salience.  ' In Nehiyawewin, only a proximate NP (as opposed to an obviative NP) may be a discourse topic. 6  17  Cf. Kinkade (1989) with respect to 'topical object' structures in Salish.  49  (32.a) Jim e-pakamahw-iko-t Joe-wa Joe (obv) hit Jim (prox)  INVERSE  b) Subject Agent  > >  Object Patient  Syntactic Hierarchy Semantic Hierarchy  3'-obv  <  3-prox  Person Hierarchy  c)  The direction of the GR hierarchies in (b) remain constant with respect to each other. It is the direction of the Person hierarchy in (c) which is non-aligned with the other hierarchies. In the process, the lower person, who is the "natural" object/ patient, is topicalized by making it the subject/agent of the verb. English passive constructions also have a topical object, but the process is different. In an English passive, the patient (as sentential topic) is in subject position. The agent (the usual candidate for subject and topic) is relegated to an oblique phrase following the verb. (33.a) Jim was hit [by Joe].  English PASSIVE  b) Subject  >  Oblique  Syntactic hierarchy  c) Patient  <  Agent  Thematic hierarchy  In the English passive, the patient is in subject position, while the agent is in an oblique position and is introduced by a preposition. The syntactic roles change and are separated from their usual thematic counterparts, i.e., subject is no longer agent, as illustrated in (33). In effect, the two GR hierarchies are misaligned and it is this misalignment which is marked by the passive structure in English. To reiterate, English active vs. passive depends on a misalignment between the Syntactic (subject/ object) hierarchy and the Thematic (agent/patient) hierarchy.  50  Nehiyawewin passives (see (31 .b)), like English passives, are restricted to the misalignment of the subject/object and the agent/patient hierarchies as illustrated below. However, in these situations, the verb is inflected only for its subject (i.e., the patient) — oblique arguments are not marked in the verbal morphology. (34.a) Jim e-pakamahw-iht Jim conj-hit-PASS.3  b) Subject Patient c) 3-prox  18  PASSIVE (Nehiyawewin)  0 Agent 0  In other words, in the Nehiyawewin passive, the patient is the subject — just as in English. This constitutes a non-alignment of components within the GR hierarchy itself, i.e., between the Syntactic and the Thematic hierarchies. But since the oblique agent argument is not marked on the verb, there is no direction marking. In other words, the verb is detransitivized and there can be no alignment violation with respect to the Person hierarchy. The only person in (c) is the 3-proximate subject.  19  2 0  Nehiyawewin verbs may have applicative inflection whereby an animate indirect object replaces the direct object of the verb; however, there is no inflection for a demoted subject (a chomeur in the relational grammarian's terms). 18  19  An obviative 3rd-person can also be the subject in a passive sentence (see Appendix A).  D. James (p.c.) points out that the issues involved with inverse vs. passive in Nehiyawewin may be more complex than I have shown them to be above. For example, there are some questions raised in the literature as to whether examples like (3l.b) are indeed passives. (My analysis accepts the prevailing claim that (3l.b) is passive.) James makes particular reference to Wolfart (1991) for a discussion of some of the issues. 20  51  2.4 Summary and Comments In this chapter we have seen that the Alignment Conditions contribute much to the understanding of reference tracking in Nehiyawewin. We have seen how the proximate/ obviative hierarchy and the animacy hierarchy form extensions of the Person hierarchy. The Person hierarchy, in turn, works in conjunction with to the GR hierarchies to mark the contrast between direct and inverse. The inflectional morphology provides information with respect to the persons involved. The direct/inverse marking designates their grammatical roles — in accordance with the Alignment Conditions — to specify the grammatical relationships between two arguments in a single transitive clause. (35) Alignment Conditions a)  When the verb is marked DIRECT, the Person Hierarchy (2>1>3>3'>0>0') and the Grammatical Relations hierarchy (subject/agent > object/patient) are aligned.  b)  When the verb is marked INVERSE, the Person Hierarchy (2>1>3>3'>0>0') and the Grammatical Relations hierarchy (subject/agent > object/patient) are not aligned.  Inverse structures as discussed above do not occur in English, but do occur in Nehiyawewin. Since inverse does not involve disruption of the GR hierarchies, the inverse Nehiyawewin verb retains its active (vs. passive) status and the verb is inflected for both arguments. We have also seen that passive structures and inverse structures involve misalignment between two different sets of hierarchies: (36)  a)  Active vs. passive depends on a misalignment between the two GR hierarchies, i.e., between the Syntactic (subject > object) hierarchy and the Thematic (agent > patient) hierarchy.  52  b)  Direct vs. inverse depends on a misalignment of the Person hierarchy in relation to both of the GR hierarchies.  Active vs. passive misalignment occurs in both English and Nehiyawewin; however, the results are different in each of these languages. In English, which relies on structural positions for verbal arguments, a passive structure has the patient as subject and an oblique phrase containing the agent. In Nehiyawewin, verbal arguments are indicated in the agreement morphology. The demoted agent argument cannot be inflected on the verb — only the patient argument (as subject) is part of the verbal AGR. The theoretical point to be drawn from this discussion is that the direct and inverse marking is characteristic of V T A transitive verbs — and where this marking does not exist, there is only a single argument position, i.e., the subject. Secondly, direct and inverse, by definition, range over a section of the Person hierarchy, as in (2 > 1) or (3 > 3'), and inverse (3 < 1) ~ where each participant has a different value in the Person hierarchy. Given this "range", disjoint reference is obligatory with transitive verbs; in other words, coreferential combinations like *(3 <> 3) and *(1 < > 1), where both participants have the same value in the Person hierarchy, are impossible. This makes predictions with respect to Binding Conditions A and B of the Binding Theory. Condition A states that an anaphor must be bound in its governing category. In other words, the [+anaph] constituent '...self must be bound by the subject: (3 <> Anaph).  53  By definition, then, the verb cannot have direct or inverse morphology and therefore, it cannot be a transitive verb (VTA). It follows, then, that there can be no such thing as an A bound lexical anaphor in argument (object) position of a verb in Nehiyawewin. Condition B states that a pronoun must be free in its governing category. In other words, the object of a transitive verb cannot have the same reference as its subject. Since any two third-person arguments or a transitive (VTA) verb cannot both be proximate, that means that that the following configuration is not possible, i.e. , *(3 <> 3) where both participants have the same value on the Person hierarchy. In other words disjoint reference is obligatory in a transitive clause and Condition B is a given.  21  Bar el (1997) has looked at binding in Nehiyawewin and has found evidence that Condition C holds in the language. M y own data in this regard, rather perversely, concentrates on situations where Condition C is violated.  21  54  Chapter 3  WHERE IS THE WH-WORD?  3.0 Introduction In this chapter, we begin our investigation of wh-questions in Nehiyawewin — an investigation which will span the remaining chapters. The following examples are representative: (l.a)  b)  awina ana ka-wapam-a-t John-a who that rel-see-dir-3 John-obv Who is it that saw John? awini-wa Mary e-wT-wikim-a-t who-obv Mary conj-intend-marry-dir-3 Who is Mary going to marry?  M y investigation of the nature of Nehiyawewin wh-questions is embedded within a theory of natural language which holds that, although languages may display surface differences of various kinds, they share a core set of principles in accordance with Universal Grammar. Under this view, one expects that the properties of wh-questions that have been found to hold in other languages should also be active in Nehiyawewin. Wh-questions have the following characteristics (van Riemsdijk and Williams 1986:100): (i) there is a wh-word in Spec CP; (ii) a gap is involved; (iii) the wh-word is related to the gap by movement; (iv) the relation between the wh-word and the gap is subject to subjacency; and (v) the relation between the wh-word and the gap is unbounded. This chapter seeks to establish whether properties (i) to (iii) hold of Nehiyawewin wh-questions.  55  The last two properties are examined in Chapter 6. The questions that will preoccupy us are: (i) Where is the wh-word located at S-structure? and (ii) Is there wh-movement? One way in which languages display surface variation is in how they satisfy properties (1) and (ii), namely the presence of a wh-word in Spec CP and the presence of a corresponding gap. In languages like English, the wh-word is evidently in Spec CP (as evidenced by the presence of do in Comp, i.e., the head (C) of CP) and there is an obvious gap: (2)  Whoi did John see ti? However, in languages such as Chinese, the wh-word is in argument position (Cheng  1991:123) and there is no apparent gap: (3)  botong kan-wan-le sheme Botong read-finish-ASP what 'What did Botong finish reading?'  ;  At first glance, the Chinese data seems to contradict the claims in (i) and (ii) above, according to which all wh-questions have their wh-word in Spec CP with a corresponding gap in an A position. To see this more clearly, compare the English and Chinese S-structures given in example (4):  56  English  wh-t  Chinese (Huang 1982)  wh-word  Although Chinese wh-words are not in Spec CP at S-structure, various analyses (Huang 1982, Cheng 1991) have proposed that they move to that position at Logical Form (LF). This preserves the generalization that all wh-questions involve a relation between a wh-word in Spec CP and a gap. In English-type languages, this movement applies at Sstructure and the wh-word appears in Spec CP at S-structure, as in (5.a). This is representative of languages that have overt wh-movement. In Chinese-type languages, the wh-word is in situ at S-structure, with covert wh-movement applying at LF as in (5.b).  English (S-structure movement)  Chinese (LF movement)  57  Note that in both analyses, the wh-word is generated in an argument position. If the wh-word does not move at S-structure, then the wh-word remains in its base-generated, i.e. in situ  position. If the wh-word moves at S-structure, then there is a relation between it and  a gap in A-position. (The gap is represented as a trace (t), and the relation is indicated by coindexation.) In light of the distinction between wh-/'« situ languages and wh-movement languages, a question that naturally arises concerning Nehiyawewin wh-questions is where wh-movement applies: at LF or at S-structure. Answering this question requires that we examine the relation of the wh-word to the clause it is associated with. In this chapter, I argue that: (i)  In Nehiyawewin, the wh-word cannot be in situ, i.e., it is not in argument position of a verb;  (ii)  In Nehiyawewin, the wh-word cannot be in Spec CP of a clause, i.e., it does not move out of an argument position to the Spec CP.  If correct, these two claims lead back to the question we started with: What is the position of the wh-word? Given that wh-words are a kind of NP, one possibility is that whwords would be positioned in the same way as ordinary NPs. As discussed in Chapter 1, in Nehiyawewin, ordinary NPs are analyzed as being adjoined to the clause and are co-indexed with a pro in argument position, as in (6. a). If wh-words are similarly licensed, this would lead to the configuration in (6.b). (6.a)  [NPj [^...proi..]]  b)  *[ whi [n- ...proi...]]  58  I argue that (6.b) is impossible, which leads to a third claim about the nature of Nehiyawewin wh-questions: (iii)  In Nehiyawewin, the wh-word is not adjoined to IP and co-indexed with pro (or other empty category) in argument position (cf. Reinholtz and Russell 1995 on Swampy Cree)  The remainder of this chapter is devoted to motivating the three claims made above. Section 3.1 surveys the surface properties of Nehiyawewin. In the course of doing this, it emerges that wh-words are not licensed in the same way as ordinary NPs; i.e., they are not adjoined to IP. Section 3.2 shows that the in  situ  hypothesis does not apply to Nehiyawewin:  wh-words are not in A-position at S-structure. In Section 3.3,1 show that the wh-movement hypothesis does not apply, i.e., wh-words do not move from an argument position to Spec CP. Section 3.4 introduces the null-operator hypothesis, which forms the basis of the rest of the thesis. Although Nehiyawewin wh-words do not undergo wh-movement per se, they are linked with a null operator, and it is this null operator which undergoes movement.  3.1 Nehiyawewin Wh-questions This section introduces the surface characteristics of Nehiyawewin wh-questions. These include: (i)  the sentence-initial position of the wh-word (Section 3.1.1);  (ii)  the agreement of the wh-word with the A-position it is construed with (Section 3.1.2). This involves the presence or absence of proximate/  59  obviative agreement, according to the status of a corresponding argument as represented in the verbal AGR. (iii)  the choice of the complementizer on the verbal clause associated with the wh-word (Section 3.1.3). The complementizer on the verb may be e- or kd-.  3.1.1. The Position of the Wh-word A general characteristic of Nehiyawewin wh-questions is the sentence-initial location of the wh-phrase as in (7.a) — the [* NP.. V.. wh- ] and [* V.. wh- .. NP ] orderings as shown in (b) and (c) are ungrammatical: (7.a)  awini-wa John ka-ocem-a-t who (obv) John (prox) rel-kiss-dir-3 Who did John (prox) kiss?  b)  *John ka-ocem-a-t awini-wa John (prox) rel-kiss-dir-3 who (obv) Who did John (prox) kiss?  c)  *ka-ocem-a-t awini-wa John REL-kiss-dir-3 who (obv) John (prox) Who did John (prox) kiss?  In addition, in a Nehiyawewin sentence with two or more clauses, the sentence-initial whword may be construed with an argument two or three clauses away as in (8), where the whword is construed as the object of the embedded verb 'kiss'. (8)  1  awina e-itwe-yan e-iteyiht-am-an John e-ocem-a-t who conj-say-2 conj-think-th-2 John conj-kiss-dir-3 Who did you say you think John kissed? B.497  Examples like (8) give the appearance of long-distance wh-movement. We will see in Chapter 5 that the wh-word always moves locally. 1  60  The fact that wh-words are restricted to clause-initial position is significant in a language in which (i) ordinary NPs order freely within the clause, as shown in (9); and (ii) NPs are restricted to their own clause as shown in (10). (9. a)  John e-ocem-a-t Mary-wa John (prox) conj-kiss-dir-3 Mary-obv John kissed Mary.  SVO  b) e-ocem-a-t Mary-wa John  VOS  c) e-ocem-a-t John Mary-wa  VSO  d) John Mary-wa e-ocem-a-t  SOV  e) Mary-wa John e-ocem-a-t  OSV  f)  OVS  Mary-wa e-ocem-a-t John  The following example illustrates the clause boundedness of NPs.  John  cannot  precede the main clause verb, as shown in (10). (10)  *John ni-kiskeyiht-e-n e-ocem-a-t Mary-wa John 1-know s.t.-th-sg conj-kiss-dir-3 Mary-obv I know that John kissed Mary.  This contrast between regular NPs (which order freely and are clause bound) and wh- NPs (which are sentence initial and not clause bound) is correlated with differences between the wh-words and ordinary NPs. NPs typically agree with the proximate/obviative value of the pro in the argument position with which they are construed. This agreement is necessary in order to identify the role of a given NP in any sentence with two or more third-persons. However, agreement is sometimes unmarked on wh-words.  61  3.1.2 Wh-Agreement A wh-word may show agreement with the argument position that it is construed with. In particular, it may agree with the proximate/ obviative status of the A-position. As discussed in Chapter 2, the involvement of two third-persons requires that one argument be marked proximate (3) and the other argument be marked obviative (3'). This is illustrated in (11). (11)  John e-ocem-a-t Mary-wa John conj-kiss-dir-3 Mary-obv John (prox) kissed Mary (obv).  3 —»3'  As interrogative NPs, awina 'who' (prox) and awinihi or awiniwa 'who' (obv) show the same contrast for proximate/obviative features as would any other lexical NP. In (12.a), the wh-word is proximate and represents the subject of the verb with a (3') object. In (b), the wh-word is proximate and represents the object argument (with a 2nd-person subject). (12.a) awina ka-ocem-a-t Mary-wa who (prox) rel-kiss-dir-3 Mary-obv Who (prox) kissed Mary (obv)? b) awina ka-wapam-at who (prox) rel-see-2>3 (dir) Who did you see? J.417  3-wh —> 3'  2 —»3-wh  The wh-word in both sentences in (13) is constued with the obviative object of the following verb. The expected agreement pattern occurs in (a) where the wh-word is marked obviative. However, the unmarked wh-word shown in (b) is also possible. The default (proximate) form awina 'who' is usually acceptable where there is no chance of ambiguity. (13.a) awini-wa John ka-ocem-a-t who-obv John rel-kiss-dir-3 Who did John (prox) kiss?  3 —» 3'-wh  62  b) awina John ka-ocem-a-t who John rel-kiss-dir-3 Who did John (prox) kiss?  3 —» 3'-wh cf.B.497  In both examples, the proximate NP John identifies the proximate subject argument of the verb.  2  In (13 .b) the non-agreeing wh-word occurs by itself in the wh-expression. It is also possible for a non-agreeing wh-word to occur with a deictic such as ana 'that', as in (14). As will be discussed in greater detail in Chapter 5, wh-deictic combinations are usually (but not obligatorily) non-agreeing, and they appear to require the kd- complementizer. An agreeing form is given in (c). (14.a) awina ana John ka-ocem-a-t who that one John rel-kiss-dir-3 Who is it that John kissed? b) awina ana John ka-wT-wikim-a-t who that one John rel-fut-marry-dir-3 Who is it that John is going to marry?  3 - » 3'-wh  3 —> 3'-wh D.263  c) awini-wa anihi John ka-wi-wTkim-a-t who-obv that(obv) John rel-fut-marry-dir-3 Who is it that John will marry?  3 —> 3'-wh  D.261  There are two sub-dialect forms for the obviative wh-word, i.e., awini-wa and awinihi 'who (obv)' as shown in (15).  3  By contrast, the absence of an overt NP in (i) with an unmarked wh-word could result in ambiguity as shown: (i) awina ka-ocem-a-t who rel-kiss-dir-3 3 —> 3' Who kissed him/her? Who did s/he kiss?  2  63  Proximate  Obviative  Mary 'moose'  napew Mary mos  napew-a Mary-wa mos-wa  •who'  awina  awini-hi awini-wa  (15)  'what'  OR  2 .4 kikway'  kikway  There are two sub-dialect forms for the obviative wh-word (examples in (i) and (ii)). The older form awinihi occurs in N . Alberta and is typically represented in the grammars (cf. Wolfart 1973). (i) awini-wa John ka-wi-wikim-a-t 3 - » 3' -wh who-obv John rel-intend-marry-dir-3 Who is it that John will marry? D.261 3  (ii)  awinihi John ka-ocem-a-t who (obv) John (prox) rel-kiss-dir-3 Who did John (prox) kiss? cf.B.497  3 —» 3' -wh  The usual form found among speakers in more southerly regions of Saskatchewan and Alberta is The obviative -wa suffix is the same suffix which typically appears on lexical NPs, as shown in the list in (15) which shows the forms of'who' together with examples involving ordinary NPs. Note that, while the -wa pattern in (i) inflects wh-words like ordinary NPs (e.g., awini-wa, Mary-wa), the -hi pattern in (ii) treats wh-words like determiner expressions (awini-hi, ani-hi 'that (obv) one'). At present, it is not known whether the inflectional pattern of the wh-words has any consequences for the structure of wh-questions. In this regard, it is potentially significant that more southerly sub-dialects use e- complementizer more freely and even employ independent mode verb forms with wh-questions; these are the same dialects which inflect wh-words like ordinary NPs (awini-wa). The more conservative (northerly) dialect does use the ecomplementizer, but not as freely as in southern varieties, and the independent mode forms are much more restricted (i.e., to rhetorical questions); these are the dialects which inflect wh-words like determiner expressions (awini-hi). awini-wa.  Inanimate NPs do not contrast for proximate/obviative. The form kikwdy-a is given in Wolfart (1973) as representing the plural form, which typically collapses with obviative for inanimate NPs.  4  64  The wh-words given in (15) are those which are construed with argument positions. There are also wh-adverbials (cf. Section 3.2..2), but they do not show wh-agreement, and so are not relevant to the issue at hand. In the examples above, we have seen that the wh-word either shows agreement [+AGR] with its corresponding argument on the verb or is unmarked or [-AGR] with respect to the proximate/obviative status of its referent in the wh-question. In other words: (i)  A [+AGR] wh-word occurs in the obviative form awiniwa or awinihi if its referent in the following clause is obviative, or as (unmarked) awina if its referent is proximate.  (ii)  5  A [-AGR] wh-word occurs in the unmarked form awina.  Morphologically, a wh-word has two forms: the unmarked proximate form awina, and the marked obviative form awiniwa or awinihi, depending on the sub-dialect. Moreover, an obviative argument may be linked to either an obviative wh-word or an unmarked whword. This establishes that the unmarked form of the wh-word is not specified as [+proximate], but rather is a default form, i.e., unspecified for the proximate/obviative distinction. In addition, the use of a default non-agreeing wh-word occurs in every subdialect; and although individual speakers may differ in how they use the non-agreeing form, everybody uses it.  I will focus my arguments on wh-words construed with obviative argument positions. The proximate cases which use the awina neutralize the cases of [+AGR] and [-AGR] because proximate is morphologically unmarked. The analyses for the [+AGR] and [-AGR] forms for obviative cases can be generalized to the proximate examples (see Chapter 5). 5  65  In summary, we have seen: i) that wh-words always occur sentence initially; and ii) that wh-words may agree or not with the argument position they are construed with, i.e., they may be [+AGR] or [-AGR].  3.1.3 Choice of Complementizer We now consider the evidence with respect to complementizer choice in wh-questions. There are three possibilities: kd- complementizer, e- complementizer, or no complementizer at all, as illustrated in (16). In all of these examples, the wh-word agrees with a proximate argument. 16.a)  b)  awina ka-ocem-a-t John-a 3-wh —> 3' who (prox) rel-kiss-dir-3 John-obv Who is it that kissed John (obv)? // Who kissed John? awina e-ocem-a-t John-a who (prox) conj-kiss-dir-3 John-obv Who kissed John (obv)?  c) awina ekote ki-wapam-a-w who there 2-see s.o.-dir-3 Who did you see there? D.66  3-wh —> 3'  2 —> 3-wh  (16.a) and (16.b) there is contrast with respect to element is the choice of kd- or ecomplementizer (or conjunct marker as they are referred to in Algonquian grammars, cf. 6  Wolfart & Carroll (1981), Ahenakew (1987a), for example). Another possibility is for the  Speakers (all dialects) often explain the difference between e- and kd- as being a matter of present vs. past tense respectively. However, in my experience this tense distinction is consistently disregarded in the elicited sentences. The e- form can be elicited using either past or present tense. 6  66  wh-word to occur with a verbal clause that has no complementizer, the so-called Independent Mode shown in (16.c). As we shall see, this third possibility is much more restricted. We first examine conjunct mode complementizers in more detail, looking at the function of the kd- complementizer in a variety of structures (Section; then we will turn our attention to the e- complementizer (Section We close with a brief discussion of the independent mode (Section It should be noted that, in this section, only examples with agreeing wh-words are given. This is done in order to present the range of complementizer variation more clearly. (See Chapter 5 for discussion and analysis of how wh-agreement interacts with complementizer variation.) Complementizer Art-  Complementizer kd- is not obligatory in elicited wh-questions; however, kd- occurs more frequently with more northerly conservative speakers (like Bill) who gives the kd- form in Nehiyawewin with the simple elicited question "Who kissed John". One speaker (Donna) uses the kd- form only with the focussed English 'who is it that...' elicitation form. The complementizer kd- is termed by Ellis (1983) a restrictive subordinator. Complementizer kd- is obligatory in clefted or focussed NP constructions in Nehiyawewin (as in (17.a) and (b)) and also in relative clauses as shown in (17.c); otherwise, the usual complementizer is e-. (17.a) [ekoni Mary-wa] John the very one Mary-obv John It's Mary that John saw.  ka-wapam-a-t rel-see-dir-3  67  b) [John ana ] ka-wapam-a-t Mary-wa o-kawT-yi-wa John that one rel-see-dir-3 Mary-obv 3-mother-obvP-obv It's John that saw Mary's mother. J.750 c) [naha napew ka-sakih-a-t Mary-wa ] ocem-e-w that (dist) man rel-love-dir-3 Mary-obv kiss-dir-3 [That man who likes Mary] he kissed her. cf.B.74 What do focussed NP constructions and relative clauses have in common? In English, both these constructions may involve empty operator movement to Spec CP; and both involve an NP located outside the clause as antecedent of the operator-variable chain. In Nehiyawewin, focus constructions and relative clauses have an obligatory kdcomplementizer. Note that this complementizer does not occur in ordinary complement-type subordinate clauses in the language. Therefore, there must be some link between its obligatory occurrence in relative or cleft-type constructions and the operator movement typical of these structures.  7  Structures with deictics such as ana 'that' are also found with wh-phrases, and the presence of the deictic seems to have a focussing effect. With this form of the wh-phrase, only the kd- complementizer is licit, as illustrated in (18). (18.a) awina ana ka-ocem-a-t John-a who that rel-kiss-dir-3 John -obv Who is it that kissed John? D. 17 b) *awtna ana e-ocem-a-t John-a who that conj-kiss-dir-3 John-obv Who is it that kissed John? cf.D.17  .3-wh—> 3'  3-wh —> 3'  When a clefted English sentence is elicited as in (17.a), the Nehiyawewin translation always has kd- and some form of [NP NP] structure sentence initially, (a) has ekoni Mary-wa 'Mary (is) the very one'. Alternately the cleft might be Mary-wa ekoni 'the very one (is) Mary' or some speakers use a form with a deictic as shown in (b). We will see in Chapter 6 that deictics like ana/anihi 'that (one)' have focussing properties. 7  68  The contrast in (18) is very important. The occurrence of the deictic is associated with the focussed or relativized structures in (17), which are environments requiring operator movement. Given that wh-questions must also have some form of operator movement, we must assume that operator movement is allowed with both e- and kd- complementizers. Complementizer kd- also occurs (though not obligatorily) with other forms of the whexpression; for example: (19.a) awinihi Mary ka-pikiskwat-a-t 3 —> 3'-wh who (obv) Mary rel-speak to-dir-3 Who is Mary talking to? cf. B.294 Who is he [(the one) that Mary is talking to]? b) awinihi e-itwe-yan ka-pakamahw-a-t who(obv) conj-say so-2 rel-hit s.o.-dir-3 Who did you say that s/he hit? B.36  3 —> 3'-wh  It is plausible that, in general, kd- clauses involve operator movement in the verbal clause (as reflected in the alternate glosses given above): this is consistent with the fact that they occur in relative clause environments and focus environments as well as wh-questions. In addition, the parallel between clefted focus constructions (see (17)) and wh-clefts (see the wh- examples in (14) and (18)) suggests that wh-questions can be parsed into two clauses ~ a copular-type nominal clause typical of cleft constructions plus the verbal clause. Within the copular structure, the first part constitutes the wh-word while the second part contains the DP subject — with the verbal clause adjoined somewhere inside the nominal clause. Much of Chapter 4 is devoted to motivating this claim.  69 Complementizer eWhile complementizer kd- is obligatory in relative clauses and focussed NP constructions, it is not obligatory in wh-questions. In fact, a wh-question may occur with either a kd- or an ecomplementizer. (20.a) awinihi Mary ka-ocem-a-t who (obv) Mary rel-kiss s.o.-dir-3 Who did Mary kiss? D.136.b  3 —> 3'-wh  b) awinihi Mary e-ocem-a-t who (obv) Mary conj-kiss s.o.-dir-3 Who did Mary kiss? D.136.a  3 —» 3'-wh  The problem which presents itself here is: if wh-questions with kd- share something in common with clefted or focussed NPs, then what are the structural properties of whquestions when the e- complementizer is used? In addition to appearing with wh-questions, the e- complementizer occurs with ordinary complement-type clauses (21.a), and sometimes even in main clauses (21.b). (21.a) kahkiyaw napew iteyiht-am-(w) e-takahkapewi-t all man think-th-(3) conj-handsome-3 Every man thinks he is handsome. D.191.b  b)  mohkoman e-ohci-nipa-iso-t knife conj-with-kill-reflex-3 He killed himself with a knife  D. 114  While the kd- complementizer clearly correlates with operator environments, the status of the e- complementizer is less clear, especially as regards the possibility of null-operator movement. This question is taken up in detail in Chapter 5.  70  A s with the kd- complementizer, the contrasting examples in (22) show that the whword is also restricted to clause-initial position with e- complementizer. (22.a) awinihi Mary e-wa-wapam-a-t who (obv) Mary conj-redup-see s.o.-dir-3 Who is Mary seeing (i.e., dating)? b) *Mary e-wa-wapam-a-t awinihi M a r y conj-redup-see s.o.-dir-3 who (obv) Who is Mary seeing (i.e., dating)?  3 —> 3'-wh D . 132 3 —> 3'-wh D . 132  c) * e-wa-wapam-a-t awinihi Mary 3 —> 3'-wh conj-redup-see s.o.-dir-3 who (obv) Mary Who is Mary seeing (i.e., dating)? D.132 If the wh-word is construed with an A-position in an embedded clause, there are two possibilities: either the wh-word occurs in clause-initial position o f the embedded clause (23.a), or it occurs in sentence-initial position (23.b). (23.a) namoya ni-kiskeyiht-e-n [awina e-wlhtamaw-a-t] neg 1-know s.o.-dir-lsg [who conj-tell s.o.-dir-3] I don't know who told him. B.393 b) awina e-itwe-yan e-pakamahw-isk who conj-say so-2 conj-hit s.o.-3>2 Who did you say hit you? B.141 (23.b) appears to involve long-distance movement o f the wh-word itself. However, in my analysis in Chapter 5,1 argue that there is long-distance movement o f a null operator (within the e- clause) — and that the e- complementizer occurs in environments where two clauses are conjoined.  71 Independent mode The evidence provided for wh-questions above involves the conjunct mode which is much more commonly used — not only for wh-questions, but in general (see Chapter 2). However, wh-questions in Nehiyawewin may also occur with the independent mode which does not involve an overt complementizer. This is a marginal form for wh-questions and there is considerable variation among speakers. More conservative speakers (i.e., Bill) use this form only for rhetorical questions (24.b). Younger speakers (like Donna), on the other hand, use the form more freely as in (24.a), for example: (24.a) awina ekote ki-wapam-a-w who there 2-see s.o.-dir-3 Who did you see there?  D.66  b) awina kiskeyiht-am-(w) tanehki ka-sipwete-t who know s.t.-th-(3) why conj-leave-3 Who knows why he left? B. 169 As suggested above, the interpretation available to the independent forms is different from the conjunct forms. Older speakers typically use the independent form for questions which are either rhetorical or irrealis in nature, i.e., 'Who loves his mother?' ~ or in questions like (b) which can be interpreted in the idiom (of English, at least) as non-questions. The embedded question with 'why', as usual, involves kd-; however, it is the independent initial verb which is of interest here. Depending on the situation and the intonation/stress used, (24.b) can have the following interpretations in English: (i)  Which one of you knows why he left?  72  (ii)  Nobody really knows why he left.  (iii)  Does anybody know why he left?  In the case of the younger speaker (Donna) who uses the form more liberally, the contrast between the independent and the conjunct forms seems to involve differences in presupposition. According to her description, the questioner has no knowledge of or preconceived notion about the issue which is being questioned using the independent mode. It is an "out of the blue" question, i.e., there is no discourse-linked interpretation (Pesetsky 9  1987). Another speaker (Mary Ann) feels the use of the independent mode is most appropriate if the event (i.e., a fight, for example) is happening now, while the conjunct forms (both e- and kd-) are used to refer to some event in the past. The primary focus of this thesis concerns the properties of conjunct mode whquestions with kd- and e- complementizers. The syntax and semantics of independent mode wh-questions is an area which requires much more research.  3.1.4 Summary In summary, we have seen that wh-questions show the following properties: (i)  the wh-phrase is clause initial.  Note in the translations that the quantifiers nobody and anybody may be used to translate awina, suggesting that this use of awina in Nehiyawewin is related to polarity. Compare this with evidence for Moose and Swampy Cree in Section 3.2.4 below.  8  A discourse-linked interpretation requires a discourse context to determine the precise referent or meaning. 9  73  (ii)  the wh-word agrees or not with with its corresponding argument in the verbal complex.  (iii)  a wh-question uses either complementizer, e- or kd- in the conjunct mode. (The independent mode, though also used, is less common.)  In the following sections, we consider three possible hypotheses concerning the relation of the wh-word to the verbal clause. In Section 3.2,1 look at the wh- in situ hypothesis. In Section 3.3,1 look at the wh-movement hypothesis; and in Section 3.4,1 consider the Null-operator movement hypothesis.  3.2 The Wh-m situ Hypothesis In this section, we will consider some of the characteristics of wh-in situ languages which are evidenced in Nehiyawewin and other Cree dialects, specifically Swampy Cree (the N-dialect spoken in northern Manitoba and Ontario) and Moose Cree (the L-dialect spoken around Moose Factory, Ontario). In a wh-w situ language, a wh-word remains in argument position of the verb at Sstructure rather than moving to Spec CP. This analysis has two parts: first, the wh-word is in A-position; and second, the wh-word does not move. According to Cheng (1991), wh-w situ languages have the following properties: (i)  the wh-word is in A-position (at S-structure);  (ii)  yes/no questions are associated with a question (Q) particle;  (iii)  wh-questions are associated with a wh- Q particle; and  74  (iv)  wh-words are polarity items, i.e., they are ambiguous between an interrogative interpretation 'what' and an indefinite interpretation ' something/anything'.  A language which exhibits all of these properies is Chinese, which is generally considered to be a wh- in situ language. Wh-words occur only in argument positions, (25). Yes/no questions are associated with a sentence-final Q-particle ma, (26.a). Wh-questions are associated with an optional sentence-final Q-particle ne, (26.b). Wh-words may also be interpreted as indefinites if they are under the scope of an operator, such as negative mei, as in (27), or the yes/no Q-particle ma in (26.a). (25)  botong kan-wan-le sheme Botong read-finish-ASP what 'What did Botong finish reading?'  (26.a) jialuo mai-le sheme ma Jialuo buy-asp what QY/N Did Jialuo buy anything? b) shei mai-le sheme (ne) who buy-ASP what (Q h) 'Who bought what?' W  (27)  jialuo mei-you mai sheme Jialuo not-have buy what Jialuo did not buy anything? The remainder of this section examines to what extent these four properties are  attested in Nehiyawewin. I will also discuss material from Swampy Cree and Woods Cree dialects — both of which show stronger evidence with respect to the ambiguity of wh-words.  75  3.2.1 The Wh-word is not in Argument Position We have noted in Chapter 1 that Nehiyawewin is a free-word-order language, and we have observed the obligatory clause-initial position for wh-words in the language. However, given that overt NPs are themselves arguably not in A-position (at least at S-structure, see Chapter 1, Section 1.2.1), the fact that wh-words have a fixed position does not necessarily indicate they are not occupying an A-position. In order to determine whether Nehiyawewin is, or is not, a wh- in situ language, it is therefore necessary to examine the other wh- in situ properties.  3.2.2 Yes/No Question Particles Nehiyawewin has some of the properties of wh-/w situ languages. For example, there is a Qparticle ci for yes/no questions. (28) Nehiyawewin (Plains Cree): a)  awiyak ci e-pahpi-t somebody Q conj-laugh-3 Did anybody laugh? B.128  b)  Mary ci eka e-sakih-a-t ni-tosk-astotin Mary Q neg conj-love-th-3 1-new-hat Doesn't Mary like my new hat? B.331  The Q-particle typically occurs in second position in the sentence. The corresponding Qparticles for Swampy Cree is nd and in Moose Cree it is na.  76  (29) Swampy Cree: ki-(t)asoniyam-in na 2-have money-sg Q Do you have any money? (30) Moose Cree: (Ellis 1983:29) swap-ihk na ihta-w store-loc Q be there-3 Is she at the store?  3.2.3 Wh-Question Particles According to Cheng's analysis, if there is a Q-particle for yes/no questions, then there must also be a wh-question particle. Cheng allows for the possibility that the wh- Q-particle may be overt or null (non-overt). At first glance, Nehiyawewin seems to counterexemplify this claim, since wh-words such as awina 'who' and kikwdy 'what' do not co-occur with a Qparticle. (31) Nehiyawewin (Plains Cree): a)  awini-wa John e-wikim-a-t who-obv John conj-marry-dir-3 Who did John marry? J.798  3 —» 3'-wh  b)  kikway John ataw-e-w what John buy-th-3 What did John buy?  3 —> 0-wh B.519  However, although wh-words construed with A-positions are not associated with a Qparticle, wh-adverbials do contain a morpheme which is a candidate for a Q-particle. The  77  table in (32) illustrates the prefix morpheme tan- which, when added to an independentlyoccurring adverbial, creates an adverbial wh-word. This could be argued to be a Q-particle: (32) Wh-Adverbials vvhtantantantantantan-  Adverbial + + + + +  ite ita isi ispT tahto ?  Wh-adverbial  'there, thither' = 'there' = 'thus' = 'then' = 'as many' =  tanite tanita tanisi tanispT tantahto tanehki  'where, whither' 'where' 'how' 'when' 'how many' 'why'  The same morpheme tan- is used with pronominal agreement suffixes to form a paradigm for 'which (one)'. The table in (33) shows this set (in the first column) in conjunction with the 'who/what' paradigm and with a demonstrative paradigm: (33) 3 3pl 3' 0 Opl  'which'(one) tan- i tan- i-ki tan- i-hi tan- i-(ma) tan- i-hi  'who/what' awin- a awtn- i-ki awin- i-hi / -wa kikway kikway-a  'this' aw-a 6-ki 6-hi 6-ma 6-hi  Consistent with Cheng (1991), in yes/no questions, a Q-particle is evident in the form of second-position ci. However, on the basis of the surface evidence, it remains unclear whether or not wh-words, as a class, are associated with a Q-particle. Given that whadverbials contain the Q-particle tan-, the status of awin- 'who' and the kikway 'what' series, which appear to lack an overt Q-particle, becomes moot. On the basis of the surface evidence, one could conclude that wh-questions lack a Q-particle, and that Nehiyawewin  78  therefore does not have the properties of wh-in situ laguage. The surface morphological evidence is consistent with either analysis.  10  3.2.4 Ambiguity of wh-words Another of the characteristic features of in-situ wh-languages is that wh-words exhibit polarity for interrogativity (Cheng 1991:123 ft). In other words, they may have an interrogative interpretation or they may be indefinite. At least the potential for this ambiguity exists in some dialects of Cree, e.g., Moose Cree kekwdn 'what, something' (Ellis 1983:70-71, 127). Examples (34.a) and (b) illustrate the interrogative use of inanimate kekwdn 'what' in the proximate and obviative forms respectively. The wh-word occurs sentence initially and always with the conjunct form of the verb. The indefinite use of kekwdn 'something' 'something' is shown in (c) with independent mode. (34) Moose Cree: a)  kekwan n'tawel't-am-an What want s.t.-th-2sg What do you want?  b)  (conjunct Vb)  kekwaliw n'tawel't-ah-k what (obv) want s.t.-th-3 What does he want?  (conjunct Vb)  Under either analysis, there remains the question of why wh-words associated with adverbial positions are consistently associated with the wh-typing morpheme {tan-), while wh-words associated with an argument position (e.g., awin- 'who' and kikwdy 'what') do not have an (overt) wh-typing morpheme. 10  79  c)  ki-n'taweTt-e-n na kekwan 2-want s.t.-th-2sg Q. something Do you want something?  (independent Vb)  Here the indefinite pronoun occurs after the verb. In other words, the position of the word kekwdn disambiguates between its two possible meanings. Example (c) is a yes/no question, and the Q-particle na occurs in second position in the sentence. The following examples illustrate the same pattern for the wh-word awenihkdn 'who/somebody' for Moose Cree: (35) a)  b)  Moose Cree: awenihkan weyapam-at anta who see+I.C.-2>3 there Whom do you see there?  (conjunct Vb)  ni-wapam-a-w awenihkan walawttimi-hk l-see-dir-3 something outside-loc I see somebody outside.  (independent Vb)  Once again, the wh-word awenihkdn verb. The indefinite awenihkdn  'who' occurs sentence initially and with the conjunct  'somebody' shown in (b) occurs after the verb which is in the  independent mode. The position of the wh-word disambiguates between the interrogative and the indefinite meaning. In addition, the wh-question uses the conjunct form of the verb while the independent verb occurs in the examples provided for indefinite readings. In Swampy Cree, only (proximate/obviative) kekwdn/kekwdniw  'what/something' is  ambiguous while awena 'who' has only interrogative meaning (contrast awiyak 'somebody'). (cf. Glossary of Wolfart 1988; Cook-Neff (undated).)  80  According to Reinholtz and Russell (1995), it is solely the Independent vs. Conjunct verb strategy which is used to disambiguate between the two interpretations of kekwdn 'what, something' in the Swampy Cree examples in (36). (Presumably, however, the word ordering in the (a) example is free.) (36) Swampy Cree: a)  kekwan ki-ki-wapahten what 2-past-see it You saw something. * What did you see?  (independent vb.)  b)  kekwan ka-ki-wapaht-an what rel-past-see it-2  (conjunct vb.)  What did you see? Regardless of the sentence-initial position of kekwdn, the above sentence with the independent verb form can only have the indefinite reading. Wh-questions with kekwdn 'what' always occur with the conjunct verb form, while the indefinite reading kekwdn 'something' can be obtained only with the independent form of the verb. The other set in Swampy Cree is awena 'who' and awiyak 'somebody'. In Nehiyawewin (Plains Cree), neither wh-word ('who' or 'what') has a form which is homophonous with the indefinite pronominal. The awina 'who' awiyak 'somebody' forms are similar to the Swampy Cree set. Both examples involve the conjunct mode of the verb. (37) Nehiyawewin (Plains Cree): a)  awina e-pah-pahpi-t who conj-redup-laugh-3 Who is laughing? MA.259 ^Someone is laughing.  b)  81  * e-pah-pahpi-t awina conj-redup-laugh-3 who *Who is laughing? ^Someone is laughing.  The interrogative awina 'who' is restricted to clause-initial position. With clause-final position as in (b), the sentence is ungrammatical with either interrogative or indefinite interpretation. With indefinite awiyak 'somebody', either ordering is grammatical, i.e., it acts like an ordinary NP. (38) Nehiyawewin (Plains Cree): a)  awiyak e-pah-pahpi-t b) somebody conj-redup-laugh-3 *Who is laughing? Somebody is laughing. cf.MA.259  e-pah-pahpi-t awiyak conj-redup-laugh-3 somebody *Who is laughing. Somebody is laughing.  The interrogative awina 'who' and indefinite awiyak 'somebody', although related, are clearly distinct forms. A similar pattern is attested for the inanimate wh-word kikwdy 'what', which contrasts with the indefinite kikwey 'something'. Although, kikwdy 'what' is segmentally similar to kikwey 'something', the vowel quality is consistently distince (d vs. e).  n  The wh-  word is restricted to clause-initial position, and is only compatible with a wh-interpretation. (39) Nehiyawewin (Plains Cree): a)  kikway ataw-e-w what buy-th-3 What did he buy? cf.B.519 *He bought something.  b)  * ataw-e-w kikway buy-th-3 what *What did he buy? *He bought something.  The forms as reported for Nehiyawewin in various sources are: Wolfart (1973) Ahenakew (1987) Blain 'what' kikway kikway kikway 'something' kikway kikway kikwey In the form which I write as kikwey 'something', I have used the vowel quality I hear invariably from both Alberta and Saskatchewan speakers. I expect that Ahenakew's version, kikway, sounds the same as my version. Wolfart's form, kikwdy, seems to suggest a stress shift. 11  82  As for the indefinite kikwey 'something', like other NPs it may appear either before or after the verbal complex, and is compatible only with an existential interpretation. (40) Nehiyawewin (Plains Cree): a)  kikwey ka-wThtamaw-iti-n b) something fut-tell it-inv-l>2 I will tell you something. M A . 197.a *What did I tell you?  ka-wlhtamaw-iti-n kikwey fut-tell it-inv-l>2 something I will tell you something *What did I tell you?  It is evident that Nehiyawewin has distinct interrogative and indefinite forms. However, the wh-word and the corresponding indefinite are so similar that there is no doubt they were once the same. While Nehiyawewin wh-words are never homophonous with indefinites, other Cree dialects display either partial or complete homophony, as can be seen from the chart in (41). Like Nehiyawewin, Swampy Cree has distinct forms for 'who' and 'somebody', awena and awiyak respectively. However, the swampy Cree forms for 'what' and 'something' are homophonous, both being realized as kekwan.  Moose Cree shows  complete homophony: the forms for 'who' and somebody' are homophonous (awenihkan), are the forms for 'what' and 'something' (kekwdn). (41) Wh-words and Indefinites Dialect  Wh-word  Indefinite  Nehiyawewin: (Plains Cree)  awina kikway  •who' 'what'  awiyak kikwey  'somebody' 'something'  Swampy Cree:  awena kekwan  'who' 'what'  awiyak kekwan  'somebody' 'something'  Moose Cree:  awenihkan kekwan  'who' 'what'  awenihkan kekwan  'somebody' 'something'  83  as  If the dialectal differences summarized in (41) are taken into consideration, then one is led to the conclusion that, historically, wh-words and indefinites were homophonous. However, at least for Nehiyawewin, this is no longer the case synchronically. Consequently, with respect to Cheng's claim that a wh- in situ language will have homophonous wh-words and indefinites, one can conclude that Nehiyawewin wh-words do not have this property. In addition, even the Cree dialects where (partial or complete) homophony exists between whwords and indefinites, the wh-environment is distinguished by: (i) obligatory initial position, and (ii) the use of the conjunct verb form. By contrast, indefinites behave like ordinary NPs: (i) they can occur either before or after the verbal complex, (ii) they freely occur with the independent form of the verb.  12  3.2.5 Evaluating the W h - in situ Hypothesis The table in (42) summarizes Cheng's wh-/« situ diagnotics for the three Cree dialects discussed in this section: Nehiyawewin, Swampy Cree and Moose Cree. (42)  Nehiyawewin Q-Particle: Y / N Q-Particle: WhAmbig: who/someone what/something Free Ordering of wh-word  yes ? no no no  Swampy C .  Moose C .  yes ? no yes no  yes ? yes yes no  Whether or not these forms constitute polarity items is a matter for further study. However, note in Section 3.1.3..3 with respect to independent mode in wh-questions, that in some subdialects there is an indefinite interpretation in "rhetorical" interpretations of a question. 12  84  No Cree dialect clearly satisfies all of the wh- in situ diagnostics listed in (42). For example, with respect to the claim that, in a wh- in situ language, both yes/no questions and wh-questions will be associated with a Q-particle, the surface evidence is inconclusive: although yes/no questions are uncontroversially marked by a second-position Q-particle (ci in Nehiyawewin, na in Moose Cree and nd in Swampy Cree), the existence of a Q-particle in wh-questions is moot. With respect to the claim that a wh- in situ language will have wh-polarity items (i.e., forms that are ambiguous between a wh-interpretation and indefinite interpretation), there is considerable dialectal variation: wh-forms and indefinites are homophonous in Moose Cree, partially homophonous in Swampy Cree, and completely distinct in Nehiyawewin. Finally, inasmuch as a wh- in situ analysis predicts that wh-words will occupy the same positions as ordinary NPs, note that wh-words are constrained to clause-initial position (in contrast to the free ordering of ordinary NPs, (cf. Section 3.1). This strongly suggests that whatever position wh-words occupy, it is not the same position occupied by ordinary NPs. In sum, Nehiyawewin is not a wh- in situ language.  3.3 The Wh-movement Hypothesis In order to properly evaluate whether or not Nehiyawewin wh-questions are formed by whmovement, it is important to distinguish (overt) wh-movement from null-operator movement, a distinction which was introduced in Chapter 1. English relative clauses exploit both strategies (43) but English wh-questions are only compatible with overt wh-movement, (44).  85  (43.a) This is [ the man [ who; [ I talked to t  ]]]  b) This is [ the man [Op; [ I talked to t; ]]] (44.a)  [Who; did [John see ti ]]]  b) *[Opi did [John see t; ]]] This section (3.3) considers whether there is overt wh-movement in Nehiyawewin, and concludes that there is not. However, wh-questions must have some form of operator movement in order to have wh-interpretation. In the following section, I argue that, if Nehiyawewin wh-questions do not have overt wh-movement, then they must have nulloperator movement, i.e., the wh-word is linked to a null operator.  3.3.1  Wh-movement as A ' - b i n d i n g  Overt wh-movement involves the movement of a wh-word out of an argument position to Spec CP. The wh-word in operator position c-commands and binds its trace in an operatorvariable relationship. Because the wh-word occupies a non-argument (A ) position, this is an 1  instance of A'-binding. (45)  Who; did John see tj?  In classical government and binding theory (Chomsky 1981, 1982, 1985), a wh-trace is the non-overt counterpart of an R(eferential)-expression. At first glance, treating the trace of wh-movement as a kind of R-expression might seem counterintuitive. The treatment of wh-traces as R-expressions is motivated by binding theory. Ordinary R-expressions are free  86  within the domain of their clause, a so-called Condition C effect of binding theory. The same is true of wh-traces, (46).  13  (46.a) *[Shei saw Mary; ] b) *[who i did [she; see t; ] With the system of NP features proposed by Chomsky (1981), ordinary R-expressions and wh-traces are both defined as [-pronoun, -anaphor]. They are distinguished from each other by means of the features [± overt]: R-expressions being [+overt] and wh-traces being [-overt]. This classification fits into the system of NP types summarized in (47).  (47). NP features Pron  Anaph  Overt  +  -  +  +  -  -  + +  + -  lex. anaphors/reflexives NP trace  +  R-expr. // Resump. pron. wh-trace//pro  pronouns pro  14  + +  + +  + -  PRO  Example (46.b) illustrates Strong Crossover: a wh-word cannot be co-indexed with a pronoun that c-commantds it. The relevance of Strong Crossover to the wh-movement hypothesis is taken up in more detail below (Section 3.3.3). 13  pro is typically classified as a [-overt] pronoun ([+pron] [-anaph]) as seen above. We have seen that a resumptive pronoun ([-pron] [-anaph]) occurs in an A-position A'-bound by a leftdislocated NP (cf. Cinque 1990). Therefore, the question arises whether the empty category in Aposition coindexed with an IP-adjoined overt NP might be its [-overt] counterpart. 14  87  Note that movement of a wh-word creates the operator-variable chain <wh, t>. By contrast, in null-operator movement, it is the movement of a pro which creates the operator variable chain (cf. Browning 1982, Cinque 1990). Unlike a quantificational operator which has a range, the null-operator has no range and therefore requires an antecedent in order to fix its range. In Nehiyawewin, the clearest evidence of the absence of overt wh-movement involves the prohibition of multiple wh-questions.  3.3.2 Multiple Wh-questions In a language where wh-words are generated in argument position — independent of whether the wh-word remains in situ or subsequently undergoes movement (as in English) — the occurrence of multiple wh-words is possible. In wh- in situ languages such as Chinese (cf. Cheng 1991), both wh-words occur in argument positions, (48). Another possibility is for both wh-words to undergo overt movement, e.g. the Slavic languages (Rudin 1988) as well as Mohawk, an Iroquoian language (Baker 1996:92, fn.38). Mohawk examples are provided in (49). A third possibility is for one wh-word to move and the other to remain in situ, e.g., English where the wh-word in subject position moves (string vacuously) at S-structure, while the wh-word in object position remains in situ and moves at LF, (50). (48)  W h - in situ:  Chinese  shei mai-le sheme (ne) who buy-ASP what (Qwi.) 'Who bought what?'  88  (49)  Multiple Wh-movement: Mohawk  a) tak-hrori uhka nahotv wa'-e-hninu-' 2sS.imper/l/sO-tell who what fact-FsS-buy-punc Tell me who bought what. b) tak-hrori nahotv uhka wa'-e-hninu-' 2sS.imper/l/sO-tell what who fact-FsS-buy-punc Tell me who bought what. (50)  Wh-movement & W h - in situ: English Who bought what? Thus, multiple wh-questions differ according to whether the language exploits a  uniform in situ strategy (e.g. Chinese), a uniform movement strategy (e.g. Mohawk), or a mixed strategy (English). However, not all languages permit multiple wh-questions. In Berber and Italian, Calabrese (1984, 1987) has observed that multiple whquestions are not allowed. Example (a) is from Italian and (b) is from Berber (Calabrese 1987:118): (51 .a) *Che cosa hai dato a chi? What did you give to whom? b) *maymi m-ay t-sghu terbalt? why what-that 3fs.bought girl Why did the girl buy what? Calabrese has argued that the absence of multiple wh-words is a diagnostic for clefts in those languages. In Calabrese's analysis, a wh-word must be a focussed NP, i.e., Who is it  89  and states, "the cleft position is an argument internal to the copular predicate, and  [that...];  in a copular predicate there can be only one argument (1987:118).  1,15  Multiple wh-questions are disallowed in Nehiyawewin. Consider the following examples: (52.a) * awina e-itwe-t kikway who conj-say so-3 what Who said what?  cf.J.478  b) * awina ka-ptkiskwat-a-t awtna-wa who rel-speak to s.o.-dir-3 who-obv Who spoke to whom? D.283 A multi-clausal sentence can have two wh-words; however, they cannot both be associated with the same verb. (53)  awina kiskeyiht-am-(w) tanehki ka-sipwete-t who know s.t.-th-(3) why rel-leave-3 Who knows why he left? B. 169  Each wh-word occupies clause-initial position in a separate clause. I therefore propose that Nehiyawewin wh-phrases, like focussed NPs, are clefted. Consider the following clefted sentences: (54) a)  Clefted NPs: eko John [Mary ka-wapam-a-t] the very one John Mary rel-see-dir-3 It is John [that Mary saw].  It stands to reason that both arguments in a sentence cannot be clefted or focussed — unless, of course, one of them can remain in argument position at S-structure as in the following English example: No, [it is ME ] who hit HIM. In this sentence, thefirst-personpronoun is focussed structurally in a cleft (and possibly also with stress), while the object HTM is focussed by stress alone while remaining in object position. 15  90  b)  John ana [Mary ka-wapam-a-t] John that (one) Mary rel-see-dir-3 It is John [that Mary saw].  c)  Wh-Cleft: ana [Mary ka-wapam-a-t] who that (one) Mary rel-see-dir-3 Who is it [that Mary saw]? awina  In the bolded clefted structures in the English glosses, the NPs John and who are in an equative relation (which requires a copula) with a dummy subject, /'/. Nehiyawewin does not use a copula, so the equative structures in Nehiyawewin in both cases simply involve two NPs. Rather than a dummy pronoun, I propose that Nehiyawewin can use an empty pronoun pro as one of the nominals — with or without deictic ana 'that' as shown in (c). These issues will be discussed at length in Chapters 4 and 5. In summary, I have shown that wh-words in Nehiyawewin are always clause initial ~ a  fact which argues against the possibility of wh-w situ. I have also argued that the absence  of multiple wh-questions provides evidence that wh-words do not originate in the argument positions of a verb. In line with Calabrese's (1984, 1987) claim that theabsence of multiple wh-questions in a language is indicative of wh-clefts, I propose that Nehiyawewin wh-words are clefted.  16  In the following chapters, several patterns of wh-questions are discussed. My analysis provides different structural accounts for the different patterns ~ and not all of them are, in fact, "clefted". Therefore, the use of the term here is to be taken loosely to signify that the wh- NP does not occur internal to the verbal clause. 16  91  In other words, the wh-word is contained in a equative structure, i.e., "Who is it [that...]'. As we shall see in Chapter 4, this nominal clause may stand alone or be combined with another verbal clause in a variety of patterns. The absence of multiple wh-questions in Nehiyawewin suggests there is no whmovement.  3.3.3 Strong Crossover Another diagnostic of wh-movement is Strong Crossover (SCO), introduced above in Section 3.3.1. The term describes the strong ungrammatically which results if a wh-trace is bound by a c-commanding pronoun. Representative examples of SCO are given in (55). (55.a) * Whoi did [shej see ti ] b) * Who; did [she; think [ she; saw ti ]] c) * Whoi did [she; think [t; saw you ]] Unfortunately, SCO cannot be used as a test for wh-movement in Nehiyawewin because of the confounding factor of the direct/inverse contrast.  This claim is not new. Wolfart (1973:34) states that the wh-word awina 'who' i) may stand alone as a complete utterance; ii) may occur in an equational sentence; or iii) it may function predicatively "with a conjunct clause depending on it", i.e., a verbal clause with e- or kd- complementizer or "conjunct marker". As noted in Chapter 1, clefted wh-questions have also been proposed by Johns (1982) for Ojibwa and by Reinholtz and Russell (1995) for Swampy Cree — though they do not consider the consequences of this proposal in detail. 17  92 Strong Crossover in Nehiyawewin Strong Crossover effects cannot be checked in Nehiyawewin, as shown in the following examples. In this respect, it is important to recall (from Chapter 2) the role played by the direct/inverse morphology as well as the distinction between proximate and obviative third persons. The example in (56.a) involves three argument positions with two third persons, one being proximate and the other obviative. The coindexed pronominals have proximate status, while Mary, the other third person, is obviative. Proximate status is assigned initially to the pronominal subject in the main clause of (56.a) for two reasons: (i)  18  In the elicitation context, it is the first third person used so it will typically be assigned the unmarked status.  (ii)  Add to this the fact that this argment is the main clause subject (i.e., sentencial topic) and the fact that it is a pronoun and must be "old information' — all these are properties of proximate third person.  (56.a)  Hei said [ Mary likes him; ] prox  obv  \  Inverse  In the subordinate clause, the subject Mary is obviative and is disjoint in reference from the matrix clause subject (and obligatorily disjoint in reference with the subordinate clause object). The object pronoun — which is coindexed with the (proximate) subject of the  Also, in a discourse context, there can only be one proximate third person in a given span, whereas there ma be more than one obviative third person. Therefore, coreference is a natural (though not obligatory) interpretation of proximate arguments (cf. Grafstein 1984). 18  93  main clause -- must also be proximate. Accordingly, the verb in the subordinate clause with an obviative subject and a proximate object will be inverse. The Nehiyawewin equivalent for the sentence in (a) with the coreference shown is given in (56.b). (56.b) e-itwe-t Mary-wa e-miyweyim-iko-t conj-say-3 Mary-obv conj-like-inv-3 He; said Mary likes him; prox  obv^  ^pr°  3 —» 3' Inverse  x  Inverse The corresponding wh-question is shown in (57). The SCO English gloss is ungrammatical; however, the Nehiyawewin sentence is perfectly good with the licit English interpretation shown. (57)  awina e-itwe-t Mary-wa e-miyweyim-iko-t who conj-say-3 Mary-obv conj-like-inv-3 Whoj did he; said Mary likes him; ? »  3 —> 3' Inverse  Who; tj said Mary likes him; ?  Since the wh-word is proximate and the subject of the main clause is proximate, they must be the same person.  3.3.4 Weak Crossover Weak Crossover (WCO) defines situations where wh-movement takes place over a complex DP containing an embedded coindexed pronoun as in the examples in (58.b-c). The simple pronominal DP in subject position of the SCO example in (58.a) is replaced by a possessive DP as subject in (b) and a relative clause DP as subject in (c).  94  (58.a) * Who; did [ [ she ] p see t ] :  D  b) * Who; did [[.hen mother ]  SCO  ;  DP  seet;]  WCO (possessive DP)  c) * Who; did [ [ the woman [ that hei loves ]CP ]DP see tj ] WCO (relative clause) The pronoun in the complex DP does not c-command the trace; but neither does the trace < command the pronoun. In both WCO examples (b) and (c), the coindexation shown is ungrammatical. The WCO evidence is characterized by a subject/object asymmetry with respect to the extraction site. In the ungrammatical examples above, wh-movement takes place object position (passing over'the DP in subject position). When the wh-extraction is from subject position with the complex DP in object position, the sentences are good, as shown in (59). (59.a) Who; [ ti saw [ her, mother ]? b) Whoi [ ti saw [ [ the woman [ that hej loves ] ]? WCO is used as evidence for wh-movement in English; therefore, these WCO examples in other languages should show the same grammaticality contrasts. If the results are different, then this must reflect a difference (structural or otherwise) between the languages in question. In the course of the following discussion, we will see evidence for three sets of contrasts with respect to WCO: (i) Mohawk is different from English; (ii) Nehiyawewin is different from English; and (ii) Nehiyawewin is different from Mohawk. The details of WCO and my analysis for Nehiyawewin are discussed in more detail Chapter 6. The point here is simply to illustrate the difference between Mohawk and Nehiyawewin with respect to wh-movement. Mohawk, like Nehiyawewin, is a rich head-  95  marking language. For both languages, it is claimed that overt NPs do not occupy argument positions at S-structure but are adjoined to IP. This includes possessor DPs and relative clause DPs. Baker (1996) claims that there is overt wh-movement in Mohawk; and we will compare evidence in the two languages. Possessor DPs: Mohawk vs. Nehiyawewin In this sub-section, we see that the WCO facts for both Mohawk and Nehiyawewin differ from the English WCO facts involving possessor phrases. Possessive DPs in Mohawk show no contrast for WCO effects; both are grammatical, in contrast to the subject/object asymmetry found in English. Compare the two sets of examples in Mohawk and English (as represented in the gloss). (60)  19  Mohawk:  a)  Uhka wa'-te-shako-noru'kwanyu-' rao-skare' who fact-dup-MsS/FsO-kiss-punc MsP-friend Whoj kissed his; girlfriend?  b)  Uhka wa'-te-shako-noru'kwanyu-' ako-skare' who fact-dup-MsS/FsO-kiss-punc FsP-friend *Who; did her; boyfriend kiss?  The (a) example is grammatical in both Mohawk and English. The (b) example should not be good, but (b) is good in Mohawk.  In the Mohawk example, the possessor phrase may be located preceding the verb without affecting the results (M. Baker, p.a). 19  96  Nehiyawewin DPs, like those in Mohawk, are adjoined to IP; and WCO also does not hold in Nehiyawewin possessor phrases (Blain 1992). (61)  Nehiyawewin:  a)  awina ka-nawaswat-a-t o-tem-a who REL-chase-dir-3>3' 3-dog-obv Whoi is chasing hisi (own) dog?  b)  awina o-tem-a ka-nawaswat-iko-t who 3-dog-obv REL-chase-inv-3'>3 *Who; is his; dog chasing? B.152 OR,  Whose; (own) dog is chasing him;?  (pref. Eng. gloss)  There is no subject/object asymmetry with respect to the grammaticality of these sentences in either Mohawk or Nehiyawewin in contrast to English. The specifics of Baker's (1996) analysis for Mohawk are discussed in Chapter 6 ~ as is my own analysis for Nehiyawewin. However, Baker claims that there is overt wh-movement in Mohawk; and the evidence so far shows that Mohawk and Nehiyawewin behave in the same manner. Relative Clause DPs: Mohawk vs. Nehiyawewin In this sub-section, we see that the WCO facts for both Mohawk differ from both English and Nehiyawewin WCO facts involving relative clauses.  20  Baker shows that there is no subject/object asymmetry in Mohawk relative clauses and that bound-variable readings for both are ungrammatical. The Mohawk example in (62.a) represents the object-related relative clause which is good in English but bad in  We see only evidence for relative clause in object position. We will see in Chapter 6 that the Nehiyawewin relative clauses in subject position allow two alternative interpretations ~ in contrast to the English evidence. 2 0  97  Mohawk. The (b) example illustrates that the same example in Nehiyawewin is good. It can be concluded from this contrast in the WCO evidence that Mohawk and Nehiyawewin are different. (62.a) *Uhka wa'-t-huwa-noru'kwanyu-' ne rukwe ne ruwa-nuhwe'-s who fact-dup-FsS/MsO-kiss-punc N E man N E FsS/MsO-like-hab Who; kissed the man she; likes? b)  awina ka-ocem-a-t anihi napew-a ka-takakeyim-a-t who rel-kiss-dir-3 that man-obv rel-like-dir-3 Who; kissed the man that shej likes? J.388 Baker's account of the Mohawk weak crossover evidence uses a parasitic gap analysis  to account for the difference between possessor NPs (which are always good) and relative clauses (which are always bad). In the account of his analysis in Chapter 6, we will see that the relevant property of parasitic gaps is that the c-command capacity of the operator in the gap construction is sensitive to subjacency. Relaltive clauses provide a subjacency environment while possessive DPs do not. Given the contrast in (62), this analysis is not available in Nehiyawewin. We have a three distinct situations with respect to the WCO evidence seen above. As shown in (63), Nehiyawewin and Mohawk contrast with English with respect to WCO with possessor phrases; and Nehiyawewin and English contrast with Mohawk with respect to the relative clause evidence for WCO. (63)  WCO Contrasts English  Nehiyawewin  Mohawk  Possessor Phrase  WCO  no WCO  no WCO  Rel. Clause (Object)  no WCO  no WCO  WCO  98  In summary, we have seen from the wh-movement evidence that Mohawk and Nehiyawewin are different in two respects: (i) multiple wh-words are possible in Mohawk but not in Nehiyawewin, and (ii) the WCO evidence is different between the two languages. As noted above (and as will be discussed in Chapter 6), Baker (1996) argues that there is overt wh-movement in Mohawk; therefore, this cannot be the case for Nehiyawewin. There is one more possibility to consider: that wh-questions involve null-operator movement.  3.4 The Null-operator Movement Hypothesis As discussed above, wh-questions require an A'-chain involving a operator and a trace/ variable. This means there has to be an operator and the operator must move. The operator may be the overt wh-word or it may be null; and in either case, the movement may occur either at S-structure or at LF.  (64) illustrates the four possible combinations of these  variables. (64)  Nehiyawewin Wh-operators: Overt Operator  Null Operator  SS movement  * wh-movement  kd- complementizer  L F movement  *wh- in situ  e-  complementizer  We have seen that there is no overt operator in the verbal clause of a Nehiyawewin whquestion, i.e., there is neither S-structure wh-movement of an overt wh-operator nor is there L F movement of an in situ wh-word. That leaves the possibility of a null operator. In  99  Chapter 5,1 discuss null operator movement in more detail and argue for the pattern of movement illustrated in (64).  3.4.1 The Antecedent Condition on Null-operator Movement A wh-word in operator position has inherent semantic content by virtue of its interrogative force: as such it restricts the range of the trace/variable it is coindexed with, (62. a). But a null-operator (pro) does not have inherent semantic content, and so cannot restrict the range of the trace/variable with which it is coindexed. Consequently, the null operator requires an antecedent in order to fix the range of its variable. If the antecedent of the null-operator is a wh-word, this yields a wh-question interpretation, (65.b). (65.a) [ p whi [n>... t; ... ] C  b) [whi ] [CP Op; [ ... ti ... ] IP  This predicts that a wh-word undergoing movement as in (62.a) will be structurally distinct from a wh-word which serves as the antecedent to a null operator. The moved whword is contained in the same clause as the variable that it binds and is predicted to exhibit the usual properties of A'-chains. The antecedent wh-word is outside of the CP which contains the < Op, t > chain. The Nehiyawewin evidence is consistent with (b): wh-words are not in argument position, nor do they move to Spec CP of the verbal clause they are construed with.  100  If Nehiyawewin wh-questions involve null-operator movement, in addition to the standard properties of A'-chains listed above and copied in (i) to (v) below, we also expect them to have a wh-antecedent outside the CP which contains the operator as shown in (vi). (i)  there is an operator in Spec CP  (ii)  a gap is involved  (iii)  the wh-word is related to the gap by movement  (iv)  the relation between the wh-word and the gap is subject to subjacency  (v)  the relation between the wh-word and the gap is unbounded  (vi)  the null operator is linked to a wh-antecedent  In Chapter 4,1 will argue that this wh-antecedent is the predicate of a nominal clause, an [NP (is) NP] construction with a subject and a predicate, both nominals.  3.5 Comments The results of this chapter establish what Nehiyawewin wh-questions are not. In the following chapters, we look at what Nehiyawewin wh-questions are. The first step is to show that wh-words are licensed as predicates of a nominal clause (Chapter 4). Then I argue that this nominal clause is in turn associated with the verbal clause within which null operator movement has applied (Chapter 5). I establish that, despite surface appearances, in Nehiyawewin, the first three properties of wh-constructions hold: the wh-word is in Spec CP (of a nominal clause); there is a gap (in the nominal clause); and the wh-word is related to the gap by movement (specifically, predicate-fronting).  101  The final chapter turns its attention to the remaining two properties of wh-questions as they pertain to Nehiyawewin, namely subjacency effects and (un)boundedness. In Chapter 6,1 will discuss Baker's analysis for Mohawk in more detail as well as presenting my analysis to account for the Weak Crossover evidence in Nehiyawewin.  102  Chapter 4 NOMINAL CLAUSES 4.0 Introduction Chapter 3 began our investigation of wh-questions in Nehiyawewin. We saw that wh-words are not in  situ  in argument position of the verb ~ nor is there overt wh-movement from an  argument position. In fact, I have argued that the wh-word never originates inside a clause containing a verb, but that the wh-word is generated in a position outside of and preceding the verbal clause. I refer to this sentence-initial position as a "clefted" position until the structures can be determined. The goal of this chapter is to investigate the structure of clefted NPs, including wh-words, in Nehiyawewin.' The following examples involve clefted/focussed constructions in Nehiyawewin, for example: (l.a)  John eko [CP ka-sipwehte-t ] John the very one rel-leave-3 [It's John] that left.  b)  eko John [ p ka-sipwehte-t ] the very one John rel-leave-3 [It's John ] that left. C  In these focussed structures, the two NPs, i.e., John and the pronominal eko 'the very one (previously mentioned)', may alternate in their positions preceding the kd- clause (analyzed as a relative clause with null-operator movement). Though both sentences are elicited in the same manner as shown by the gloss, either NP may be in initial position.  103  Another way of focussing an NP involves the deictic ana 'that (one)' as shown in (2). With a deictic DP, only  John  can be in initial position of the nominal clause; the clause-initial  deictic in (b) is not grammatical in a focussed construction, as shown. (2.a)  John ana [ CP ka-sipwehte-t ] John that (one) rel-leave-3 [It's John ] that left.  b)  *ana John [ p ka-sipwehte-t ] that (one) John rel-leave-3 *[It's John ] that left. C  The examples in (3) show a clefted/focussed wh-phrase with deictic ana 'that one'. Again, only the wh-word may occur in initial (predicate) position. The reverse ordering as shown in (b) is ungrammatical. (3.a)  b.  awina ana [CP ka-sipwehte-t ]] who that (one) rel-leave-3 [Who is it] that left? *ana awina [ p ka-sipwehte-t ]] that (one) who rel-leave-3 [Who is it] that left? C  The examples above include both focussed NPs and wh-clefts; and in all the examples, the clefted structures in the English glosses involve a nominal clause in which both the subject and the predicate are nominal constituents.  In all cases, the nominal clause is associated with  an operator-variable structure in the verbal clause. I propose the same analysis for Nehiyawewin, i.e., the two nominal constituents occur in a separate nominal clause which is followed by a verbal clause with the relativizing kd- complementizer. Nehiyawewin, like many languages, does not have a copula in any of these nominal clause constructions (cf. Dechaine 1993).  104  It is necessary to study nominal clause structures in Nehiyawewin for the following reason. Given that the wh-word does not originate inside the clause, the wh- NPs must be licenced by some other means, i.e., by predication inside a nominal clause. This is what occurs in the case of focussed wh-examples in English as shown in (3); and this will be my analysis for all argument-type wh-words in Nehiyawewin. 1  In this Chapter, we will be looking at nominal clauses in both English (Section 4.1.1) and in Nehiyawewin (Section 4.1.2). In the course of the investigation, I will show that the ordering in Nehiyawewin nominal clauses is predicate initial, and that predicate-subject asymmetries in Nehiyawewin reveal a referential hierarchy similar to that for English (Heggie 1988). In this hierarchy, deictic DPs are the most referential and indefinites, including whwords, are the least referential. In Section 4.1.3,1 discuss agreement between the NP constituents of nominal clauses. Predicational agreement in nominal clauses without a verbal constituent (copula) is typically restricted to the inherent lexical features of a noun, e.g., gender and number. However, there is also proximate/ obviative agreement between the subject and predicate in Nehiyawewin nominal clauses. I use this additional agreement feature to motivate predicate fronting — which accounts not only for the agreement facts but also for the predicate-initial ordering in Nehiyawewin. As a result, the sentence-initial wh-word is part of the more general operation of predicate fronting in all nominal clauses.  Argument-type wh-words include only the who and what forms. Adverbial wh-words are not part of this analysis. As non-arguments (without pro in A-positions), it may be the case that they can move to the Spec CP position as in English. Harriida Demirdache (p.c.) notes that this contrast exists with Egyptian Arabic wh-words. 105 1  Based on the analysis of nominal clause structures (including wh-phrases) in this chapter, we will go on to provide an analysis for wh-questions in Chapter 5.  4.1 Nominal Clauses A nominal clause is a sentence in which both the subject and the predicate are nominals, i.e., [NP is NP] as in (4). (4.a) b)  [DP John] is [DP the chief]. [DP The morning star ] is [ p the evening star ] D  There are two types of nominal clause constructions discussed at length in the literature (cf. Higgins 1973, Rapoport 1987, Heggie 1988, Moro 1990, Williams 1994); these are (i) equative nominal clauses, and (ii) predicative nominal clauses. We will look at examples of both types and discuss the properties of each.  4.1.1 English Nominal Clauses In order to provide an analysis for the clefted structures, we must consider the properties and structures involved with the simpler nominal clause as in (4.a). To set the scene, we will consider the theoretical issues by looking at nominal clause structures in English. Nominal clauses involve two nominal expressions in a subject-predicate relationship. One can view these structures in terms of the properties of the subject, which is usually more referential than the predicate (Heggie 1988), or, conversely, at properties of the nominal predicate (Rapoport 1987). Heggie (1988:106) argues for the following hierarchy of reference, based on the subject-predicate asymmetry:  106  Hierarchy of Reference: (Heggie (1988:106)  (5)  deictic > names > definite descriptors > indefinites  [subject] more referential  —> [predicate] less referential  Proceeding from left to right, the constituents are decreasingly referential and more predicate-like. According to this hierarchy, deictic DPs have the highest referentiality ~ they actually point to someone in the discourse. Names and definite descriptors, by definition, refer to a specific individual or unique thing in the discourse ~ while indefinites do not. This hierarchy allows us to compare the properties of equative and predicative nominal clauses. Predicative Nominal Clauses  Thefirsttype of nominal sentence we will look at is the predicative construction. In predicative sentences, there is an asymmetry between the referentiality or rigidity of two nominals. The subject DP must be more rigid/referential than the predicate. This is illustrated in the following sentences: (6. a) b)  [ j He] [ sub  pred  is a moron],  Ubj Mary] [ d is a genius]. pre  The examples above illustrate the canonical (subject-predicate) order (Ruwet 1982) for English. In (a), the pronoun has no intrinsic reference but its referent must be identified in the discourse and is therefore specific. However, the indefinites  a moron  and a  genius  describe a property of a person — and they cannot be referential. The indefinite cannot be in subject position, for example:  107  (7.a) b)  *[ j A moron] sub  *[subj  [pred  A genius] [  pred  is he/him], is Mary],  One of the diagnostics for predicative structures involves agreement features. In Italian, the copula can agree only with the more referential DP (i.e., the DP which is more salient with respect to subjecthood) even when it is not in the canonical subject-initial position (cf. Moro 1990:15). In (8.a), the copula sono 'are' agrees in number with the following DP loro 'them' rather than with the DP la causa 'the cause' which precedes the copula. (8.a) L a causa sono loro the cause (sg) are them (pi). b)  *La causa e loro the cause is them  Pronouns must have a referent in the discourse and are the equivalent of a name — which according to Heggie's hierarchy in (5) are more referential than definite descriptors. By contrast, the copula in English always agrees with the initial DP (which is analyzed as the subject on the basis of linear ordering). (9.a)  b)  (i)  The problem is them.  (ii)  *The problem are them,  (i)  They are the problem.  (ii)  T h e y is the problem.  2  For further discussion of the asymmetries between the DPs, I refer the reader to Ruwet (1982), Moro (1990) and Rapoport (1987) which look at a wider range of evidence. 2  108  Clefting is used as a diagnostic to show which nominal in a predicative structure is more referential (Heggie 1988:80). Consider the possibilities for the predicational sentence in (10) which has the canonical subject-predicate ordering in asymmetrical nominal structures, only the more referential DP in the canonical ordering may be clefted; and names are more referential than possessor DPs. Therefore, in (b), the more referential subject may be clefted while in (c), the predicative nominal cannot be clefted. Note that [...] marks the original position of the clefted constituent. (10.a)  [Subj John Smith]  b) It's  [  pre(J  is my doctor].  [Subj John Smith] that... is my doctor.  c) *It's [pred my doctor] that John Smith is .... Now we will consider the inverse ordering (Ruwet 1982) with the two DPs reversed as in (1 l.a). In the inverse ordering (cf. Heggie 1988), the less referential initial DP is reanalyzed as the subject as shown in (a). As we see, neither the less referential (subject) DP in (b) nor the more referential nominal (as predicate) in (c) can be clefted. (1 l.a)  [Subj M y doctor] [pred is John Smith ].  b) *It's [subj my doctor ] that... is John Smith. c) *It's [pred John Smith ] that my doctor is .... According to the hierarchy in (5), deictic DPs are the most referential; and accordingly, only the more referential deictic DP can be clefted. A nominal sentence with canonical subject-predicate ordering is shown in (12.a) (Ruwet 1982). Only the the more referential subject nominal can be clefted as seen in (b) and (c). (12.a)  Ubj  That man over there]  [pred  is Jack Jones].  109  b)  It's  [Subj that man over there] that... is Jack Jones.  c) *It's [pred Jack Jones] that that man over there is. The following set involves the inverse ordering with the less referential DP in initial position. As in (11) neither DP can be clefted. (13.a)  [SUbj Jack Jones] [pred is that man over there].  b) *It's [subj Jack Jones] that... is that man over there. c) *It's  [pe r d that man over there] that Jack Jones  is ....  The patterns we have seen above reveal the asymmetry between the two nominals in predicative structures. In the canonical ordering, the more referential subject DP can be clefted while the predicate DP cannot be. In the inverse ordering where the more referential DP is in the predicate position, neither DP can be clefted. Diagnostics which are sensitive to the properties of the predicative constituent include: i) the albeit test, and ii) though preposing. The examples in (14) are in the 3  canonical ordering with the referential nominal as subject. In (a), the indefinite predicate a fool  can be qualified by the albeit phrase; while (b) shows that a name in predicate position  cannot be. (14.a) b)  [SUbj Mixal] [pred  is a fool ], albeit cunning.  *[subj The chair of the  department]  [pred is Jane Smith ], albeit on leave.  Similarly with the though preposing test. The though preposing diagnostic requires dislocation of the predicate of a copular clause and contrasts two property-denoting characteristics which may be applied to the subject DP. These examples illustrate that a  3  This test is credited to Ken Hale (Rapoport 1987:133 ft) 110  property-denoting nominal predicate (whether indefinite or definite) may be preposed with though, as shown in (a) and (b). In (c), the preposed predicate is adjectival. (15.a)  [pred A fool]  though Mixal is  b)  [Pred The chief] though Mary is  c)  [Pred Proud ]  though Tali is  she is cunning. she is well liked. she is kind.  The sentence in (a), for example, the subject is Mary and preposing of the predicate a fool contrasts two properties of Mary, i.e., she is a fool, and she is cunning. On the other hand, in the inverse sentences in (16), the fully referential DPs (which are not property denoting) cannot be preposed over the less referential DP. (16.a)  *  [pred Jane Smith] though the chair is  she is stupid.  b)  * [pred That woman over there ] though Tali is  c)  * [pred Mary ] though the  chief is  I didn't recognize her.  she is well liked.  Similarly with the albeit test in (17): (17.a)  [Subj Mary ] [pred  b) * [subj The chief  is the chief], albeit a woman.  [pred is Mary],  albeit a Woman.  Again, the albeit phrase contrasts properties of the two nominals. Though both Mary and the definite DP, the chief, may be referential (cf. Williams 1994:42), they need not be. Of the two DPs, only the chief may denote a property (insofar as it represents a position which any person may attain). The two DPs are asymmetric and this asymmetry is captured in Rapoport's tests. In summary, we have seen evidence for Heggie's (1988) hierarchy of reference in (5), which is repeated here:  111  (5)  Hierarchy of Reference: ( Heggie (1988:106) deictic > names > definite descriptors > indefinites [subject] more referential  —> [predicate] less referential  The least referential DPs are indefinites. We have seen that they may occur only as property-denoting predicates, and inverse sentences with the indefinite DP in initial position are not generally felicitous (unless both nominals are indefinite). The other three categories — definite DPs, names and deictics — are more flexible and generally can occur in either subject or predicate position in a copular sentence in English. However, in the canonical ordering, they occur in accordance with the order shown in the hierarchy in (5). In other words, a deictic DP is the canonical subject in relation to a name or a definite DP; and a name is the canonical subject in relation to the definite DP, but the canonical predicate in relation to the deictic DP. The hierarchy in (5) will be used as a guide in the investigation of nominal clause structures in Nehiyawewin. Equative Nominal Clauses In equative sentences, both nominal consituents are equally referential (Heggie 1988). In 4  other words, there is no asymmetry between the subject and the predicate; for example, in (18.a) both terms refer to the same star. In the context of the movie, Chinatown, (b) have the same specific referent. These sentences are equational:  4  See also the work of Moro (1990) following Higgins (1973). 112  both DPs in  (18.a) The morning star is the evening star. b) M y daughter is my sister. The first DP is in subject position while the second is in predicate position in the copular structure. Both of the DPs must have a specific referent and that referent must be one and the same person/thing in both cases. In equative sentences, the two DPs are 5  interchangeable. (19.a) i)  The evening star is the morning star,  ii)  The morning star is the evening star,  i)  M y sister is my daughter,  ii)  M y daughter is my sister.  b)  We will be looking at different kinds of tests to determine the properties of the constituent DPs. For example, Heggie (1988:80) tests for the more referential DP using a cleft construction (revealing that the less referential predicate nominal cannot undergo clefting). If the two DPs are symmetrical (equally referential), then we predict that there will be no asymmetries in the clefting patterns regardless of which constituent is in subject position. Take one of the equative sentences in (19. a), for example. As the illustrations in (20) show us, the subject nominal in (b) may be clefted but the predicate in (c) cannot be: (20.a)  is  [pred the  evening star],  It's  [subj the morning star] that is the evening star.  c) * It's  [pred the evening star] that the morning star is.  b)  5  [Subj The morning star]  Alternatively, both DPs can be equally non-specific, i.e., A man is a human being. 113  Now if we reverse the same two nominals, the clefting test will show exactly the same results, i.e., the subject nominal may be clefted while the nominal in predicate position cannot be. This shows that there is no asymmetry between the two nominals. (21.a) [subj The evening star] is [pred the morning star]. b) It's [Subj the evening star] that is the morning star. c) * It's [pred the morning star] that the evening star is. In equational sentences the predicate DP cannot be clefted; only the subject DP (i.e., the argument DP) can be clefted. Rapoport (1987) tests for the property-denoting characteristics of nominals. Given the examples above where both DPs are equally referential, the predicate DP is not property denoting. This should be reflected in tests for property denoting characteristics of predicate DPs. Using Rapoport's  though  preposing test on the predicate DPs in (19.b), we see that this  is indeed the case. Both of the sentences in (22) are both bad, i.e., neither DP as a preposed predicate can be property denoting. (22.a) *My sister though my daughter is, I love her. b) *My daughter though my sister is, I love her. In examples in (20) to (22), there is no asymmetry between the two nominal constituents in each set, i.e., both have the same referential properties.  4.1.2 Nehiyawewin Nominal Clauses Now we will consider Nehiyawewin nominal clauses. But first, we look at the types of nominal constituents in Nehiyawewin and classify them according their characteristic  114  features. This occurs in Section In Section, we look at the behaviour of deictic DPs in simple nominal clauses. Deictic DPs universally have a special status insofar as they are the most referential DPs. Therefore, deictic DPs in Nehiyawewin only occur as the subject in nominal structures while non-deictic DPs may occupy either subject or predicate position. We will then consider more complex examples involving clefted NPs and a relative clause. These may involve clefted NPs in Focussed NP constructions; and clefted wh-words in wh-questions. These discussions are in Section Then we will look at the agreement between subject and predicate in nominal structures and provide an analysis for the evidence. Classifying Nominal Constituents Nehiyawewin does not have obligatory determiners. There is no distinction between definite and indefinite DPs. The only determiners available are a set of deictic determiners shown in (23). Deictic determiners point to a specific individual and include information for three degrees of distance, proximate vs. obviative, gender (animate (3) vs. inanimate (O)), and number. (23) Demonstratives (Wolfart 1973:33):  3 3p 3' 0 Op  This  that  that yonder  awa oki ohi  ana aniki anihi anima anihi  naha neki nehi nema nehi  /v  6  oma ohi  oma 'this' is also the default or unmarked determiner used often in conjunction with animate as well as inanimate nominals (cf. Ahenakew 1987.b).  6  115  Before we can begin our discussion of nominal clauses, we must consider the types of nominal constituents available in Nehiyawewin and rank them. The table in (24) contains the nominals (overt and non-overt) which are commonly found in the language and groups them according to their patterns of interaction. (24) Nominal Constituents a) Indef:  [+Quant] [-wh]  awiyak 'somebody'  indefinite pronoun  [+Quant] [+wh]  awina  wh-word  b) Descriptors and Names:  'who'  iskwew 'woman'  bare noun  John  proper name  niya  T, me'  personal pron.)  ni-simis 'my sibling'  possessive phrase  eko  pronoun  'the very one'  (previous mentioned) c) Deictics and pro. [-deictic] [+deictic]  7  [DP 0 pro  (discourse) null DP  ]  ana napew  'that man'  DP  ana pro  'that one'  DP  The Indefinite [+Quant] category in (a) is the least referential and includes wh-words and indefinites like 'somebody, something'.  8  We will see that [+wh] indefinites are always  (nominal) predicates in the context of nominal clauses. Descriptors and Names in (b) are more referential than indefinites and less referential than deictic DPs and pro. In the absence of non-deictic determiners, the bare nouns are  Bare nouns can have either a definite or indefinite interpretation depending on the context (the parameters have yet to be determined).  7  Truly Quantified NPs like kahkiyaw ndpew 'every man' are not discussed here though they are certainly part of the inventory.  8  116  unmarked with respect to NP vs. DP status (and definiteness) and must be classified by other means. This is an area which requires future study. (i)  Overt pronominals refering to the speech participants,, niya T and kiya 'you' are usually considered to be definite. These pronominal forms serve a 9  predominantly emphatic function. They never occur as arguments of a verb ~ though they may occur (as subjects or predicates) in nominal clause structures. In either capacity they have DP status because they require a discourse referent. (ii)  The other members of this set include names, possessive DPs, and the eko set of pronominals . These are referential and/or definite and have DP status. 10  (iii)  According to the analysis in Longobardi (1994), common nouns such as iskwew  'woman' denote a kind of entity, animate or inanimate. In a discourse  context Nehiyawewin NPs may get definite or indefinite interpretation. In a verbal clause they are licensed by pro in argument position. However, we will see that a lexical noun may occur in subject position of a nominal clause. In the context of an elicited nominal clause, a deictic determiner is typically  Third-person wiya 's/he' does not have the same occurrence patterns as the 1/2 pronouns and its use appears to be restricted to the function of an intensifier (Blain 1994). The eko set fills this gap in the paradigm. 9  This pronoun belongs to the following in the discourse: 3 prox sg., 0 sg. 3 prox pi., 3 obv, 0 pi. 10  set which specify a (previously mentioned) person/thing ~ ewako or eko — ekonik -ekoni 117  provided by Nehiyawewin speakers for the subject nominal in a nominal clause environment.  11  The most referential category in the table in (24) is (c), which involves deictics and pro.  A deictic DP may have an overt NP or an empty category, pro as illustrated in (25).  12  Deictics and pro are distinguished by the fact that they cannot function as the predicate in a nominal clause. (25)  DP D  NP  I  I pro  ana  napew 'man' When teamed with an overt determiner, pro has the interpretation '(this/that) one' rather than the normal 's/he', 'him', 'them', etc. Empty pronominal pro as an argument in A-position of a verb has [+definite] status. As an argument in subject position of a nominal clause, pro (without an overt deictic determiner) is analysed as having a null determiner as shown in (26). If there is no deictic, then as the subject of a nominal clause these NPs would require a null determiner. We will see below that this rarely (if ever) happens and a deictic determiner is typically inserted by the speaker in these situations even if the sentence has not be elicited with a deictic. More research is required to determine whether a referential (DP) interpretation is really possible with these NPs. 11  In other words, I do not see these demonstratives as proforms in and of themselves. They require a nominal constituent, empty or overt, to modify. Baker (1996) claims that, in discontinuous DPs, the demonstrative and the nominal are both coindexed with the pro in argument position. In my analysis, the lexical NP (with or without a deictic determiner) is coindexed with pro in A-position. The demonstrative (+ pro) is an independent DP which must be licensed in some other manner. A perusal of narratives suggests that nominal clauses are a common occurrence (cf. Dechaine (to appear)). In other words, these deictic DPs may often occur in nominal clauses with other nominal constituents. 12  118  (26)  DP  / \  D I 0  NP I  pro  The examples in (27) illustrate the DP structures in (25): the (bolded) constituent ana ndpew  'that man' in (a) is a full DP, and  ana 'that (animimate/proximate)  one' in (b) is a DP  without an overt NP. (27.a) ana napew e-wapam-a-t Mary-wa that man conj-see-dir-3 Mary-obv That man saw Mary. b) iyikohk e-itahtopipone-t ana, kiseyiniw ana, ... as much conj-be so old -3 that (one), old man that (one),.. "As old as that old man was, ..." (Ahenakew 1987a: 104) 13  Literally: As old as that one was, that's the old man,... The more complex (b) example is from a text and requires some explication. The deictic ana 'that (one)' in the first clause (preceding the comma) constitutes the DP ana (pro) 'that one'. After the comma, there is a nominal clause kiseyiniw  ana,  for which I provide the literal  translation, 'that is the old man'. (We will see below that nominal clauses in Nehiyawewin are predicate initial.) Subject and Predicate Ordering According to Heggie's (1988) hierarchy, deictic DPs in English are strongly referential; and as a result, these DPs may only occur as subject in the canonical ordering of copular sentences. We will look at the corresponding Nehiyawewin structures with respect to the three  13  literally:  itahto-pipone-  'be so many winters' 119  categories in (24): (a) Indefinites, (b) Descriptors and Names, and (c) Deictics and pro. We must first establish the canonical ordering for subject and predicate; and then we will contrast members of all the various subcategories. The following hierarchy is adapted for the predicate-subject ordering in Nehiyawewin and shows the three categories we will be discussing. (28) Nehiyawewin Hierarchy of reference: indefinites <— Descriptors & Names <— Deictics and pro [predicate] <less referential  [subject] more referential  The first pair of examples involves a deictic DP (which is more referential than any other nominal constituent) and illustrates the predicate-initial ordering of Nehiyawewin nominal clauses. (29) involves the ordering of the deictic ana 'that' and a possessive phrase which is a member of the Descriptors and Names category We see that these two constituents may constitute a single DP, as in (a), or a nominal clause structure with two DPs, as in (b), depending on their ordering. With the ordering in (a), the deictic is the determiner of a DP introducing nisimis. The nominal clause reading is not available, as indicated, and this example requires a verb in order to complete the. sentence. When the order is reversed, as in (b), it constitutes a nominal clause. Observe in example (29.b) that there is no copula in Nehiyawewin nominal clauses. (29.a) ana ni-sTmis  Vb  that 1-younger sibling That younger sibling of mine ^ M y younger sibling is that one.  D. 182.b  120  b) ni-sTmis ana 1-younger sibling that That (one) is my younger sibling.  D. 182.a  The descriptive nominal nisimis 'my younger sibling' precedes the more referential deictic DP subject ana (pro) 'that one'. Given that the order in a nominal clause is not reversible (as illustrated in (a)), I propose that this example constitutes a predicative (asymmetric) nominal clause structure with the first nominal constituent as the predicate. Furthermore, this example establishes the predicate-initial ordering of nominal clauses in Nehiyawewin. The nominal clause examples with deictic DP subject show that the predicate-subject ordering as shown in (30.a) is licit while subject-predicate ordering as in (b) is not. (30.a) Nom.Clause DP I  b)  * Nom.Clause  DP  DP DP  I  I  PRED SUBJ  I  SUBJ PRED  In other words, the predicate is clause initial in a nominal clause. Three Paradigms We have established that nominal clauses are predicate-initial; now we will provide examples for three paradigms of nominal contrasts, namely: I.  [+wh] Indefinite  -  Descriptors & Names  II.  [+wh] Indefinite  -  Deictics and pro  III.  Descriptors & Names -  Deictics and pro  We will also look briefly (in Section at a fourth set, i.e., Descriptor - Descriptor. 121  We will see that, when both NPs are Descriptors without a discourse context — or when the sentence has tense/aspect, one of the DPs typically is verbalized. Paradigm I: [+wh] Indefinite - Descriptors & Names In this section, we discuss the properties of the various descriptors which occur as subject of the nominal clause. Equating them with an indefinite provides insights into their referential properties. The examples involve a name in (31), a personal pronoun in (32), eko 'the very one' in (33), a possessor phrase in (34), and a bare noun in (35). In a nominal clause which is elicited with a non-deictic NP (subject) and equated with an indefinite like awina 'who', a deictic determiner is often introduced (i.e., ana Mary 'that Mary') in the subject of the Nehiyawewin sentence. This suggests that non-deictic Descriptors and Names (especially those without a discourse context) do not have referential DP status in their bare form. The following example involves a name and was presented with the interjected ana 'that' as shown. The reverse ordering in (b) is ungrammatical. (3l.a) [p d awina] [ubj ana Mary ] who that Mary Who is 'Mary'? cf.D.292 re  S  b)  *anaMary awina that Mary who That'Mary'is who?  cf.D.292  In (32.a -b) a default all-purpose oma 'this (inan)' occurs optionally with the 2-person subject. As shown in (c), the reverse word order is not grammatical; the wh-predicate must be clause initial. (32.a) awina (oma) kiya who (this) you Who are you?  D.22  122  b)  [pred awina] [SUbj  kiya] who you Who are you? B.433  c)  *kiya awina you who Who are you?  B.433  The 2nd-person speech participant is referential enough to occur (albeit optionally) in the bare form. In (33.a), eko 'the very one' the third-person counterpart of 1st- and 2nd-person pronouns, also has referential status.  14  By definition, this pronominal refers to some person  in the discourse; hence its ability to refer. The reverse ordering in (b) is not licit. (33.a)  [pred awina] [subj  eko]  who the very one Who is he (i.e., the one we were talking about)? b) *eko the very one  B.399.b  awina who  Who is he (i.e., the one we were talking about)?  B.399.b  Deictic DPs never occur with discourse dependent eko 'the very one'. Possessor phrases have referential status via the possessor, who must refer either directly or via a referent in the discourse. However, a deictic is optionally inserted with a possessor phrase — in this case, the default 6ma 'this one' . The order shown in (b) is not good. (34.a) awina (oma) ki-simis who this 2-younger sibling Who is your younger sibling?  cf.MA.467  b) *(6ma) ki-simis awina this 2-younger sibling who Who is your younger sibling ?  14  Third-person wiya serves other functions in the syntax, as suggested above. 123  Regular non-refering lexical NPs are always translated with a deictic determiner such as ana 'that' as in (35.a). The reverse ordering in (35.b) is not grammatical.  15  (35.a) [pred awina] [SUbj ana iskwew] who that woman Who is the woman? MA.457  b)  *ana iskwew awina that woman who Who is the woman?  Alternatively (i.e., in a context which precludes a deictic DP), a non-refering non-deictic NP may be realized as a verb with verbal morphology. (36)  awina ota e-okimahkan-iwi-t who here conj-chief-vb-3 Who is the chief here? MA.350.b The above evidence shows that, without a discourse context, there is the option if not  a preference for a deictic determiner to cooccur with nominals in the Descriptors and Names category when they occur in argument (subject) position of a nominal clause. In the case of common nouns, it appears that the deictic is obligatory in this environment.  16 Paradigm II: [+wh] Indefinite - Deictics and pro The examples in (37) and (38) involve an elicited deictic DP (vis a vis examples in the previous section where the deictic was not elicited as such). The example in (37.a) involves  The example in (i) involves the inanimate wh-word kikway 'what', illustrating that the same type of structures apply to the inanimate wh-word. The (obligatory) inanimate deictic oma 'this' is inserted as subject followed by a pause (comma), and the questioned word is in apposition. In Section 4.1.3 and in Chapter 5,1 discuss the issue of agreement between subject and predicate and my analysis will show the appositional NP adjoined to IP. (i) [pred kikway] [sui,j oma], "iskwew" what this, "woman" What is an "iskwew"? 15  16  This argues for the non-argument status of (at least this group of) overt lexical NPs. 124  a regular lexical DP. Example (37.b) shows that the reverse ordering is not grammatical. The examples in (38) do not have an overt NP. (37.a)  (38.a)  [pred awina] [subj  ana napew] who that man Who is that man? MA.452  [pred awina] [subj ana ]  b)  ^ana napew awina that man who Who is that man? MA.452  b)  * a n a  who that (one) Who is that one? B.399  a w i n a  that who Who is that one?  cf.B.399  The wh-phrase in (39) is a bare wh-predicate. Every predicate must have a subject (as the the gloss illustrates); accordingly, the subject must be pro. (39)  [pred awina ] [subj pro] who Who is it/he?  B.527 Paradigm III: Descriptors & Names - Deictic DP In this paradigm, one of the nominals is from the set of Descriptors & Names while the other, the subject DP, contains a deictic determiner. In the first sub-set, the deictic DP contains a non-overt nominal. In the next sub-set, the deictic DP contains an overt nominal. Deictic DP (Non-overt Nominal) The first example has a proper name as predicate and (as it happens) comes in the form of a yes/no question with the Q-particle ci inserted in second position between the two nominals. Again, the reverse ordering with a clause-initial deictic DP is not grammatical: (40.a)  [pred Bill]  ci [subj ana pro] Bill Q that (one) Is that Bill? LET.6  b) * ana ci Bill that Q Bill Is that Bill?  125  The next example involves a personal pronoun as predicate. The sentence was not elicited in this manner (i.e., with the appositional DP), but the Nehiyawewin translation was presented as in (a). In example (a), the subject is ana 'that (one)' in the segment preceding the comma, which is literally: 'mine/my one that one . This is followed by the DP 'that dog" in 1  apposition. Observe that the deictic DP is always the second nominal (subject) in the nominal clause, i.e., the ordering in (b) is not licit: (41.a)  [pred niya] [subj  ana atim mine that (one), that dog That's my dog. J. 16  b)  anapro],  * ana niya , ana atim that mine , that dog That's my dog. cf.J. 16  The next set contains the pronominal eko/ewako 'the very one' as the predicate. The deictic cannot be in predicate position as shown in (b). (42.a) ewako oma the very one this (one) This is the one. This is it!  b) D.246.a  * oma ewako this (one) the very one This is the one. This is it!  The predicate is a possessive DP in the following set: (43.a)  [pred ni-atim] [subj  ana pro ] l.Poss-dog that (one) That's my dog. MA.477  b)  * ana ni-atim that l.Poss-dog That's my dog.  It is more difficult to get a bare noun as predicate in an elicited example. Ahenakew (1987b: 153) provides the following example with a bare noun as predicate; the reverse ordering in (b) provides a deictic DP rather than a nominal clause. (44.a)  [pred mohkoman] [sui,j oma] knife This is a knife. ^This knife....  b)  this  126  oma mohkoman this knife This knife ^This is a knife.  According to Mary Ann (R.M. Dechaine, p. a), this type of example requires a show-and-tell situation. Out-of-the-blue elicitations do not appear to provide the referential context required; thus, they are more difficult to obtain. Another example from a text supports this claim. In (c) (Ahenakew 1987a: 104), the bolded nominal clause has a bare noun kiseyiniw  'old man' in  predicate position. Given the context of the story, the predicate DP has a referential reading. c)  iyikohk e-itahtopipone-t ana, kiseyiniw ana, ... as much conj-be so old-3 that (one), old man that (one),.. "As old as that old man was, ..." (Ahenakew 1987a: 104) Literally: As old as that one was, that's the old man,... Deictic DP (Overt Nominal) This category, with an overt lexical DP as subject, does not contain a full range of examples due to gaps in the data available. The example in (45) involves a name as the predicate DP while in (46) the predicate is a personal pronoun. In both cases, the deictic DP is restricted to subject position. (45)  [pred John] [SUbj awa okimahkan] John this chief John is chief. M A . 54 Literally: This chief is John.  (46.a)  [Pred niya] [subj  awa okimaw] I this chief/boss I'm the chief. cf.MA.352  b)  * awa okimaw niya this chief/boss I I'm the chief.  The English glosses are the sentences which were elicited. In (45), John is the subject of the English sentence. In the Nehiyawewin translation, the surface ordering is maintained; however, John is the predicate DP.  127  The sentence in (47) has three DPs -.- the pronominal eko 'the very one', the DP ana napew  'that man', and Bill — in the Nehiyawewin translation of an English sentence involving  two DPs.  I propose (see analysis in Section that the third NP is adjoined to IP in the  nominal clause structure, in apposition to the subject DP. Again, the reverse ordering with the deictic DP as predicate is impossible, as shown in (b). (47.a)  b)  [pred eko ]  [ j ana napew], Bill the very one that man Bill That man is Bill. LET.4 Literally: That man is him, Bill. sub  *[pred  ana napew] [ eko] , Bill that man the very one Bill Bill is that man. LET.4 subj  In all the Paradigm III examples above, there is an elicited deictic DP in subject position together with another nominal (descriptor or name) as predicate. In the following section, we will look at what happens when both of the nominal constituents are elicited as a descriptor or name. Descriptor - Descriptor In Section above, we saw evidence in examples in (44) that bare nouns require a discourse context in order to occur as a nominal predicate. In elicited sentences, this discourse context is missing and this is reflected in the examples below. This is seen again in  (48) where the bare NP predicate in (a) is problematic in this elicited sentence — Mary Ann preferred the version in (b) with a verbalized form for elicited DP, the chief.  128  okimahkan] [ i,j awa ni-mis ] chief this l.Poss-sister The chief is my sister. MA.348 Literally: This sister of mine is chief  (48.a) ??[ d pre  b)  su  [ e-okima-wi-t ] awa ni-mis conj-chief-vb-3 this l.Poss-sister The chief is my sister. MA.348 Literally: She is chief, this sister of mine.  Not only is the deictic DP always in subject position, but the subject in these examples ~ including names and possessive DPs — almost invariably occurs with a deictic determiner in elicited sentences. The question arises: Is this due to the fact that, in nominal clauses, the overt lexical N P can occur in subject position (i.e., in argument position) as opposed to verbal clauses, which have pro in A-positions?  17  Other examples include (49) in which both nominals were elicited as lexical DPs with the indefinite DP being the predicate, i.e., the chief is a woman. The elicited English predicate in (a) is verbalized while the other DP becomes the predicate of a separate nominal clause, as indicated in the literal translation. Reversing the deictic and the bare nominal,as in (b), creates a deictic DP structure with the translation as shown. (49.a)  okimahkan ana e-iskwew-i-t chief that conj-woman-vb-3 The chief is a woman. MA.345.b Literally: That one is the chief & she is a woman.  b) ana okimahkan e-iskwew-i-t that chief conj-woman-vb-3 That chief is a woman ^That one is the chief &  The absence of determiners in Nehiyawewin and the semantics of NPs is an area which requires study (cf. Matthewson 1996 on Salish, for example). 129  The nominal clause gloss (i.e., 'that one is the chief) for the first two constituents in (b) is not grammatical. The evidence suggests that i f both nominal expressions are property denoting, then one must be verbalized. Another situation which requires a verb rather than a lexical D P involves a sentence with tense or aspect. There is no overt aspect in (50.a), John is chief, and the sentence has a nominal clause structure with a deictic D P as subject. With overt aspect as in (50.b), John used to be chief, the second N P chief'is derived as a verb. John is the predicate in a sentence-intial nominal clause structure with awa (pro) 'this (one)' as subject D P — resulting in a biclausal sentence which is illustrated in (c) with a literal translation as shown. (50.a)  18  John] [ bj awa okimahkan] John this chief John is chief. M A . 54  [p j re<  su  b)  John esa awa John I understand this John used to be chief.  e-kl-okimahkan-i-t conj-perf-chief-vb-3 MA.55  c)  [CP John esa awa pro] & [ p e-kT-okimahkan-i-t] Literally: John , I understand, is this one & he used to be chief. C Summary W e have seen that nominal clauses are predicate-initial in Nehiyawewin. W e have also seen that predicate-subject asymmetries in Nehiyawewin reveal a referential hierarchy similar to that for English (Heggie 1988). In this hierarchy, deictic D P s are the most referential and indefinites, including wh-words, are the least referential. Within that context, wh-words are  Particles like esa 'I understand' are often inserted in Nehiyawewin sentences including nominal clauses. They are assumed to be adjoined to IP and do not affect the analysis. 130  always in predicate position and deictic DPs and pro are always in subject position of a nominal clause structure. These asymmetries are captured in the Table in (51). I have also indicated with respect to the set in (I) the relative tendencies to insert a deictic determiner with the Descriptor/Name in subject position. (51) NP!=PRED I: wh-Indef - Desc/Name  * Desc/Name wh-word  NP = SUBJ 2  wh-word Desc/Name deictic name deictic noun (deictic) pronoun (deictic) poss. phrase — eko  II: wh-Indef - Deictic/pro  * Desc/Name wh-word  wh-word Desc/Name  III: Desc/Name - Deictic DP  * Deictic DV/pro Desc/Name  Desc/Name Deictic DP/pro  pro  Desc - Desc  (- context)  Desc + verbalizer  deictic Desc.  (+context)  Desc.  deictic Desc.  Within the Descriptors & Names category of (see list in (24)), all the constituents, except for bare nouns, may occupy either subject or predicate position. This implies that these constituents should be interchangeable when they come together in a nominal clause situation. In the case of bare nouns, however, they are also restricted in their use as predicates unless they have a discourse context. On the other hand, they ~ like any other 131  descriptor and/or name — always require a deictic determiner when in subject position. It has also been observed that the strong evidence for deictic determiners with subject D P s may be due to the fact that, in nominal clauses, overt DPs are allowed in argument position.  4.1.3. Agreement and Structure in Nominal Clauses In Section 4.1.2, we looked at sentences involving two lexical NPs — especially those involving regular nouns; and we observed the tendency to derive a property-denoting NP as a verbal predicate rather than a nominal one. In particular, clauses marked for tense or aspect require a verbal predicate to register this temporal agreement together with person agreement for the arguments involved.  Nehiyawewin nominal clauses, as noted above, do not have a  19  copula to register this agreement; therefore, verbal morphology is required on the verb. W e have seen, in (1) above, that Nehiyawewin nominal clauses in focussed N P structures frequently involve eko 'the very one (previously mentioned)' and another nominal. (1) is repeated here as (52).  The ordering o f nominals is reversible as shown in (52.a) and  (52.b). (52.a)  b)  John] [ j eko ] John the very one It was John that sang. [p d r e  sub  eko ] [ ubj John ] the very one John It was John that sang. [p d i e  S  [ p C  [ p C  ka-nikamo-t ] rel-sing-3 B.222 ka-nikamo-t ] rel-sing-3  In the process, the ordering possibilities are more free, i.e., rather than obligatory clause-initial predicate, the ordering of verb and adjoined N P s is more flexible. 19  132  In both structures, the initial DP is the predicate while the second DP is in subject position. There is only one third-person involved in this sentence, i.e., John, and the NP is proximate. In the nominal clause, both the subject and the predicate are proximate.  20  The following examples show a possible variation within the nominal clause of focus constructions. In (53), there are two third-persons: John is proximate, and the focussed NP Mary-wa  is obviative. The predicate ekoni 'the very one (obv)' in (53.a) agrees with the  obviative status of the subject of the nominal clause, Mary-wa. Mary-wa (53.a)  b)  Mary-wa  Note also that ekoni and  are interchangeable as shown in (b). [ d ekoni ] [ j Mary-wa] [ p ka-wapam-a-t John] the very one(obv) Mary-obv rel-see-dir-3 John It was Mary that John saw. cf.B.343 | ) r e  sub  [pred Mary-wa ] [ bj ekoni ] Mary-obv the very one(obv) It was Mary that John saw. SU  C  [ p C  ka-wapam-a-t John] rel-see-dir-3 John cf.B.343  is the antecedent for the obviative object pro in argument position of the verbal  clause. In this section (i.e., 4.1.3), we will look at the kind of agreement with occurs between the subject and predicate nominals both in regular nominal clauses and in nominal wh-clauses. I will then propose a structure for nominal clauses.  The verbal clause with kd- complementizer (and null-operator movement) is a relative clause. I will argue below that a relative clause can be hosted in the nominal clause in a position adjoined to IP.  20  133 Nominal Clause Agreement We will look at four types of agreement in nominal clauses: agreement for number, gender, proximate/obviative, and person agreement.  As noted above, we will include regular  nominal clauses and nominal wh-clauses. We will compare briefly, in Section the agreement possible with a verbal constituent. Agreement for Number In (54) and (55), we see that the subject and predicate in nominal clauses are marked for number. The (a) examples show singular (unmarked) agreement while the subject and predicate in the (b) examples are plural forms. The wh-examples are given in (55). (54.a) eko [ bj ana pro ] the very one that That's him/the one we were speaking about. SU  b)  (55.a)  b)  ekonik [ b j aniki pro ] the very ones those That's them.  J.887  awina] [ j ana (pro) ] who (sg) that (one) Who are that one?  B.399  su  [p ed r  suu  awiniki] [ aniki (pro) ] who (pi) those (ones) Who are those guys? B. 149 [  p r e  d  subj Agreement for Qender There is no masculine/feminine gender distinction in Nehiyawewin; however, there is animate/inanimate "gender" distinction. The following examples illustrate that the subject  134  and predicate in nominal clauses agree for animacy. The NP in (56.a) is inanimate (with an inanimate deictic form) while (b) involves an animate NP with a corresponding animate deictic form: (56.a) mohkoman [ b j oma knife this(inan) This is a knife. su  pro]  Ahenakew (1987b: 153)  b) Bill ci [ ubj awa pro] Bill Q this (anim) Is this Bill? cf.LET.6 S  The same contrast exists in wh-clauses; example (57.a) with kikwdy 'what' involves the inanimate wh-word while (b) with awina 'who' represents animate gender. (57.a)  b)  kikway] [ b j oma ] what this (inan) What is this? cf.MA.459 [p d r e  su  awina] [ bj ana napew ] who that man Who is that man? MA.452 [pred  SU  The number and gender agreement shown represent the typical features involved in subjectpredicate agreement. The proximate/obviative agreement is a language specific feature. Agreement for Proximate/Obviative The examples in (58) involve possessor phrases with first- and third-person possessor in (a) and (b) respectively. The (first-person) possessed dog in (a) is proximate (unmarked); in (b), with a third-person possessor, the dog is marked for obviation. In each case, the deictic in subject position reflects this contrast. (58.a)  ni-atim] [ j ana pro ] l.Poss-dog that (one) That's my dog. MA.477 [p d r e  sub  135  b)  o-tem-a ] 3.Poss-dog-obv That's his dog.  anihi pro ] that (one) cf.MA.477  [pred  [ bj SU  The same contrast occurs in the wh-clauses in (59). With the second-person possessor in (a), the nominal -simis 'younger sibling' is proximate. With a third-person possessor, the possessed nominal would be obviative as shown in (b). (59.a)  awina] [ ki-simis ] who your younger sibling Who is your sister? MA.455 [pred  subj  awina] [ j o-simis-a ] who his-younger sibling-obv Who is his younger sister? [p d i e  sub  The examples in (60) show obviative agreement between subject and predicate. The obviative-marked nominal clause form in (a) does not typically occur by itself as shown, i.e., it would require a special context. In the full wh-question in (b), this context is provided. This example involves an obviative referent in the following clause. The [+AGR] wh-phrase is marked obviative accordingly. (60.a)  b)  awini-wa ] [ j anihi ] who-obv that (obv) one Who is that? M . A . 377.b [pred  sub  awini-wa] [ j anihi [ John ka-wi-wikim-a-t ]] who-obv that(obv) one John rel-fut-marry-dir-3 Who is it that John is going to marry? D.261 [pred  sub No Agreement for Person We have seen above that there is subject-predicate agreement in Nehiyawewin for number, animate/inanimate gender, and for proximate/obviative. On the other hand, there is never  136  person agreement between subject and predicate nominals in a nominal clause. This type of agreement is restricted to a verbal constituent. The absence of agreement is evidenced in the Nehiyawewin examples in (61), in contrast to the English glosses in which the copula has different forms for the first- and third-person subjects. (61.a)  b)  John] [ j awa okimahkan] John this chief John is chief. M A . 54 Literally: This chief is John.  [pred  sub  niya] [ j awa okimaw] I this chief/boss I'm the chief. cf.MA.352  [pred  sub  Similarly, the wh-phrase is not marked for person agreement as illustrated in (62). The wh-word awina 'who' pronominal has the same form with the third-person in (a) and the second-person in (b). (62.a)  [p d re  awina] [  subj  eko ]  who the very one Who is he? (the very one we mentioned) b)  MA.458  awina] [ j kiya ] who you Who are you? B.433 [pred  sub  The wh-word is always the standard pronominal form while the subject in (b) is secondperson. The subject-predicate agreement which occurs between nominals in a nominal clause is termed Predicational agreement. In the following subsection, predicational agreement is contrasted to verbal agreement.  137 Verbal Agreement Verbal agreement is richer than predicational agreement and may include all of the agreement categories listed. For example, consider the paradigms in Appendix A for Nehiyawewin. In (63), (a) illustrates the person agreement forfirstand third persons with the conjunct form of an intransitive (VIA) verb, nikamo- 'to sing'. The examples in (b) show number agreement with the third person singular and plural. (63.c) shows the contrast between 3-proximate agreement and 3'-obviative. (63.a) Person:  b) Number:  c) Prox/obv:  e-nikamo-yan  T am singing'  e-nikamo-t  's/he is singing'  e-nikamo-t  's/he is singing'  e-nikamo-t-ak  'they are singing'  e-nikamo-t  's/he (prox) is singing'  e-nikamo-t-a  's/he (obv) is singing'  Note that gender (animate/inanimate) in Nehiyawewin generally determines the verb type (for example VTA vs. VTI) and this in turn selects different verb steins or different verbfinal morphemes. Three types of Agreement Typically, temporal and aspectual properties are associated with states and events and are expressed in conjunction with a verbal predicate. In English, the copula in nominal clauses 21  Evidence for tense-marking of nominals occurs in Coast Salish (cf. Burton 1996) signifying, for example, that a person is dead, i.e., my late grandfather. There is a similar morpheme in Ojibwa (Nichols 1980) and in Potowatomi (Hockett 1966). As noted in Wolfart (1973:31) this morpheme /epan/ no longer exists in Nehiyawewin. 138  21  ~ as well as in some verbal paradigms — functions as the verbal component to instantiate person agreement (and tense). In Nehiyawewin, there is no equivalent of a copula in nominal clauses; therefore, a tensed clause requires a verb (see (50) above). Predicational agreement occurs on non-verbal lexical predicates, i.e., nouns and adjectives, and is typically restricted to number and gender agreement between a subject and predicate. This contrast is schematized in (64). 22  (64. a)  Agreement Verbal  Predicational  Number Gender Person  number gender  In Nehiyawewin, there is the added category of proximate/obviative marking which is not part of the usual paradigm for predicational marking. (Note that only obviative is marked with a suffix while proximate is unmarked or 0-marked.) This third type of agreement, for proximate/obviative, is shown in (64.b). (64.b)  Agreement  Verbal Number Gender Person  Predicational  Proximate/obviative  number gender  This is also the type of agreement which occurs between nouns and their modifiers/adjectives and determiners within a DP. In Nehiyawewin, deictic determiners also agree with their NP for proximate/obviative status.  22  139  This three-way contrast with respect to agreement is not restricted to Nehiyawewin; similar contrasts are found in Semitic languages, for example, in Modern Hebrew (Rapoport 1987, Doron 1983, Dechaine 1993). (65)  Agreement  Verbal Number Gender Person  Predicational  Pronominal  number gender  The Hebrew examples given in (66) (from Rapaport 1987) show three degrees of agreement. In (66.a), the two nominal constituents agree for number and gender. This is restricted to predicative sentences, i.e., where the predicate NP is property denoting. The pronominal agreement in (b) (the third-person pronoun marking number and gender) is optional in predicative sentences but obligatory in an equative nominal sentence with null tense. Nominal sentences with past or future tense require a verbal constituent (copula) as shown in (c). (66.a) Dani more ba-universita. Dani teacher in-university Dani [is] a teacher at the university.  Present tense Number/gender agreement  b)  Dani hu more ba-universita. Dani 3sm teacher in-university Dani [is] a teacher at the university.  Present tense Pronominal agreement (number/gender)  c)  Dani haya more ba-universita. Dani be.Pst teacher in-university Dani was a teacher at the university.  Past tense Verbal agreement  140  In Rapaport's (1987) analysis, the equative and verbal examples in (b) and (c) require an IP projection with the pronominal agreement morpheme occurring in Infl. Nehiyawewin nominal clauses do not have a separate morpheme constituent in Infl containing agreement information. Gender is determined by the choice of wh-word, for example, awina 'who' and kikwdy 'what' or by the choice of determiner from the deictic paradigm shown above in (23). Plural forms are marked on the nominals with -ak 'plural' suffix  while singular is unmarked or 0-marked. I propose an IP structure for nominal clauses in Nehiyawewin. The subject and  predicate agree for gender and number (= predicational agreement). In my analysis, proximate/obviative agreement is obtained by movement of the predicate nominal to Spec CP of the nominal clause. This also accounts for predicate-initial ordering in Nehiyawewin nominal clauses. Nominal Clause Structures In Chapter 1,1 have adopted a basic head-initial spec-initial phrase structure for Nehiyawewin. This means that the basic IP structure of a nominal clause has the structure in (67.a). In this configuration, there is predicational agreement. This accounts for the basic agreement between the subject and predicate DPs for gender and number. (67.a)  IP  1=0 [SUBJ]  DP  [PRED]  141  In this structure, the subject (in Specifier position) precedes the predicate; however, it was observed above that nominal clauses are predicate-initial. I propose that this ordering is derived via predicate (XP) fronting to a position higher than IP, i.e., to Spec CP, as shown in (b) (cf. Dechaine (to appear)). (67.b)  CP PRED j  1=0  ti  [PRED] [SUBJ] Unlike in Hebrew, there is no evidence for a distinction between predicational nominal clauses and equational ones. In both cases the two nominals agree within the clause for proximate/obviative — though the clause as a whole may be unmarked (i.e., morphologically proximate by default) while the argument to which it refers in a following clause is obviative. This obligatory agreement does not appear to be significant with respect to the 23  distinction between predicative and equative clauses, as evidenced in Hebrew. However, this is yet another area which requires further study. Analysis of Nominal Clauses Given this analysis, we will look at a range of nominal clause examples and provide structures for them. The first set of examples is shown in (68).  Proximate NPs are unmarked (which also constitutes the default form) while obviative NPs are marked by -(w)a suffix.  23  142  (68.a) ni-snnis ana 1-younger sibling that That is my younger sibling. D. 182.a b)  eko ana napew Bill the very one that man Bill That man is Bill. LET.4  All of the examples have null tense, and the basic structure involves an IP projection as in (67.a). In (68.a) above, the predicate is proximate and the proximate form of the deictic is used — both are the unmarked forms. Given these constituents, the underlying form of the sentence is represented as in (69.a), with the subject DP preceding the predicate. (69.a)  ana pro 'that'  [SUBJ]  1=0  DP | ni-simis 'my younger sibling'  [PRED]  The subject and predicate nominals agree inherently for the lexical features animate and singular. In my analysis, the predicate DP raises to Spec CP for proximate agreement, leaving a trace of the moved DP in the predicate position as shown in (b). b)  CP ni-simisi  \IP DP  I ana pro  \  r  /  \  I  ti 'That is my younger sibling.'  [PRED] [descriptive]  [SUBJ] [deictic] 143  Clearly this movement analysis requires motivation. As noted above, nominals agree inherently for number and gender, i.e., predicational agreement. However, Nehiyawewin nominals in nominal clauses also agree for proximate/obviative status. It is this non-inherent agreement which triggers predicate raising to Spec CP for Spec-head agreement. Hence the obligatory predicate-initial ordering in Nehiyawewin nominal clauses. For example (68.b), I propose the structure in (70.a) where the third NP is adjoined to IP in the nominal clause structure: (70.a)  CP  Bill  ana N P 1= 0 'that' | napew 'man' SUBJ  DP | eko 'the very one'  PRED  This tree is represented in the bracketted structure in (b). b)  [CP  [rp [IP [ ana napew j p D  [ eko ] p ] Bill ]] D  In this example, the DP Bill is in apposition to ana napew 'that man' subject of the nominal clause. Again, the predicate raises for proximate agreement with its subject in (71 .a).  144  (71.a)  CP  eko; 'the very one' Bill DPj  A  ana N P I 'that' I napew 'man' PRED [descriptive]  DP  'That man is Bill.  ti  SUBJ [deictic]  Deictic DPs are the most referential and must occur as the subject of a nominal clause. As we have seen above, the deictic DP is always in second position, i.e., Nehiyawewin nominal clauses are predicate initial. The movement shown in (71 .b) accounts for this predicate-initial position. This tree is represented in the bracketed structure in (b). b)  [ p eko; [IP [IP [ ana napew ] C  D P  [t ] p ] Bill ]] ;  D  A more literal translation of the above would be: 'That man, Bill, is the very one (we were talking about)' or, reflecting the Nehiyawewin predicate-initial ordering: ' The very one is that man, Bill.'  The predicate in (72) shows obviative agreement between the subject and predicate DPs and a relative clause in apposition to the subject Mary-wa. (72)  ekoni [NP Mary-wa ] the very one(obv) Mary-obv It was Mary that John saw.  [cp ka-wapam-a-t John]] rel-see-dir-3 John cf.B.343  The same configuration accounts for the obviative agreement.  I argue that the relative  clause is IP-adjoined. The following diagram represents the output after predicate movement  145  in the nominal clause. The predicate ekoni 'that very one (obv)' shows agreement with the obviative status of the subject Mary-wa.  In the IP-adjoined (headless) relative clause with  kd- complementizer, there is null-operator movement, as shown. (73)  CP  The same analysis accounts for wh-clauses. The examples in (74) involve proximate agreement in (a) and obviative agreement in (b) with a possessor phrase as subject. The possessor is 3-proximate, and the possessee is therefore obviative. (74.a) awina ana Mary who that Mary Who is 'Mary'?  D.292  b) awtni-wa anihi o-kawiy-a who-obv that 3.Poss-mother-obv Who is his mother? I have suggested above that the determiner on subjects like Mary in (a) may be due to the fact that overt NPs are in argument (subject) position in nominal clauses; and as such,  146  they require a determiner. The only determiners available in Nehiyawewin are deictics. This is, however, an issue which requires further study.  24  Example (74.b) has a possessor phrase with a deictic determiner, as shown in the tree in (75.a). The possessor is 3-proximate, and the possessee is obviative. The wh-word shows agreement for the obviative subject. (75.a)  IP \ r  DP  A  / \  anihi NP 1=0 'that' | o-kawiy-a  DP | awina  3-mother-obv  'who'  [SUBJ]  [PRED]  The wh-word awina 'who' agrees with the subject for the lexical features animate and singular. The predicate DP raises to Spec CP for obviative agreement, leaving a trace of the moved DP in the predicate position as shown in (b). This tree is represented in (c). b)  CP awiniwai \ I P anihi I o-kawiy-a [PRED] [descriptive]  t; 'Who is his mother?'  [SUBJ] [deictic]  For example, it has been suggested (R-M.Dechaine p.c.) that this sentence might be literally: 'Who is she (that one), Mary'. This assumes a comma (= pause) preceding If there was one, it was not recorded in my notes. 24  147  Mary.  c)  [cp awlniwaj  [P [ anihi okawiya ]DP [ t; ]DP ] ]  4.1.4 Summary I have argued that wh-words can never be part of a verbal clause; rather, they occur in nominal clause constructions. In this chapter, I have looked at nominal clauses, in both English and in Nehiyawewin. In the course of the investigation, I have established that: (i) the ordering in Nehiyawewin nominal clauses is predicate initial; and (ii) that predicate-subject asymmetries in Nehiyawewin reveal a referential hierarchy similar to that of Heggie (1988). In this hierarchy, deictic DPs are the most referential and indefinites, including wh-words, are the least referential. Hence the wh-word is always in predicate position of the nominal clause. Given the underlying Spec-initial (i.e., subject-initial) structure proposed in Chapter 1 for phrasal projections (XP), the predicate-initial ordering of nominal clauses presents a problem. This problem is addressed via (iii), the agreement properties of nominal clauses. Predicational agreement in nominal clauses without a verbal constituent (copula) is typically restricted to the inherent lexical properties of a noun, e.g., gender and number. However, we have seen that there is also proximate/obviative agreement between the subject and predicate in Nehiyawewin nominal clauses. This additional agreement feature (iv) forces predicate fronting - accounting, not only for the agreement facts, but also for the predicate-initial ordering in Nehiyawewin. In conclusion, the sentence-initial wh-word arises via the more general operation of predicate fronting in all nominal clauses.  148  Chapter 5 CLEFTS AND CONJUNCTS  5.0 Introduction In this chapter, I provide an analysis of Nehiyawewin wh-questions. This analysis will account for simple nominal wh-clause structures as in (1) and for bi-clausal questions as in (2). The nominal clause subject may be an overt deictic like ana 'that one' as in the (a) examples, or null pro as in the (b) examples. In the biclausal examples in (2), the complementizer in the verbal clause may be kd- as in (a) or e- as in (b). (La) [p d awina] [ j ana pro ] who that (one) Who is he? // Who is that (one)? re  sub  B.399.a  b) awina [ p pro ] who Who is it/he? D  (2.a)  [cp awina ana ] [CP ka-sipwehte-t] who that rel-leave-3 Who is it that left?  b) [awina [DP pro ]] [cp e-sipwehte-t] who conj-leave-3 Who left?  I have argued in Chapter 4 that the nominal-clause subject and predicate agree for number, gender (animate/inanimate) and for proximate/obviative status. In Section 5.1, we will look briefly at proximate/obviative agreement across clauses. In wh-questions of more than one clause, the wh-phrase is the antecedent for an argument (pro) in a following clause. (More specifically, the wh-phrase is the antecedent of the null operator-variable chain which results from the movement ofpro (as null operator) to Spec CP of the verbal clause.) We  149  will see that the nominal clause subject and predicate — although they must agree with each other — may agree ([+AGR]) or not ([-AGR]) with the proximate/obviative status of a pro in argument position of the verb. This is discussed in Section 5.1. The analysis for wh-questions with kd- clauses is provided in Section 5.2.1 while the analysis for e- clauses is provided in Section 5.2.2. This latter section also provides a preliminary discussion of null operator movement with e- clauses.  5.1 Agreement Across Clauses In Chapter 4, we saw that nominal clauses may host a third NP or a relative clause as an adjunct within the clause. The adjoined relative clause contains an operator-variable chain which is coindexed with the wh-antecedent. The first evidence we will look at involves agreement between the wh-phrase and the proximate/obviative status of the referent in the following clause. Obviative is the marked status and involves a suffix (-hi or -(w)a) on the wh-word realized as awini-hi or awini-wa 'who'. The obviation marking provides the only overt evidence for the presence or absence of [AGR]. Therefore, we will provide an analysis based on the obviative agreement, which can then be generalized to proximate (unmarked) agreement. We will look at two types of Wh-questions: (i)  Wh-questions with kd- clauses: awina (ana) [ ka-... ];  (ii)  Wh-questions with e- clauses:  awina [ e- ... ]'.  The first set involves a nominal clause with kd- complementizer in the verbal clause. In this set, the nominal clause may have a deictic subject or a non-deictic subject. The deictic wh-  150  expression awina  ana  'who is that (one)' always takes kd- complementizer in the verbal  clause. The wh-expression may agree ([+AGR]) or not ([-AGR]) with the proximate/ 1  obviative status of the referent in the verbal clause. In (3), the unmarked [-obv] whexpression in the nominal clause has an obviative referent in object position of the adjoined clause. The [-AGR] agreement facts are illustrated in (b). (3.a)  b)  awina ana John ka-wT-wikim-a-t who that (one) John rel-fut-marry-dir-3 Who is it that John is going to marry? awina ana  [-obv]  [-AGR] 3 —» 3'-wh D.263  [Johnj... proj ... marry...prO[ \, ] ] +0  [-AGR]  •.  V  1  The bolded constituents constitute the nominal clause subject and predicate while the remainder of the sentence is the relative clause involving two third persons. The verb is direct with a 3-proximate subject and a 3'-obviative object. The proximate NP John is coindexed with the pro in subject position; therefore, it is the unidentified obviative object which is the referent of the wh-phrase. The wh-word and the determiner of the DP subject in the nominal clause — although they agree with each other as subject and predicate — are both unmarked for obviation and are therefore [-AGR] with respect to their referent in the verbal clause. I argue in Section 5.1.1 that the nominal clause hosts the relative clause via adjunction to IP.  One consultant insisted that an e- clause is not grammatical; and another consultant, when presented with an example using e-, hesitated and then simply stated her preference for kd-. As noted above and as we will see in Chapter 6, this wh-phrase seems to have a focussing effect. 1  151  In example (4), the (bolded) nominal clause constituents are [+AGR] with the coreferent object of the verb. (4.a) awini-wa anihi John ka-wi-wikim-a-t who-obv that John rel-fut-marry-dir-3 Who is it that John is going to marry? b)  [+AGR] 3 —» 3'-wh D.261  awina-wa^anihi [Johnj... proj...marry...pro  l+obv]  [+obv]  ]  [+AGR]  The verbal clause in (4) is identical to that in (3). The only difference is in the [+obv] features of the subject and predicate in the nominal wh-clause. In the nominal clause, the subject and predicate agree with each other for obviation; and together, they agree ([+AGR]) with the obviative referent of the verbal clause. In my analysis, the relative clause is adjoined to IP within the nominal clause in both (3) and (4), as represented in the structure in (5).  2  (5)  CP = relative clause  DP  Johnj... pro-y.. ka-wi-wikim ... t; 'marry'  pro  The relative clause in (5) is shown with a pro head. In contrast with complement type e- clauses, relative clauses (by virtue of the operator-variable chain) require an antecedent —  This is essentially the structure attributed to Ojibwa wh-questions in Truitner and Dunnigan (1972).  2  152  specifically, an N P .  3  In Section 5.2.1,1 consider some of the issues involved and alternative  adjunction sites for relative clauses.  4  An overt determiner like ana 'that' is not obligatory in wh-questions, as illustrated in the examples in (6). In all the examples, the referent in the relative clause is obviative. In (a), the wh-word is [-AGR], the default/unmarked form (= proximate). The verb in the relative clause is direct with a proximate subject (coindexed with John) and an obviative object which is coreferent with the wh-expression. In contrast, the wh-word is [+AGR] in (b). (6.a)  b)  awina  [John ka-ocem-a-t]  [-AGR]  who (obv) John rel-kiss-dir-3 Who did John kiss?  3 —» 3'-wh  awini-wa [John ka-ocem-a-t] who (obv) John rel-kiss-dir-3 Who did John kiss?  [+AGR] 3 —> 3'-wh  The examples in (7) are both [+AGR]. In (a), there is an NP John coindexed with the subject pro in the verbal clause. There is no lexical NP in (b); however, the wh-phrase is marked obviative and agrees only with the object pro in the verbal clause. (7.a)  awinihi [John ka-pakamahw-a-t] who (obv) John rel-hit s.o.-dir-3 Who(obv) is it that John(prox) hit? cf. B.360.b  [+AGR] 3 —> 3'-wh  As discussed in Williamson (1987, and references therein), the head of a relative clause is indefinite — the relative clause restricts the reference of the head and contributes the definite interpretation. 3  Whether or not all Nehiyawewin relative clauses are restrictive is an issue which must be addressed.  4  153  b)  awinihi [ka-nisokamow-a-t] who (obv) rel-help s.o.-dir-3 Who is it that she is helping? B.338  [+AGR] 3 —> 3'-wh  5  As we have seen in Chapter 3, the verbal clause in a wh-question in Nehiyawewin is not always a relative clause with kd- complementizer. Complementizer e- may also be used in the verbal clause in wh-questions. The examples in (8) involve a non-deictic DP as subject of the nominal clause together with an e- clause. These examples reveal the same [+/-AGR] contrast with e- complementizer. (8.a)  b)  awina  ana napew  e-pa-pakamahw-a-t]  [-AGR]  who that man (prox) conj-redup-hit-dir-3 Who is that man hitting? cf.MA. 158.b Literally: [Who is he; ] & [ that man is hitting him; ]  3 —> 3'-wh  awini-hi Mary e-wa-wapam-a-t who (obv) Mary conj-redup-see s.o.-dir-3 Who (obv) is Mary seeing (i.e., dating)? D.132 Literally: [Who is he; ] & [ Mary is seeing him; ]  [+AGR] 3 —> 3'-wh  The CP-adjoined e- clause is interpreted in the literal translations as being conjoined to the nominal clause.  5.2 Analysis of Wh-questions The section above has provided us with evidence for all possible combinations with respect to: i) the form of the wh-phrase; ii) the complementizer used; and iii) the [+/-AGR] features  The verb -niso-kamow- was used by Bill to translate 'help' and is listed thus with the above spelling in Faries (1938:95). Literally, it breaks down: niso-k-amaw ?'two- ? -benef-'  5  154  possible between the wh-phrase and the referent in a following clause. In this section, I provide an analysis of the following categories of wh-questions. In the table in (9), the wh-phrases in (I) are analyzed as clefted constructions and are characterized by the fact that the verbal clause has kd- complementizer. Wh-phrases occuring with a deictic constituent in the subject DP occur obligatorily with kd- complementizer, while in [-deictic] wh-phrases, kd- complementizer is an option. Otherwise, the conjunctive examples in (II) occur with ^-complementizer. In both clefted and conjunctive situations, agreement of the nominal clause DPs may be [+AGR] or [-AGR] for the proximate/obviative status of the coreferring constituent in the verbal clause. The two categories, clefts and conjuncts, are distinguished by the their adjunction site with respect to the nominal wh-clause. The clefted examples, as noted in (5) above, are adjoined to IP (Section 5.2.1). Conjuncts, as we will see (Section 5.2.2) are adjoined to CP.  155  (9)  Clefts and Conjuncts Wh-phrase (+ deictic)  I:  Clefts  awina  (- deictic)  Complementizer  [AGR]  ana ka-  awinihi anihi  [-/+AGR]  awiniwa anihi awina awinihi  ka-  [-/+AGR]  e-  [-/+AGR]  awiniwa awina  II: Conjuncts  awinihi awiniwa  It should be noted that not every speaker uses the entire range, but they all seem to be available and grammatical.  6  5.2.1 Wh-questions with kci- clauses In this section we look at wh-questions with kd- complementizer.  In addition, wh-questions may also occur with the independent form of the verb. This category I have not fully researched, and I have, therefore, not included it above. 6  156 Adjunction Sites for kd- clauses There is operator movement in kd- clauses at S-structure. The relative clause cannot stand alone as an independent clause; it must be hosted by the nominal clause ~ either adjoined to the NP subject in the nominal clause as in (10.a) or adjoined to IP as in (b).  IP  CP = relative clause  ti We have these two possible adjunction sites in the nominal clause for relative clauses. We also have two possible patterns of [+AGR] and [-AGR]. Proximate/obviative agreement is not obligatory between the nominal clause constituents and the pro (cum null operator) in the relative clause. Given the above, we have two possible accounts in (11) for the [+/-AGR] facts:  157  (11.a) That [+/-AGR] reflects structural differences, i.e.: [+AGR] = NP adjunction [-AGR] = IP adjunction b) That there is no structural correlation. I argue for (b) on the basis of the following two arguments: (i) that e- clauses show the same [+AGR] and [-AGR] facts — and e- clauses are uniformly CP adjoined to the nominal clause (see discussions in Section 5.2.2); and (ii) that, with a structural account as in (1 l.a), one would expect differences in interpretation. For example, with IP adjunction, the relative clause would be interpreted as an appositive structure while, with N P adjunction, the relative clause would have restrictive interpretation. However, this is not the case; the interpretation is always that of a restrictive relative. There is no contrast in the interpretation between [+AGR] and [-AGR] examples in this respect; and there can be no correlation between the adjunction site and the restrictive/nonrestrictive interpretation of relatives. I therefore adopt the configuration in (lO.b) which generalizes a uniform IPadjunction site for all kd- clauses. I propose, further, that this constitutes a clefted structure. This [+/-AGR] configuration is illustrated in (12). (12)  [CP awina; [ [IP (ana) pro; tj ] [n> pro [ p Opi Johnj pro, ... hit... ti ] ] ] | [+/-obv ] | | [+obv] | C  The IP-adjunction site in (12) allows the independent proximate/obviative status of the subject nominal, i.e., the subject may be either [+AGR] or [-AGR].  158 Clefted Wh-phrase with ka- Complementizer Clefted wh-phrases occur with a deictic determiner in the subject DP ([+deictic]) or without a deictic determiner ([-deictic]). In the following set, the wh-phrase may be [-AGR] or [+AGR] in relation to the referent in the accompanying clause. The following examples are [+AGR] with an obviative referent in the following clause. Example (13.a) involves a deictic while (b) does not. (13.a) awini-wa anihi [ Op; [ John ka-wi-wikim-a-t ] ] who-obv that (one) John rel-fut-marry-dir-3 Who is it that John will marry? D.261 Literally: Who is that [ Op; [ John will marry tj ] ] b)  awinihi [ Op; [ ka-nisokamow-a-t ] ] who (obv) rel-help s.o.-dir-3 Who is it that she is helping? B.338 Literally: Who is it [ Op; [ pro is helping t; ] ]  [+AGR] 3 ->3'-wh  [+AGR] 3 ->3'-wh  By generalization from the obviative examples above, the examples in (14) are [+AGR] for the proximate (subject) argument in the verbal clause. (The unmarked status of proximate does not distinguish between marked and unmarked examples.) The (a) example involves a deictic while the (b) and (c) examples do not. (14.a) awina ana [ Op; [ ka-ocem-a-t John-a ] ] who that (one) rel-kiss-dir-3' John-obv Who is it that kissed John? D. 17 Literally: Who is that [ Op; that [ t; kissed John ] ] b)  c)  awina [ Op; [ ka-wapam-isk ] ] who rel-see-dir-3>2 Who is it that saw you?  [+AGR] 3-wh->3'  [+AGR] 3-wh -> 2 D.106  awina [ Opi [ ka-pikiskwat-it]] who rel-speak to-3>l Who is talking to me? J.29  [+AGR] 3-wh -> 1  159  In the following examples with a deictic DP subject, the wh-phrase is [-AGR] with the obviative referent in the following verbal clause. (15.a)  awina ana [ Op; [John ka-wi-wikim-a-t]] who that (one) John REL-intend-marry-dir-3 Who is that one that John will marry? D.263 Literally: Who is that [ Op; [ John will marry t; ] ]  [-AGR] 3 —> 3'-wh  b)  awina ana [ Op; [ kahkiyaw aniki ka-wapam-a-t-ik]] who that (one) all those rel-see-dir-3-pi Who is it they all saw? D.251 .b Literally: Who is that [ Op; [ they all saw ti ] ]  [-AGR] 3.pl->3'-wh  The example in (16) is without a deictic and the wh-word is also [-AGR]. (16)  awina [ Opi [John ka-ocem-a-t] who (obv) John REL-kiss-dir-3 Who did John kiss?  [-AGR] 3 —> 3'-wh  Literally: Who is it [ Op; [ John kissed ti ] ] In (13.a), for example, there is agreement within the nominal clause awini-wa  anihi  'who is that (obviative)' between the wh-predicate and the deictic DP in subject position; and both in turn agree [+AGR] with the obviative status of the referent in the verbal clause. In my analysis, predicate fronting registers this agreement on the wh-word. The operator movement inside the relative clause completes the link identifying the argument referred to in the wh-question. (17) represents (13.a), repeated below, with the relative clause adjoined to IP.  160  (13.a) awlni-wa anihi John ka-wi-wikim-a-t Who is that [ Op: [ John will marry tj ] ] (17.a)  CP  ./\ awiniwa; •who' IP DP  anihi 'that'  NP | pro,  IP / T  I  DP | wh-t  wikim'marry' b)  tj  [CP who; [IP [IP that pro t; ] [cp Op; [ Johnj proj will marry t; ] ] ] ] | [+obv] | |_ [+obv] | In the following example, representing (14.b), the anaphoric element in the following  clause is proximate — which is not morphologically marked. The above analysis is generalized to the proximate example in (18) as [+AGR].  161  (14.b) awina [ Op; [ ka-wapam-isk ] ] Who is it that saw you? (18a)  DP 0  NP  T I  pro\  DP wh-ti  wapam-  b)  [CP who;  I  [IP [IP0 pro  [-obv]  t; ] [ p Opi C  |  |  [ U see pro.2  pro.2  ]]]]  |  The object pro.2 is second person; and because there is only one third person involved, it is 3-proximate in agreement with the wh-word (and the pro subject) in the nominal clause. The diagrams in (19) illustrate example (15.a), which is [-AGR] with the default form of the wh-word. The relative clause is adjoined to IP inside the nominal clause. The proximate subject John in the relative clause is coindexed with the subject pro, and the remaining pronominal argument moves to Spec CP creating the A'-chain for the null-operator which is associated with kd- clauses. There can be no ambiguity as to the referent of the whphrase.  162  (15.a)  awina ana John ka-wi-wikim-a-t Who is that [ Opi [ John will marry t; ] ]  (19.a)  y\ ana 'that'  DP NP |  pro\  T  / \ I  DP | wh-t  wikim' marry' b)  [CP who; [IP [n> that pro t\ ] [ p Op; [Johnj proj will marry ti ] ] ] ] | [-obv] | | [+obv] | C  The examples in (19) are [-AGR] with an obviative referent which is the marked form with overt agreement morphology. In summary, I have proposed the following for relative clauses with kdcomplementizer. Relative clauses cannot occur as an independent CP and are uniformly hosted in an IP-adjoined position inside the nominal wh-clause in Nehiyawewin wh-questions. In this configuration, the wh-phrase is clefted. Wh-phrases involved with kd- clauses may have a deictic determiner in the subject DP of the nominal clause. If there is a deictic, then the kd- complementizer is obligatory in the  163  verbal clause. If there is no deictic in the subject DP, then the kd- complementizer is not obligatory. In all cases, there is agreement for proximate/obviative between the subject and wh-predicate in the nominal clause. However, the agreement of the nominal clause constituents with the proximate/obviative value of the anaphoric element in the following verbal clause is not obligatory. The nominal clause constituents may be morphologically unmarked while referring to a [+obviative] argument of the verb.  7  Before leaving this section, we can speculate as to why e- complementizer is impossible with NPs introduced by a deictic determiner. Recall that when the subordinate clause has e- complementizer, it is interpreted as conjoined with the matrix nominal whclause: (i) Who is he; & Mary likes him;. Note that anaphora across coordinate clauses yields the semantics of a restrictive relative clause, i.e., (i) is equivalent to (ii) 'Who is the one who Mary likes'. Now, if the subject in the nominal clause were introduced by a deictic determiner, and the subordinate clause were introduced by e- complementizer, then the resulting interpretation would be that of an appositive relative: (iii) 'Who is [ that pro ]•, & Mary likes him;' (Ross 1967; also see Williams 1988). In sum, anaphora across coordinate clauses, where the antecedent is a DP with a deictic determiner, would not yield the semantics of a restrictive relative but rather that of an appositive relative. We can therefore suggest that e8  The reverse is not possible, i.e., the nominal clause constituents cannot be marked [+obviative] while the referent in the following clause is [-obviative].  7  D. James (p.c.) observes that, in fact, typical relative clause environments in Moose and Plains Cree, for example, may have e- complementizer, i.e., with a plural head noun, for example. She cites the Plains Cree example: ni-ki-wdpam-d-wak mitoni e-iyinisi-cik 8  164  complementizer is illicit with deictic subject DPs in the nominal clause because the resulting structure would not have the interpretation of a restrictive relative. Conversely, kdcomplementizer is licit with a deictic subject DP in the nominal clause precisely because kaclauses are unambiguously interpreted as relative clauses restricting the reference of the subject i.e., Who is [ the [ >#> pro; [ that Mary likes t; ]]].  9  Next we turn to wh-questions which have an e- complementizer.  5.2.2 Wh-questions with e- clauses Clauses with e- complementizer typically occur as complement clauses; however, they also occur in Nehiyawewin wh-questions.  10  In my analysis, there is no necessary operator  movement in e- clauses; an e- clause can stand alone as a CP and does not require an antecedent (= head). As a result, a clause with an e- complementizer cannot be hosted inside the nominal clause in the same manner as a kd- clause.  11  The alternative is clausal adjunction  to CP. The two configurations in (20) show the difference between an e- clause adjoined to CP in (a) and an IP-adjoined relative clause in (b).  I-saw-them really they-be intelligent beavers. T saw some very intelligent beavers' (Ahenakew 1982:2). Presumably, this would be an example of an appositive relative. amiskwak  9  10  Thank you to Hamida Demirdache for this insightful suggestion. As noted in Appendix C, e- clauses can also occur in a main clause context.  " One might also argue that a verb may select a clausal complement; however, a nominal predicate cannot.  165  (20. a)  CP CP  / \ / \  CP CP  IP  D PD  i  Subj  I  \[e- clause ]  IP  [.  r  p  DP  / \  DP  r/. clause] [kd-  r  ~»D  Ti  CP  Subj I  Pred  Pred  The CP adjunction to CP configuration in (a) is symmetrical — the typical conjunction configuration [XP & XP] — while the DP adjunction to IP in (b) is asymmetrical. In my proposal for e- clauses, the e- clause is CP adjoined (= conjoined) to another CP. The question arises: why is there an obligatory anaphoric relation between the subject of the nominal clause and an argument in the adjoined/conjoined e- clause, as illustrated in the examples in (21). (21.a) [ Who is pro; ] & [ Mary kissed him;] b) *[ Who is pro; ] & [ Mary kissed hiirij (= s.b. else) ] I argue that it is null operator movement in the conjoined CP which forces the anaphoric relation between the two arguments. A truly quantificational operator like every has a range, i.e., it ranges over a group of possible referents. A null operator, on the other hand, has no range; it therefore requires an antecedent (cf. Chomsky (1982, 1986), Williams 1988)). The evidence for this lies in a variety of structures; for example: (i) purpose clauses, (ii) parasitic gaps, and (iii) corelatives. Before'we look at these structures in Section,  166  however, there is a point of clarification which must be made with respect to operator movement. Null-Operator Movement In the preceding discussion, I have been refering to null-operator movement at either Sstructure or at L F in Nehiyawewin wh-questions. In Chapter 3,1 noted that wh-questions require an A'-chain involving an operator and a trace/ variable. The operator may be either overt or null — and the operator movement may be overt (at S-structure) or non-overt (at LF). The possible combinations are illustrated in the following table (copied from (61), Chapter 3). (22)  Nehiyawewin Wh-operators: Overt Operator  Null Operator  SS movement  *wh-movement  kd- complementizer  L F movement  *wh- in situ  e-  complementizer  I noted that, since there is no overt wh-movement, there must be null-operator movement; and I proposed (as in (22)) that the null operator moves at different levels according to the complementizer used: at S-structure with kd- complementizer, and at L F with an e- clause. Kd- complementizer is obligatory in other null-operator environments (focussed NPs and relative clauses) and is also used in a variety of patterns in wh-questions. On the other hand, e- occurs normally in non-operator environments but may also occur in wh-questions. In other words, operator movement with e- complementizer must be an option — it occurs if  167  required for wh-interpretation of the wh-question. The question then arises: at which level(s) does operator movement occur with respect to these two complementizers? Example (23) illustrates the possible combinations with respect to complementizers and the level at which operator movement occurs. (23)  Nehiyawewin Null-operator Movement kd- clauses  SS movement  S  L F movement  X  both  X X  e- clauses Wh-Q  Comp.  X  X  •/  X  I propose that the obligatory occurrence of kd- in null-operator environments like focussed constructions and relative clauses — as opposed to its absence in non-operator environments — precludes [X ] L F movement with this complementizer. Therefore, kd- is uniquely associated with S-structure movement. With e- complementizer, operator movement does not occur in regular complement clauses, but only when forced, as in wh-questions. The two situations are not overtly marked by a contrast in the choice of complementizer. I propose that LF operator movement takes place only in the cases triggered by a wh-environment. Thus, there is a three-way contrast between null operator movement at S-structure with kd- complementizer, LF movement of the null operator with e- complementizer in wh-environments, and the absence of operator movement in non-wh- e- clauses.  168 Null Operator Constructions W e will look briefly at three types o f structures to illustrate the role o f null operator movement in forcing anaphoric relations between an argument in an adjoined/conjoined clause with an antecedent in a preceding clause. The examples provided are from English and from Hindi. Purpose Clauses Consider the following examples from Rizzi (1986:514). The examples in the purpose clauses in (24.a) and (b) show a gap which is anaphoric with the D P the dog in the preceding clause. It should be noted that a resumptive 'it' could occur in place o f the gap.  12  Bolded [e]  represents the empty category/gap. (24.a) John bought the dog for Bill to give e to Mary, b) John bought the dog for Bill to give bones to e. The above sentences are represented by the structures in (25) which illustrate the null operator movement in the second clause. The null operator movement creates the gap and provides the link with the coindexed D P in the first clause. (25.a) John bought the dog; [CP Opi [ for b)  Bill to give tj to Mary.]]  John bought the dog; [cp Op; [ for [IP Bill to give bones to ti.]]  Recall in Chapter 3 example (47), the non-overt counterpart o f a resumptive pronoun is pro; and pro is the constituent which becomes a null operator (Cinque 1990, Browning 1992). 12  169 Parasitic Gaps A Parasitic Gap is the trace o f an empty operator which is parasitic on another operator movement. Parasitic gaps occur in adjoined or conjoined clauses where ah operator movement in the main clause is shadowed by a (coindexed) operator movement in the second clause. According to the analysis, this operation involves two distinct operator-variable chains (cf. Chomsky (1982, 1986) for adjunct environments, and Williams (1988) for conjoined environments). (26.a) Which boy; did you warn ti [CP Op; [before striking P G ; ]]? b) Who; do you love ti [CP Op; [and want to marry P G ; ]]? c) Who; does Jane respect t; [  CP  Op, [ and admire P G ; ]]?  In all three examples, wh-movement leaves a trace (gap) in object position o f the first verb and allows a corresponding operator movement and a gap in the adjoined/conjoined clause (illustrated in (c)). Thus, there are two A'-chains, the one with the null-operator being dependent on the "real" operator. The null-operator movement explains why the P G must be anaphoric with the wh-phrase, as was the case with the purpose clauses. Correlatives Correlatives are a common phenomenon in Hindi (Dwivedi 1994:8).  13  In the following set,  the (a) example is embedded, i.e., it is in its base position adjoined to the N P which it  Correlatives as found in Hindi, for example (cf Bains 1989; Dwivedi 1994; Srivastav 1991a,b) can be much more complex structures with two heads. A comparable English example might be: 'Whichever girl saw whichever boy, she liked him.' 13  170  restricts. In (b), the relative is right dislocated. In (a) and (b), the restricted D P is vo laRkii 'the girl' and the relative marker jo 'reP is in Spec C P o f the relative clause. In the correlative example in (c), the relative marker is the determiner in the D P jo laRkii  'which  girl' -- and the phrase has quantificational force (Srivastav 1991, Bains 1989). In the original relative clause, the demonstrative vo 'the/that one' occurs in lieu o f the relative marker as a kind o f resumptive pronoun. (27.a) Embedded Relative: [vo laRkii ] jo khaRii hai lambii hai dem. girl rel standing is tall is 'The girl who is standing is tall. b)  Right-adjoined Relative: [vo laRkii ] lambii hai jo khaRii hai dem. girl tall is rel standing is 'The girl who is standing is tall. Literally: The girl is tall, who is standing.  c)  [vo laRkii ] dem. girl  [vo laRkii ] dem. girl  Correlative: (cf. similar examples in Bains 1989) <Relative> <Correlative> [jo laRkii] khaRii hai vo lambii hai rel girl standing is dem tall is 'The girl who is standing is tall. Literally: 'Which girl is standing, that one is tall'  Ijo  laRkii]  rel  girl  In the (27.c) structure, the initial /o-clause is the relative and the second vo-clause is the correlative (Dwivedi 1994:8). English correlative examples shown below are given in (Dwivedi 1994:112 ff.).  14  According to Dwivedi (1994:2), the correlative precedes the relative in the English structures, the reverse of the Hindi structure in (27. c) above. 1 4  171  (28.a) When you tell me, then I shall go. (McGregor 1977) b) What he said, that I didn't understand. c)  That which you tell me, it I will do. (McGregor 1977) In the following example (29), I attempt to schematize the relevant structure for  (27. c). B y way o f illustration, I use the quantificational whichever  man in a sentence with  transitive verbs in imitation o f the literal gloss for (27.c), i.e., 'Which girl is standing, that one is tall'. The gap in object position o f the second clause makes the null operator movement more apparent. In the first clause, the quantificational D P moves creating an A'-chain. In the second clause, there is null operator movement with a focussed pronominal as antecedent. The pronominal must be coreferential with the operator in the first clause. (29)  Whichever man; Judy married t;,  15  him; [ p Opi [I don't like ti ]] C  The point is, in order to get the coindexed reading (which is obligatory) there must be null operator movement in (29) « and in the Hindi examples in (27). It is the null-operator movement which accounts for the obligatory coindexing as shown. In summary, we have looked at examples o f purpose clauses, parasitic gaps and correlatives. A l l these are instances where there is obligatory anaphoric relation between a pronominal in an adjoined or conjoined clause with an antecedent in the preceding clause. A l l o f these structures can be shown to involve a null-operator in the dependent clause — either at S-structure or at L F . The null operator is non-quantificational and has no range; therefore,  Bear in mind that I am not claiming that this is a correlative structure per se — just that it resembles the structure and the movement which occurs in a genuine correllative. 15  172  it requires an antecedent to fix its range. It can only acquire an antecedent via predication (Chomsky 1986). Movement creates the open position that allows the rule o f predication to apply and, thus, coindex the null operator with an antecedent. The relevant example is repeated here with the L F operator movement indicated in (b). (30.a)  [ W h o is h e ] & [Mary kissed himi]  b) [ Whoi is he ti ] & [ Opi Mary kissed ti ] This provides us with an explanation for the obligatory coreference in wh-questions with an e- clause. In my analysis, the e- clause (CP) is conjoined to the nominal clause C P containing the wh-phrase, as shown in (20.a). There are two alternatives (Cinque 1990) for a null-operator analysis: (i) the null operator may be base generated in Spec C P , or (ii) there may be operator movement to Spec C P . I am assuming null-operator movement at L F , as discussed above. A s we will see in the following chapter, Nehiyawewin shows evidence for movement in island effects, including both adjunct islands and wh-islands. Wh-Questions with e- Clauses The following examples involve an e- complementizer in the verbal clause. There is no Sstructure operator movement in <?- clauses, and the verbal clause is CP-adjoined to the nominal clause. The null-operator movement occurs at L F . In (31), the anaphoric constituent in the verbal clause is obviative and the wh-phrase is [+AGR].  173  (31)  awinihi [ Mary e-wa-wapam-a-t] who (obv) Mary conj-redup-see s.o.-dir-3 Who (obv) is Mary seeing (i.e., dating)? D . 132 Literally: Who is he; & Mary is dating him;  [+AGR] 3 —> 3'-wh  The [+AGR] features o f the wh-word is evidence that the pro subject in the nominal clause in marked [+obv] in agreement with the intended referent in the following clause. The tree in (32.a) represents the S-structure o f this sentence while (b) represents the structure at L F . (3 2. a)  1 6  CP CP  /\ awinihi; 'who'  IP / \  DP  A  r  /  0 NP I  \ NP  Maryj  pro.3'  wawapam(dating)  ^pro.3',  The [+obv] agreement [3'] is marked in the nominal clause. However, in order to force the coreference between the antecedent and the anaphoric constituent in the following clause, null-operator movement is required at L F . The L F representation o f the above struture is represented in (32.b).  1 adopt the convention o f labelling the pro arguments in tree structures for their person reference, (i.e., pro.3' (= 3-obviative), pro.2, etc. for added clarity. 16  174  (32.b)  CP CP awinihi; who'  CP IP  /  Op; \  r  DP  A  0 NP pro.V\  I  A\N P wh-t;  wawapam' dating'  t;  The LF-operator movement forces the (anaphoric) coreference between the obviative object o f the verb wawdpam-  'dating' and the wh-antecedent in the nominal clause. A t L F ,  this constitutes a parasitic gap structure with two A'-chains (Chomsky 1986a:98). However, this type o f structure is also reminiscent o f Williams' (1988) A T B analysis involving conjunction. Consider the [+AGR] examples in (33) which are proximate — generalizing the analysis for the obviative forms to the proximate examples. (33.a) awina  b)  e-wikim-a-t John-a  [+AGR]  who conj-marry-dir-3 John-obv Who married John? J.799 Literally: Who is she; & shei married John  3-wh —> 3'  awina e-wihtam-isk eko acimowin who conj-tell-3>2 the very one story W h o told you that story? J.852 Literally: Who is she; & she; told you that story.  [+AGR] 3-wh —»3'  175  (33.c) awina e-itwe-t e-sakih-a-t John-a who conj-say-3 conj-love-dir-3John-obv Who said she likes John? B . 122 Literally: Who is she; & she; said [that [she; likes John]]  [+AGR] 3-wh—> 3'  The unmarked wh-word represents a proximate argument in the following clause. The Sstructure o f (c) is represented in (34). This sentence involves three clauses, the nominal whclause and two e- clauses, the second embedded as the complement in the first e- clause. (34.a)  CP  awina; who / 1  IP \  e  IP / \  sakih'love'  pro.V\  The unmarked subject in the nominal clause agrees with its intended referent in the verbal clause. The L F representation o f the above structure is shown in (34.b) and (c).  176  (34.b)  CP  CP^^ awmihii IP who' / \ DP I'  ^CP Opi  C / \ e-  IP  Move' sakihc) [ p Who; [rp pro C  I  pro.3'j  tj ] & [ p Opj [n> tj said [ p that [IP pro, likes John ]]]] C  I  C  I  II  I  Coreference between the anaphoric pro. 3 and the antecedent in the nominal wh-clause is forced by null operator movement in the top e- clause of the conjoined structure, leaving a trace in subject position. The coreferring pro. 3 in the embedded complement clause is coindexed with the variable which binds it from the matrix clause. (See Chapter 6 for discussions regarding complement clauses.) In (35.a) and (b), the unmarked wh-word is [-AGR] with its referent in the following clause.  177  (35.a) a w i n a ana napew e-pa-pakamahw-a-t who that man (prox) conj-redup-hit-dir-3 Who is that man hitting? M A . 158.b Literally: Who is he; & that man is hitting him; b) a w i n a Mary e-wapam-a-t who Mary conj-see-dir-3 W h o did Mary see? M A . 3 92 Literally: Who is hei & Mary saw him;  [-AGR] 3 -> 3'-wh  [-AGR] 3 -> 3'-wh  Recall that there is always the option o f using the default form for wh-words. The generalization seems to be that the use o f the default form increases with distance or with interference from some topic-like element between the wh-clause and the clause which is the source o f the wh-extraction. The structure o f example (35.a) above is illustrated as in (36). The unmarked form o f the subject in the nominal clause is construed with the [+obv] object in the conjoined clause. The proximate subject argument is represented by the overt D P in the clause. (3 6. a)  awina;  0  IP  AN P IA  D P ana  NP napewj  pro.3j  \  pakamahw-  V pro. 3 '•,  In order to force the coindexing o f the antecedent and the anaphoric element in the following clause, as shown in (36.b) and (c), null-operator movement occurs at L F .  178  (36.b)  CP CP  c)  CP  [CP Who; [IP pro  tj ] & [CP Opi [IP that man hit ti ]]  With some speakers, the [-AGR] examples with an e- clause have more restricted distribution. The [-AGR] bare awina 'who' wh-phrase is often associated with a sentence which has a topic-like constituent. (37)  awina  wiya  John e-ocem-a-t  [-AGR]  who E M P H John conj-kiss-dir-3 Who did John himself kiss? cf.D.32 Literally: Who is shei & John himself kissed hen  3 —> 3'-wh  The intensifying emphatic pronominal wiya 'himself (Blain 1994, 1996) only associates with a topic-like overt N P .  1 7  The complementizer e- shows conjunction with another clausal  It is clear that this is not a focussed N P structure because there is no kd- complementizer on the verb. 17  179  constituent, i.e., the wh-phrase. In the following example, an emphatic second-person pronoun affects the o f the question. (38)  18  awina kiya ki-iteyiht-e-n John e-pakamahw-a-t [-AGR] who Y O U 2-think so-th-2sg John conj-hit-dir-3 3 - » 3'-wh W h o do Y O U think John hit? D.259 Literally: Who is he; & (do) Y O U think [ John hit him; ]  The speaker would accept only the [-AGR] form o f the wh-word — no doubt due to interference from the intervening topic, which involves an intervening null operator. Compare this to the companion example with [+AGR] wh-word. Note that awina can occur with (39) but awinihi cannot occur with (3 8). (39)  awinihi ki-iteyiht-e-n  19  John e-pakamahw-a-t  who 2-think so-th-2sg John conj-hit-dir-3 Who do you think John hit? D.260 Literally: Who is he; & (do) you think [ John hit hiirij ]  [+AGR] 3 —> 3'-wh  The [-AGR] wh-word in all these examples refers to the obviative object in a following clause. In (39), the referent occurs in a clause which is the complement o f another (independent mode) clause. The structure o f example (38) is illustrated in (40); the emphatic pronoun is shown adjoined to C P in the topic position in accordance with Dahlstom (1995), and there is corresponding operator movement into Spec C P .  According to Dahlstrom (1995), overt personal pronouns (in Algonquian languages in general) typically represent a new topic (given the Independent mode of the verb ki-iteyihten 'you think s.t.', the pronoun cannot be interpreted as focussed). 18  Generally, where awinihi occurs, you can always use awina (the default form). However, the reverse is not necessarily true. 19  180  (40.a)  CP CP^^  ^ ^ C P  pakamahw'hit'  pro.3',  The unmarked subject in the nominal clause is [-AGR] with the obviative status of its intended referent in the verbal clause. The conjoined CP has the anaphoric constituent embedded in the complement clause as the object of pakamahw- 'hit'. The LF representation of the above structure is shown in (40.b). There is nulloperator movement within the complement clause ~ as well as (topic-related) null operator movement to Spec CP of the matrix clause. The operator movement in the complement clause forces the anaphoric link with the antecedent in the wh-clause, according to the obligatory pattern shown in the gloss of (38). However, agreement for [+obviative] between the embedded operator-variable chain and the wh-phrase is disallowed.  181  (40.b)  CP d ^ ^ ^ ^ ^ c p  'hit' pakamahw-  tj  This LF structure is represented in the bracketed structure in (c) for added clarity: (40.c) [CP Whoi [ proi t ] & YOU [ Op [ t say [ ti [ John hit t ]]]] ;  CP  k  [-AGR]  k  CP  ;  [+obv]  This [-AGR] form frequently occurs with other longer distance operator extractions, as shown in (41) in which the referent of the wh-word (i.e., the null-operator extraction site) even more deeply embedded. (41.a) awina e-itwe-yan e-iteyiht-am-an John e-ocem-a-t [-AGR] who conj-say it-2 conj-think-th-2 John conj-kiss-dir-3 3 —> 3'-wh Who did you say you think John kissed? B.497 Literally: Who is hei & you said [you think [ John kissed him; ]]  182  A s in (40.b) above, the L F wh-interpretation forces operator movement o f the coindexed pro o f the most deeply embedded clause to Spec C P and then to each intermediate Spec position creating a chain to Spec C P o f the matrix e- clause, as shown in (41.b). (41.b) L F : [ Who; (is) pro; t ] & [ Op [ proj say ] [ tj [ proj think ] [ t; [ John kiss ti ]]]] ;  ;  A l l the intervening subjects are accounted for — two o f them involve second person, and proximate John in the final clause is coindexed with the subject pro. In examples o f this sort, the awinihi  'who' form can usually be substituted; however, the sentence was usually  presented by the speaker with the default unmarked form o f the wh-word.  5.2.3 Summary In Section 5.2,1 have provided an anaylsis for wh-questions with kd- complementizer and with e- complementizer. With the kd- complementizer on the verb, the wh-expression may involve the bare wh-word or may include a deictic (i.e., ana 'that'). The deictic is prohibited in a wh-question with an e- clause.  20  With both e- clauses and kd clauses, we saw  examples where the wh-word is [+AGR] with its referent (the anaphoric constituent) in a following clause and we saw examples where the wh-word is [-AGR] with its referent.  Why do wh-questions with e- clauses clauses not contain a deictic determiner in the nominal wh-clause? I suggest that the [+deictic] wh-phrase forces a focus type construction (i.e., with Sstructure operator movement in the verbal clause and kd- complementizer. Focus constructions never involve e- clauses. 2 0  183  In my analysis, a kd- clause has S-structure null-operator movement (i.e, a relative clause structure). The kd- clause is hosted by the nominal clause in an IP-adjoined position.  21  A clause with e- is C P adjoined to the nominal clause C P and the two clauses get a conjoined reading. There is usually no operator movement in e- clauses ~ which typically occur as complement clauses. However, wh-questions require movement o f an operator and an operator-variable chain in order to force coreference between the anaphoric constituent in the verbal clause and the antecedent in the nominal wh-clause.  22  The two patterns for kd- clauses have been given the same basic analysis above; however, there is clear evidence (see Chapter 6) that they are different. The difference requies further study as to the semantics of these determiners in wh-questions and focus structures in general. 21  This can be interpreted as two types of relativization strategies ~ one with an overt relative clause and the other via conjunction (cf. Ross 1967, Williams 1988).  2 2  184  Chapter 6 ISLANDS AND WEAKEST CROSSOVER  6.0 Introduction In this chapter, I extend the investigation of Nehiyawewin wh-questions. Using the analysis for basic wh-questions developed in Chapter 5, we will consider the evidence for extraction from more complex structures including wh- and complex-NP Islands. In so doing, I show that the conclusions drawn from Chapter 3 — i.e., that Nehiyawewin wh-questions involve null-operator movement rather than wh-movement as in Mohawk (Baker 1996) — are the correct conclusions. In Section 6.1, we examine the evidence for extraction asymmetries between complement clauses and adjunct clauses. Adjunct clause examples involve clauses with 'because' and wh-islands occurring with wh-words, i.e.,  tdnehki  'why', and with  dsam  kispin  'whether, if. On the basis of the evidence for long-distance extraction, I conclude that Nehiyawewin complement clauses are in argument position. We will see that extractions from adjunct clauses are illicit; and, further, that extraction from wh-islands in complement position is illicit. Both these facts will be derived from Huang's (1982) Condition on Extraction Domains. In Section 6.2, we look in detail at the evidence for Weak Crossover. Baker's (1996) parasitic gap analysis for Weak Crossover in Mohawk makes the wrong predictions with respect to relative clauses in Nehiyawewin. I argue that the absence of Weak Crossover in  185  Nehiyawewin is due to the non-quantificational nature o f the operator which undergoes movement in wh-questions. The arguments are based on the analysis proposed by Lasnik and Stowell (1991) and Demirdache (1997) to account for the absence o f W C O in some structures in English. This analysis provides an account o f the absence o f W C O based on universal principles; and I extend this analysis to languages as diverse as English and Nehiyawewin. This contrasts with Baker's analysis, which is based on language-specific properties o f Mohawk.  6.1  Complement Clauses vs. Adjunct Clauses  In this section we look at the evidence for wh-extraction from complement clauses and adjunct clauses. Complement clauses in argument position allow wh-extraction while adjunct clauses do not. If all clauses in Nehiyawewin are adjoined or conjoined in the same manner, then we should not expect to get these extraction asymmetries. I conclude that, since extraction asymmetries do exist between Nehiyawewin complement clauses and adjunct clauses, Nehiyawewin complement clauses cannot be base generated in adjunct positions.  6.1.1 Long-Distance Extractions In the following examples, we will see that long-distance extractions are possible from complement clauses. Therefore, we can argue that these clauses must be in argument positions. Consider: (1 .a)  awina e-itwe-yan e-iteyiht-am-an John e-ocem-a-t who conj-say it-2 conj-think-th-2 John conj-kiss-dir-3 Who did you say you think John kissed? B.497  186  Since the complementizer in ( l . c ) is e-, the L F configuration o f this sentence has nulloperator movement from the most deeply embedded complement clause as in (l.b), represented by the tree structure as in (l.c). The operator moves from the lowest object position to the highest Spec C P successive cyclically. (l.b)  [ W h o i s p r o ; ] & Opi [you said ti [you think tj [John kissed ti ]]]. I  I  I  I  In the tree structure in (l.c), the matrix e-clause is CP-adjoined to the nominal whclause. The operator movement shown is at L F . Given that the long-distance extraction in (1) is grammatical, we can assume that complement clauses are in argument position. If they were in adjoined position, we would expect (1 .a) to be ungrammatical: it would be ruled out as a violation o f the C E D (Condition on Extraction Domains, cf. Huang 1982) which prohibits extraction from adjunct clauses.  187  proj  wh-tj  ocem'kiss'  The operator movement leaves traces in all the intermediate Spec C P positions.  188  6.1.2 Island Extractions In contrast to complement clauses, which freely allow extraction, adjunct clauses do not allow extraction.  These clauses are referred to as adjunct islands, and the Condition on  Extraction Domains ( C E D ) prohibits movement out o f adjunct clauses. O n the other hand, movement out o f wh-questions in complement position violates subjacency. In the following sections, we will be looking at three types o f islands: (i) adjunct Islands with osdm 'because' (; (ii) indirect questions introduced by a wh-phrase (; and (iii) an indirect question introduced by kispin 'if, whether' ( The examples occur with various combinations o f clauses involving complementizer kd- (relative clause) and/or e-. Adjunct Islands with osdm 'because' In the example in (2), the relative clause is adjoined to IP. It contains an osdm 'because' clause introduced by the e- complementizer which is itself adjoined to I P inside the relative 1  clause. In accordance with Huang's (1982) Condition on Extraction Domains ( C E D ) , extraction from the strong island with osdm is ungrammatical. (2)  *awina ka-mato-yan osam e-pikon-a-t kit-awasisihkan-a who rel-cry-2 because conj-break-dir-3>3' 2-doll-obv Who did you cry because ... broke your doll? B.431 [Who is proi [ Opi [ [ you cried ] [ because ti broke your doll ]]]]  Recall that CPs with e- are CP-adjoined to a preceding nominal clause (i.e., a clause without a verb). However, the clause introduced by osdm 'because' is dependent on the preceding verbal clause; hence the IP-adjunction. As noted by R - M Dechaine (p.c), VP-internal adjunction is unlikely since all VP-internal operations seem to be valency affecting (Goddard 1990) 1  189  In (3), the verbal clause has e- complementizer and is CP-adjoined to the nominal whclause. The because clause has kd- complementizer and, thus, involved S-structure operator movement. Crucially, since it is an adjunct clause inside the matrix e- clause, operator extraction from the because clause is, once again, prohibited by the C E D . (3)  * awinihi Mary e-paslkwetah-a-t John-a [osam ka-ocem-a-yi-t ] who (obv) Mary conj-slap-dir-3>3' John-obv because rel-kiss-dir-obv-3>3' Who is it that Mary slapped John because he kissed (her)? B.340 A s noted above, because clauses are adjuncts and strong islands (cf. Cinque 1990);  and extraction is disallowed regardless o f the nature o f the clause, i.e., the C E D holds at Sstructure and at L F . In the diagrams in (3.b) and (c), the clause with e- is CP-adjoined to the nominal whclause. The operator movement shown is at L F and it violates the C E D since it takes place from within an adjunct clause. (3.b)  [Who is proj] & [Op; [ Mary slapped John [because [he kissed ti]] I  *  I  In the tree below, the adjunct clause with kd- is IP-adjoined inside the matrix eclause.  2  Recall also that I argued in Chapter 5 that e- clauses are C P adjoined to nominal clauses, i.e., an e- clause cannot be hosted by a nominal clause. However, e- clauses may occur in argument position o f the verb as seen in (1) and may otherwise be adjoined as a dependent clause o f another verbal clause as in (3). 2  190  'slap'  'kiss'  A s noted above, null-operator movement is prohibited from the adjunct osam 'because' clause by the C E D . Wh-Islands A complement wh-question is an "island" for extraction, i.e., operator movement is prohibited from wh-questions.  In English, these island effects are attributed to Subjacency  (Chomsky 1973, 1986; and cf. discussion in Ouhalla 1994, for example).  (4)  Subjacency Movement cannot cross more than one bounding node in a single step, where bounding nodes are IP and D P .  191  Subjacency prohibits extraction from any complex structure like wh-questions and complex D P s (relative clauses) — both o f which have an operator in Spec C P , preventing the movement o f another wh-phrase which requires a trace in that Spec C P position.  3  Given the internal structure proposed for wh-questions in Chapter 5, what happens when a wh-word is extracted from a wh-island in Nehiyawewin? Recall that subjacency does not apply to adjuncts (Ross 1967, Chomsky 1973, Huang 1982).  4  Let us consider the structures involved in (5), for example. W e see that a relative kdclause is IP-adjoined in the nominal ' w h o ' clause and contains an embedded 'why' question which also has kd- complementizer. Operator extraction from the object position o f the tdnehki 'why 'clause is not licit. In English, extraction would be ruled out as a subjacency violation. Is the ungrammaticality o f (5.a) a subjacency violation? The structures in (b) and (c) illustrate example (5.a). (5.a)  b)  *awinihi ka-kakwecim-at Mary tanehki ka-pon-kiyokaw-a-t who (obv) rel-ask s.o.-2>3 Mary why rel-stop-visit-dir-3 *Who did you ask Mary why she stopped visiting? B.448 *Who is pro3'i [ Opi [IP [ you asked Mary ] WHYk [ she stopped visiting tj tk ]] I  *  '  *  i  The examples below illustrate subjacency violations with the wh-island example in (ii) and the relative clause example in (iii). Example (i) is licit with an intermediate trace in Spec C P o f the complement clause. The bounding nodes are indicated with an asterisk [*]. i) Which car; did [IP you think [ p t'i [c (that) [IP John would fix tj ]]]]? ii) *Which can do [IP* you wonder [cp when [ - [IP* John will fix t; ]]]]? iii) * Which can have [IP* you met [DP* someone k [CP w h o k [ e [IP* tk can fix t; ]]]]? 3  C  C  Baker (1996) uses subjacency to "derive the C E D " in his parasitic gap analysis o f Mohawk relative clauses (which are adjuncts).  4  192  .c)  **  CP  IP  A\ r A A\ DP  0  NP  pro.3',  I  DP  0 NP  NP  pro,  A\ CP  wh-ti  'ask'  'visit' kiyokaw-  ti « *  Recall that the relation between adjunct wh-phrase and its trace is not mediated by a null operator (which is not the case with argument-type wh-phrases). Nothing which follows hinges on this assumption. The verb ask in the intervening clause is a three-place verb. Semantically, there are 5  two possible referents inherent in the ask clause (other than the two arguments ~ agent and goal ~ marked on the verb): you ask Mary about someone; or alternatively you might ask Mary a question like Why did..? In both cases, these additional referents -- the oblique phrase or the wh-question — are IP-adjoined within the clause containing a transitive verb  The glossary in Wolfart and Ahenakew (1993:108) gives the following: ASK: kakwecim- VTA 'ask s.o.; make a request of s.o.; ask s.o. about (it/them).  5  193  (VTA) with agreement morphology for two animate arguments. Only the inflectionally marked arguments can be in A-positions (compare examples (5) and (9), and observe fn. 9) . There are two possible analyses of (5.c): (i) the extraction [ « * ] from the wh-island must cross two IP nodes violating subjacency (since CP2 is not available as an escape hatch ~ it is occupied by why); or (ii) this is simply a CED violation involving extraction from an adjunct clause. Given the adjunct position of the wh-question, this must be considered a CED violation. In the examples in (6), with the wh-word ktkwdyiw 'what (obv)' , the following clause 6  has e- complementizer. The e- clause is CP-adjoined to the nominal wh-clause. The embedded who (nominal) clause has an IP-adjoined kd- clause and the entire who question is in complement position in the e- clause — the question as a whole is not an adjunct — although, you will note that the verbal clause from which the extraction is to be made is itself an adjunct within the who question. The verb kiskeyiht- 'know it' is a VTI verb with an inanimate object, i.e., the wh-question. Null-operator extraction from the complement whclause is not licit. (6.a) *[kikwayiw] & [e-kiskeyiht-am-an [awina [ ka-wapaht-am-k]] what-obv conj-know-th-2sg who REL-see s.t.-th-3 What is it that you know who saw (it)? B.424.b b)  [What is pro] & [Op; [you know [ W H O is proj [Opj [tj saw ti ]]]]] 1  *  *  *  1  This form of kikwdy 'what' is apparently a variation on the obviative form (H.C. Wolfart, p.c). However, I question whether it is not an actual wh-predicate with verbal agreement for thirdperson, i.e., kikway-i-w = what-vb-3 'What is it?'. 6  194  The (Opj ...tj) movement in (6.b) and (c) below occurs at S-structure in the embedded wh-question with kd- complementizer. The matrix clause has e- complementizer; therefore, extraction from its complement would occur at L F in my analysis. The L F extraction o f the object argument is ungrammatical. Why? Because, given the syntax proposed for wh-questions in Nehiyawewin, extraction from within a wh-question will always be taking place from within an adjunct island. In (6), extraction o f the object 7  takes place from the verbal clause adjoined to the nominal wh-clause; and, as such, it is ruled out by the C E D .  (6.c)  CP  / \ kikwayiwj what'  IP /  DP  0 NP I | pro.O',  \ r  NP | wh-tj  kiskeyihtjQP 'know' / awinah  y ^  *  IP DP  A  0 NP pro  h  CP  /\ A T'  I  Op  DP  wh-t h  C'  h  ka-  IP * VP  'see it' wapaht-  tj  «  This is true whether the wh-question is itself in complement position as in (6) or in adjunct position as in (5). 7  195  To conclude, the proposal that wh-questions in Nehiyawewin involve a nominal clause (Who is pro), whose subject is anaphoric with an argument in an adjoined clause, elegantly explains why extraction from a wh-question is ungrammatical. Whatever the combination of e- and/or kd- complementizers involved, it clearly has no effect on the grammaticality o f the sentences in (5) and (6). Though, in my analysis, kd- clauses have Sstructure null-operator movement and e- clauses have L F movement, the choice o f complementizer does not affect the grammaticality o f the wh-island extractions. Wh-Islands with kispin 'if, whether' Clauses with //and whether are also classified as wh-structures. In example (7) below, the relative clause is IP-adjoined inside the nominal wh-clause. The island involves kispin 'if, whether' and has kd- complementizer with S-structure operator movement. The whether clause is an indirect question and is IP-adjoined in the 'ask' clause because it is not licenced by agreement morphology. The (S-structure) operator extraction from the (adjunct) 8  whether clause involves a C E D violation because it involves extraction from an adjunct clause. (7.a)  b)  *awina ka-kakwecim-i-yan kispin John ka-sakih-a-t who rel-ask-dir-2>l whether John rel-love-dir-3 Who did you ask me whether John loves? B.355 [Who is pro; [ Op; [ you asked me ] [ whether [John loves ti ]]]  I  i  D . James (p.c.) refers to kispin as an embedded yes/no question. Nehiyawewin typically has ci 'Q-marker' in yes/no questions. However, ci is not used in these constructions.  8  196  In (8), the 'ask' clause has e- complementizer and is CP-adjoined to the nominal whclause. The whether clause is IP-adjoined (an adjunct) in the e- clause. As in (7) above, operator extraction from the whether clause is not allowed. The same analysis holds with Operator L F operator movement out o f the adjunct island. (8.a) *awinihi e-kocim-i-yan kispin John ka-pakamahw-a-t who (obv) conj-ask-dir-2>l whether John REL-hit-dir-3>3' Who did you ask me whether John hit (him)? B.425 b)  [Who is pro;] & [Op; [you ask me [ IF [ John hit t ]]] ;  I  I  A s with the wh-island examples above, the kispin 'whether' clause is an island for operator movement at L F — a violation o f the C E D . In Section 6.1.1 above, we saw that long-distance movement is grammatical from complement clauses which shows that they are in complement position o f the verb. We have also seen that extraction from both adjunct clauses and wh-questions, violates the C E D . Extraction is not allowed from o f an embedded wh-question because, given the internal structure proposed for wh-questions, extraction is taking place from an adjoined clause (i.e., from a verbal clause adjoined to the nominal wh-clause). Escaping the Island Effects Another speaker (Jane) has a way o f bypassing the island effects in examples like those discussed above. In her alternative strategy, she reinterprets long-distance extraction as short-distance extraction. She uses the wh-phrase awina ana which requires (obligatory) kdin the following clause.  197  In (9), the verb ask is a three-place verb (compare with (5) above). In this example, the deictic ana 'that (animate one)' forces the animate "about pro" reading o f the ask clause generating a corresponding (IP-adjoined) oblique pro.  9  This pro becomes the target for  questions. In other words, ana forces the extraction from the ask clause as shown in (9.b) and (c). In these examples, a short distance extraction is forced from the matrix clause — thus avoiding the necessity for long-distance extraction from the island (wh- or adjunct). In this manner, this consultant is able to avoid the subjacency and C E D violations, and the resulting sentence is grammatical. (9.a)  b)  10  A w i n a ana ka-kakwecim-i-yan tanehki ka-sakih-a-t Mary-wa who that rel-ask-dir-2>l why rel-love-dir-3 Mary-obv Who is it that you asked me why he loves Mary? J.962 Who is that one [Opi [ [ you asked me] about ti ]]] & [Why does pro, love Mary?] I  I  The resulting grammatical sentence is then a conjunction o f two separate wh-questions. The pro in the conjoined why question is coindexed with its antecedent as shown in the bracketed structure above. The structure proposed for this sentence is shown in (9.c). The wh-phrase is coindexed with the proximate referent (aboutpro.3) in the relative clause which is adjoined to IP in the subject position o f the nominal clause.  The tanehki 'why' clause is  adjoined to C P and has the interpretation o f a conjoined clause.  Browning (1982) states: " A l l empty categories are subject to an identification requirement. When pro is not identified by a strong agreement, it has to move to an A'-position in order to be identified via a rule of predication that co-indexes it with an antecedent." 9  1 ran both sets of examples by another speaker, Mary Ann, reading the Cree and asking for grammaticality judgements. The Island extraction examples in the previous section were not good to her while the examples in this section ranged from sort-of-okay to good. 10  198  kakwecim'ask'  pro. 1  We can assume either that coindexation takes place in the same way as antecedentpro relations in other structures or that this anaphoric relation is established via operator movement. This would yield basically an ATB extraction, i.e., with two conjoined clauses -[you asked me about pro ] and [why does pro love Mary]. Coindexation between these pros and the pro of the nominal clause [Who is that pro] would be forced by null operator movement across the board (that is, in each conjunct). Note that, if this is the case, then subjacency cannot hold at LF (as is standardly assumed) since LF operator movement to the Spec of the second conjunct (i.e., in the why clause) would be a subjacency violation. As shown above, the sentence is grammatical.  199  The C E D violation in the kispin 'if, whether' examples in (10) uses similar strategies — although, in this case, the final C P (as well as the oblique phrase) is adjoined within the kdclause. (10.a) awina ana John ka-kakwecim-isk  kispin Bill-a e-wikim-a-yi-t  who that John rel-ask-3>2 i f . Bill-obv conj-marry-dir-obv-3 Who is it that John asked you if Bill married? J. 977 b) Who is that pro: [ Op; [IP [[ John asked you ] about ti ] (and) if Bill married pro, ]] I  I  The coindexed pro in the IP-adjoined kispin ' i f clause is bound by the trace in the IPadjoined oblique phrase. There are also examples which use the same tactic to avoid C E D violations in adjunct islands with osam 'because' as illustrated in (11). In the following example, the because clause with e- is adjoined to IP inside the IP-adjoined relative clause. The short-distance extraction involves an oblique pro in the matrix clause as we saw in examples above. The coindexed pro in the IP-adjoined osam 'because' clause is anaphoric with the trace in the IPadjoined oblique phrase, as shown in (1 l.b). (11.a) awina ana ka-mato-yan osam e-pikon-a-t kit-awasihkan-a who that (one) rel-cry-2 because conj-break-dir-3 2-doll-obv Who is it that you cried because he broke your doll? J.996 b) Who is that pro, [Op, [IP [ [you cried ] about t; ] because pro, broke your doll ]] I  I  In summary, we have seen that extractions are possible from complement clauses while there is no extraction from adjunct clauses. This indicates complement clauses are in argument position. We have also seen that extractions are prohibited from wh-questions both  200  as adjuncts and as complements of the matrix verb. Extraction from adjunct islands violates the CED (cf. Huang 1982). The ungrammaticality of extraction from wh-islands provides strong support for the analysis of wh-questions defended here: all wh-questions involve adjunction (either to IP or CP) with respect to the nominal wh-clause itself. Hence the CED uniformly accounts for the evidence presented here.  6.2 Weak Crossover vs. Weakest Crossover  In this section we look at the evidence for Weak Crossover (WCO) in Nehiyawewin. Weak Crossover is a test for wh-movement. We noted in Chapter 3 that Baker's (1996) parasitic gap analysis accounts for the WCO evidence in Mohawk. However, this parasitic gap analysis makes the wrong predictions for Nehiyawewin which does not show evidence of WCO effects. In Baker's analysis, there is wh-movement of the wh-words from argument position to Spec CP. For Nehiyawewin, however, I look to the analysis in Lasnik and Stowell (1991) which accounts for the absence of WCO in some constructions in English with an analysis which depends on the nature of the operator. This analysis is termed Weakest Crossover.  6.2.1 Weak Crossover  Weak Crossover is a diagnostic for wh-movement within a clause. Consider the English sentences in (12), (a) shows a pronoun embedded in a possessive NP subject and coindexed with an NP in object position. This example is licit because the object NP is not bound by the pronoun inside the complex NP. The wh-movement in (b) leaves a trace in object position of  201  the clause creating a Weak Crossover violation.  11  When the possessive N P is in object  position as in (c), the wh-movement in (d) does not show W C O effects. (12.a) b) c) d)  [His; dog] chased Johnj. * W h o i did [his; dog] chase ti? John; chased [hisi dog]. Who; tj chased [his; dog]. The same evidence occurs with relative clauses in (13).  When the complex N P is in  subject position and the wh-variable precedes the pronoun as in (b), once again the sentence is ungrammatical. The subject/object asymmetry holds for English relative clause examples as shown by the grammaticality o f the example in (d). (13.a) b)  [The woman he; loves] chased John . ;  * Whoi did [the woman he, loves] chase ti?  c)  John; chased [the woman he; loves]?  d)  Who; ti chased [the woman he; loves]? The configuration which violates W C O is shown in (14) where the coindexed  pronoun precedes the trace o f the moved wh-word. (14)  * W h o ; . . . hei... ti?  These phenomena are descriptively captured in the Leftness Condition (Chomsky 1976, Koopman and Sportiche 1982) -- a version o f which is presented here:  (15)  The Leftness Condition: A wh-trace cannot be co-indexed with a pronoun to its left.  For various versions o f Weak Crossover, see Chomsky (1976), Koopman and Sportiche (1982), and Reinhart (1983:122), for example. 11  202  I do not depend on any specific theory of WCO. However, because the analysis in the following sections should hold regardless of the principles of the grammar used to derive WCO, I use the descriptive generalization as stated in The Leftness Condition in (15). Weak Crossover in Mohawk  Mohawk is a rich head-marking language with pro in argument positions and overt NPs adjoined to IP. Baker (1996) claims that there is wh-movement in Mohawk and thus predicts Weak Crossover effects. However, Baker shows that there are no subject/object asymmetries with respect to either possessor phrases (which never trigger WCO) or relative clauses (which always trigger WCO). WCO tests show that possessor phrases are grammatical in subject and in object position as shown in (16): (16.a) Uhka wa'-te-shako-noru'kwanyu-' rao-skare' who fact-dup-MsS/FsO-kiss-punc MsP-friend Whoi kissed his; girlfriend? b) Uhka wa'-te-shako-noru'kwanyu-' ako-skare' who fact-dup-MsS/FsO-kiss-punc FsP-friend *Whoi did hen boyfriend kiss? The (a) example is grammatical in Mohawk (as it is in English). Example (b) is ungrammatical in English and should not be good in Mohawk, which also has overt whmovement (Baker 1996); however, as seen above, it is perfectly good in Mohawk. On the other hand, relative clauses examples are ungrammatical in both positions: 17.a)  *Uhka wa'-t-huwa-noru'kwanyu-' ne rukwe ne ruwa-nuhwe'-s who fact-dup-FsS/MsO-kiss-punc NE man NE FsS/MsO-like-hab Whoi kissed the man shei likes?  203  b) *Uhka wa'-ti-shako-noru'kwanyu-' ne rukwe ne shako-nuhwe'-s who fact-dup-MsS/FsO-kiss-punc NE man NE MsS/FsO-like-hab *Who; did the man who likes her, kiss? In Baker's analysis, the wh-word moves out of its argument position to Spec CP. If DPs were in A-positions, there should be a subject/object asymmetry as in English. However, the proposal that DPs are not in argument positions does not explain why WCO is absent with possessor phrases but always occurs with relative clauses. Baker proposes a parasitic gap analysis to capture the WCO facts for both possessive DPs and relative clauses. Parasitic Gaps in Mohawk Baker's (1996) account of WCO examples in Mohawk uses a parasitic gap analysis. Parasitic gaps (PG), as the term suggests, are gaps which are parasitic on existing operator-variable gaps (cf. Chomsky 1986, Williams 1988, for example). As a result, PGs can occur only in conjunction with operator constructions, for example, wh-questions and relative clauses. Within this context, they commonly occur in structures which involve coordination or some other type of linking word; for example: (18.a) Who does Mary love t and want to marry PG? b) Who is the man [that Mary loves t] and wants to marry PG? c) Who does John want to marry t because he loves PG/her d) Poirot is a man [ you distrust t ] when you meet PG/him. e) Who did pictures of PG annoy t? In (a) to (d), the clauses with the parasitic gap are adjoined or conjoined to the clause containing the wh-movement. The final example (e) contains a kind of possessor phrase.  204  Baker's analysis with parasitic gaps accounts for the difference between possessor NPs and relative clauses. The preceding discussion bears on the issue in two areas: First, in Mohawk, all overt lexical NPs are adjuncts. As shown in the examples in (18), adjunct clauses (and conjunct clauses) are typical extraction sites in a parasitic gap analyses. Secondly, the relevant adjuncts in the Mohawk examples include both possessor NPs and relative clauses associated with pro arguments of the verb. Of these two types of adjuncts, relative clauses are islands for parasitic gaps. Parasitic gaps involve movement, and movement cannot occur out of a relative clause without violating subjacency. In this way, Baker uses subjacency ~ via the parasitic gap analysis ~ to derive the CED effects which he obtains with complex relative clause DPs.  12  Thus, possessive DPs, as seen in (16) above, show no contrast for WCO effects; both are grammatical, in contrast to the subject/object asymmetry found in English. The following example shows Baker's parasitic gap analysis of possessor DPs. In this analysis, the possessor is not a pro, but a parasitic gap, i.e., the trace of an empty operator. The movement of the "real" wh-operator in the clause is shadowed by the movement of an empty operator in the parasitic gap construction (based on the parasitic gap analysis of Chomsky (1982, 1986). The bracketted structure in (19.b) represents (16.b). In this structure, each operator — the real one and the shadow one — separately binds its variable in the IP and in  CED violations are involved with extractions from adjunct clauses. In none of Baker's (1996) examples is the primary extraction from the adjunct phrase or clause. Rather the whmovement is out of argument position of the verb while the "extraction" in the adjunct clause is secondary extraction of a null-operator, i.e., a parasitic gap. The purpose of the extraction is to obtain coindexation with an empty category (pronominal) in conjunction with the WCO configuration. 12  205  the adjoined D P respectively, circumventing a W C O configuration. This produces the grammatical interpretation in (a) in Mohawk. (19.a) U h k a wa'-te-shako-noru'kwanyu-' ako-skare' Who; did hen boyfriend kiss? b) [ p who; [IP [IP pro kiss t ] [ Opj [NP P G ; friend] ]] C  ;  While the wh-operator A'-binds its trace in the clause, the shadow/empty operator A'-binds the empty position in the adjoined D P , and weak crossover is avoided in this possessive D P structure. The example in (16.a) with the complex D P associated with the object argument is analyzed in the same manner. As seen in examples in (17) above, Baker shows that the Mohawk relative clause examples cannot have a bound reading. Both the subject and the object examples are ungrammatical in Mohawk. Consider the structure in (20), representing example (17.a). (20.a) *Uhka wa'-t-huwa-noru'kwanyu-' ne rukwe ne ruwa-nuhwe'-s Who; kissed the man she; likes? b) [CP whoj [IP [IP ti kiss pro ] [ Opi [NP man [ p [n> P G i likes him ]]] C  I  I  I  *  k  * I  The ungrammatically o f the relative clause example is derived via the subjacency violation with the parasitic gap extraction. Baker (1996:82) describes the ungrammaticality o f this structure as follows: "[PGi] cannot be a parasitic gap, because it is not subjacent to a potential operator position with scope over N P . ... Neither can [PG;] be a bound pronoun, k  because it is not c-commanded by the trace o f the wh-phrase. [changes mine]" The structure and the analysis is the same in (21), which represents (17.b). The following example is ungrammatical in Mohawk and in English:  206  (21.a)  *Uhka wa'-ti-shako-nonj'kwanyu-' ne rukwe ne shako-nuhwe'-s *WhOi did the man who likes her; kiss?  b) [CP who; [IP [IP pro kiss t; ] [ Op; [HP man [ I  I  I  CP  [ip he likes P G ]]]  *  ;  *  I  Again, subjacency is violated by the null-operator movement out o f the relative clause island. Thus, the parasitic gap analysis accounts for the ungrammaticality o f both relative clause examples. The wh-movement in Baker's account provides the 'real' operator movement which is necessary for the parasitic gap analysis — an analysis which provides Baker with an account o f the absence o f W C O in the possessive N P examples in (16) and the presence o f W C O in the relative clause cases in (17). Baker's analysis cannot be used for Nehiyawewin for two reasons; the first is empirical and the second is conceptual. Empirically, the evidence is different for Nehiyawewin, i.e., relative clause examples are grammatical. Compare the same sentence in M o h a w k (22.a) and in Nehiyawewin (22.b). There are no W C O effects in Nehiyawewin when the offending pronoun is embedded in an object relative clause. (22.a) Mohawk: *Uhka wa'-t-huwa-noru'kwanyu-' [ne rukwe ne ruwa-nuhwe'-s] who fact-dup-FsS/MsO-kiss-punc N E man N E FsS/MsO-like-hab Who; kissed the man she; likes?  b) Nehiyawewin: awina ka-ocem-a-t [anihi napew-a ka-sakih-a-t] who REL-kiss-dir-3 that(obv) man-obv REL-love-dir-3 Who; kissed the man she; loves? J.388 Conceptually, I choose an analysis which can uniformly account for the absence o f W C O in languages like English as well as Nehiyawewin and M o h a w k as opposed to a  207  language-specific analysis as proposed for M o h a w k in Baker (1996).  In particular, I argue  that it is the nature o f the operator that is responsible for the absence o f W C O . A nonquantificational operator does not trigger W C O , as evidenced in Nehiyawewin as well as in English (cf. Lasnik and Stowell 1991).  6.2.2 Weakest Crossover Lasnik and Stowell argue that Weak Crossover occurs only with the movement o f true quantifiers and wh-elements as shown in (23). In the Weak Crossover example in (a), the wh-movement is at S-structure. Example (b) has movement and W C O effects at L F .  (23) Weak Crossover: a)  *Whoi does his; son love ti?  b)  S-structure:  *Hisi son loves everyone;.  *LF:  [ p Everyone; [n> his; son loves t; ]] C  The examples in (24) represent Weakest Crossover (Lasnik and Stowell 1991); in these examples, we will observe three properties: (i)  These constructions involve movement o f a null-operator rather than a quantificational wh-operator.  (ii)  A null-operator does not range over a set (as do quantificational operators); rather, its range is fixed via coindexation with an antecedent.  (iii)  The examples are grammatical and do not exhibit W C O . This concurs with the analysis o f Lasnik and Stowell (1991) (hereafter, L & S ) wherein the  208  absence of WCO in the following constructions in English is attributed to the nature of the null operator. (24) a)  Weakest Crossover: (Lasnik and Stowell 1991) 13  Topicalization of a non-quantificational NP: John; [ Op; [ I believe his; mother loves t; ]] I  b)  I  Focus/clefts: It is this book; [Op; [that I got its; author to read from ti ]]  c)  Parasitic gap: i  I  Who; did [Op; [ his; mother's stories about t; ] annoy (ti)] I  I  In the (b) structure, the clefted constituent is base-generated in CP or a position adjoined to CP and is accompanied by null operator movement to Spec CP leaving an A'-trace in argument position. Topicalized element in (a) shows null-operator movement. Even though there is a gap (indicating movement), and there is a coindexed pronominal between the gap and the operator position, neither of these examples shows Weak Crossover effects. The wh-movement in the main clause in (c) triggers a shadow (parasitic) movement in the possessor phrase. Within the possessor phrase, there is a coindexed pronominal between the gap and the operator position -- but there is no Weak Crossover violation.  The structures which provide evidence of Weakest Crossover include any environment which resultsfromnull-operator movement. L&S (1991) also show examples involving pseudo-clefts, tough movement, and restrictive relatives. 13  209  For these and other cases of "weakest crossover", L&S (1991) make the following descriptive generalization: (25.a) WCO does not follow exclusively from the structural relation between the pronoun and the variable A'-bound by the same operator. WCO arises only in contexts where a pronoun is locally A'-bound at LF by a true quantifier ranging over a possibly nonsingleton set. b) Weakest Crossover: Null epithets do not trigger Weak Crossover effects.  14  For Demirdache (1997:231), on the other hand, it is not the nature of the variable (i.e., whether it is a null epithet) which determines the absence of WCO. But rather, it is the nature of the relation between the variable and the anaphoric element to its left; WCO does not occur when this relation is one of Strict Coreference. Demirdache (1997) Demirdache's analysis is devoted to maintaining the status of WCO as a diagnostic for movement. Her analysis allows for both alternatives in any situation, i.e., the WCO interpretation and a Weakest Crossover interpretation. Demirdache supports L&S's (1991) claim that "there are no A'-binding configurations that are immune to WCO". It is simply that some situations offer an alternative solution. Demirdache's analysis of L&S's (1991) English examples is based not on the status of the variable or on movement, but on the status of relations between the constituents involved. When there is an anaphoric relation of strict coreference, then WCO effects can be  A trace of a non-quantificational operator is termed by L&S (1991) a null epithet — which is the null counterpart of an anaphoric epithet like "the bastard" and "the idiot" (Demirdache (1997: note 18)). 14  210  circumvented. Strict coreference will occur in situations involving an empty operator and an anaphoric pronoun since, following Browning (1982) and Cinque (1990), Demirdache analyses null operator as pro. Consider the sentence in (26.a) which involves a parasitic gap and does not trigger W C O . In order to see how the analysis works, consider only the bolded parasitic gap structure in (b) — disregarding the actual wh-movement in the matrix clause. The key factor is the relationship between the Operator, the pronominal possessor in the N P and the trace and the operator. They are all coreferential pronominal forms whose relationship is one o f strict coreference.  15  Demirdache derives the analysis from Chomsky's (1992) copy theory o f  movement with the crucial assumption that the copy remains at L F until it is deleted. (26.a) Who did his mother's stories about P G annoy?  b) Who; did [ Opi [ his; mother's stories about t; ] annoy t; ]? pro; pro; I  pro; |  < copy  c) Who; did [ Op; [ his; mother's stories about tj ] annoy? ] *LF:  pro;  pro;  t;  Demirdache (1997) analyzes a null operator as pro following Browning (1992). Demirdache (fh. 2 4 . ) states: "For arguments that the empty category in parasitic gap constructions is pro, see also Cinque (1990) who analyses a parasitic gap as a pro A'-bound by a base-generated nulloperator. Browning (1982) motivates her analysis as follows: All empty categories are subject to an identification requirement. When pro is not identified by a strong agreement, it has to move to an A'-position in order to be identified via a rule o f predication that co-indexes it with an antecedent." 15  211  d) Whoi did [ Opi [ his; mother's stories about tj ] annoy? ] I  ^LF:  pro;  I  I  t;  pro;  The steps in the analysis are as follows: (i)  The first step is to copy the constituent which is being moved. In (b), the pronominal complement o f the preposition about is copied — one pro moving as the null operator and the other pro remaining in situ. This gives us the sequence o f three coindexed pronominals.  (ii)  Then, under the condition o f identity, one o f the pros must be deleted to form the movement chain. The configuration in (c) derives from the deletion o f the pro in final position leaving a trace as shown in (c). A W C O configuration occurs and the L F interpretation is bad.  (iii)  If the other (leftmost) pro is deleted under identity with the pronominal operator in Spec C P leaving the trace in the intermediate position, as in (d), then there is no W C O violation and the L F interpretation is good. The latter is the Weakest Crossover interpretation.  B y contrast, the movement via copying o f a quantificational wh-word, as shown in the sentence in (27) without a parasitic gap, does not result in strict coreference between the three coindexed constituents. It is the wh-operator itself which is copied and there is no relation o f strict coreference (i.e., identity) between the pronoun and the quantificational operator in Spec C P . There is only one choice for deletion as shown in (b) and the resulting configuration is ungrammatical as shown.  212  (27.a)  *Whoi does [his; mother love who; ] I  Q-whj  I  pro*  I  Q-whi  I  b)  I  *Whoi does [ his; mother love t; ] I  *LF:WCO  Q-whi I  I  I  pro;  tj I  This contrast between quantificational wh-operators and null operators represents the essence o f Lasnik & Stowell's (1991) generalization in (25). Empty operator movement provides the environment for the strict coreference (identity) relations which are the basis o f Demirdache's (1997) analysis as outlined above. Only when there is this anaphoric relation o f strict coreference can W C O effects be circumvented. Supporting evidence for this claim is provided by Demirdache (1997) with examples in (28) involving resumptive pronouns (RP). A n R P is defined by Demirdache as the in situ counterpart o f a null-operator; "it is a pronominal operator whose range is fixed by an antecedent via predication." Unlike a null operator, however, it moves at L F as opposed to S-structure. In the Arabic examples in (28), the S-structure in (a) is grammatical with the two coreferential pronominal forms in situ. The copy stage is illustrated in (b) with three coindexed pronouns. The L F in (c) with the rightmost pronominal deleted results in a W C O configuration. The L F in (d), with the leftmost pronoun deleted under identity, avoids the W C O configuration (data from Demirdache 1991, 1997). (28.a)  S-structure [CP" kull walad, [ C [IP 'urn-ulu bithibb-uh; every boy mother-his loves-him ""Every boy, his mother loves him'  213  b)  COPY:  [cpkull walad every boy  [ lU; [IP um-uh bithibb- t; ]]] mother-him lovec  prorij c) * L F  [CP kull walad [c hu; [IP um-uh bithibbevery boy mother-him love0 i  prorii  P  d) ^LF  pron;  pron;  ti  [ p kull walad [c hu; [n> um- t; bithibb hu ]]] motherevery boy love him C  0  protii  P i  In Arabic examples below, when the pronoun in the possessor phrase is replaced by an (overt) anaphoric epithet, W C O is triggered. Example (29) involves a Q P antecedent while the N P in (30) is a proper name. In both examples, after the copy mechanism in (b) has been done, the only constituent available for deletion under identity is the pronoun in object position. W C O cannot be avoided in the resulting L F configurations in (c) for both examples. (29.a)  *kull walad;, 'um il /uimaar; bithibb-uh; every boy mother def-donkey loves-him *'Every boy, the donkey's mother loves him'  b) C O P Y :  *kull walad;, [ hu; [ 'um il ftumaar; bithibb-uh;  c) (30.a)  *LF:  I pron;  I NP;  I pron;  Op;  NP;  t;  *xaled;, 'um il ftumaar; bithibb-uh; Xaled mother def-donkey loves-him *'Xaled, the donkey's mother loves him'  b) C O P Y : *xaled;, [ hu; [ 'um il /jumaar; bithibb-uh; pron;  NP;  pron; 214  c) *LF:  Opi  NPi  ti  In the following examples from Hebrew, overt movement of the RP in (b) triggers WCO in relative clauses. The example in (a) without S-structure movement is grammatical. According to Demirdache's (1997) account, the (a) example has two possible configurations at LF, the ungrammatical WCO counterpart of (b) and the licit Weakest Crossover alternative as discussed in conjunction with the examples above. (3l.a)  ha-'is se 'im-Oj 'ohevet 'oto; the-man that mother-his loves him *'the man that his mother loves him.'  b) * ha-'is oto; 'im-Oi 'ohevet ti the-man him mother-his loves *'the man that his mother loves him.'  The examples above show how Demirdache's analysis works for Semitic languages.  6.2.3 Evidence for Nehiyawewin Baker (1996) uses a parasitic gap analysis to account for the evidence involved with complex NPs in Mohawk wh-questions. We saw above that the Nehiyawewin evidence for relative clauses is different. We will look at Nehiyawewin examples with possessive phrases and relative clauses. Recall that the wh-word is in the nominal wh-clause and there is null operator movement in the verbal clause. Possessive DPs There is no evidence for Weak Crossover in conjunction with Nehiyawewin possessor phrases — the subject and object examples are both grammatical. 215  (32.a) awina ka-nawaswat-a-t o-tem-a who rel-chase-dir-3 3-dog-obv Who; is chasing his; (own) dog? b)  3 —» 3' (direct)  awina o-tem-a ka-nawaswat-iko-t who 3-dog-obv rel-chase-inv-3'>3 *Whoj is hisi dog chasing? B.152.a  3'—>3 (inverse)  16  The example in (32.b) is represented as in (33). (33.a) b) C O P Y :  awina [ — [ o - tenia [ka-nawaswat-iko-t]]] Who; is [ pro, [ pro-, - dogk [ pro  k  chasing pro, ]]]  t c) * L F = W C O  Whoi [ O p ; C O M P [pro,-dog  l  Op; d) ^ L F  I  I  k  pro,  Whoi [ O p ; C O M P [ ti - dogk Op;  ti  [pro chasing l, ]]] k  I  t; [pro,, chasing pro,]]] pro,  According to the Weakest Crossover analysis outlined above, the pronominal argument is copied as shown in (b). The copy in operator position and its original in object position o f the clause remain at L F where there is now a string of three identical pro, constituents. This creates a situation o f strict coreference with three identical and coindexed pro constituents  In Nehiyawewin, the verb morphology can specify the coreference relationship. The following example forms a pair with (32.b) above which together represent the two possibilities: (i) awina o-tem-a ka-nawaswat-a-yi-t who 3-dog-obv rel-chase-dir-obv-3 3' —»3" Who; is hisj dog chasing ti? Whose; dog is chasing hitrij? (alt. Eng. gloss) The inverse direction on the verb in (32.b) (3' —»3) designates the proximate argument (i.e., the dog's owner) as the person being chased. In the (i) example, the morphology is direct and means that the object argument is someone other the the dog's owner. 16  216  with two possible LF interpretations. In the first as shown in (c), the copied pro, is deleted and the resulting configuration creates WCO effects. Alternatively, given strict coreference of the three constituents, the pro, in possessor position of the subject DP is deleted under identity as in (d). This yields the licit LF configuration which is referred to as Weakest Crossover. Since all three pro-, constituents in (b) are identical semantically, the switch does not alter the meaning; and because all the constituents are null, the PF output is never affected. The same analysis is available for example (32.a). Relative Clauses Relative clause examples are also grammatical in wh-questions. The following example contains a relative clause coindexed with a pro in object position of the verb. These are always good. Note that the matrix clause with kd- is a relative clause IP- adjoined in the nominal clause. The subordinate relative clause is adjoined to IP in the matrix verb clause. (34.a) awina ka-ocem-a-t anihi napew-a ka-takakeyim-a-t who rel-kiss-dir-3 that man-obv rel-like-dir-3 Who; kissed the man that she; likes? J.388 b) [ Who; is [IP [ pro; t; ] [ Op [n> [ tj kissed ] [that manj ] [ Opj [ she; likes tj ]]]]]] This sentence is represented by the tree structure in (34.c). In the object relative clause, I show anihi  ndpew-a  'that man' and the relative clause as separate IP adjunctions. The  relevant constituents are bolded.  217  (34.C)  IP  \  /  anihi that'  \  VP / \ ti V* ocem'kiss'  CP;  DP,  / \  Op,  NP | napewaj man'  ka-  C IP / \ VP / \ pro,  proj  V  takakeyim'like'  tj  Note that the operator movement o f the matrix subject leaves a trace in Spec IP where the subject picks up Case. The pro, in the IP-adjoined relative clause can be interpreted as a variable bound by the trace in Spec IP once we define binding in terms o f "exclusion" rather than c-command. Under the exclusion version o f binding, ti binds pro because the first node dominating t; (IP) does not exclude pro since there is a segment o f IP which dominates p r o . (35) L F :  Opi  17  1 8  In sum, there are no W C O effects with the configuration in (35). t  ;  pro;  A excludes B iff no segment o f A dominates B . The same c-command possibilities are not available for operator movement out o f object position (i.e., the V P dominates the object trace and excludes pro). This would predict that there is subject/object asymmetry in Nehiyawewin for extraction from adjoined relative clause N P s as (36.a) illustrates. However, we shall see that this W C O violation can be circumvented as (36.b) illustrates. 17  18  218  The null Operator of the matrix relative clause locally binds its own trace. The pro, in the IPadjoined relative clause is bound by the trace as shown in (35) and is coindexed with the operator-variable chain. Relative Clauses Associated with Subject When the relative clause is associated with the subject, the situation is not so straight forward. Consider the following examples.  19  (36.a) *awmihi napew ka-sakih-a-t ka-ocem-a-t who (obv) man rel-love-dir-3 rel-kiss-dir-3 *Whoi did [the man who loves her;] kiss? D.256 b) awina ana ana napew ka-sakih-a-t ka-ocem-a-t Who that that man rel-love-dir-3 rel-kiss-dir-3 "Who is it that the man loves and kisses?" MA.372 The bolded relative clause in both cases precedes the matrix verb. Both sentences were 20  composed by myself with the intention that the bolded relative clause represents the subject of the main verb. Otherwise, the two verbs both have kd- complementizer and are the same in every way. Yet the two structures are quite different as shown in (37) and (38.a). Note that, though the glosses are different, they convey the same information, i.e., the man loves her and he is kissing her. In (37), the sentence — as presented and glossed by me ~ was rejected by the speaker as ungrammatical. In my gloss, I had in effect imposed the WCO  19  The form of the wh-phrase does not, to my knowledge, affect or determine the gloss provided.  As in Mohawk, the relative may precede or follow the verb without altering the grammaticality judgements. In fact, the preferred word order for relative clauses in wh-questions is following the verb. 20  219  structure as illustrated in (37). Note that the DP napew represents the antecedent of the subject relative clause. (37) [Whoi|>rOi tj ][ Opi [n> [the manj [Opj [ tj loves hen] kiss tj ]]]]] *LF=WCO  Op;  proi  t  ;  In (36.b), represented in (38), a very similar sentence is presented by me without a gloss. The speaker accepted the sentence and provided her own gloss as shown in (36.b), thereby imposing her own structure as shown in (38). (38)  [Who;  [pro\  [ Opi [IP the manj loves  tj  ] & [ Op; [a>  pro  y  kisses  tj ]]]]]  The licit structure assigned to the string involves coordination of two flauses with ATB extration: Null operator movement takes place across the board in each conjunct. Thus, there are two A'-chains, each with a null operator. Note that the relative clause antecedent the gloss of (38) is not the same as in (37). (38) is a relative clause with no overt head; napew  represents the subject of the verb in the relative clause (to love). In (37),  napew  represents the antecedent of the corresponding subject relative clause. These structures require some clarification. First we will discus's what makes example (37) "bad" and then v will see how example (38) is "good". The bad example (36.a) (repeated here as (39.a)) represents the Weak Crossover interpretation of the sentence. (39.a) *aw?nihi napew ka-sakih-a-t ka-ocem-a-t who (obv) man REL-love-dir-3 REL-kiss-dir-3 *Who; did [the man who loves hen] kiss? D.256 (39.a) has a WCO interpretation at LF, and is represented by the tree structure in (40).  220  (39.b)  [Whoi [pro,  V, ] [ Opi [IP [the marij [Opj [ tj loves hen] kiss t; ]]]]]  *LF=WCO  I  I  I  Op;  pro;  ti  In the tree structure in (40) the coindexed pro, in the IP-adjoined subject relative cannot be interpreted as a variable bound by the trace (tj) in the matrix clause because the V P dominating tj "excludes"  pro,  21  sakih'love'  pro,  In the tree as shown, pro, precedes the matrix clause trace creating the W C O linear configuration as shown in (39.b). However, linear order is not the relevant issue. Even i f the relative clause subject followed the main verb, it is still attached higher in the structure (that is, it would still be IP adjoined) and W C O would still apply. Pro, would not be bound by tj because the V P dominating tj excludes pro,. Thus, the same hierarchical relations obtain. 2 1  221  Why is the Weakest Crossover interpretation not available? With  ndpewa  'man (obv)'  as head of the relative clause as shown, the Weakest Crossover configuration as shown in (41) is not possible because the null-operator movement out of the relative clause violates subjacency constraints [*]. This is illlustrated in the tree in (41.b).  22  (41.a) [Who;  [pro-,  X\ ] [ Op; [IP [DP  |  *LF=Subjacency  the manj [Opj [ tj loves hen] kiss ti ]]]]]  * *  *  Opi I  |  ti *  |  pro;  I  The chain ( Op; ti) is not licit because movement has crossed three bounding nodes.  sakih'love'  ti  This is similar to Baker's (1996) parasitic gap analysis for Mohawk. 222  Neither option — the WCO interpretation nor the the Weakest Crossover interpretation which yields a subjacency violation — is licit. Now we will consider the licit structure in (38) which bypasses WCO. The (36.b) example (repeated here as (42)) represents the licit interpretation. (42)  awina ana ana napew ka-sakih-a-t ka-ocem-a-t Who that that man REL-love-dir-3 REL-kiss-dir-3 "Who is it that the man loves and kisses?" MA.372 23  The licit gloss for this sentence was provided by the speaker (Mary Ann). The configuration shown in (43.a) is represented in the tree structure shown in (43.b): (43.a) [Who; [pro, [ Opi [IP the manj loves t; ] & [ Op; [IP prc-j kisses ti ]]]]] (43.b)  CP ana NP I pro,  NP Op; wh-t;  CP C  Op  (  C  ka napewaj 'man'  sakih'love'  ocem- tj 'kiss'  Similarly, two relative clauses conjoined in a narrative context: (i) ekoni anihi kikway [kikwey] mistahi ka-mayi-tota-ko-yahk ka-pikon-iko-yahk;... these those thing(s) greatly rel-bad-do to s.o.-inv-lpl rel-break-inv-lpl "These are the things that greatly hurt us and break us;..." (Wolfart & Ahenakew 1993:74).  23  223  When the gloss of the given Nehiyawewin sentence is left to the interpretation of the speaker, the licit structure shown above emerges. In this structure there are two parallel relative clauses (both clauses have kd- complementizer) each with the same obviative referent extracted via the null-operator movements. The coordination of two relative clauses provides an ATB (across-the-board) parasitic gap structure like that discussed in Williams (1988). However, Chomsky-style, there are two separate A'-chains. In Williams (1988) ATB parasitic gap structure, a single (overt) operator creates a gap in two parallel (conjoined) structures, i.e., an single headed operator chain with two variable chains. Chomsky (1986) maintains two separate and independent structures, the original A'-chain with an overt head and the parasitic A'-chain with a null head.  6.3 Comments In this chapter we have seen that there is clear contrast between long-distance extraction from embedded complement clauses and the absence of extractions from adjunct islands. These contrasts provide evidence that complement clauses are in argument position. I have established, further, that extraction from within wh-islands is ungrammatical. This fact provides strong support for the analysis of wh-questions defended in Chapter 5. In particular, both extraction from adjunct clauses and extraction from wh-islands are ruled out as CED violations (Huang 1982).  24  On the other hand, we have seen that there are strategies  More research is required with other types of adjuncts. The evidence available isfromdata gathered some years ago and there are gaps in the paradigm. 24  224  We then looked at the Weak Crossover evidence with complex NPs including relative clauses. Baker's (1996) parasitic gap analysis for complex DPs cannot be applied to Nehiyawewin because it incorrectly predicts that all relative clause examples will be bad. In fact, the opposite is true. I have proposed an alternative analysis to account for the absence of WCO effects in both possessor DPs and relative clauses. I have reanalyzed the absence of WCO in Nehiyawewin as an instance of Weakest Crossover. This is a phenomenon found in a wide range of languages including English (Lasnik and Stowell 1991) which occurs whenever movement of a non-quantificational operator gives rise to conditions of strict coreference in a string of pronominal elements. I have shown that this analysis accounts for the possessive DP examples. Moreover, it explains why object-related relatives never involve WCO while subject-related relatives contrast for good and bad interpretations (though both have the same meaning). In other words, these Nehiyawewin sentences are structurally ambiguous. Using the same surface string, one structure triggers WCO but the other structure does not.  25  There are a number of variations (i.e., involving different combinations of complementizers) for which I have only partial paradigms and which require further investigation. The evidence presently shows that these examples, too, provide both good and bad interpretations (and structures). 25  225  Chapter 7 CONCLUSIONS  7.0 Conclusions I have investigated and provided an analysis for the structure of Nehiyawewin whquestions. I have proposed that wh-words are not generated in argument position of a verb. Rather, the wh-phrase occurs as the predicate NP in a separate nominal clause structure. The subject of the nominal clause is linked to an A-position in a subordinate clause in one of two ways, either via clefting of the wh-phrase or by conjunction. When the subordinate clause has kd- complementizer, the resulting structure is a relative clause which restricts the reference of the subject, i.e., Who is that/the [one that Mary likes?]. When the subordinate clause has e- complementizer, the clauses are interpreted as conjoined, i.e., Who is he & Mary likes him^ t  Wh-questions require movement, and I have argued that movement of a null operator occurs in all Nehiyawewin wh-questions. The choice of complementizer determines not only the structure of the wh-question, but also the level of the syntax at which operator movement takes place. With complementizer kd-, there is movement at Sstructure. When the complementizer is e-, there is LF operator movement in the subordinate clause — which forces the anaphoric relation beween the wh-word and an argument in the conjoined clause.  226  This investigation of wh-questions has also provided an account of nominal clause structures in Nehiyawewin. In my analysis, obligatory predicate fronting accounts for the proximate/obviative agreement between the subject and predicate DPs — as well as for the obligatory initial position of the wh-word. We have seen, further, that there is a contrast between complement clauses and adjuncts with respect to long-distance null-operator movement. Complement clauses allow long-distance extraction while adjunct clauses do not allow extraction, in accordance with Huang's (1982) Constraint on Extraction Domains. Based on these asymmetries, I propose that complement clauses are in argument positions in Nehiyawewin. I have argued, further, that the ungrammaticality of extraction of wh-clauses in complement position provides strong support for the analysis of wh-questions defended here; i.e., that both extraction from adjunct clauses and extraction from within wh-islands can be uniformly ruled out as CED effects. With respect to null-operator movement, I also argue that WCO may be avoided because there is no movement of a truly quantificational operator. Nehiyawewin whquestions are immune to WCO in the same way as null-operator structures in English (i.e., parasitic gaps) are immune to WCO (see Lasnik and Stowell 1991). My analysis is based, in particular, on Demirdache's (1997) analysis of Weakest Crossover. In addition, in Chapter 2, I have used an analysis based on hierarchies to provide an account of the direct/inverse system characteristic of Algonquian languages. The treatment of the data via a system of hierarchies and the alignment of those hierarchies  227  provides insights into the issue of inverse vs. passive which have not been available previously. During the course of this work, I have suggested areas which require further research. One area which invites investigation involves the question of (i) why NPs are prohibited from appearing in A-position; and (ii) whether they are base generated in A' position or scrambled out of an A-position into an A'-position. In other words, is Nehiyawewin a Pronominal Argument Language (PAL)? I expect that this PAL issue will revolve around the fact that Nehiyawewin has no regular determiner system — other than the deictic set discussed in Chapter 3. As noted in Chapter 4, overt DPs occur in subject position of the nominal clauses (i.e., non-verbal clauses which have no verbal-agreement morphology to identify pronominal arguments); and we have observed that speakers almost invariably supply a deictic determiner for these arguments. What are the semantics and syntax of the NP/DP distinction? Many of the questions arising out of this investigation of wh-constructions hinge on these issues. We have seen that there are two forms for the obviative wh-word awini-wa vs. awini-hi  'who'. The standard (more conservative) form of the suffix is -hi which  categorizes wh-words along with other sets in the pronominal paradigms. The -wa suffix typically occurs on nouns. What is the significance of this shift and how does it affect wh-strategies, if at all? I have observed a number of other sub-dialectal differences which are set down in (1), i.e., the dialects with -wa tend to use e- complementizer more than kd— as well as using the independent mode of the verb more freely. Are these issues  228  related? Also, What is the significance of the variation across dialects and across speakers as shown in the table in (1)? (1)  Wh-question Patterns: Eastern Cree  Nehiyawewin Dialect I: awini-hi  * kaA  e-  Dialect II: awini-wa  ka-  ka-  le-  Te-  i Ind  f Ind  I have suggested in Chapter 3 that the absence of kd- (in all but past tense) in Moose Cree wh-questions, for example, is related to the issue of Initial Change (or IC, a morphophonological ablauting process). Moose Cree, like the neighbouring Ojibwa language, appears to have retained the more overt expression of the IC process. The IC process is used as a focussing device and is, in fact, represented in the e- conjunct/ complementizer. By contrast, the e- complementizer is absent from ordinary complement clauses in Moose Cree. I propose that Nehiyawewin has grammaticalized the process in the form of kd- and e- complementizers which is not tense-related (see discussion in Appendix D). However, this proposal requires further research. There is also the issue of argument-type wh-words, i.e., those corresponding to who and what — as opposed to adverbial wh-words like why and when. It is possible that,  229  since the adverbial wh-words are not associated with an argument position, they may be generated inside the verbal clause and move to Spec CP in Nehiyawewin as they do in English. This would contrast with wh-words associated with argument positions, which are generated in a separate nominal clause in my analysis. This contrast exists with Egyptian Arabic wh-expressions (Demirdache p.c). However, I have not investigated the issue and this needs to be done. These are all issues which have been brought to light in the course of my investigation of Nehiyawewin wh-constructions. This analysis of Nehiyawewin wh-constructions is presented within the framework of generative grammar. As such, it provides a significant contribution to this growing body of Algonquian literature in this framework — as well as to the theoretical literature on wh-constructions and the structure of Native American languages in general.  230  REFERENCES Ahenakew, Freda, ed. (1987a) wdskahikaniwiyiniw-dcimowina. Stories of the House People. Publications of the Algonquian Text Society, Winnipeg: The University of Manitoba Press. Ahenakew, Freda. (1987b) Cree Language Structures: A Cree Approach. Winnipeg:  Pemmican Publications Inc. Ahenakew, Freda and H.C. Wolfart, eds. (1991) kohkominawak otdcimowiniwdwa. Our Grandmothers' Lives as Told in their Own Words. Saskatoon: Fifth House  Publishers. Bains, Gurupreet. (1989) Complex Structures in Hindi-Urdu: Explorations in Government  and Binding. Ph.D. dissertation. New York University. Baker, C. Lee. (1970) Notes on the Description of English Questions. Foundations of Language 6:197-219. Baker, Mark C. (1985) The Mirror Principle and Morphosyntactic Explanation. Linguistic Inquiry 16:373-416. Baker, Mark C. (1988) Incorporation: A Theory of Grammatical Function Change.  Chicago: University of Chicago Press. Baker, Mark C. (1991) On some Subject/object Non-asymmetries in Mohawk. Natural Language and Linguistic Theory 9(4):537-576.  Baker, Mark C. (1996) The Polysynthesis Parameter. ms. McGill University. Bar-el, Leora. (1997) Nonconfigurationality in Plains Cree. Ms. University of British Columbia. Barss, Andrew, Kenneth Hale, Ellavina Perkins and Margaret Speas. (1989) Logical Form and Barriers in Navajo. In C.T.J. Huang and R. May, eds., Logical Form: CrossLinguistic Perspectives. Dordrecht: Kluwer Academic Publishers. Benveniste, Emile. (1971) Problems in general linguistics. Translated by Mary Meek. Coral Gables. Blain, Eleanor M. (1992) Binding Effects in Plains Cree. ms. University of British Columbia.  231  Blain, Eleanor M. (1993) In Search of the Structure of Cree. ms. University of British Columbia. Blain, Eleanor M. (1994) Emphatic Pronouns in Plains Cree. ms. University of British Columbia. Blain, Eleanor M. (1995a) Emphatic wiya in Plains Cree. In D. Pentland, ed., Papers of the Twenty-sixth Algonquian Conference. Winnipeg: University of Manitoba Press. Pp. 22-34 Blain, Eleanor M. (1995b) Are wh-words Clefted in Plains Cree? Paper presented to the Canadian Linguistics Association meeting, Montreal, June, 1995. Blain, Eleanor M. (1996a) Cleft Focus and Predicate Focus in Cree, paper presented at the S.S.I.L.A., San Diego, CA. January, 1996. Blain, Eleanor M. (1996b) Restrictive complementizer kd- in Plains Cree. Paper presented to the Workshop on Structure and Constituency in Native American Languages, University of Manitoba, Winnipeg. Blain, Eleanor M. (1997) The Covert Syntax of Wh-Questions in Plains Cree. BLS 22, University of California, Berkeley. Bloomfield, Leonard. (1928) The Plains Cree Language. International Congress of Americanists, Proceedings 22:2:427-431.  Bloomfield, Leonard. (1934) Plains Cree Texts. Publications of the American Ethnological Society, 16, New York. Bloomfield, Leonard. (1946) Algonquian. Harry Hoijer et al., eds. Linguistic structures of native America. Viking Fund Publications in Anthropology. Pp.685-129. Bloomfield, Leonard. (1993) Sacred Stories of the Sweet Grass Cree. National Museum of Canada, Bulletin 60, Ottawa (Original Publication 1930). Saskatoon: Fifth House. Brittain, Julie, (1995) Two Negative Morphemes in Sheshatshit Montagnais (Innu-aimun): apu and ekd. Paper presented at the Twenty-seventh Algonquian Conference, N.Carolina, October 1995. Browning, Margeurite. (1992) Nidi Operator Constructions. Ph.D. dissertation, MIT. Burton, Strang, (to appear) Past Tense on Nouns as Death, Destruction, and Loss. In K. Kusomoto ed., Proceedings of NELS 27: 65-77.  232  Calabrese, Andrea. (1984) Multiple questions and focus in Italian. In W. de Geest and Y. Putseys eds., Sentential Complementation. Dordrecht: Foris. Pp. 67-74. Calabrese, Andrea. (1987) Focus Structure in Berber: A comparative Analysis with Italian. In M. Guerssel and K. Hale, eds., Studies in Berber Syntax, Lexicon Project.  Working Papers, Center for Cognitive Sciences, Cambridge: MIT Press. Pp. 103120. Campana, Mark, (1996) The Conjunct Order in Algonquian. Canadian Journal of Linguistics 41 (3): 201 -234.  Cheng, Lisa Lai Shen. (1991) On the Typology of Wh-Questions. Ph.D. dissertation, M.I.T. Chomsky, Noam. (1973) Conditions on Transformations, in S. R. Anderson and P. Kiparsky eds., A Festschrift for Morris Halle. New York: Holt, Rinehart & Winston. Chomsky, Noam. (1976) Conditions on Rules of Grammar, Linguistic Analysis 2:303-351. Chomsky, Noam. (1977) On Wh-Movement. In Culicover et al, eds. Formal Syntax. New York: Academic Press. Pp. 71 -131. Chomsky, Noam. (1980) On Binding, Linguistic Inquiry, 11:1 -46. Chomsky, Noam. (1981) Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. (1982) Some Concepts and Consequences of the Theory of Government  and Binding. Linguistic Inquiry Monograph Six, Cambridge: MIT Press. Chomsky, Noam. (1986) Barriers. Cambridge: MIT Press. Chomsky, Noam. (1992) A Minimalist Program for Linguistic Theory. MIT Occasional Papers in Linguistics 1, MITWPL, Cambridge, Massachusetts. Chomsky, Noam. (1995) The Minimalist Program. Cambridge: MIT Press. Cinque, Guglielmo. (1990) Types of A'-Dependencies. Linguistic Inquiry Monograph Seventeen. Cambridge: MIT Press. Clarke, Sandra, Marguerite MacKenzie and Deborah James, (1993) Preverb Usage in Cree/Montagnais/Naskapi. In William Cowan, ed., Papers of the Twenty-fourth Algonquian Conference. Ottawa: Carleton Press. Pp.32-45.  233  Cook-Neff, Stella, (undated) A Cree Dictionary; Itewina masinayikan. Manitoba Department of Education. Culicover, Peter, Thomas Waslow and Adrian Akmajian, eds. (1977) Formal Syntax. New York: Academic Press. Dahlstrom, Amy. (1986) Plains Cree Morphosyntax. Ph.D. dissertation, University of California, Berkeley. Dahlstrom, Amy. (1988) Independent Pronouns in Fox. In William Shipley ed., In Honor of Mary Haas from the Haas Festival Conference on Native American Linguistics.  Berlin: Mouton de Gruyter. Dahlstrom, Any. (1995) Topic, Focus and other Word Order Problems in Algonquian. The  Belcourt Lecture, University of Manitoba. Voices of Rupert's Land, Winnipeg. Dechaine, Rose-Marie. (1993,) Predicates Across Categories: Towards a Category-Neutral  Syntax. Ph.D. dissertation. University of Massachusetts, Amherst. Dechaine, Rose-Marie. (1996) What Algonquian Morphology is really like: Hockett Revisited. Paper presented at the Workshop on Structure and Constituency in Native American Languages, University of Manitoba, Winnipeg. Dechaine, Rose-Marie. (1997) Modifier Flexibility in Plains Cree. Paper presented at the Workshop on structure and constituency in Native American Languages, University of Manitoba, March, 1997. Dechaine, Rose-Marie, (to appear) Nominal Predication in Plains Cree. In David Pentland, ed., Procedings of the Twenty eighth Algonquian Conference. Winnipeg: University  of Manitoba Press. DeLancey, Scott. (1981) An Interpretation of Split Ergativity and Related Patterns. Language 57:3:626-9. Demirdache, Hamida. (1997) Dislocation, Resumption and Weakest Crossover. Materials on Left Dislocation. Anagnostopoulou, Elena, Henk van Riemsdijk, and Frans Zwarts, eds. Linguistics Today 14:193-231. Amsterdam: John Benjamins. Di Sciullo, Anna Maria and Edwin Williams. (1987) On the Definition of Word. Linguistic Inquiry Monographs 14. Cambridge: M.I.T. Press. Doron, Edit. (1982) On the Syntax and Semantics of Resumptive Pronouns. In R. BleyVroman, ed., Texas Linguistics Forum 19, Austin, Texas. Pp. 1-48.  234  Dryer, Matthew S. (1992) A Comparison of the Obviation Systems of Kutenai and Algonquian. ms. SUNY at Buffalo. Dwivedi, VeenaDhar. (1994) Syntactic Dependencies and Relative Phrases in Hindi.  Ph.D. dissertation, University of Massachusetts Amherst. Ellis, C. Douglas. (1971) Cree Verb Paradigms. International Journal of American Linguistics 37:76-95. Ellis, C. Douglas. (1983) Spoken Cree. Revised edition, Edmonton: Pica Pica Press. Erteschik-Shir, Nomi. (1993) The Dynamics of Focus Structure, ms. Ben Gurion University of the Negev, Israel. Faries, R. ed. (1938) A Dictionary of the Cree Language. Toronto: Church of England of Canada. Fassi Fehri, A. (1982) Linguistique arabe: forme et interpretation. Rabat: Publications de la Faculte des Lettres et Sciences Humaines. Frantz, Donald G. (1966) Person Indexing in Blackfoot. International Journal of American Linguistics 32:50-58. Frantz, Donald G., (1976) Equi-subject clause union. BLS 2:179-187. Frantz, Donald G., (1978) Copying from complements in Blackfoot. Eung-Do Cook and Jonathan D. Kaye, eds., Linguistic studies of native Canada. Pp. 89-109. Frantz, Donald G., (1979) Grammatical relations in universal grammar. SIL, Universiry of North Dakota Workpapers 23, Supplement. Huntington Beach, CA. Frantz, Donald G. (1991) Blackfoot Grammar. Toronto: University of Toronto Press. Goddard, Ives. (1990) Primary and Secondary Stem Derivation in Alonquian. International Journal of American Linguistics, Vol. 56:4:449-483. Grafstein, Ann. (1984) Argument Structure and the Syntax of a Non-Configurational  Language. Ph.D. dissertation, McGill University. Grafstein, Ann. (1989) Disjoint Reference in a "Free Word Order" Language. In Donna B. Gerdts, and Karin Michelson, edd., Theoretical Perspectives on Native American  Languages. Buffalo: SUNY Press. Pp.63-175.  235  Haegeman, Liliane. (1991) Introduction to Government and Binding Theory. Cambridge: Basil Blackwell Inc. Hale, K e n . (1983) Warlpiri and the Grammar o f Non-Configurational Languages, Natural Language and Linguistic Theory, 1:5-47. Halle, Morris and Alec Marantz. (1993) Distributed morphology and the pieces o f inflection. In K e n Hale, and S. J. Keyser eds., The View From Building 20. Cambridge: M I T Press. Pp. 111-176. Heggie, Lorie A . (1988) The Syntax of Copular Structures. Ph.D. dissertation, University o f Southern California. Higginbotham, James. (1983) Logical form, binding, and nominals. Linguistic Inquiry 14:395-420. Higginbotham, James. (1985) On semantics. Linguistic Inquiry 16:547-93. Higginbotham, James. (1987) Indefmiteness and predication. In Eric Reuland and Alice ter Meulen, eds., The Representation of (In)Defmiteness. Cambridge: M I T Press. Pp.43-70. Higgins, F . R . (1973) The Pseudocleft Construction in English. Ph.D. dissertation, M . I . T . Hirose, Tomio. (1997) On the Inchoative Morpheme -payi in Plains Cree. M s . University o f British Columbia. Hockett, Charles F. (1966) What Algonquian is really like. International Journal of American!Linguistics 32:59-73. Huang, James C - T . (1982) Logical Relations in Chinese and the Theory of Grammar. Ph.D. dissertation, M . I . T . Jacobs, J., et al, edd. (1993) Syntax: An International Handbook of Contemporary Research. Berlin: Walter de Gruyter. James, Deborah. (1991) Preverbs and the Function of Clauses in Moose Cree. The Belcourt Lecture, University o f Manitoba. Voices o f Rupert's Land, Winnipeg. Jelinek, Eloise. (1984) Empty Categories, Case, and Configurationality, Natural Language and Linguistic Theory, 2:39-76.  236  Johns, Alana. (1981) Relative Clauses and Related Phenomena in Ojihway. M.A. Thesis,  University of Ottawa. Johns, Alana. (1982) A Unified Analysis of Relative Clauses and Questions in Rainy River Ojibwa. In William Cowan, ed. Papers of the thirteenth Algonquian Conference.  Ottawa: Carleton University Press. Jolley, Catherine A. (1982) On the Plains Cree Passive: an Analysis of Syntactic and Lexical Rules. In B. D. Joseph, ed. Grammatical Relations and Relational Grammar  ??  Kayne, Richard. (1984) Connectedness and binary branching. Dordrecht: Foris. Kayne, Richard. (199'4) The Antisymmetry of Syntax. Linguistic Inquiry: Monograph Twenty-five. Cambridge: MIT Press. Kinkade, M. Dale. (1989) When Patients are Topics: Topic Maintenance in North American Iandian Languages. Papers of the 24th International Conference on Salish  and Neighboring Languages, University of British Columbia. Klaiman, M. H. (1992) Inverse Languages. Lingua 88:227-61. Koopman, H. and D. Sportiche. (1982) Variables and the Bijection Principle, The Linguistic Review, 2:139-60. Kuno, Susumo. (1976) Subject, Theme and the Speaker's Empathy — a Reexamination of Relativization Phenomena. Linguistic Inquiry, 417-44. Lasnik, Howard and Tim Stowell. (1991) Weakest Crossover. Linguistic Inquiry 22:687720. Longobardi, Giuseppe. (1994) Reference and proper names: a theory of N-movement in syntax and logical form. Linguistic Inquiry 25:609-65. Lyons, John. (1977) Semantics. Vols. 1 and 2. Cambridge. Matthewson, Lisa. (1996) Determiner Systems and Quantificational Strategies: Evidence  from Salish. Ph.D. dissertation. University of British Columbia. Matthewson, Lisa and Charlotte Reinholtz. (1996) The Syntax and Semantics of Determiners: A Comparison of Salish and Cree. Papers for the 31 st International Conference on Salish and Neighbouring Languages. UBC.  237  McGinnis, Martha, (1996) Is There Syntactic Inversion? Evidence from Ojibwa, paper presented at the Conference and Workshop on Structure and Constituency in Native American Languages, University of Manitoba, Winnipeg, March, 1996. M o r o , Andrea. (1990) The Raising of Predicates: Copula, Expletives and Existence. M s . Universita Venizia and M . I . T . Nichols, Joanna. (1986) Head-marking and dependent-marking grammar. Language 72:56119. Nichols, John D . (1980) Ojibwe Morphology. Ph.D. dissertation, Harvard University. Nishigauchi, Taisuke. (1986) Quantification in Syntax. Ph.D. dissertation, University o f Massachusetts, Amherst. Ouhalla, Jamal. (1994) Introducing Transformational Grammar: from Rules to Principles and Parameters. London: Edward Arnold. Pentland, David H . and H . Christoph Wolfart, eds. (1982) Bibliography o f Algonquian Linguistics. Winnipeg: University o f Manitoba Press. Pesetsky, David. (1987) Wh-in-Situ: Movement and Unselective Binding. In Eric J. Reuland and Alice G . B . ter Meulen eds. The Representation of (In)definiteness. Cambridge: M I T Press. Pp.98-129. Rapoport, Tova Rebecca. (1987) Copular, Nominal, and Small Clauses: A Study of Israeli Hebrew. Ph.D. dissertation, M . I . T . Reinhart, Tanya. 1983 Anaphora and semantic interpretation. Chicago: University o f Chicago Press. Reinholtz, Charlotte and Kevin Russell. (1995) Quantified N P s in Pronominal Argument Languages: Evidence from Swampy Cree. Proceedings of the 25th North Eastern Linguistic Society, Philadelphia: University of Pennsylvania. Reuland, Eric J. and Alice G . B . ter Meulen, eds. (1987) The Representation of (Injdefiniteness. Cambridge: M I T Press. Rigau, Gemma. (1982) O n the Avoid Pronoun Principle and the Emphatic Operator. M s , Universitat Autonoma de Barcelona/MIT.  238  Rigau, Gemma. (1989) Conexity Established by Emphatic Pronouns. Text and Discourse Connectedness, in Maria-Elisabeth Conte, Janos Petofi and Emel Sozer, eds. Amsterdam: Benjamins. Pp. 191-205. R i z z i , L . (1986) N u l l Objects in Italian and the Theory of pro. Linguistic Inquiry. 17:3:501-557 Rochemont, Michael S. (1986) Focus in Generative Grammar. Philadelphia: John Benjamins. Rogers, Jean. (1978) Differential Focusing in Ojibwa Conjunct Verbs: on Circumstances, Participants, or Events. International Journal of American Linguistics, 44:3:167179. Ross, J. (1967) Constraints on Variables in Syntax. Ph.D. dissertation,  M.I.T.  Rouveret, A . and J.-R. Vergnaud. (1980) Specifying Reference to the Subject: French Causatives and Conditions on Representations. Linguistic Inquiry 11:97-202. Rudin, Catherine. (1988) O n multiple questions and multiple W H fronting. Natural Language and Linguistic Theory 6:445-502. Russell, Kevin and Charlotte Reinholtz. (1995) Hierarchical Structure in a N o n configurational Language: Asymmetries in Swampy Cree. Paper presented at W C C F L X I V , L o s Angeles, C A . March, 1995. Russell, Kevin and Charlotte Reinholtz. (1996) Non-configurationality and the SyntaxPhonology Interface, paper presented at W C C F L X V , Irvine, C A . March, 1996. Ruwet, Nicolas. (1982) Grammaire des Insultes, et autres etudes. Paris: Editions du Seuil. Siewierska, Anna. (1993) On the Interplay of Factors in the Determination o f Word Order. In J. Jacobs, A . von Stechow, W . Sternefeld and T. Venneman, edd., Syntax: An International Handbook of Contemporary Research. Berlin: Walter de Gruyter. Pp.826-47. Silverstein, Michael. (1976) Hierarchy o f Features and Ergativity, in R . M . W . Dixon, ed., Grammatical Categories in Australian Languages, Canberra: Australian Institute o f Aboriginal Studies. Pp.113-171. Starks, Donna. (1993) The Effect o f Information Status on Word Order: Evidence from Woods Cree. Linguislica atlantica 15:165-173.  239  Stowell, Tim. (1989) Subjects, specifiers and X-bar theory. Mark Baltin, and Anthody Kroch, eds. Alternative conceptions of phrase structure. Chicago: The University o f Chicago Press. Stowell, Tim. (1993) Tense, ms. University o f California, L o s Angeles. Thompson, Chad L . (1989) Voice and Obviation in Athabaskan and Other Languages. Ph.D. Dissertation, University o f Oregon. Tomlin, Russell S., and Richard Rhodes. (1992). Information Distribution in Ojibwa. In Doris L . Payne, ed., Pragmatics of Word Order Flexibility (Typological Studies in Language 22.) Philadelphia: John Benjamins. Truitner, Kenneth and Timothy Dunnigan. (1972) Wh-Questions in Ojibwe. Chicago Linguistics Society. Chicago. Pp.359-367. van Riemsdijk, H . and E . Williams. Cambridge: M I T Press.  (1986) Introduction to the Theory of Grammar.  Wahba, Abdel-Faheem Batran. (1984) Wh-Constructions in Egyptian Arabic. P h . D . dissertation, University o f Illinois at Urbana-Champaign. Watanabe, Akira. (1992) Subjacency and S-Structure Movement o f Wh-in-situ. Journal of East Asian Linguistics 1:255-291. Williams, Edwin. (1988) The A T B Theory o f Parasitic Gaps. ms. Williams, Edwin. (1994) Thematic Structure in Syntax. Linguistic Inquiry Monograph Twenty-three. Cambridge: M I T Press. Williamson, Janis S. (1987) A n Indefiniteness Restriction for Relative Clauses in Lakhota. In Eric J. Reuland and Alice G . B . ter Meulen eds., The Representation of (In)definiteness. Cambridge: M I T Press. Pp. 168-190. Wolfart, H . Christoph. (1973) Plains Cree: A Grammatical Study. Transactions ofthe American Philosophical Society, Volume 63, Part 5, Philadelphia. Wolfart, H . Christoph. (1991) Passives with and without Agents. In H . C . Wolfart, ed. Linguistic Studies Presented to John L. Finlay. Memoir 8, Algonquian and Iroquoian Linguistics. Winnipeg. Pp. 171-190.  240  Wolfart, H. Christoph. (1996) Sketch of Cree, an Algonquian Language. In Ives Goddard, ed., Languages, William C. Sturtevant, ed., Handbook of North American Indians,  Vol 17. Washington, D.C., Smithsonian Institution. Wolfart, H.Christoph and Freda Ahenakew, eds. (1993) kinehiydwiwininaw nehiyawewin. The Cree Language is our Identity. The LaRonge Lectures of Sarah Whitecalf. Publications of the Algonquian Text Society. Winnipeg: University of Manitoba Press. Wolfart, H. Christoph and Janet F. Carroll. (1981) Meet Cree. Edmonton: University of Alberta Press. Zubin, David. (1979) Discourse Function of Morphology: the Focus System in German. In Talmy Givon ed. Syntax and Semantics 12. Discourse and Syntax. New York:  469-504.  241  APPENDIX A INFLECTIONAL PARADIGMS VII Intransitive Inanimate (subj) Stem: mihkwa- 'be red  1  Independent inan.sg. inan.pl inan.obv.sg inan.obv.pl  mihkw-a-w mihkw-a-w-a mihkw-a-yi-w mihkw-a-yi-w-a  Conjunct e-mihkw-a-k e-mihkw-a-k-i e-mihkw-a-yi-k e-mihkw-a-yi-k-i  VAI Intransitive Animate (subj) Stem: pimipahta - 'run'  1 2 l.pl.excl. 21.pl.incl. 2.pl 3 3pl 3" (obv)  Independent  Conjunct  ni-pimipaht-a-n ki-pimipaht-a-n ni-pimipaht-a-nan ki-pimipaht-a-(na)naw ki-pimipaht-a-(na)waw  e-pimipaht-a-yan e-pimipaht-a-yan e-pimipaht-a-yahk e-pimipaht-a-yahk e-pimipaht-a-yek  pimipaht-a-w pimipaht-a-w-ak pimipaht-a-yi-w-a  e-pimipaht-a-t e-pimipaht-a-c-ik e-pimipaht-a-yi-k  VTI Transitive Inanimate (obv) Stem: wapaht- ' see (something)'  1 2 l.pl.excl. 21.pl.incl. 2.pl 3 3pl 3' (obv)  Independent  Conjunct  ni-wapaht-e-n ki-wapaht-e-n ni-wapaht-e-nan ki-wapaht-e-(na)naw ki-wapaht-e-(na)waw  e-wapaht-am-an e-wapaht-am-an e-wapaht-am-ahk e-wapaht-am-ahk e-wapaht-am-ek  wapaht-am-(w) wapaht-am-w-ak wapaht-am-iyi-w-a  e-wapaht-ahk e-wapaht-ahk-ik e-wapaht-am-iyi-k  242  Theme Markers: Vb.class I: -6-1/2 subj (Ind) -am- elsewh.  Vb. class II: -aIII: - 0 -  VTA Transitive Animate (obi): DIRECT Paradigm Stem: sekih- 'to frighten (someone)' subj-->obj  Independent  Conjunct  2-1 2pl-l 2(pl)-lpl.ex  ki-sekih-i-n (2/lsg) ki-sekih-i-nawaw ki-sekih-i-nan  e-sekih-i-yan e-sekih-iy-ek e-sekih-iy-ahk  1-3 l-3pl 1-obv  ni-sekih-a-w ni-sekih-a-w-ak ni-sekih-im-a-w-a  e-sekih-ak (portmant) e-sekih-ak-ik e-sekih-im-ak  2-3 2-3pl 2-obv  ki-sekih-a-w ki-sekih-a-w-ak ki-sekih-im-a-w-a  e-sekih-at (portmant) e-sekih-a-c-ik e-sekih-im-at  lpl.ex-3 lpl.ex-3pl lpl.ex-obv  ni-sekih-a-nan ni-sekih-a-nanak ni-sekih-im-a-nan-a  e-sekih-a-yahk e-sekih-a-yahk-ik e-sekih-im-ahk  21pl.in-3 21pl.in-3pl 21pl.in-pbv  ki-sekih-a-naw ki-sekih-a-nawak ki-sekih-im-a-naw-a  e-sekih-a-yahk e-sekih-a-yahk-ok e-sekih-im-a-yahk  2pl-3 2pl-3pl 2pl-obv  ki-sekih-a-waw ki-sekih-a-wawak ki-sekih-im-a-waw-a  e-sekih-a-yek e-sekih-a-yek-ok e-sekih-im-a-yek  3-obv 3pl-obv obv-obv  sekih-e-w sekih-e-w-ak sekih-e-yi-w-a  e-sekih-a-t e-sekih-a-c-ik e-sekih-a-yi-t  (further obviative) 3-obv' 3pl-obv'  sekih-im-e-w sekih-im-e-w-ak.  e-sekih-im-a-t e-sekih-im-a-c-ik  243  VTA Transitive Animate (obi): INVERSE Paradigm (Stem: sekih- 'frighten s.o.') Subj—> obv 1-2 l-2pl lpl.ex-2(pl)  Independent ki-sekih-iti-n ki-sekih-iti-nawaw ki-sekih-iti-nan  Conjunct e-sekih-it-an e-sekih-it-ak-ok (port) e-sekih-it-ahk  3-1 3pl-l obv-1 inan-1  ni-sekih-ik ni-sekih-ik-w-ak ni-sekih-iko-yi-w-a ni-sekih-iko-n  e-sekih-it (port) e-sekih-ic-ik e-sekih-iyi-t (port) e-sekih-iko-yan  3-2 3pl-2 obv-2 inan-2  ki-sekih-ik ki-sekih-ik-w-ak ki-sekih-iko-yi-w-a ki-sekih-iko-n  e-sekih-isk (port) e-sekih-isk-ik e-sekih-iy-isk (port) e-sekih-iko-yan  3-lpl.ex 3pl-lpl.ex obv-lpl.ex inan-1 pi. ex  ni-sekih-iko-nan ni-sekih-iko-nan-ak ni-sekih-iko-nan-a ni-sekih-iko-nanaw  e-sekih-iko-yahk e-sekih-iko-yahk-ik e-sekih-iko-wa-yahk e-sekih-iko-yahk  3-21pl.in 3pl-21pl.in obv-21 pi. in inan-21 pi. in  ki-sekih-iko-naw ki-sekih-iko-naw-ak ki-sekih-iko-naw-a ki-sekih-iko-nanaw  e-sekih-iko-yahk e-sekih-iko-yahk-ok e-sekih-iko-wa-yahk e-sekih-iko-yahk  3-2pl 3pl-2pl obv-2pl inan-2pl  ki-sekih-iko-waw ki-sekih-iko-waw-ak ki-sekih-iko-waw-a ki-sekih-iko-nawaw  e-sekih-iko-yek e-sekih-iko-yek-ok e-sekih-iko-wa-yek e-sekih-iko-yek  obv-3 obv-3pl obv-obv  sekih-ik(o-w) sekih-ik-w-ak sekih-iko-yi-w-a  e-sekih-iko-t e-sekih-iko-c-ik e-sekih-iko-yi-t  inan-3 inan-3pl  sekih-iko-w /(ow) sekih-iko-w-ak  e-sekih-iko-t e-sekih-iko-t-ik  244  Paradigm: Passive  Stem: sekih 'frighten'  1 2 lpl.ex lpl.in 2pl 3 3pl obv  Independent Conjunct ni-sekih-ikawi-n e-sekih-ikawi-yan ki-sekih-ikawi-n e-sekih-ikawi-yan ni-sekih-ikawi-nan e-sekih-ikawi-yahk ki-sekih-ikawi-nanaw e-sekih-ikawi-yahk ki-sekih-ikawi-nawawe-sekih-ikawi-yek sekih-a-w e-sekih-iht sekih-a-w-ak e-sekih-ihc-ik sekih-im-a-w-a e-sekih-im-iht  Indefinite Subject (AI,TI Stem: pimipahta 'run') pimipahta-niniw  e-pimipahta-hk  245  APPENDIX B NUMBER AND GENDER  1.0 Proximate/obviative & Gender in Nehiyawewin Overt NPs are classified according to animate vs inanimate gender. Animate nouns include all humans and animals, and also some items which are typically inanimate in English; for example, kon- 'snow', otapanaskw- 'vehicle', ospwakan- 'pipe', kdwiy'porcupine quill, and soniydw- 'money'. (See glossary in Ahenakew (1987.a) for more examples.) Dechaine (1996) (based on Hockett 1966) proposes the following schema for the Numbergender system of Nehiyawewin (reproduced here from above): GENDER  (1)  Inanimate  Animate Local Addressee [+2]  Non- ocal Speaker [+1] [3]  [-2,-1] [3']  [3"]  [0] [0']  obviation contrasts The proximate/obviative contrast occurs only in conjunction with third-person — animate or inanimate. Within that category, there is a further breakdown with respect to animate vs inanimate lexical NPs and personal pronouns; and between nominal inflection and verbal inflection. Consider the following:  246  Nominal Agreement Verbal Agreement  (2) (i)  Animate NPs  prox/obv  prox/obv  (ii)  Inanimate NPs  n/a  prox/obv  n/a  prox/obv  (iii)  Personal pronouns  The proximate/obviative contrast is marked in the agreement morphology on the verb in all three cases. This marking is obligatory with TA (transitive animate) verbs in particular, but typically occurs on other verbs also. With respect to nominal agreement, we will look at paradigms in the three categories in (2). (i)  Animate NPs  The proximate/obviative contrast is illustrated in the NPs in (3); simple NPs and names are shown with obviative being marked by suffix -(w)a. The third column shows the plural forms of the common nouns; you will note that plural does not distinguish between proximate and obviative: (3)  3 (prox) 3'(obv) Mary Mary-wa 'man' napew napew-a 'duck' sisip sisip-a 'pipe' ospwakan ospwakan-a John John-a  3pl napew-ak sisip-ak ospwakan-ak  A common noun NP may take either (but not both) of the suffixes. Possessed NPs are illustrated in (4). In possessor NPs, the status of the animate possessor is always indicated while the obviative status of the possessee is marked only if the possessee is animate gender.  247  (4) Possessive NPs (Animate): Prox/obv (Wolfart 1973:31)  3 (prox) ni tern my horse ki •tern your horse  3'(obv)  ni-tem-a my horse(s) ki-tem-a your horse(s) o-tem-a his (3) horse(s) o-tem-iyiw-a his (3') horse(s) tem-inan our horse ni-tem-inan-a our horse(s) m ki tem-inaw our(2.1) horse ki-tem-inaw-a our(2.1) horse(s) ki tem-iwaw your (pi) horse ki-tem-iwaw your(pl) horse(s) o-tem-iwaw-a their horse(s)  The third-person possessor is proximate while the possessee is obviative; and again, the obviative examples are not marked for plural as shown in (5). (5) Possessive NPs (Animate): Sing/plur (Wolfart 1973:31)  3sg ni-tem ki-tem o-tem-a o-tem-iyiw-a ni-tem-inan ki-tem-inaw ki-tem-iwaw o-tem-iwaw-a  my horse your horse his (3) horse(s) his (3') horse(s) our horse our (2.1) horse your (pi) horse their horse(s)  3pl ni-tem-ak ki-tem-ak  my horses your horses  ni-tem-inan-ak our horses ki-tem-inaw our (2.1) horses ki-tem-iwaw your (pi) horses  Either suffix may occur on the stem; but when the possessor is third-person, only the obviative suffix may be used regardless of whether there is one horse or more than one.  (ii)  Inanimate NPs  Inanimate nouns, on the other hand are marked only for singular and plural. Contrast the examples for my shoe/shoes and his shoe/shoes, for example:  248  (6) Possessive NPs (Inanimate): Sing/plur (Wolfart 1973:31) Osg ni-maskisin ki-maskisin o-maskisin o-maskisin-iyiw ni-maskisin-inan ki-maskisin-inaw ki-maskisin-iwaw o-maskisin-iwaw  Opl 'my shoes' ni-maskisin-a 'my shoe' 'your shoes' 'your shoe' ki-maskisin-a 'his.3 shoes' 'his.3 shoe' o-maskisin-a 'his.3' shoes' 'his.3' shoe' o-maskisin-iyiw-a 'our shoes' ni-maskisin-inan-a 'our shoe' 'our.2/1 shoe' ki-maskisin-inaw-a 'our.2/1 shoes' 'your shoe' ki-maskisin-iwaw 'their shoe' o-maskisin-iwaw-a  'your shoes' 'their shoes'  Note that the plural marker for inanimate NPs is homophonous with the obviative marker for animate NPs as seen in (4) above. The following example (7) illustrates the absence of obviative marking on inanimate NPs. The -iyiw suffix on the noun represents the obviative status of the possessor (see paradigm in (6)). 1  (7)  Mary e-otin-am-k o-wiyas-iyiw Mary conj-take-th-3 3-meat-(3')poss Mary took his meat. B.862  On the other hand, as noted by Wolfart (1973:29), the obviative status of the NP is retained on the verb in (8). (8)  e-wanit-a-t ka-mihkwe-yi-k o-pakiwayan conj-lose-th-3 rel-be red-obv-3 3-shirt He lost his red shirt. J.622  It should be noted that some Algonquian languages do show the proximate/obviative contrast on inanimate NPs, i.e., in some dialects of Ojibwa, the obviative suffix -ini occurs on inanimate nouns (Grafstein 1984). 1  249  (iii) Personal Pronouns: Sing/plur As indicated above, inanimate NPs and all overt pronominal NPs contrast only for singular and plural. The complete list of personal pronouns is as follows: (9)  Personal Pronouns  1st.  Singular niya  2nd. 3rd.  kiya wiya  Plural niyanan (lpl.exclusive) kiyanaw (2.1 inclusive) kiyawaw wiyawaw  Except for the 2.1 inclusive plural, all three persons show only singular/plural contrast. First- and second-person (and the plural) forms typically occur in emphatic contexts and also in conjunction with another NP in a compound NP. However, this does not appear to be the case with singular third-person wiya. The absence of a 3-obviative form, for one thing, means that the distinction is lost between two different third-persons in a sentence or discourse. 2  Cf. Frantz (1991) with regard to personal pronouns in Blackfoot. In that Algonquian language, first- and second-person pronouns show proximate/obviative contrast while the third-person pronoun has only the obviative form (see Dechaine (1996)). 2  250  APPENDIX C CONJUNCT VS. INDEPENDENT Verbs occurs in either the Conjunct Mode or the Independent Mode — the two major paradigms of verbal inflection discussed in this work. The conjunct paradigm appears to be the form most commonly used — both in elicited sentences and in a narrative or story-telling context. The conjunct mode may occur in both main clauses and in subordinate clauses. There are two conjunct markers (complementizers): e- and ka-. We have seen that ka- is syntactically motivated while e- is the "unmarked" conjunct. The Independent paradigm may occur only in the main clause of a sentence. However, the use of Conjunct vs. Independent forms in a matrix clause is not clearly defined along syntactic lines. 3  There appear to be some semantic and/or discourse related issues involved. In the context of elicited sentences, the use of independent forms in matrix clauses is optional. Elicited sentences may be interpreted as "out of the blue" sentences, i.e., they have no discourse context. As such, some speakers use the independent form more liberally in this context than do other speakers. When independent forms do occur, it is often in conjunction with sentences involving first- or second-person arguments. 4  Historically, the Independent paradigm is a more recent development ~ and originally was used to represent subordinate clauses (Richard Rhodes, p.a). Note that the inflectional morphology in the Independent Mode mimics that found in possessor phrases. Further evidence exists with negation particles: the same form used to negate NPs, i.e., namoya 'not', is typically used with the Independent Mode verbs while Conjunct Mode verbs have a separate form, kka 'did not'. 3  In a textual/narrative context, the Independent mode is typically used with background sentences. In other words, with sentences which do not advance the story line but which rather provide information added by the story teller to explain or clarify some issue or perhaps to express a personal opinion on the content of the story. The foregrounded story is typically (though not exclusively) rendered in the conjunct mode.  4  251  The following example illustrates these possibilities:  Main Clause Independent  S  Conjunct e-  ^  Subordinate Clause  S  The main clause/subordinate clause distinction is clearly not the only component involved with the two paradigms.  252  APPENDIX D  COMPLEMENTIZERS  kd-: Complementizer or Past Tense D. James (p.c.) points out that, in Moose Cree wh-questions, kd- means 'past tense and is not used in present and future tense questions. Blain (1996) discusses the issue of kd- in wh-questions contrasting a dialect of Ojibwa (Johns 1982) with Moose Cree (James 1991) and Nehiyawewin (Plains Cree). The evidence for wh-questions in shown in (1): 1  (1) Wh-Questions Ojibwa  PRESENT:  PAST:  FUTURE  Initial Change (IC)  Moose Cree  Nehiyawewin (Plains Cree) Independent Conjunct  e- / IC  e- = COMP ka- = COMP  ka-  ka7ki7 + IC 'past'  ki + IC 'past'  kehkih + IC 'past'  wa7wi7 + IC 'fuf  keka + IC •fuf  wehwih + IC 'fuf  5  e-(kih-) ka-(kih-) COMP + ('past')  ka-wihCOMP + 'fuf  These Nehiyawewin forms were obtainedfromthe more conservative speaker (Bill) from Northern Alberta — and were double checked. These forms are most unusual and do not occur in the other dialects I have tested. They involve the process of IC in conjunction with the independent form of the verb ~ and IC normally does not occur with independent mode. As a result, these forms are suspect and require further research with other speakers of that northern dialect. The past tense form by itself would be particularly suspect; however, the corresponding form for future tense (as noted by Dave Pentland, p.c.) provides considerable support for my claim that these forms actually represent an instance of Initial Change in the dialect. Initial Change is no longer a robust process in Nehiyawewin; and as a result, it may take more erratic forms in the language as the usage wanes. 5  253  The bolded forms represent past tense in wh-questions in Ojibwa, Moose Cree, and in Independent forms of Nehiyawewin. The variability in the form of the (bolded) "complementizer" in past tense for the three languages in (1) reflects the form of their respective past tense morpheme. In these three cases, I claim that the past tense forms represent the process of Initial Change on the past-tense morpheme in the language in question. Hence the past tense meaning. In other words, the kd- shown for Moose Cree is not the same complementizer" as occurs in the (conjunct mode) Nehiyawewin wh-questions or in the relative clause examples for all three languages in (2). Nehiyawewin wh-questions in the conjunct mode use kd- in every tense — along with the appropriate tense marker. 6  The table in (2) shows the Complementizers used in all three languages for relative clauses and focus constructions. You will note that the form is standardized as kd- in all three languages. (The [7] represents glottal stop.) 7  (2) Relative Clauses and Nehiyawewin (Plains Cree) Wh-environment Ojibwa PRESENT: PAST: FUTURE  ka-  ka-ki7COMP + 'past' ka-wi7 COMP + •fut*  Moose Cree ka-  Nehiyawewin (Plains Cree) Rel Clause Whka-  ka-kl-  COMP + 'past' —  ka-  ka-(kih)-  ka-(kih)-  COMP + (past)  COMP + (past)  ka-wih-  ka-wih-  COMP + 'fut'  COMP + 'fut'  Initial Change is a morphophonological process which ablauts the first vowel in the verb complex. According to Rogers (1978), Initial Change functions to focus an argument or a "condition" on the clause — orfromanother perspective, IC subordinates a clause to a constituent or some condition of its context in the discourse.  6  This complementizer is analyzed as being derived histori-callyfromthe process of Initial Change on *PA ki- The source of reconstructed *ki- varies as 'past tense' ki- in some analyses (cf. Wolfart 1973, for example) and ki- the preverb form of kiwi 'around' (attributed to Goddard in Clarke et al (1993), and Pentland p.c). 7  254  The last column shows that the Nehiyawewin complementizer for wh-questions in Conjunct mode fits into this pattern. For a complete discussion of the data, cf. Blain (1996). Given the evidence above, I argue that kd- in Nehiyawewin does not distinguish tense per se. But rather it is some notion of presupposition or reference to some specific event or person or thing involved in the restrictive clause which carries with it this sense of realized time. The forms given for Nehiyawewin are based on observations from one speaker from the Northern Alberta dialect — though the general claims with respect to kdcomplementizer hold for other speakers as well. More research and investigation is required in the Northern Alberta dialect.  255  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items