THE SEMANTICS OF DETERMINERS: DOMAIN RESTRICTION IN SKWXWU7MESH by Carrie Samantha Gillon B.A., The University of British Columbia, 1998 M.A. , The University of Toronto, 1999 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY i n THE F A C U L T Y OF G R A D U A T E STUDIES (Linguistics) THE UNIVERSITY OF BRITISH COLUMBIA August 2006 © Carrie Samantha Gillon, 2006 Abstract In this thesis, I investigate the properties of determiners in Skwxwu7mesh (Squamish) Salish. Determiners in Skwxwu7rriesh behave significantly differently from the definite determiner the in English. Skwxwu7mesh lacks a definite/indefiniteness distinction; all DPs can be used in both familiar and novel contexts, and are not required to refer to a unique entity. Instead, Skwxwu7mesh determiners are split along deictic/non-deictic lines. I argue that deictic features on the determiners have consequences for the grammar in terms of (i) scope and (ii) implicature of uniqueness. If a DP is deictic, (i) it can take wide scope and (ii) any sentence containing it will carry an implicature of uniqueness. If a DP is non-deictic, (i) it must take narrow scope and (ii) any sentence containing it does not carry an implicature of uniqueness. I claim that non-deictic DPs are composed via Restrict and deictic DPs via Specify (cf. Chung and Ladusaw 2004). There is therefore no correlation between more structure and wide scope, but rather a correlation between features and wide scope. Deictic features allow DPs to take wide scope; the lack of features prevents DPs from taking wide scope. Determiners in Skwxwu7mesh are quite different from determiners in better-known languages. Do determiners share anything in common cross-linguistically? I argue that Skwxwu7mesh determiners and English the are both associated with domain restriction (cf. von Fintel 1994). Both non-deictic and deictic DPs are sensitive to the context in which they are used; in familiar contexts,, they (usually) refer to the set of entities under discussion. Non-deictic DPs, which in terms of scope behave like bare nouns, must differ from bare nouns in this respect. Bare nouns (in languages which use articles) cannot be used in familiar contexts. They can only introduce new discourse referents. Non-deictic DPs can introduce new discourse referents, but can also refer to previously introduced discourse referents, and can also be used partitively. Skwxwu7mesh determiners must be associated with domain restriction, whereas bare nouns cannot be. I propose there is a strict correlation between the syntax and semantics: if a determiner occupies D, it has domain restriction in its representation. ii Table of Contents Abstract ii Table of Contents iii List of Tables vii List of Figures viii Abbreviations ix Acknowledgements xi Dedication xiii Chapter One: Introduction 1 1 Introduction 1 2 The background 4 2.1 The problem in Skwxwu7mesh 4 2.2 The problem in English 8 2.3 Why are these two problems related? 8 3 Preview of the analysis 9 4 Outline of the thesis 11 Chapter Two: The Background on Skwxwu7mesh 13 1 Introduction 13 2 Methodology 15 3 Morphology 16 4 Syntax 18 5 D-determiners 23 5.1 Obligatory D-determiners 23 5.2 D-determiners and demonstratives do not co-occur 26 5.3 D-determiners and quantifiers or numerals do co-occur 27 5.4 Gender 31 6 Previous discussions of the determiner system 32 6.1 Kuipers' (1967) description of the Skwxwu7mesh determiner system 32 6.2 Jacobs' (1997) analysis 36 7 Deixis 38 7.1 Anchor 39 7.1.1 Anchor for the D-determiners 40 7.1.2 Anchor for the demonstratives 41 7.2 Spatial demarcation 42 7.2.1 Distal 43 7.2.1.1 The distal D-determiner 43 7.2.1.2 The distal demonstratives 44 7.2.2 Neutral 45 7.2.2.1 The neutral D-determiner 46 7.2.2.2 The neutral demonstrative 46 iii 7.2.3 The medial demonstrative 46 7.2.4 Proximal 48 7.2.4.1 The proximal D-determiner 48 7.2.4.2 The proximal demonstrative 49 7.3 Region configuration: (in)visibility 49 7.3.1 The invisible D-determiner 49 7.3.2 The invisible demonstratives 50 7.4 Summary 51 Chapter Three: Theoretical Background 53 1 Introduction 53 2 Presuppositions vs. implicatures 54 2.1 Presuppositions 54 2.2 Conversational implicatures 55 3 The DP hypothesis 57 4 Background on definiteness 58 4.1 Definiteness in English: uniqueness or familiarity? 59 4.2 Evidence for uniqueness 59 4.2.1 Assertion versus presupposition of existence 59 4.2.2 Presupposition of uniqueness 61 4.2.3 Assertion of uniqueness : 64 4.3 Evidence for familiarity 65 4.4 Deriving uniqueness from familiarity 67 4.5 Deriving familiarity from uniqueness 68 4.6 Summary • 69 5 Domain restriction 69 5.1 Why domain restriction? 70 5.2 Quantifiers and domain restriction 70 5.3 Bare nouns and (the lack of) domain restriction 71 6 Foreshadowing the analysis 73 6.1 Deriving familiarity 73 6.2 Specify and Restrict 77 6.2.1 Specify 77 6.2.2 Restrict 78 6.3 The Skwxwu7mesh D-determiners 79 7 Conclusion 80 Chapter Four: The deictic D-determiners 81 1 Introduction 81 2 No novelty/familiarity distinction in Skwxwu7mesh 83 2.1 Discourse new vs. discourse old 83 2.2 Hearer new vs. hearer old 85 3 Uniqueness in Skwxwu7mesh 88 3.1 No assertion of uniqueness 88 iv 3.2 Implicature of uniqueness 89 3.3 Implicature of maximality 90 3.4 Summary 94 4 Domain restriction and the deictic D-determiners 94 4.1 Why choice functions? 94 4.2 Domain restriction and implicature of uniqueness 96 4.3 Domain restriction and quantifiers 100 4.4 Domain restriction and demonstratives 101 5 Co-reference effects 105 5.1 Cross-sentential co-reference 105 5.2 Co-reference and the effect of transitivizers 109 6 Deictic features I l l 7 Summary 116 Chapter Five: The non-deictic D-determiner kwi 118 1 Introduction 118 2 Structure and scope 120 2.1 Structure = wide scope? 120 2.2 Semantic features = wide scope 122 3 The position of kwi 124 3.1 Kwi is a D-determiner 125 3.1.1 Syntactic evidence that kwi is a D-determiner 125 3.1.2 Semantic evidence that kwi is a D-determiner 127 3.1.2.1 Kwi does not presuppose familiarity/novelty 127 3.1.2.2 Kwi does not assert uniqueness.. 129 3.2 N o split D P domain in Skwxwu7mesh 130 4 The analysis: kwi as a non-deictic D-determiner 134 4.1 The non-deictic D-determiner kwi 135 4.2 Narrow scope and kwi 142 4.3 Non-deictic D-determiners: composition via Restrict 143 4.4 Predictions 145 4.4.1 Restrict and word order : 146 4.4.2 Contexts for non-deictic D-determiners 147 4.4.3 Proper names 148 4.4.4 Individuation, subordinating, and non-deictic D-determiners 149 4.4.5 Non-deictic D-determiners as complementizers 150 4.5 Summary 151 5 Featureless D-determiners and the context 152 5.1 D-determiners matter 153 5.2 Bare nouns 156 5.2.1 Bare nouns, narrow scope and Restrict 156 5.2.2 Bare nouns and the lack of context 159 5.3 The contribution of D-determiners to DPs 160 5.4 Implications 163 6 Alternative analyses? 168 6.1 Kwi is not like ku in St'at'imcets 168 v 6.2 Kwi is not Longobardi's existential D 170 6.3 Kwi is not an expletive D-determiner 171 6.4 Kwi is not a Skwxwu7mesh-type quantifier 172 6.5 Kwi and ta are not different kinds of definite determiners 173 7 Conclusions and Implications 175 Chapter Six: Typological Implications 177 1 Introduction 177 2 Implications for other Salish languages 178 2.1 Salish languages with potentially non-deictic D-determiners 180 2.1.1 St'at'imcets ; '. 180 2.1.2 Sechelt 184 2.1.3 Lushootseed 187 2.1.4 Musqueam 190 2.1.5 Upper Chehalis 194 2.1.6 Cowlitz 195 2.1.7 Secwepemctsin 197 2.1.8 Nuxalk 199 2.1.9 Me?kepmxcin 201 2.2 Salish languages lacking non-deictic D-determiners 203 2.2.1 Straits 203 2.2.2 Upriver Halkomelem 206 2.3 Summary 208 3 Implications for English 209 3.1 Distinguishing D-determiners from quantifiers in English 211 3.1.1 Evidence from the cardinal/proportional readings of weak quantifiers 212 3.1.2 Evidence from domain restriction 215 3.1.3 The (lack of) evidence for every 218 3.2 Distinguishing D-determiners from demonstratives 218 3.3 Distinguishing D-determiners from indefinite articles 222 3.3.1 The semantic contribution of indefinite articles 222 3.3.2 The syntactic position of indefinite articles 224 3.4 Summary 225 4 Implications for "articleless" languages 226 5 Implications for Maori 228 6 Conclusion 231 References 177 Appendix: Key to the Skwxwu7mesh Orthography 243 vi List of Tables Table 1.1: The D-determiner system of Skwxwu7mesh 9 Table 2.1: The Salish language family 14 Table 2.2: The D-determiner system of Skwxwu7mesh 15 Table 2.3: The demonstrative system of Skwxwu7mesh 15 Table 2.4: The D-determiner and demonstrative system of Skw_wu7mesh 32 Table 2.5: The D-determiner and demonstrative system of Skwxwu7mesh 36 Table 2.6: The D-determiner system of Skwxwu7mesh j 36 Table 2.7: The D-determiner system of Skwxwu7mesh 1 39 Table 2.8: The demonstrative system of Skwxwu7mesh 39 Table 2.9: The D-determiner system of Skwxwu7mesh 51 Table 2.10: The demonstrative system of Skwxwu7mesh 52 Table 4.1: The D-determiner system of Skwxwu7mesh 112 Table 5.1: Number of occurences of D-determiners in the texts collected by Kuipers (1967).. 148 Table 5.2: Number of occurences of demonstratives in the texts collected by Kuipers (1967). 148 Table 5.3: The article system of Fering 173 Table 6.1: The St'at'imets D-determiner system 182 Table 6.2: The St'at'imets D-determiner system 183 Table 6.3: The D-determiner system of Sechelt 184 Table 6.4: The D-determiner system of Sechelt 187 Table 6.5: The D-determiner system of Lushootseed 187 Table 6.6: The D-determiner system of Lushootseed 190 Table 6.7: The Musqueam D-determiner system 190 Table 6.8: The Musqueam D-determiner system 194 Table 6.9: The D-determiner system of Upper Chehalis 194 Table 6.10: The D-determiner system of Upper Chehalis 195 Table 6.11: The D-determiner system of Upper Chehalis 195 Table 6.12: The D-determiner system of Upper Chehalis 197 Table 6.13: The D-determiner system of Secwepemctsin 197 Table 6.14: The D-determiner system of Secwepemctsin 199 Table 6.15: The "a-type" D-determiner system of Nuxalk 199 Table 6.16: The "the-type" D-determiner system of Nuxalk 199 Table 6.17: The D-determiner system of Nuxalk 200 Table 6.18: The "a-type" D-determiner system of Nuxalk 200 Table 6.19: The "the-type" D-determiner system of Nuxalk 201 Table 6.20: The D-determiner system of Nle?kepmxcin 201 Table 6.21: The D-determiner system of Nle?kepmxcin 203 Table 6.22: The determiner system of Lummi 203 Table 6.23: The determiner system of Saanich 204 Table 6.24: The D-determiner system of Upriver Halkomelem 206 Table 6.25: The D-determiner system of Upriver Halkomelem 208 Table 6.26: Salish languages and non-deictic D-determiners 209 Table 6.27: The article system of of Maori 228 vn List of Figures Figure 2.1 : Speaker and relative distances from objects 39 Figure 2.2 : Distal object 43 Figure 2.3 : Medial object 47 Figure 2.4 : Proximal object 48 Figure 2.5 : Relative distances between distal, medial and proximal objects 51 viii Abbreviations Skwxwu7mesh sg = singular 1 = first person tr = transitivizer 2 = second person 3 = third person Blackfoot abs = absolutive agreement 3' = third person obviative act — active nonaffirm = non-affirmative mood appl = applicative caus = causativizer Cowlitz comp = complementizer imperf = imperfective conj = conjunction stat = stative aspect dem = demonstrative det = determiner Fering dir = directed towards A = A-article emph = emphatic D = D-article erg = ergative agreement m = masculine f = female foe = focus particle Inuktitut fut = future abl = ablative case imper = imperative ind = indicative mood impf = imperfective inst = instrumental case inch = inchoative inter = interrogative mood indep = independent pronoun loc = locative case intr = intransitivizer part = participial mood instr = instrument irr = irrealis Lummi lc = limited control link = linking particle loc = locative predicate neg = negation Lushootseed nom = nominalizer hab - habitual o = object perf = perfective obi = oblique prog = progressive pass = passive sbj = subjunctive prefix pi = plural stat = stative aspect poss = possessive agreement prog = progressive Mandarin prox = proximal auxiliary sfp = sentence final particle pst = past Q = yes/no question particle Maori redup = reduplicant A = aspect refl = reflexive art = article rl = realis auxiliary du = dual s = subject gen = genitive sbj = subjunctive/conjunctive agreement Ident = identity copula ix Pred = predicate T = tense Norwegian def = definite Saanich accom = accompany conjee = conjecture prob = probably real = realized Secwepemctsin aor = aortive evid = evidential St'at'imcets appar = apparently compl = completive particle conj = conjunctive morphology deic = deictic dimin = diminuative reduplicant exis = existence hyp = hypothetical prog = progressive quot = quotative wh = wh-morphology Thompson enj = conjunctive particle instr = instrument rprt = reportative Upper Chehalis part = particle Upriver Halkomelem aux = auxiliary Acknowledgements There are so many people to thank I don't even know where to begin. When I first began looking at the determiners of Skwxwu7mesh, Peter Jacobs said, "You're brave." How right he was. It's a good thing I didn't realize how brave until it was too late to turn back, or I may never have discovered the beauty of the Skwxwu7mesh system. I owe a huge debt of gratitude to all of the elders who provided me with the judgments necessary to write this thesis in the first place (in the order I met them): the late Eva Lewis, the late Yvonne Joseph, the late Lawrence Baker, the late Tina Cole, Margaret Locke, Lena Jacobs and the late Frank Miranda. Huy chap a. All of you had an incredible sense of humour. Our tapes are full of laughter. Thanks also for sharing your stories with me. I learned a lot about the language, but also about the history and culture. I also owe a huge debt of gratitude to everyone on my committee. Thanks to Lisa Matthewson, supervisor extraordinaire. I honestly doubt I'd have made it this far without your encouragement, comments, and incredible energy. Your commitment to linguistics and the St'at'imcets community is awe-inspiring. Thanks to Hamida Demirdache, for encouraging me to work on Skwxwu7mesh instead of St'at'imcets way back in 1997, for inviting me to visit Nantes last year, and for always providing encouragement, right from the beginning. Your energy is still missed here at UBC. Thanks to Hotze Rullmann, for forcing me to think about semantics more carefully. My thinking has greatly benefited from all of our discussions. Thanks to Martina Wiltschko, for trying to make a better writer out of me. I hope you succeeded. Thanks also for holding amazing parties, and for introducing me to real schnapps. Thanks to my external examiner, Greg Carlson, for insightful comments on this work. If only I had the time to address them all here! Nothing gets done in a department of our size without an incredible secretary (or two). Thanks especially to Edna Dharmaratne, for keeping things running and for helping all the graduate students out, whenever we ask. She really does make the department work, and keeps many of the graduate students from despair! Thanks also to Sarah Chen, for making all of our lives just that much easier. Thanks to all my professors at UBC over the years. I especially want to thank the late M . Dale Kinkade, my very first linguistics professor. He introduced me to (Moses-Columbian) Salish, and I was hooked. I am truly sorry I didn't finish this before his death. Thanks to Henry Davis, who also encouraged me to work on Salish, and who keeps everyone on their toes. Rose-Marie Dechaine taught me that all data is interesting. Doug Pulleyblank kept me from drowning in my own paperwork. Thanks also to Laura Downing, Pat Shaw, Guy Carden and Bryan Gick who all taught me something important during my PhD. Financially, I was supported by a Killam Predoctoral Fellowship; thanks to UBC for awarding me that. Thanks also to SSHRC, Henry Davis, Lisa Matthewson, and Martina Wiltschko for financially supporting some part of this research. Thanks also to all my professors from the University of Toronto, especially Alana Johns, who taught me a lot about Inuktitut, and introduced me to Buffy. Thanks also to Diane Massam, Elizabeth Cowper, Elan Dresher, Bob Binnick, Jack Chambers, Keren Rice, and Ron Smyth. Thanks for making me welcome in big bad Toronto/teaching me new things about linguistics. Thanks to all the graduate students who made my life more liveable. Thanks to Linda Tamburri (Lalitadevi), who I can always talk to, no matter how long it's been, and who was a xi comrade from day 1. Thanks to Leora Bar-el for amazing parties and for sharing war stories with me. I think I was better-prepared for the thesis battle only because of your experience. Thanks to Rachel "Dub" Wojdak for nicknaming everyone, but also for your insane ideas about linguistic mascots. Thanks to Peter "Jacbos" Jacobs, for setting up most of my elicitations, for helping me in my work, and for your amazing sense of humour. Evil is the new good. Thanks also to Hurly Burly Adam Werle for the last minute discusssion of the deictic distinctions. Thanks to Naomi Sawai, who worried too much, to Solveiga Armoskaite, who is the best officemate anyone could hope for, to James Thompson, who rides W A Y too much, Jason "JCB" Brown, who is always the calm in the eye of the storm, to Tyler Peterson, for the sharing of stories, to Yunhee Chung, for your amazing optimism, to Sonja Thoma, for your energy, to Calisto Mudzingwa, for appreciating the eternal fight over science, to Michele Kalmar, for sharing my love of bellydance, to Ryan Waldie, for having an amazing t-shirt collection, to Masaru Kiyota, for being very kind to everyone, and to Karsten Koch, for the long discussions on NIe?kepmxcin. Thanks to everyone else who shared this experience with me: Noriko Yamane-Tanaka, Leszek Barczak, Marion Caldecott, Mario Chavez-Peon, Shujun Chin, Christie Christodoulou, Gene Chu, Clare Cook, Donald Derrick, Atsushi Fujimori, Fusheini Hudu, Yoko Ikegami, Kristin Johannsdottir, Susie Jones, Jeff Muehlbauer, Yumiko Nakamura, Nahal Namdaran, Martin Oberg, Dominique Quis, Randy Sharp, Amelia Silva, Olga Steriopolo, Oladiipo Ajiboye, Maxine Batiste, Fiona Campbell, Suzanne Gessner, Jennifer Glougie, Andrea Gormley, Eun-sook Kim, Ramona McDowell, Jeremy Perkins, Christine Ravinski, Ad Ruangjaroon, Scott Shank, Kayono Shiobara, and last, but not least, Ian Wilson, who beat me to the finish line. I can't thank you all enough. Thanks to all those outside of linguistics, who contributed to my sanity. Jose Soto, my one-time boss and good friend, thanks for the parties and the Venezuelan food. Joydeep Sengupta, the workaholic, thanks for the sari, and for feeding all of us wonderful food. Thanks to Brian de Alwis and Cheryl Wiramanaden for many drinks, dinners and discussions. Cheryl, finish already, so we can graduate together. Thanks to the brunch bunch: Deborah Bleakley, Desiree Sattler and April Wessel. April especially shares my love of thrift stores and nerdly obsessions. Thanks to Dave Tompkins, for hosting many tv/movie nights and for always driving me home. Thanks to Becky Solomon, for sharing my love of 6 Feet Under. Or, perhaps more accurately, for enjoying my enjoyment of 6 Feet Under. Thanks to Kirsten Baker-Williams, for adding a little bit of drama every week. You also should finish already so we can graduate together. Thanks to Erin Young, for long conversations on the nature of academic life. I hope we all find the balance. Thanks to Vighen Pacradouni, for making my PhD seem short, and for your words of wisdom on the thesis process. Thanks to Davor Cubranic and Nicole Dorphan for the games nights. Thanks to Joey Kish for the pasta nights. Thanks to all of my family for financial and emotional support over the years. Everyone should have a roommate/sister as good as Bronwyn. Thanks to Tristan, for being the weird one (or else, I'm sure it would be me). Thanks to my parents, for a million things, but especially for letting us rent their retirement condo. (I'm moving out soon, I hope!) Thanks to my cousin Christie Laidlaw for sharing a love of Sha Lin noodle soup. Take care out there in Afghanistan. Thanks to Kitty, aka Wheezy, aka Lemon, for being the cutest, crankiest half-hairless cat I'll ever know. To Suresh Pillai, aka Hash, aka The Scientist, thanks for being there at just the right time. I'm not sure I would have finished without your support these past two years. I also enjoyed our many arguments since I first began the PhD. Is linguistics a science yet? xii for all the Skwxwu7mesh elders Chapter One: Introduction 1 Introduction Th is thesis addresses the quest ion o f what it is semant ica l ly that a determiner can do. B y 'determiner ' , I do not refer to a l l general ized quant i f ier-creat ing elements, as B a r w i s e and Coope r (1981) do. Instead, I refer to a subset o f noun phrase in i t ia l elements - a subset more s im i la r to those that have been ca l led 'a r t i c les ' . 1 (See a lso G ius t i 199.1 w h o argues that the term 'determiner ' obscures the structural and categor ial di f ferences between the many types o f 'determiners ' ) . In order to d is t inguish between the more tradi t ional use o f the term 'determiner ' (wh ich inc ludes demonstrat ives, numerals and quant i f iers) and the mean ing intended here, I use the term D-determiner. The ma in c l a i m o f this thesis is that a l l D-determiners share a c o m m o n core semantics: doma in restr ict ion (cf. Westerstahl 1984 and v o n F in te l 1994). T h i s doma in restr ict ion, imp lemented by a contextual var iable C , restricts the doma in o f quant i f icat ion. There is a long-standing debate about the pos i t ion o f doma in restr ict ion. Stanley and Szabo (2000) and Stanley (2002) argue that doma in restr ict ion is associated w i th the noun , w h i l e v o n F in te l (1994) and M a r t i (2003) argue that doma in restr ict ion is associated w i th the D-determiner or quanti f ier. G i a n n i k i d o u (2004) argues that domain restr ict ion is associated w i th the nomina l i n Eng l i sh and the D-determiner in S t 'a t ' imce ts Sa l i sh . I argue instead that, un iversa l ly , on l y D-determiners are associated w i th doma in restr ict ion. In this thesis, I focus on the D-determiner system o f S k w x w u 7 m e s h Sa l i sh (also k n o w n as Squamish) . S k w x w u 7 m e s h D-determiners prov ide ev idence for the c l a i m that D-determiners are associated w i th doma in restr ict ion. A l l o f the D-determiners i n S k w x w u 7 m e s h are sensit ive to the context they are used in . A l l D-determiners in S k w x w u 7 m e s h can be used to refer to a prev ious ly ment ioned referent. Th i s is because they are a l l associated w i th doma in restr ict ion. One o f the D-determiners dif fers f rom the rest o f the system in that it must take narrow scope (kwi), s im i la r l y to bare nouns in languages l i ke E n g l i s h (Car l son 1980), M a n d a r i n 1 However, I do not analyze the English indefinite article a as belonging to this set. See Chapter 6 for discussion. 1 (Rullmann and You 2003), Brazilian Portuguese (Muller 2005) and Blackfoot (Glougie 2000). I claim that kwi DPs compose via Restrict, as can bare nouns (cf. Chung and Ladusaw 2004). However, kwi DPs and bare nouns differ in one crucial aspect: kwi DPs can be used to refer to previously mentioned referents, and can also be used partitively. Bare nouns (in languages with overt D-determiners) cannot do this. I claim that bare nouns lack D-determiners and are unable to refer to a previously mentioned discourse referent. Kwi DPs can refer to a previously mentioned referent (or be used partitively) precisely because they have a D-determiner (kwi). The behaviour of the D-determiner kwi also provides us with evidence that there is no strict correlation between the presence of more structure and the ability to take wide scope. Instead, I argue that a lack of features is correlated with the obligation to take narrow scope. Skwxwu7mesh D-determiners behave significantly differently from those of better studied languages, and also differ in their semantics from those of St'at'imcets (cf. Matthewson 1998, 1999), a related Salish language. I will assume that the English D-determiner the asserts the uniqueness of its referent. Skwxwu7mesh D-determiners do not assert the uniqueness of their referents. Instead, sentences containing (most) DPs are associated with an implicature of uniqueness of the referent of the DP. The DPs are normally interpreted as referring to the unique entity which matches the NP description; however, this uniqueness can be cancelled. On the basis of the differences between English and Skwxwu7mesh, I argue that D-determiners can vary with respect to (i) whether they are used in only familiar contexts or both familiar and novel contexts, (ii) whether they assert uniqueness, and (iii) whether they carry deictic information. (This is not meant to be an exhaustive list: other information may also be encoded.) For example, Skwxwu7mesh D-determiners (i) can occur in both familiar and novel contexts (as can all Salish D-determiners; see Matthewson 1998) (ii) do not assert uniqueness, and (iii) can carry deictic information.2 The English D-determiner the, on the other hand, can only occur in familiar contexts, asserts uniqueness and is not deictic. The properties listed above are not all independent of one another. I argue that the familiarity effects seen with English the are derived from the domain restriction and the assertion of uniqueness. Skwxwu7mesh lacks familiarity effects, which I derive from the lack of assertion 2 Kwi differs from the rest in not carrying any deictic information (as I show in Chapter 5); however, it behaves like the other determiners with respect to familiarity and assertion of uniqueness. 2 of uniqueness. Deictic information is independent of the other two: theoretically, a language could have D-determiners which asserted uniqueness and encoded deictic information. 3 Skwxwu7mesh D-determiners provide us with evidence that the effects of definiteness in English must be teased apart. Both English and Skwxwu7mesh D-determiners share domain restriction. However, the lack o f assertion o f uniqueness in Skwxwu7mesh allows the D -determiners to be used in novel contexts, unlike English the. I extend the property of domain restriction to all D-determiners, no matter what other properties they may have. (1) A l l D-determiners have domain restriction in their denotation. In this thesis I propose a semantic definition of a D-determiner, which I crucially link to the syntax. In the traditional semantics literature, 'determiner' refers to anything that creates a generalized quantifier from a predicate (see, e.g. Barwise and Cooper 1981). That analysis makes no reference to the syntax of determiners. For example, more than one is treated as a determiner. This is unexpected i f all determiners occupy the same head (since more, than, and one are all themselves heads). I w i l l argue in this thesis that there is a link between the syntax and the semantics of D-determiners: i f an element has a particular semantics, it occupies D , and i f an element occupies D , it w i l l have that particular semantics. For example, all cannot be a D -determiner because it does not occupy D . We can see this in (2) below. Assuming that the occupies D , all may not also occupy D . (2) A l l the men walked. I argue that the position D is strictly tied to one particular meaning. (3) Domain restriction is only introduced by D-determiners. I thus argue that D-determiners have a semantic 'core'. Some researchers have claimed that the syntactic position D is associated with certain distinctions (definiteness, specificity, etc.), but do not share a particular core semantics (see Matthewson 1998, for example). Unlike English D-determiners, Salish D-determiners (including Skwxwu7mesh) do not encode definiteness (Matthewson 1998, Gi l lon 2003; see also Chapter 4). This led Matthewson to conclude that the position D does not have the same semantics across all languages. Here I argue against this and 3 Demonstratives are potential examples of this. However, I do not treat demonstratives as D-determiners. See Chapter 6 for discussion. 3 instead argue that all D-determiners share something in common. That is, even though D-determiners may not have the exact same semantics (as they can vary with respect to assertion of uniqueness, for example), they share a core semantics. I also argue against Matthewson's (1998) claim that Salish D-determiners do not access the common ground of the discourse. Salish D-determiners, according to Matthewson, have a particular setting of a Common Ground Parameter, as in (4). (4) Common Ground Parameter Determiners may access the common ground of the discourse Yes: {English,...} No: {Salish,...} (Matthewson 1998) Matthewson appealed to this in order to capture the fact that Salish D-determiners do not encode definiteness. However, I argue that all D-determiners in Skwxwu7mesh are sensitive to the context they are used in and that they are able to access the common ground, via domain restriction. I derive the lack of definiteness effects from the lack of assertion of uniqueness instead. 2 The background In this section, I outline the issues that arise in both Skwxwu7mesh and English. Skwxwu7mesh lacks a definite/indefinite distinction, or a non-assertion of existence/assertion of existence distinction (cf. Matthewson 1998). English has a definite/indefinite distinction. The problem is how to relate these two systems. Do they share anything in common? 2.1 The problem in Skwxwu7mesh Skwxwu7mesh DPs do not encode a definite/indefinite contrast. In (5)a, the DP can be used in a novel context, where the referent has not been previously mentioned. This same effect is found in (5)b, where the DP is used in an existential context. (5)c can be used following either of (5)a or b. Here the hearer is familiar with the referent, as it has already been introduced. The use of 4 the DPs in novel or familiar contexts is not affected by the choice of D-determiner. There are eight D-determiners in Skwxwu7mesh (four gender-neutral and four female);4 here I show the gender-neutral D-determiners in both novel and familiar contexts. (All data is from my own fieldwork, unless otherwise noted.) (5) a. Chen kw'ach-nexw ti/ta/kwa/kwi swi7ka. Isg.s look-tr(lc) det man 'I saw a man.' (novel context) b. Tsi7 ti/ta/kwa/kwi swi7kana7 ta lam'. exist det man loc det house 'There's a man in my house.' (novel context) c. Na kw'ay' ti/ta/kwa/kwi swi7ka. rl hungry det man 'The man is hungry.' (familiar context) Given that there are so many D-determiners in Skwxwu7mesh, the question arises as to what distinctions are encoded by these D-determiners. A possible analysis has already been provided by Matthewson (1998), the first in-depth study of Salish D-determiners. Her main claim is that Salish D-determiners (with the exception of Straits Salish) encode an 'assertion of existence' distinction. However, Skwxwu7mesh D-determiners, as we shall see, do not encode this distinction.. The 'assertion of existence' distinction is one between those D-determiners which assert the existence of their referent and those that do not. Matthewson argues that St'at'imcets (along with most Salish languages) makes this distinction. In St'at'imcets, most of the D-determiners assert the existence of their referent. In both (6)a and b, the D-determiner ti...a asserts the existence of the book. Thus, the DP tipu'kwa 'the/a book' cannot take narrow scope with respect to the non-factual operator kelh 'might' in (6)b. (6) a. tecwp-min-lhkan ti piikw-a lhkunsa. buy-appl-lsg.s det book-det today 'I bought a/the book today.' 3x, book (x), I bought x today 4 One of the female D-determiners is not often used, as I discuss in Chapter 5. 5 b. tecwp-min-lhkan kelh ti piikw-a natcw. buy-appl-lsg.s might det book-det tomorrow 'I might buy a/the book tomorrow.' 3x, book (x), I might buy x tomorrow (St'at'imcets; Matthewson 1998) In St'at'imcets, there is one D-determiner (ku) which does not assert the existence of its referent. According to Matthewson, this non-assertion of existence D-determiner cannot occur in a declarative sentence; ku requires the presence of a non-factual operator, as in (7)a.5 (7) a. tecwp-min-lhkan kelh ku pukw natcw. buy-appl-lsg.s might det book tomorrow 'I might buy a book tomorrow.' b. * tecwp-min-lhkan ku ptikw lhkiinsa. buy-appl-lsg.s det book today (I bought a book today) (St'at'imcets; Matthewson 1998) Matthewson claims that this restriction on non-assertion of existence D-determiners to these environments holds in other Salish languages. While she does not explicitly claim that Skwxwu7mesh D-determiners encode this distinction, she suggests that most Salish languages encode this distinction. However, it cannot be the correct analysis for Skwxwu7mesh. There is no 'non-assertion of existence' D-determiner in Skwxwu7mesh. A l l D-determiners are equally available in declarative sentences as well as in those with non-factual operators. However, different DPs can take different scope with respect to an operator. Kwa DPs must take wide scope, ti and ta DPs can take either wide or narrow, and kwi must take narrow scope. (8) a. Chen silh7-an ti/ta/kwa/kwi sts'ukwi7. Isg.s buy-tr det fish 'I bought a/the fish.' b. A : Nii chexw silh7-an kwa sts'ukwi7? rl.Q 2sg.s buy-tr det fish 'Did you buy a/the fish?' (wide scope) B: # Haw, hawk sts'ukwi7. neg be. not fish 'No, there weren't any fish.' 5 It is also licit in object position of morphologically intransitive verbs (Davis and Matthewson 2003). I do not discuss this issue here. 6 ti/ta sts'ukwi7? det fish (wide or narrow scope) kwi det sts'ukwi7? fish (narrow scope) B: Haw, an tl'i7. neg very dear 'No, it was too expensive.' c. A: Nu chexw silh7-an rl.Q 2sg.s buy-tr 'Did you buy a/the fish?' B: Haw, hawk sts'ukwi7. neg be. not fish 'No, there weren't any fish.' B: Haw, an tl'i7. neg very dear 'No, it was too expensive.' c. A: Nu chexw silh7-an rl.Q 2sg.s buy-tr 'Did you buy a fish?'6 B: Haw, hawk sts'ukwi7. neg be.not fish 'No, there weren't any fish.' B: # Haw, an tl'i7. neg very dear 'No, it was too expensive.' The D-determiner kwi is therefore the closest candidate to a 'non-assertion of existence' D-determiner; however, as shown in (8)a, it can occur in declarative sentences. It cannot be a non-assertion of existence D-determiner. The question then is: what distinctions are encoded by Skwxwu7mesh D-determiners? I will argue that deictic distinctions are the only relevant factor and that the D-determiner kwi has no deictic features at all. Matthewson (1998) has already shown that deictic distinctions are found in all Salish languages, but here, the deictic distinctions will receive primary focus. Further, I will derive the differences between the D-determiners in Skwxwu7mesh directly from the presence or absence of deictic properties. 6 The lack of a definite interpretation will be shown to derive from kwi's obligatory narrow scope. See Chapter 5 for more discussion. 7 The non-deictic D-determiner kwi behaves the same in this respect as he in Maori (cf. Chung and Ladusaw 2004). See Chapter 6 for more discussion. 7 2.2 The problem in English English does have an indefinite/definite distinction. A is only felicitous in novel contexts, as in (9)a and b, whereas the is (usually) only felicitous in familiar contexts.8 Novel contexts are contexts where the hearer is not acquainted with the referent of the DP. Familiar contexts are contexts where both the speaker and the hearer are acquainted with the referent of the DP. (9) a. I saw a man. (novel context, *familiar context) b. There's a man in my house, (novel context, *familiar context) c. I saw the man. (familiar context, *novel context) Definite D-determiners have been associated with many different properties: assertion of existence and uniqueness (Russell 1998[1905]), presupposition of existence and uniqueness (Strawson 1998[1950], Kadmon 1992), familiarity (Christophersen 1939, Heim 1988), inclusiveness (Hawkins 1978), etc. Indefinite articles have also been associated with many different properties: as existential quantifiers (Russell 1998 [1905]), as choice functions (e.g., Reinhart 1997), or even as presupposing uniqueness (Percus 1998). Given that both indefinites and definites have been associated with a presupposition of uniqueness, what is the difference between a and the? I argue that the shares a core semantics with other D-determiners, and that a does not. In fact, I argue that a is not a D-determiner at all. The goal of this thesis is a unified semantics for D-determiners. There is much variation between English and Skwxwu7mesh (and between any two unrelated systems). However, I show that there is commonality between these disparate systems: the presence of domain restriction in D. 2.3 Why are these two problems related? Skwxwu7mesh and English seem to be very different from each other. The definite D-determiner in English bears little obvious relation to the many D-determiners in Skwxwu7mesh.The D-determiners in Skwxwu7mesh cannot be analyzed using the traditional notions of familiarity or uniqueness. However, they do appear to serve a similar function as other D-determiners. The 8 In some cases, the can be used in novel contexts (Heim 1988). See Chapter 3 for more discussion. 8 question then is , what, i f anything, do D-determiners i n Skwxwu7mesh and E n g l i s h have i n c o m m o n ? D o m a i n restr ict ion appears to p lay a role i n both Skwxwu7mesh and E n g l i s h . I pursue the idea that d o m a i n restrict ion is shared b y D-determiners cross- l inguis t ica l ly . T h e analysis I prov ide for Skwxwu7mesh is (partial ly) appl icable to E n g l i s h . T h e properties o f the D-determiners i n Skwxwu7mesh p r o v i d e us w i t h evidence for universa l c la ims about D-determiner denotations, but also for a specif ic analysis o f the E n g l i s h definite D -determiner. I p r o v i d e a u n i f i e d analysis o f the core properties shared b y a l l D-determiners i n both languages, as w e l l as an analysis o f the cross- l inguist ic differences i n D-determiner semantics. 3 P rev iew of the analysis I argue that there are two differences between Skwxwu7mesh and E n g l i s h D-determiners . U n l i k e E n g l i s h the, Skwxwu7mesh D-determiners have deict ic features a n d are not used sole ly i n definite contexts. T h e deict ic features are g iven i n the table b e l o w . D e i c t i c N o n - d e i c t i c N e u t r a l P r o x i m a l D i s t a l , i n v i s i b l e gender-neutral ta t i k w a k w i feminine lha tsi k w e l h a kwes T a b l e 1.1: T h e D-determiner system o f Skwxwu7mesh. D-determiners i n both E n g l i s h and Skwxwu7mesh share d o m a i n restr ict ion, however . I assume that the definite/indefinite d is t inct ion i n E n g l i s h arises f r o m the assertion o f uniqueness encoded b y the, roughly f o l l o w i n g H e i m and v o n F i n t e l (2001). I argue that the difference between the indefinite article a and the definite D-determiner the does not arise sole ly due to the assertion o f uniqueness; the difference also arises f r o m d o m a i n restr ict ion, w h i c h is introduced by the D-determiner the (but not the indefinite art ic le a). D-determiners i n Skwxwu7mesh do not assert uniqueness. I argue that i n Skwxwu7mesh a subset o f the D-determiners instead are associated w i t h an implicature o f uniqueness. Uniqueness is therefore s t i l l a factor i n this language. H o w e v e r , one o f the D-determiners is not associated w i t h an impl icature o f uniqueness. Therefore, uniqueness is not a necessary part o f the semantics o f D-determiners. I predict that unless there is a contrast i n the D-determiner system o f 9 any given language (along deictic lines, for example), all D-determiners will be associated with uniqueness: either by assertion, or by implicature. I derive the familiarity effects of English the from the assertion of uniqueness and the presence of domain restriction. The forces the referent to be the maximal member of the contextually restricted domain that matches the NP description. The only way to know if the domain contains this unique member is if the referent is familiar to the hearer (in the spirit of Kadmon 1992). I provide the denotation of the below. (10) [[the] = XP max(kx [P(x) A C ( X ) ] ) I argue that indefinite articles are not D-determiners, and do not introduce domain restriction over their NP. Therefore, even if they presuppose uniqueness (as argued by Percus 1998), the speaker does not assume that the hearer knows the referent. The presupposition of uniqueness in and of itself is not enough to force familiarity; it is only the interaction between the presupposition or assertion of uniqueness and domain restriction. I argue that the lack of familiarity effects in Skwxwu7mesh arises from the lack of assertion of uniqueness. The domain restriction provided by the D-determiner can be accommodated (cf. Lewis 1979) because the speaker does not need to assert the uniqueness of the referent. The hearer therefore does not have to be familiar with the referent. I provide the denotation of the deictic D-determiners below. I provide a choice function analysis of the deictic D-determiners. (11) [[ta] = XP f(kx [P(x) A C(x)]) The non-deictic D-determiner kwi also does not assert uniqueness, nor does it presuppose familiarity. Unlike deictic DPs, non-deictic DPs must take narrow scope. I claim that it does this because it composes with the predicate via Restrict (cf. Chung and Ladusaw 2004). (12) [[kwi]] = XP Xx [P(x) A C(x)] I argue that the core semantics of D-determiners is domain restriction; much of the rest of the semantics may vary. In (13) below, I give the denotations for English the, and Skwxwu7mesh deictic D-determiners and the non-deictic D-determiner..All contain domain restriction. 10 (13) a [the]] = AP max(Xx [P(x) A C ( X ) ] ) b. pa]] = XP f(kx [P(x) A C(x)]) c. _kwi_ = AP A,x [P(x) A C(x)] This is unlike bare nouns, which do not have domain restriction. (14) D introduces domain restriction; NPs lack domain restriction 4 Ou t l i ne of the thesis This thesis has the following structure. In Chapter 2,1 provide some background information on Skwxwu7mesh, including the D-determiner system, and other morphological, syntactic or semantic information necessary for understanding the data presented in this thesis. I show that there are no argument nominals smaller than a DP in Skwxwu7mesh. I also provide preliminary analysis of the deictic D-determiners, as well as evidence for the deictic features in Skwxwu7mesh. In Chapter 3,1 provide the theoretical background for this thesis. This includes a discussion of presupposition and implicature, the debate on definiteness in English, and the background on domain restriction of DPs. I also provide an overview of my analysis for English in order to compare to the analysis of Skwxwu7mesh in Chapter 4. In Chapter 4,1 show that Skwxwu7mesh D-determiners can be used in both novel and familiar contexts. This has already been shown for St'at'imcets D-determiners as well (Matthewson 1998). I also show that Skwxwu7mesh D-determiners do not assert the uniqueness of their referent; instead, sentences containing a deictic DP carry an implicature of uniqueness of the referent of that DP. I relate the implicature to the domain restriction associated with the D-determiner: it is simpler to assume the speaker is talking about the unique referent in the context, unless he or she (or the context) gives you reason to believe otherwise. In Chapter 5,1 argue that there is a non-deictic D-determiner in Skwxwu7mesh. This non-deictic D-determiner can also be used in novel and familiar contexts. However, sentences containing a non-deictic DP do not carry an implicature of uniqueness of the referent of the DP. I 11 argue that this is because the non-deictic D-determiner does not have deictic features. This lack o f deictic features has other consequences for the grammar: a D P headed by this D-determiner must take narrow scope with respect to any quantifier or operator because it composes via Restrict. DPs headed by this D-determiner behave somewhat like bare nouns in other languages. However, this D-determiner also introduces domain restriction, unlike bare nouns. This domain restriction allows the non-deictic D-determiner to be interpreted partitively. In Chapter 6,1 discuss some typological issues that arise from my analysis. I suggest that non-deictic D-determiners are present in many other Salish languages. I make predictions as to how those non-deictic D-determiners should behave. M y analysis of Skwxwu7mesh also forces me to conclude that English D-determiners are also associated with domain restriction. I raise the question of whether quantifiers are also associated with domain restriction, and suggest that they are not. I argue instead that there is a set of elements which are domain restrictors over their N P , and that this set of elements are D-determiners. I argue that these elements occupy D. I also extend the analysis to other unrelated languages. I conclude that N P s cannot be associated with domain restriction and that only DPs are forced (where possible) to refer to previously mentioned referents. 12 Chapter Two: The Background on Skwxwii7mesh 1 Introduction The purpose of this thesis is to explore the semantics of Skwxwu7rnesh D-determiners, and to . compare them to English D-determiners.1 In this chapter I provide some background to facilitate understanding of the proposals made in the next few chapters. In particular, I give an overview of the syntax and morphology of Skwxwu7mesh as these topics are necessary for understanding the data and analysis provided later in the thesis. I begin with the language family. Skwxwu7mesh is a Central (or Coast) Salish language spoken in southwestern British Columbia. The list of Salish languages is given in Table 2.1 below. Languages marked with * are extinct. ' Recall that this term does not include quantifiers, numerals or demonstratives. 13 Branch Language Dialects Nuxalk (Bella Coola) Central Salish Comox Sliammon, Klahoose, Homalko, Island Comox Pentlatch* Sechelt Skwxwu7mesh (Squamish) Halkomelem Chilliwack/Upriver Halkomelem, Musqueam, Nanaimo/Cowichan Nooksack* Northern Straits Semiahnioo, Saanich, Lummi, Songish, Samish, Sooke Klallam Lushootseed Northern, Southern Twana* Tillamook* Tsamosan Upper Chehalis Cowlitz* Lower Chehalis Quinault* Satsop, Oakville, Tenino Interior Northern St'at'imcets (Lillooet) Nie?kepmxcin (Thompson) Secwepemctsin (Shuswap) Mount Currie/Lower Lillooet, Fountain/Upper Lillooet Eastern, Western Southern Okanagan Moses-Columbian Kalispel Coeur d'Alene Northern, Southern/Colville Spokane, Kalispel, Flathead Table 2.1: The Salish language family (adapted from Thompson and Kinkade 1990: 34-35) The D-determiner systems of some of these languages will be addressed in Chapter 6. Skwxwu7mesh is extremely endangered. There are fewer than twenty speakers remaining. I worked with seven native speakers (five female and two male) in order to gather the data necessary for this dissertation. The speakers did not always have the same judgments; where there is speaker variation, I note it below. In §2,1 provide information on the methodology used to gather the data for this dissertation. In §3,1 provide some background on the morphology of Skwxwu7mesh. Salish languages, including Skwxwu7mesh, are quite different from English syntactically and morphologically. Salish languages are radically head-marking languages; arguments are obligatorily marked on the predicate via pronominal agreement morphology. Pronominal agreement morphology is any affix attached to a verb marking the subject or object of the clause. 14 In §4,1 provide some background information on the syntax of Skwxwu7mesh. Null arguments (both object and subject) are commonly found in texts and in conversation. Where DPs are used, the word order is relatively free, with some important restrictions discussed below. In §5,1 provide basic information on the D-determiner system. For example, D-determiners are obligatory on any overt argument. In §6,1 discuss previous analyses of the Skwxwu7mesh D-determiner system. In §7,1 provide detailed information about the D-determiner system, including a re-analysis of the structure of the Skwxwu7mesh D-determiner system. I provide evidence for the deictic features of the deictic D-determiners and demonstratives, and show the contexts where the deictic D-determiners and demonstratives can be used. The deictic features are provided in the tables below. Deictic Non-deictic Neutral Proximal Distal, invisible gender-neutral ta ti kwa kwi feminine lha tsi kwelha kwes Table 2.2: The D-determiner system of Skwxwu7mesh. Neutral, invisible Proximal Medial Distal Unmarked Invisible gender-neutral number-neutral kwiya(wa) ti(wa) tay' kwetsi plural kwiyawit iawit itsiwit kwetsiwit kwawit feminine kwiya(wa) tsiwa alhi kwelhi Table 2.3: The demonstrative system of Skwxwu7mesh. 2 Methodology In gathering the data for this thesis, subtle judgments about meaning were required from the speakers. In order to get this kind of information, it was necessary to set up different contexts. This was often done in English. I would then provide the speakers with a sentence, and ask if that sentence made sense in the context given. Sometimes, I provided the context in Skwxwu7mesh. Where the context was provided in Skwxwu7mesh, the entire discourse is provided in the examples. I then asked the speakers if the discourse I had given them made sense to them. In other cases, pictures were shown to the speakers. I sometimes elicited comments on the pictures 2 The suffix -wa is only licit if the referent is human. 15 by asking them to describe the situation. Other times I offered Skwxwu7mesh sentences and asked the speakers if that sentence could be used to describe the picture. I re-elicited the same sentences in different sessions, to test whether the judgments were firm. 3 M o r p h o l o g y Skwxwu7mesh is a radically head-marking language. Head-marking languages indicate syntactic relationships via agreement morphology on the head of the phrase (see Nichols 1986, Baker 1996). Pronominal agreement morphology appears on predicates, as in (1). (1) Na ch'aw-at-ts-as ta swi7ka. rl help-tr-lsg.o-3erg det man. 'The/a man helped me.' Skwxwu7mesh displays split-ergative properties. First and second person follow a nominative-accusative pattern (2), whereas third person follows an ergative pattern (3). (2) a. Chen ch'aw-at-umi. Isg.s help-tr-2sg.o 'I helped you.' b. Chexwch'aw-at-ts. 2sg.s help-tr-lsg.o 'You helped me.' c. Chen imesh. Isg.s walk 'I walked.' (3) a. Na ch'aw-at-ts-as. rl help-tr-lsg.o-Serg 'S/he helped him/her.' b. Chen ch'aw-at-0. Isg.s help-tr-3abs 'I helped him/her.' c. Na imesh-0. rl walk-3abs 'S/he walked.' 16 Assuming that all arguments are marked on the verb, third person absolutive is marked by zero morphology. For the remainder of the thesis, I will not mark third person absolutive morphology in the glosses. Possessors are marked on the head. The possessive morphology is affixed to the head noun (Kuipers 1967).3 (4) a. ta-n skwemay' b. ta skwemay'-chet det-lsg.poss dog det dog-lpl.poss 'my dog' 'our dog' c. ta e-skwemay' d. ta skwemay'-yap det 2sg.poss-dog det dog-2pl.poss 'your (sg) dog' 'your (pi) dog' e. ta skwemay'-s f. ta skwemay'-s-wit det dogSposs det dog-3poss-3pl 'his/her dog' 'their dog' Plurality is also marked on the noun. This is done via a CoC- reduplicant (Kuipers 1967). (5) a. mixalh b. mex-mixalh bear redup-bear 'bear' 'bears' However, the unmarked form of the noun can still get a plural interpretation. "Usually number remains unspecified" (Kuipers 1967: 100). (6) a. Chen kw'ach-nexw ta push. Isg.s look-tr(lc) det cat 'I saw a cat/the cat/cats/the cats.' b. Chen kw'ach-nexw ta pesh-push. Isg.s look-tr(lc) det redup-cat 'I saw (the) cats.' Transitive predicates are usually marked as such by transitive morphology, which are often called transitivizers. 3 The first person singular possessor morphology usually attaches to the preceding D-determiner. 17 (7) a. Kw'elh ta stakw. spill det water 'The water spilled.' b. Chen kw'elh-at ta Isg.s spill-tr det 'I spilled the water.' stakw. water Transitivizers in Salish encode degrees of "control" (Thompson 1979). Kuipers (1967) originally characterized these as volitional versus non-volitional. However, I will refer to them as control versus limited control transitivizers. The control transitivizers include -t, - Vt,4 -(a)n and -s. The limited control transitivizer is -nexw. An example of the difference in meaning is given below. (8) Chen xep'-t ta xel'-tn. Isg.s break-tr det write-instr 'I broke the pencil.' b. Chen xep'-nexw ta xel'-tn. Isg.s break-tr(lc) det write-instr 'I accidentally broke the pencil.' In (8) a, the agent of the action has full control of the situation. The agent intended to break the pencil. However, in (8)b, the agent did not intend to break the pencil, and the limited control transitivizer encodes this. 4 Syntax Clauses in Skwxwu7mesh minimally contain a predicate and a particle or pronoun. Overt DPs are optional. (9) a. Na ch'aw-at-ts-as lha slhanay'. rl help-tr-lsg.o-Serg det.f woman. 'The/a woman helped me.' b. Na ch'aw-at-ts-as. rl help-tr-lsg.o-3erg 'S/he helped me.' 4 Here the V represents an echo vowel, which matches the vowel of the stem it attaches to. 18 c. Chen ch'aw-at. Jsg.s help-tr 'I helped him/her.' (10) a. Na ts'its'ap' ta swi7ka. rl work det man 'The/a man worked.' b. Na ts'its'ap'. rl work 'S/he worked.' c. Chen ts'its'ap'. Isg.s work 'I worked.' The particle na, here glossed as 'realis', is obligatory, or at least highly preferred when the subject is third person, as shown in (1 l)a.5 If there is another particle in the clause, such as the imperfective marker wa, na is not required, for most speakers, as in (1 l)b. Both can co-occur, as in(ll)c. (11) a. ? Ts'its'ap'. work 'Worked.' Consultant's comment: "Not a full sentence. " b. % Wa ts'its'ap'. impf work 'S/he is working.' c. Na wa ts'its'ap'. rl impf work 'S/he is working.' Skwxwu7mesh is a predicate-initial language. It allows both VSO and VOS word orders (Kuipers 1967, Currie 1997, Gillon 1998b). 5 The semantics of na are not fully understood. The term 'realis' is misleading. Na can occur in irrealis contexts, such as questions. (i) Nu[=na+u] chexw ts'its'ap'? rl.Q [rl Q] 2sg.s work 'Did you work?' More research is required into this particle. 19 (12) a. Na ch'aw-at-as lha Vanessa ta Peter. rl help-tr-3erg det.f Vanessa det Peter 'Vanessa helped Peter.' (VSO) b. Na ch'aw-at-as ta Peter lha Vanessa. rl help-tr-3erg det Peter det.f Vanessa 'Vanessa helped Peter.' (VSO) Post-verbal word order is free for most speakers.6 Pre-verbal DPs are also possible. Arguments can be clefted. If transitive subjects are clefted (as in (13)b), the ergative morphology is missing/deleted (Kuipers 1967, Gillon 1998a; see also Gerdts 1988 for the same facts in Halkomelem). This is the same pattern of morphology found in relative clauses. Clefts are often introduced by nilh. Kroeber (1999) calls these 'introduced clefts'. (13) a. Nilh ta swi7ka na ts'its'ap'. foe det man rl work 'It's the man that worked.' b. Nilh ta swi7ka na ch'aw-at-ts. foe det man rl help-tr-lsg.o 'It's the man who helped me.' c. Nilh ta swi7ka na ch'aw-at-an. foe det man rl help-tr-lsg.erg 'It's the man that I helped.' Extraction of transitive subjects can also occur without the focus particle nilh. These types of constructions are referred to as 'pseudo-clefts' in Gerdts (1998) and 'bare clefts' in Kroeber (1999).7 In(14), ta swi7ka 'the man' must be interpreted as the subject of the clause, because the object is marked by first person pronominal agreement morphology (-ts-). 6 Animacy, context and the use of proper names all seem to play roles in determining post-verbal word order. 7 It may be that intransitive subjects can also undergo this process, but it is impossible to tell on the surface. (i) Ta swi7ka na ts'its'ap'(-0). det man rl work-3abs 'The man worked.' 'It's the man who worked.' Recall that the pronominal agreement for third person is null (or -0). If this agreement is present and then deleted, it makes no overt difference. See Roberts (1999) for discussion of the presence or absence of agreement morphology in clefts in St'at'imcets. See also example (16) below. 20 (14) Ta swi7ka na ch'aw-at-ts. det man rl help-tr-lsg.o 'It's the man who helped me.' Objects, on the other hand, may not be clefted without the use of nilh. In (15), ta swi7ka must be interpreted as the object of the lower clause, because the subject is marked by first person pronominal agreement morphology (-an). (15) *Ta swi7ka na ch'aw-at-an. det man rl help-tr-lsg.erg (It's the man that I helped.) Subjects may also precede the verb without being clefted (i.e. without the loss of ergative morphology). SVO is a possible word order (Currie 1997). Some speakers use this order more frequently than others do; however, it is available to all speakers. (16) Lha Vanessa na kw'ach-nexw-as ta Peter. det.f Vanessa rl look-tr(lc)-3erg det Peter ' Vanessa saw Peter.' (SVO) This must be distinguished from A'-extraction (as in (14)), because the ergative morphology is still present. This fronted position is not available in embedded clauses for subjects. VSO and VOS are the only acceptable word orders. (17) a. Chen tsut [kwi-s-e-s kw'ach-nexw-as ta swi7ka Isg.s say comp-nom-rl-3poss look-tr(lc)-3erg det man lha slhanay']. det.f woman (i) 'I said that the man saw the woman' (VSO) (ii) 'I said that the woman saw the man.' (VOS) b. * Chen tsut [kwi-s-e-s ta swi7ka kw'ach-nexw-as Isg.s say comp-nom-rl-3poss det man look-tr(lc)-3erg lha slhanay']. det.f woman (SVO) For more information on word order and fronting in Salish, see Kroeber (1999). 21 Clauses with two DPs are rare in discourse or texts. Familiar referents are often null. After the DP kwetsi mixalh 'a bear' is introduced into the story, the bear can be referred back to using a null pronoun. (18) ...N-s-na lsg.poss-nom-rl Chet men lpl.s just men just men kw'ach-nexw-an just look-tr(lc)-lsg.erg lha7n, chet men approach lpl.s just kwum, n-s-na ashore lsg.poss-nom-rl kwetsi mixalh. dem bear nam' ch'imi, n-s-na go close lsg.poss-nom-rl men nam' tl'ich-it-an. just go stalk-tr-lsg.erg '...Then I spotted a bear. We approached and went up close, and then I went ashore and sneaked up on it.' (Kuipers 1967: 240) Familiar referents do not need to be null, however. The speaker can still use a full DP. Later in the story, a full DP kwetsi mixalh 'the bear' is used for the same familiar bear. (19) ...Na7-kw mi ch'it kwetsi mixalh rl-already come close dem bear 'The bear had come close up to us.' Ihe-lha7n-t-umulh-as. redup-approach-tr-lpl.o-Serg (Kuipers 1967: 241) A full DP can even be found in a sentence immediately following one with the same DP. I hem'i. mixalh bear prox come s-e-s men lhxilsh nom-rl-3poss just stand (20) N-s-na men chem'usn-t-an ti lsg.poss-nom-rl just meet-tr-1 sg.erg det 7n-s-na men nam' ch'it, lsg.poss-nom-rl just go be.near kwetsi mixalh lhe-lha7i... dem bear redup-approach 'Then I went to meet the bear that was coming on. I went right up to it, and then the bear stood up on its hind legs and approached...' (Kuipers 1967: 241) It is therefore only a tendency that familiar DPs are null. (See Gerdts and Hukari 2003 for rates of overt DPs in Halkomelem.) Within DPs, the NPs can also be null, as long as there is a demonstrative, adjective, quantifier or numeral. A D-determiner cannot occur without a following NP. (21) a. N-s-tl'i7. kwetsi. lsg.poss-nom-dear dem 'I want that' 22 b. N-s-tl'i7 ta h iy i . lsg.poss-nom-dear det big 'I want the big one.' c. Chen kw'ach-nexw i7xw. Isg.s look-tr(lc) all 'I saw all, everything, everyone.' d. Chen kw'ach-nexw an'us. Isg.s look-tr(lc) two 'I saw (the) two (of them).' e. * Chen kw'ach-nexw ta. Isg.s look-tr(lc) det 5 D-determiners In this section, I provide the necessary background for the behaviour of D-determiners in Skwxwu7mesh: where they occur, their co-occurrence restrictions, and the gender distinctions that they encode. They often behave similarly to those of St'at'imcets (see Matthewson 1998). In the discussion below, I point out where the two languages differ. In this section, I show that D-determiners are obligatory on arguments, D-determiners and demonstratives do not co-occur, and that D-determiners and quantifiers can co-occur. I also show that gender is encoded by the D-determiners. 5.1 Obligatory D-determiners D-determiners are obligatory in argument position in Skwxwu7mesh (unless a quantifier or numeral is present; see §5.3). 23 (22) a. Na wa sik kwi/ta kalaka. rl impf fly det crow (i) 'Crows fly.' (ii) 'The crow is flying.'8 b. * Na wa sik kalaka. rl impf fly crow (23) a. Ha71h-s-t-as kwi/ta swi7ka lha slhanay.' good-caus-tr-3erg det man det.f woman (i) 'Men like women.' (ii) 'A/the man likes a/the woman.' b. * Ha71h-s-t-as swi7ka lha slhanay.' good-caus-tr-3erg man det.f woman (24) a. Ha71h-s chen kwi/ta mixalh. good-caus Isg.s det bear 'I like (the) bear/bears.' b. * Ha71h-s chen mixalh. good-caus Isg.s bear Bare plurals are not licit. (25) a. Ha71h-s chen kwi/ta mex-mixalh. good-caus Isg.s det redup-bear 'I like (the) bears.' b. * Ha71h-s chen mex-mixalh. good-caus Isg.s redup-bear Not even mass nouns (26), proper names (27), or independent pronouns (28) may be bare. (26) a. N-s-tl'i7 kwi/ta stakw. lsg.poss-nom-dear det water 'I want (some)/the water.' b. * N-s-tl'i7 stakw. lsg.poss-nom-dear water 8 The generic reading is only obtained when translating from the English (and is available for both ta or kwi). When the speaker is asked to translate the S k w x w u 7 m e s h back into English, the episodic reading is given. This is true of all examples, regardless of the determiner involved. 24 (27) a. Chen kw'ach-nexw ta Peter.9 Isg.s look-tr(lc) det Peter 'I saw Peter.' b. * Chen kw'ach-nexw Peter. Isg.s look-tr(lc) Peter (28) a. Nilh ta ens.10 foe det lsg.indep 'It's me.' b. * Nilh ens. foe lsg.indep Independent pronouns vary across Salish; in St'at'imcets, for example, the independent pronouns cannot occur with D-determiners (Matthewson, p.c), whereas Upriver Halkomelem independent pronouns do (Wiltschko 2002). D-determiners are not licit in predicate position. (29) a. Slhanay' lha Kirsten. woman det.f Kirsten 'Kirsten is a woman.' b. * .Lha slhanay' lha Kirsten. det.f woman det.f Kirsten D-determiners are only found on arguments. 9 Longobardi (1994) speculated that the lack of determiners on proper nouns in English is what allows them to be "scopeless" and rigidly referring. This can only be the case in a language where determiners are not normally found on proper names, as in English. (i) I would like to meet a Bronwyn some day. (ii) The Tristan I talked to last night was born in Penticton. In Skwxwu7mesh, proper names (introduced by a D-determiner) behave like determinerless proper names in English in that they are rigidly referring. (iii) Haw k-'an i kw'ach-nexw ta Peter. neg irr-lsg.sbj prox look-tr(lc) det Peter T didn't see Peter.' * I didn't see anyone named Peter Instead, to get this interpretation, the proper name must lack a D-determiner. (iv) Haw k-'an i kw'ach-nexw kwi swi7ka s-na-s Peter. neg irr-lsg.sbj prox look-tr(lc) det man nom-call-Zposs Peter T don't know anyone named Peter.' 1 0 Unlike English, where pronouns appear to occupy D (Postal 1969, Longobardi 1994), here the pronoun must occupy NP. 25 5.2 D-determiners and demonstratives do not co-occur Unlike in some Salish languages, in Skwxwu7mesh D-determiners and demonstratives cannot co-occur. (30) a. * Chen kw'ach-nexw kwetsi ta/ti/kwa/kwi mixalh. Isg.s look-tr(lc) dem det bear (I saw that bear) b. * Chen kw'ach-nexw ta/ti/kwa/kwi kwetsi mixalh. Isg.s look-tr(lc) det dem bear c. * Chen kw'ach-nexw tay' ta/ti/kwa/kwi mixalh Isg.s look-tr(lc) dem det bear d. * Chen kw'ach-nexw ta/ti/kwa/kwi tay' mixalh Isg.s look-tr(lc) det dem bear e. * Chen kw'ach-nexw tiwa ta/ti/kwa/kwi mixalh Isg.s look-tr(lc) dem det bear f. * Chen kw'ach-nexw ta/ti/kwa/kwi tiwa mixalh Isg.s look-tr(lc) det dem bear (31) a. Chen kw'ach-nexw kwetsi mixalh. Isg.s look-tr(lc) dem bear 'I saw that bear.' b. Chen kw'ach-nexw tay' mixalh Isg.s look-tr(lc) dem bear 'I saw that bear.' c. Chen kw'ach-nexw tiwa mixalh Isg.s look-tr(lc) dem bear 'I see this bear.' In St'at'imcets, D-determiners and demonstratives can co-occur. (32) Lan-lhkan tu7 wa7 paqw-ens takem iz' i pukw-a. already-lsg.s compl impf look-tr all dem det.pl book-exis 'I already looked at all these books.' (St'at'imcets; Matthewson 1998) 26 5.3 D-determiners and quantifiers or numerals do co-occur D-determiners can co-occur with quantifiers and numerals. Both weak (kex 'many/lots') and strong (i7xw 'all') quantifiers can precede or follow the D-determiner. (33) a. Chen kw'ach-nexw i7xw ta Isg.s look-tr(lc) all det 'I saw all the dogs.' skwem-kwemay'. redup-dog b. Chen kw'ach-nexw ta Isg.s look-tr(lc) det 'I saw all the dogs.' i7xw skwem-kwemay' all redup-dog c. Chen kw'ach-nexw kex ta Isg.s look-tr(lc) many det 'I saw lots of dogs.' skwem-kwemay'. redup-dog d. Chen kw'ach-nexw ta Isg.s look-tr(lc) det 'I saw lots of dogs.' kex skwem-kwemay'. many redup-dog This is different from St'at'imcets, where only the strong quantifiers may precede or follow the 12 D-determiner; post-verbal weak quantifiers must follow the D-determiner. (34) a. it'-em i cw7it-a smulhats. sing-intr pi.det many-det woman 'A lot of women sang.' b. * qwatsats cw7it i sk'wemk'uk'wmi'it-a. leave many pi.det children-exis (Many children left) c. qvlqvl-ts-min-lhkan zi7zeg' ta sqaycw-a bad(redup)mouth-appl-Isg.s each det man-exis 'Each man I saw, I swore at.' ats'x-en-an. see-tr-lsg.conj d. kwan-lhkan ku mule lhel-ti take(tr)-Isg.s det stick from-det 'I took a stick from each of the children. zi7zeg'-a sk'uk'wmit. each-exis child (St'at'imcets; Matthewson 1998) " The order of D-determiner and weak quantifier is variable, but within an elicitation session, the speakers choose one order over another. 1 2 Preverbal weak quantifiers must precede the D-determiner in St'at'imcets (Matthewson 1998). 27 Numerals in Skwxwu7mesh must follow the D-determiner. (35) a. Chen kw'ach-nexw tsi/ta/kwi an'us mixalh Isg.s look-tr(lc) det two bear 'I saw (the) two bears/two of the bears.' b. * Chen kw'ach-nexw an'us tsi/ta/kwi mixalh Isg.s look-tr(lc) two det bear Quantifiers and numerals may occur without the presence of a D-determiner; however, the presence of a D-determiner is strongly preferred. (36) a. Chen kw'ach-nexw i7xw mixalh. Isg.s look-tr(lc) all bear 'I saw all the bears.' b. Chen kw'ach-nexw kex mixalh. Isg.s look-tr(lc) many bear 'I saw many bears.' c. Chen kw'ach-nexw an'us mixalh. Isg.s look-tr(lc) two bear 'I saw (the) two bears/two of the bears.' This is unlike St'at'imcets, where a D-determiner is always required, regardless of whether a quantifier or numeral is present. (37) a. q'welaw'-em takem i syaqts7-a. pick.berries-intr all pi.det woman-det 'All the women pick berries.' b. * q'welaw'-em takem smelhmulhats. pick.berries-intr all woman(redup) c. i cw7aoz-as kw-s cin'-s, ziiqw-as when.past negSsg.conj det-nom long.time-3sg.poss die-3sg.conj i n7an'was-a ucwalmicw wa7 zwat-en-an. pi.det two (human)-det person prog know-tr-lsg.conj 'Not long ago two people died that I knew.' d. * ats'x-en-lhkan n7an'was smulhats. see-tr-lsg.s two(human) woman (I saw two women) (St'at'imcets; Matthewson 1998) 28 When a DP containing a quantifer/numeral occupies a pre-predicate position, the D-determiner is usually dropped. (38) a. An'us slhanay' na ts'its'ap'. two woman rl work 'Two women worked.' b. I7xw slhen-lhanay' na ts'itsap'. all redup-woman rl work 'All the women worked.' c. Kex slhen-lhanay' na ts'its'ap'. many redup-woman rl work 'Many women worked.' One speaker allowed the D-determiner to be dropped even without the presence of other functional material in this fronted position.13 (39) Kalaka wa k'exk'ix, crow impf black 'Crows are black.' I should note that the examples in (36) and (38) are ruled out by Jelinek's (1995) analysis of Salish. Jelinek claims that there are no D-quantifiers in Salish. (Note that this term is completely independent of my use of the term D-determiner.) The distinction between D-quantification and A-quantification is a distinction between quantification over individuals and quantification over events, times, or situations (Partee et al. 1987). The D stands for 'determiner', but D-quantification can refer to any DP-internal quantification (that is, the quantifier does not have to occupy D). The A stands for adverbs, auxiliaries, affixes, and argument-structure adjusters (Partee 1991). An example of each is given below. (40) a. Most birds eat insects. (D-quantification) b. Kitty usually eats cat food. (A-quantification) 1 3 Demirdache et al (1994) argue that Skwxwu7mesh does not allow DPs to occur without the presence of a D-determiner, even when quantifiers or numerals are present, in examples like (36). It may be that we asked different speakers, and each group had different judgments regarding the presence or absence of D-determiners. It is my impression that speakers prefer the D-determiners, but do not require them. 29 In (40) a, the quantifier most ranges over individuals who are birds, and in (40)b, the quantifier usually quantifies over situations where my cat eats. Jelinek (1995) argues that Straits only has A-quantification. According to her, the universal quantifier behaves like an adverbial quantifier in that it can unselectively bind variables throughout a sentence. Jelinek provides an example from Lummi, a dialect of Straits. (41) mekw=0 'ew poq tso speqoq. all=3abs link white det sprout 'They are all/completely white, the flowers.' (Lummi; Jelinek 1995: 514) However, the quantifiers in (42) range over individuals, and not situations. (42) a. Chen kw'ach-nexw i7xw push. Isg.s look-tr(lc) all cat 'I saw all the cats.' (at a particular time) *I always saw cats b, Chen kw'ach-nexw kex push. Isg.s look-tr(lc) many cat 'I saw many (of the) cats.' * I often see cats/a cat. This is also true when the determiner is present. (43) a. Chen kw'ach-nexw i7xw ta push. Isg.s look-tr(lc) all det cat 'I saw all the cats.' (at a particular time) *I always saw cats b, Chen kw'ach-nexw kex ta push. Isg.s look-tr(lc) many det cat 'I saw many (of the) cats.' * I often see cats/a cat. The data in (42) above also appear to be counter-evidence to Matthewson's (1998) claim that there are no Det-quantifiers in Salish. Unlike D-quantifiers, Det-quantifiers must occupy D. However, the data in (42) are only apparent counterexamples because all quantifiers in Skwxwu7mesh can always co-occur with D-determiners (as shown in (43)). Recall that this is not the case for English. In English, some quantifiers cannot co-occur with a D-determiner. (44) a. I saw every cat. 30 b. * I saw the every cat. c. * I saw every the cat. In Salish (including Skwxwu7mesh), there are no quantifiers which cannot co-occur with a D-determiner. Therefore I still propose, in spite of the data in (42), that there are no Det-quantifiers in Skwxwu7mesh. 5.4 Gender The female D-determiners are used when the referents are female humans or animals, as shown in (45)a. Otherwise, gender-neutral determiners are used (Kuipers 1967), (45)b and c. a. Chen kw'ach-nexw lha/tsi/kwelha staw'xwelh Isg.s look-tr(lc) det.f child 'I saw a/the girl.' b. Chen kw'ach-nexw ta/ti/kwa staw'xwelh. Isg.s look-tr(lc) det child 'I saw a/the boy.' c. Chen kw'ach-nexw ta/ti lapat.14 Isg.s look-tr(lc) det cup 'I saw a/the cup.' The female D-determiners are not obligatorily used with female referents, however, as shown in (46). (46) Chen kw'ach-nexw ta/lha slhanay'. Isg.s look-tr(lc) det/det.f woman 'I saw a/the woman.' Gender-neutral D-determiners are licit in any context; female determiners can but do not have to be used for female referents. The D-determiner kwa is more restricted than the other D-determiners. See §7.2.1.1 for more discussion. 31 6 Previous discussions of the determiner system The determiner system of Skwxwu7mesh was first described by Kuipers in his 1967 grammar of the language. I will retain some aspects of his analysis and put it in more formal terms in Chapters 4 and 5. In this section I also discuss Peter Jacobs' analysis of the determiner system as recorded in Currie (1997). 6.1 Kuipers' (1967) description of the Skwxwu7mesh determiner system Kuipers' (1967) original insight (that I will build upon in this thesis) is that proximity and (non-)presence are encoded in the Skwxwu7mesh D-determiner and demonstrative systems.151 give his system in the table below. He divides the system into definite and indefinite forms; the definite forms into present and non-present; and the (non)-present into weak and strong. (All of these terms will be explained below.) Definite Indefinite Present Non-present Weak Strong Weak Strong Proximal Distal plain ta (tl'a)16 ti tay' kwa kwetsi kwi feminine lha (tl'a) tsi alhi kwelha kwelhi kwes Table 2.4: The D-determiner and demonstrative system of Skwxwu7mesh (adapted from Kuipers 1967:137). Kuipers states that "[fjhe definite forms are used for objects which are individually identified for the speaker in an independent way" (1967: 137). That is, the referents are known to the speaker. Some examples of this are given below. In (47)a, for example, the speaker has seen the snake; in (47)b, however, the speaker has not seen any snake, and therefore the "indefinite" D-determiner kwi is used. 1 5 1 continue to make a distinction between D-determiners and demonstratives. 1 6 The determiner tl'a is the oblique version of ta or lha when the NP is a proper name or pronoun (Kuipers 1967). For all other determiners and common nouns, if the DP is marked oblique, the oblique marker t- is added. 32 wa lesiw'ilh impf under (47) a. Yuu chaxw, na take, care 2sg.emph rl kwetsi elhkay'. dem snake 'Careful, there is a snake under the stone.' b. Yuu chaxw, iw'ayti na take, care 2sg.emph maybe rl t-ta smant kwi elhkay'. obl-det stone det snake 'Careful, there may be a snake under the stone.' (48) a. Sat-shit-ka ta stakw. give-appl-imper det water 'Give him the water!' b. Sat-shit-ka kwi stakw. give-appl-imper det water 'Give him (some) water!' t-ta obl-det smant stone wa lesiw'ilh impf under (Kuipers 1967: 138) (Kuipers 1967: 138) As Kuipers himself notes, the "definite forms" are not equivalent to the definite determiner in English (a point that I will discuss further in Chapter 4). Within the category he labels definite, Kuipers makes a distinction between referents which can be pointed out in the speech-situation (present) and referents which cannot be pointed out in the speech-situation (non-present). He also notes that the present form is used when the DP refers to a class of individuals, rather than a particular individual. (49) a. Na wa n-s-7ip'akw'alh ta rl impf Isg.poss-nom-scared det 'I'm afraid of bears.' b. Chen ki-s ta slhem'xw. Isg.s bad-caus det rain 'I dislike rain.' mixalh. bear (Kuipers 1967: 139) Kuipers also notes that the present form can also be used for referents which are absent, especially in texts. That is, ta can be used for referents which are not in the same vicinity as the speaker (e.g. not in the same room). The absent form kwa cannot be used for referents in the same vicinity as the speaker, and can only be used for absent referents. Kuiper claims that the present forms are "unmarked"; the absent "marked". (See §7 for more discussion and data, where I provide an analysis of this phenomenon.) 33 The "strong" and "weak" determiners behave differently syntactically. Only the "strongr determiners (ti/tsi, tay Valhi, and kwelhi/kwetsi) can occur without an NP. 17 (50) a. Chen kw'ach-nexw tay'/kwetsi/alhi/kwelhi/ti mixalh. Isg.s look-tr(lc) dem bear 'I saw that/this bear.' b. Chen kw'ach-nexw tay'/kwetsi/alhi/kwelhi/ti. Isg.s look-tr(lc) dem 'I saw that/this.' The "weak" determiners may not. (51) a. Chen kw'ach-nexw ta/ti/kwa/lha/tsi/kwelha Isg.s look-tr(lc) det 'I saw the bear.' b. * Chen kw'ach-nexw ta/ti/kwa/lha/tsi/kwelha. Isg.s look-tr(lc) det mixalh. bear On the basis of this difference, Kuipers suggest that the "strong" determiners are demonstratives. This is a universal definition of demonstratives which I adopt for the remainder of the thesis. (52) If a deictic determiner can occur without a following NP, it is a demonstrative, and not a D-determiner.18 I continue to refer to the "weak" determiners as D-determiners. According to Kuipers, the D-determiners can be used with unique referents (the sun, for example), including proper names and pronouns.19 This is true. (53) a. c. ta snekwm det sun 'the sun' ta Tarn det Tom 'Tom' b. Na encha kwelha rl where det.f 'Where is your mother?' d. kwa Tarn det Tom 'Tom' 20 chesha7? mother 1 7 As we shall see below, tsi cannot occur without an NP anymore. 1 8 1 assume that all demonstratives have some deictic information. Other determiners may also occur without a following NP, such as quantifiers and pronouns (in English). However; these are unlikely to have deictic information and I do not treat them as demonstratives. 1 9 He does not say whether demonstratives can also be used with unique referents. 2 0 The second person possessive morphology (e-) is often lost, presumably because of the quality of the vowel (schwa). 34 e. ta ens f. ta new det lsg.indep det 2sg.indep T 'you (sg)' g. ta nimalh h. ta new-yap det lpl.indep det 2indep-2pl 'we' 'you (pi)' (Kuipers 1967: 140) Kuipers claims that ta (the present, gender-neutral determiner) can be used for previously mentioned (i.e., familiar) non-unique referents. He further claims that this use of ta is only allowed if the referent has already been previously mentioned using a demonstrative. However, this cannot be correct, as ta can be used for novel referents, as I will show in Chapter 4. Examples of novel ta can also be found in the texts in Kuipers (1967). Within the "present" category of the demonstratives, Kuipers identifies a proximal-distal opposition, but does not discuss which contexts each of ti and tay' can be used in. (54) ti i tay' . dem conj dem 'this one and that one' (Kuipers 1967: 140) He claims that there are also a few independent forms (those that cannot occur with following NPs), which he only briefly discusses. The element -wa is usually added to the demonstrative ti if it occurs without an NP. (55) tay' i tiwa i tsiwa dem conj dem conj dem.f 'that one and this one and this one (f)' (Kuipers 1967: 140) Other elements which Kuipers claims can only be used without NPs are ia-wit, itsi-wti, kwetsi-wit and kwa-wit. I add them to his determiner/demonstrative table, given below. 22 2 1 He does not say explicitly which demonstratives are used in these introductory cases, but I assume he means kwetsi, which is often - though not always - used for novel referents. 2 21 do this because they do behave like the other demonstratives, in that they can occur with an NP. 35 Definite Indefinite Present Non-present D-determiner Demonstrative D-determiner Demonstrative Proximal Distal plain singular ta (tl'a) ti/tiwa tay' kwa kwetsi kwi plural ia-wit itsi-wit kwa-wit kwetsi-wit feminine lha (tl'a) tsi/tsiwa alhi kwelha kwelhi kwes Table 2.5: The D-determiner and demonstrative system of Skwxwu7mesh (adapted from Kuipers 1967:137-143). 6.2 Jacobs' (1997) analysis Another analysis of the determiner system was done by Peter Jacobs. I provide this here to compare with Kuipers' analysis, and for comparison with my own analysis, given in §7. Jacobs re-analyzes the determiner system on the basis of his own fieldwork, as below (given by Currie 1997). Unlike Kuipers, Jacobs treats the demonstratives separately from the D-determiners, because of their different behaviour, shown above in (50) and (51). Here I provide the D-determiner system. Potentially Visible Invisible Visible Non-Visible Proximal Distal non-feminine ti ta kwa kwi feminine tsi lha kwelha kwes Table 2.6: The D-determiner system of Skwxwu7mesh (Currie 1997:31; as suggested by Peter Jacobs). The D-determiners, instead of being split along "defmite'Vindefinite" lines (i.e., whether the speaker knows the referent or not), are split into potentially visible and invisible. A potentially visible referent would be something the speaker may have previously seen. An invisible referent, on the other hand, would not have been seen by the speaker at any time. (I discuss these issues further in Chapter 5.) The potentially visible D-determiners are then further split into visible and non-visible, and the visible into proximal and distal. Jacobs' analysis differs from Kuipers' in another way. Unlike Kuipers, Jacobs treats ti and tsi as D-determiners, rather than demonstratives. This is because tsi cannot occur without a following NP. 36 (56) a. Chen kw'ach-nexw tsi slhanay'. Isg.s look-tr(lc) det.f woman 'I saw a/the woman.' b. * Chen kw'ach-nexw tsi/tsi. Isg.s look-tr(lc) det.f The second reason has to do with the interaction of stress and ti. Ti can be stressed or unstressed. The difference between stressed ti and unstressed ti is audible because the vowel quality changes. In Skwxwu7mesh, what is represented by HI is pronounced [e] in stressed positions (Kuipers 1967, Bar-el and Watt 1998); I'd is pronounced [i] in unstressed positions.23 If ti is pronounced [ti], then it must be unstressed. If ti is pronounced [te], then it stressed. Unstressed ti also behaves like a D-determiner as it cannot occur on its own. (57) a. Chen kw'ach-nexw ti swi7ka. [ti swe?qa] Isg.s look-tr(lc) det man 'I saw a/the/this man.' b. * Chen kw'ach-nexw ti. [ti] Isg.s look-tr(lc) det Stressed // behaves like a demonstrative, as it can occur without a following NP. (58) a. Chen kw'ach-nexw ti swi7ka. [te swe?qa] Isg.s look-tr(lc) dem man T saw a/the/this man.' b. Chen kw'ach-nexw ti.* [te] Isg.s look-tr(lc) dem 'I saw this one.' In the next section, I provide my own descriptions of the Skwxwu7mesh D-determiner and demonstrative systems. I provide more evidence for deictic features, and show that neither Kuipers' nor Jacobs' characterizations capture all of the data. In particular, the obligatory narrow scope of the non-deictic D-determiners cannot be captured by an "indefinite" or "invisible" analysis of kwi. There are rare cases of an unstressed vowel having the stressed vowel quality (Bar-el and Watt 1998). 37 7 De ix is The previous descriptions of the determiner system captured the fact that deictic features, such as presence, or visibility, play a role in Skwxwu7mesh. Here I delve deeper into the deictic features of the determiner system. The term "deixis" can be used to refer to many different notions, including person deixis, space deixis, time deixis, social deixis, etc. (see Fillmore 1997 [1975]; Lyons 1979; Levinson 1983). The common feature in all of these is the notion of distance, anchored to the speech actors, or utterance. This distance can involve distance in time, space, social hierarchies, etc. Here I will be focusing on space and time deixis, as these are the only notions relevant to the determiners of Skwxwu7mesh. Spatial deixis is especially relevant here. Deixis is often assumed to apply only to demonstratives rather than D-determiners, in the nominal domain (see Imai 2003, for example). However, in Skwxwu7mesh, deixis is a feature of both the demonstratives and D-determiners. In this section, I provide evidence that deixis is relevant to both demonstratives and D-determiners. Deictic elements can differ along many different axes. Here I follow Iami (2003) in assuming that there are three parameters: 1) anchor, 2) spatial demarcation, and 3) referent and region configuration.24 1) The anchor can be speaker (typically), hearer, both, or someone or something else. 2) The space can be divided by relative distance (proximal, medial, and distal, for example) or by notions such as up/down, uphill/downriver, north/west/south/east, etc. 3) The configuration of the referent and the region can involve motion, visibility, posture and the overlap between the referent and the region. In gathering most of the data in this section, I placed objects at certain distances away from the consultants. In Figure 1, the Xs mark various distances from the speaker. The rectangle is representative of a room, as that is the size of the area where the elicitation was conducted. The rectangle may be representative of the speaker's visual field; more elicitation outdoors would have to be undertaken to test this hypothesis. 2 4 Imai argues that there are four. I ignore his fourth parameter (function) as it does not seem to be relevant for the Skwxwu7mesh determiner system. 38 X X X object object object Figure 2.1 : Speaker and relative distances from objects I then asked if the particular sentence was felicitous in the context. For each piece of data given in the next few sections, the context is given next to the English gloss. On the basis of the data given below, I argue for the following categorizations of the D-determiner and demonstrative systems in Skwxwu7mesh. Deictic Non-deictic Neutral Proximal Distal, invisible gender-neutral ta ti kwa kwi feminine lha tsi kwelha kwes Table 2.7: The D-determiner system of Skwxwu7mesh. Neutral, invisible Proximal Medial Distal Unmarked Invisible gender-neutral number-neutral kwiya ti, tiwa tay' kwetsi plural kwiyawit iawit itsiwit kwetsiwit kwawit feminine kwsa tsiwa alhi kwelhi Table 2.8: The demonstrative system of Skwxwu7mesh. There are a number of differences between this analysis and the ones provided by Kuipers and Jacobs. First, I do not analyze the D-determiners along present/non-present or potentially visible/invisible lines. Instead, I distinguish non-distal D-determiners from distal D-determiners I also distinguish between 'distal' and 'distal, invisible'. The difference between "neutral" and "medial" features is discussed below. 7.1 Anchor The anchor is the reference point for deictic elements: the base to which referents are related. Crosslinguistically, the anchor for deixis is typically the speaker, although the hearer is the 39 anchor in some languages (Imai 2003). In the next sections, I show that the speaker is the anchor for both the D-determiners and the demonstratives in Skwxwu7mesh. 7.1.1 Anchor for the D-determiners In Skwxwu7mesh, the anchor is the speaker. This can be seen with body parts. The speaker can use either proximal ti or neutral ta to refer to their own body parts, but only neutral ta for someone else's. (See §7.2 for more discussion of the fact that proximal ti and neutral ta can often be used interchangeably.) (59) a. Na mi ptim ti-n s7atsus. rl come swell det-lsg.poss face 'My face is puffy/swollen.' b. Na mi piim ta-n s7atsus. rl come swell det-lsg.poss face 'My face is puffy/swollen.' c. Na mi pum ta e-s7atsus. rl come swell det 2sg.poss-face 'Your face is puffy/swollen.' d. * Na mi pum ti e-s7atsus. rl come swell det Isg.poss-face If the hearer were the anchor, we would expect that one of the D-determiners would only be used for the hearer's body parts (and not for the speaker's). The fact that the speaker is the anchor can also be seen in other contexts. For example, if 25 the referent is closer to the speaker than the hearer, either proximal ti or neutral ta may be used. (60) a. Chen takw-an ta stakw. Isg.s drink-tr det water 'I drank the water.' (water near speaker) b. Chen takw-an ti stakw. Isg.s drink-tr det water (water near speaker) If the referent is closer to the hearer than the speaker, then only the neutral ta can be used. 2 5 These examples do not permit the use of kwa; not all NPs can co-occur with kwa. (See §7.2.1.) It may also be a problem with the choice of example, since the water now occupies the same position as the speaker. 40 (61) a. Chen takw-an ta stakw. Isg.s drink-tr det water 'I drank the water.' (water near hearer) b. * Chen takw-an ti stakw. Isg.s drink-tr det water (water near hearer) Furthermore, if the referent is far from the speaker and the hearer, only neutral ta is licit. (62) a. Chen takw-an ta stakw. Isg.s drink-tr det water 'I drank the water.' (water far from speaker and hearer) b. * Chen takw-an ti stakw. Isg.s drink-tr det water (water far from speaker and hearer) Again, if the hearer were the anchor, we would expect a different D-determiner choice for the context in (61) versus the context in (62). That is, we would expect that at least one of the D-determiners would be used for referents close to the hearer, and that another D-determiner would be used for referents far from the hearer. 7.1.2 Anchor for the demonstratives The anchor for the demonstratives is also the speaker. If the referent is held by the speaker, only the proximal demonstrative ti can be used. (63) a. Chen takw-an ti stakw. Isg.s drink-tr • dem water 'I drank this water.' (near speaker; holding cup) b. * Chen takw-an tay' stakw. Isg.s drink-tr dem water 'I drank that water.' (near speaker; holding cup) c. * Chen takw-an kwetsi stakw. Isg.s drink-tr dem water (near speaker; holding cup) If the referent is within grasping reach, then either proximal ti or medial tay' is licit. The distal demonstrative kwetsi cannot be used. 41 (64) a. Chen takw-an ti stakw. Isg.s drink-tr dem water 'I drank this water.' (near speaker; within reach) b. Chen takw-an tay' stakw. Isg.s drink-tr dem water 'I drank that water.' (near speaker; within reach) c. * Chen takw-an kwetsi stakw. Isg.s drink-tr dem water (near speaker; within reach) If the referent is far from the speaker, regardless of the relative distance to the hearer, then only the distal demonstrative kwetsi is acceptable. (65) a. * Chen takw-an ti stakw. Isg.s drink-tr dem water 'I drank this water.' (far from speaker; near or far from hearer) b. * Chen takw-an tay' stakw. Isg.s drink-tr dem water 'I drank that water.' (far from speaker; near or far from hearer) c. Chen takw-an kwetsi stakw. Isg.s drink-tr dem water 'I drank that water.' (far from speaker; near or far from hearer) Again, if the hearer were the anchor, then we would expect distance from the hearer to affect the choice of demonstrative. 7.2 Spatial demarcation The determiners mark out space by relative distance: proximal, neutral and distal. The choice of a Skwxwu7mesh determiner is directly tied to the distance between the object and the speaker. The examples above have already shown that distance is encoded; however, here I will show it more systematically. I begin with the distal category, the furthest from the speaker. 42 7.2.1 Distal The distal D-determiner and the distal demonstrative behave differently. The behaviour of each is shown below. Neither the distal D-determiner nor the distal demonstrative can be used to refer to objects that are within reach or half-way across a room. An object on the other side of the room (or further) from the speaker is usually considered to be distal, as shown in Figure 2.2. Figure 2.2 : Distal object However, distance is not the only important factor in the determiner system. Invisibility also plays a role, as is shown in § 7 . 3 . 7.2.1.1 The distal D-determiner The distal D-determiner kwa can only be used if the referent is not in the same vicinity (e.g., if the referent is not in the same room) as the speaker. If a person is not in the room, the speaker can choose to use the distal D-determiner kwa. If the person is in the room, kwa cannot be used. (66) a. Kw'ay' kwa Bill. hungry det Bill 'Bill is hungry.' (Bill not in room) b. Kw'ay' ta Bill. hungry det Bill 'Bill is hungry.' (Bill in room) c. * Kw'ay' kwa Bill. hungry det Bill (Bill in room) If someone has arrived from somewhere else, and they wish to name the place, they must use the distal determiner. 43 (67) a. Men yalh s-en mi tl'ik tina7 t-kwa Skwxwu7mesh. just finally nom-lsg.sbj come arrive from obl-det Skwxwu 7mesh 'I just arrived from Squamish.' b. * Men yalh s-en mi tl'ik tina7 tl'a Skwxwu7mesh. just finally nom-lsg.sbj come arrive from obi.det Skwxwu7mesh c. * Men yalh s-en mi tl'ik tina7 t-ti Skwxwu7mesh. just finally nom-lsg.sbj come arrive from obl-det Skwxwu 7mesh Kwa cannot be used for referents that are proximal to the speaker. There is a further complication with kwa. This determiner can only be used if the referent is interesting enough to warrant the use of it. For example, kwa can be used for people and places. However, it can only be used for animals if the particular animal has been made interesting enough. (68) a. * Chen kw'ach-nexw kwa mixalh. Isg.s look-tr(lc) det bear b. Chen kw'ach-nexw kwa mixalh wa an kw'ay'. Isg.s look-tr(lc) det bear impf very hungry 'I saw a bear that was very hungry.' (elicited by Elizabeth Currie) If the animal is not "interesting", the neutral determiner is used instead. (69) Chen kw'ach-nexw ta mixalh. Isg.s look-tr(lc) det bear 'I saw the bear.' (not in room; invisible) What counts as "interesting enough" is unclear at this point. Further research into this behaviour is required. 7.2.1.2 The distal demonstratives There are two distal demonstratives: kwetsi(wit) and kwawit. The distal demonstrative kwetsi behaves very differently from the distal determiner kwa. Similarly to the determiner, kwetsi also cannot be used for referents that are near the speaker. 4 4 (70) *Chen kw'ach-nexw kwetsi swi7ka. Isg.s look-tr(lc) dem man (near speaker) However, the demonstrative kwetsi can be used for referents that are closer to the speaker than the determiner kwa can be. (71) a. * Chen kw'ach-nexw kwa swi7ka. Isg.s look-tr(lc) dem man (in room, far from speaker) b. Chen kw'ach-nexw kwetsi swi7ka. Isg.s look-tr(lc) dem man 'I see that man.' (in room, far from speaker) Kwetsi also cannot be used for place names, unlike kwa. (72) *Men yalh s-en mi tl'ik tina7 t-kwetsi Skwxwu7mesh. just finally nom-lsg.sbj come arrive from obl-dem Skwxwu7mesh Kwawit, on the other hand, is like kwa in that it can only be used for referents that are remote from the speaker. (73) a. Chen kw'ach-nexw kwawit swi7ka. Isg.s look-tr(lc) dem.pl man 'I saw those men.' (far from speaker, not in room) b. * Chen kw'ach-nexw kwawit swi7ka. Isg.s look-tr(lc) dem.pl man (far from speaker, in room) Simply referring to one feature "distal" is not enough to explain the data in Skwxwu7mesh. This will also be discussed in §7.3. 7.2.2 Neutral There are two elements which can be used to refer to entities at any location: the D-determiner ta and the demonstrative kwiya. I call these neutral because they are not used for referents which cannot be located at all. They are only used for referents which can be located or were locatable at some point by the speaker. Neutral D-determiners are therefore still deictic. In Chapter 5,1 discuss the non-deictic D-determiner which can be used for referents that cannot be located. 45 7.2.2.1 The neutral D-determiner The D-determiner ta can be used for (nearly) any referent. If the referent is in the same location as the speaker (near or far), or was at some earlier point visible to the speaker, ta may be used. (74) a. Chen kw'ach-nexw ta Isg.s look-tr(lc) det 'I see the man.' swi7ka. man (man near speaker) b. Chen kw'ach-nexw ta swi7ka. Isg.s look-tr(lc) det man 'I saw the man.' (man no longer near speaker; possibly no longer visible) I therefore treat ta as neutral. This is similar to the locational adverb da 'there' in German, which Imai (2003) claims is neutral, in contrast to proximal hier 'here' and distal dort 'there'. 7.2.2.2 The neutral demonstrative There is a demonstrative that Kuipers did not mention. This demonstrative kwiya may also be used for referents which are close or far from the speaker. (75) Chen tkwaya7n kwiya kw'i7xwm. Isg.s hear dem owl 'I heard an owl.' (near speaker/in middle distance/far from speaker) I also treat this demonstrative as neutral. As I will show, this demonstrative can only be used for invisible referents. See §7.3 for more discussion. 7.2.3 The medial demonstrative Medial objects are usually out of reach from the speaker's grasp, but are not as far away as a distal object. For example, a medial object may be halfway across the room from the speaker. 46 Figure 2.3 : Medial object Unlike the neutral determiner ta, the medial demonstrative tay' is truly medial (i.e. in the middle distance from the speaker; approximately 3 feet away). It is not neutral, and can only be used for referents that are somewhat close to the speaker. (76) a. Chen kw'ach-nexw tay' swi7ka. Isg.s look-tr(lc) dem man 'I see the man.' (halfway across the room) b. * Chen kw'ach-nexw tay' swi7ka. Isg.s look-tr(lc) dem man (across the room) The medial demonstrative tay' cannot be used when the speaker is holding or touching the referent. (77) a. P'ek' tay' lapat. white dem cup 'That cup is white.' b. * P'ek' tay' lapat. white dem cup c. P'ek' ti lapat. white dem cup 'This cup is white.' (within reach) (in hand of speaker) (in hand of speaker, or near speaker) The feature medial must be present in the demonstrative system; however, only neutral is present in the determiner system. 47 7.2.4 Proximal Proximal objects are usually those within reach of the speaker (e.g. within arms-length or closer), or in the hand of the speaker. Figure 2.4 : Proximal object Unlike the distal and medial/neutral categories, the proximal determiner and the proximal demonstrative behave similarly. 7.2.4.1 The proximal D-determiner The proximal D-determiner ti can be used only if the referent is located very close to the speaker. For example, if someone has just arrived somewhere, the proximal D-determiner must be used with the place name. (78) a. c. Men yalh s-en mi tl'ik ti eslha7an. just finally nom-lsg.sbj come arrive det eslha7m 'I just arrived in Eslha7an (a part of North Vancouver).' Men yalh s-en mi tl'ik ta eslha7an. just finally nom-lsg.sbj come arrive det eslha7kn Men yalh s-en mi tl'ik kwa eslha7an. just finally nom-lsg.sbj come arrive det eslha7kn The proximal D-determiner cannot be used if the referent is moderately or very far away from the speaker. (79) Chen kw'ach-nexw ti swi7ka. Isg.s look-tr(lc) det man 'I see the man.' (near speaker) 48 b. * Chen kw'ach-nexw ti swi7ka. Isg.s look-tr(lc) det man (in the middle distance/far away from speaker) 7.2.4.2 The proximal demonstrative The proximal demonstrative must also be used where the referent is very close to the speaker. (80) a. Chen kw'ach-nexw ti(wa) swi7ka. Isg.s look-tr(lc) dem man 'I see this man.' b. * Chen kw'ach-nexw ti(wa) swi7ka. Isg.s look-tr(lc) dem man (near speaker) (in the middle distance/far away from speaker) Both the proximal D-determiner and demonstrative must be used for referents that are close to the speaker. 7.3 Region configuration: (in)visibility In Skwxwu7mesh, there are three elements that must only be used for invisible referents: the distal D-determiner kwa, the neutral demonstrative kwiya(wit) and the distal demonstrative kwawit. Cross-linguistically, distal elements are more likely to also be invisible (Fillmore 1982). 7.3.1 The invisible D-determiner The distal, invisible determiner kwa is only used for invisible referents. (81) a. Chen kw'ach-nexw kwa Peter. Isg.s look-tr(lc) det Peter 'I saw Peter.' (no longer visible, in a different location) b. * Chen kw'ach-nexw kwa Peter. Isg.s look-tr(lc) det Peter (Peter is in room or Peter is still visible in another room) This D-determiner cannot be used for referents which are close to the speaker, even if the referent is invisible. It cannot be simply an invisible D-determiner. 49 (82) a. Na kw'ay' kwa Peter. rl hide det Peter 'Peter is hiding.' (in a different location) b. * Na kw'ay' kwa Peter. rl hide det Peter (in the same room) If the referent is not important enough to use kwa (see §7.2.1), then ta is used instead, even if it invisible and distal. (83) P'ek' ta lapat. white det cup 'The cup is white.' (within reach/in middle distance/far away, not visible) The distal demonstrative, unlike the distal D-determiner, can be used for visible referents. (84) a. Chen kw'ach-nexw kwetsi Peter. Isg.s look-tr(lc) det Peter 'I saw Peter.' (no longer visible) b. Chen kw'ach-nexw kwetsi Peter. Isg.s look-tr(lc) det Peter (Peter is in room or Peter is still visible in another room) The distal feature has different effects in the two systems. I assume that more features are involved: kwa must also have an invisibility feature which the demonstrative kwetsi lacks. 7.3.2 The invisible demonstratives There are two invisible demonstratives: kwiya(wit) and kwawit. The invisible demonstrative kwiya(wit), unlike the invisible D-determiner kwa, is not distal, but instead neutral. (85) a. Chen tkwaya7n kwiya-wit na wa kwikwi. Isg.a hear dem-3pl rl impf talk 'I heard them talking.' (invisible to speaker and very close to speaker/in same room/outside room) b. * Chen tkwaya7n kwiyawit na wa kwikwi. Isg.a hear dem-3pl rl impf talk 'I heard them talking.' (visible to speaker) 50 The invisible demonstrative kwiya can be used for referents in any location, as long as the speaker is unable to see them, and is able to hear them. The distal invisible demonstrative kwawit can only be used if the referent is far from the speaker and invisible. (86) a. Chen tkwaya7n kwawit na wa kwikwi. lsg.a hear dem.pl rl impf talk 'I heard them talking.' (invisible to speaker and outside room) b. * Chen tkwaya7n kwawit na wa kwikwi. lsg.a hear dem.pl rl impf talk 'I heard them talking.' (invisible to speaker and inside room) 7.4 Summary Distal, medial and proximal objects have varying degrees of distance between them and the speaker. X X X distal object medial object proximal object Figure 2.5 : Relative distances between distal, medial and proximal objects Neutral objects can be anywhere in this field, or invisible to the speaker. Invisible objects must be invisible. The theoretical status of all of these features will be discussed in Chapter 4. The non-deictic D-determiner kwi, which I have not discussed here, is analyzed in Chapter 5. Below I repeat the analysis of the D-determiners argued for in this chapter. Deictic Non-deictic Neutral Proximal Distal, invisible gender-neutral ta ti kwa kwi female lha tsi kwelha kwes Table 2.9: The D-determiner system of Skwxwu7mesh. 51 Neutral, invisible Proximal Medial Distal unmarked invisible gender-neutral number-neutral kwiya(wa) ti(wa) tay' kwetsi plural kwiyawit iawit itsiwit kwetsiwit kwawit feminine kwsa(wa) tsiwa alhi kwelhi Table 2.10: The demonstrative system of Skwxwu7mesh. 52 Chapter Three: Theoretical Background 1 Introduction This thesis is primarily about D-determiners:1 their universal function and their language-specific properties. The main claim of this thesis is that D-determiners always introduce domain restriction over their NP, regardless of what other properties they may have. Their function is to constrain the set introduced by the NP to a set of contextually salient individuals. However, they may have additional properties. For example, some D-determiners assert the uniqueness of their referent. D-determiners can also encode deictic information, such as proximity. Deictic information is crucially distinct from domain restriction: deixis provides information about the location of referents, and domain restriction includes only contextually salient individuals, regardless of their physical location. The deictic information interacts with the domain restriction. Both deixis and domain restriction narrow down the domain, but in different ways. In §2 ,1 discuss the status of presuppositions and conversational implicatures. This is important background which is necessary to understand the difference between English and Skwxwu7mesh determiners. In §3,1 provide the background on the notion of definiteness. I will discuss two opposing theories of definiteness: familiarity (anaphoricity) and uniqueness. This is also important background information for understanding the difference between Skwxwu7mesh and English determiners. In §4,1 discuss domain restriction and its place in the semantics of nominals. I provide evidence that domain restriction must be introduced in the functional domain of a DP (cf. von Fintel 1994) and not by a nominal (contra Stanley 2002, Stanley and Szabo 2000). This background is necessary to understand in which ways English and Skxwxu7mesh determiners behave similarly. In §5,1 provide the analysis of the English D-determiner the; this is in preparation for the next chapter, which will provide a different analysis for Skwxwu7mesh D-determiners. I argue 1 Recall that the term D-determiners does not include demonstratives, numerals or quantifiers. 53 that domain restriction and uniqueness interact: in a language where D-determiners assert uniqueness, the domain restriction wi l l force the D P to be familiar; in a language where sentences containing D-determiners only carry an implicature o f uniqueness, the domain restriction w i l l not force familiarity. §6 concludes the chapter. 2 Presuppositions vs. implicatures The available interpretations for any given D P are determined in part by presuppositions or implicatures. Here I provide an overview of presuppositionality and implicatures. 2.1 Presuppositions Presuppositions are pragmatic inferences, distinct from entailments, which can be drawn from the use of sentences. These inferences are tied to particular constructions or lexical items (Levinson 1983). For example, manage foXpresupposes that the agent attempted to X ; ( l )a presupposes (l)b. (1) a. Davor managed to get a job. b. Davor tried to get a job. Presuppositions are taken for granted, rather than entailed. One test for presupposition involves the survival of the presupposition under negation. Entailments, under negation, can be altered; presuppositions usually "survive". 2 ( l )a entails that Davor got a job (2)b, but (2)a does not. However, both ( l )a and (2)a presuppose that Davor tried to get a job (i.e. (l)b). (2) a. Davor didn't manage to get a job. b. Davor got a job. 2 Sometimes these presuppositions do not survive. (i) Davor didn't manage to get a job because he wasn't even trying. I do not discuss these cases in this thesis. 54 In §3 ,1 discuss more cases of presupposition involving D-determiners. See Frege (1997[1892]), Strawson (1998 [1950]), Stalnaker (1973, 1974), Karttunen (1974), Levinson (1983) among many others for more discussion. 2.2 Conversational implicatures Conversational implicatures (commonly referred to as implicatures). are pragmatic inferences that are not tied to any particular construction or lexical item (Levinson 1983).3 Instead, they are drawn from the uses of sentences, on the assumption that speakers observe certain rational principles governing conversation (Grice 1975). Grice specifies four conversational maxims that speakers generally obey to create effective and cooperative communication. (3) a. The maxim of Quality try to make your contribution one that is true, specifically: (i) do not say what you believe to be false (ii) do not say that for which you lack adequate knowledge b. The maxim of Quantity (i) make your contribution as informative as is required for the current purposes of the exchange (ii) do not make your contribution more informative than is required c. The maxim o f Relevance make your contributions relevant d. The maxim of Manner be perspicuous, and specifically: (i) avoid obscurity (ii) avoid ambiguity (iii) be brief (iv) be orderly (from Levinson 1983: 101-102) Implicatures can arise on the basis of these maxims in at least two ways. First, an implicature can arise from the assumption that the speaker observes the maxims. (4) a. I went into a house, implicates: b. The house was not the speaker's house. (by Quantity) 3 However, Chierchia (2001) and Fox (2004) both argue that implicatures are calculated compositionally. 55 (5) a. April likes chocolate, implicates: b. The speaker believes that April likes chocolate. (by Quality) (6) a. ' A: I'm out of gas. * B: There's a gas station down the block. implicates: b. The gas station is or might be open, and has gas to sell. (by Relevance) Secondly, they can arise when the speaker deliberately violates a maxim. Irony is an example of this. (7) A: John didn't come to pick me up as promised today. B: What a great friend you have there. (7) appears to be an obvious case of a violation of the Maxim of Quality (as the speaker cannot possibly believe his or her statement and be sincere about it). The hearer assumes that the speaker is cooperative, and takes the utterance to convey the exact opposite of what it literally says. Another example is given below. (8) War is war. (8) is a tautology and seems to blatantly violate the Maxim of Quantity. Sentences like (8) should not communicate anything to the hearer. However, they do. The Maxim of Quantity requires that speakers be informative; if the speaker says something that appears to be uninformative, the hearer makes an assumption that the speaker is in fact saying something informative. In this case, the sentence means something like 'horrible things always happen in war, and you can't do anything about it'. Implicatures are different from presuppositions in that they are cancelable, or do not even arise when contradicted. B's utterance in (9) does not have the implicature associated with (4)a. (9) A: Did you stay outside all day today as planned? B: No, I had to enter a house because I got so cold. In fact, my house was the closest, so I just went home. Implicatures arise from the maxims, given a certain context. For instance, what counts as "the right amount" of information depends on the context. The context itself can prevent an implicature from arising that would otherwise arise. See Chapter 4 for some more examples of implicatures and cancellation. 56 3 The D P hypothesis In much of the traditional syntactic and semantic literature on English, what has been considered to be a determiner includes the set o f all functional elements that can precede the N P within the nominal domain (excluding adjectives). (10) a. I watched the/a/one/each/every/that swan swim across the lake, b. I watched the/two/those swans swim across the lake. Unt i l Abney (1987), nominals were commonly assumed to be NPs , and to have a noun head. The noun could take a complement and had a specifier position that could host D-determiners, numerals, demonstratives and quantifiers. (11) ^ N P ^ D/Q/Num/Dem N ' N Abney (1987) argues instead that determiners are the head of nominals: NPs are the complement of the head D . (12) D/Q/Dem/Num N P He analyzes all of the pre-NP elements (cardinal numerals, quantifiers, demonstratives, and articles) as occupying the same position: D . 57 (13) D P D N P f a ) I some I i the \ I each I I f/iar J F o l l o w i n g A b n e y (1987), I assume that most nomina ls are D P s , rather than N P s and that D -determiners are the head o f a D P . U n l i k e A b n e y , I do not assume that quant i f iers, demonstrat ives and numerals have the same syntact ic status as D-determiners (see also G ius t i 1993, 1994, 1995). (14) D P D N P 4 Background on definiteness The behav iour o f S k w x w u 7 m e s h D P s provides us w i th ev idence that def ini teness reduces to the interact ion o f (i) assert ion o f uniqueness and (i i) doma in restr ict ion. B o t h must be present in the denotat ion to produce defini teness. I w i l l argue be low that doma in restr ic t ion is a lways present when a D-determiner is present; a l l D P s ( in any language) w i l l therefore i nvo l ve doma in restr ict ion. A s I w i l l show in the next chapter, S k w x w u 7 m e s h D-determiners lack an assert ion o f uniqueness. S k w x w u 7 m e s h D-determiners are used in both nove l and fami l ia r contexts, and therefore the defini teness effects seen in a language l i ke E n g l i s h are m iss ing . T h i s sect ion w i l l address the arguments for dif ferent analyses o f def in i teness. T h e analysis I adopt in this thesis w i l l be g iven in §5. 58 4.1 Definiteness in English: uniqueness or familiarity? While the semantic contribution of definiteness is not agreed upon, most authors seem to agree that definiteness is a primitive of the grammar: a DP is either definite or indefinite. The debate is mainly divided into two camps. Many researchers argue that some form of uniqueness drives the definiteness effects we see (Frege 1997[1892], Russell 1998 [1905], Hawkins 1978, 1991, Abbott 1999, Kadmon 1992, among many others). Others argue that familiarity is encoded by definite DPs (Christophersen 1939, Heim 1988, Prince 1981, Prince 1992, among others). However, there are some who argue that more features are required to describe English (de Jong 1987). De Jong (1987) in particular claims that there are three categories of DPs in English: definite, indefinite, and something in between. Definiteness for her must be decomposed into two features: uniqueness and presupposition of existence. The behaviour of D-determiners in Skwxwu7mesh sheds new light on the English debate. Once we look at a language which lacks familiarity effects, we can see more clearly that definiteness is composed of more than just uniqueness or just familiarity. I will therefore also argue that definiteness is not a primitive of the grammar. However, instead of features, I will appeal to domain restriction in §5. Domain restriction is necessary to describe the Skwxwu7mesh facts, as well as the English. I will decompose definiteness into two parts: (i) domain restriction and (ii) assertion of uniqueness. The familiarity effects will be derived from these two parts. 4.2 Evidence for uniqueness In the philosophical literature, both of the original analyses of definiteness (Frege 1997 [1892] and Russell 1905 [1998]) viewed uniqueness as being relevant to the interpretation of any definite description. In Russell's case, the uniqueness of the referent was asserted, and in Frege's case, it was presupposed (in modern terms). 4.2.1 Assertion versus presupposition of existence In order to understand most of the uniqueness analyses, a digression on presupposition or assertion of existence is necessary. First, presupposition or assertion of existence is not 59 equivalent to familiarity. Familiarity and presupposition of existence both involve knowledge the speaker assumes the hearer has; however, familiarity also is related to the discourse structure. Second, the discussion on existence and uniqueness is usually conflated. In the next three sections, I try to tease these notions apart, so that the analysis given in §6 is clear. The Russellian analysis of the sentence in (15)a is given in (15)b. Both the existence and the uniqueness of the referent are asserted. (15) a. The king of France is wise. b. 3x [king-of-France(x) & Vy [king-of-France(y) —> y=x] & wise (x)] The sentence in (15) asserts all of (16). (16) a. There is a king of France. b. There is not more than one king of France. c. This individual is wise. According to Russell, the falsity of any of (16) entails that (15)a is false. On this view, both the uniqueness and the existence of the king of France are logical entailments of the sentence containing the DP 'the king of France'. Under a Russellian analysis, (17) (the negative counterpart to (15)) should be ambiguous between two readings: one where the entire proposition is negated, and one where the unique king of France exists, but is not wise, as shown in (18). (17) The king of France is not wise. (18) a. -'[Bx [king-of-France(x) & Vy [king-of-France(y) —» y=x] & wise (x)]] b. 3x [king-of-France(x) & Vy [king-of-France(y) —» y=x] & -'[wise (x)]]. However, if there is no king of France, (15) is neither true nor false, according to Strawson (1998 [1950]). That is, even though (16)a is false, (15) is not automatically judged by speakers to be false. It is instead judged to be neither true nor false (see von Fintel 2004). Strawson further claimed that the description 'the king of France' does not assert that there is a king of France, but rather refers to him (see also Searle 1969). Frege's analysis was similar to 60 Strawson's (and over half a century earlier): the existence and uniqueness of the definite is presupposed. As we saw in §2.1, presuppositions "survive" under negation. (19) a. The king of France is wise. true if there is exactly one king of France and he is wise false if there is exactly one king of France and he is not wise truth-valueless if there is not exactly one king of France b. The king of France is not wise. true if there is exactly one king of France and he is not wise false if there is exactly one king of France and he is wise truth-valueless if there is not exactly one king of France Both sentences in (19) can be judged true if there is a king of France. Frege also argues that the sentences are only judged true if there is exactly one king of France. While existence does not appear to be asserted, it is not clear whether uniqueness is also not asserted. For example, if there is no King of France, (17) does appear to be truth-valueless. However, if there are two or more kings of France, (17) seems to have a different status: the hearer wants to force one king to be more salient or prominent than another.4 In the next two sections I discuss analyses which presuppose or assert the uniquness of the referent. 4.2.2 Presupposition of uniqueness Presupposition of uniqueness is argued to be the distinction between a and the by many researchers (Hawkins 1978, 1991; Abbott 1999; Kadmon 1992, 2001, and many others). They argue that a does not presuppose the existence of a unique individual matching the description of the NP, whereas the does. (20) a. The king visited me. b. A king visited me. In (20)a, there is only one king in the context; in (20)b, there can be many different kings. It would be infelicitous to use (20)b in case there is only one king. This effect can also be seen in negative contexts. 4 See von Fintel (2004) for more discussion of judgments of presupposition failures. 61 (21) a. I didn't visit the king. b. I didn't visit a king. In (21)a, there must be a unique king; in (21)b, there does not. In fact, there may not even be any kings. The force of uniqueness that the provides is shown in example (22)a below, where the fact that there could be other alternatives is metalinguistically negated by the speaker using the instead of a (cf. Horn 1985). A strongly implicates the existence of alternatives in (22)b. (22) a. That wasn't a reason I left Pittsburgh, it was the reason. b. He was a friend; I had others. (Abbott 1999) However, Lyons (1999) and Hawkins (1991) both argue that a is not strictly non-unique. When it is used, the speaker often implicates that there may be more than one. The fact that non-uniqueness is not required is shown in examples like the one below. (23) I saw a man yesterday. At the beginning of a conversation, the man is being introduced. However, there may be no other men relevant to the discourse. He may be, for the purposes of this conversation, unique. But a is usually used to introduce a new referent, as we will see in the next section. An interesting case is one where there is only one possible referent. Here, the use of the indefinite article is infelicitous. (24) a. # A sun appeared. b. The sun appeared. Hawkins (1991) argues that the presupposes a unique referent; the use of a only implicates that there is more than one entity satisfying the description of the NP in the context. As Heim (1991) notes, sentences like (24)a are not predicted to be bad, since a does not presuppose non-uniqueness, and in fact cannot. This can be seen in examples like (25) below. (25) A pathologically curious neighbor of mine broke into the attic. (Heim 1991) 62 If a presupposed the non-uniqueness of the DP, this sentence should presuppose that I have at least two pathologically curious neighbours. However, (25) does not presuppose this. Heim argues that we are forced to make an extra assumption for why a is bad in situations where the referent is known to be unique, like the sun, at least on our Earth. She posits a maxim: "Make your contribution presuppose as much as possible!" If the hearer knows that the speaker has reason to presuppose the uniqueness of the referent, the speaker cannot use a form which does not presuppose the uniqueness. Heim formulates this need to use the where the presuppositions are satisfied, as in (26). (26) In utterance situations where the presupposition for [the Q 4 is already known to be satisfied, it is not permitted to utter [a Q (Heim 1991) The use of a implicates that the speaker is not able to use the, because s/he does not know if there is exactly one referent. In a case like (24)a, both speaker and hearer know that there is exactly one referent, and this knowledge conflicts with the implicature that the speaker does not have enough information to use the. If the speaker does not obey the maxim, the hearer will expect there to be another referent. If the hearer does not know there is exactly one referent, as in (25), the use of a is licit. So far we have only discussed singular definites. The uniqueness analysis for singular definites can be extended to plural and mass definite DPs. There are a number of names associated with 'plural uniqueness': the maximal set (Sells 1985) and inclusiveness (Hawkins 1978) among them. I will refer to uniqueness and maximality (following Kadmon 1992) for the singular and plural instances for the remainder of this thesis. Uniqueness and maximality are essentially the same thing: the maximal individual is also the unique member that includes the join of all the atoms (the supremum - Link 1983), and the unique individual is also the maximal individual, which happens to be an atom. That maximality is relevant to the in plural and mass noun cases can be seen below. (27) a. Yesterday a bunch of children were playing in the yard. I saw the children again today. b. A: I bought some milk today. I don't want it to go bad. Did you put away the groceries? B: Most of them, but I drank the milk. 63 The set of children in the second sentence of (27)a is the set of all o f the previously mentioned children. In (27)b, speaker B must have drunk all of the milk that was purchased today. 4.2.3 Assertion of uniqueness Not all researchers argue that uniqueness is presupposed. Instead, they claim that uniqueness is asserted. For example, L ink (1983) and von Fintel and Heim (2001) both posit analyses which assert the uniqueness of the referent. L ink ' s (1983) definition of the (which is basically Russellian) is given below in (28). The asserts the both the uniqueness and the existence of the referent. (28) the = XQ XP 3y [Q (y ) A V X [ Q ( X ) - x l l y ] A P(y)] His analysis captures both uniqueness and maximality: the definition of the allows for singular, plural, and mass nouns. He claims that the nouns themselves provide the singular, plural, or mass interpretation of the DP . The part of the formula Vx [Q (x ) - xny] is the assertion of uniqueness/maximality of the referent. The predicate n is the individual part relation (or i-part); xny means that x is an i-part ofy. This i-part x o f y cannot be null . In a singular case (as in (29)), y is an atom because singular predicates like child only have atoms in their denotation, x must therefore be equal to y. This is equivalent to the i-operator, which demands that the referent be unique. (29) The child = XP 3y[child'(y) A Vx[child '(x) - xny] A P(y)] XP 3y[child'(y) A Vx[child '(x) - x = y] A P(y)] AP 3y[y = ix child'(x) A P(y)] In a plural case (as in (30)), y cannot refer to an atom, because plural predicates like children only have sums of atoms in their denotations. Instead, y refers to the maximal individual sum of all the atoms in the predicate. (30) The children = AP 3y[®child'(y) A Vx[®child'(x) - xny] A P(y)] AP 3y[y = ix fchild'fx) A Vz[*child'(z) - znx] ] A P(y)] 64 In the case of a mass noun (as in (31)), y cannot refer to an atom because mass predicates like water do not have atoms in their denotations.5 (31) The water = XP 3y[water'(y) A Vx[water'(x) - xITy] A P(y)] In all cases, the DP will refer to the supremum (the unique atom, the plural object consisting of all atoms, or the entire mass) of the set denoted by the NP. von Fintel and Heim (2001) also appeal to an analysis which asserts the uniqueness of the referent. Their mechanism differs from Link's in that it presupposes existence, rather than asserting it. Their analysis is similar in that it can handle singular, plural, or mass nouns. In the formula in (32)a below, anything before the period is presupposed, and anything following is asserted (following the notation in Heim and Kratzer 1998). (32) a [the]] = XP: 3x e P(x) = 1. max(P) b. max(P) := the unique x such that P(x) = 1 & Vy[P(y) = 1 —» y < x] Max(P) is the maximal individual (i.e. the supremum) that P is true of; it is undefined if there is no unique individual. I provide their analysis here, because I build upon their idea in §5. 4 .3 Evidence for familiarity The competing theory of definiteness is that of familiarity (Christophersen 1939, Karttunen 1976, Grosz 1977, Clark and Haviland 1977, Clark and Marshall 1981, Heim 1988, Prince 1981 and 1992). Heim (1988) argues that the and a in English track discourse referents by marking a distinction between novel and familiar referents. Under a familiarity analysis, definites can only be used when both the speaker and hearer are familiar with the referent and indefinites can only be used when the hearer is not familiar with the referent. There are two ways that a referent can be familiar to the hearer: the referent can either be discourse-old, in which case the referent has an antecedent in the discourse, or hearer-old, in which case the referent is part of the shared knowledge of the world (Prince 1992). 51 am glossing over some of the finer details of how mass predicates work. See Link (1983) for more details. 65 (33) A: I saw a cat lurking around my garden last night. (discourse-new) B: Where is the cat now? (discourse-old) (34) I saw the moon last night. (discourse-new, hearer-old) If the DP does not have an antecedent in the discourse, and is not part of the shared knowledge of the world, the nominal must be indefinite. (35) #1 saw the cat lurking around my garden last night. (discourse-/hearer-new) If the referent does have an antecedent, the nominal must be definite. (36) A: I saw a cat lurking around my garden last night. (discourse-new) B: #Where is a cat now? (discourse-old) However, there are exceptions to the claim that definites must always be familiar (that is, not every definite has a referent that is discourse- or hearer-old). (37) Watch out, the dog will bite you. (Heim 1988) The sentence in (37) can be used in a context where there was no previous mention of a dog, even if the dog is not in sight, or the hearer does not know that the dog exists. Heim argues that in this case, the hearer accommodates the presupposition of familiarity (following work by Lewis 1979). The speaker can assume that the hearer will be able to accommodate the new information provided by 'the dog'. The definition of accommodation is given below. (38) Accommodation: if at time t something is said that requires presupposition p to be acceptable, and if p is not presupposed just before t then - ceteris paribus - presupposition p comes into existence. (Lewis 1979: 172) Accommodation obviously does not happen in all cases, or the speaker should be able to use (35) out of the blue. It can be accommodated if the hearer has reason to believe the speaker has a cat 66 (that perhaps ran away). However, i f the hearer knows of no cat that could be part of the discourse context, the hearer has a right to ask "which cat?"6 4.4 Deriving uniqueness from familiarity Each hypothesis captures some intuition about how the is used by speakers. However, it is not immediately clear how either of them captures the intuition that is central to the opposing hypothesis. The familiarity hypothesis must explain why there is an intuition that the is associated with uniqueness, especially in novel contexts. In (39) below, the hammer must refer to the only contextually relevant hammer; the only hammer within reach, for example. This sentence can be used in novel contexts. (39) Pass me the hammer. Heim (1988) argues that this is not a systematic effect. Instead, she claims that there is only an intuition that definites are typically unique. She derives this intuition from her familiarity theory and Gricean constraints. In Heim's system, definites must be familiar. A definite must be associated with a discourse referent which is already present in the representation of the discourse before it can be processed. The introduction of a discourse referent for the definite can be triggered by a DP in the previous discourse, by something salient in the context (as in (39)), or even by accommodation (see previous section). A speaker can only use a familiar DP if it is clear which discourse referent is intended to be its referent. If there are equally likely candidates, the hearer is faced with ambiguity. This is a violation of Grice's maxim of manner. (40) Maxim of Manner: 1. Avoid obscurity of expression. 2. Avoid ambiguity. 3. Be brief (avoid unnecessary wordiness). 4. Be orderly. This can be seen in an example like (41). 6 This is essentially the 'wait-a-minute' test proposed by von Fintel (2004). 67 (41) John has a cat and a dog. The cat's/*the pet's name is Felix. (Heim 1988: 384-5) The definite the cat is preceded by two indefinite nominals whose discourse referents could serve as its antecedent. The cat is appropriate because only one discourse referent fits its descriptive content. The pet, on the other hand, is inappropriate because there are two equally likely antecedents, and the hearer will be unable to resolve the ambiguity. In this system, the cat does not have to refer to the only cat John has, or the only salient one in the context. It just has to be linked with the discourse referent of a cat. The uniqueness effect is therefore indirect; uniqueness is not part of the meaning of the. Instead, the uniqueness is forced by the requirement to avoid ambiguity when locating a discourse referent that satisifies the familiarity requirement imposed by the. Uniquely identifying descriptions, which are hearer- and discourse-new, pose a larger problem for the familiarity hypothesis. In (42)a, the hearer need not know that Mary had bought a car, and in (42)b, the referent does not exist yet. In neither case is the referent familiar. Yet both of these cases are felicitous. (42) a. Mary's just gone for a spin in the new car she bought. (Lyons 1999) b. I will build the first space ship to take humans to Mars. Similar examples are given below. (43) a. The only whale in this ocean/the moon/the prime minister b. That whale underneath us These are all uniquely referring expressions. In order to make these examples fit a familiarity analysis, accommodation of the referents is necessary.7 Familiarity alone cannot account for these facts. 4.5 Deriving familiarity from uniqueness On the other hand, the uniqueness hypothesis must be able to explain the intuition that the is usually used when the referent is familiar, such as in (44) below. 7 (42)b is also a problem for analyses which presuppose or assert the existence of the referent. 68 (44) Mary had a cat on her lap. She was petting the cat, and it was purring. Kadmon (1992), who works in the Discourse Representation Theory (DRT) of Kamp (1981) and Heim (1988), argues that definites are obligatorily unique, and that they presuppose the existence of their referent. She also claims that it is possible to derive the "anaphoric nature" of definites from their uniqueness. According to her, definite DPs do not have to presuppose familiarity. Kadmon argues that if you assign a new variable to a definite DP (that is, you interpret it as a novel discourse referent), it necessarily violates uniqueness. This is because the variable that has been introduced is brand new and there is nothing predicated of it, so there is no way to guarantee the uniqueness of the value of the variable. She does not provide examples of how this would work; nevertheless, I adapt the idea in §6 below. 4.6 Summary Both uniqueness and familiarity seem to be relevant to definites in English. It is difficult to describe all properties of the in terms of one or the other. However, in the next section, I will show that D-determiners introduce domain restriction over their NP. This is shown to be the missing link: I show that domain restriction is necessary to account for the behaviour of English the and that familiarity is a byproduct of the interaction between uniqueness and domain restriction. 5 Domain restriction In this section, I provide background on domain restriction, as well as some argumentation for the position of domain restriction. Particularly, I argue that domain restriction must be provided by some functional superstructure of the NP. 69 5.1 Why domain restriction? It has been argued that DPs are sensitive to the context in which they are uttered (Westerstahl 1984; von Fintel 1994, 1998, 1999; Marti 2003, among others). This is because DPs (usually) cannot refer to all individuals in the world that match the NP description. For example, in (45 )a, the men does not (normally) refer to all men in the world. Instead, it refers to the set of contextually salient men. Similarly, in (45)b, the man cannot refer to the only man in the world; Q it can only refer to a man who is unique in the context. (45) a. The men were laughing, b. The man was laughing. Westerstahl (1984) claims that the is itself domain restriction, and nothing more. I will not adopt this, as uniqueness also plays a role. 5.2 Quantifiers and domain restriction According to some, quantifiers introduce unpronounced domain restriction variables ranging over properties of individuals (Westerstahl 1984; von Fintel 1994, 1998, 1999; Marti 2003).9 von Fintel claims that strong quantifiers restrict the domain of the NP that is quantified over. In this way, strong quantifiers are context-dependent. (46) The dinner guests had rhubarb pie for dessert. Everyone developed a rash. (von Fintel 1998:2) In the example above, everyone does not quantify over all the individuals in the world; in fact, it cannot quantify over all the individuals in the world. Instead, it is restricted to the dinner guests who had rhubarb pie for dessert. Formally, the domain of the quantifier is restricted to those dinner guests by an unpronounced element (C) that is introduced by the quantifier. In the example below, the domain of the quantifier every is restricted to the freshmen in the context. 8 Attempts to make uniqueness more 'realistic' (see Kadmon 1992) involved contextual dependence. 9 1 claim that D-determiners are (at least in some languages) the pronunciation of this domain restriction. 70 (47) Every freshman is from out of state, every [C & freshman] [out of state] every Xx [C(x) & freshman (x)] [Xx [out of state(x)]] (von Fintel 1999:3) This unpronounced element C is of type and is interpreted via intersective predicate modification with the NP predicate (which is also of type ). C is the characteristic function of the set of individuals that are under discussion: in this context, this set might include all the participants in the relevant undergraduate semantics class. 5 . 3 Bare nouns and (the lack of) domain restriction In general, DPs appear to be associated with domain restriction. However, there is a debate over where the domain restriction appears. Stanley and Szabo (2000) have argued that nouns are associated with domain restriction. It follows from their analysis that bare nouns are also associated with domain restriction. Here I show that this cannot be correct. If nouns themselves were to introduce domain restriction, we would expect bare nouns to also introduce domain restriction. However, bare nouns do not seem to show the same sensitivity to the context as other nominals do. In the following example, the bare noun bears does not refer back to the set introduced by some bears. In the generic case in (48)a, bears must refer to all the bears in the world. In (48)b and c, bears must introduce a new group of bears, which sounds strange following a discussion of the first group of bears without some notification of the change in topic. (48) a. I saw some bears last night. They were wandering around Stanley Park. Bears like to hang around the park. b. I saw some bears last night. They were wandering around Stanley Park. # I shot bears. c. I saw some bears last night. They were wandering around Stanley Park. # Bears were eating garbage. If I want to refer back to the original set of bears, I must use a D-determiner or demonstrative, as in (49). 71 (49) a. I saw some bears last night. They were wandering around Stanley Park. The/those bears like to hang around the park. b. I saw some bears last night. They were wandering around Stanley Park. I shot the/those bears. c. I saw some bears last night. They were wandering around Stanley Park. The/those bears were eating garbage. If I want to introduce a new set of bears, I must notify the hearer by using a partitive. (50) a. I saw some bears last night. They were wandering around Stanley Park. I shot some other bears. b. I saw some bears last night. They were wandering around Stanley Park. Some other bears were eating garbage. Breheny (2003) also argues on independent grounds that nouns cannot introduce domain restriction. (51) Every fake philosopher is from Idaho. (Kratzer 2004; ascribed to Breheny 2003) Let the domain for the DP every fake philosopher be the set of Americans. The sentence in (51) may only get the interpretation hi (52)a. However, if the domain restriction is associated with the noun itself, the sentence should get the interpretation in (52)b. This is an impossible interpretation. (52) a. Every American fake philosopher is from Idaho. b. Every fake American philosopher is from Idaho. Stanley and Szabo's (2000) analysis cannot be correct. The contextual restriction must be introduced by some higher functional projection than the NP. I will argue in the next chapter that this is D, at least in Skwxwu7mesh.10 Kratzer (2004) argued that quantifiers could not be associated with domain restriction since languages never appear to have overt domain restriction. However in Skwxwu7mesh the determiners are overt domain restrictors. 72 6 Fo reshadow ing the analysis In order to understand the analysis of Skwxwu7mesh D-determiners in Chapters 4 and 5, it is necessary to understand domain restriction and modes of composition. First I describe how domain restriction works in English, in order to compare to Skwxwu7mesh in the next two chapters. Secondly, I provide an overview of Chung and Ladusaw's (2004) two modes of composition, Specify and Restrict because I analyze deictic and non-deictic DPs as composing via Specify and Restrict, respectively. 6.1 Deriving familiarity I have shown above that bare nouns cannot introduce domain restriction. In the next chapter, I will argue that only D-determiners can introduce domain restriction. I argue here that English DPs assert the uniqueness/maximality of their referent. This assertion interacts with domain restriction to create the familiarity effects we see in English. I show in Chapters 4 and 5 that Skwxwu7mesh D-determiners are associated with domain restriction. If we assume the same for English, plus assertion of uniqueness, the familiarity effects can be accounted for. It is difficult to decide, on English-internal grounds, which analysis works best for definiteness. However, the mechanisms discussed above cannot be extended to Skwxwu7mesh and are therefore not universally valid. Familiarity effects are not found in Skwxwu7mesh, and so any analysis of D-determiners which crucially rests on familiarity will not be extendable to Skwxwu7mesh. Presupposition of existence and presupposition or assertion of uniqueness are also not found in Skwxwu7mesh, so any analysis which only rests on these effects will also not be extendable to Skwxwu7mesh. However, the analysis in this thesis, that all D-determiners are associated with domain restriction, is potentially universally valid. I argue that the familiarity effects in English arise from domain restriction and the assertion of uniqueness. If a DP must be unique, as with English definite DPs, then the referent will be restricted to the intersection of the domain restriction and the set denoted by the NP. I argue that if a D-determiner asserts uniqueness, the DP must refer to the intersection of C and the NP. This is similar to Kadmon's (1992) analysis of definiteness. She argues that if you use a definite DP in a novel context, the DP has to be assigned a new variable. Since this variable is 73 brand new, there is no w a y to guarantee that it is unique. Rather than appeal ing to a D R T representation, as K a d m o n does, I argue that the definite D-determiner the has d o m a i n restrict ion i n its representat ion;" this d o m a i n restrict ion must contain the unique element that matches the descriptive content o f the N P . I f it does not contain a unique element that matches the 12 descript ion, the D P is infel ic i tous. T h i s is because the d o m a i n restr ict ion must contain a l l o f D e . U n t i l the context has been narrowed, C must contain the entire set o f i n d i v i d u a l s i n the w o r l d . There can be no unique i n d i v i d u a l that satisfies the N P descr ipt ion. I a m adapting the f o r m u l a g i v e n by v o n F i n t e l and H e i m (2001) b y a d d i n g d o m a i n restrict ion ( C ) to the representation. I do not assume that the presupposes existence; I o n l y adopt the assertion .of uniqueness. (53) [[the] = XP max(A.x [P(x) A C ( X ) ] ) I do not adopt the presupposit ional part o f their analysis because once w e adopt d o m a i n restrict ion, presupposit ion o f uniqueness is redundant. C is inherently presupposi t ional ; it is a free variable, w h i c h is l i n k e d to the context. In what f o l l o w s , I w i l l consider a number o f different cases: n o v e l examples o f s ingular and p lura l definite D P s , examples w i t h s ingular and p lura l definite D P s where C contains one i n d i v i d u a l , and examples w i t h singular and p l u r a l definite D P s where C contains more than one i n d i v i d u a l . I begin w i t h a n o v e l use o f a s ingular D P . H e r e , the cannot be used. I assume that the d o m a i n restrict ion includes the entire d o m a i n o f entities ( D e ) , because the d o m a i n has not been narrowed b y anything i n the discourse. (54) a. # I saw the bear. (novel) Qhebear = D e 1 3 b. [ the b e a r ] = max(A,x [bear'(x) A C ( X ) ] ) = 0 ' ' Unlike Marti (2003), I do not argue that the domain restrictor occupies a separate syntactic node. 1 21 assume that it includes all of D e and not, say, all entities that exist right now, because it is always possible to talk about deceased entities. (i) The cat liked to walk around. (now deceased cat) Nothing in the sentence gives us the information that the cat is no longer alive; the only way that the DP could refer to the right cat is if C included deceased entities. 1 3 Westerstahl (1984) argues that domain restriction must be different for each DP; for the sake of simplicity I am providing the domain restriction for the relevant DP. 74 Because C contains all bears in the domain D e , the intersection of bear and C contains the same individuals as bear. There is no maximal individual that belongs to both bear and C . Plural definites are slightly different. The cannot be used in a novel context for plural DPs either, but the result is different. ( 5 5 ) a. # I saw the bears. (novel)) Qhe bears = D e b. [the bears]] = max(Xx [®bear'(x) A C(X)]) = D e Here, the context set again contains all bears in the domain D e ; the intersection of bears and C is the sum of all bears. The sentence / saw the bears then can only be true if I saw all of the bears in the world, which is extremely unlikely. People do not normally have the opportunity to see all the bears in the world, especially at one time. Pragmatically, hearers know that the domain should be narrowed, but without any other information, they do not know how to narrow the domain. In cases where the domain includes one bear, the DP will refer to that bear. The intersection of C and the set provided by bear is the bear in the domain. ( 5 6 ) a. I saw the bear. C t h e bear = {bear;} b. [the bear]] = max(Xx [bear'(x) A C(X)]) = bean If the DP is plural, but the domain only includes one bear, the DP cannot refer to that bear. This is because the predicate %ear only provides individual sums of members of bear. There are no atomic individuals in @bear. There are also no individual sums in C . The intersection of C and %ear is null. ( 5 7 ) a. # I saw the bears. C t h e bear = {bear;} b. [the bear]] = max(Xx [®bear'(x) A C ( X ) ] ) = 0 In cases where the domain includes more than one bear, a singular DP cannot be used. This is because the predicate bear only includes atomic individuals. There is no maximal individual in the intersection of C and bear. ( 5 8 ) a. # I saw the bear. Qhe bear = {bean, bearj, beark} 7 5 b. [[the bear] = max(U [bear'(x) A C ( X ) ] ) = 0 If the DP is plural, and the domain includes more than one bear, the DP will be felicitous. This is because the intersection of C and ®bear will be individual sums of the predicate bear. Max will choose the maximal individual of that set. (59) a. I saw the bears. C t h e bears = {beari, bearj, beark} b. [[the bear] = max(A,x [®bear'(x) A C(X)]) = bearj+bearj+beark If the hearer is given enough information to decide that the referent is unique, it is no longer necessary that the referent be familiar (cf. Hawkins 1991, Kadmon 1992). (60) a. Mary went out with the man she met yesterday. C ( h e man = {Mary} b. [[the man she met yesterday] = max(A,x [man-she-met-yesterday'(x) A C ( X ) ] ) = mani C t h e man = {Mary, man;} Hearers can narrow the domain C; but they can only do so if they have enough information to do so. Under most circumstances, they will not be able to tell how to narrow the domain enough for the DP refer to a unique individual. They will not normally accept a definite DP in a novel context, because they feel uncertain as to the contextual domain. The familiarity effects seen in English derive from domain restriction and the assertion of uniqueness. It is therefore possible that only one feature of the (domain restriction or uniqueness) is relevant to other languages. I address such a language in Chapter 4.1 show that Skwxwu7mesh D-determiners are associated with domain restriction, but do not assert the uniqueness of their referent. Under the analysis provided here, the fact that definites are (usually) used in familiar contexts is no longer part of the lexical entry of the. Instead, it falls out from the fact that the provides domain restriction over its NP and that it asserts the uniqueness of its referent. The domain C must intersect with the set of the NP. The lexical entry for the must include assertion of uniqueness, since any definite DP refers to the unique individual/maximal set matching the description denoted by the NP. 76 6.2 Specify and Restrict Chung and Ladusaw (2004) argue that there are two modes of composition for indefinites: Specify and Restrict. My analysis of Skwxwu7mesh DPs draws upon their analysis of Maori indefinites. 6.2.1 Specify The first mode of composition that I discuss is Specify. Specify is essentially another term for choice function. It type-shifts the property denoted by the NP to an individual, where the individual is the output of a choice function (Chung and Ladusaw 2004; cf. Reinhart 1997, Winter 1997, Kratzer 1998 and Matthewson 1999, among others). The function variable assigns an individual to the property supplied by the NP. The individual saturates the argument of the predicate. (61) a. A dog barked. b. EC {Xx [bark'(x)], CF([dog'(y)])) = 3f[bark'(f(dog'))] Chung and Ladusaw argue that existential closure of the choice function can apply at any point in the derivation.14 This allows an indefinite to take any scope with respect to an operator. (62) a. A dog didn't bark. b. 3f^[bark'(f(dog'))] (wide scope) c. —i 3f [bark'(f(dog'))] (narrow scope) The existential closure only applies to save the structure. The existential closure takes place at any point in the derivation (above or below negation), in order to provide closure over the variable over choice functions. I analyze the Skwxwu7mesh deictic DPs as composing with the predicate via Restrict. I do this because the deictic DPs are able to escape the scope of negation, but do not necessarily take wide scope. 1 4 As they note, for some languages, the existential closure must take place at the highest point. See, for example, Matthewson (1999) for arguments that Stat'imcets DPs are closed off at the highest point. 77 (63) Haw k-'an i yelx-t ta swi7ka. neg irr-lsg.sbj prox find-tr det man 'I didn't find a man.' 3f^[find'(f(*man'))(I)] - i 3 f [find'(f(*man'))(I)] (wide scope) (narrow scope) Further, deictic DPs can escape islands. (64) i7xw ta nexw7usialh wa7 ek' seselkw all det teacher impf fut sad [u if k irr huya7-as leaveSsbj ta s7ixwelh]. det child 'All the teachers will be sad if a child leaves.' 3x [*child'(x) A [leave'(x) -> Vy [*teacher'(y) A sad'(y)]]] (wide scope) The DP therefore cannot be undergoing QR (see Fodor and Sag 1982 and Ruys 1992). 6.2.2 Restrict Restrict differs from Specify in that it does not saturate the argument position of the predicate. If an argument is composed by Restrict, it is interpreted differently. "In this mode, the property argument is interpreted as a restrictive modifier of the predicate" (Chung and Ladusaw 2004: 6). The domain of the predicate is thereby restricted to elements that have the property introduced by the object. Restrict does not change the type of the predicate. The verb feed is of type < e , < e , t » , and the type of feed plus a Restrict DP is still < e , < e , t » . The internal argument of the predicate must still be saturated via some other process; they do this by appealing to existential closure or by function application of another argument. Chung and Ladusaw also assume that once an argument has been targeted by Restrict, it can be demoted. This has the effect of "flipping" the order of arguments, as in (66)a. The argument does not have to be demoted, however. In (66)b, the object argument is saturated by the DP Fido before the subject argument is saturated. (65) Restrict (Xy Xx [feed'(y)(x)], dog') Xy Xx [feed'(y)(x) A dog'(y)] (Chung & Ladusaw 2004: 5) 78 (66) a. Restrict (Xy Xx [feed'(y)(x)], dog') Xx Xy [feed'(y)(x) A dog'(y)] F A (Xx Xy [feed'(y)(x) A dog'(y)], John) Xy [feed'(y)(John) A dog'(y)] E C Xy [feed'(y)(John) A dog'(y)]) = 3y [feed'(y)(John) A dog'(y)] "John dog-fed." b. Restrict (Xy Xx [feed'(y)(x)], dog') Xy Xx [feed'(y)(x) A dog'(y)] FA (Xy Xx [feed'(y)(x) A dog'(y)], Fido) Xx [feed'(Fido)(x) A dog'(Fido)] F A (Xx [feed'(Fido)(x) A dog'(Fido)], John) feed'(Fido)(John)) A dog'(Fido)] "John dog-fed Fido." They argue that existential closure can take place at any point before the event argument is closed off (the VP level). This is different from function application, where the argument saturates the argument position of the predicate. (67) F A (Xy Xx [feed'(y)(x)], Fido) Xx [feed'(Fido)(x)] It is also different from Specify. (68) Specify (Xy Xx [feed'(y)(x)], CF ([dog'(y)])) Xx [feed'(f(dog')(x)] F A (Xx [feed'(f(dog')(x)], John) [feed'(f(dog'))(John)] EC ([feed'(f(dog'))(John)] 3f[feed'(f(dog'))(John)] I analyze the non-deictic DPs as composing via Restrict in Chapter 5.1 do this because the non-deictic DPs take obligatory narrow scope. 6.3 The Skwxwu7mesh D-determiners In the next two chapters, I analyze deictic DPs as composing via Specify, and non-deictic DPs as composing via Restrict. In terms of composition, Skwxwu7mesh DPs behave quite differently 79 from English definite DPs. However, like the English D-determiner the, I analyze all the D -determiners in Skwxwu7mesh as having domain restriction in their representations. ( 6 9 ) a [the]] = AP max(Xx [P(x) A C (X)]) b. [ta_ = AP f(A.x [P(x) A C(x)]) c. [kwi]] = AP A.x [P(x) A C(x)] The fact that none of the D-determiners in Skwxwu7mesh assert the uniqueness of their referent allows them to be used in both novel and familar contexts. 7 Conclusion In this chapter, I have provided some of the background necessary for understanding the analysis of the Skwxwu7mesh D-determiners, which wi l l be given in the next two chapters. I have also provided an analysis of the English D-determiner the. This analysis shares something in common with the analysis for Skwxwu7mesh: domain restriction. 80 Chapter Four: The deictic D-determiners 1 Introduction In the last chapter, I provided an analysis for English the. I argued that the asserts the uniqueness of the referent. I also argued that it (as well as all other D-determiners) includes domain restriction (C) in its denotation; this C is intersected with the denotation of the NP. With this in mind, we can now turn to the D-determiners in SkwxwuVmesh.1 In this chapter, I will provide an analysis of the deictic D-determiners ta/lha, ti/tsi, and kwa/kwelha. The non-deictic D-determiner is analyzed in Chapter 5. As I suggested in Chapter 3, a D-determiner that does not assert the uniqueness of its referent can be used in both novel and familiar contexts. In Skwxwu7mesh, the deictic D-determiners are be used in both novel and familiar contexts (§2). Unlike the in English, the deictic D-determiners in Skwxwu7mesh also do not assert the uniqueness of their referents, as I show in §3. However, sentences with deictic DPs usually carry an implicature of uniqueness. This implicature does not arise if the pragmatics otherwise do not allow it, and it can also be canceled in certain contexts, discussed below. In §4,1 provide an analysis of the interaction of domain restriction and lack of uniqueness in Skwxwu7mesh. The fact that the deictic D-determiners need not be familiar can be derived from the fact that, as we saw in Chapter 2, all arguments require D-determiners. This requirement forces the D-determiners to not assert the uniqueness/maximality of their referent. If it were the case that the D-determiners asserted their uniqueness, it would be difficult to begin conversations, as the domain C would never be narrowed enough for the DP to refer to a unique individual. For example, in (1), the DP the bear cannot refer to any particular bear because C contains all bears in the domain D e . The intersection of bear and C contains the same individuals as bear. There is no unique individual that belongs to both bear and C. ' Recall that the term D-determiners excludes quantifiers, numerals and demonstratives. 2 These are gender-neutral and feminine pairs. 3 Proper names are (usually) familiar, but this familiarity is not marked by the choice of determiner. 81 (1) a. # I saw the bear. (novel) C thebear = D e b. phe bearJJ = max(A,x [bear'(x) A C(X)]) = 0 A D-determiner that does not assert the uniqueness of the referent of its DP allows the hearer to narrow the domain C to an entity that satisfies the descriptive content of the NP because the hearer does not need to assume that it is a unique entity. The domain includes all bears, but here the function variable assigns an individual to the property supplied by the NP. (2) a. Chen kwelash-t ta. mixalh. (novel) C t amixaih = D e Isg.s shoot-tr det bear 'I shot a bear.' b. Jta mixalh] = f(5U [bear'(x) A C ( X ) ] ) = bearj C = {bean} A D-determiner that carries an assertion of uniqueness, on the other hand, cannot be understood, unless the hearer knows how to narrow the domain. In §5,1 discuss co-reference effects of Skwxwu7mesh DPs. DPs are usually co-referent across clauses and sentences, unless pragmatics forces them to refer to different entities. I argue that this a result of having domain restriction. The deictic D-determiners do not assert uniqueness, nor must they be used in familiar contexts; instead they encode deictic information. The D-determiners have [proximal], [neutral], [distal] and [invisible] features, as I showed in Chapter 2. (3) a. DP b. DP c. DP d. DP 4 D NP D NP D NP D NP [prox] [neut] [dist, invis] | | | | kwi ti ta kwa I treat this information in terms of presuppositional features in §6. In this chapter I argue that domain restriction is present in the representation of all deictic D-determiners. 4 See Chapter 5 for arguments that kwi lacks deictic features. 82 2 No novelty/familiarity distinction in Skwxwu7mesh The distinction between familiar and novel DPs is lacking in Skwxwu7mesh (Gillon 2003). This can be seen in both elicitation and textual contexts. The deictic D-determiners can be translated as either indefinite or definite, depending on the context.5 (4 ) Chen kw'ach-nexw ta/ti/kwa swi7ka. Isg.s look-tr(lc) det man 'I saw a/the man.' They are used when the speaker is familiar with the referent; it does not matter if the hearer is familiar with the referent or not (Kuipers 1967; see also Chapter 4 ) . In the next two sections, I will show that neither discourse old/new nor hearer old/new (Prince 1992; see also Chapter 3 for discussion) is encoded by the D-determiners. 2.1 Discourse new vs. discourse old The deictic D-determiners can be used in both discourse-new and discourse-old contexts. The deictic D-determiner to can be used to introduce both novel and familiar referents. In the example below, 'the barrel (full of molasses)' ta k'ek'ilas and 'molasses' ta mlashis are first introduced in the story, using the D-determiner to. (5) Uyulh-shit-em-wit ta k'ek'i7as si7ich' ta mlashis. canoe-appl-pass-Spl det barrel full det molasses 'A barrel of molasses was put aboard for them.' (discourse-new) (Kuipers 1967: 238) In the next example, the referent 'the big basket' is introduced using demonstrative kwetsi. Later in the text, the D-determiner ta is used to refer back to the now-familiar basket. 5 In elicitation contexts, ta, ti and kwa are often given definite interpretations, even in cases where native speakers of English would not use definites. As we will see below, the speakers provide definite (non-partitive) translations for the D-determiners even where the context provided should not allow it. In these cases, native English speakers would be required to use a partitive. A possible explanation for this difference is that the Skwxwu7mesh consultants learned English too late to acquire the uniqueness assertion of the. This is evidence that we cannot rely on translations as a clue to definiteness, but must instead rely on the contexts in which sentences are used. 83 (6) Na mi uys kwelhi hiyi slhanay', chem-chem'a7s-t-as rl come in dem.f big woman redup-carry.on.back-tr-3erg kwetsi hiyi sitn. dem big basket 'A big woman came in, carrying a large basket on her back.' ... s-e-s men tsexws-t-as ta staw'xwelh txwta7 t-ta sitn. nom-rl-3poss just throw-tr-3erg det child into obl-det basket ' . . .and she threw the children in the basket...' (discourse-old) (Kuipers 1967: 219-220) This pattern is consistent in elicited data as well. In the example below, the DP ta mixalh can introduce the referent ('a bear'), and can be used to refer back to the familiar referent ('the bear'). (7) Chen-t wa i-7imesh. Chen kw'ach-nexw ta mixalh. Isg.s-pst impf redup-walk Isg.s look-tr(lc) det bear 'I was walking. I saw a bear.' (discourse-new) Ta mixalh na mi ch'i-ch'ay-s-t-ts-as. det bear rl come redup-follow-caus-tr-lsg.o-3erg 'The bear followed me.' (discourse-old) Similarly, // can be used in both discourse-new and -old cases. In the example below, which is taken from a text, the new referent 'what he thought to be a floating island with trees on it' is introduced using the deictic D-determiner ti. Later in the text, the now-familiar island can be referred back to using ti. (8) Na te7usem (t-)ta ni7ch, na kw'ach-nexw-as rl look (obl-)det sea rl look-tr(lc)-3erg ti tsiin-t-as sesi7x skwtsa7s stita7 ti stsek-tsek. det call-tr-3erg float island on det redup-tree 'He looked out at the sea and saw what he thought to be a floating island with trees on it.' (discourse-new) ...S-e-s men shich'an-t-as-wit ti tsiin-t-as-wit nom-rl-3possjust circle-tr-3erg-3pl det call-tr-3erg-3pl skwtsa7s. island 'Then they circled around this would-be island.' (discourse-old) (Kuipers 1967: 236-7) This is seen in elicitation as well. 84 (9) Chen-t wa i-7imesh. Chen kw'ach-nexw ti mixalh. lsg.s-pst impf redup-walk Isg.s look-tr(lc) det bear 'I was walking. I saw a bear.' (discourse-new) T i mixalh na mi ch'i-ch'ay-s-t-ts-as. det bear rl come redup-follow-caus-tr-lsg.o-3erg 'The bear followed me.' (discourse-old) The distal D-determiner kwa can be used in discourse-new and discourse-old cases as well. The example below is of a discourse-new DP 'a spot upstream'. (10) Na txwta7 t-kwa s(hjiw-s... rl come obl-det upstream-3poss 'He came to a spot upstream....' (discourse-new) (Kuipers 1967: 234) In the example below, the D-determiner kwa is used in a discourse new context (where the spot upstream is introduced to the discourse), and then in a discourse old context (where that same spot is referred back to). (11) S-e-s men nam' kwetsi schishn txw-nam' t-kwa nom-rl-3poss just go dem messenger dir-go obl-det txw-7umich, nam' k'ap'-n-m kwi tina7 dir-upstream.region go include-tr-pass det from t-kwa Chiakmesh. obl-det Chiakmesh 'Then messengers were sent upstream as far as Chiakmesh.' (discourse-new) S-e-s men mi wukw' kwetsi-wit tina7 t-kwa hiw... nom-rl-3poss just come down dem-3pl from obl-det upstream 'Then the people from upstream came down...' (discourse-old) (Kuipers 1967: 236-7) The D-determiners do not differentiate between a discourse-new or -old environment. However, this is not the only possible way to encode familiarity. Familiarity effects are also sensitive to hearer-new and hearer-old status (Prince 1992). 2.2 Hearer new vs. hearer old Theoretically, the deictic D-determiners could encode a hearer-new versus hearer-old contrast. However, all of the deictic D-determiners can be used in both hearer-new and hearer-old contexts (Prince 1992; see also Chapter 3). 85 The D-determiner ta does not distinguish between hearer-old and hearer-new referents. It can introduce a new referent to the hearer, as in (5), repeated below. (5) Uyulh-shit-em-wit ta k'ek'i7as si7ich' ta mlashis. canoe-appl-pass-3pl det barrel full det molasses 'A barrel of molasses was put aboard for them.' (hearer-new) (Kuipers 1967: 238) This same D-determiner can occur with place names, such as the place name St'a7mes, proper names, such as Xwech 'tal', as well as the sun and the moon. All of these are hearer-old referents. (12) a. Tina7 tl'a6 St'a7mes ta Xwech'tal'. from obi.det St 'a?'mes det Xwech'tal' 'Xwech'tal' came from the village St'a7mes.' (hearer-old) (Kuipers 1967: 230) b. An tutaw ta snekwem/ta lhkaych. very bright det sun/det moon 'The sun/the moon is very bright.' (hearer-old) The D-determiner ti can also be used to refer to both hearer-old and hearer-new referents. It can be used when the speaker introduces a new referent to the hearer. (13) Chen kwelash-t ti mixalh kwi chel'aklh. An iyim. Isg.s shoot-tr det bear det yesterday . very strong 'I shot a bear yesterday. It was very strong.' (hearer-new) It can also be used to refer to a referent known to the hearer, such as ti tmixw 'the ground/earth', or for a place name, such as ti eslha7an. (14) a. Nilh melh kwi-s-e-s-kw xwakw'i-wit na tsut-wit foe then comp-nom-rl-3poss-already got.drunk-3pl rl say-3pl kwi-s-e-s shich'-an-tsut i7xw, s-kwekwi-7in'-tsut-s comp-nom-rl-3poss circle-tr-refl all nom-behave-tr-refl-3poss ti tmixw. det earth 'Then when they got drunk they thought that everything was spinning around and that the ground was moving in all directions.' (hearer-old) (Kuipers 1967: 239) 6 Recall that the determiner ta and the oblique marker t combine to make tl 'a with proper names and pronouns (Kuipers 1967). 86 b. Men yalh s-en mi tl'ik ti eslha7an. just finally nom-lsg.sbj come arrive det eslha7kn 'I just arrived in Eslha7an (a part of North Vancouver).' (hearer-old) The D-determiner kwa can be used to introduce a referent to the hearer, as in kwa siws 'a spot upstream'. It can also be used when the referent is known to the hearer, as in kwa Skwxwulmesh 'Squamish', or kwa Chiakmesh, both place names, as well as for proper names, such as Bill. (15) a. Na txwta7 t-kwa s(h)iw-s (hearer-new) rl come obl-det upstream-3poss t-kwa Skwxwu7mesh kwi-s-e-s kew. (hearer-old) obl-det Squamish comp-nom-rl-3poss descend 'He came to a spot upstream from Squamish when he descended.' (Kuipers 1967: 233) b. S-e-s men nam' kwetsi schishn txw-nam' t-kwa nom-rl-3poss just go dem messenger dir-go obl-det txw-7umich, nam' k'ap'-n-m kwi tina7 dir-upstream. region go include-tr-pass det from t-kwa Chiakmesh. obl-det Chiakmesh 'Then messengers were sent upstream as far as Cheakamus.' (hearer old) • (Kuipers 1967: 236-7) c. Na lulum kwa Bill. rl sing det Bill 'Bill sang.' ' (hearer-old) The D-determiner kwa does not distinguish between hearer-old or -new referents. Further, all of the deictic D-determiners can be used in existential contexts. The DPs in existential contexts are both hearer and discourse new. (16) Tsi7 ta/ti/kwa sha7yuna7 ta-n lam'. exist det ghost loc det-lsg.poss house 'There's a ghost in my house.' (discourse-new; hearer-new) The D-determiners therefore do not make reference to either the hearer-new/old distinction or the discourse-new/old distinction. In the next section, I show that the D-determiners also do not assert uniqueness. 87 3 Uniqueness in Skwxwu7mesh The D-determiners in Skwxwu7mesh do not assert uniqueness, unlike the D-determiner the in English. Skwxwu7mesh is not totally devoid of uniqueness effects, however. Sentences containing a deictic DP usually carry an implicature of uniqueness, which can be canceled in certain contexts (§3.2 and §3.3). 3.1 No assertion of uniqueness Unlike English the, Skwxwu7mesh D-determiners do not assert the uniqueness of their referent. For example, the D-determiner ta can be used in a context where the DP is not the unique referent, as in (17) below. There were two cups, equidistant from the speaker. They were exactly the same shape, size and colour. Neither was more salient than the other. In this context, (17) is perfectly felicitous. (It should be noted that the speaker is not asking for both of the cups.) (17) Mi7-shit-s chexw ta lapat. come-appl-caus 2sg.s det cup 'Bring me one of the cups.' (translated as 'bring me the cup') Consultant's comment: "You 're not asking for a specific one. " This same effect can be seen with mass nouns and plurals. In (18)a, the DP ta slhum' 'the/some soup' does not have to refer to the entire mass of contextually relevant soup. In (18)b, the DP ta skwelkwelam 'the/some berries' also does not have to refer to all of the contextually relevant berries. (18) a. Chen huy'-s ta slhum'. Tsi7-xw ta slhum' na7 Isg.s finish-caus det soup exist-still det soup loc ta nkwi7stn. det pot 'I ate some soup. There's still some soup in the pot.' (translated as 'I ate the soup and there's still some soup in the pot.) 88 b. Chen huy'-s ta skwel-kwelam, welh na7 Isg.s finish-caus det redup-berry conj loc ta na pukw-i7. S-en men haw k-'an det rl mould-inch nom-lsg.sbj just neg irr-lsg.sbj i huy'-s ta na piikw-i7. prox finish-caus det rl mould-inch 'I ate some of the berries, but some of them were mouldy, so I didn't eat the mouldy ones.' (translated as 'I ate the berries...') This is different from the, which I argued in Chapter 3 asserts uniqueness. Instead, sentences with deictic D-determiners only carry an implicature of uniqueness, as I show in the next section. 3.2 Implicature of uniqueness Although the D-determiners do not assert uniqueness, most sentences with a deictic DP carry an implicature of uniqueness. In the example below, ta mixalh refers to one bear. This sentence carries the implicature that it is the only bear in the context. (19) Chen kwelash-t ta mixalh kwi chelaklh. Isg.s shoot-tr det bear det yesterday 'I shot a bear yesterday.' This implicature arises because of the presence of domain restriction. If the D-determiner is associated with domain restriction, then the easiest way to interpret the DP is if the intersection of the domain restriction and the set of the NP gives the unique individual that is the referent of the DP. This implicature of uniqueness can be cancelled, as we saw in (18) above. Another example is given below. (20) a. Chen kwelash-t ta/tsi mixalh kwi chelaklh. Chen kw'ach-nexw Isg.s shoot-tr det bear det yesterday rl look-tr(lc) ta/tsi chanat mixalh, welh na tl'iw'-numut-wit. det three bears conj rl escape-refl-3pl 'I shot a bear yesterday. I saw three bears, but some escaped.' 89 Chen mukwts kwa si-wi7ka welh haw k-'an i Isg.s kiss det redup-man conj neg irr-lsg.sbj prox mukwts kwa John. kiss det John 'I kissed some of the men, but I didn't kiss John.' (translated as 'I kissed the men, but I didn't kiss John.') 3.3 Implicative of maximality Mass nouns behave very similarly to the singular count nouns. A mass DP usually refers to the entire mass. (21) Chen huy'-s ta slhum'. Isg.s finish-caus det soup 'I ate the soup.' (all of the soup in your bowl, for example) Like singular count nouns, this implicature of uniqueness can be canceled. This can be seen in the examples below.7 (22) a. Chen huy'-s ta slhum'. Tsi7-xw ta slhum' na7 Isg.s finish-caus det soup exist-still det soup loc ta nkwi7stn. det pot 'I ate some soup. There's still some soup in the pot.' (translated as 'I ate the soup and there's still some soup in the pot.) b. Chen takw ta/ti stakw. Isg.s drink det water 'I drank some of the water.' (translated as 'I drank the water') Context: I drank some water from my cup, but left some behind. Sentences containing a deictic DP carry an implicature of uniqueness: the hearer expects the referent to be unique in any given context, unless the context rules that uniqueness out, or the implicature is canceled. They cannot carry an assertion of uniqueness, as the DP is often interpreted as non-unique. 7 Examples of mass nouns with kwa are difficult (if not impossible) to construct. As discussed in Chapter 3, kwa is only used when the referent is "interesting" enough. 9 0 Plural nouns behave similarly to the singular and mass nouns. When a plural referent is introduced into the discourse, a following DP with the same descriptive content will (usually) refer to the entire group. If a plural DP is introduced, it is difficult to get a subset of that group, even when the speaker uses a number-neutral DP. For example, in (23), the DP lha slhanay' cannot refer to a subset of the group. (23) Xa7iitsn slhanay' na mi uy's. four woman rl come inside ?? Chen kwikwi-s lha slhanay'. Isg.s talk-caus det.f woman 'Four women came in. I talked to all of the women/*one of the women/*some of the women.' Instead, the DP must refer to the maximal set denoted by the NP. Sentences containing plural deictic DPs carry an implicature of the maximality of their referents. (24) a. Chen nam ch'aafl'am kwi chel'aklh. Chen kw'ach-nexw Isg.s go hunt/track det yesterday Isg.s look-tr(lc) tsi xa7iitsn mixalh. S-en men kwelash-t det four bear nom-lsg.sbj just shoot-tr ta/tsi mex-mixalh. det redup-bear 'I went hunting yesterday. I saw four bears. I shot all the/*some of the bears.' Consultant's comment: "How can you shoot all four bears? " b. Xa7iitsn swi7kana mi uy's. four man rl come inside Chen mukwts kwa si-wi7ka. Isg.s kiss det redup-man 'Four men came inside. I kissed all the men.' This effect is often seen in 'out-of-the-blue' cases. The consultants typically offer an 'all of the' translation. 8 As the DP refers to a plural referent, the speakers prefer the DP to be marked plural, (i) X a 7 u t s n slhanay' na mi uy's. four woman rl come inside Chen kwikwi-s lha slhen-lhanay'. Isg.s talk-caus det.f redup-woman 'Four women came in. I talked to all of the women/*one of the women/*some of the women.' Regardless, only the plural interpretation is available for the DP lha slhanay'. 91 (25) a. Chen huy-s ta skwel-kwelam. Isg.s finish-caus det redup-berry 'I ate all the berries.'9 b. Chen mukwts kwa si-wi7ka. Isg.s kiss det redup-man 'I kissed all the men.' This effect is strong, but it too can be canceled. For example, in (26)a, the DP ta mexmixalh 'the bears' is interpreted as referring to all four bears until the clause welh na tl'iw'numut ta nch 'u7 mixalh 'but one bear managed to escape' cancels the implicature of maximality.10 When collecting this piece of data, one of the speakers translated every sentence as I gave the Skwxwu7mesh. She translated sen men kwelasht ta mexmixalh as 'I shot all the bears'. Only after I finished the last clause welh na tl'iwnumut ta nch 'u7 mixalh 'but one bear escaped' did the consultants understand sen men kwelahst ta mexmixalh as 'I shot some of the bears'. (26) Chen nam ch'aatl'am kwi chel'aklh. Chen kw'ach-nexw Isg.s go hunt/track det yesterday Isg.s look-tr(lc) ta xa7utsn mixalh. S-en men kwelash-t det four bear nom-lsg.sbj just shoot-tr ta mex-mixalh, welh na tl'iw'-numut det redup-bear conj rl escape-refl ta nch'u7 mixalh. det one bear 'I went hunting yesterday. I saw four bears. I shot some of the bears, but one of them escaped.' (translated as '...I shot the bears, but one of them escaped.') Similarly, in (27), ta skwelkwelam is interpreted as referring to all of the berries, until the second clause welh na7 ta na pukwi7 'but some of them were mouldy' is added. 9 In this case, the plural can refer to all different kinds of berries. 1 0 In English, shot allows the sentence to be true if the animal is only wounded. In S k w x w u 7 m e s h , however, the verb kwelasht strongly implies the death of the animal. 92 (27) Chen huy-s ta skwel-kwelam, welh na7 Isg.s finish-caus det redup-berry conj loc ta na pukw-i7. S-en men haw k-'an det rl mould-inch nom-lsg.poss just neg irr-lsg.sbj i hiiy-s ta na pukw-i7. prox finish-caus det rl mould-inch T ate some of the berries, but some of them were mouldy, so I didn't eat the mouldy ones.' (translated as 'I ate the berries...') In cases where the number is unmarked, the effect is the same. (28) a. Chen huy-s ta skwelam, welh na7-t Isg.s finish-caus det berry conj loc-pst ta na pukw-i7. S-en men haw k-'an det rl mould-inch nom-lsg.sbj just neg irr-lsg.sbj huy'-s ta na piikw-i7. finish-caus det rl mould-inch 'I ate some berries, but some of them were mouldy, so I didn't eat the mouldy ones.' (translated as 'I ate the berries...') b. Chen hiiy-s ti skwelam, welh na7-t Isg.s finish-caus det berry conj loc-pst ti na pukw-i7. S-en men haw k-'an det rl mould-inch nom-lsg.sbj just neg irr-lsg.sbj huy'-s ti na piikw-i7. finish-caus det rl mould-inch 'I ate some berries, but some of them were mouldy, so I didn't eat the mouldy ones.' (translated as 'I ate the berries...') c. Xa7iitsn swi7kana mi uys. Chen mukwts kwa swi7ka four man rl come inside Isg.s kiss det man welh haw k-'an i mukwts kwa John. conj neg irr-lsg.sbj prox kiss det John 'Four men came inside. I kissed some of the men, but I didn't kiss John.' (translated as 'I kissed the men, but I didn't kiss John.') Skwxwu7mesh plurals are different from definite plurals in English in that they do not assert the maximality of their NP. 93 3.4 Summary D-determiners in Skwxwu7mesh do not assert the uniqueness/maximality of their referent. However, sentences containing deictic DPs carry an implicature of uniqueness which can be canceled or does not arise, if the situation does not allow it. The deictic D-determiners are different from the definite D-determiner in English, which asserts the uniqueness of its referent. The last remaining D-determiner (the non-deictic determiner) is also not a candidate for a definite D-determiner (see Chapter 5). If there is no set of definite D-determiners in Skwxwu7mesh, then by definition there can be no definite/indefinite distinction. 4 Domain restriction and the deictic D-determiners So far, I have shown that deictic D-determiners are unlike English the in that they lack assertion of uniqueness. In general, however, sentences containing deictic D-determiners carry an implicature of uniqueness. In this section, I argue that domain restriction is necessary to capture this implicature. The deictic D-determiners have something in common with the: domain restriction. 4.1 Why choice functions? Before I provide the analysis of the deictic D-determiners as being associated with domain restriction, it is necessary to explain the formalism I have adopted. Here I explain why I adopt a choice function analysis of the deictic D-determiners. I treat the deictic D-determiners in Skwxwu7mesh as choice functions (cf. Reinhart 1997, Winter 1997, Matthewson 1999, among many others). (29) _ta_ = AP f(Xx [P(x) A C(X)]) The function variable assigns an individual to the property supplied by the NP. Existential closure can apply at any point in the clause. In other words, they compose via Specify (cf. Chung and Ladusaw 2004). 94 (30) [ta mixalh]] = f(kx [bear'(x) A C(X)]) The DP ta mixalh will refer to an individual which is assigned by the choice function. This individual will also be a member of the set C . The choice of individual will therefore not be random, but determined by the context. I treat the deictic D-determiners as choice functions for two reasons. First, because deictic DPs can escape islands, as mentioned in Chapter 3. In example (31) below, the DP ta s7ixwelh 'a child' can take wide scope with respect to the universal quantifier. The DP escapes the conditional clause. (31) i7xw ta nexw7usialh wa7 ek' seselkw [u k huya7-as all det teacher impf fut sad if irr leaveSsbj ta s7ixwelh]. det child 'All the teachers will be sad if a child leaves.' 3x [*child'(x) A [leave'(x) - » Vy [*teacher'(y) A sad'(y)]]] Second, in novel contexts, the choice function becomes necessary in order for the DP to refer to a particular individual, and not the entire domain of entities matching the description of the NP. (32) a. Chen kwelash-t ta mixalh. (novel). C T A mixalh = D e Isg.s shoot-tr det bear 'I shot a bear.' b. [ta mixalh]] = f(A,x [*bear'(x) A C(X)]) = bean C T Amixalh = {bear;} There is a problem with this analysis, however. Most deictic DPs (ta and // DPs) can take narrow scope with respect to negation ((33)a), but must take wide scope with respect to other operators ((33)b and c). (33) a. Haw k-'an i yelx-t ta swi7ka. neg irr-lsg.sbj prox find-tr det man 'I didn't find a man.' 3f^[find'(f(*man'))(I)] - .3 f [find'(f(*man'))(I)] 95 b Chanat-alh s-en melyi three-times nom-lsg.sbj get.married 'I married a/the man three times.' ta det swi7ka. man 3f 3times [marry'(f(*man'))(I)] * 3times 3f [marry'(f(*man'))(I)] c i7xw slhen-lhanay' na mukwts all redup-woman rl kiss 'Every woman kissed a/the child.' 3f Vy [Voman'(y) A kiss'(f(*child'))(I)] * Vy 3f [swoman'(y) A kiss'(f(*child'))(I)] ta det s7ixwelh. child The scopal behaviour of the deictic DPs is mysterious. I have adopted a Specify analysis of choice functions for Skwxwu7mesh DPs. Choice functions, under a Specify analysis, should either be existentially closed (i) at any point of the derivation, or (ii) at the top-most point, depending on the language (Chung and Ladusaw 2004). However, in Skwxwu7mesh, the existential closure appears to apply either at the top of the clause, or at the top of the sentence. If this is true, I must assume that negative sentences are bi-clausal, as argued by Davis (2005) (contra Gillon 2002), and that the existential closure can apply at the highest level of the embedded clause, or the highest level of the sentence More research is required into this behaviour. In this thesis I focus on the ability of the deictic DPs to take wide scope, and not the exact position of their scope. 4.2 Domain restriction and implicature of uniqueness In Skwxwu7mesh, none of the D-determiners assert the uniqueness of their referents. However, sentences containing deictic D-determiners carry an implicature of uniqueness. I argue that this follows from domain restriction. This restriction is part of the denotation of the D-determiners, as given in (34)b below. I compare this denotation to the one I gave for the in Chapter 3. (34) a [the] = AP max(A.x [P(x) A C(x)]) b. _ta_ = AP f(A.x [P(x) A C(x)]) Domain restriction is a necessary part of the denotation of Skwxwu7mesh D-determiners just as much as they are of the English definite D-determiner. This is because DPs in 96 Skwxwu7mesh are also sensitive to the context they are used in. For example, across sentences, DPs must refer to the same individual. (35) a. Ghen nam ch'aafl'am kwi chel'aklh. S-en men Isg.s go hunt/track det yesterday nom-lsg.sbj just kw'ach-nexw ta mixalh. S-en men look-tr(lc) det bear nom-lsg.sbj just kw'elash-t ta mixalh. shoot-tr det bear 'I went hunting. I saw a bear. I shot the bear.' b. Sen men kw'achnexw ta mixalh. C t a mixalh = D e c. [ta mixalh] = {bear;} d. Sen men kw'elasht ta mixalh. C t a mixalh = {bean} e. [ta mixalh] = {bean} In (36)a, the DP ta slhum' 'some soup' refers to the same soup under discussion, not another, entirely new, pot of soup. Nor does it mean all of the soup in the world. (36) a. Chen hiiy'-s ta slhum'. Tsi7-xw ta slhum' na7 Isg.s finish-caus det soup exist-still det soup loc ta nkwi7stn. det pot 'I ate some soup. There's still some soup in the pot.' (translated as 'I ate the soup and there's still some soup in the pot.) b. Chen huy's ta slhum'. C t a sihum' = D e c. [ta slhum'] = {soupi} d. Tsi7-xw ta slhum' na7 ta nkwi7stn. Qasihum' = {soupi} e. [ta slhum'] = {soup;} Similarly, in (37), the DP ta mexmixalh 'the bears' refers to the set of bears already under discussion, not a wholly new set of bears, or the entire set of bears in the world. In (37)b, the DP tsi xa7utsn mixalh 'four bears' is used in a novel context. The domain restriction is the entire domain. However, in (37)d, the DP tsi mexmixalh 'the bears' is used in a familiar context, and 97 the domain is restricted to the previously introduced bears. The DP refers to all four of those bears (37)e. (37) a. Chen nam ch'aatl'am kwi chel'aklh. Chen kw'ach-nexw Isg.s go hunt/track det yesterday Isg.s look-tr(lc) tsi xa7utsn mixalh. S-en men kwelash-t det.f four bear nom-lsg.sbj just shoot-tr tsi mex-mixalh. det.f redup-bear 'I went hunting yesterday. I saw four bears. I shot all the/*some of the bears.' b. Chen kw'ach-nexw tsi xa7utsn mixalh. C t S i xa7tsn mixalh = D e c. psi xa7utsn mixalh]] = bean, bear], beark, bean d. Sen men kwelasht tsi mexmixalh C t S i mixalh = {bean, bearj, beark, bean} e. psi mexmixalh] = bean+bean+beark+bean Domain restriction normally forces the DP to refer to the set of elements already under discussion. In a context where a bear has been introduced, the DP ta mixalh 'the bear' normally refers to that same bear. (38) a. S-en men kw'elash-t ta mixalh. nom-lsg.sbj just shoot-tr det bear 'I went hunting. I saw a bear. I shot the bear.' C t a mixalh = {bean} b. pa mixalh] = f(Xx [*bear'(x) A C(X)]) = bean In a context where more than one bear has been introduced, the DP ta mexmixalh usually refers to the maximal individual sum of bears (39) a. S-en men kwelash-t ta mex-mixalh. nom-lsg.sbj just shoot-tr det redup-bear 'I shot all the bears.' C t a mexmixalh - {bean, bearj, beark, bean} b. pa mexmixalh] = f(A,x [®bear'(x) A C(X)]) = bean+bearj+beark+bean As mentioned in Chapter 2, the 'singular' DP is not singular, but rather number-neutral. In a context where more than one bear has been introduced, the DP ta mixalh will also usually refer to the maximal individual sum of bears. 98 (40) a. S-en men kwelash-t ta mixalh. nom-lsg.sbj just shoot-tr det bear 'I shot all the bears.' C T A mixalh - {bearj, bearj, beark, beari} b. pa mixalh] = f(A.x [*bear'(x) A C(X)]) = bearj+bearj+beark+beari If a D-determiner does not assert uniqueness, the hearer does not need to be familiar with the referent. The hearer does not need to narrow the domain to ensure that the DP is unique. In novel contexts, C includes D e , and the function variable assigns an individual to the property supplied by the NP. (41) a. Chen kwelash-t ta mixalh. (novel) C t a mixa lh = D e Isg.s shoot-tr det bear 'I shot a bear.' b. pa mixalh] = f(A,x [*bear'(x) A C(x)]) = bean C t amixaih = {bean} This is similar to the analyses of Matthewson (1999) and Giannakidou (2004), who both argue that domain is narrowed by the choice function itself. I argue that the choice function does not always narrow the domain. The choice function can narrow the domain as well (as we saw in cases like (26) above). However, it will not further narrow the domain unless there is a reason for it to do so. Matthewson (1999) and Giannakidou (2004) also did not address the question of how the choice function narrowed the domain. Without C, the choice function could theorectically choose any individual, or set of individuals. Having C in the representation of the D-determiners allows us to predict that the DP will refer to the individual already in the discourse to the property supplied by the NP. DPs are not definites in Skwxwu7mesh, as I've argued above, but in familiar contexts, they do behave more like definites, in that they usually refer to a previously introduced discourse referent. However, I do not adopt Giannakidou's (2004) analysis of St'at'imcets, where the DPs are argued to be definite. DPs in Skwxwu7mesh do not behave like definites. However, they do behave like some intermediate category, with definite-like behaviour in familiar contexts. Giannakidou's intuition that DPs in St'at'imcets are definite-like in familiar contexts is explained by the presence of domain restriction. 99 In the cases where an implicature of uniqueness does not even arise, as in (42)a, the function variable also assigns an individual to the property supplied by the NP. (42) a. Mi7-shit-s chexw ta lapat. C t a i a p a t = {cupj, cupj} come-appl-caus 2sg.s det cup 'Bring me one of the cups.' (translated as 'bring me the cup') b. _ta lapat] = f(Xx [*cup'(x) A C(X)]) = cup; or cupj Here, I claim that the pragmatics force the speaker to use ta to refer to a single cup (but neither one in particular), because it would be strange to ask for more than one cup in the context where I am asking for a cup to use to drink out of.11 4.3 Domain restriction and quantifiers Domain restriction is part of the representation of deictic D-determiners in Skwxwu7mesh. What about other elements? In English, quantifiers and demonstratives also appear to involve domain 12 restriction. I show here that quantifiers in Skwxwu7mesh do not involve domain restriction. As I showed in Chapter 2, quantifiers can co-occur with D-determiners. (43) a. Chen kw'ach-nexw i7xw ta push. Isg.s look-tr(lc) all det cat 'I saw all the cats.' b. Chen kw'ach-nexw kex ta push. Isg.s look-tr(lc) many det cat 'I saw many cats.' In these cases, I argue that the domain restriction is provided by the D-determiner. (44) a. S-en men kwelash-t i7xw ta mex-mixalh. nom-lsg.sbj just shoot-tr all det redup-bear 'I shot all of the bears.' C t a mexmixalh = {bean, bearj, beark, bean} b. _ta mexmixalh] = f(A.x [®bear (x) A C(X)]) - bean+bearj,+beark+bean 1 1 Under the right circumstances, this sentence can be used to refer to both cups (i.e. when I am washing dishes, and want to collect all dirty cups, plates, etc.). 1 2 On the basis of the Skwxwu7mesh discussion, I argue in Chapter 6 that quantifiers in English are not associated with domain restriction. 100 c. _i7xw ta mexmixalh] = XP Vy < f(Xx [sbear'(x) A C(X)) [atom(y) -> P(y) = 1 ] 1 3 However, as I also showed in Chapter 2, quantifiers do not obligatorily occur with D-determiners. (45) a. Chen kw'ach-nexw i7xw push. Isg.s look-tr(lc) all cat 'I saw all the cats.' b. Chen kw'ach-nexw kex push. Isg.s look-tr(lc) many cat 'I saw many (of the) cats.' I claim that the D position is still present in these cases. I do this because the sentences have the same interpretation regardless of the presence or absence of the D-determiner. I also do this because the speakers usually give examples with the D-determiner present. It is also more coherent to suggest that only one position is associated with domain restriction than to suggest that sometimes i7xw is associated with domain restriction, and sometimes it is not. 4.4 Domain restriction and demonstratives Demonstratives, unlike quantifiers, appear to be associated with domain restriction. As I showed in Chapter 2, they never occur with a D-determiner. (46) a. Chen kw'ach-nexw tay' push. Isg.s look-tr(lc) dem cat 'I saw that cat.' b. * Chen kw'ach-nexw tay' ta push. Isg.s look-tr(lc) dem det cat c. * Chen kw'ach-nexw ta tay' push. Isg.s look-tr(lc) det dem cat They behave differently than D-determiners, however, in that they can refer to individuals from within the domain, without any extra signalling. In the example below, lha slhanay' must refer to 1 3 1 am treating i7xw as a distributive operator. Another possibility is to treat it as forcing a "good fit" over the domain (see Brisson 2003). 101 all of the women. Speakers generally prefer the plural reduplicant when the referent is plural, especially when the number involved is over three. The example (47)a below is degraded because of this. (47)b is perfect. (47) a. Xa7iitsn slhanay' na mi uys. four woman rl come inside ?? Chen kwikwi-s lha slhanay'. Isg.s talk-caus det.f woman 'Four women came in. I talked to all of the women.' b. Xa7utsn slhanay' na mi uys. four woman rl come inside Chen kwikwi -s alhi slhanay'. Isg.s talk-caus dem.f woman 'Four women came in. I talked to one of the women/that woman.' This effect can be seen with all of the demonstratives. (48) a. Xa7iitsn swi7kana mi uys. four man rl come inside ?? Chen kwikwi-s ta swi7ka. Isg.s talk-caus det man 'Four men came in. I talked to all of the men.' b. Xa7iitsn swi7kana mi uys. four man rl come inside Chen kwikwi -s tay'/kwetsi swi7ka. Isg.s talk-caus dem man 'Four men came in. I talked to one of the men/that man.' In these contexts, the D-determiner ti and the demonstrative ti(wa) can also be semantically teased apart. (49) a. Xa7utsn swi7kana mi uys. four man rl come inside ?? Chen kwikwi-s ti swi7ka. Isg.s talk-caus det man 'Four men came in. I talked to all of the men.' b. Xa7iitsn swi7kana mi uys. four man rl come inside Chen kwikwi -s ti/tiwa swi7ka. Isg.s talk-caus dem man 'Four men came in. I talked to one of the men/that man.' 102 This difference between D-determiners and demonstratives has also been shown in Nuxalk (Bella Coola) by Davis and Saunders (1975). The question in (50) can be answered by (50)a or b, but not c. (50) A: kaks ti-?imlk ti-sp-ct? which det-man hitSabs-Isg.s 'Which man hit you?' B: a. ti-?imlk-tayx det-man-dem 'this man' b. ta-?imlk-tax det-man-dem 'that man' c. # ti-?imlk-tx det-man-det 'the man' This is because the D-determiner "...cannot be used to single out a particular member of a set" (Davis and Saunders 1975: 846). I argued in Chapter 2 that alhi is a medial demonstrative. This deictic information and the descriptive content of the DP must be intersected to refer to the appropriate individual. The question is, does alhi also provide domain restriction over the NP? There are two possibilities: either the demonstrative does (51)b, or there is a null D which provides the domain restriction (51) c. (51) a. [[slhanay'TJ = k[woman'(x)] b. [alhi slhanay']] = f(A,x [woman'(x) A medial-from-speaker'(x) A C(X)]) c. [alhi slhanay'TJ = k[woman'(x) A medial-from-speaker'(x)] [0D alhi slhanay'TJ = f(^x[woman'(x) A medial-from-speaker'(x) A C(X)]) Much like with D-determiners, if a DP has a demonstrative, it usually refers to a previously introduced discourse referent. Unlike D-determiners, demonstratives can refer to one member out of a previously introduced plural DP. 103 (52) a. Chen kw'ach-nexw lha slhanay'. Isg.s look-tr(lc) det.f woman An nach'im' alhi slhanay'. very pretty dem.f woman 'I saw a woman;. That woman; is pretty.' b. Chen kw'ach-nexw an'us slhanay'. Isg.s look-tr(lc) two woman An nach'im' alhi slhanay'. very pretty dem.f woman 'I saw two women. That woman is pretty.' Domain restriction still plays a role in demonstratives. There is indirect evidence that Skwxwu7mesh demonstratives occupy a different position from the D-determiners. As I showed in Chapter 2, D-determiners may precede or follow quantifiers. Demonstratives, however, may only follow. (53) a, (54) . Chen kw'ach-nexw Isg.s look-tr(lc) 'I saw a lot of bears.' ta det kex many mex-mixalh. redup-bear b. Chen kw'ach-nexw Isg.s look-tr(lc) 'I saw a lot of bears.' kex many ta det mex-mixalh. redup-bear c. * Chen kw'ach-nexw Isg.s look-tr(lc) kwetsi kex dem many mex-mixalh. redup-bear d. Chen kw'ach-nexw Isg.s look-tr(lc) 'I saw a lot of bears.' kex many kwetsi mex-mixalh. dem redup-bear a. Chen kw'ach-nexw Isg.s look-tr(lc) 'I saw all the bears.' ta det i7xw all mex-mixalh. redup-bear b. Chen kw'ach-nexw Isg.s look-tr(lc) 'I saw all the bears.' i7xw all ta det mex-mixalh. redup-bear c. * Chen kw'ach-nexw Isg.s look-tr(lc) kwetsi i7xw dem all mex-mixalh. redup-bear 104 d. Chen kw'ach-nexw i7xw kwetsi mex-mixalh. Isg.s look-tr(lc) all dem redup-bear 'I saw all the bears.' Demonstratives are different from the D-determiners in that they can more easily refer to subsets of the set of NPs given by the context; they also appear to occupy a different position. I discuss the issue of the position of demonstratives further in Chapter 6 and postulate that they occupy a different position from D-determiners based on cross-linguistic data. I also claim that demonstratives obligatorily co-occur with a null D-determiner, which provides the domain restriction over the NP. 5 Co-reference effects The presence of domain restriction on the D-determiners strongly predicts cross-sentential co-reference of DPs with the same descriptive content. Generally, this is the case. Here I discuss a few cases in detail. 5.1 Cross-sentential co-reference DPs are normally co-referent across clauses (or sentences), as we saw in §3.3. We can also see this in the examples below, where the DPs must refer to the same individual. It is impossible to conjoin a positive statement with its negative counterpart, using the same D-determiner. This is the Law of Contradiction (Heim and Kratzer 1998): only elements of type e (i.e., entities) obey this law.14 (55) a. # Na huya7 ta swi7ka i haw k-'as i huya7 rl leave det man conj neg irr-3sbj prox leave ta swi7ka. det man 'The man left and the man didn't leave.' Consultant's comment: "It's a contradiction. " Some quantifiers also are infelicitous in these contexts. 105 b. # Na huya7 ti swi7ka i haw k-'as i huya7 rl leave det man conj neg irr-3 sbj prox leave ti swi7ka. det man 'The man left and the man didn't leave.' c. # Na huya7 kwa swi7ka i haw k-'as i huya7 rl leave det man conj neg irr-3sbj prox leave kwa swi7ka. det man 'The man left and the man didn't leave.' (56) a. # Chen kwelash-t ta mixalh i haw k-'an i Isg.s shoot-tr det bear conj neg irr-lsg.sbj prox kwelash-t ta mixalh. shoot-tr det bear 'I shot the/a bear and I didn't shoot the/a bear.' b. # Chen kwelash-t ti mixalh i haw k-'an i Isg.s shoot-tr det bear conj neg irr-lsg.sbj prox kwelash-t ti mixalh. shoot-tr det bear 'I shot the/a bear and I didn't shoot the/a bear.' It is also impossible to use different D-determiners. (57) a. # Na huya7 kwa swi7ka i haw k-'as i huya7 rl leave det man conj neg irr-3 sbj prox leave ta swi7ka. det man 'The man left and the man didn't leave.' b. # Na huya7 ti swi7ka i haw k-'as i huya7 rl leave det man conj neg irr-3sbj prox leave ta swi7ka. det man 'The man left and the man didn't leave.' c. # Na huya7 ti swi7ka i haw k-'as i huya7 rl leave det man conj neg irr-3sbj prox leave kwa swi7ka. det man 'The man left and the man didn't leave.' 106 When the speaker makes it explicit that he or she is talking about two different situations, sentences like the ones in (55) above become felicitous. The two DPs must refer to the same individual. (58) Haw k-'as i huya7 ta/ti/kwa swi7ka kwi chel'aklh. neg irrSsbj prox leave det man det yesterday Na nam' huya7 ta/ti/kwa swi7ka ti stsi7s. rl go leave det man det today 'The man didn't leave yesterday. The man left today.' This is expected if the D-determiner is associated with domain restriction. (59) a. Na huya7 ta swi7ka.. C t a s w i7ka = De rl leave det man 'The man left.' b. pa swi7ka_ = f(X.x [*man'(x) A C(X)]) = man; C T A Swi7ka = {mam} c. Haw k-'as i huya7 ta swi7ka. C t a S w i 7 k a = {man;} neg irr-Ssbj prox leave det man 'The man didn't leave.' d. pa swi7ka]] = f(A.x [*man'(x) A C(X)]) = man; C t a S wi7ka = {man;} This co-reference across sentences does not always arise. Pragmatically, the two DPs cannot refer to the same individual in some cases. For example, the two occurrences of the DP ta mixalh below must refer to two different bears. (60) Chen kwelash-t ta mixalh kwi chel'aklh. Isg.s shoot-tr det bear det yesterday S-en men kiyat kwelash-t ta mixalh ti stsi7s. nom-lsg.poss just again shoot-tr • det bear det today 'I shot a bear yesterday and I shot a bear today.' Consultant's comment: "Calls for a drink; you shot two bears. " Each bear is killed on a different day: one yesterday and one today. In English, the D-determiner the, which asserts uniqueness, is not (usually) felicitous in this environment.15 1 5 As mentioned above in footnote 9, the Skwxwu7mesh word kwelasht strongly implies the death of the animal, much more than in English. The sentence (60) cannot mean that I shot and wounded the animal one day, and then shot it again another day. 107 (61) #1 shot the bear yesterday, and I shot the bear today.16 In (60), the DP is unique in each situation (one bear per day, for example), but it is not unique in the discourse context. Similarly, in example (62) below, the pragmatics force the two DPs to refer to different individuals. (62) a. Nu7-kw kwetk na xutsnalhsha7 ta swi7kai xwe7axw rl-already past rl forty det man conj not.yet k-'as wetl'ch' ta swi7ka. irr-3sbj twenty det man 'The man is past forty and the other man is not twenty yet.' b. An tl'aktaykwem ta swi7kai etsim ta swi7ka. very tall det man conj small det man 'There's a tall man and a short man.' In (55) (the law of contradiction cases), the DPs must co-refer, but the pragmatics should force them not to co-occur. I argue that within the same situation, the hearer expects the speaker to refer to the same individuals, unless they are overtly marked otherwise. If the speaker is talking about two different situations, or two different states, then the DPs will still co-refer, unless the pragmatics forces the DPs to refer to different individuals. I argue that co-reference is usually expected because of domain restriction. However, since the deictic D-determiners do not assert maximality, co-reference is not always forced. When the pragmatics disallow co-reference (and the speaker is obviously talking about two different situations), the DPs do not co-refer. When the speaker is talking about a single situation, the DPs are forced to co-refer, even if the pragmatics should disallow it. If the hearer can add entities to the domain when forced to by the pragmatics, why can't they do it every time? Why is there a preference for the DP to be co-referent across clauses? I argue that the hearer has to do less work if the intersection C and the set provided by the NP happen to give the referent. If the referent is instead a smaller group, the only way that the hearer will know this is if the speaker explicitly tells them to ignore some members of the intersection. 1 6 This sentence is felicitous if every day exactly one bear walks by the speaker. Normally the speaker doesn't shoot any of the bears, but yesterday he or she shot the bear (that walked by) and today he or she shot the bear (that walked by). 108 In the example below, the choice function will be existentially closed off at some point in the derivation. This choice function will pick out some set of three bears from the four. (17) a. S-en men kwelash-t ta mex-mixalh, nom-lsg.sbj just shoot-tr det redup-bear welh na tl'iw'-numut ta nch'u7 mixalh. conj rl escape-refl det one bear 'I shot some of the bears, but one of them escaped.' C T A mexmixalh = {bean, bearj, beark, bear,} b. _ta mexmixalh] = f(A.x [sbear'(x) A C(X)]) 5.2 Co-reference and the effect of transitivizers The choice of transitivizers has an effect on co-reference effects in Skwxwu7mesh. Certain transitivizers do not allow different reference cross-sententially, even where there are two different situations, and the pragmatics should force the DPs to refer to different referents. As I showed in Chapter 2, the transitivizer encodes the amount of control the agent has over the situation. In most cases, if the transitivizer encodes control of the agent, cross-sentential DPs can refer to different referents. In the example below, there are two bears that are killed (one per situation); the pragmatically odd interpretation where the same bear is killed is disallowed only because there is a more pragmatically viable interpretation available. (63) Chen kwelash-t tsi mixalh kwi chel'aklh s-en men kiyat Isg.s shoot-tr det.f bear det yesterday nom-lsg.sbj just again kwelash-t tsi mixalh ti stsi7s. shoot-tr det.f bear det today 'I shot a bear yesterday and I shot a bear today.' However, if the transitivizer encodes limited control of the agent, cross-sentential DPs must refer to the same referent. In the example below, the DP tsi mixalh must refer to the very same bear. (64) #Chen kw'uy-nexw tsi mixalh kwi chel'aklh s-en men kiyat Isg.s die-tr(lc) det.f bear det yesterday nom-lsg.sbj just again kw'uy-nexw tsi mixalh ti stsi7s. die-tr(lc) det.f bear det today #'I managed to kill the bear yesterday and I managed to kill the bear today.' 109 Only the very pragmatically odd reading, where the bear comes back to life only to be killed again, is allowed. Similarly, in example (65), the speaker must refer to the same man with ti swi7ka or ta swi7ka. s-en men kiyat nom-lsg.sbj just again (65) Chen kw'ach-nexw ti/ta swi7kakwi chel'aklh Isg.s look-tr(lc)(lc) det man det yesterday kw'ach-nexw ti/ta swi7kati stsi7s. look-tr(lc)(lc) det man det today 'I saw the man yesterday and I saw the man today.' Consultant's comment: "it's the same man " This looks like an effect of assertion of uniqueness. However, the uniqueness cannot be asserted, because the same non-maximality effects are found with the limited control transitivizer as were found with the control transitivizer in §3. (66) a. Chen hiiy-nexw Isg.s finish-tr(lc) na7 ta loc det slhum'. ta det soup nkwi7stn. pot Tsi7-xw exist-still ta det slhum' soup 'I ate the soup, and there's still some soup left in the pot.' Chen nam' ch'aatl'am kwi chel'aklh. det yesterday mixalh. S-en Chen kw'ach-nexw Isg.s look-tr(lc) men kw'uy-nexw Isg.s go hunt ta xa7utsn det four bear nom-lsg.poss just die-tr(lc) ta mex-mixalh,- welh na tl'iw-numut det redup-bear conj rl escape-refl ta nch'u7 mixalh. det one bear 'I went hunting yesterday. I saw four bears. I killed the bears, but one escaped.' Co-reference is therefore only indirectly tied to uniqueness. In English, co-reference is required across the discourse, because the D-determiner asserts uniqueness. In Skwxwu7mesh, co-reference is required within a situation, but is not required in different situations, unless the limited control transitivizer is used. Clearly, more work needs to be done on the effects of the transitivizers in Skwxwu7rnesh. I set these facts aside and focus on the lack of assertion of uniqueness associated with the D-determiners themselves. 110 6 De ic t i c features So far I have argued that the deictic D-determiners do not assert uniqueness and are not associated with the effects of familiarity seen in a language like English. I have also argued that they introduce domain restriction over their NP, as do all D-determiners. Here I will provide an analysis of their deictic information. Recall that the deictic D-determiners are constrained in their use by the location relative to the speaker. The proximal D-determiners were only licit when the referent was within reach of the speaker. In (67), proximal ti can only be used if the referent is close to the speaker. (67) a. P'ek' ti lapat. white det cup 'The cup is white.' (near speaker; in hand of speaker) b. * P'ek' ti lapat. white det cup 'The cup is white.' (halfway across room; across room) The neutral D-determiners were licit when the referent was visible, invisible, close to the speaker, or far from the speaker. (68) P'ek' ta lapat. white det cup 'The cup is white.' (near speaker; in hand of speaker; halfway across room; across room; invisible to speaker) The distal, invisible D-determiners were only licit when the referent outside of the room and invisible to the speaker. Bill. Bill (Bill not in room and invisible) Bill. Bill (Bill in room; Bill not in room and visible) Unlike the other D-determiners, kwa has two features. I summarize the necessary features for the determiner system in the table below. (69) a. Kw'ay' kwa get.hungry det 'Bill is hungry.' b. * Kw'ay' kwa get. hungry dist 111 Deictic Non-deictic Neutral Proximal Distal, invisible gender-neutral ta ti kwa kwi female lha tsi kwelha kwes Table 4.1: The D-determiner system of Skwxwu7mesh. These features are instantiated as in the example below. (70) a. DP DP c. DP .17 D [prox] ti NP D [neut] ta NP D NP [dist, invis] kwa d. ^ D P ^ D NP kwi The demonstratives are associated with some of the same features. Recall that tay', unlike ta, can only be used for middle distance referents, and that kwetsi, unlike kwa, can be used for visible, distal objects. Kwiya can be used for a referent at any distance, as long as the referent is invisible; kwawit can only be used for plural invisible, distal referents. (71) a. ti(wa) b. tay' c. kwetsi d. kwiya e. kwawit [prox] [med] [dist] [neut, invis] [dist, invis, pi] What is the status of these features? I assume that these features are presuppositional, following Schlenker's (2002) analysis of pronouns. Schlenker assumes that pronouns are the spell out of bundles of person, gender and number features. (72) a. /he/ <- [-author, -hearer, +masculine, -plural] b. /she/ « [-author, -hearer, -masculine, -plural] c. Ill ~ [+author, -hearer, -plural] d. /you/ ~ [-author, +hearer] Similarly, I assume that the D-determiners and demonstratives in Skwxwu7mesh spell out the bundle of gender (where relevant), number (where relevant) and deictic features. (73) a. b. c. d. /ti/ /tsi/ /ta/ /lha/ [proximal] [proximal, female] [neutral] [neutral, female] 1 7 See Chapter 5 for arguments that kwi lacks deictic features. 112 e. /kwa/ [distal, invisible] f. /kwelha/ [distal, invisible, female] a. /tiwa/ [proximal] b. /tsiwa/ [proximal, female] c. /tayV [medial] d. /alhi/ [medial, female] e. /kwetsi/ [distal] f. /kwelhi/ [distal, female] g- /kwiya/ [neutral, invisible] h. /kwsa/ [neutral, invisible, female] i. /iyawit/ [proximal, plural] j- /itsiwit/ [medial, plural] k. /kwetsiwit/ [distal, plural] 1. /kwawit/ [distal, invisible] m. /kwiyawit/ [neutral, invisible, plural] As I am focusing on the D-determiners in this thesis, I will not discuss the features of the demonstratives further. I define these features anaologously to Schlenker's analysis of the pronominal features. (75) a. proximal(x) is true iff s(x) is close to the speaker. Otherwise it is false. b. neutral(x) is true iff s(x) is locatable to the speaker. Otherwise it is false. c. distal(x) is true iff s(x) is far away from the speaker. Otherwise it is false. d. invisible(x) is true iff s(x) is invisible to the speaker. Otherwise it is false. e. female(x) is true iff s(x) is female. Otherwise it is false. The D-determiners are then only felicitous if their features match the context of use. Ti, for example, is only felicitous if the referent is close to the speaker. Lha is only felicitous if the speaker is able to locate the referent and if the referent is female. These D-determiners cannot involve features like [±author] or [±hearer], as the determiners can be used for all persons. As we saw in Chapter 2, pronouns co-occur with ta. (76) a. ta ens b. ta new det lsg.indep det lsg.indep 'I/me' 'you (sg)' These D-determiners also do not encode [±plural], as they can co-occur with singular (strictly speaking, number neutral) NPs or plural NPs. 113 (77) a. ta push b. ta pesh-piish det cat det redup-cat 'a cat/the cat(s), cats' '(the) cats' They also do not encode [±masculine], as only female human and animal referents are marked via gender on the D-determiners. (78) a. Ha71h-s chen lha slhanay'. like-caus Isg.s det.f woman 'I like the/a woman.' b. Ha71h-s chen ta swi7ka. like-caus Isg.s det man 'I like the/a man.' c. Ha71h-s chen ta lapat. like-caus Isg.s det cup 'I like the/a cup.' d. * Ha71h-s chen lha lapat. like-caus Isg.s det.f cup Unlike Schlenker, I assume that features are privative, rather than binary. Binary features would make incorrect predictions for some of the data in Skwxwu7mesh. For example, the [female] feature cannot be reinterpreted as [+female] with a [-female] counterpart. Female referents may co-occur with non-female D-determiners (as shown in Chapter 2). (79) a. An tl'aktay'kwem lha slhanay'. very tall det.f woman 'The woman is very tall.' b. An tl'aktay'kwem ta slhanay'. very tall det woman 'The woman is very tall' If the use of ta presupposed a [-female] referent, (79)b should be infelicitous. Similarly, if ta were [-proximal], it could not be used in cases where the referent were close, as with body parts. (80) a. Chen lha7n ti-n kwek'tan. Isg.s touch det-lsg.poss shoulder 'I touched my shoulder.' 114 b. Chen lha7n ta-n kwek'tan. Isg.s touch det.n-lsg.poss shoulder 'I touched my shoulder.' If ta were [-distal], it could not be used in cases where the referent was far away from the speaker. (81) a. An ha71h ta swi7ka. very good det man 'The man is good.' (man in room) b. An ha71h ta swi7ka. very good det man 'The man is good.' (man outside room) Similarly, if ta were [-invisible], it could not be used in cases where the referent was invisible to the speaker. (82) a. Chen ha71h-s ta-n push. Isg.s good-caus det-lsg.poss cat 'I like my cat.' (visible to speaker) b. Chen ha71h-s ta-n push. Isg.s good-caus det-lsg.poss cat 'I like my cat.' (out of sight of speaker) Even for the proximal and distal D-determiners, the features cannot be binary. If the feature were merely [±proximal], we would expect the distal, invisible D-determiner to used for referents that were relatively close (say, in the same room), but hidden. This is not the case. (83) a. Na kway ta Peter na7 ta rl hide det Peter loc det 'Peter is hiding inside.' b. * Na kway kwa Peter na7 ta rl hide det Peter loc det Uys. inside (speaker inside the same room) uys. inside (speaker inside the same room) Similarly,, if the feature were [±distal], we would expect the proximal D-determiner to be used for object that were in the middle distance. This is also not the case. 115 (84) a. P 'ek ' ti lapat. white det cup 'This/the cup is white.' (speaker holding cup, or cup very close to speaker) b. * P 'ek ' ti lapat. white det cup (cup in middle distance) Finally, i f the feature were [±invisible] for the proximal and distal D-determiners, we would also expect the proximal D-determiner not to be used for invisible referents. This is also not the case. (85) a. Chen lha7n ti-n kwek't 'an. Isg.s touch det-lsg.poss shoulder 'I touched my shoulder.' b. N a aa ti-n kweT. rl hurt det-lsg.poss stomach ' M y stomach hurts.' I therefore assume all of the features are privative, rather than binary. The advantage of treating these as features is that classes can be created across the demonstratives and D-determiners. The proximal D-determiners and demonstratives are used for referents that are very close to the speaker. The distal D-determiners and demonstratives only share one feature ([distal]), but they can be used in overlapping circumstances. I f the referent is invisible and far from the speaker, then either can be used. Further, given a Schlenker-type analysis, features end up having the same import as presuppositions. 7 Summary In this chapter, I have made the following claims. (86) a Skwxwu7mesh deictic D-determiners do not presuppose the familiarity of their referents. b Skwxwu7mesh deictic D-determiners do not assert the uniqueness of their referents. Sentences containing deictic determiners carry an implicature of uniqueness. c. Skwxwu7mesh deictic D-determiners spell out deictic and gender features. 116 d. Skwxwu7mesh deictic D-determiners have domain restriction in their representations. I also argued that the lack of presupposition of familiarity is a direct result of the lack of the assertion of uniqueness. As long as the DP does not need to be unique, the domain does not need to be shared between speaker and hearer. I argued that DPs are usually co-referent across clauses because they have domain restriction in their representation. Unless the DP cannot refer to a previously introduced referent (for pragmatic reasons), the domain will be restricted to the previously introduced referents. The determiners are deictic in nature; they encode distance from the speaker. I claim that all of the determiners discussed so far have features. I claim that ta has features, and is not underspecified. That this is necessary will be seen in the next chapter. There are two exactly opposite proposals that attempt to account for the data in Salish. Matthewson (1998) argues that Salish DPs are indefinite, based on their ability to occur in both novel and familiar contexts. Demirdache (1997) and Giannakidou (2004) both argue that Salish DPs are definite, based on their behaviour in familiar contexts. The analysis given in this chapter can capture the data that each proposal was trying to account for. 117 Chapter Five: The non-deictic D-determiner kwi 1 Introduction This chapter addresses the non-deictic D-determiner kwi and its place in the Skwxwu7mesh D-determiner system.1 Specifically, I explain why kwi behaves so differently from the rest of the D-determiners both syntactically and semantically. I argue in this chapter that the Skwxwu7mesh D-determiners all occupy the same position: D. I also show that the difference between these D-determiners is purely semantic; the deictic D-determiners provide the nominal with information that the non-deictic D-determiner does not. Specifically, the deictic D-determiners minimally provide the nominal with at least one deictic feature ([proximal], [neutral], [distal] and/or [invisible]); the non-deictic D-determiner does not provide the nominal with any of these features. I will argue that this difference has implications for the interpretations available to any DP in the language. Deictic DPs can take wide scope, because the deictic features do not allow the DP to compose via Restrict. I have argued in the last chapter that the D-determiners differ featurally, as below. (1) D NP [dist, invis] kwa All of the D-determiners have deictic features associated with them, marking proximity to the speaker, with the exception of kwi. This lack of featural content forces kwi to be interpreted differently from the deictic D-determiners. Crucially, the deictic features allow the DP to be able take wide scope, unlike the non-deictic D-determiner kwi, which must take narrow scope. An example of this behaviour is in negative contexts, where kwi must take narrow scope with respect to the negation, and the other D-determiners can take wide scope. 1 Recall that the term D-determiner does not include demonstratives or quantifiers. 118 (2) a. H a w k - 'an i s i lh7-an ta/ti s t s ' ukw i7 . neg irr-lsg.sbj prox buy-tr det fish 'I d idn ' t buy a fish.' (w ide o r nar row) b. H a w k - 'an i s i lh7-an kwa s t s ' ukw i7 . neg irr-lsg.sbj prox buy-tr det fish 'I d idn ' t buy a fish.' (w ide on ly ) c. H a w k - ' an i s i lh7-an kwi s t s ' ukw i7 . neg irr-lsg.sbj prox buy-tr det fish 'I d idn ' t buy a fish.' (narrow on ly ) I show that kwi ob l igator i ly takes narrow scope wi th respect to other quant i f iers (such as i7xw ' a l l ' ) and operators (such as u, the quest ion part icle) as w e l l (§3.2) . (3) a. N a mukw-s- t -as i 7 x w s lhen- lhanay ' kwi s t a w ' x w l h . rl kiss-caus-tr-3erg all redup-woman det child ' E v e r y w o m a n k issed a ch i l d . ' (narrow on ly ) b. N u 2 c h e x w k w ' a c h - n e x w kwi m i x a l h ? rl.Q 2sg.s look-tr(lc) det bear ' D i d you see a bear? ' (narrow on ly ) I argue that this is a result o f the lack o f deict ic features (§3) ; deict ic features do not a l l ow D P s to compose v i a Restr ict . Th is chapter has the f o l l ow ing structure. In §2,1 d iscuss some background on the nature o f structure and its potent ial impact on scope. In §3,1 p rov ide ev idence for treat ing kwi as a D -determiner, rather than occupy ing a lower funct ional project ion. §4 p rov ides an analys is o f kwi as a non-deic t ic D-determiner and shows what facts it can account for. §5 discusses the impact o f a featureless D-determiner. I w i l l show that the non-deic t ic D-determiner kwi is not semant ica l ly nu l l . Kwi introduces a contextual restr ict ion over its N P , just as any other D-determiner does. §6 shows that other potent ial analyses w i l l not wo rk for S k w x w u 7 m e s h . §7 conc ludes the chapter. 2 Nu comes from na u, the realis particle + Q. In slower speech, speakers are more likely to use the full form. 119 2 Structure and scope Structure has been argued to be related to scope (see Perez-Leroux and Roeper 2003, for example). However, Skwxwu7mesh provides evidence that there is not a one-to-one correspondence to the amount of structure and the scope-taking abilities of a nominal. Here I discuss the idea that wide scope should be related to structure and compare it to the idea that semantic features drive the ability to take wide scope. 2.1 Structure = wide scope? It is a reasonable hypothesis that nominals with more structure should be able to take wider scope than those with less structure. The presence of more structure might force a nominal to end up in a different position at LF (if it moves via QR), or it might create an argument out of a predicate. An argument could theoretically be existentially closed off at any point (assuming a particular version of a choice function analysis). Assuming that determiners signal the presence of more structure, we might expect that determiners also signal the ability to take wide scope. Indeed, the implicit assumption of most researchers seems to be that having a determiner is a necessary and sufficient condition for the ability to take wide scope. Perez-Leroux and Roeper (2003), for example, argue that "the semantic interpretation of bare nominals depends on their minimal syntactic structure". The data in Skwxwu7mesh shows that this cannot be true: there is one full DP that cannot take wide scope.31 will argue that a full DP in any language can take wide scope, but only if it has featural specifications. Determiners are usually associated with the ability to take wide scope. For example, the is usually interpreted with wide scope with respect to quantifiers and operators.4 (4) a. I didn't eat the apple. b. Every woman kissed the child. 3 Maori potentially also has a full DP which must take narrow scope (Chung and Ladusaw 2004). See Chapter 6 for discussion. 4 It can be interpreted with narrow scope under quantification in certain circumstances. (i) Every womanj kissed [the child she; loved most]. (narrow scope) This is not possible with a DP lacking a bound pronoun. (ii) Every woman kissed the child. (wide scope) 120 c. I always kiss the child. Any nominal headed by a is also able to take wide scope (i.e., it is not restricted to narrow scope).5 (5) a. I didn't eat an apple. (wide or narrow) b. Every woman kissed a child. (wide or narrow) c. I always kiss a child. (wide or narrow) On the other hand, bare plurals are restricted to narrow scope (Carlson 1977). (6) a. I didn't eat apples. (narrow) b. Every woman ate apples. (narrow) c. I always eat apples. (narrow) Data in English, such as (4)-(6) above, suggest that the presence of a (non-polarity) determiner is a sufficient condition for being able to take wide scope. The inverse also appears to be true: the data suggests that the lack of this functional projection forces the nominal to take obligatory narrow scope. However, this cannot be the correct generalization. Kwi is a determiner, and yet kwi DPs can only receive narrow scope with respect to negation, a quantified DP, or a question particle, as we saw in (2)c and (3). Neither bare nouns, nor kwi DPs may take wide scope. Having a determiner therefore is not a sufficient condition for the ability to take wide scope.6 The literature on narrow scope nominals (or weak indefinites) implicitly assumes that they are "barer" than full argument types. That is, a nominal which only takes narrow scope will have less structure than a nominal which can take wide scope. This can be seen in English, where the obligatorily narrow scope nominal only has a plural marker; any nominal introduced by a can take wide scope. 5 As we will see below, I do not treat a as a D-determiner. However, it potentially introduces more structure than a bare noun. 6 The non-deictic D-determiner kwi also behaves like the polarity item any in English, in taking obligatory narrow scope. However, unlike any, it is not limited to polarity and/or free choice environments. (i) Chen kw'ach-nexw kwi mixalh. Isg.s look-tr(lc) det bear 'I saw a bear.' I therefore dismiss the analysis of kwi as a polarity item immediately. 121 (7) a. John didn't see spots on the floor. (Carlson 1980:19) (neg > 3, *3 > neg) b. John didn't see a spot on the floor. (neg>3, 3 >neg) This idea, that narrow scope is linked to less structure, is more explicitly expressed by Borthen (2003): What I expect to find [crosslinguistically] is that reduced indefinites are more likely than corresponding nominals with determiners to ... take narrow scope... (Borthen 2003: 341) This does seem to be a tendency across languages. However, being "reduced", or lacking the D position is not a necessary condition for taking obligatory narrow scope. As I showed in Chapter 2, there are no bare nouns in Skwxwu7mesh (see also §3). However, there are nominals that take obligatory narrow scope: kwi DPs. Determiners do not force the ability to take wide scope. In any language, the lack of a determiner is a sufficient condition for obligatory narrow scope, but it is not a necessary condition. That is, if a nominal has a determiner, it may still obligatorily take narrow scope. Instead, I claim that the presence of semantic features determines the scope possibilities. 2.2 Semantic features = wide scope I show in §4 that having certain semantic features (such as deixis) is a necessary and sufficient condition for a D-determiner to have the ability to take wide scope. I also show that if the determiner lacks those features, it must take obligatory narrow scope. If a nominal lacks a determiner altogether, these semantic features cannot be present. This is why bare nouns must take obligatory narrow scope. These three possibilities are shown below (see also Farkas and de Swart 2004, for a similar point). (8) a. if DP -> able to take wide scope, b. if DP -> obligatory narrow scope [±F] c. if NP —» obligatory narrow scope 122 This is similar to Beghelli and Stowell's (1997) analysis of quantifier phrases and their potential scope. They claim that different QPs are associated with different features and that these features drive movement at LF. 7Ps move to their scope positions so that they can check their features. In this thesis, I abstract away from the potential scope positions, and focus on the fact that the lack of features is associated with the lack of the ability to take wide scope. Nominals must take wide scope if they are definite (i.e., they assert maximality), as we have seen with the cases in English. (9) I married the man three times. (*3X > 3, 3 > 3X) A nominal can also take wide scope if it has quantificational force. (10) I married every man three times. (3X > V , V > 3X) If a nominal has almost any functional projection, it can take wide scope. (11) a. I married a man three times. (3X>3, 3>3X) b. I married sm man three times. (3X>3,3 >3X) Nominals which have deictic information (such as ta DPs in Skwxwu7mesh) can take wide scope. In the example below, the DP ta swilka 'a/the man' must take wide scope. (12) Chanat-alh s-en melyi ta swi7ka. three-times nom-lsg.sbj get.married det man 'I married the man three times.' (*3X>3, 3>3X) I argued in Chapter 4 that the deictic D-determiners involved a choice function. (13) pa] = X? f(Xx [P(x) A C(x)]) 7 1 have assumed a choice function analysis of the deictic D-determiners in Chapter 4. L F movement is not relevant for my analysis. However, Beghelli and Stowell's analysis could still be relevant for the point at which existential closure applies to the choice function. 123 The deictic DPs technically then do not take wide scope; rather, the choice functions are existentially closed off at some point in the derivation. In this case, the choice function must be existentially closed off at the highest point in the derivation (above chanatalh 'three times'). 8 (14) 3 f 3times [marry'(f(man'))(I)] I use the short-hand "wide scope" to refer to the position that the choice function can be existentially closed off. The non-deictic D-determiner is unlike the deictic D-determiners in that it must take narrow scope. (15) Chanat-alh s-en melyi kwi swi7ka. three-times nom-lsg.sbj get.married det man 'I married a man three times.' (3X > 3, *3 > 3X) I argue that this is because it lacks any deictic features (§4). 3 T h e pos i t ion of k w i M y claim is that D-determiners lacking deictic features (like kwi) w i l l take obligatory narrow scope. I show in this section that kwi is a D-determiner (and not some other functional head) on the basis of its distribution. That is, it behaves like the other elements which are more obviously D-determiners. It also behaves more like the English D-determiner the in terms df its distribution than the English indefinite article a. I also argue that kwi creates an argument out o f an N P , as it is used in argument positions, and not in non-argument positions. Semantically, kwi also behaves like the deictic D-determiners, in terms of its lack of presupposition of uniqueness and lack of familiarity effects. I also show that a split D domain (as in Szabolcsi 1994) is unnecessary to describe the facts in Skwxwu7mesh. A split D domain reduces to the featural account I provided in Chapter 4. Recall that in some environments, deictic DPs can take narrow scope. I abstract away from that issue here. 124 3.1 Kwi is a D-determiner In this section, I argue that kwi is a D-determiner, syntactically and semantically. 3.1.1 Syntactic evidence that kwi is a D-determiner I argue that kwi is a D-determiner on the basis of its distribution. It occupies the same position as other D-determiners do. As I showed in Chapter 2, there is a group of elements in Skwxwu7mesh which are obligatory in argument position.9 This group includes kwi. As long as one of these elements introduces the NP, it is licit in argument position. There are no bare nominals in Skwxwu7mesh in subject (16) or object position (17). (16) a. Na wa sik kwi/ta kalaka. rl impf fly det crow (i) 'Crows fly.' (ii) ' The crow is flying.' 1 0 (17) a. Na wa sik kalaka. rl impf fly crow (Crows fly) Ha71h-s-t-as kwi/ta swi7ka lha good-caus-tr-3erg det man det.f (i) 'Men like women.' (ii) 'A/the man likes a/the woman.' Ha71h-s-t-as good-caus-tr-3erg Ha71h-s chen good-caus Isg.s 'I like the bear/bears. swi7ka man kwi/ta mixalh. det bear slhanay.' woman lha slhanay. det.f woman Ha71h-s good-caus chen mixalh. Isg.s bear Secondly, i f kwi were not in D, but rather in a different position (either lower or higher), we might expect it to co-occur with the other D-determiners. This is impossible, in either order. 9 The picture is actually more complicated than this, as Skwxwu7mesh allows arguments to be introduced solely with a numeral. I address this in Chapter 2. 1 0 The generic reading is only obtained when translating from the English (and is true for both ta or kwi). When the speaker is asked to translate the Skwxwu7mesh back into English, the episodic reading is given. This is true of all examples, regardless of the D-determiner involved. 125 (18) a. * Chen kw'ach-nexw ta/kwa/ti kwi mixalh Isg.s look-tr(lc) det kwi bear b. * Chen kw'ach-nexw kwi ta/kwa/ti mixalh Isg.s look-tr(lc) det det bear Co-occurance restrictions may arise from some other source, however. In English, for example, a and the also may not co-occur, despite the fact that a and the may not occupy the same position (Perlmutter 1970, Epstein 1999, Lyons 1999, Borer 2005). In Epstein's system, a occupies some position within NumP." We can show that this is not a viable position for kwi because kwi cannot occur in non-argument positions, unlike a in English. (19) a. Slhanay' lha Kirsten. woman det.f Kirsten 'Kirsten is a woman.' b. * Kwi slhanay' lha Kirsten. det woman det.f Kirsten Moreover, kwi can co-occur with numerals (20), unlike a in English (21). (20) Chen kw'ach-nexw kwi nch'ii7 skwemay'. Isg.s look-tr(lc) det one dog 'I saw one dog.' (21) a. I saw one dog. b. * I saw one a dog. c. * I saw a one dog. In English, determiners can co-occur with numerals (22), leading us to conclude that kwi is also a determiner. (22) a. I saw the three dogs. b. ? I saw some three dogs. (archaic) All of the Skwxwu7mesh D-determiners must precede any other head in the noun phrase, including possessive morphology, adjectives, numerals, etc. This is also true of kwi. 1 ' See Chapter 6 for more discussion of the position of a versus the. 126 a. Chen kw'ach-nexw ta-n an'us hiyi skwemay'. Isg.s look-tr(lc) det-lsg.poss two big dog 'I saw my two big dogs.' b. Chen kw'ach-nexw kwi-n an'us hiyi skwemay'. Isg.s look-tr(lc) kwi-lsg.poss two big dog 'I saw my two big dogs.' c. * Chen kw'ach-nexw an'us-n ta/kwi hiyi skwemay' Isg.s look-tr(lc) two-lsg.poss det/kwi big dog d. * Chen kw'ach-nexw an'us ta/kwi-n hiyi skwemay' Isg.s look-tr(lc) two det/kwi-lsg.poss big dog e. * Chen kw'ach-nexw an'us ta/kwi hiyi-n skwemay' Isg.s look-tr(lc) two det/kwi big-lsg.poss dog The non-deictic D-determiner kwi also does not occur in other non-argument positions, such as inside complex predicates. This is unlike St'at'imcets, where ku (the closest element to kwi) can occur in many non-argument positions. See §6.1 for discussion. On the basis of the above facts (kwi introduces an argument, and never a predicate or other non-arguments), I assume that kwi must be in D position. It must be a real D-determiner, and not the head of a lower functional projection. 3.1.2 Semantic evidence that kwi is a D-determiner Despite the difference between the non-deictic and deictic D-determiners, kwi behaves like the deictic D-determiners in that it can be used in both familiar and novel cases. It also does not carry a presupposition of uniqueness. 3.1.2.1 Kwi does not presuppose familiarity/novelty Further evidence that kwi occupies D comes from the properties it shares with the rest of the D-determiner system. The non-deictic D-determiner can be used in both novel and familiar contexts. For example, the D-determiner kwi can be used in an existential context, which is both discourse and hearer new (see Prince 1992 and Chapters 3 and 4 for more discussion). 1 2 The possessive morphology usually encliticizes to the first element in the DP, which happens to be the D-determiner. The fact that the possessive morphology is hosted by something else is not relevant to the ungrammaticality, which can be seen in the next example. 127 (24) Tsi7 kwi sha7yuna7 ta-n lam'. exist det ghost loc det-lsg.poss house 'There's a ghost in my house.' (discourse-new; hearer-new) In one of the texts in Kuipers (1967), the D-determiner kwes (the feminine non-deictic D-determiner) is used to introduce the speaker's wife.13 This is an instance of a discourse-new referent, and possibly a hearer-new referent. The same D-determiner is used when the speaker refers back to the same referent. (25) N-s-na men k'anatsut-nit-an kwetsi snexwilh-chet lsg.poss-nom-rl just return-appl-lsg.erg dem canoe-lpl.poss s-men tsun-t-an kwes n-skw'u7-t: ... nom-just tell-tr-lsg.erg det.f 1 sg.poss-wife-pst 'Then I returned to our canoe and told my wife: ...' (discourse-new) ...N-s-na men k'anatsut nam' t-ta n-snexwilh, lsg.poss-nom-rl just return go obl-det lsg.poss-canoe n-s-na men wilk'-t-an kwes n-skw'u7 -t: ... lsg.poss-nom-rl just ask-tr-lsg.erg det 1 sg.poss-wife-pst 'I returned to my canoe and asked my wife: ...' (discourse-old) (Kuipers 1967: 241) Another novel use of kwi can be seen below. (26) Na=kw hem'i syetsmkwi s-es hem'i kwi stl'alkm wa rl=already come report comp nom-[rl]3poss come det monster impf nan-t-em Sinulhkay'. name-tr-pass S. 'News was received that a monster named Sinulhkay' was coming.' (discourse-new) (Kuipers 1967: 230) In elicitation contexts, kwi can be used for novel or familiar referents. (27) Chen-t wa i-7imesh. Chen kw'ach-nexw kwi mixalh; lsg.s-pst impf redup-walk Isg.s look-tr(lc) det bear 'I was walking. I saw a bear.' (discourse-new; hearer-new ...Na mi chi-chay-(t)-ts-as kwi mixalh. rl come redup-follow-tr-lsg.o-3erg det bear 'The bear followed me.' (discourse-old; hearer-old) 1 3 Unfortunately, this is not the best example of a novel use of kwi, because the hearer could presumably accommodate the speaker's wife. 128 As with the deictic D-determiners, kwi can be used in novel and familiar contexts. The semantics are consistent with the other D-determiners in Skwxwu7mesh. 3.1.2.2 Kwi does not assert uniqueness As with the deictic D-determiners, kwi does not assert the uniqueness/maximality of its referent. For example, in(28) below, there may be many cups in the cupboard; the speaker is only asking for any one of the cups. (28) Mi7-shit-s chexw kwi lapat. come-appl-caus 2sg.s det cup 'Bring me a cup.' Similarly, in (29)a, kwi slhum' 'soup' does not have to refer to the entire mass of soup, and in (29) b, kwi skwelkwelam 'berries' does not have to refer to all of the berries. (29) a. Chen hiiy'-s kwi slhum'. Tsi7-xw ta slhum' na7 Isg.s finish-caus det soup exist-still det soup loc ta nkwi7stn. det pot 'I ate some soup. There's still some soup in the pot.' b. Chen huy'-s kwi skwel-kwelam, welh na7 Isg.s finish-caus det redup-herry conj loc ta na pukw-i7. S-en men haw k-'an det rl mould-inch nom-lsg.poss just neg irr-lsg.sbj i huy'-s ta na p\ikw-i7. prox finish-caus det rl mould-inch 'I ate some of the berries, but some of them were mouldy, so I didn't eat the mouldy ones.' (translated as 'I ate the berries...') As with the deictic D-determiners, kwi does not assert the uniqueness of its referent. It can, however, be used to refer to all members of a previously introduced set. (30) Chen nam' ch'aafl'am. Chen kw'ach-nexw kwi xa7iitsn mixalh. Isg.s go hunt/track Isg.s look-tr(lc) det four bear Chen kwelash-t kwi mex-mixalh. Isg.s shoot-tr det redup-bear 'I went hunting. I saw four bears. I shot all of the bears/some of the bears.' Sentences containing kwi DPs do not carry an implicature of uniqueness, unlike sentences containing deictic DPs. 129 3.2 No split D P domain in Skwxwu7mesh So far I have shown that kwi is a D-determiner. However, the possibility that the other D-determiners occupy a different position than kwi is still possible. The deictic D-determiners could occupy a different position, adjacent to the position that kwi occupies. There is no overt evidence for an analysis like this, because, as we saw above, the deictic and non-deictic D-determiners never co-occur. This section discusses whether a structural difference could drive the semantic differences outlined in §2. If structure alone can do this work for us, then I do not need to argue that the semantic differences follow from a featural difference. A different structure could still derive obligatory narrow scope. I will show in this section that the structural analysis is not possible for Skwxwu7mesh. In §4,1 will provide evidence that, for independent reasons, kwi must lack deictic features. An example of a structural analysis is Szabolcsi's (1994) of double determiners in Hungarian and Greek. She distinguishes between two determiner positions in Hungarian: (i) a higher 'subordinator' position, which she claims is analogous to C(omp) - both 'enable the clause or noun phrase to act as arguments' (p.214) and (ii) a lower quantifier/demonstrative position, which is analogous to T(ense). The subordinating determiner is labeled simply as 'D'. The quantifier/demonstrative determiner is labeled as 'Det'. I provide her structure below. (31) a(z) DP (N+I)^ DetP N+I Det' Det (Szabolcsi 1994: 214) The DetP is the complement of the (inflected) noun, whereas the D head takes the NP as its complement. These details are not relevant to the main point of the split between subordinating and deictic D-determiners. For simplicity and ease of exposition, I will argue against a simpler 130 version of Szabolcsi's structure, as in (32), where D s u b refers to the 'subordinator' position, and Ddeictic refers to the quantifier/demonstrative position. (32) / \ D s u b DP^ Ddeictic NP That (at least) two positions are necessary in some languages can be seen in the examples below. Greek DPs can have both a subordinating determiner (to - 'the') and a quantifier determiner (kathe 'every'). Hungarian can also have both a subordinating determiner (a - 'the') and a quantifier or deictic determiner (minden 'each', e^en 'this', or melyik 'which'), as long as there is some intervening syntactic material. Rumanian, St'at'imcets and Colloquial Norwegian (among many other languages) also allow a D-determiner and a demonstrative to co-occur. (33) a. t o 1 4 kathe pedhi det every chUd 'every child' (Greek; Szabolcsi 1994:213) b. a [vel-ed valo] minden/e^en/melyik talalkozas the with-2sg being every/this/which meeting 'every/this/which meeting with you' (Hungarian; Szabolcsi 1994:219) c. om-ul acesta man-det dem 'this man' (Rumanian; Giusti 1993: 111) d. nilh-s qwatsats-s ti7" t i sqaycw-a. foc-nom leave-3sg.poss dem det man-exis 'And then the man left.' (St'at'imcets; van Eijk and Williams 1981: 58) e. den herre klokka det here watch 'this watch' (Colloquial Norwegian; Cheng and Sybesma 1999: 539) Szabolcsi further claims that there are two options for determiners in a language. That is, the subordinating and quantifier/demonstrative positions can be occupied by two different 1 This is the accusative form of the D-determiner. 131 morphemes (e.g., Hungarian or Greek determiners), or they may be conflated into one morpheme (e.g., English the, every, etc.). The deictic and non-deictic D-determiners in Skwxwu7mesh could be analyzed as occupying two distinct positions: deictic D and subordinating D respectively (see Gillon 2004). (34) a. DP b. D s u b DP D s u b NP Ddeictic N P kwi ta/kwa/ti The deictic D-determiners provide the N P with deictic information, and it seems likely that that would be their position. There is also some evidence that kwi is a subordinator. It is also used as a complementizer of embedded clauses (Kuipers 1967, Gillon 2002), unlike the deictic D-determiners. (35) a. Chen lhchiws [kwi-n-s wa ts'its'ap']. Jsg.s tired comp-1sg.poss-nom impf working 'I am tired of working.' b. * Chen lhchiws [ta/kwa/ti-n-s wa ts'its'ap']. Isg.s tired det-1 sg.poss-nom impf working A problem that is raised by this analysis is that the non-deictic D-determiner cannot co-occur with the deictic D-determiners, as shown in (18), partially repeated here. (17) a. * Chen kw'ach-nexw ta kwi mixalh. Isg.s look-tr(lc) det kwi bear b. * Chen kw'ach-nexw kwi ta mixalh. Isg.s look-tr(lc) det det bear Further, quantifiers can co-occur with any of the D-determiners, suggesting that kwi and the rest of the system are in the same position.15 1 5 To be fair, the examples with i7xw 'all' are not very telling, because i7xw is probably generated higher than either determiner position (under the assumption that the D-determiners can move past the Q). On the other hand, kex 'many' should occupy the deictic/quanitifier position, at least under certain assumptions. The point here is that the D-determiners seem to behave as a class. 132 (36) a. i. Chen kw'ach-nexw i7xw ta/kwi mex-mixalh. Isg.s look-tr(lc) all det redup-bear 'I saw all the bears.' ii. Chen kw'ach-nexw ta/kwi i7xw mex-mixalh. Isg.s look-tr(lc) det all redup-bear 'I saw all the bears.' b. i. Chen kw'ach-nexw ta/kwi kex mex-mixalh. Isg.s look-tr(lc) det many redup-bear 'I saw lots of bears.' ii. Chen kw'ach-nexw kex ta/kwi mex-mixalh. Isg.s look-tr(lc) many det redup-bear 'I saw lots of bears.' A further problem is found in the Korean data Szabolcsi herself provides: demonstratives can co-occur with quantifiers in Korean (37)a. This is also true of Skwxwu7mesh (37)b and c. (37) a. i/ku motun salan dem/det every person 'all the(se) people' b. Chen kw'ach-nexw kex kwetsi-wit Isg.s look-tr(lc) many demSpl 'I saw lots of bears.' c. Chen kw'ach-nexw i7xw kwetsi-wit Isg.s look-tr(lc) all dem-Spl 'I saw all those bears.' (Korean; Scabolcsi 1994: 213) mex-mixalh. redup-bear mex-mixalh. redup-bear (Skwxwu7mesh) (Skwxwu7mesh) It appears that Szabolsci would actually need three functional projections: D s u b , Ddeictic, and Dquantifier (or Q), each of which can be conflated. Hungarian would conflate Ddeictic and Q, Korean would conflate D s u b and Ddeictic, and English would conflate all three. In that case, we predict a fourth type of language which conflates none of them. St'at'imcets appears to be this kind of language. 16 (38) lan-lhkan tu7 wa7 paqw-ens already-lsg.s compl be look-tr 'I already looked at all these books.' takem iz' i ptikw-a. all dem det book-exis (St'at'imcets; Matthewson 1998) 1 6 Szabolcsi would have to claim that the demonstratives and quantifiers can move past the D-determiner head. In Chapter 6,1 make a similar claim. 133 While it appears Szabolsci is correct in that languages need more structure than just a single D position for D-determiners, quantifiers and demonstratives (including Skwxwu7mesh), there does not seem to be any evidence for a split between the deictic and non-deictic D-determiners in the position that they occupy. The only way to extend Szabolcsi's analysis to the D-determiners in Skwxwu7mesh would be to conflate the two positions in the case of the deictic D-determiners, and for the non-deictic D-determiner to occupy the higher, subordinating D, as shown in (39). This would mean that conflation is a lexical choice. While there is no evidence against this analysis of the D-determiners, it reduces to being equivalent to the featural account I have provided in Chapter 4. The head D would have both labels (which are presumably associated with some kind of semantic information) when occupied by the deictic D-determiners, and only the subordinate label when occupied by kwi. I claim that deictic D-determiners have deictic features, but that these features are not projected in the syntax. Thus, the two D-determiner types occupy the same position. In Skwxwu7mesh, it is only the presence or absence of features in the DP domain that has semantic effects, not syntactic structure. 4 The analysis: kwi as a non-deictic D-determiner In this section, I show that kwi is not associated with any deictic features. It can only be used where the speaker cannot locate the referent, or does not wish to provide the location of the referent. 134 4.1 The non-deictic D-determiner kwi The D-determiner kwi does not have deictic features. It can be used for referents at varying distances away from the speaker. It can be used for internal body parts (40)a, referents which may be in the same room (40)b, referents which are relatively far away (40)c, or non-existent referents (40)d. (40) a. Na pum kwi-n kw'el'. rl swell det-lsg.poss stomach 'My stomach is swelling.' b. N-s-tl'i7 kwi shukwa. lsg.poss-nom-dear det sugar 'I want some sugar.' (sugar might be on counter, or in cupboard, for example) c. Chen kat kwi smanit. Isg.s climb det mountain 'I climbed a mountain.' (not necessarily nearby) d. N-s-tl'i7 kwi-n-s yeltx kwi kwtams. lsg.poss-nom-dear comp-1 sg.poss-nom find det husband 'I want to find a husband.' If the speaker can locate the referent, s/he will usually use another D-determiner or demonstrative, as in (41) 17 (41) N-s-tl'i7 ta shukwa. lsg.poss-nom-dear det sugar 'I want the sugar.' (sugar is on the table, for example) 1 7 This difference is even found in wh-questions, as in (i) and (ii) versus (iii). (i) Stam ti na wa ta7-s-t-axw? what det rl impf make-caus-tr-2sg.erg 'What are you making?' (lit: what is the thing you are making?) "Here the speaker is questioning an addressee whom sees [sic] at work on something." (Kuipers 1967: 138) (ii) Stam ta na wa takw-an-t-axw? what det rl impf drink-tr-tr-2sg.erg 'What are you drinking?' (speaker can see the liquid that the hearer is drinking, but cannot identify it) (iii) Stam kwi na wa kw'ach-nexw-axw? what det rl imperf look-tr(lc)-2sg.erg 'What did you see?' (speaker did not see the object) St'at'imcets only allows ku (the closest equivalent to kwi) to be used in wh-questions (Matthewson, p.a). 135 In some cases, the speaker may be able to locate the referent and still choose to use kwi (as in (40)a, b and c). In these cases I argue that the speaker can pretend not to know where the referent is located because there is no visible counter-evidence to their claim that they cannot locate the referent. In (40)a, for example, the speaker's stomach is not visible to the hearer. In (40)b, the speaker can use the non-deictic D-determiner kwi because he or she is asking for a part of the mass of sugar. If the referent is not locatable by the speaker (because, for example, it is not seen by him/her, or it may or may not exist), kwi must be used. (42) a. Nam' yelx-t kwi u7tis. go find-tr det egg 'Go find some eggs!' b. Yuu chaxw, take.care 2sg.emph t-ta smant kwi obl-det stone det iw'ayti na maybe rl elhkay'. snake wa impf 'Careful, there may be a snake under the stone. Chen wa yelx-t kwes18 slhanay' Isg impf find-tr det.f woman tl'a ens. obi.det lsg.indep 'I am looking for a woman to work for me.' Tsi7 u kwi e-lam'? exist Q det 2sg.poss-house 'Do you have a house?' (lit: Is there a house of yours?) N-s-tl'i7 kwi-n-s lsg.poss-nom-dear comp-lsg.poss-nom 'I want to build houses.' kwi comp ta7 make lesiw'ilh under (Kuipers 1967: 138) s-ts'its'ap'-s nom-work-3poss (Kuipers 1967: 138) kwi det lem-lam'. redup-house 1 8 This is the original feminine form of kwi. It appears to have been lost. The feminine forms are not particularly stable: ta is often used for females, especially when they are pluralized. (i) Chen kw'ach-nexw ta slhen-lhanay'. Isg.s look-tr(lc) det redup-woman 'I saw the women.' 136 f. Tsi7 u kwi e-men'-men? exist Q det 2sg.poss-redup-child 'Do you have any children?' (lit: Are there children of yours?) If the deictic D-determiner ta is used instead, the referent is locatable. By 'locatable', I mean that the speaker at some point knew where the referent was located, even if the speaker cannot locate the referent at the time of speaking. If the referent is locatable, the DP often receives a definite interpretation when translated into English. In some cases, the referent is not locatable to the speaker, but ta is still licit. I argue this is because ta allows the DP to take narrow scope. The fact that this is not the best choice of determiner can be seen in the variable judgments. In (43)f, for example, ta may not be used. (43) a. Nam' yelx-t ta u7iis. go Jind-tr det egg 'Go find the egg!' b. Iw'ayti na wa lesiw'ilh t-ta smant ta elhkay'. maybe rl impf under obl-det stone det snake 'Maybe the snake is under the stone.' c. Chen wa yelx-t lha slhanay' kwi s-ts'its'ap'-s Isg impf find-tr det woman comp nom-work-3poss tl'a ens. obi.det lsg.indep 'I am looking for the woman who works for me.' d. Tsi7 u ta e-lam'. exist Q det 2sg.poss-house 'Do you own a house?' (lit: Is there a house of yours?) e. N-s-tl'i7 kwi-n-s ta7 ta lem-lam'. lsg.poss-nom-dear comp-1 sg.poss-nom make det redup-house 'I would like to make houses.' f. * Tsi7 u ta e-men'-men? exist Q det 2sg.poss-redup-child The non-deictic D-determiner is also used for things like sna 'name', or when introducing one's name, which are both non-locatable. 137 (44) a. Peter kwi n-s-na.19 Peter det 1 sg.poss-nom-call 'My name is Peter.' b. * Peter ta-n s-na. Peter det-Isg.s nom-name c. Chen wa nan-t-em t-kwi Sxaltxw. Isg.s impf call-tr-pass obl-det Sxaltxw T am called Sxaltxw.' (Kuipers 1967: 138) d. * Chen wa nan-t-em tl 'a Sxaltxw. Isg.s impf call-tr-pass obl.det Sxaltxw e. Kwi s-we7u Pita nam' hewa7 tl'a ens. det nom-call Peter go accompany obl.det lsg.indep 'The one called Peter is to accompany me.' (Kuipers 1967: 138) This lack of deictic information is often represented in the English gloss as an emphasized a. (45) Ha71h-s chen kwi mixalh. good-caus Isg.s det bear 'I like a bear.' Complex numerals also take kwi. This is expected since numerals are not locatable. (46) Upen i kwi nch'u7 ten conj det one 'eleven' If the referent is plural, and there is a chance these individuals might not be in the same location as each other, the speaker often chooses to use kwi. Most of the deictic D-determiners may be used as well, but it is not the first choice. (47) a. Chanat kwi n-lem-lam'. three det lsg.poss-redup-house 'I have three houses.' (lit: my houses are three) 1 9 The first person possessive marking does not always encliticize to kwi; in this case, it procliticizes to the following word. This marking seems to always encliticize to the other D-determiners. I do not know if this is a significant difference. 138 b. Chanat ta-n lem-lam'. three det-lsg.poss redup-house 'I have three houses.' c. Chanat ti-n lem-lam'.20 three det-lsg.poss redup-house 'I have three houses.' Often, deceased relatives are introduced by kwi as well. As they are no longer locatable, it makes sense to use the non-deictic D-determiner. The "past tense" marker -t is also used in these constructions. (See Burton 1997 for a discussion of the equivalent of this marker in Halkomelem.) (48) a. kwi n-kwupits-t det 1' sg.poss-older.sibling-past 'my deceased older brother' (Kuipers 1967: 138) b. kwi Tina-t det Tina-pst 'the late Tina' If the referent is still alive, kwi cannot be used. This is because the speaker knows that the referent is located somewhere, even if the speaker does not know the exact location. The distal D-determiner can be used if the referent is not visible (and therefore the exact location is likely to be unknown to the speaker). (49) a. Chen kw'ach-nexw ta Peter. Isg.s look-tr(lc) det Peter 'I saw Peter.' b. Chen kw'ach-nexw kwa Peter Isg.s look-tr(lc) det Peter 'I saw Peter.' (not in same room and not visible to speaker) b. * Chen kw'ach-nexw kwi Peter. Isg.s look-tr(lc) det Peter The referents in cases like (48) are as identifiable to the speaker as the referent in (49) is, so identifiability cannot be the relevant feature (or lack thereof). "In some cases the mere 2 0 The distal D-determiner is ungrammatical here, for independent reasons, as I discussed in Chapter 2. The D-determiner kwa can only be used for referents that are human, or somehow made more "interesting". 139 impossibility of the object's conceivably being pointed out by the speaker allows or necessitates the use of an indefinite form [i.e. kwi - CG], even though the object is independently identified by the speaker" (Kuipers 1967: 138). In the same vein, referents that do not yet exist also must be introduced by kwi (50)a, as well as referents which may never have existed (50)b.21 (50) a. Wa chaxw ek' xi-xi-t-em impf 2sg.emph fut redup-laugh-tr-pass t-kwi a-7aw't stelmexw. obl-det redup-future people 'The future generation will be laughing at you.' b. Na7 t-kwi kwekwin' wa yan'-t-m loc obl-det long, time impf take.care-tr-pass ta staw'xwlh yiiu-as-wit, haw k-w-'as det children take.care-3erg-3pl neg irr-impf-3sbj p'i7-t-as-wit kwi stam tina7 t-kwi get-tr-3erg-3pldet what from obl-det haw k-w-'as lhk'i7-s-t-as-wit. neg irr-impf-3sbj know-caus-tr-3erg-3pl 'In the old days they used to warn the children to be careful not to accept anything from anyone they didn't know.' (Kuipers 1967: 219) The fact that kwi is non-locating can also be seen in cases where pictures are involved. Despite involving the same environment given purely with words, as soon as there is a picture to look at, kwi is ungrammatical. In the example below, there were a number of girls, and I was trying to say something about one girl in particular. In this case, kwi is normally given. Instead, the demonstrative was required. (51) a. Na wa seselkw alhi (slheny'-ullh). rl impf lonely/sad dem.f woman-young 'She/the girl's unhappy.' b. * Na wa seselkw kwi slheny'-ullh. rl impf lonely/sad det woman-young 2 1 People who already do exist but do not yet bear a relationship to you are introduced by ta. (i) Chen chem'-us-n ta-n kwtams ek'. Isg.s meet-face-tr det-lsg.poss husband fut 'I met my husband-to-be.' (ii) * Chen chem'-us-n kwi-n kwtams ek'. Isg.s meet-face-tr det-lsg.poss husband fut This is because the speaker can locate the referent at the time of the meeting. 140 In this case, the deictic D-determiners are only licit if the DP is interpreted as referring to the entire group, as I discussed in Chapter 4. The speakers prefer a number-neutral DP to refer to a singular entity, and so the sentence in (52) is dispreferred. (52) ??Na wa seselkw lha slheny'-ullh. rl impf lonely/sad det.f woman-young 'The girls are unhappy.' Past and future time periods are also introduced by kwi. (53) a. c. kwi chel'aklh det yesterday 'yesterday' kwi xaw's det new 'next month' lhkaych' moon d. kwi tepanu det year 'last year' kwi kwekwin' det long, time 'a long time ago' While we might expect distal D-determiners to be used for time distant from the present, only kwi can introduce non-present times. Consultants do not have any intuition as to what a deictic D-determiner + chel 'aklh would even mean. (54) a. * Chen kw'ach-nexw ta Isg.s look-tr(lc) det b. * Chen kw'ach-nexw ta Isg.s look-tr(lc) det mixalh kwa chel'aklh. bear det yesterday mixalh ta bear det chel'aklh. yesterday The only time that can be introduced with a deictic D-determiner is a time period during the present day. (55) a. ti s-tsi7-s b. ti natlh det nom-exist-3poss det morning 'today' 'this morning' (lit: the it is being there) c. ti txw-na-nat det dir-redup-night 'tonight' The obvious question raised by all of this is why time is not locatable in the same way space is. Time and space are often linked, especially in Salish languages. However, time is still 141 more abstract than space, and it is not a necessary result that once something marks distance it will then mark time. Languages should be able to use many different resources to mark time; D-determiners would be one possible way. Within the D-determiner system, the language could still mark whether time was locatable to them or not. Skwxwu7mesh has chosen to mark time as non-locatable; hence kwi is used to introduce non-present time periods. In all cases where the speaker chooses not to locate the referent (either because s/he cannot, or because it is unimportant), the DP is introduced by the non-deictic D-determiner. If the speaker can and wants to locate the referent, any of the other D-determiners or demonstratives can be used instead. In this section, I will show how lacking deictic features can derive the behaviour of kwi. The behaviour discussed here includes narrow scope interpretations, restricted word order, the lack of (non-)uniqueness, and the lack of familiarity or novelty. I also discuss the lack of number marking on kwi, to show that it lacks any potential feature. 4.2 Narrow scope and kwi Any DP introduced by kwi takes narrow scope with respect to many different quantifiers and operators. It takes narrow scope with respect to negation (56). The sentence in (56)a can be continued by the sentence in (b), where there can be no possible referent, but not by (c). (56) a. Haw k-'an i kw'ach-nexw kwi mixalh. neg irr-lsg.sbj prox look-tr(lc) det bear 'I didn't see a bear.' b. Hak mixalh. be.not bear 'There weren't any bears.' c. # Na kway. rl hide 'It was hidden.' It also takes narrow scope under a quantified subject DP (57) or an adverbial quantifier (58). 142 (57) Na mukst-s-t-as i7xw slhen-lhanay' kwi staw'xwlh. rl kiss-caus-tr-3erg all redup-woman det child 'Every woman kissed a (different) child.' ( V > 3, *3 > V ) (58) a. Lhik' chen wa mukwts-t kwi swi7ka. always Isg.s impf kiss-tr det man 'I always kiss a man.' (always > 3, *3 > always) b. Chanat-alh s-en melyi kwi swi7ka. three-times nom-lsg.sbj get.married det man ' I married a man three times.' (3X > 3, *3 > 3X) DPs introduced by kwi also take narrow scope under intensional verbs, as in (59).The sentence in (59) a can be continued by (59)b, but this sentence can only be interpreted to mean that I was unsucessful in finding any boy, not a specific one. (59) a. Chen wa yelx-t kwi swi7ka-7ullh. Isg.s impf lookfor-tr det man-young 'I am looking for a boy.' b. Haw chen k-alh m'i kw'ach-nexw. neg Isg.s irr-times come look-tr(lc) 'I didn't see one.' narrow * 'I didn't see him.' wide 4.3 Non-deictic D-determiners: composition via Restrict Non-deictic D-determiners do not have any features associated with them. This means that they must take obligatory narrow scope. In order to derive this, I appeal to the notion of Restrict (Chung and Ladusaw 2004), which only allows narrow scope interpretation. Narrow scope nominals in Skwxwu7mesh are composed via Restrict.22 In (60), Restrict adds the property of the NP mixalh ('bear') as a restriction on the argument of the predicate kw'achnexw ('see'), leaving that argument unsaturated. 2 2 Werle (2000) argues that St'at'imcets ku is a marker of predicate modification. This is a very similar approach to Restrict. The analysis for kwi will also apply to ku. 143 (60) Chen kw'ach-nexw kwi mixalh. Isg.s look-tr(lc) det bear 'I saw a bear.' (I bear-saw) Existential closure is required to resolve the unsaturated argument of the predicate. (61) VP 3y [see'(yXI) A bear'(y)] Xy [see'(y)(I) A bear'(y)] DP I VP Xx Xy [see'(y)(x) A bear'(y)] Xy Xx [see'(y)(x)] V see' DP bear' Xz [bear'(z)] Restrict The DP kwi mixalh and the verb kw 'achnexw are composed together in such a way that the predicate becomes something like 'bear-see'. Narrow scope nominals, on this analysis, are predicates. The D-determiner does not change the type of the NP predicate. The type of a Restrict-type nominal is therefore . The structure of a narrow scope nominal with a featureless D-determiner is given in (62). (62) DP A,x [bear'(x)] [to be revised] D NP Xx [bear'(x)] D in these cases does not change the type of the NP. The reason why kwi DPs compose via Restrict, instead of some other semantic composition (e.g. choice function/Specify) is because kwi is a non-deictic D-determiner. I claim that only featureless D-determiners can compose via Restrict. If a D-determiner has deictic features, it must be interpreted via Specify. This is because the deictic features are not compatible with a predicative interpretation. Non-deictic D-determiners must be composed via Restrict; anything composed via Restrict must take narrow scope. Deictic D-determiners cannot be composed via Restrict because 2 3 The syntax of the clause in Skwxwu7mesh still needs more research (see Davis 1999 for a discussion of word order in St'at'imcets); I ignore the clause above the VP level. Obviously, to get verb-initial order from the tree here, the verb must raise past the subject. The issue of word order raises many questions of its own. Skwxwu7mesh word order deserves its own dissertation. 144 they are associated with features that do not allow them to be interpreted as a predicate. Anything that does not compose via Restrict can have a wide scope interpretation. Rullmann and You (2003) argue that bare nouns must take narrow scope. They further argue that bare nouns are number-neutral, and suggest that low-scope indefinites can compose via Restrict because they are number-neutral. I extend this idea to deictic features in Skwxwu7mesh. The data in Skwxwu7mesh provide evidence that deictic features do not allow DPs to compose via Restrict. Because kwi DPs are composed via Restrict, they are forced to take narrow scope. This is because the variable in the predicate must be existentially closed within the VP (following Diesing's 1992 insight). For example, under negation, kwi cannot take wide scope. (63) Haw k-'as i silh7-an-t-as kwi sts'ukwi7 ta Peter. neg irr-3sbj prox buy-tr-tr-3erg det fish det Peter 'Peter didn't buy a fish.' - . [3x[fish (x) & buy (x) (P)]] (64) NegP - 3y[buy'(y)(p) A fish'(y)] ^ V P ^ 3y[buy'(y)(p) A fish'(y)] EC VP ty[buy'(y)(p) A fish'(y)] DP V Xx Xy[buy'(y)(x) A fish'(y)] Peter Xy Xx[buy'(y)(x)] V DP Xz[fish'(z)] buy' fish' Restrict This is because the object must be closed off long before the negation can apply. The nominal is within the nuclear scope of negation. 4.4 Predictions The analysis of kwi as a non-deictic D-determiner that composes via Restrict makes certain predictions about its use. Certain word orders are more likely to be used for obligatorily narrow scope nominals (§4.4.1). Similarily, certain contexts will be more likely to allow non-deictic D-145 determiners than others (§4.4.2). The fact that proper names can co-occur with kwi is unexpected for a Restrict analysis of kwi (§4.4.3); I discuss the consequences of proper names for the theory of Restrict. I also discuss consequences for Cheng and Symbesma's (2005) discussion of determiners (§4.4.4), as well as the use of kwi as a complementizer (§4.4.5). 4.4.1 Restrict and word order The analysis of kwi makes certain predictions about its word order. Chung and Ladusaw (2004) argue that Restrict nominals must be closed off within the VP. In that case, we expect that these types of nominals cannot move outside of the VP. As we saw in Chapter 2, word order in Skwxwu7mesh is fairly free, in general. VSO, VOS and SVO are all viable word orders in Skwxwu7mesh. Unlike with the deictic D-determiners, a DP introduced by the non-deictic D-determiner may not be fronted. That is, SVO word order is not acceptable when the DP is introduced by kwi. (65) (66) b. Na lulum [kwi slheny'-ullh]. rl sing det woman-young 'A girl was singing/sang.' [Kwi slheny'-ullh] na lulum. det woman-young rl sing Na lulum [lha/tsi/kwelha rl sing det.f 'A/the girl was singing.' slheny'-ullh]. woman-young non-deictic D non-deictic D deictic D deictic D b. [Lha/tsi/kwelha slheny'-ulh] na lulum. det.f woman-young rl sing 'A/the girl was singing.' A DP introduced by a non-deictic D-determiner cannot be fronted because it must be composed via Restrict. In order for the variable in the predicate to be existentially closed off, it must be closed off within the VP. 146 (67) IP ,24 ??? Xy [girl'(y)] DP; girl' VP EC VP 3x [sing'(x) A P(x)] 9? X VP Xx [sing'(x) A P(x)] DP V Xx [sing'(x)] tj sing' While some languages allow their nominals to move back to inside the VP at LF (e.g. English), Skwxwu7mesh does not.25 4.4.2 Contexts for non-deictic D-determiners I have argued that non-deictic D-determiners are used when the speaker cannot locate the referent. Contexts where a non-deictic D-determiner is more likely to be used are under negation, under quantifiers, in questions, and in intentional contexts. (68) a. Haw k-'as i teh-im' kwi lam'. neg irr-3sbj prox make-act. intr det house 'S/he didn't build a house.' b. Lhik' na teh-im' kwi lem-lam'. always rl make-act.intr det redup-house 'S/he's always making houses.' c. Nu chexw kwi e-lam'? • rl.Q 2sg.s det 2sg.poss-house 'Do you have a house?' d. N-s-tl'i7 kwi-n-s teh-im' kwi lam'. 1 sg.poss-nom-dear comp-lsg.poss-nom make-act.intr det house 'I want to build a house.' In factive contexts, it is more likely that the speaker will be able to locate the referent, so deictic D-determiners are more likely to be found. 2 4 IP may not be the right label for the category where subjects move to. More work is required on the functional domain of the clause in Skwxwu7mesh. 2 5 This seems to be consistent in languages with relatively free word order, which tend to mark scope by means of surface word order. 147 In the five texts collected by Kuipers (1967), this trend can be seen. The neutral D-determiners are used far more than any other D-determiner, as is to be expected for a neutral element. The proximal D-determiner is the next most common D-determiner. The non-deictic D-determiner is used more than the distal D-determiner. Recall that the distal D-determiner must be used for invisible referents, and is usually only used for place names or for reference to humans. The low number reflects these other contraints. proximal26 neutral distal, invisible non-deictic gender-neutral 22 121 8 16 female 0 1 0 3 total 22 122 8 19 Table 5.1: Num )er of occurences of D-determiners in the texts collected by Kuipers (1967). That distal elements are used in the texts can be seen in the number of distal demonstratives used, which are by far the most of any of the demonstratives. proximal medial distal gender-neutral 3 2 85 female 0 19 13 plural 1 0 0 total 4 21 98 Table 5.2: Number oi ~occurences of demonstratives in the texts collected by Kuipers (1967). More important is the contexts in which kwi and kwes are used: in more than half the cases (11/19), they are used in non-factive environments, such as under negation, with the future marker, and in conditional clauses. The neutral D-determiner ta, on the other hand, is rarely used in non-factive contexts (9/122). 4.4.3 Propernames If, as I am arguing, kwi must compose via Restrict, then I predict that proper names should not be able to co-occur with kwi. However, we already saw that this was perfectly licit, as long as the referent was deceased. 2 6 As Kuipers originally characterized this as a "strong" determiner (a demonstrative), this count may be inflated. I discounted the one instance where ti occurred without a following NP, where is most certainly behaving as a demonstrative. 2 7 This count does not include the instances of kwi as a complementizer. 148 (69) Chen lhk'i7-s kwi Tina-t. Isg.s know-caus det Tina-pst T knew the late Tina.' The sentence in (69) should mean something like 'I Tina-knew'. Worse, the non-deictic DPs in this case can take wide scope. (70) Haw k-'an i lhk'i7-s kwi Tina-t. neg irr-lsg.sbj prox know-caus det Tina-pst 'I didn't know Tina.' * 'I didn't know any Tinas.' This is a potential problem for my analysis so far. However, I claim that proper names are already of type e (and are therefore scopeless). The D-determiner does not have any effect on this; kwi does not change Tina into a predicate. If it is already of type e, it cannot compose with the predicate via Restrict. What does this tell us about Restrict? Restrict appears to be a repair strategy, brought in when there is nothing to change the type of the NP to e, or to type-shift via a choice function (Specify). The non-deictic D-determiner cannot change the type of its NPs, nor does it allow the DP to compose via Specify. The predicate is left with no choice but to compose via Restrict, and to force the nominal to become a predicate modifier. 4.4.4 Individuation, subordinating, and non-deictic D-determiners Cheng and Sybesma (2005) claim that the most basic function of a determiner is its deictic function. They further claim that the "subordinating" function of determiners (i.e., the ability of determiners to create arguments) and the "individuating" function (i.e., the ability to pick out an individual from the NP predicate) both arise from the deictic function. "We think that these two functions which D is supposed to perform (individuation, syntactic subordination) are closely related to, or even different manifestations of, a more fundamental property of the DP domain: its deictic property - the property to be able to refer at all" (Cheng and Sybesma 2005). They claim that there is a division of labour between the lexical and functional domains: lexical units describe and functional units refer. The subordinating function for them appears to be dependent 2 8 N o t e t h a t , as f a r as I c a n t e l l , t h e y a r e o n l y r e f e r r i n g t o D - d e t e r m i n e r s , a n d n o t q u a n t i f i e r s , w h i c h d o n o t r e f e r t o i n d i v i d u a l s . 149 on the rest: the individuation arises from the deictic features, and the subordination arises from the individuation. I schemetize this idea below. (71) a. If a determiner subordinates, then it must individuate, b. If a determiner individuates, then it must be deictic. However, I explicitly claim that kwi lacks deictic features. It does, however, create an argument. Therefore, the deictic features cannot be driving the subordinating function. On the other hand, the individuating function does seem to arise from the deictic features, as kwi does not pick out referents in the same way the rest of the D-determiner system does. I assume that individuation can involve type shift from type to e; in other words, be a choice function. Kwi does not do either of these things. Adapting Cheng and Sybesma's (2005) description of the functions of D, I claim instead that D must at least subordinate (thereby creating an argument). It may also individuate, but only if it has deictic features. The two functions of D-determiners (subordination and individuation) must therefore be separate. 4.4.5 Non-deictic D-determiners as complementizers I argued above that kwi is only used for referents that cannot be located in space. I also showed that kwi is used as a complementizer in Skwxwu7mesh. (72) Ha71h kwi-s paym-chet. good comp-nom rest-lpl.poss 'It'd be good if we rest.' I also argued that this was not because kwi occupied a different position than the other D-determiners. If they do occupy the same position, then why is it only kwi that is used as a complementizer? I argue that kwi is used because it does not locate in space. Events are locatable in time and space; however, states are difficult to. locate in space. The most likely candidate for a complementizer from the D-determiner system then is one with the fewest number of features: the non-deictic D-determiner then is the best choice, as it has none. I predict that if temporal locatablity were to be encoded in the complementizer system, then other D-determiners would likely be used. 150 4.5 Summary The D-determiner kwi does not encode information about uniqueness or familiarity. If this analysis is correct, then it raises the question of what, if any, features kwi does have. Further, if the D-determiner does not encode any features, then what does it mean to be a D-determiner? I have argued that kwi is non-deictic. Previous analyses of Skwxwu7mesh have not addressed this issue directly. The labels of "indefinite" and "invisible" have been applied to kwi (by Kuipers 1967 and Currie 1997, respectively). However, these descriptions fall to explain why 29 kwi only takes narrow scope, and kwa only takes wide, as we saw in (2), repeated here. (2) a. Haw k-'an i silh7-an ta/ti sts'ukwi7. neg irr-lsg.sbj prox buy-tr det fish 'I didn't buy a fish.' (wide or narrow) b. Haw k-'an i silh7-an kwa sts'ukwi7. neg irr-lsg.sbj prox buy-tr det fish 'I didn't buy a fish.' (wide only) c. Haw k-'an i silh7-an kwi sts'ukwi7. neg irr-lsg.sbj prox buy-tr det fish 'I didn't buy a fish.' (narrow only) I argue that kwi does not have any features. Only the non-deictic analysis proposed here can account for this data. The D-determiner kwi does not encode any features which could force it to be individuated in any way (such as [proximal], [invisible], etc., or assertion of uniqueness). I claim that these features force deictic DPs to compose with the predicate via Specify. I further claim that the lack of a choice function is what forces a non-deictic DP to take narrow scope. If a D-determiner (or any other part of the functional domain) provides any of these features, it must be able to take wide scope, because it composes via Specify. If a D-determiner asserts uniqueness, then it is forced to take wide scope. This raises the question of why this correlation between lack of features and scope should exist in the first place. The intuitive answer is that nominals which compose via Restrict are still predicates, regardless of whether they are introduced by a D-determiner or not. Featureless D-2 9 Recall that kwa can only be used if the referent is human, or has been made "interesting" enough. This need to be interesting may be why kwa must take wide scope. See Tunstall (1998) for discussion of the link between being "interesting" and taking wider scope. 151 determiners do not change the type of the NP. The link between featureless D-determiners and scope is therefore indirect; featureless D-determiners (and nominals lacking determiners altogether) must compose via Restrict. Anything else will compose normally. 5 Featureless D-determiners and the context So far, I have shown that kwi DPs must take narrow scope. I argued that this was because kwi DPs lack features and that this prevents them from being able to take wide scope. Bare nouns have also been argued to take obligatory narrow scope (Carlson 1980). Potentially, these also lack features that prevent them from taking wide scope. However, this tells us nothing about whether bare nouns are in fact "bare" or not. I have argued above that structure, in itself, does not matter. I claimed that the presence of features allowed a nominal to take wide scope. (73) a. if DP -> able to take wide scope, b. if DP —> obligatory narrow scope [±F] c. if NP —> obligatory narrow scope Instead of the schema in (73) above, it could be that bare nouns be introduced by a null, featureless D. The non-deictic DPs and bare nouns would be predicted to behave essentially the same. (74) a. if DP -> able to take wide scope, b. if DP ~» obligatory narrow scope I [±F] In this section, I show that this cannot be the case, on the basis of the availability of partitive readings with kwi. 152 5.1 D-determiners matter So far, it appears that kwi DPs and bare nouns (bare plurals and bare numberless nominals) behave similarly.30 They both involve fewer features than ta DPs (or other DPs with features). Bare nouns get narrow scope interpretations, at least in English (Carlson 1980), Chinese (Rullmann and You 2003), Brazilian Portuguese (Muller 2005) and Blackfoot (Glougie 2000), just as kwi DPs do. It could be possible that kwi DPs and bare nouns are even more similar: that the structure involved in both cases is the same. Both bare nouns and kwi DPs could have a D position that is featureless. This featureless D position would mark the NP as composing via Restrict and would explain the scope facts for both. (75) a. bare noun b. full DP DP DP D NP D NP 0 | kwi | N N There would be three possibilities here, were we to adopt the structure in (75)a for bare nouns: i) kwi has meaning and the null D is semantically null, ii) both the null D and kwi are semantically null or iii) both the null D and kwi have meaning. I argue that the first two possibilities are untenable. As Wiltschko (to appear) argues, phonologically null elements must have meaning. The null D position should have some semantics. This leaves us with the third possibility: that kwi and the null D have the same semantics.31 I argue that this third possibility is also untenable. This is because bare nouns and kwi DPs have different semantics. I argue that bare plurals in English are actually barer than full DPs. Bare plurals lack the D position. This has implications for how and where bare plurals can be used. 3 0 Bare numberless nominals are nominals without any overt number marking which can receive either a singular or plural interpretation, as in Mandarin (Rullmann and You 2003), or Brazilian Portuguese (Muller 2005). 3 1 However, see Sobin (1985), among others, who argues that there are null expletives in Slavic. 153 (76) a. bare noun b. full DP N P 3 2 DP N D NP kwi | N I argue that kwi contributes something semantically, based on its behaviour in certain contexts, and that bare nouns lack this semantics. I extend this idea by postulating that the correlation between a D and the semantics of kwi is universal. All D-determiners in Skwxwu7mesh share the core semantics of domain restriction. (77) a. M = APf(Xx[P(y)AC(x)]) b. pwiJJ = XP Xx [P(y) A C(x)] I also argued that English the contains this core semantics. (78) [[the]] = XP max(Xx [P(x) A C(X)]) I therefore hypothesize that all D-determiners share this core semantics. Bare nouns must introduce a new discourse referent. DPs must refer back to a previously introduced referent. (79) a. I saw four women walk into the store. # Women were looking for clothes. b. I saw four women walk into the store. The women were looking for clothes. Kwi DPs behave like other DPs in Skwxwu7mesh in that they can refer back to a previously introduced referent. Further, unlike the deictic D-determiner ta, kwi can refer to a part of the set already introduced in the discourse as (80). That is, kwi can have a partitive reading. (80) Xa7titsn slhanay' na mi uys. four woman rl come inside Chen kwikwi -s k w i slhanay'. Isg.s talk-caus det woman 'Four women came in. I talked to one of women.' 3 2 1 am agnostic as to the presence of any intervening syntactic nodes, such as NumP or (|)P (see Dechaine and Wiltschko 2002 for arguments that some nominal phrases are bigger than NP but smaller than DP). It is only necessary that bare nouns lack D-determiners. 1 5 4 The D-determiner kwi does not have to refer to a new referent, as we might expect with an obligatorily narrow scope nominal. When a deictic D-determiner is used instead, the DP strongly prefers to be interpreted as referring to the entire set, as we saw in Chapter 4. (81) Xa7utsn slhanay' four woman ?? Chen kwikwi-s Isg.s talk-caus na mi uys. rl come inside lha slhanay'. det.f woman 'Four women came in. I talked to all of the women.' When the context allows for only one individual, then a deictic D-determiner may be used. If the context for (82) has more than one boy (two boys and one girl), then kwi must be used. Otherwise, if there is only one boy (two girls and one boy), ta can be used. (82) a. Chanat men' -men wa three redup-child impf Chen kwikwi-s Isg.s talk-caus kw'shetsut na7 play be. there kwi swiw'lus. det boy ta det b. ta det atsk. outside atsk. outside 'Three children were playing outside. I talked with a boy. Chanat men'-men wa kw'shetsut na7 three redup-child impf play be.there Chen kwikwi-s ta swiw'lus. Isg.s talk-caus det boy 'Three children were playing outside. I talked to the boy.' Other examples of kwi being used partitively are given below. (83) a. Chen wa lhem-nta schi7i. Isg.s impf pick-tr det strawberry Chen hiiy-s kwi schi7i. Isg.s finish-caus det strawberry 'I picked strawberries. I ate one strawberry.' b. Chen teh-im' ta slhum'. Isg.s make-act.intr det soup Chen huy-s kwi slhum'. Isg.s finish-caus det soup 'I made some soup. I ate some of the soup.' The ability for kwi to be interpreted partitively, unlike the deictic D-determiners, arises from its lack of deictic features. As I discussed above, non-deictic D-determiners can only be 155 used in contexts where the speaker cannot locate the referent. Partitive contexts allow the use of kwi because the speaker does not locate the indvidual within the group. The speaker is behaving as if he or she cannot locate the referent; the referent can be any member of the group. If the speaker wishes to provide information about the location, then he or she must use a demonstrative. The use of kwi in the cases above is only acceptable because the speaker is not providing information about the location. Unlike demonstratives, which can also be used partitively, kwi does not refer to a particular object, that the hearer will also likely be able to locate. This behaviour of kwi, where it can be used in partitive contexts, is in opposition to the behaviour of bare nouns, as we will see below. I argue that D-determiners must be interpreted with respect to some contextual domain in a way that bare nouns are not. The D-determiner kwi restricts the domain of its NP. Empirically, we can see that bare nouns cannot have their domain restricted. I claim that this is a result of the lack of the necessary syntactic apparatus: a D-determiner. 5.2 Bare nouns Bare nouns provide crucial evidence that D-determiners do something important. This is because they do not have the same properties as DPs. In some respects, they behave similarly to kwi DPs. However, they behave differently in a crucial way. 5.2.1 Bare nouns, narrow scope and Restrict Like kwi DPs, bare nouns (when interpreted existentially) are interpreted with narrow scope (84). Bare plurals, bare singulars (or bare numberless nominals) and incorporated nouns all take narrow scope with respect to negation in each example. (84) a. bare plural John didn't see spots on the floor. (Carlson 1980:19) (neg > 3, *3 > neg) 156 bare singular ni-maats-iyapi-hpa 1-neg-pst.see. intr-lnonaffirm 'I didn't see an eagle.' (neg > 3, *3 > neg) piita. eagle (Blackfoot; Glougie 2000:127) bare singular El ninito no trajo det boy neg brought 'The boy didn't bring a ball.' (neg > 3, *3 > neg) bare singular anu kitaab nahiiN paRhegii. Anu book not read 'Anu won't read any book.' (neg > 3, *3 > neg) pelota. ball (Spanish; Miller and Schmitt 2005: 92) (Hindi; Dayal 1999) bare numberless nominal Joao nao viu mancha Jodo neg see spot 'Joao didn't see spots on the floor.' (neg > 3, *3 > neg) (Brazilian Portuguese; Schmitt and Munn 1999) no on chao. floor f. incorporated noun Juuna Kaali-mit allagar-si-nngi-l-a-q. J.abs K-abl letter-get-neg-ind-[intr]-3sg ' Junna didn't get a letter/letters from Kaali.' (Inuktitut; Bittner 1994:118) (neg > 3, *3 > neg) This is in contrast to other nominals in each system. In some languages, the full DPs can take narrow or wide scope. (85) a. John didn't see a spot on the floor. (Carlson 1980:19) (neg > 3, 3 > neg) b. El ninito no trajo una pelota. det boy neg brought a ball 'The boy didn't bring a ball' (Spanish; Miller and Schmitt 2005: 92) (neg > 3, 3 > neg) 157 c. anu ek/koii kitaab nahiiN paRhegii. Anu one/some book not read-F 'Anu won't read a book.' (Hindi; Dayal 1999) (neg> 3, 3 >neg) In other languages, full DPs or non-incorporating nouns must only take wide scope. (86) a. ni-maats-ino-a-waatsiiks om-i piita 1 -neg-see-3-nonaffirm dem-3' eagle 'I didn't see the/an eagle.' (Blackfoot; Glougie 2000:127) (*neg > 3, 3 > neg) b. Taqqialu-up tuktu taku-lau-nngit-t-a-(ng)a Taqqialu. -erg caribou(abs) see-pst-neg-part-ftr]-3sg.erg. 3sg.abs 'Taqqialu didn't see a caribou.' (Inuktitut; Wharram 2003:39) (*neg > 3, 3 > neg) Having a narrow scope option does not force all other nominals to take wide scope. In all the above cases, the bare nominals are forced to take narrow scope because, like kwi, they lack the features that would allow them to take wide scope. They do not encode deictic information, number, or, as we shall see below, familiarity. I claim that, like kwi DPs, bare nouns take obligatory narrow scope because they compose via Restrict (following Chung and Ladusaw 2004). In sentences lacking any other operator, bare nouns (like kwi DPs) will take narrow scope with respect to existential closure. I do not, therefore, adopt Carlson's (1980) account of bare nouns in English. Carlson claims that bare nouns denote kinds, which are a type of individual (see also Carlson 1989 and Chierchia 1998). Due to the realization relation adopted by Carlson, the kind analysis and the Restrict analysis end up truth-conditionally equivalent. However, the kind analysis does not apply straightforwardly to Skwxwu7mesh. This is because kind readings of DPs are difficult to get. The speakers will produce sentences like (87), but when asked for a translation of the sentence (87)a, they will always give an episodic translation. For (87)b, they often give a "specific" reading. (87) a. Na wa sik kwi/ta kalaka. rl impf fly det crow (i) 'Crows fly.' (ii) 'The crow is flying.' 158 b. Ha71h-s chen kwi/ta pesh-push. good-caus Isg.s det redup-cat (i) 'I like cats.' (ii) 'I like the cats.' Note that either ta or kwi is used to translate an English generic sentence. I therefore cannot adopt Carlson's analysis of bare nouns for kwi DPs. As the Restrict and kind-plus-realization analyses are truth-conditionally equivalent, I adopt a Restrict analysis for English for consistency with Skwxwu7mesh. (See Wilkinson 1991, Diesing 1992, Gerstner and Krifa 1993, and Kratzer 1995 for arguments that bare nouns are ambiguous between a kind reading and an indefinite reading.) 5.2.2 Bare nouns and the lack of context As I have already argued in Chapter 3, bare nouns are not sensitive to the context. They cannot refer to a subset of a previously mentioned set. This is 'non-specificity' as defined by Enc (1991). In (88) - (90), a bare noun cannot be used after the set under discussion has already been introduced. (88) bare plurals A: There are five children playing in the yard. B: What are they doing? A: # Boys are digging in the sand. (89) incorporated nouns A: Nillataartitsivim-mi tallima-nik manne-qar-p-u-q. fridge-loc five-inst.pl egg-have-ind-fintrj'-3sg 'There are five eggs in the fridge.' B: Jensi-p uku-nannga qassi-t neri-ssa-v-a-i? Jensi-erg dem-abl.pl how.many-abs.pl eat-fut-inter-[tr]-3sg.3pl 'How many from those will Jensi eat?' A: # Jensi marlun-nik manni-tu-ssa-q.33 Jensi.abs two-inst.pl egg-eat-fut-ind-[intr]-3sg 'Jensi will eat two eggs.' * 'Jensi will eat two of the eggs.' (West Greenlandic; van Geenhoven 1998) 3 3 This is not independently bad because of the stranded marlun; Inuktitut allows this kind of incorporation (van Geenhoven 1996). 159 (90) bare singular ?? Det var igjen mange sykler etter salget, sa jeg ga it was left many bikes after sale-def so I gave sykkel til Kari. bike to Kari 'There were many bikes left after the sale, so I gave Kari a bike (any bike).' (Norwegian; Borthen 2003: 29-30) There are some bare nouns that can be used anaphorically. However, this only occurs in languages that lack overt D-determiners.34 In Hindi, for example, bare nouns can receive a definite interpretation. The bare plural bacce 'children' in the second sentence appears to refer to the entire set introduced in the first sentence. (91) kuch bacce andar aaye. bacce bahut khush the. some children inside came children very happy were 'Some children came in. The children were very happy.' (Hindi; Dayal 1999) I set these examples aside here, and discuss articleless languages further in Chapter 6. Bare nouns are unable to be used in partitive contexts. They are usually used in novel contexts. Bare nouns in languages which otherwise use overt D-determiners, or in languages which allow noun incorporation, cannot refer back to a subset or the entire set. 5.3 The contribution of D-determiners to DPs If a D-determiner can be featureless, and yet still allow a nominal to be used in a partitive context, then what exactly is the D-determiner doing? We know that D-determiners can do at least the following four things: i) create an argument out of a predicate, ii) assert maximality, iii) encode specificity/non-specificity or deictic features (or some other contrast), and iv) restrict the domain from which the individual is picked out. The non-deictic D-determiner in Skwxwu7mesh provides us with evidence that all D-determiners restrict the domain. It does not have any features, and is composed via Restrict, yet it still has domain restriction in its denotation. (92) a. p w i ] = XP Xx [P(x) A C ( X ) ] 3 4 Mithun (1984) discusses examples of incorporated nouns that also receive definite interpretations. I set them aside and leave them for further research. 160 b. [kwi mixalh] = Xy [*bear'(y) A C(y)] On the basis of this, I argue that D-determiners must at least create a syntactic argument out of a predicate (cf. Higginbotham 1985, Szabolsci 1987, 1994; Stowell 1989; Longobardi 1994) and restrict the domain. The purely syntactic requirement (that the D-determiner create a syntactic argument) is obviously not a universal requirement. Some languages (like Skwxwu7mesh) require a D-determiner to create a syntactic argument. Kwi does not change the semantic type of the nominal, and yet is required because Skwxwu7mesh requires that all arguments have a D. This must be language-specific and not related to any semantic function. In English, for example, there is no such syntactic requirement, and so it allows bare nouns as arguments. (93) a. I saw bears. b. [bearsJJ = Xx [%ear'(x)J c. [I saw bears] = 3y [see'(y) A ®bear'(y)J The non-deictic D-determiner kwi creates an argument (as it is one of a set of elements that are obligatory with arguments), and restricts the domain of the NP. Bare nouns, while being arguments, do not have a restricted domain. As we saw above, kwi DPs must be different from bare nouns, since kwi can be used in instances where the domain must be restricted. (94) a. Chen men sk'i7-s kwi men nch'e-nch'u7. Isg.s just know-caus det just redup-one 'I knew one of them.' 3y [know'(y)(I) A one'(y) A C(y)J b. Chen kwikwi-s kwi swiw'lus. Isg.s talk-caus det boy 'I talked to a boy.' 3y [talk'(y)(I) A *boy'(y) A C(y)J This can and must be extended to all uses of kwi. If there is a D, this D must introduce C. 161 (95) Chen kw'ach-nexw kwi mixalh. Isg.s look-tr(lc) det bear 'I saw a bear.' 3y [see'(y)(I) A *bear'(y) A C(y)] Bare nouns are not context-dependent, and therefore cannot introduce C. This is due to the lack of a D-determiner. (96) I saw bears. 3y [see'(y) (I) A @bear'(y)] * 3y [see'(y) (I) A ®bear'(y) A C(y)] Earlier, I provided no denotation for kwi whatsoever ((62), repeated here). (62) kwi DP A,x [bear'(x)] [to be revised] D NP Xx [bear'(x)] However, in this section, we have seen evidence for kwi having domain restriction in its denotation. As it lacks any other features, the non-deictic kwi is only a function which introduces C. (97) ^JDP^ Xz [*bear'(z) A C ( Z ) ] AP Xy [P(y) A C(y)] D NP A.x [*bear'(x)] I I kwi mixalh C is a predicate of type , and is the domain restriction on the NP. The non-deictic D-determiner is still different from the deictic D-determiners. It stands out as different from the rest in that it can more easily refer to a subset of the NPs in the given context: that is, it get a partitive reading without any extra work. This D-determiner can "see" inside of the set. I claim that it able to do this because it is non-deictic. When a speaker uses kwi they are making a claim that they cannot locate the referent. If the referent is within a larger group, the speaker is not locating the referent individually, just making a claim that that referent is somewhere within that previously mentioned group. 162 p 5 . 4 Implications So far, I have made two main claims. First, if a nominal is not associated with any features, such as deixis, presupposition of uniqueness, etc., it must take obligatory narrow scope. Having more structure does not mean being able to take wider scope; having more features does. Structure does not determine scope - only the presence or absence of features do. (98) a. if DP —» able to take wide scope, b. if DP —» obligatory narrow scope I [±F] c. if NP —• obligatory narrow scope This means that the presence of the head D, by itself, does not matter for scope purposes, only the features associated with that D projection. On the other hand, my second claim is that bare nouns are structurally different from kwi DPs: they lack the D head. Bare nouns then are not DPs, but rather NPs (or something in between). (99) D introduces domain restriction; NPs lack domain restriction. All D-determiners must introduce C - that is, they must be constrained by the context. If there is no context, and the D-determiners do not assert the uniqueness of their referents, then the referents can be accommodated into C. Bare nouns are not be constrained by the context because they lack a D-determiner. All of this raises questions about the nature of D-determiners. Are there any semantically null D-determiners? That is, do bare nouns have null D-determiners, present only to create a syntactic argument? The answer must be no, if D-determiners really do introduce C. Vergnaud and Zubizarreta (1992) claim that expletive determiners are used in constructions where they are syntactically required, but provide no semantic information. An example of an expletive determiner is given below, in the Construction of Inalienable Possession in languages like Spanish. (See also §6.4, where I show their analysis cannot apply to kwi.) (100) Los ninos volvieron la cabeza. (Spanish; Baauw 2001:3) det.pl boys turn det head 'The boys turned their heads.' In (100), la cabeza 'the head' does not refer to a specific head in the discourse. Instead, it is interpreted distributively with respect to the subject of the sentence. 163 In cases such as these, we can reanalyze the expletive determiner as a D-determiner that introduces C: here the domain would be the set of boys. The D-determiner does not enforce a singular reading here, because the DP is bound by the subject DP. (101) DP max(^z [head'(z) A C(z)])35 max(Xy [P(y) A C(y)]) D NP Xx [head'(x)] . I I la cabeza It is not clear why all languages cannot do this, however. A second environment where these 'expletive' determiners are used is the generic environment. In languages which do not allow bare plurals to be generic, a determiner is inserted. (102) Los leones son carnivoros. (Spanish; Baauw 2001:3) det.pl lions are carnivores 'Lions (in general) are carnivores. The "expletive" determiner here could be a D-determiner, "restricting" the domain to the entire domain of lions. Recall that the domain includes all members of the domain of individuals (De) in novel contexts. C is not narrowing the domain in these cases. The G E N operator (Krifka et al 1995) allows us to understand the generic DP as including any lion-entity. (103) G E N [max (Xx [*lion'(x) A C ( X ) ) ] [*carnivore'(x)]] C l o s |e0nes = {De} Then the question becomes why generics in other languages are not introduced by a D-determiner. English generics sometimes lack D-determiners. (104) a. Crows are black. b. * T h e crows are black, (for generic reading) c. T h e crow is black. 3 51 assume that Spanish la asserts maximality; it is only crucial that the D-determiner have domain restriction. 3 6 The real question might be why these languages are not required to have possessive morphology, since possessives are arguably a kind of D-determiner anyway. So these kinds of determiners are not 'expletive', but do lack person features. 164 If, as I claim, all D-determiners have domain restriction, (104)c must also involve domain restriction, whereas (104)a must not. In the formulae below, example (104)a corresponds to (105)a and (104)c corresponds to (105)b. (105) a. G E N [*crow'(x)] [black'(x)] b. G E N [max(A,x [crow'(x) A C(X))]] [black'(x)]] For languages which require arguments to be marked as such in the syntax (like Skwxwu7mesh or French), the necessity of the presence of the D-determiner is explained. English bare singulars are not licit arguments, and so the presence of the D-determiner is also explained. The point is merely that D-determiners still have the same semantic core, regardless of the context they are used in. 3 7 Finally, 'expletive' determiners are used for proper names. (106) a. O Nikos agapai ti Maria. det Nikos loves det Mary 'Nikos loves Mary.' (Greek; Marinis 1997: 171) b. Chen kw'ach-nexw lha Kirsten. Isg.s look-tr(lc) det.f Kirsten 'I saw Kirsten.' (Skwxwu7mesh) My analysis raises the question of the correct treatment of proper names, and touches on large and unsolved problems in the literature on proper names. Here I offer some speculations of how proper names and domain restriction could interact. In example (106)b above, the D-determiner lha provides the context by which the name can be evaluated, and gives us the particular Kirsten we are talking about. Therefore, there could always be more than one Kirsten in the context. The D-determiner lha narrows the context down 3 7 1 predict that in familiar contexts, the "generic" DPs will no longer be interpreted generically, but rather as referring to a previously introduced referent. This prediction is bom out in English. (i) The lion is a carnivore. (generic) (ii) I saw a lion, a panda and a lizard at the zoo yesterday. T h e lion is a carnivore, (familiar) However, it is not bom out in French. (i) J'ai vu des ours hier soir. Ils erraint dans Stanley Park. I. have seen • of.the bears last night they wander in Stanley Park J'adore les ours. I.love det bears 'I saw some bears last night. They were wandering around in Stanely Park. I love bears.' (Hamida Demirdache, p.c.) The status of generics in languages is still unclear, and deserves more discussion than that given here. 165 to the one under discussion. If there is only one such person in the world (like Gwyneth Paltrow's baby Apple, who is presumably the only person with that name), the domain restriction would end up being vacuous. Historically, it is unclear whether people ever shared names in the Skwxwu7mesh community. Within the last 200 years, it was uncommon for people to share the same name, but it did still occur, perhaps because some of the names were lost (Peter Jacobs, p.c.) Even if there had been a ban on name-sharing, the D-determiner would still behave as if there were no such ban. Matthewson (1998) claims that names are predicates in Salish; on this view the D-determiner would already be necessary to create an argument. However, I claimed above that names are not predicates, but entities. I also claim that the requirement that all arguments have D-determiners is still playing a role. Again, Salish languages require arguments to be marked syntactically as such, regardless of their semantic type. I further claim that the D-determiner is not only creating a syntactic argument, it is also narrowing the domain. (107) a. PiaJJ = XP fkx [P(y) A C ( X ) ] b. [lha Kirsten] = f(?ix [Kirsten(y) A C ( X ) ] ) This leads us to another question. What about languages which lack D-determiners with proper names (as in English)? There are two options for the analysis of D-determinerless proper names. Proper names could differ across languages, where some require a D-determiner to create an argument (like Salish), whereas others do not require a D-determiner. This is similar to the expletive analysis. As a result the proper name in a Salish-type language would also have its domain narrowed, and in an English-type language, the proper name would not have its domain narrowed. (108) Argument parameter: Setting A: all arguments require D-determiners (regardless of semantic type) Setting B: arguments do not require D-determiners This analysis may explain why some languages seem to allow D-determiners, but do not force them. If a language has Setting B, the language is free to add D-determiners to any argument (including proper names). An example of a language which allows D-determiners (but does not require them) is Italian. 166 (109) a. Gianni mi ha telefonato. (Longobardi 1994:622) Gianni me has called 'Gianni called me up.' b. II Gianni mi ha telefonato. det Gianni me has called 'Gianni called me up.' Here the D-determiner can be overt or covert. I take this as evidence that the D-determiner position is always available for proper names. This leads to an analysis where all proper names have D-determiners. In this case, I predict that phonetically null D-determiners would restrict the domain of "articleless" proper names. Longobardi (1994) argued that null determiners were possible; however, the interpretation of his null D was existential. This is the wrong result for a proper name, under any normal assumption. Instead, he claimed that proper names moved to the empty D position, substituting for the D position. (110) John If all proper names have (c)overt D-determiners, the N must not substitute for the D position because the D position must be able to restrict the domain. Further, this null D would only be phonetically null, not semantically null. So if there is N to D movement, it will involve adjunction, rather than substitution. (Ill) max(Xx [P(x) A C(x)]) D NP John N D max(kx [P(x) A C(x)]) John The fact that proper names appear to move to D (in Italian) could be motivated by a ban on phonetically null D positions. 167 6 A l te rna t i ve analyses? Now that I have demonstrated what claiming that kwi is a non-deictic D-determiner can explain, it is necessary to show that any other analysis will run into problems when accounting for the data. 6.1 Kwi is not like ku in St'at'imcets An obvious analogy to kwi in Skwxwu7mesh is ku in St'at'imcets, described by Matthewson (1998). There are immediately apparent differences between these two elements. First, ku can only be used in polarity contexts (112), or as the object in morphologically intransitive clauses (113). (112) a. tup-un'-as s-John punch-tr-3erg nom-John 'John hit a policeman.' b. * tup-un'-as s-John punch-tr-3erg nom-John ti det ku det plismen-a. policeman-exis plismen. policeman (St'at'imcets; Matthewson 1998) (113) cuz' k'ac-cal ku going, to dry-act.intr det 'Lemya7 is going to dry saskatoons.' stsaqwem kw s-Lemya7. saskatoon det nom-Lemya7 (St'at'imcets; Davis and Matthewson 2003) St'at'imcets ku may also be used on non-arguments, such as inside complex predicates, as we saw in §2.1. This is very different from Skwxwu7mesh kwi, which may only be used with arguments. I repeat the example (31) below. (31) a. [gelgel(ku) sqaycw] strong det man 'John is a strong man.' b. [iyim swi7ka] ta strong man det 'John is a strong man.' c. * [iyim kwi swi7ka] strong det man kw-s det-nom John. John ta John. det John John. John (St'at'imcets; Matthewson 1998) (Skwxwu7mesh) (Skwxwu7mesh) 168 Ku may also appear inside a relative clause, on the head, provided it is head-final. In Skwxwu7mesh, head-final relative clauses do not seem to be grammatical anymore. (See Kuipers 1967 for examples of head-final relative clauses.) In initial position, kwi may not co-occur with a deictic D-determiner. (114) a. ats'x-en-lhkan [ta [xzum-a see-tr-lsg.s det big-exis 'I saw a big bird.' Chen kw'ach-nexw [ta Isg.s look-tr(lc) det 'I saw the man who is big. Chen kw'ach-nexw [ta Isg.s look-tr(lc) det (ku) spzuza7]]. det bird (St'at'imcets; Matthewson 1998) swi7ka[na hiyi]]. man rl big (Skwxwu7mesh) kwi swi7ka[na hiyi]]. det man rl big (Skwxwu7mesh) St'at'imcets ku may also be used with demonstratives. This is not true of kwi • 38 (115) a. b. tecwp-min-lhkan buy-appl-lsg.s 'I bought that car.' Chen silh7-an Isg.s buy-tr 'I bought that fish. Chen silh7-an Isg.s buy-tr til ku dem det kaoh. car (St'at'imcets; Matthewson 1998) tay' sts'ukwi7. dem fish tay' kwi sts'ukwi7. dem det fish (Skwxwu7mesh) (Skwxwu7mesh) Furthermore, ku may be used inside certain adverbial clauses. This is also not true of kwi. (116) a. b. uxwal'-lhkan (ku) xwem. go.home-Isg.s det fast 'I went home right away, quickly. Chen tskwatsut Isg.s run 'I ran quickly.' Chen tskwatsut Isg.s run ts'ats'i7x. quickly kwi ts'ats'i7x. det quickly (St'at'imcets; Matthewson 1998) (Skwxwu7mesh) (Skwxwu7mesh) 3 8 However, demonstratives and D-determiners never co-occur, so this difference may be irrelevant. 169 Whatever ku is, it is not the same as kwi. Some of the properties of ku may arise from the fact that it is also non-deictic (see Matthewson 1998, who argues exactly this), but it must also have other properties (such as polarity). See also Chapter 6 for more discussion of ku. 6.2 Kwi is not Longobardi's existential D Since kwi takes obligatorily narrow scope, we may expect that it is the pronounced version of Longobardi's (1994) existential D-determiner. His description of the behaviour of the empty determiner in Italian is given below. (117) Empty determiners may occur at S-Structure in Italian only under the following conditions: a. They are restricted to plural or mass nouns like several other determiners. b. They are subject to a lexical government requirement like other empty heads. c. They receive an indefinite interpretation corresponding to an existential quantifier unspecified for number and taking the narrowest possible scope (default existential). (Longobardi 1994:617) Longobardi also claims that the empty D-determiner cannot be used for proper names, days, months, etc. because they do not have the existential reading, or mass/plural reading required by empty D. Assuming that this is true of all languages, any empty existential D-determiner should have the properties listed in (117). However, a pronounced version of this existential D-determiner should not be subject to a lexical government requirement, as it is no longer empty. The putative Skwxwu7mesh existential D-determiner should have the following features. (118) kwi may occur at S-Structure in Skwxwu7mesh only under the following conditions: a. It is restricted to plural or mass nouns. b. It receives an indefinite interpretation corresponding to an existential quantifier unspecified for number and taking the narrowest possible scope. (118)b does seem to be true for kwi. However, (118)a does not seem to be true, as kwi can occur with singular count nouns. As we have seen, some languages allow bare singulars (or numberless nominals), so this is not necessarily an argument against treating kwi as an existential D-170 determiner. A more compelling argument comes from the fact that kwi can be used to introduce proper names, if the referent is dead, or time periods, as long as it is not the present day ( § 3 ) . Calling kwi an existential D-determiner would also not explain the lack of deictic features or its abilitity to be used partitively. Further, Longobardi does not explain why his existential D-determiner must take narrow scope. In this chapter I have appealed to Restrict as an explanation for the narrow scope behaviour. 6.3 Kwi is not an expletive D-determiner Vergnaud and Zubizarreta (1992) claim that in certain contexts languages require an expletive D-determiner. I have already claimed that there is no such thing as an expletive D-determiner; however, if expletive D-determiners really did exist, we might predict that kwi was the expletive version of the deictic D-determiners. There are many reasons not to think this is correct. First, expletive determiners are used for the Construction of Inalienable Possession, generics and names. However, there is no equivalent of the Construction of Inalienable Possession in Skwxwu7mesh. Possessive morphology is always preferred. (119) a. Na kexw-en-tsut ta rl gather.together-tr-refl det ta stsatsi7n-s. det bloodSposs 'The people donated their blood.' stalmexw people kwi-s comp-nom sat-shit-as give-appl-3erg Na kexw-en-tsut rl together-tr-refl kwi stsatsi7n-s. det bloodSposs ta det stalmexw people kwi-s comp-nom sat-shit-as give-appl-3erg When translating from English, "generic" DPs can be introduced by either kwi or ta, suggesting that either both are 'expletive', or that Skwxwu7mesh DPs are not equivalent to kinds. (120) a. Ha71h-s good-caus 'I like cats. chen ta push. Isg.s det cat 171 b. Ha71h-s chen kwi push. good-caus Isg.s det cat 'I like cats.' Finally, proper names are not introduced by kwi, unless the referent is dead (as we saw in §3). (121) a. Na ilhen ta John. rl eat det John 'John ate.' b. * Na ilhen kwi John. rl eat det John We also saw above that kwi introduces some meaning to the NP: the contextual set. For all of these reasons I conclude that the D-determiner kwi is not an expletive determiner. 6.4 Kwi is not a Skwxwu7mesh-type quantifier Another potential analysis of kwi is that it is a quantifier of some kind. Skwxwu7mesh has two quantifiers that are found in DPs: ilxw 'all' and kex 'many'. The non-deictic D-determiner kwi does not behave like either of them: kwi can be used with the verb of existence tsi7. (122) a. c. Tsi7 kwi sha7yuna7 ta-n lam'. exist det ghost loc det-lsg.poss house 'There is a ghost in my house.' Tsi7 ta/kwi kex sha7yuna7 ta-n exist det many ghost loc Tsi7 i7xw ta/kwi sha7yuna7 ta-n exist all det ghost loc lam 39 det-lsg.poss house lam'. det-lsg.poss house One of the quantifiers kex, like the numerals, can also be used in predicate position, unlike kwi. (123) a. Kex kwi n-skwem-kwemay'. many det 1 sg.poss-redup-dog 'I have many dogs.' 3 9 Matthewson (1998) argues that, in argument position, weak quantifiers in St'at'imcets are only given a proportional reading, and never a cardinal reading. If this is also true in Skwxwu7mesh, then the fact that the weak quantifier is as ungrammatical as the strong quantifier in this construction is explained. 172 b. Kex ta-n skwem-kwemay'. many det-lsg.poss redup-dog 'I have many dogs.' c. * Kwi ta-n skwem-kwemay'. det det-lsg.poss redup-dog (Intended meaning: I have some dogs) The quantifier i7xw 'all' also cannot be used in predicate position. (124) * i7xw kwi/ta skwem-kwemay'. all det redup-dog (Indended meaning: I have all the dogs.) However, it is not possible to analyze kwi as a vague 'all'. The meaning of kwi is, if anything, the opposite of 'all', as it can be used to refer to a subset of the context. 6.5 Kwi and ta are not different kinds of definite determiners The fact that kwi can be used in familiar contexts (i.e. the partitive contexts) means that it could be some kind of definite determiner. Of course, we would have to redefine what "definite" means, as all Skwxwu7mesh determiners can occur in novel contexts, including existential sentences. Assuming, for the moment, that we can redefine definiteness in this way, kwi and ta could be different kinds of definite determiners, where kwi is used partitively and ta is used to refer to the entire set. Some German dialects make a distinction between two definite determiners: North Frisian (Ebert 1971), Bavarian (Scheutz 1988), Armen (Heinrichs 1954) and Mdnchengladbach (Hartmann 1982). One of the determiners, called the A-article by Ebert (1971), is used if the referent is known to the speaker by world knowledge (i.e. for uniquely referring DPs or generic DPs). The other one, called the D-article, is used for anaphoric and deictic use, where the referent has been introduced in the discourse. A-article D-article singular masculine a di feminine/neutral at det plural a don Table 5.3: The article system of Fering (adapted from Ebert 1971). 173 The D-article can be used anaphorically (125)a, or where the referent is visible in the physical context (125)b. (125) a. Oki hee an hingst keeft. Di hingst haaltet. Oki have? a horse bought D.m.sg horse lame 'Oki bought a horse. The horse was lame.' b. Don kaater kleesi D.pl cat.pl scratch 'The/those cats are scratching.' The A-article is used with unique objects (126)a, generics (126)b, or referents which are unique within a situation (126)c. (126) a. a san A.m.sg sun 'the sun' b. * di san D.m.sg sun (Frisian; Ebert 1971:160) b. A kaater kleesi. A.pl cat.pl scratch 'Cats scratch' c. Ik skal deel tu a kuupmaan. / must go to A.m.sg grocer 'I must go (down) to the grocer.' If Skwxwu7mesh determiners made the same distinction, we would predict that the determiner used for the uniquely referring NPs (such as sun) should be used in generic contexts (if they indeed exist). However, generic contexts allow either determiner. (127) Ha71h-s chen ta/kwi push. good-caus Isg.s det cat 'I like cats.' Further, we would predict that different determiners should be used in anaphoric and deictic contexts from the uniquely referring DPs. However, the deictic determiner ta is used in anaphoric contexts and for uniquely referring NPs. The non-deictic D-determiner kwi can be 174 used in anaphoric contexts, but not for uniquely referring DPs, because the speaker is able to locate them. (128) a. Na te7us-em (t-)ta ni7ch'. rl look.out-intr (obl-)det sea 'He looked out at the sea.' (Kuipers 1967:236) b. * Na te7us-em t-kwi ni7ch'. rl look.out-intr obl-det sea (129) a. An tutaw ta lhkaych'. very bright det moon 'The moon is bright.' b. * An tutaw kwi lhkaych'. very bright det moon The distinction cannot be along these lines. 7 Conclusions and Implications I have argued that the D-determiner kwi is crucially non-deictic. Further, I have argued that it is a non-deictic D-determiner with no other features (such as assertion of uniqueness). This has implications for how the D-determiner can be interpreted. I have argued that it is composed via Restrict, which entails that it must be interpreted with narrow scope. It also has implications for which subject position it may occupy (crucially, a lower subject position). I have also argued that kwi, as a D-determiner, must have domain restriction in its denotation. ( 1 3 0 ) [kwiTJ = XP Xx [P(y) A C ( X ) ] This allows kwi to be used to refer to a previously introduced discourse referent, unlike bare nouns. When a kwi DP is used in a sentence, that sentence does not carry an implicature of uniqueness, unlike sentences containing deictic D-determiners. I argued that this lack of 175 uniquness is derived from the lack of deictic features. The non-deictic D-determiner can be used partively precisely because it has no deictic features. 176 Chapter Six: Typological Implications 1 Introduction In this chapter I discuss the ramifications of the analysis of Skwxwu7mesh D-determiners that I have provided in Chapters 4 and 5. There are two main questions. First, do other Salish languages have similar D-determiner systems? Second, if D-determiners in Skwxwu7mesh provide domain restriction over their NPs, and quantifiers do not, then the question becomes, is this universal or language specific? In §2 ,1 discuss the implications of my analysis for other Salish languages. Some Salish languages appear to have non-deictic D-determiners, while others do not. The potentially non-deictic D-determiners appear to behave much like kwi. That is, they have D-determiners that are used in questions, under negation, in complex numerals, or in any other contexts when the speaker is unable (or unwilling) to locate the referent. These D-determiners are often used as complementizers, because events are not physically locatable in the same way entities are. A further prediction for a non-deictic D-determiner would be that they could be used partitively; no evidence for this is found in any of the grammars. This is probably because the data involved are not the usual kinds of data elicited in the production of grammars. In order to get partitive judgments, the right contexts must be set up by the researcher. Some Salish languages appear to lack a non-deictic D-determiner. They do not have a D-determiner which behaves like kwi. All D-determiners in these systems are used for potentially locatable referents. I claim in §3 below that, even in English, only the D position is associated with domain restriction. I also provide some indirect evidence that English determiners have domain restriction in their denotation and quantifiers do not. In §4 ,1 discuss the implications of my analysis for languages without (overt) D-determiners. I claim below that covert D-determiners are present in certain contexts (e.g., in definite contexts). 177, In §5 I discuss the implications for Maori . Maori has a functional element which forces the nominal to take narrow scope, but is not used in familiar contexts. I claim that this element cannot be a D-determiner. In §6 ,1 conclude the thesis with remaining questions. 2 Implications for other Salish languages M y analysis of Skwxwu7mesh involves four main claims. (1) a. all D-determiners in Skwxwu7mesh involve domain restriction. b. no D-determiners in Skwxwu7mesh assert uniqueness c. most D-determiners have deictic features and can take wide scope d. one D-determiner lacks deictic features and must take narrow scope None of my claims can be properly tested on other Salish languages without fieldwork, especially the first two. However, all Salish languages with D-determiners appear to encode deictic features (Matthewson 1998). The final claim (that Skwxwu7mesh has a non-deictic D -determiner) seems to apply to some of the other Salish languages (though not all). Non-deictic D-determiners are non-locatable, and lack any deictic features. I claimed that this resulted in kwi DPs being composed via Restrict, and taking obligatory narrow scope. I predict that non-deictic D-determiners in other Salish languages w i l l often be found in questions and under negation, but also in contexts where the speaker is unable to locate the referent. If a non-deictic D-determiner co-occurs with an operator, I also predict it w i l l take narrow scope with respect to that operator. Non-deictic D-determiners w i l l also be more likely to be used as complementizers. I argued in Chapter 5 that the non-deictic D-determiner kwi is the D-determiner used as a complementizer because it is used for referents that cannot be physically located by the speaker. Embedded clauses cannot be physically located. Recall that in Skwxwu7mesh the deictic D -determiners only locate in space, not in time. Other languages have deictic D-determiners that can be used to locate referents in time as well as space (see Kinkade 1964 for Upper Chehalis; 178 Davis and Saunders 1975 for Nuxalk; Demirdache 1996 for St'at'imcets; K o c h 2006 for Me?kepmxcin). In languages like that, I predict that the deictic D-determiners can be used as complementizers, and that they could encode a present/past tense distinction. Without doing extensive fieldwork, it is impossible to tell for any language whether the D-determiners locate in time as well as space. I attempt to show that the non-deictic D-determiners are often used as complementizers. Matthewson (1998) argues that some Salish languages have a polarity D-determiner:1 specifically, Sechelt, St'at'imcets, Secepemctsin and Nuxalk. Matthewson's analysis and the one given in Chapter 5 make some o f the same predictions. A polarity D-determiner should be used in questions and under negation; a non-deictic D-determiner w i l l also be used in questions and under negation. However, a polarity D-determiner should not be found in factive environments. A non-deictic D-determiner can (but does not have to) be found in factive environments. However, our analyses end up being the same for the non-deictic D-determiners in some languages: those that have a polarity D-determiner. Matthewson (1998) predicted that polarity D -determiners were non-deictic, but did not address the question of whether all non-deictic D -determiners were polarity items. (2) If a D-determiner is only found in polarity environments then it is non-deictic If a D-determiner is non-deictic, must it be a polarity item? I argued in Chapter 5 that kwi is non-deictic but not a polarity item. M y system allows for two kinds o f non-deictic D-determiners: polarity items or plain non-deictic D-determiners. (3) If a D-determiner is only found in polarity environments then it is non-deictic. If a D-determiner is non-deictic, it can be: i) a polarity item, or ii) plain. Below I present each of the D-determiner systems that have been described well enough to test for a non-deictic D-determiner. 1 Matthewson (1998) uses the term 'non-assertion of existence', rather than 'polarity'. See §2.1.1 for discussion of non-assertion of existence. 179 2.1 Salish languages with potentially non-deictic D-determiners Only a subset of Salish languages appears to have a non-deictic D-determiner. These D-determiners seem to share some behaviour with kwi. They are found in questions and under negation, and are also often found in contexts when the speaker cannot locate the referent. 2.1.1 St'at'imcets The St'at'imcets D-determiner system is the best-described of all the other Salish languages. Here I briefly discuss the determiner system and compare it to Skwxwu7mesh. Matthewson (1998) argues that St'at'imcets lacks an indefinite/definite distinction. Most of the D-determiners can be used in both novel and familiar contexts. For example, in (4)a, the referent smem 'lhatsa 'a girl' is introduced in the story. In (4)b, the same D-determiner is used on the now-familiar DP. (4) a. huy'-lhkan ptakwlh, ptakwlh-min lts7a ti smem'lhats-a... going.to-Isg.s tell.story tell.story-appl here det woman(redup)-det 'I am going to tell a legend, a legend about a girl...' (novel) b. wa7 ku7 ilal lati7 ti smem'lhats-a. prog quot cry deic det woman(redup)-det 'The girl was crying there.' (familiar) (St'at'imcets; van Eijk and Williams 1981: 19, cited by Matthewson 1998) On the basis of this, Matthewson argues for a Common Ground Parameter to distinguish between Salish and English D-determiners. English D-determiners, according to her, can access the common ground, while Salish D-determiners cannot. (5) Common Ground Parameter Determiners may access the common.ground of the discourse Yes: {English,...} No: {Salish,...} (Matthewson 1998) The analysis of Skwxwu7mesh D-determiners that I have argued for in this thesis is incompatible with the Common Ground Parameter. I therefore argue that the CGP cannot be the source of the difference in meaning between English and Salish D-determiners. I have argued that all D-180 determiners access the common ground via domain restriction. The difference in meaning instead arises from the (lack of) assertion of maximality. Instead of an indefinite/definite distinction, Matthewson argues for a non-assertion of existence/assertion of existence distinction. (6) Assertion of existence (informal definition) the speakers' intent to 'refer to' or 'mean' a nominal expression to have non-empty references - i.e. to 'exist' within a particular universe of discourse (i.e. not necessarily within the real world) (Givon 1978: 293-4) There are two major differences between an assertion of existence D-determiner and a non-assertion of existence D-determiner. First, assertion of existence D-determiners take wide scope and non-assertion of existence D-determiners take narrow scope with respect to some operator. (7) a. cuz' " tsa7cw kw-s Mary lh-t'iq-as going, to happy det-nom Mary hyp-arrive-3conj ti qelhmemen'-a. det old.person(dimin)-det 'Mary will be happy if an elder comes.' = 3x [elder (x) & [come (x) —> happy (Mary)]] (wide scope) 4 [3x [elder (x) & come (x)]] —» happy (Mary) (narrow scope) b. cuz' tsa7cw kw-s Mary lh-t'iq-as going, to happy det-nom Mary hyp-arrive-3conj ku qelhmemen'. det old.person(dimin) 'Mary will be happy if any elder comes.' 4 3x [elder (x) & [come (x) —> happy (Mary)]] (wide scope) = [3x [elder (x) & come (x)]] - » happy (Mary) (narrow scope) (St'at'imcets; Matthewson 1999: 90) Second, non-assertion of existence D-determiners are more restricted in terms of the environments which they can occur in. Non-assertion of existence D-determiners can only be used in non-factive environments, such as under negation, in questions, or under other operators. They cannot occur in factive environments. (8) a. cw7aoz kw-s ats'x-en-as ku sqaycw. neg det-nom see-tr-3erg det man 'S/he didn't see any men.' 181 b. ats'x-en-lhkacw ha ku sqaycw? see-tr-2sg.s Q det man ' D i d you see a man/any men?' c. ats'x-en-as k ' a ku sqaycw. see-tr-3erg appar det man ' S/he must have seen a man.' d. * ats'x-en-as ku sqaycw. see-tr-3erg det man 'S/he saw a man.' (St'at'imcets; Matthewson 1999: 88) Assertion of existence D-determiners can be used in any environment. (9) a. cw7aoz kw-s az'-en-as ti sts'uqwaz'-a neg det-nom buy-tr-3erg det fish-det kw-s Sophie. det-nom Sophie 'Sophie didn't buy a fish.' (= 'There is a fish which Sophie didn't buy.') (St'at'imcets; Matthewson 1999: 91) b. tup-un'-as ha ti sam7-a s-John? hit-tr-3erg Q det white.person-exis nom-John ' D i d John hit a white man?' (St'at'imcets; Matthewson 1998) c. kan-as kelh qwal'iit-s-as k Mary ti naplit-a. wh-3conj might talk-caus-3erg det Mary det priest-det 'Mary might talk to a priest.' (= 'There is a priest who Mary might talk to.') (St'at'imcets; Matthewson 1999: 91) d. tecwp-min-lhkan ti piikw-a lhkunsa. buy-appl- Isg.s det book-det today 'I bought a/the book today.' (St'at'imcets; Matthewson 1998) Matthewson (1998) also argues that the determiner system of St'at'imcets has assertion of existence determiners, which encode deictic distinctions (as well as number). assertion of existence non-assertion of existence present absent remote ku (kwelh) -plural ti...a ni...a ku...a +plural -collective i...a nelh...a kwelh...a +collective ki...a Table 6.1: The St'at'imets D-determiner system (Matthewson 1998). 182 Matthewson (1998) argues that the non-assertion of existence determiner ku lacks deictic features. I tentatively reanalyze the D-determiner system below. proximal medial, invisible distal, invisible non-deictic (polarity) singular . ti...a ni...a ku...a ku plural -collective i...a nelh...a kwelh...a (kwelh) +collective ki...a Table 6.2: The St'at'imets D-determiner system. I am not only reanalyzing the D-determiner system to reflect the basic deictic/non-deictic contrast, but I am also adopting the assumption that the features argued for in Skwxwu7mesh are universal, following (Imai 2003). I am therefore renaming the particular deictic distinctions that previous authors have used, for all of the Salish languages discussed in this section. The non-deictic D-determiner ku cannot merely lack deictic features, however. Matthewson (1998) argues that ku is a non-assertion of existence determiner because it can only occur in non-factive environments. (10) a. tecwp-min-lhkan kelh ku buy-appl-lsg.s might det 'I might buy a book tomorrow.' piikw natcw. book tomorrow tecwp-min-lhkan ku buy-appl-lsg.s det (I bought a book today) piikw lhkunsa. book today (St'at'imcets; Matthewson 1998) In Matthewson (1999), she recast ku as a polarity determiner. I adopt this terminology here. I claim that ku is a non-deictic, polarity D-determiner. This appears to be only one of three such D-determiners in the Salish family. M . Dale Kinkade (p.c) suggested that St'at'imcets may have borrowed this form from the Coast Salish languages.2 This may explain why it has a more restricted distribution than kwi. I predict that ku should be used partitively. However, this is not the case (Lisa Matthewson, p.c). There are two potential reasons for this. One is that it is a polarity item. Because of the environments it is used in, the partitive reading may be difficult to get. The other, 2 However, both Secwepemctsin and Nie?kepmxcin have a k in their non-deictic D-determiner. These are in the same branch of Salish (Northern Interior) as St'at'imcets. Further, the Nie?kepmxcin non-deictic D-determiner also appears to be a polarity item (see below). 183 more compe l l i ng , reason is that St 'a t ' imcets has a part i t ive quant i f ier nukw ' some o f the ' , and therefore m a y have taken over the part i t ive uses o f ku. St 'a t ' imcets D-determiners can phys ica l l y locate a referent; they can also locate a referent i n t ime (Demi rdache 1996, Ma t thewson 1998). A subset o f the D-determiners can be used as complement izers . (11) a. ama ti s - t ' iq -s-a s-Gert ie . good comp nom-arrive-3sg.poss-exis nom-Gertie 'It is good that Ger t ie came. ' b. zwat-en-as kw-s 4 qacwecw-s-as t i q i l ' q - a k w - s Henry . know-tr-3erg comp-nom break-caus-3erg det chair det-nom Henry ' S h e knows that Hen ry broke the chai r . ' (S t 'a t ' imcets ; Ma t thewson 1998) F o r more detai ls on the St 'a t ' imcets system, see Ma t thewson (1998). The ma in s imi la r i ty between our two systems is that the non-de ic t ic and/or po lar i ty D -determiners should take narrow scope. 2.1.2 Sechelt The Sechelt D-determiner system is s imi lar to the S k w x w u 7 m e s h system. Beaumont (1985) c la ims that there is a three-way dist inct ion (aside f rom gender) in the D-determiner system: ( i) v is ib le or inv is ib le , ( i i) inv is ib le and ( i i i ) unspec i f ied or abstract. v is ib le or inv is ib le inv is ib le unspec i f ied or abstract non- female te che she female, s ingular l he s Table 6.3: The D-determiner system o f Sechelt (adapted f r om Beaumon t 1985: 25). The unspec i f ied or abstract D-determiner is a good candidate for a non-de ic t ic D-determiner. " I f the speaker is referr ing to something that is not "real" (or "actual"), that is, something that he can or could not (or doesn V want to) identify specifically in a physical sense, he uses she, shen, etc." (emphasis o r ig ina l ; Beaumont 1985: 53). Th is is very s imi la r to the analys is o f kwi I gave in 3 For a conflicting view, see Matthewson (2005). 4 Kw is equivalent to ku (Davis and Matthewson 1997). 5 Beaumont also mentions another female D-determiner in his index, but never directly addresses it: tse, which is only used for visible female referents. 184 the previous chapter. Kwi is used for referents which the speaker either cannot or does not want to locate. She can be used in many of the same environments as kwi is used. For example, it is used in questions (12)a, when the speaker is unable to locate the referent (12)b, and in complex numerals (12)c. (12) a. ?e sxatl'-a she stamas? 2sg.poss want-Q det what 'Do you want anything/something?' b. kuku-am-chen. ne sxatl' she s?iwuts. thirst-intr-Isg.s lsg.poss want det water 'I'm thirsty. I want some water.' c. ?upan ?iy she pala ten conj det one 'eleven' (Sechelt; Beaumont 1985: 52-53) d. she shashishalhem det sechelt.language 'the Sechelt language' (Sechelt; Beaumont 1985) Matthewson (1998) argues that she is a non-assertion of existence determiner. However, this cannot be right, as she can be used in factive environments (13).6 (13) a. mi-la ?e she-ms sni! come-imper obi det-lpl.poss place 'Come to our place.' (our general location) t'i tsu ?e she fact go obi det 'He went to Deserted Bay.' ts'iinay. Deserted.Bay t'i-sht tsu ?imash ?e she fact-lpl.s go walk obi det 'He went walking (somewhere) in the forest.' salnachiya. forest (generally inside it) (Sechelt; Beaumont 1985: 53) 6 Better counter-examples to Matthewson's claim would involve non-oblique arguments. 185 The above examples show that, like kwi, she can be used when the referent is located within another location. (This is also similar to the use of kwi when referring to an object within a previously mentioned set.) (14) Na7 t-kwi n-lam' ta-n yasakw. loc obl-det 1sg.poss-house det-lsg.poss hat 'My hat is in my house.' If a deictic D-determiner is used instead, the particular location is focused on. (15) a. mi-la ?e te/che-ms sni! come-imper obi det-lpl.poss place 'come to our place.' (emphasis on house as a known physical thing) b. kw'enit-chen te ts'unay. see-lsg.s det Deserted.Bay 'I see Deserted Bay.' c. t'i tsu ?e te/che salnachiya. fact go obi det forest 'he went walking in the forest.' (thinking of actual forest) (Sechelt; Beaumont 1985: 54) I take it that she is non-locating, and therefore non-deictic, while the other D-determiners are deictic. Unlike kwi, she is not used a complementizer. The element kwe is instead, and is only used as a complementizer. (16) a. ne sxatl' kwe-n s-?itut. lsg.poss want comp-lsg.poss nom-sleep 'I want to sleep.' b. may-stexw-chen kw'e s-k'uk'ukit. bad-caus-Isg.s det.2sg.poss nom-kiss 'I don't like it that you are kissing him/her.' (Sechelt; Beaumont 1985: 153) Lacking deictic features therefore does not ensure that the D-determiner will be used as a complementizer. Assuming the kwe and kwi derive from the same source, it may be that kwe lost 186 its status as a D-determiner, and she was brought in to do the work of the non-deictic D-determiner.7 I present my tentative reanalysis of the Sechelt D-determiner system below. neutral distal, invisible non-deictic gender-neutral te che she female, singular 11 le Table 6.4: The D-determiner system of Sechelt. As I discussed in Chapter 2, neutral D-determiners or demonstratives are used for referents which are locatable, regardless of their distance from the speaker (see Imai 2003). The Sechelt D-determiner te appears to be equivalent to the Skwxwu7mesh ta, which can be used for proximal, medial, distal and invisible referents. I predict that the non-deictic D-determiner she would obligatorily take narrow scope. I also predict that it should be able to be used partitively. This needs to be tested. 2.1.3 Lushootseed Hess (1995) claims that the Lushootseed determiner system can be split along three lines (aside from gender): unique reference, neutral or non-contrastive and hypothetical/remote. unique reference neutral/ non-contrastive hypothetical/remote non-female ti/so8 to k w i female, singular tsi/sa tso kwsi Table 6.5: The D-determiner system of Lushootseed (adapted from Hess 1995: 77). The hypothetical/remote D-determiners are the most obvious candidate for a non-deictic D-determiner. Like Skwxwu7mesh kwi, the Lushootseed D-determiner kwi can be used in questions (17)a and under negation (17)b. 7 Henry Davis (p.c.) suggests this may be a plausible analysis, as Skwxwu7mesh III may have derived from schwa. 8 The two forms reflect a difference in dialect. In each case, the first entry is from Northern Lushootseed; second from Southern Lushootseed. 187 (17) b. ?ah ?u k wi gw-ad-pispis. there Q det sbj-2sg.poss-cat 'Do you have a cat?' x w i? k w i gw9-d-scqi. neg det sbj-1 sg.poss-sockeye 'I don't have [any] sockeye.' (Lushootseed; Hess 1995: 94) (Lushootseed; Hess 1995: 94) Matthewson (1998) did not address the status of the hypothetical/remote D-determiner in Lushootseed, but it is clear that k "z and k wsi also cannot be non-assertion of existence D-determiners. They can be used in factive environments (18). (18) a. b. b-s-lil-cut cad tul'-?al k w i bakw sp'aJUft. ?al progstat-far-refl Isg.s from-loc det all worthless loc ti swatixwtsd. det world 'I am keeping myself from all the worthlessness in the world.' (Lushootseed; Hess 1995: 84) bl i? k wi bsqsad ?9 different det beak obi 'The beak of a raven is different.' k w i det ?ss-t'igwid ?al k w i dadatut. stat-thank loc det morning 'Thank someone in the morning.' qaw qs. raven (Lushootseed; Hess 1995: 95) (Lushootseed; Hess 1995: 83) In all of the cases in (17) and (18), the referent cannot be located in space, or in time. Example (18)c refers to any morning, not a particular morning. Hess does not provide the equivalent of (18)a-c with the deictic D-determiners; I predict they are not licit in these environments. Like kwi, the Lushootseed k "I can also act as a complementizer. (19) a. x w i? k w i gwa-s-u-xaab-s. neg comp sbj-nom-perf-cry-3poss 'He doesn't cry.' (Lushootseed; Hess 1995: 96) 188 b. ck ' w aq id cad 3U-baliic k w i g w a - d - s - u -?9 i t x w always Isg.s hab-forget comp sbj-lsg.poss-nom-perf-feed ti d-sqwobay?. det lsg.poss-dog 'I always forget to feed my dog.' (Lushootseed; Hess 1995: 96) Again, this is likely a result of the fact that the deictic D-determiners cannot locate referents in time. It is difficult to tell which features the potentially deictic D-determiners ti/tsi and ta/tso have. It is also difficult to tell whether by "unique" Hess intended the interpretation of "unique" used in this thesis. That is, it is unclear whether the D-determimers ti and tsi assert the uniqueness of their referents. From the data provided by Hess, it appears that ts is proximal and ti is distal. Whether this is accurate would need to be tested systematically. In (20), the D -determiner ts is used for referents which appear to be proximal to the speaker. (20) a. b-s-t 'ag wt cod lii-?al ta stiqiw. prog-stat-on.top Isg.s via-locdet horse ' I 'm riding on the horse.' (Lushootseed; Hess 1995: 84) b. dogwas dx w -?al ta x w dog w i g w sa l i . inside.tr dir-loc det bag 'Put [something] into the bag.' (Lushootseed; Hess 1995: 83) In (21), the D-determiner ti is used for referents which appear to be distal from the speaker. (21) a. lu-tolawil cad dx w -?al ti x w uyubal?tx w . irr-run Isg.s dir-loc det store ' I ' l l run to the store.' (Lushootseed; Hess 1995: 83) b. b -s -q ' i l cod li l-?al ti l i lud. prog-stat-ride Isg.s via-loc det train ' I am traveling by train.' (Lushootseed; Hess 1995: 84) I present my tentative reanalysis of the Lushootseed D-determiner system below. 189 proximal distal non-deictic gender and number neutral t9 ti/S9 k w i female, singular tS9 tsi/s9 k w s i 9 Table 6.6: The D-determiner system of Lushootseed. I predict that the non-deictic D-determiners k "I and k wsi should obligatorily take narrow scope. I also predict that the non-deictic D-determiners should be able to be used partitively. 2.1.4 Musqueam Suttles (2004) claims that the Musqueam determiner system can be split along three lines (aside from gender and case): (i) present and visible, (ii) nearby and invisible and (iii) remote or hypothetical. present, visible nearby, invisible remote or hypothetical non-female t9 (te9) k w 0 9 , k w 9 , k w k w 9 , k w female 09 te, k w te , i, k w f kwS9 oblique K Table 6.7: The Musqueam D-determiner system (adapted from Suttles 2004: 340). The most obvious candidates for non-deictic D-determiners are k w 9 , k w , and k w s9 , the remote or hypothetical D-determiners. These D-determiners share some properties with kwi, and with ku, the St'at'imcets polarity non-deictic D-determiner. Matthewson does not address whether Musqueam has non-assertion of existence D -determiners (as the grammar had not yet been published), but the candidates would be the same as for the non-deictic D-determiners. Both my analysis and Matthewson's analysis make similar predictions in that the remote or hypothetic D-determiners should be used in questions (22)a and under intensional operators (22)b and c (and take narrow scope with respect to the operators). (22) a. stem k W 3 sk w ix-s ti?i? what det name-3poss dem 'What is the name of this?' (Musqueam; Suttles 2004: 348) 9 The female non-deictic D-determiner is likely used in sentences like "I'm looking for a woman", where it is clear what gender the referent should have, were one to be located. 190 b. na-s-[c-]li R w a pay. 1 sg.poss-nom-do-value det pie 'I want some pie.' (Musqueam; Suttles 2004: 345) c. ?oy kw-s ne'm-ct sgwq-t kw pgwgy good det-nom go-lpl.poss seek-tr det flounder [?o] to cawcow. obi det offing 'We'd better go look for flounders offshore.' (Musqueam; Suttles 2004: 345) However, unlike a non-assertion of existence analysis, my analysis also, predicts that the remote or hypothetical D-determiners can be used in factive contexts. (23) a. ni [?a] kwa n9-sii?a?aqwt be.there obi det 1sg.poss-rear 'behind me (in a canoe).' b. nem can [?o] kwa spgixgn.10 go Isg.s obi det pasture 'I'm going to the pasture (way off, out of sight).' (Musqueam; Suttles 2004: 344) In the above two examples, the referents are in principle locatable to the speaker, but are unseen by the speaker. It appears that, like kwi, the remote or hypothetical D-determiners can be used for locatable referents, but only if there is no visible evidence contradicting the speaker's use of a non-deictic D-determiner. If a deictic D-determiner is used instead, the referent must be locatable, either by sight, or by shared knowledge.11 (24) a. ni [?a] ta ng-TeGaqgn be.there obi det Isg.poss-front 'in front of me' (in canoe) 1 0 In all other cases, this was spelled spaixan. " Many of the examples involve obliques. However, these sentences were pronounced without the oblique marker; I retain Suttles' analysis of the underlying form. (29)a is an example of the non-deictic D-determiner in a more obvious argument position. 191 b. ne'm can ts spsixan. go Isg.s obi det pasture 'I'm going to the pasture (not far and visible).' c. ne'm can k w 6s spaixsn. go Isg.s obi det pasture 'I'm going to the pasture (out of sight).' (if known to the hearer) (Musqueam; Suttles 2004: 344) Similarly to kwi, the remote or hypothetical D-determiners can be used for deceased referents, (25)a and b. (25) a. kw3 ns-sil'a det lsg.poss-grandparent 'my late grandfather/great uncle' b. k w ss na-sil'a det lsg.poss-grandparent 'my late grandmother/great aunt' (Musqueam; Suttles 2004: 343) A deictic D-determiner may also be used, but only with the past tense marker on the noun. (26) a. k w 03 na-sil's-i det lsg.poss-grandparent-pst 'my late grandfather/great uncle' b. fe ns-sil'a-i det lsg.poss-grandparent-pst 'my late grandmother/great aunt' (Musqueam; Suttles 2004: 344) The remote or hypothetical D-determiners can also be used for non-present times (27)a and for complex numerals (27)b. (27) a. k w caleqs* det yesterday 'yesterday' (Musqueam; Suttles 2004: 345) b. ni-wi sisrh K ?apan i kw9 naca?. be.there-already wearing.off obi.det ten and det one 'It's half past eleven.' (Musqueam; Suttles 2004: 350) 192 In the above cases, I predict that deictic D-determiners are not licit.1 2 All of the examples in (22), (23), (25), and (27) have referents which are non-locatable, and which have equivalent uses in Skwxwu7mesh. The remote or hypothetical D-determiners can also be used in cases which are more similar to St'at'imcets ku. Like ku, (and unlike kwi), k W 9 can be used with adverbials. (28) b ?alq9l [?a] k W 3 xwam! come exit obi det fast 'He came out right away.' (Musqueam; Suttles 2004: 346) Two more cases of kwa as non-deictic involve introducing words that are unknown to the hearer (29) a, and comparing the referent to another (non-located) class'(29)b. (29) a. kwikw3xt-am k w 9 mSesi name-pass det msOdl 'It is called maOai [possibly Indian Hemp, Apocynum cannabinumY b. t3?a-max [?s] Kw3 swsyqe?. resemble obi det man 'It looks like a man.' (Musqueam; Suttles 2004: 345) Like kwi, k w 9 is the only D-determiner which is used as a complementizer. (30) na-s-[c-]/U k w 9 ns-s-nehi. 1 sg.poss-nom-do-value comp 1 sg.poss-nom-go 'I want to go.' (Musqueam; Suttles 2004: 345) Suttles claims that within the "nearby, invisible" class of D-determiners, the kwfaD-determiners are more distal than the fa D-determiners. I suggest that it may be possible to analyze the Musqueam D-determiner system into a four-way system, like that of Skwxwu7mesh. I present my tentative reanalysis of the Musqueam D-determiner system below. 1 2 However, Upriver Halkomelem allows both the equivalent of the "remote or hypothetical" D-deteminer and the equivalent of the "present, visible" D-determiner in numerals. See §2.2.2. 193 proximal medial, invisible distal, invisible non-deictic gender-neutral to (teo) kw09, kw9, kw kw9, kw female 09 kwi9, kwi kwS9 oblique 1 Table 6.8: The Musqueam D-determiner system. I predict that the non-deictic D-determiners Itwd, #"and kwsd should obligatorily take narrow scope. I also predict that the non-deictic D-determiners should be able to be used partitively. 2.1.5 Upper Chehalis Kinkade (1964) presents a four-way split in the Upper Chehalis D-determiner system: (i) by speaker, (ii) near speaker, (iii) not near speaker and (iv) indefinite. by speaker near speaker not near speaker indefinite -female tit ?it tat t +female tic, cic ?ic tac, cac c is (adapted from Kinkade 1964). Table 6.9: The D-determiner system of Upper Chehal The "indefinite" D-determiners are good candidates for non-deictic D-determiners. The "indefinite" D-determiners are often used in generic (31)a or future contexts (31)b and c. (31) a. ...nkws-sxw9qww-ans ?a* t panXiS. hab-hungry-Isg.s at det wintertime 'I am always hungry in the wintertime.' ...wi xwaqu?t and all det i t fut det ssa'malaxw people ?6-cs s5Us. one winter wi and i fut t det musa sleep '...all the people will sleep for one winter.' wi ?ai t ponXaqa-m n i qitaciri. and at det springtime part fut arrive 'And in the springtime, daylight will come.' (Upper Chehalis; Kinkade 1983: 255-257) 194 However, the "indefinite" D-determiners cannot be non-assertion of existence D-determiners because they can be used in factive contexts. (32) a. to-malti t sqitaci short det daylight 'Daylight and darkness are short. b. kwaxwmisi"rt tit scotxwh reach det Bear 'They reached Bear one day.' . ca t conj det skwoswn. darkness ?ai t ?6-cs sqitaci. at det one day (Upper Chehalis; Kinkade 1983: 255-256) I therefore suggest that the "indefinite" D-determiners are non-deictic, and not non-assertion of existence. The gender-neutral, non-deictic D-determiner can also be used as a complementizer. (33) ?aqw miita t sa?s tit well neg comp make det 'The world will not be made thus.' tarns 1 t qicx. world fut det thus (Upper Chehalis; Kinkade 1983: 256) Below I provide my tentative reanalysis of the Upper Chehalis determiner system. proximal medial distal non-deictic gender-neutral tit ?it tat t feminine tic, cic Tic- tac, cac c Table 6.10: The D-determiner system of Upper Chehalis. I predict that the non-deictic D-determiners t and c should obligatorily take narrow scope. I also predict that the non-deictic D-determiners should be able to be used partitively. 2.1.6 Cowlitz Kinkade (2004) claims that there is a four-way distinction in the Cowlitz D-determiner system. The Cowlitz system is very similar to the Upper Chehalis system given in the previous section. by speaker near speaker not near speaker indefinite non-feminine tit ?it tat t femine cic ?ic cac c Table 6.11: The D-determiner system of Upper Chehalis (adapted from Kinkade 2004: 254) 195 The "indefinite" D-determiners are good candidates for non-deictic D-determiners. They can be used in intensional contexts, and in questions. (34) a. ?ac-qins put-n ?aci ta qi t-1 stat-want know-tr stat.Q ?? qi comp-pst s...makwuyq \ t qawam'. s...bake. in. ashes obi det camas 'They want to know how we baked camas in a steampit.'13 (Cowlitz; Kinkade 2004: 253) b. ?ac-qins nam-n kn sali t ?amxkwu-yumx stat-want jinish-tr Isg.s two det basket-pl \ tit pan-lix. obi det time-cold 'I want to finish two baskets this winter.' (Cowlitz; Kinkade 2004: 269) c. fix na c ?a-kuwi? now Q det 2sg.poss-wife 'Is she your wife?' (Cowlitz; Kinkade 2004: 270) d. Well, uh k w i c Frances t nx-am'tan-i? how.many det.f Frances det pl-child-3poss 'Well, how many children does Frances have?' (Cowlitz; Kinkade 2004: 268) The "indefinite" D-determiners cannot be non-assertion of existence D-determiners because they are found in factive environments. (35) a. tit qi xapa-n-ani ka- t qi s-xapa-t-s det qi dry-3o-3poss? where det qi nom-dry-tr-3poss t slalas. det deer 'drying rack where he dries the deer meat' (Cowlitz; Kinkade 2004: 251-252) b. ?aqa n t'aqi-stamt t xe'w'i ?ai...p'entmx kal then and find-lpl.s det trail at...beside on taw'ai ts.... maqwm. big det prairie 'And then we find a trail at...beside a big prairie.' (Cowlitz; Kinkade 2004: 266) The question marks are in the original. The morphemes qi and s were also glossed as qi and 5. 196 c. s-pusa-w-n t mus-awmis-i. imperf-swell.up-imperf det eye-pl-Sposs 'His eyes swelled up.' (Cowlitz; Kinkade 2004: 269) I therefore suggest that the "indefinite" D-determiners are non-deictic. The non-feminine, non-deictic D-determiner can also be used as a complementizer. (36) miita t n-s-piit-n. neg comp 1 sg.poss-know-tr 'I don't know.' Below I provide my tentative reanalysis of the Cowlitz D-determiner system. proximal medial distal non-deictic gender-neutral tit ?it tat t feminine cic Tic- cac c Table 6.12: The D-determiner system of Upper Chehalis. I predict that the non-deictic D-determiners t and c should obligatorily take narrow scope. I also predict that the non-deictic D-determiners should be able to be used partitively. 2.1.7 Secwepemctsin Kuipers (1974) argues that the determiner system of Secwepemctsin has a three-way split between (i) actual-determinate present, (ii) actual-determinate absent and (iii) hypothetical-indeterminate. actual-de terminate hypothetical-indeterminate present absent absolutive y 1 k relative ill tk/lk (seldom tke) Table 6.13:1 'he D-determiner system of Secwepemctsin (adaptec I from Kuipers 1974: 57). The hypothetical-indeterminate D-determiners are good candidates for non-deictic D-determiners. Kuipers (1974) argues that the hypothetical determiners are found especially in interrogative, imperative, conditional, and negative sentences, as well as sentences with a future reference. Both a non-assertion of existence analysis and a non-deictic analysis would predict this. 197 (37) a. ta? xwum lyi? k snewt. neg at.all dem det wind 'There was no wind.' 14 (Secwepemctsin; Kuipers 1974: 105) ta? tyi? k qw9nim9qi neg dem det mosquito 'There are no mosquitoes.' nM? Ik syist one det camp.overnight 'one (more) night of camping' ckenm tlyi? me? sckwnem do dem jut get (Secwepemctsin; Kuipers 1974: 82) (Secwepemctsin; Kuipers 1974: 57) tk ?sx?rin-kt? det food-lpl.poss 'Should we get some of it for our food?' (Secwepemctsin; Kuipers 1974: 106) e. ...me? ckwnem-kt llune tk swewi. fut get-lpl.poss dem det fish '...We'll catch some fish over there.' (Secwepemctsin; Kuipers 1974: 106) The hypothetical D-determiners are also used as complementizers; so are the "actual" D-determiners. (38) a. ...ye-akwe k s-ptihsms:... and-3s comp nom-think '...and he thought: ...' b. yyi? y s-wikts nkwii? t sitqt y qwanim3qi... dem comp nom-see one det day det Mosquito 'One day, he saw Mosquito...' (Secwepemctsin; Kuipers 1974: 96) The hypothetical D-determiners can also be used in factive environments. (39) a. ...y s?etwn yenke yi? k xwuxwltn... det Crane evid dem det whistle '.. .Crane was the whistle...' (Secwepemctsin; Kuipers 1974: 104) S9su'q v ...m-twkemins yi? n-cenamn tk aor-see dem to-Chinese det grouse '...he sold them [crows] to the Chinese for grouse.' 1 4 1 have changed some o f his symbols to more recognizable symbols. 198 (Secwepemctsin; Kuipers 1974: 103) c. d. ...me? ciin yi? k sqwyic y?elye. fut say dem det rabbit dem '...I'll say, here are some rabbits.' (Secwepemctsin; Kuipers 1974: 103) m-cun-s-as yi? Ik aor-say-tr-3erg dem det 'He thought it was skimmings.' styewlkwle. skimming (Secwepemctsin; Kuipers 1974: 92) I present my tentative reanalysis of the Secwepemctsin D-determiner system below. proximal distal (invisible?) non-deictic absolutive y 1 k relative . . t / l tk/lk (seldom tke) Table 6.14: The D-determiner system of Secwepemctsin. I predict that the non-deictic D-determiners k and tk should obligatorily take narrow scope. I also predict that the non-deictic D-determiners should be able to be used partitively. 2.1.8 Nuxalk There are two previous analyses of the Nuxalk determiner system. I present both of these here. Nater (1984) argued that there were two sets of determiners: those that are usually translated as 'a' and those that are usually translated as 'the'. He does not argue that these represent an indefinite-definite distinction. close remote non-female singular ti- ta-plural (w)a-, 0 tu-, ta-female singular tsi- lha-, 7ilh-Table 6.15: The "a-type" D-determiner system of Nuxalk (adapted from Nater 1984: 41). close remote non-female singular ti-...-tc ta-...-tx plural wa-...-ts tu-...-txw female singular tsi-...tse lha-...-7ilh Table 6.16: The "the-type" D-determiner system of Nuxalk (adapted from Nater 1984: 43). 199 Davis and Saunders (1975) present a different analysis of the determiner system.15 First, they do not analyze the "a-type" determiners at all (although they do present some examples of them). Secondly, they include the demonstratives, as shown below. proximal space, present time middle space, near/past present distal dista space, time I II III IV V VI -female singular ti...tx tL.'tayx ta-'tax ta..i ta...tx ta....tax plural wa...c wa...?ac ta-'taxw ta..i ta...txw ta...tux +female singular ci...cx ci....cayx ia-?iia?i* ia...i ia...?ii ia...?ii demonstrative? no yes yes no no yes Table 6.17: The D-determiner system of Nuxalk (adapted from Davis and Saunders 1975: 846). On the surface, there does not appear to be any candidates for a non-deictic D-determiner, or a non-assertion of existence D-determiner. However, there is evidence that the close "a-type" D-determiners are non-deictic. Davis and Saunders (1974) claim that the proximal prefixes are ungrammatical in a declarative sentence. (40) *knsmak ti-?imlk. work det-man The sentence in (40) "...is unacceptable because declarative utterances presuppose... speaker witness, but this contradicts ti-?imlk, that expresses the claim the speaker has never seen the man" (Davis and Saunders 1974:31). On the basis of this, Matthewson (1998) argues that the proximal prefix is a non-assertion of existence determiner. In my terms, it would be a non-deictic, polarity D-determiner. I tentatively reanalyze the Nuxalk D-determiner system below. distal non-deictic gender-neutral singular ta- ti-plural tu-, ta- (w)a-, 0 female singular lha-, 7ilh- tsi-Table 6.18: The "a-type" D-c eterminer system of Nuxalk. 1 5 Nater (1984) and Davis and Saunders (1975) use different orthographies. 200 proximal medial distal gender-neutral singular ti...tx ta..i ta...tx plural wa...c ta..i ta...txw female singular ci...cx la...! la...?rl Table 6.19: The "the-type" D-determiner system of Nuxalk. I predict that the non-deictic D-determiners ti-, wa- and tsi- should obligatorily take narrow scope. I also predict that the non-deictic D-determiners should be able to be used partitively. 2.1.9 Nie Tkepmxcin Koch (2006) claims that there are five determiners in Nie?kepmxcin. specific; present, visible remote (in space or time) unrealized/irrealis direct he, a, 0 1(9) k oblique t tk Table 6.20: The D-determiner system of Me?kepmxcin (adapted from Koch 2006: 131). The oblique "determiner" can co-occur with any of the other determiners (Koch p.c); I take this to mean that the oblique is not a D-determiner and instead occupies a higher functional head. The obvious candidate for a non-deictic D-determiner is the unrealized/irrealis determiner k. Matthewson did not discuss Nle?kepmxcin; however, her analysis can potentially be applied to k. This D-determiner can be found in negative contexts, in imperative contexts, in questions, and under evidentials. This is expected if the D-determiner is a non-assertion of existence D-determiner or a non-deictic D-determiner. I also predict that the non-deictic D-determiner will take narrow scope (as it appears to in (41)a). (41) a. to-te? ke s-te?-s. redup-neg det nom-somethingSposs 'They didn't have anything.' (Nle?kepmxcin; Thompson and Thompson 1992: 200) 201 b. ksl-t-et-e tu? k c'y'e! detach-tr-2sg.s-imper away.from det basket 'Take it out of some basket or other!' (Nle?kepmxcin; Thompson and Thompson 1992: 156) c. swet k (?)upi-t-m us le s-q'wiyt. who det eat-tr-indef cnj det nom-fruit 'Who ate those berries?' d. la?xans kn ekwu ta-k s-q'wiyt. eat Isg.s rprt obl-det nom-fruit 'They tell me I ate some kind of berries [I do not remember].' (N4e?kepmxcin; Thompson and Thompson 1992: 154) It appears that A: is a non-assertion of existence D-determiner or polarity item, as it is almost exclusively found in non-factive sentences (Koch, p.c.).16 It also appears to lack deictic features, as in the examples above, the referents are not located. 17 Most of the D-determiners can be used as complementizers as well. (42) a. y'e t-e s-nik'-e-s good obl-comp nom-cut-appl-Serg 'It is a good thing that he cut [the undergrowth back].' (Nle?kepmxcin; Thompson and Thompson 1992: 173) b. ke?e k e?-s-xwuy' nes? is.it.that? comp 2sg.poss-nom-fut go 'Will you go?' (N4c?kepmxcin; Thompson and Thompson 1992: 174) 1 6 The two types of counterexamples are objects of morphologically intransitive predicates (i) and "descriptives" (ii) (Koch 2006). (i) xwuy' xe? n'-t-sem-s i n-sinci? FUT dem give-tr-1 sg.o-3erg det I sg.poss-younger.br other tk k'atnfm'-tn. det rod.fish-instr 'My younger brother is gonna' give me a fishing rod.' (ii) y'e xe?9 tk piiti tk pfkca-s e Mary. good dem det pretty det picture-3sgposs det Mary 'It's a good pretty picture of Mary.' (Koch 2006) These appear to be good candidates for a Restrict-type D-determiner. 1 7 The specific, present, visible determiner apparently cannot introduce an embedded clause without the oblique marker. 202 c. s-c'oq?ewr-c I xwuy' nxwesi(t)-tn-s. nom-canoe-3poss comp fut travel-instr-3poss 'It was his conveyance that he was going to travel in.' (NIe?kepmxcin; Thompson and Thompson 1992: 173) As the D-determiners can locate in time as well as space (Koch 2006), this is expected. The non-deictic D-determiner is associated with future tense in (42)b; the deictic D-determiners are both associated with past tense in (42)a and c. I present my tentative reanalysis of the Nie?kepmxcin D-determiner system below. proximal distal non-deictic (polarity) he, o, 0 1(0) k Table 6.21: The D-determiner system of Nle?kepmxcin. I predict that the non-deictic D-determiner k should obligatorily take narrow scope. I also predict that the non-deictic D-determiner should be able to be used partitively. 2.2 Salish languages lacking non-deictic D-determiners Not all Salish languages appear to have a non-deictic D-determiner. Some languages only have deictic D-determiners. I present these languages here. 2.2.1 Straits Jelinek and Demers (1994) argue that in the Lummi dialect of Straits there is a four-way split in determiner system between (i) proximal, visible, (ii) neutral, (iii) distal/out of sight and (iv) remote. proximal, visible neutral distal/out of sight remote +female sl'o so kwo kwsa general tl'a CO kwo kwco Table 6.22: T be determiner system of Lummi (adapted from Jelinek and Demers 1994: 717). 203 The most likely candidates for non-deictic D-determiners are the remote determiners. However, none of these determiners are D-determiners. They can occur without a following NP, and are therefore demonstratives. (43) lerj-t son k wsa. see-tr Isg.s dem.f 'I saw her, that one.' (Lummi; Jelinek and Demers 1994: 717) It is unclear what feature(s) differentiate(s) between the distal and the remote demonstratives. However, demonstratives are, by definition, deictic. I predict that there will not be a non-deictic demonstrative in this language (or any other). Montler (1984) argues that there is a two-way split between the determiners in the Saanich dialect of Straits: (i) not invisible and generally present,18 and (ii) invisible, remote. not invisible, or generally present invisible, remote non-feminine tss, tb k w S3 feminine, singular 0 3 k w 0 3 Table 6.23:1 fhe determiner system of Saanich (adapted from Montler 1984: 225) The most likely candidates for non-deictic D-determiners are the invisible, remote determiners. Like kwi, they can be used for non-locatable referents (44). (44) k w3 c-teb?-3n, ?i? Tslqsla? son ?3 k w s ?ebn. comp have-money-Isg.s accom buy Isg.s obi det house 'If I had money, I'd buy a house.' (Saanich; Montler 1984: 240) However, these D-determiners can also be used for referents which are locatable (45)a and b. For example, in (45)a, the speaker is hiding in a canoe, and knows that his father is on shore, cooking food. The referent is locatable by sound. In (45)d, the referent is locatable to the speaker by sight: his child has just jumped up from hiding in front of him. (45) a. k w l q w 9 l ' cts kwss s-qw3bn kwss ns-men. real ready prob det nom-barbecue det lsg.poss-father 'My father's BBQ must be ready.' (speaker is hiding from father and can smell the cooking.) (Saanich; Montler 1984: 226) 1 8 M o n t l e r a l s o s u g g e s t s t h a t t h e g e n e r a l l y p r e s e n t d e t e r m i n e r s c a n b e f u r t h e r s p l i t i n t o tsa ' p a r t i c u l a r i n d i v i d u a l ' v e r s u s th ' n e a r l o c a t i o n ' . I i g n o r e t h e s e d i f f e r e n c e s h e r e . 204 c. waseTs ' kwsa sqexa?. bark det dog 'The dog (not visible) is barking.' kwan-naxw san kw03 sieni?. see-tr Isg.s det woman 'I saw the woman.' (She's not here now) (Saanich; Montler 1984: 228) (Saanich; Montler 1984: 226) d. cane nii yaxw kwsa na-rjane?. goodness, it.is conjee det 1 sg.poss-offspring 'Goodness, it's my child!' (Saanich; Montler 1984: 245) Kwi can only be used when the referent is out of sight. (46) a. Na7-ch' huy kwi-s kwukw-s rl-evid finish comp-nom cookSposs 'My father must be finished cooking.; b. * Na7-ch' huy kwi-s kwukw-s rl-evid finish comp-nom cookSposs c. Chen kw'ach-nexw kwelha slhanay'. Isg.s look-tr(lc) det.f woman 'I saw a/the woman.' (out of room) (Skwxwu7rnesh) Further, the determiners can be used without a following NP (Timothy Montler, p.c). kwa-n man. det-lsg.poss father (can smell the BBQ) (Skwxwu7mesh) kwi n-man. det 1 sg.poss-father (Skwxwu7mesh) (47) a. xcit san 9a. know Isg.s dem.f I know her. b. xcit san tsa. know Isg.s dem I know him. c. xcit san kwsa. know Isg.s dem 'I know him (not visible).' (Saanich; Montler, p.c.) I therefore assume that the invisible, remote determiners are distal demonstratives, rather than non-deictic D-determiners. 205 2.2.2 Upriver Halkomelem Galloway (1993) argues that there is a three-way split between the determiners in Upriver Halkomelem: (i) present and visible, (ii) near, not visible, and (iii) remote. present, visible near, not visible remote number-neutral masculine, neutral te kwthe kw'e feminine the se, kwse kw'the plural ye proper names tr Table 6.24: The D-determiner system of Upriver Halkomelem (adapted from Galloway 1993: 387 and Wiltschko 2002). The most likely candidates for non-deictic D-determiners or non-assertion of existence determiners are the remote D-determiners kw 'e and kw 'the, as the term "remote" could be a reference to the lack of locatability. (48) a. Li (ye) qex kw'e siyolh li kw'a lalem? loc (pi) many det wood loc det.2sg.poss house Ts there lots of wood at your house?' Stam kw'e s-tl'i? what det nom-want 'What do you want?' L s-tl'i lsg.poss nom-want 'I want water.' kw'e qo: det water (Upriver Halkomelem; Galloway 1993: 388-389) Like Skwxwu7mesh kwi, the remote D-determiners are also used for deceased referents, in complex numerals, and to refer to time. However, the near, not visible D-determiners can also be used for deceased referents, as in (49)a, and the present, and the masculine, visible D-determiner can be used in complex numerals, as in (49)b. (49) a. b. kwthe/kw'e-1 sila:-lh det/det-lsg.poss grandparent-pst 'my late grandfather' (Upriver Halkomelem; Galloway 1993: 388-389) 'opel qas ten and 'twelve' te/kw'e det/det 'isale two (Upriver Halkomelem; Galloway 1993: 406) 206 c. kw'e tselaqelh(-elh) det yesterday (-pst) 'yesterday' (Upriver Halkomelem; Galloway 1993: 389) d. kw'(e) spelwalh det year 'last year.' (Upriver Halkomelem; Galloway 1993: 389) The remote D-determiner kw 'e can also be used as a complementizer. (50) Tsel 'ats-lexw kw'e-s q'ay-lexw-es te swiyeqe te spath. Isg.s hear-tr comp-nom kill-tr-3erg det man det bear 'I heard that the man killed a bear.' (Upriver Halkomelem; Galloway 1993: 395) The remote D-determiners in Upriver Halkomelem cannot be non-assertion of existence D-determiners because they can be used in non-factive contexts. (51) a. Ts'tl'em kw'e swiyeqe.19 jump det man 'The man jumped.' b. Kw'ets-lexw te spath kw'e swiyeqe. see-tr det bear det man 'The man saw the bear.' (Upriver Halkomelem) (52) a. Setqtst-es kw'e pipe. UghtSerg det paper 'He lights paper [on fire].' (Upriver Halkomelem; Galloway 1993: 374) b. Le thiyqw-t-es kw'e sth'ekw. aux dig-tr-3erg det worm 'He dug for worms.' (Upriver Halkomelem; Galloway 1993: 389) c. Sta'a kw'e stl'oqwi. like det fish 'It's like a fish.' (Upriver Halkomelem; Galloway 1993: 389) However, the remote D-determiners also cannot be non-deictic D-determiners. They can be used for place names, which the speaker should be able to locate, for proper names, and to make reference to proximal locations. 1 9 These data are from my own fieldwork. 207 (53) a. Tl'a-l-su And-lsg.poss-so ' So I reached Vancouver.' tes kw'e lhq'alets. reach det Vancouver (Upriver Halkomelem; Galloway 1993: 388) k w ' det Bill. Bill (Upriver Halkomelem; Galloway 1993: 388) b. Le qal-t-em aux rob-tr-pass 'Bill was robbed.' c. 'i kw 'e 16. here det here 'Here, in this place.' (Upriver Halkomelem; Galloway 1993: 399) I therefore assume that the remote D-determiners are distal, rather than non-deictic. proximal medial, invisible distal, invisible number-neutral gender-neutral te kwthe kw'e female the se, kwse kw'the plural ye proper names tr 2.3 Summary Deictic features are rampant throughout the determiner systems of Salish. Systems differ as to whether they have only demonstratives (such as Lummi) or whether they have both D-determiners and demonstratives (such as the rest of the languages discussed here). They also differ as to whether they have a non-deictic D-determiner or not. 2 0 Upriver Halkomelem and Musqueam are dialects of the same language (Halkomelem). It may be that the Musqueam non-deictic D-determiner is also a distal determiner. More research is required. j208 Does the language have determiners? non-deictic D-determiners? polarity D-determiners? Skwxwu7mesh yes yes no St'at'imcets yes yes yes Sechelt yes probably no Lushootseed yes probably no Musqueam yes probably no Upper Chehalis yes probably no Cowlitz yes probably no Secwepemctsin yes probably no Nuxalk yes probably yes Nie?kepmxcin yes probably yes Straits Lummi no n/a n/a Saanich yes no n/a Upriver Halkomelem yes no n/a Table 6.26: Salish languages and non-deictic D-determiners I predict that all of the non-deictic D-determiners would obligatorily take narrow scope. (Recall that in the Skwxwu7msh texts, only 9/122 of the deictic D-determiner ta was used in non-factive environments, suggesting that kwi is preferred in these contexts in order to get the narrow scope reading. I expect similar numbers in languages with non-deictic D-determiners.) I also predict that the non-deictic D-determiners would be able to be used partitively, because the speaker does not locate the referent within the group. The D-determiners should also be used in non-partitive familiar contexts, because by definition, D-determiners have domain restriction in their denotations. For most languages, I am unable to tell if these predictions hold, because the grammars do contain this level of detail. Extensive fieldwork is required in order to test my claims. 3 Implications for English The implications for English are very different than those for Salish languages. English does not encode deictic features on the D-determiner the. Instead, the analysis of Skwxwu7mesh that I have provided in this thesis raises some interesting questions as to what counts as a D-determiner. The term "determiner" is often used as a catch-all for articles, demonstratives and quantifiers, especially in English. In this section, I question whether other "determiners" occupy 209 the same position as D-determiners (D), and whether they have the same semantics as D-determiners.21 I have argued that D-determiners in Skwxwu7mesh have domain restriction in their representations. I also argued that they shared this property with English the, and extended this to all languages with overt D-determiners. I provide the denotations of the and ta in (54)a and (54)b, respectively. The domain restriction in each case is bolded. (54) a [the] = XP max(Xx [P(x) A C(x)]) b. [ta] = XP f(Xx [P(x) A C(x)]) What about quantifiers, demonstratives and indefinite articles? Do they also have domain restriction in their representations? In Chapter 4,1 argued that Skwxwu7mesh quantifiers do not have domain restriction in their denotation. This is because quantifiers and D-determiners can co-occur. What does this tell us about English? There are three possible analyses of English. First, English could be significantly different from Skwxwu7mesh (and other languages) in that it conflates the D and Q positions into one head (as argued by Szabolcsi 1994). Secondly, English could have the same structure as Skwxwu7mesh; that is, it could have both Q and D heads, and the D head could introduce domain restriction. The third potential analysis is somewhere in between; some quantifiers could be conflated, while others could not. In the next section, I provide data that suggests that the first analysis is unlikely. There is indirect evidence that some English quantifiers co-occur with a null determiner, in some contexts (see Matthewson 2004, who also argues this). However, it is difficult to determine if all quantifiers must behave this way. 2 1 Westerstahl (1984) also argued that the should be treated differently from the rest of the determiners (i.e. differently from the quantifiers). However, he argued that the should not be treated as a determiner, but instead simply domain restriction. I also argue that the has domain restriction in its representation, but it also must be more than simply domain restriction. (Some reference to uniqueness is required.) I also argue that the is a determiner, and that quantifiers belong to a different domain. 210 3 .1 Distinguishing D-determiners from quantifiers in English In a very gross sense, quantifiers and D-determiners behave semantically similarly, in that they create arguments out of predicate NPs (at least on the surface) in English. However, on a much more subtle level, they do something quite different. The goal of this thesis is to elucidate the special semantics of the D-determiners. Here I will show that quantifiers do not share the same position or the same semantics, even in English. In many languages, quantifiers do not create arguments out of predicates (Matthewson 2001, 2004). D-determiners, quantifiers and demonstratives (or some subset) can co-occur with each other. If D-determiners create arguments out of predicates, then surely quantifiers cannot be doing this as well in these languages. Once the D-determiner has created an argument, the quantifier will not apply to a predicate. Even in English, D-determiners and quantifiers behave semantically quite differently. Although Barwise and Cooper (1981) and others treat them as a unified category of functions of type « e , t > , « e , t > , t » (from sets to sets of sets), I make the distinction between quantifiers (which are functions of type « e , t > , « e , t > , t » ) and D-determiners, which are functions of type « e , t > , e > (from sets to entities), or do not change the type at all (such as kwi). In much of the traditional syntactic and semantic literature on English, what has been considered to be a determiner includes the set of all functional elements that can precede the NP within the nominal domain. (55) a. I watched the/a/one/each/every/that swan swim across the lake. b. I watched the/two/those swans swim across the lake. For example, (Abney 1987) analyzes all of these elements (cardinal numerals, quantifiers, demonstratives, and articles) as occupying the same position: D. 211 (56) However, I have shown in this thesis that this cannot capture the data in Skwxwu7mesh. Here I extend the claim that D-determiners occupy a different syntactic position than other determiners to English. 3.1.1 Evidence from the cardinal/proportional readings of weak quantifiers I suggest that proportional quantifiers occupy a higher position than D-determiners do, as shown in (57)a, and that cardinal quantifiers occupy an adjective position (57)b and c. (Partee 1987 argues that weak quantifiers in adjective position are unambigously cardinal.) (57) a. QP b. NP c. Q DP AP N D NP many many children the D NP AP N 0 children many children This analysis can account for two facts: (i) that (some) weak quantifiers can co-occur with D-determiners and (ii) that cardinal quantifiers can occur in existential sentences, and proportional quantifiers cannot. Most weak quantifiers can co-occur with the D-determiner the, demonstratives, possessors, and pronouns.22 2 2 There is at least one case where a strong quantifier can co-occur with a determiner. (i) The genie granted his every wish. Not all weak quantifiers can co-occur with determiners or demonstratives (Jackendoff 1977). (ii) * The some elves left. 212 (58) a. D Q fFred's] fmany ] "{the r" "{few }• dwarfs I those I IseveralJ IwhichJ (Jackendoff 1977: 104) b. W e few linguists have a lot of work to do. Crucially, the D-determiner can co-occur with most weak quantifiers. This can be captured by the analysis below. (59) a. DP b. DP c. DP D NP D NP D NP the the / \ the AP NP AP N A N few dwarfs several dwarfs many dwarfs When a weak quantifier occurs without a D-determiner, demonstrative, possessor or pronoun, the weak quantifier is ambiguous between a proportional and cardinal reading (Milsark 1979). (60) Many children ran around. i. There were many children who ran around. (cardinal) ii. Many of the (contextually salient) children ran around. (proportional) Under the proportional reading, the quantifier quantifies over a contextually salient set of individuals; I argue that the contextual set is introduced not by the quantifier, but by D. In the example below, for expositional clarity I abstract away from the types and treat the DP as type . Max has the same denotation as before, but here it returns a set instead of an individual. (61) QP XQ Xy rfchild'Cv) A C (y ) A Q(v)l > n] / \ rchild'(y) A C(y)| XP XQ Xy riP(v) A Q(y)| > n ] Q DP max(Xz [®child'(z) A C(z)]) |P(y)| many max(A,x [P(x) A C(X)]) D NP Xy [echild'(y)] 0 children This structure also allows us to understand why the proportional reading of weak quantifiers cannot be used in existential sentences. The null D position is associated with domain restriction 213 and assertion of uniqueness. The existential sentence is incompatible with the assertion of uniqueness.23 (62) a. There were many children in the garden. (cardinal) b. # There were the many children in the garden. c. # There were M A N Y children in the garden. (proportional) Many children is ambiguous between a cardinal reading (which is licit in existential readings) and a proportional reading (which is not) (Milsark 1979). The proportional reading is, however, not equivalent to the many X, as can be seen in familiar contexts. (63) a. I saw children wandering in the halls. The many children were chewing gum. b. I saw children wandering in the halls. Many children were chewing gum. In example (63)a, the many children refers to all of the children introduced in the previous sentence. However, in (63)b, many children refers to a subset of. the set of children introduced in the previous sentence. We therefore must distinguish between weak quantifiers in adjectival position, and those that are higher. This is not evidence that the weak quantifier (when it has a strong reading) occupies a different position than a determiner, however. The weak quantifier, when it is not adjectival, could be in the head of D. This would be a conflated analysis of the Q and D heads. In the next section, I address the possibility of a conflated analysis, and show that it cannot be correct. I argued in Chapter 4 that quantifiers in Skwxwu7mesh occupied a position above D; here I argue that quantifiers in English can occupy a position above D (in which case they receive a proportional reading), or below D (in which case they receive a cardinal/adjectival reading.) 2 31 argue that the existential is incompatible with the assertion of uniqueness of the D position, rather than with the domain restriction because in Skwxwu7mesh, the determiners (which I have already argued are associated with domain restriction) are licit in existential contexts. (i) Tsi7 ta/kwa/ti/kwi sha7yu na7 ta-n lam'. exis det ghost be.there det-lsg.poss house 'There's a ghost in my house.' 214 (64) QP DP c. NP Q many DP D 0 NP children D NP AP the many AP N many children N children I argue that weak quantifiers can only be associated with a proportional reading if they take a DP complement. 3.1.2 Evidence from domain restriction I therefore argue against a conflation analysis (cf. Szabolsci 1994) of quantifiers. Quantifiers, in the system developed here, do not occupy a D/Q position, but rather a Q position, separate from D. I claim that strong or proportional quantifiers attach above D. Indirect evidence that (most) quantifiers cannot occupy a conflated Q/D position comes from Stanley and Szabo (2000). The evidence they present shows that the quantifier itself cannot be associated with domain restriction, and that the domain restriction must be located somewhere lower than the the quantifier. They argue that their evidence shows that the NPs themselves are associated with domain restriction, but, as I showed in Chapter 3, that position is untenable. Bare nouns cannot be used to refer back to a previously mentioned referent. Instead, they can only be used to introduce a new referent. (65) a. I saw some bears last night. They were wandering around Stanley Park. Bears like to hang around the park. b. I saw some bears last night. They were wandering around Stanley Park. # I shot bears. c. I saw some bears last night. They were wandering around Stanley Park. # Bears were eating garbage. Stanley and Szabo's evidence that quantifiers themselves cannot be associated with domain restriction is given in example (66). (66) Most people regularly scream. They are crazy. (Stanley and Szabo 2000: 257) 215 There are two readings associated with the second sentence in (66): one where the pronoun they refers to all of the people in the domain (a certain village, for example), and one where it refers to those people in the village who regularly scream. They claim that this is evidence that people is associated with the domain restriction. For the first reading, they claim that "there is no single node in the logical form whose associated semantic value is the set of people in the village", if the domain variable is associated with most. If the nominal is associated with the domain restriction, however, there is a single node (the NP). (67) a. QP b. Q NP Q NP most+C people most people+C Stanley and Szabo also claim that the second reading cannot be captured by having domain restriction associated with most. They appeal to Neale's (1990) analysis of they, where it is proxy for a description which is reconstructable from the logical form of the first sentence. (68) If x is a pronoun that is anaphoric on, but not c-commanded by a non-maximal quantifier '[Dx:Fx]' that occurs in an antecedent clause '[Dx:Fx](Gx)', then x is interpreted as '[the x: Fx&Gx]. (Neale 1990: 266) According to them, if the domain restriction is associated with most, they should be interpreted as [the x: person(x) & regularly-scream (x)], which should mean everyone in the universe who regularly screams (rather than everyone in the village who regularly screams). If NPs cannot be associated with domain restriction, and quantifiers like most cannot be associated with domain restriction either, then the question becomes: where is the domain restriction? My analysis of Skwxwu7mesh determiners and quantifiers can be extended to English to solve this problem. For the first reading of (66) (where they refers to all of the villagers), we need a single node whose associated semantic value is the set of villagers. 216 (69) XP (= the set of people in the village/in C) This single node must be DP: I have already shown that the D position is associated with domain restriction in both English and Skwxwu7mesh. (70) DP (= the set of people in the village/in C) NP people Similarly, the second reading (where they refers to the villagers who regularly scream) can be solved by the presence of a D position. The (as in [the x: Fx&Gx]) is precisely the element which contains domain restriction under the approach in this thesis. The structure provided in (70) accounts for this second reading, assuming that they is determined as in (68). This argument also applies to weak quantifiers, such as many. (71) Many people regularly scream. They are crazy. The second sentence in (71) is ambiguous in the same way that (66) is. Stanley and Szabo, then, provide us with evidence that domain restriction involves a lower head than Q, but not necessarily the noun itself. Since the nominal can be shown independently not to be associated with domain restriction, we are forced to assume a null D position, which is itself associated with domain restriction. In the case of strong quantifiers, this null D must be obligatory; however, with weak quantifiers, only the proportional reading would be associated with a D position. (72) a. QP b. QP c. NP A P N many children NP people 217 This is contra von Fintel (1994), who claimed that no weak quantifiers introduce C / 4 The analysis of strong quantifiers in (72)a explains why quantifiers and determiners co-occur in some languages; the position is always available. Matthewson (1998) argues that only a subset of quantifiers introduce domain restriction. She argues that only a subset of quantifiers occupy D, and it is those quantifiers which also introduce domain restriction. Here I argue that no quantifiers introduce domain restriction, because none of them occupy D. 3.1.3 The (lack of) evidence for every I have argued above that weak quantifiers (like many) take DP complements when they are interpreted proportionately. I have also argued that at least strong quantifiers (like most) also (obligatorily) take DP complements.25 However, there is a lack of evidence for some strong quantifiers that they occupy a different position from D (like every). Some languages do distinguish between the equivalent of every and the D position. (73) D Q to26 kathe pedhi det every child 'every child' (Greek; Szabolcsi 1994:213) It is therefore possible that English does as well, covertly. Matthewson (2001) argues that every in English is not itself quantificational and occupies D. In Matthewson (1998), she argues instead that every conflates D and Q. I argue for the strongest hypothesis that every does not occupy D, and co-occurs with a D position. 3.2 Distinguishing D-determiners from demonstratives So far, I have shown that quantifiers (for the most part) must be distinguished from D-determiners. The difference between D-determiners and demonstratives, however, is more subtle. 2 4 Strictly speaking, I agree with this. However, I claim that no quantifiers restrict the domain by themselves. 2 51 treat of as meaningless, introduced for syntactic reasons. However, Giannikidou (2004) argues that of is meaningful. 2 6 This is the accusative form of the determiner. 218 The distinction I appeal to here is that D-determiners usually cannot refer to subsets of previously introduced sets, whereas demonstratives always can.2 7 For example, in discourse, if a group referent has been introduced, the D-determiner the can only be used to refer to the supremum of the set (as discussed in Chapter 3). (74) Pass me the hammers. In (74), the speaker must be referring to the entire set of contextually relevant hammers. If the speaker wants to refer to a subset of the salient group of hammers he or she is forced to choose between a partitive and a demonstrative. (75) a. Pass me two of the hammers. b. Pass me those hammers. Informally, demonstratives can be used to refer to referents (often using a pointing gesture) from a larger set. D-determiners can never be used this way. Instead, they must refer to the entire set denoted by NP that are given in any context. (76) a. Look at those penguins on the other side of the room. That penguin just stole some guy's dinner! (pointing one out) b. Look at those penguins on the other side of the room. #The penguin just stole some guy's dinner! c. Look at those penguins on the other side of the room. The penguins just stole some guy's dinner! The DP in (76)b should refer to the unique referent conforming to the NP description; however, there is more than one potential referent. The DP in (76)c refers to the maximal set of referents conforming to the NP description. Demonstratives, on the other hand, are able to refer to subsets. If the context contains 20 girls scattered throughout a room, and I wish to refer to a subset of the girls in the room, I am forced to use a demonstrative. I can also use the demonstrative to refer to the entire set, but only if the situation allows that choice (if I am outside the room containing the girls, for example). 2 7 In S k w x w u 7 m e s h , the non-deictic D-determiner can refer to a subset, much like demonstratives can. However, the reason kwi can refer to a subset is due to its lack of deictic features. 2 8 As shown in Chapter 4 , in S k w x w u 7 m e s h this is only an implicature. However, the implicature never arises with demonstratives. 219 (77) a. Those girls are really rambunctious. (= all 20 girls, or 5 girls in a corner, etc.) b. The girls are really rambunctious. (= all 20 girls) Informally, demonstratives appear to refer to a subset from a previously introduced set, whereas D-determiners can only refer to the entire set. I assume that this referring to a subset of the given set involves a (overt or covert) pointing gesture. It is still possible, however, that the demonstratives in English denote deictic features and domain restriction. This would be similar to the denotation of the Skwxwu7mesh deictic D-determiners. In languages other than English, there is evidence that demonstratives must be distinguished from D-determiners, as I showed in Chapter 3. For example, in Michif, a D-determiner is required if a demonstrative is used (Rosen 2003), as in (78). (78) awa la fij dem det girl 'that girl' (Michif; Rosen 2003: 41) As in the cases with the quantifiers, I argue that the D-determiner provides the domain restriction: a narrowing of the relevant set to the salient individuals which match the descriptive content of the NP. The demonstrative also narrows the set, but only to the set of individuals which are close or far (depending on the demonstrative; see also Chapter 4 for discussion of deictic features) from the anchor (usually speaker). I assume that the demonstrative is an adjective (following Giusti 1993 and Bernstein 1997); the set denoted by awa intersects with the set denoted by fij. (79) a. [awa] = set of individuals which are distal with respect to the speaker b. [fij] = set of girls c. [awa fij] = set of girls which are distal with respect to the speaker. d. DP (= set of salient girls, distal to the speaker) (= set of girls distal to the speaker) 220 Only the D-determiner provides C (the contextually salient set); this C intersects with the set provided by awa fij.291 assume that la has the same semantics as the does.30 (80) a. P j ] =Xx[girl'(x)] b. [awa fij]] = Xx [girl'(x) A distal-from-speaker'(x)] c. [la awa fij]] = max(A,x [girl'(x) A distal-from-speaker'(x) A C ( X ) ] ) There is no direct evidence that English (or Skwxwu7mesh) also has a D-determiner that co-occurs with demonstratives.31 However, given that quantifiers do seem to occupy a different position from D-determiners, even in English, and that many languages distinguish between demonstratives and D-determiners, I suggest that demonstratives always occupy a different position from D-determiners. This can explain why, in some languages, D-determiners are overtly required: the position D is always present. A D-determiner is needed to introduce domain restriction; demonstratives and quantifiers are unable to do this. I assume that English has the same structure, and that the D-determiner must be null (as it is with quantifiers). (81)a has the same interpretation as (80)c. set of salient girls, distal to the speaker) set of girls distal to the speaker) (81) a. those girls b. DP (= D 0 FP (= DemP NP those girls 2 9 For independent reasons, Rosen (2003) argues that the demonstrative originates in a position lower than the determiner and raises to a fronted position. 3 0 Only the domain restriction is critical for the analysis presented in this thesis; more research is required to determine if la asserts the uniqueness of its referent. 3 1 There is some data that suggests that demonstratives may occupy a different position. D-determiners are preferred over demonstratives when occuring with the adjectival version of weak quantifiers. D-determiners and demonstratives are equally good in partitive constructions. (i) The many girls were dancing by the fire. (ii) ?? Those many girls were dancing by the fire. (iii) Many of the girls were dancing by the fire. (iv) Many of those girls were dancing by the fire. 221 Languages may choose to allow this position to be covert because the presence of the demonstrative suggests the presence of higher structure. 3.3 Distinguishing D-determiners from indefinite articles I have argued that D-determiners (i) occupy D, and (ii) obligatorily include domain restriction in their denotation. The question so far has been what counts as a D-determiner? I have argued that quantifiers and demonstratives do not occupy D, nor include domain restriction in their denotation. Therefore, they cannot be D-determiners. In this section, I argue that indefinite articles are also not D-determiners, for both syntactic and semantic reasons. 3.3.1 The semantic contribution of indefinite articles A is semantically quite different from the. Similarly to the definite D-determiner, the indefinite articles cannot be used to refer to a referent that belongs to a previously introduced set. However, they cannot refer to the entire set either. For example, in (82)a, a penguin cannot refer to a member of the set of penguins introduced by those penguins; neither can sm penguins refer to a subset of the set of penguins in (82)b. In both cases, the nominal would have to introduce another referent, which is odd, given that the speaker has just pointed out a salient group of penguins. In (82)c, a penguin cannot refer to the same penguin introduced by that penguin; neither can sm penguins refer to the entire set of penguins introduced by those penguins in (82)b. (82)' a. Look at those penguins on the other side of the room. # A penguin just stole that guy's dinner! b. Look at those penguins on the other side of the room. #Sm penguins just stole that guy's dinner! c. Look at that penguin on the other side of the room. # A penguin just stole that guy's dinner! Unlike the definite D-determiner, a and sm can only be used to introduce new referents. I have argued above that the definite D-determiner cannot refer to a member of a previously introduced set because its domain is set by the context. A definite DP must refer to 222 the entire set given in the context. I argue here that the indefinite article can only introduce new referents because it does not introduce domain restriction. I claim that an indefinite DP has no access to the context. There are, however, examples of nominals referring to a member or members of a previously mentioned discourse set using a or sm. In (83)a, a student can refer to one of the students who were standing outside the factory gate, and in (83)b, sm tires can refer to some of the tires of the car Fred bought last week. (83) a. Some students were standing outside the factory gate. Bill kept his eye on them. After a little while, a student came up to him and asked him his name. b. Fred bought a car last week, and then sold sm tires to his friend. (adapted from Hawkins 1978: 174) I claim that this is similar to accidental co-reference. The speaker is introducing a new referent, but it can refer to a member of the group if the group has been made/is less salient. In (83)a, the group of students is "demoted" as a salient group by the introduction of Bill. This allows a student to introduce a new referent which can just happen to be a member of the original group. In (83)b, the tires of the car that Fred bought are never mentioned as a discourse topic at all. This 32 allows sm tires to (indirectly) refer to the tires of the car. Further, the examples in (83) do not have to refer to a member of the discourse set (Hawkins 1978). In fact, some English speakers disprefer the reading where the referent is part of the previously introduced set. These speakers are forced to use a partitive to force the partitive reading.33 (84) a. Some students were standing outside the factory gate. Bill kept his eye on them. After a little while, one of the students came up to him and asked him his name. b. Fred bought a car last week, and then sold some of the tires to his friend. I argue this is because nominals with a and sm do not have access to the context. 3 2 In fact, if sm tires refers to the tires of the car, it can refer to all of them, or all of them plus some other tires, from a different car. 3 3 1 find it especially difficult to interpret sm tires partitively; other speakers cannot interpret either the plural or singular examples partitively. 223 3.3.2 The syntactic position of indefinite articles My analysis forces me to claim that a occupies a different position than the. On independent grounds, the indefinite article a is argued to occupy a different position than that of the (Epstein 1999, Lyons 1999, Borer 2005). One of Epstein's arguments is based on the distribution of the, two, such and a. The must precede the cardinal two (85)a and two must precede such (85)b. However, such must precede a (85)d. The and a occupy different syntactic positions.34 (85) a. The two cars are safe. b. Two such cars are safe. c. Most of the cars here are unsafe. But a couple of cars have been built more sturdily. The two such cars are safe. d. Such a car is safe. e. * Such the car is safe.35 Epstein argues that a occupies a lower projection than the (NumP). While the syntactic evidence given here is not strong, it is at least consistent with the idea that a and the occupy different positions. It is also not important exactly which position the indefinite article occupies, only that it does not occupy D (and is therefore not a D-determiner). The analysis given in this thesis provides us with a way to explain the observation that a seems to be different from the, both syntactically and semantically. There is a unified syntactic and semantic constraint on D-determiners. A and sm do not occupy D, nor do they include domain restriction in their denotation. Therefore, neither of them are D-determiners. There is still a remaining problem: if other indefinites can co-occur with D, why can't a co-occur with the? One way around this problem is to claim that the and a can co-occur, but the effects are masked. Perlmutter (1970) argues that a is the unstressed variant of one, since wherever stressed a would be expected, one is found instead. 3 4 This argument does not work in other Germanic languages (Greg Carlson, p.c). The syntactic arguments may be different in each language. 3 5 Independently, however, the and such cannot co-occur (Bresnan 1973), unless something follows the, such as first, only or numerals (Landman 2006). (i) * The such car is safe. (ii) The only such car is safe. (iii) The one such car is safe. (iv) The first such car was safe. 224 (86) a. I bought a book. b. * I bought a book. c. * I bought one book. d. I bought one book. If this is correct, the and a can co-occur, but only if the numeral receives stress. (87) I bought the one book. However, in certain contexts, a can still be stressed, as shown in (88). (88) a. That wasn't a reason I left Pittsburgh, it was the reason. b. He was a friend; I had others. (Abbott 1999) On the other hand, it is only possible to stress a when it is explicitly contrasted with the alternatives, and is metalinguistically negated by the use of the (cf. Horn 1985). (89) a. * That wasn't a reason I left Pittsburgh. I therefore adopt this analysis of a as unstressed one. 3.4 Summary I argue that only elements which occupy D and have domain restriction in their denotations are D-determiners. I provided indirect evidence that quantifiers do not occupy D, even in English. Demonstratives do not occupy D, at least in some languages; I claimed that English demonstratives could be analyzed the same way. Indefinite articles were also shown not to occupy D, nor to include domain restriction in their denotation. I claimed that only elements which are constrained by the context in a very particular way can be called D-determiners. I make the strong claim that D is sensitive to the context and that nothing else is. b. * He was a friend. 225 (90) If a nominal is introduced by a D (overtly or covertly), it will be restricted by C. If a nominal lacks D, it will not be restricted by C. Bare nouns are not restricted by the domain because they lack a determiner. Only quantifiers under a cardinal reading, indefinite nominals and bare nouns lack a D-determiner, which in turn means they lack domain restriction. (91) a. strong/proportional Q b. cardinal Q/indefmite36 QP QP O DP 0 NP D NP full DP d. bare noun DP NP D NP By my arguments given above, it follows that English has only two D-determiners: the and the null D that co-occurs with quantifiers. However, other languages (such as Skwxwu7mesh) have more than one overt D-determiner. There is no reason why a language should have more than one overt D-determiner, unless other features are encoded, such as deictic features. I therefore argue for the special status of D, not only in Skwxwu7mesh, but in English as well. I argue that D-determiners occupy a different position from quantifiers and demonstratives. (92) Determiners are D-determiners iff they occupy D. 4 Implications for "articleless" languages My claim that only D is sensitive to the context has implications for languages which lack overt D-determiners. Unfortunately, I cannot do these languages justice here, and it must remain a 3 6 1 am ignoring the exact position of a, which may be in a different position than the weak quantifiers. Nothing hinges on the exact position of any of these; the only restriction is that they cannot occupy D . 226 topic for future research. I briefly outline the predictions of my analysis for articleless languages here. On the surface, languages like Mandarin Chinese do not have D-determiners. However, bare nouns can get definite interpretations in many contexts ((93)b and c), as well as indefinite ((93)a) (Cheng and Sybesma 1999). (93) a. Hufei mai shu qu le. Hufei buy book go sfp 'Hufei went to buy a book/books.' b. Hufei he-wan-le tang. Hufei drink-finish-LE soup 'Hufei finished the soup.' c. Gou yao guo malu. dog want cross road 'The dog wants to cross the road.' (Mandarin; Cheng and Sybesma 1999: 510) Indefinite interpretations are expected for NPs; as I have argued, bare nouns are not associated with C, and must introduce a new set to the discourse. It is the definite interpretation which concerns us here. How, under the analysis I have given in this thesis, can a bare noun be interpreted as a definite? The analysis presented in this thesis allows for only two analyses of languages like Mandarin Chinese. If these nominals are in fact bare NPs, then the only way that they can be co-referent is "accidentally", as with indefinite nominals in English. We would then expect readings where the nominal refers to the previously introduced referent and other referents. The other potential analysis is to claim that, when the nominals are interpreted as definites, there is a null D, and it is that D which is supplying the domain restriction. (94) a. ^ D P ^ b. D = max(A,x [P(x) A C(x)])37 D NP 0 gou 'the dog' 3 7 Again, I am assuming that the null determiner asserts the uniqueness of its referent. This is not necessarily true, and needs to be tested. 227 This is the analysis which I adopt here. There is no syntactic evidence for this functional structure, but semantically, it is a more coherent picture of the split between the indefinite and definite readings. 5 Implications for Maori There are many other languages that have determiners which would be relevant to the discussion here. I restrict the discussion to one more language: Maori. Maori has six different articles which potentially occupy D. Three of these articles are 38 indefinite and four are definite (Chung and Ladusaw 2004). indefinite definite aforementioned singular he, tetahi te taua plural etehi nga aua Table 6.27: The artic e system of of Maori (ac apted from Chung and Ladusaw 2004: 23-33). In the system developed in this thesis, indefinite articles cannot occupy D. Anything that occupies D should have domain restriction in its denotation. Semantically, the indefinite articles do not appear to introduce domain restriction over their NP. They can only introduce new referents (Chung and Ladusaw 2004). In (95)a, etehi 'some' in the first clause must refer to a different set of individuals than the etehi 'others' in the second clause. Similarly, in (95)b, he must refer to a different part of the fort than the previously mentioned part (tend wahi 'that place'). (95) a. Ka moe etehi, ka ara ko etehi ki taratou mahi. T sleep art.pl T awake Ident art.pl to their work 'While some slept, others stayed awake to keep up the work.' Bauer (1993) also argues that there is a separate D-determiner a for proper names and pronouns. 228 b. Taraua mahi he whakaputa ake i tena wahi their.du work Pred.a appear up at dem place o te pa, ka rere, a ka whakaputa ake of det fort T rush and T appear up he wahi ano. art place again 'They appear at [that] part of the fort, then rushed to another part.' (Maori; Jones and Biggs 1995: 191, cited by Chung and Ladusaw 2004: 33) Furthermore, two of the indefinite articles (tetahi and eteh'i) appear to be demonstratives or weak quantifiers, as they can occur without a following NP, as shown for the plural in (95)a. They do not occupy D. What about the remaining indefinite article, he? It has a much more restricted distribution than the other indefinite articles do (Hale and Hohepa 1969; Chung and Ladusaw 2004). Chung and Ladusaw discuss four differences: (i) only he can act as a pivot of an existential sentence, (ii) he cannot follow a preposition, (iii) he can only introduce an internal subject and (iv) he nominals are predicational. The most telling difference between tetahi and he is the final difference. The article he is predicational, whereas the other indefinite articles are identificational. (96) a. Ko tetehi pakanga kaha tena. Ident art battle strong dem 'It was a fierce battle.' (Maori; Jones and Biggs: 369, cited in Chung and Ladusaw 2004: 62) b. He korero ataahua tena. art speech beautiful dem 'That's a beautiful saying.' (Maori; Karetu 1974: 61, cited in Chung and Ladusaw 2004: 63) c. * Tetehi korero ataahua tena. art speech beautiful dem (Maori; Chung and Ladusaw 2004: 64) Tetahi nominals cannot be used as predicates, whereas he nominals can. This is strikingly similar to the use of a in English in predicate position. A nominals can be used in predicate position, while the English quantifier some is degraded. (97) a. I am a linguist. b. ?? We are some linguists. 229 Recall that Skwxwu7mesh kwi, which I showed to be a D-determiner in Chapter 5, also cannot be used in predicate position. (98) a. Slhanay' lha Kirsten. woman det.f Kirsten 'Kirsten is a woman.' b. * Kwi slhanay' lha Kirsten. det woman det.f Kirsten It is plausible that he is not a D-determiner, and occupies a position in NumP, like English a. Assuming this is correct, Maori has only four D-determiners: te, nga, taua and aua. This language also makes a distinction between (potentially) wide scope nominals and obligatorily narrow scope nominals (Chung and Ladusaw 2004). Tetahi or etehi allow the nominal to take wide or narrow scope; he nominals can only take narrow scope. (99) a. Kaore tetahi tangata i waiata mai. T.not art person T sing to.here 'A (particular) person didn't sing.' (= There was a person who didn't sing; wide) b. Kaore and tetahi tangatakia taha i te ara. T.not yet art person T pass on det path 'No one had yet passed along the track.' (H.M. Ngata 1994: 304; narrow) c. Kaore he take kotahi. T.not art reason one 'There's no reason at all.' (lit. there is not one reason; narrow) I have already suggested that none of these are D-determiners. However, my analysis predicts that there should be a difference between the articles that take wide or narrow scope and the article which only takes narrow scope. If tetahi or etehi are demonstratives, they should have deictic features. As they do not contrast with other indefinite demonstratives, the feature involved would have to be [neutral] (following Imai 2003). On the other hand, if they are weak 39 quantifiers, the quantificational feature would allow the nominals to take any scope. He, which I have suggested occupies a position lower than D, would have to lack all of these features. 3 9 See Bauer (1993) who suggests tetahi and etehi are quantifiers on the basis of the fact that they can take partitive structures. 230 Furthermore, o n l y the definite and aforementioned articles w o u l d o c c u p y D and have d o m a i n restrict ion i n their denotations. T h e rest o f the system w o u l d not. 6 Conclusion In this thesis I have made the f o l l o w i n g c la ims. (100) a. A determiner is a D-determiner i f f it occupies D . b. D o m a i n restrict ion is o n l y associated w i t h the p o s i t i o n D . True bare nouns do not have d o m a i n restriction. c. D-determiners share a core semantics ( d o m a i n restrict ion); D-determiners m a y have other features (such as assertion o f uniqueness, deict ic features, etc.) d. Skwxwu7mesh D-determiners are split into two groups: those that have deict ic features, and those that d o not. e. D e i c t i c features force D P s to be able to take w i d e scope/be c o m p o s e d v i a Specify . N o n - d e i c t i c D P s must take narrow scope. f. Restr ict is a last-resort c o m p o s i t i o n type: non-deict ic D P s must be c o m p o s e d v i a Restr ict because they lack features that w o u l d otherwise force t h e m to compose v i a Specify . g. Def in i te articles are D-determiners; indefinite articles are not. T h e c l a i m s h a d i m p l i c a t i o n s for other S a l i s h languages. W h i l e some o f the m a i n c l a i m s c o u l d not be tested without f ie ldwork , the presence o f a non-deict ic D-determiner is at least a plausible analysis o f m a n y o f the S a l i s h languages for w h i c h there is enough data. S o m e languages do not have anything l i k e a non-deict ic D-determiner. These c la ims also had impl icat ions for what is a D-determiner i n other languages, i n c l u d i n g E n g l i s h . T h e behaviour o f weak quantifiers was e x p l a i n e d i n terms o f the presence or absence o f the D pos i t ion . A t least some strong quantifiers were s h o w n to o b l i g a t o r i l y occur w i t h a n u l l D-determiner. I argued that demonstratives i n some languages c o u l d not be associated w i t h j (i) Ka mahue mai etahi o ona tangata. T/A leave hither art.pl gen gen.pl people 'Some of the people were left behind.' (Maori; Bauer 1993: 300) 231 domain restriction (as they co-occur with D-determiners); this analysis was extended to English and Skwxwu7mesh demonstratives. The claims were also extended to "articleless" languages, such as Mandarin Chinese. I claimed that a D-determiner must be (covertly) present in certain contexts. The determiner system of Maori was also investigated in light of my claims. Maori indefinite articles were claimed not to occupy D, based on their semantics. The data in Skwxwu7mesh also provide us with evidence for a third category: definite, indefinite and non-definite. Non-definites can be used in both novel and familiar cases, but behave much like definites in familiar contexts. 232 References Abbott, Barbara. 1999. Support for a Unique Theory of Definiteness. Proceedings from Semantics and Linguistic Theory 9:1-15. Abney, Stephen. 1987. The English Noun Phrase in its Sentential Aspect. PhD dissertation, MIT. Baauw, Sergio. 2001. Expletive Determiners in Child Dutch and Spanish. BUCLD Proceedings 25. Somerville, Mass: Cascadilla Press. Bar-el, Leora, and Linda Tamburri Watt. 1998. What Determines Stress in Skwxwu7mesh (Squamish)? Papers for the 33rd ICSNL, Seattle: University of Washington. Barwise, Jon, and Robin Cooper. 1981. Generalized Quantifiers and Natural Language. Linguistics and Philosophy 4:159-219. Beaumont, Ron. 1985. She Shashishalhem, the Sechelt Language: Language, stories, and sayings of the Sechelt Indian people of British Columbia. Penticton, BC: Theytus Books. Beghelli, Filippo, and Tim Stowell. 1997. Distributivity and Negation: The Syntax of Each and Every. Ways of Scope Taking, ed. Anna Szabolcsi, 72-107. Dordrecht: Kluwer. Bernstein, Judy B. 1997. Demonstratives and Reinforcers in Romance and Germanic Languages. Lingua 102: 87-113. Bittner, Maria. 1994. Case, Scope, and Binding. Dordrecht: Kluwer Academic Publishers. Borer, Hagit. 2005. In Name Only. Structuring Sense, Volume I. Oxford: Oxford University Press. Borthen, Kaja. 2003. Norwegian Bare Singulars. PhD dissertation, Norwegian University of Science and Technology. Breheny, Richard 2003. A Lexical Account of (implicit) Bound Contextual Dependence. SALT XII, University of Washington at Seattle (May 9-11). Bresnan, Joan. 1973. Syntax of the Comparative Clause Construction in English. Linguistic Inquiry 4:275-343. Brisson, Christine. 2003. Plurals, All, and the Nonuniformity of Collective Predication. Linguistics and Philosophy 26. 129-184. Burton, Strang. 1997. Past Tense on Nouns as Death, Destruction, and Loss. Proceedings of the North East Linguistic Society 27, ed. K. Kusumoto. Amherst: GLSA, University of Massachusetts. 231 Carlson, Greg. N. 1989. The Semantic Contribution of English Generic Sentences, Properties, Types and Meaning, G. Partee and E. Turner (eds), Dordrecht, Kluwer. Carlson, Greg N. 1980. Reference to Kinds in English. New York: Garland Publishing. Cheng, Lisa Lai-Shen and Rint Sybesma. 2005. Classifiers in Four Varieties of Chinese. The Oxford Handbook of Comparative Syntax, eds. Guglielmo Cinque and Richard S. Kayne, 259-292. New York: Oxford University Press, Cheng, Lisa Lai-Shen, and Rint Sybesma. 1999. Bare and Not-So-Bare Nouns and the Structure of NP. Linguistic Inquiry 30: 509-542. Chierchia, Gennaro. 2001. Plurality of Mass Nouns and the Notion of 'Semantic Parameter'. Events in Grammar, ed. Sam Rothstein, 53-103. Dordrecht: Kluwer. Chierchia, Gennaro. 1998. Reference to Kinds Across Languages, Natural Language Semantics 6, 339-405. Christophersen, Paul. 1939. The articles. A study of their theory and use in English. Copenhagen: Munksgaard. Chung, Sandra and William A. Ladusaw. 2004. Restriction and Saturation. Cambridge: MIT. Clark, Herbert H. , and Susan E. Haviland. 1977. Comprehension and the Given-New Contract. Discourse Production and Comprehension, ed. Roy O. Freedle, 1-40. Norwood, New Jersey: Ablex. Clark, Herbert.H., and Catherine R. Marshall. 1981. Definite Reference and Mutual Knowledge. Elements of Discourse Understanding, eds. Aravind K. Joshi, Bonnie L. Webber and Ivan A. Sag, 10-63. Cambridge: Cambridge University Press. Currie, Elizabeth. 1997. Topic Time: the Syntax and Semantics of SqwXwu7mish Temporal Adverbials. M A thesis, University of British Columbia. Davis, Henry. 2005. On the Syntax and Semantics of Negation in Salish, International Journal of American Linguistics 71: 1-55. Davis, Henry 1999. Word Order and Configurationality in St'at'imcets, Papers for the Thirty-Fourth International Conference on Salishan and Neighbouring Languages, 291-316. Kamloops, BC: Secwepmec Cultural Education Society. Davis, Henry and Lisa Matthewson. 2003. Quasi Objects in St'at'imcets: The (Semi)-Independence of Agreement and Case. Formal Approaches to Function: Papers in Honor of Eloise Jelinek, eds. Andrew Carnie, Heidi Harley and Mary Ann Willie, 76-106. Oxford: Blackwell. Davis, Philip W. and Ross Saunders. 1975. Bella Coola nominal deixis. Language 51: 845-858. 232 Davis, Philip W. and Ross Saunders. 1974. Bella Coola nominal deixis. Paper presented at the 9th International Conference on Salishan and Neighbouring Languages. Dayal, Veneeta. 1999. Bare NPs, Reference to Kinds, and Incorporation. Proceedings of SALT IX. Dechaine, Rose-Marie and Martina Wiltschko. 2002. Decomposing Pronouns. Linguistic Inquiry 33: 409-442. Demirdache, Hamida. 1997. On Descriptions in (Lillooet) Salish. Paper presented at the Colloque de Syntaxe et de Semantique de Paris. Demirdache, Hamida. 1996. The Temporal Reference of Noun Phrases in St'at'imcets. Paper presented at the Linguistic Society of America, San Diego. Diesing, Molly. 1992. Indefinites. Cambridge: MIT Press. Ebert, Karen H. 1971. Zwei Formen des bestimmten Artikels. Problem und Fortschritte der Transformationsgrammatik, ed. D. Wunderlich, 159-74. Munich: Hueber. van Eijk, Jan, and Lorna Williams. 1981. Cuystwi malh ucwalmicwts: Lillooet legends and stories. Mount Currie: Ts'zil Publishing House. Enc, Murvet. 1991. The Semantics of Specificity. Linguistic Inquiry 22: 1-25. Epstein, Melissa A. 1999. On the Singular Indefinite Article in English. UCLA Working Papers in Syntax. Farkas, Donka F. and Henriette de Swart. 2004. The Semantics of Incorporation: From Syntax to Discourse Transparency. Stanford, CA: CSLI. Fillmore, Charles. 1997 [1975]. Lectures on Deixis. CSLI. Fillmore, Charles. 1982. Towards a Descriptive Framework for Spatial Deixis. Speech, Place and Action, eds. Robert J. Jarvella and Wolfgang Klein, 31-59. New York: John Wiley and Sons. von Fintel, Kai. 2004. Would You Believe It? The King of France is Back! Presuppositions and Truth-Value Intuitions. Descriptions and Beyond, eds. Marga Reimer and Anne Bezuidenhout. Oxford: Oxford University Press. von Fintel, Kai. 1999. Quantifier Domains and Pseudo-Scope. Ms, MIT. Handout of a talk given at at Cornell Context-Dependence Conference, March 28. von Fintel, Kai. 1998. The Semantics and Pragmatics of Quantifier Domains. Ms, MIT. Handout of a talk given at the Vilem Mathesius Center in Prague, March. von Fintel, Kai. 1994. Restrictions on Quantifier Domains. PhD dissertation, University of Massachusetts, Amherst. 233 von Fintel, Kai, and Irene Heim. 2001. Focus and Clefts, handout from Pragmatics in Linguistic Theory class at MIT. Fodor, Janet D. and Ivan Sag. 1982. Referential and Quantificational Indefinites. Linguistics and Philosophy 5: 355-398. Frege, Gottlob. 1997 [1892]. On Sense and Reference. Readings in the Philosophy of Language, ed. Peter Ludlow, 563-583. Cambridge: MIT Press. Galloway, Brent Douglas. 1993. A grammar of upriver Halkomelem. Berkeley, Calif: University of California Press. van Geenhoven, Veerle. 1998. Semantic Incorporation and Indefinite Descriptions: Semantic and Syntactic Aspects of Noun Incorporation in West Greenlandic. Stanford: CSLI Publications. Gerdts, Donna. 1988. Object and Absolutive in Halkomelem Salish. New York: Garland. Gerstner, C. and M . Krifka. 1993. Genericity, Handbuch der Syntax, J. Jacobs, A. von Stechow, W. Sternefeld and T. Venneman (eds), de Gruyter, Berlin. Giannakidou, Anastasia. 2004. Domain Restriction and the Arguments of Quantificational Determiners, Proceedings of SALT XIV, ed. Robert B. Young. C L C Publications. Gillon, Carrie. 2003. The Semantic Composition of DPs in Skwxwu7mesh. Ms. University of British Columbia. Gillon, Carrie. 2002. Negation and Clause Structure in Skwxwu7mesh. Ms. University of British Columbia. Gillon, Carrie. 1998a. Extraction from Skwxwu7mesh relative clauses. Proceedings from the Northwest Linguistics Conference, held at Simon Fraser University, March 7-8, 1998. Gillon, Carrie. 1998b. C-command and quantifiers in Skwxwu7mesh Salish, late paper given at 33rd International Conference on Salish and Neighbouring Languages, University of Washington, August 8-10. Giusti, Giuliana. 1995. Heads and Modifiers among Determiners: Evidence from Rumanian. Advances in Roumanian Linguistics, eds. Guglielmo Cinque and Giuliana Giusti, 103-125. Amsterdam: John Benjamins Publishing Company. Giusti, Giuliana. 1994. Enclitic Articles and Double Definiteness: A Comparative Analysis of Nominal Structure in Romance and Germanic. The Linguistic Review 11: 241-255. Giusti, Giuliana. 1993. La sintassi dei determinanti, Padova: Unipress. Giusti, Giuliana, 1991. The Categorial Status of Quantified Nominals, Linguistische Berichte 136, 438-452. 234 Givon, Talmy. 1978. Definiteness and referentiality. Universals of Human Language, ed. J. Greenberg. Stanford, CA, Stanford University Press 4: 291-330. Glougie, Jennifer. 2000. Blackfoot "Indefinites": Bare Nouns and Non-assertion of Existence. WCCFL 19, 125-138. Grice, H. Paul. 1975. Logic and Conversation. Syntax and Semantics 3: 41-58. Grosz, Barbara J. 1977. The Representation and Use of Focus in a System for Understanding Dialogues, Stanford: Stanford Research Institute. Hawkins, John A. 1991. On (In)defmite Articles: Implicatures and (Un)grammaticality Prediction. Journal of Linguistics 27, 405-442. Hawkins, John A. 1978. Definiteness and Indefiniteness : A Study in Reference and Grammaticality Prediction. London: Croom Helm. Heim, Irene. 1991. Articles and Definiteness. Published in German as "Artikel und Definitheit." Semantics: An International Handbook of Contemporary Research, eds. Arnim von Stechow and Dieter Wunderlich, 487-535. Berlin: de Gruyter. Heim, Irene. 1988. The Semantics of Definite and Indefinite Noun Phrases. New York: Garland. Heim, Irene, and Angelika Kratzer. 1998. Semantics in Generative Grammar, Oxford: Blackwell. Hess, Thom. 1995. Lushootseed Reader with Introductory Grammar, Volume I: Four Stories from Edward Sam. University of Montana Occasional Papers in Linguistics, No. 11. Higginbotham, James. 1985. On Semantics. Linguistic Inquiry 16: 547-593. Horn, Laurence. 1985. Metalinguistic Negation And Pragmatic Ambiguity. Language 61: 121-174. Imai, Shingo. 2003. Spatial Deixis. PhD dissertation, State University of New York at Buffalo. Jackendoff, Ray S. 1977. X' syntax: a study of phrase structure. Cambridge: MIT Press. Jelinek, Eloise. 1995. Quantification in Straits Salish. Quantification in Natural Languages, eds. Emmon Bach, Eloise Jelinek, Angelika Kratzer and Barbara Partee, 487-540. Dordrecht: Kluwer. Jelinek, Eloise and Richard Demers. 1994. Predicates and pronominal arguments in Straits Salish. Language 70: 697-736. de Jong, Franciska. 1987. The Compositional Nature of (In)definiteness. The Representation of (In)definiteness, eds. Eric. Reuland and Alice ter Meulen, 286-317. Cambridge: MIT Press. 235 Kadmon, Nirit. 2001. Formal Pragmatics: Semantics, Pragmatics, Presupposition, and Focus. Oxford: Blackwell Publishers. Kadmon, Nirit. 1992. On Unique and Non-unique Reference and Asymmetric Quantification. New York: Garland. Karttunen, Lauri. 1976. Discourse Referents. Syntax and Semantics 7, ed. James McCawley, 363-385. New York: Academic Press. Karttunen, Lauri. 1974. Presuppositions and Linguistic Context. Theoretical Linguistics 1: 181-194. Kinkade, M . Dale. 2004. Cowlitz dictionary and grammatical sketch. University of Montana Occasional Papers in Linguistics 18. Missoula. Kinkade, M . Dale. 1964. Phonology and Morphology of Upper Chehalis: IV. International Journal of American Linguistics 30: 251-260. Koch, Karsten. 2006. Nominal modification in Nlhe7kepmxcin (Thompson River Salish) MIT Working Papers on Endangered and Lesser Known Languages: Studies in Salishan, eds. S.T. Bischoff, L. Butler, P. Norquest and D. Siddiqi, 113-126. Kratzer, Angelika. 2004. Covert Quantifier Restrictions in Natural Languages, talk given at the University of British Columbia. Kratzer, Angelika. 1995. Stage-level and Individual-level Predicates, The Generic Book, eds. G. Carlson and J. Pelletier, The University of Chicago Press. Krifka, Manfred, Francis Jeffry Pelletier, Gregory N. Carlson, Alice ter Meulen, Gennaro Chierchia, and Godehard Link. 1995. Genericity: An Introduction. The Generic Book, eds. Gregory N. Carlson and Francis Jeffry Pelletier. Chicago: The University of Chicago Press. Kroeber, Paul. 1999. The Salish Language Family: Reconstructing Syntax. Lincoln: University of Nebraska Press in cooperation with the American Indian Studies Research Institute, Indiana University, Bloomington. Kuipers, Aert H. 1974. The Shuswap Language. The Hague: Mouton. Kuipers, Aert H. 1967. The Squamish Language: Grammar, Texts, Dictionary. The Hague: Mouton. Landman, Meredith L. 2006. Variables in Natural Language. PhD dissertation, University of Massachusetts, Amherst. Lewis, David. 1979. Scorekeeping in a Language Game. Journal of Philosophical Logic 8, 339-359. 236 Levinson, Stephen C. 1983. Pragmatics. Cambridge: Cambridge University Press. Link, Godehard. 1983. The Logical Analysis of Plurals and Mass Terms: A Lattice-theoretical Approach. Meaning, Use, and Interpretation of Language, eds. Rainer Bauerle, Christoph Schwarze and Arnim von Stechow, 302-323. Berlin: de Gruyter. Longobardi, Giuseppe. 1994. Reference and Proper Names: A Theory of N-Movement in Syntax and Logical Form. Linguistic Inquiry 25: 609-665. Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press. Lyons, John. 1979. Deixis and Anaphora. The Development of Conversation and Discourse, ed. Terry Myers. Edinburgh: Edinburgh University Press. Marinis, Theodore. 1998. The Acquisition of Expletive Definite Articles in Modern Greek. Proceedings of ConSOLE 6, eds. T. Cambier-Langeveld, A. Liptak and M . Radford, 169-184. Marti, Luisa. 2003. Contextual Variables, PhD dissertation, University of Connecticut. Matthewson, Lisa. 2005. An Unfamiliar Proportional Quantifier. Paper given at the QP/Nom/DP Workshop, Universitat des Saarlandes, December 16. Matthewson, Lisa. 2004. Strategies of Quantification: Evidence from the World. Paper given at Strategies of Quantification, York, July 15-17. Matthewson, Lisa. 1999. On the Interpretation of Wide-Scope Indefinites. Natural Language Semantics 7: 79-134. Matthewson, Lisa. 1998. Determiner systems and quantificational strategies: Evidence from Salish. The Hague: Holland Academic Graphics. McNally, Louise. 2004. Bare Plurals in Spanish are Interpreted as Properties. Catalan Journal of Linguistics 3: 115-133. Miller, Karen and Cristina Schmitt. 2005. The Interpretation of Indefinites and Bare Singulars in Spanish Child Language. Proceedings of the Seventh Hispanic Linguistics Symposium and the Sixth Conference on the Acquisition of Spanish and Portuguese as First and Second Languages 2003. Milsark, Gary. 1979. Existential Sentences in English. New York: Garland. Mithun, Marianne. 1984. The Evolution of Noun Incorporation. Language 62: 847-894. Montler, Timothy R. 1986. An Outline of the Phonology and Morphology of Saanich, North Straits Salish. University of Montana Working Papers in Linguistics 4. Muller, Ana. 2005. Bare Nominals, Number and the Count-Mass Distinction in Brazilian Portuguese. Talk given at the University of British Columbia, October 14. 237 Nater, Hank F. 1984. The Bella Coola Language. Mercury series; Canadian ethnology service No. 92. Ottawa: National Museums of Canada. Neale, Stephen. 1990. Descriptions. Cambridge: MIT Press. Partee, Barbara. 1991. Adverbial Quantification and Event Structures. Proceedings of the Seventeenth Annual Meeting of the Berkeley Linguistic Society. University of California, Berkeley. Partee, Barbara. 1987. Noun Phrase Interpretation and Polymorphic Types. Studies in Discourse Representation Theory and the Theory of Generalized Quantifier, eds. Jeroen Groenendijk, Dick de Jongh and Martin Stokhof. Dordrect: Foris. Partee, Barbara, Emmon Bach, and Angelika Kratzer. 1987. Quantification: A Cross-Linguistic Investigation, NSF proposal, University of Massachusetts at Amherst. Percus, Orin J. 1998. Aspects of A. PhD dissertation, MIT. Perlmutter, David. 1970. On the article in English. Progress in Linguistics, eds. M . Bierwisch and K. Heidolph,. The Hague: Mouton. Perez-Leroux, Ana T., and Thomas Roeper. 1999. Scope and the Structure of Bare Nominals: Evidence from Child Language, Linguistics 37, 927-960. Postal, Paul. 1969. Anaphoric Islands. Papers from the Fifth Regional Meeting Chicago Linguistic Society, eds. Robert I. Binnick et. al., 209 - 239. Prince, Ellen F. 1992. The ZPG Letter: Subjects, Definiteness, and Information-Status. Discourse Description: Diverse Linguistic Analyses of a Fund-Raising Text, eds. William C. Mann and Sandra A. Thompson, 295-325. Amsterdam: Benjamins. Prince, Ellen F. 1981. On the Inferencing of Indefinite-this NPs. Elements of Discourse Understanding, eds. Aravind K. Joshi, Bonnie L. Webber, and Ivan A. Sag, 231-250. Cambridge: Cambridge University Press. Reinhart, Tanya. 1997. Quantifier Scope: How Labor Is Divided between QR and Choice Functions. Linguistics and Philosophy 20: 335-397. Roberts, Craige. 1993. Uniqueness Presuppositions in Definite Noun Phrases, ms, The Ohio State University. Roberts, Taylor. 1999. Grammatical Relations and Ergativity in St'at'imcets (Lillooet Salish), International Journal of American Linguistics, v. 65, 275-302. Rosen, Nicole. 2003. Demonstrative Position in Michif. The Canadian Journal of Linguistics/La Revue canadienne de Linguistique 48: 39-69. 238 Rullmann, Hotze and Aili You. 2003. General Number and the Semantics and Pragmatics of Indefinite Bare Nouns in Mandarin Chinese. Ms, University of Calgary. Russell, Bertrand. 1998 [1905]. On Denoting. Definite Descriptions: A Reader, ed. Gary Ostertag, 35-49. Cambridge, MA: MIT Press. Ruys, Eddy. 1992, The Scope of Indefinites, PhD dissertation, Utrecht University. Published in the OTS Dissertation Series, Utrecht. Schlenker, Philippe. 2003. Indexicality, Logophoricity, and Plural Pronouns. Research in Afroasiatic Grammar II (Selected Papers from the Fifth Conference on Afroasiatic Languages, Paris, 2000), ed. Jacqueline Lecarme, 409-428. Benjamins. Schmitt, Cristina and Alan Munn. 1999. Against the nominal mapping parameter: bare nouns in Brazilian Portuguese. Proceedings of NELS 29. Searle, John R. 1969. Speech Acts: An Essay in the Philosophy of Language. Cambridge: Cambridge University Press. Sells, Peter. 1985. Restrictive and Non-Restrictive Modification. Report #CSLI-85-28, CSLI, Stanford University. Sobin, Nicholas. 1995. On case assignment in Ukrainian morphologically passive constructions. Linguistic Inquiry 16: 649-662. Stalnaker, Robert C. 1974. Pragmatic Presuppositions. Semantics and Philosophy, eds. Milton K. Munitz and Peter K. Unger, 197-213. New York: New York University Press. Stalnaker, Robert C. 1973. Presuppositions. Journal of Philosophical Logic 2: 447-457. Stanley, Jason. 2002. Nominal Restriction. Logical Form and Language, eds. G. Peters, and G. Preyer, 365-388. Oxford: Oxford University Press. Stanley, Jason, and Zoltan Gendler Szabo. 2000. On Quantifier Domain Restriction. Mind & Language 15: 219-261. Stowell, Tim. 1989. Subjects, Specifiers and X-bar Ttheory. Alternative Conceptions of Phrase Structure, eds. Mark Baltin and Anthony Kroch, 232-262. Chicago: The University of Chicago Press. Strawson, P. F. 1998 [1950]. On Referring. Definite Descriptions: A Reader, ed. Gary Ostertag, 135-160. Cambridge, MA: MIT Press. Suttles, Wayne. 2004. Musqueam Reference Grammar. Vancouver: UBC Press. Szabolcsi, Anna. 1994. The Noun Phrase. The Syntactic Structure of Hungarian, Syntax and Semantics; v. 27, 179-275. New York: Academic Press. 239 Szabolcsi, Anna. 1987. Functional Categories in the Noun Phrase. Approaches to Hungarian, Volume 2, ed. Istvan Kenesei, 167-189. Szeged: JATE. Thompson. Laurence C. 1979. The Control System: A Major Category in the Grammar of Salishan Languages. British Columbia Provincial Museum Heritage Record No. 4, ed., Barbara Efrat, 156-176. Thompson, Laurence C . and M . Dale Kinkade. 1990. Languages. Handbook of North American Indians - Northwest Coast, Volume 7, ed. Wayne Suttles. 30-51. Washington D . C : Smithsonian Institution. Vergnaud, Jean-Roger, and Maria Luisa Zubizarreta. 1992. The definite determiner and the inalienable construction in French and English, Linguistic Inquiry 23, v. 4, 595-652. Werle, Adam. 2000. Semantic Incorporation in Lillooet. University of British Columbia Working Papers in Linguistics 3: Proceedings of the 35th International Conference on Salish and Neighboring Languages, eds. Suzanne Gessner & Sunyoung Oh, Vancouver: U B C W P L . 219-225. Westerstahl, D . 1984. Determiners and Context Sets. Generalized Quantifiers in Natural Languages, (eds.) Johan van Benthem and Al ice ter Meulen, 45-71. Dordrecht: Foris. Wharram, Douglas. 2003. On the Interpretation of (Unjcertain Indefinites in Inuktitut and Related Languages. PhD dissertation, University of Connecticut. Wilkinson, Karina. 1991. Studies in the Semantics of Generic NPis, Ph.D. Dissertation, University of Massachussets, Amherst. Wiltschko, Martina, to appear. Expletive Categorical Features. A Case Study of Number in Halkomelem. Proceedings ofNELS35. Amherst: G L S A , University of Massachusetts.. Wiltschko, Martina. 2002. The Syntax of Pronouns: Evidence from Halkomelem Salish. Natural Language and Linguistic Theory 20, 157-195. Winter, Yoad. 1997. Choice Functions and the Scopal Semantics of Indefinites. Linguistics and Philosophy 20: 399-467. 240 Appendix: Key to the Skwxwu7mesh Orthography Orthography Kuipers orthography IPA P P P P* P' P' m m m m' m? m' t t t t' t' t' ts c ts ts' c' ts' s s s n n n n' . n? n' ch c tj ch' c' t f sh s I lh X i t l ' tr 1 1 l r 1? r k k k k' k ' k ' kw k° k w k w ' k ' ° k ' w xw x° x w k q q k' q' q' kw q° q w k w ' q '° q ' w X X X xw x° x w h h h 241 Orthography Kuipers orthography IPA w u w w' u? w' y i j y ' J" e 3 3 i i i , e, e u u U, 0, 3 a a a 7 ? ? 242