UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Determiner systems and quantificational strategies: evidence from Salish Matthewson, Lisa 1996

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

Item Metadata


831-ubc_1996-147940.pdf [ 18.14MB ]
JSON: 831-1.0087844.json
JSON-LD: 831-1.0087844-ld.json
RDF/XML (Pretty): 831-1.0087844-rdf.xml
RDF/JSON: 831-1.0087844-rdf.json
Turtle: 831-1.0087844-turtle.txt
N-Triples: 831-1.0087844-rdf-ntriples.txt
Original Record: 831-1.0087844-source.json
Full Text

Full Text

DETERMINER SYSTEMS A N D QUANTIFICATIONAL STRATEGIES: EVIDENCE F R O M SALISH by Lisa Matthewson B.A. , Victoria University of Wellington, 1987 B .A. (Hons), Victoria University of Wellington, 1989 M . A . , Victoria University of Wellington, 1991 A THESIS SUBMITTED LN PARTIAL F U L F I L L M E N T OF THE REQUIREMENTS FOR THE D E G R E E OF DOCTOR OF PHILOSOPHY in THE F A C U L T Y OF G R A D U A T E STUDIES (Department of Linguistics) We acceptdiis thesis as confprming to the required standard THE UNIVERSITY OF BRITISH C O L U M B I A April 1996 © Lisa Christine Matthewson, 1996 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shajl not be allowed without my written permission. Department of L) U\O^W\s^r\CS The University of British Columbia Vancouver, Canada DE-6 (2/88) A B S T R A C T This dissertation has three main goals: 1. To provide an analysis of the syntax and semantics of Salish determiners and quantifiers. 2. To provide an account of differences in the determiner and quantification systems of Salish and English which reduces cross-linguistic variation to a minimum, in line with a restrictive theory of Universal Grammar. 3. To assess the theoretical consequences of the analysis of Salish, including implications for the range of possible cross-linguistic variation in determiner and quantification systems, and the nature of the relationship between syntactic structure and interpretation. I give evidence that one common method of expressing quantificational notions in English is absent in Salish. While English readily allows quantifiers to occupy the syntactic position of the determiner (as in every woman, most women), Salish languages do not allow such constructions (see also Jelinek 1995). I propose that Salish and English exemplify opposite settings of a Common Ground Parameter, which states that Salish determiners may not access the common ground of the discourse. This parameter accounts not only for the absence of quantificational determiners in Salish (since quantifiers presuppose existence, and therefore access the common ground), it also derives several other differences between Salish and English determiners, such as the absence of a definiteness distinction in Salish. I further demonstrate that Salish possesses a robust system of DP-internal quantification, and that quantificational DPs in Salish function as generalized quantifiers at logical form. This means that the strong hypothesis that languages do not differ with respect to the presence or absence of generalized quantifiers is upheld (cf. Barwise and Cooper 1981). Simple DPs in Salish, unlike in English, do not function as generalized quantifiers. This result follows from the Common Ground Parameter. I give further evidence from St'at'imcets (Lillooet Salish) on the strong/weak quantifier distinction; I argue that the interpretation of weak quantifiers is derivable directly from the overt syntactic position of the quantifier. i i TABLE OF CONTENTS Abstract i i Table of Contents i i i Acknowledgements vi i Dedication vi i i INTRODUCTION 1 1. Goals of the investigation 1 2. Why study Salish? 3 2.1. Morphology 4 2.1.1. Lexical categories 6 2.2. Syntax 8 2.2.1. Phrasal categories 11 Lexical phrasal categories: NP and V P 11 Functional phrasal categories 12 2.2.2. Overt DPs appear in argument position 14 2.3. The status of the typological split 15 3. Theoretical Assumptions 15 3.1. Determiners and Determiner Phrases 15 3.2. Quantifiers 18 4. Outline of the dissertation 20 5. The Salish family 23 5.1. Subgrouping 23 5.2. A subset of Salish languages 24 C H A P T E R 1 T H E S Y N T A X A N D SEMANTICS OF SALISH DETERMINERS 26 0. Introduction 26 1. Distinctions encoded in Salishan determiner systems 28 1.1. Visibility, proximity, gender and number 28 1.2. Definiteness 32 1.2.1. Defining definiteness 32 1.2.2. Salish determiners do not encode definiteness 33 1.2.3. Evidence against a homophony analysis 38 1.3. Specificity 40 1.3.1. Defining specificity 40 1.3.2. Salish determiners do not encode specificity 43 1.4. Quantificational determiners 46 1.4.1. Defining quantificational determiners 46 1.4.2. There are no quantificational determiners in Salish 48 The absence of strongly quantificational determiners.. 48 The absence of weakly quantificational determiners... 50 Deteminers do not have quantificational force 53 1.4.3. Summary 55 1.5. Salish determiners encode assertion of existence 56 1.5.1. Assertion of existence in Sechelt 61 1.5.2. Assertion of existence in Bella Coola 63 1.5.3. Assertion of existence in Secwepemctsin 67 1.5.4. Assertion of existence in Halkomelem (Chilliwack dialect) 68 1.5.5. The absence of assertion of existence in Straits 69 1.6. Conclusions 71 2. The syntactic distribution of determiners 73 i i i 2.1. Determiners are operators which bind a variable 74 2.2. The syntactic distribution of determiners in Salish 75 2.2.1. The distribution of determiners in argument positions 76 2.2.2. The lack of determiners in (main) predicate position 80 2.2.3. Determiners with temporal nouns in adjunct positions 83 2.2.4. Accounting for the distribution of determiners 85 2.2.5. Salish determiners accord with the variable-binder analysis 88 3. Conclusions 89 C H A P T E R 2 PRESUPPOSITIONAL DETERMINERS A N D T H E C O M M O N GROUND P A R A M E T E R 91 0. Introduction 91 1. Presupposition 92 2. Discourse representation theory and File change semantics 95 3. Presuppositions induced by determiners 98 3.1. Definite determiners presuppose existence 98 3.2. Specific determiners presuppose existence 100 3.3. Quantificational determiners presuppose existence 107 3.3.1. Do weak quantifiers presuppose existence? 110 4. Salish lacks presuppositional determiners 116 4.1. Independent evidence for the lack of presuppositional Ds 118 5. The Common Ground Parameter 121 5.1. Presupposition relies on the common ground 122 5.2. Can Salish access hearer knowledge at all? 125 5.3. The Common Ground Parameter: Subset and superset languages 129 5.4. Predictions of the Common Ground Parameter 130 6. The Salish system within Discourse Representation Theory and File Change Semantics 132 6.1. The lack of definites 133 6.2. The lack of specifics 134 6.3. The lack of quantificational determiners 136 7. Conclusions 139 C H A P T E R 3 ASSERTION OF EXISTENCE 140 0. Introduction 140 1. Assertion of existence is a speaker-oriented notion 141 2. Assertion of existence within Discourse Representation Theory 143 2.1. Main vs. subordinate DRSs 143 2.2. Main vs. subordinate DRSs in English 154 2.3. Modal subordination 156 3. Assertion of existence within File Change Semantics 159 3.1. The problem of specifics 164 4. The logical representation of Salish DPs 165 5. Further empirical support for the Common Ground Parameter 168 5.1. Deictics in Salish are speaker-oriented 169 5.2. Morphological encoding of speaker knowledge 170 5.2.1. Are the speaker-knowledge particles really determiners? 176 6. Ds across the world 179 6.1. Definiteness, specificity, assertion of existence 180 7. Conclusions . 183 C H A P T E R 4 T H E ST'AT'IMCETS DETERMINER SYSTEM.. . . 185 0. Introduction 185 iv 1. Overview of the system 186 2. Assertion of existence determiners 193 3. The non-assertion of existence determiner 195 3.1. Licensing environments for the non-assertion of existence determiner 196 3.1.1. DP Operators do not license ku 196 The restriction of a quantifier appears inside the main DRS 198 3.1.2. Non-factual contexts license ku 201 Modals 202 Intensional verbs 205 Yes-no questions 206 WTi-questions 208 Negation .210 Summary 214 3.1.3. Extension of assertion of existence to clausal complements 215 3.2. Ku inside non-arguments 219 3.2.1. Complex Predicates 220 3.2.2. Relative clauses 223 3.2.3. Quantified temporal adjuncts 224 3.2.4. Demonstrative constructions 226 3.2.5. Adverbial phrases 230 3.2.6. 'Objects' of middles' 231 3.2.7. The two uses of ku 235 4. The internal syntax of DP 236 4.1. Possessors 238 4.1.1. Possessor scrambling 241 4.2. Demonstrative pronouns vs. universal quantifiers 242 5. Conclusions 243 C H A P T E R 5 DP-QUANTIFICATION IN SALISH 244 0. Introduction 244 1. D-quantification vs. A-quantification 246 2. Salish possesses DP-quantifiers 251 2.1. Universal quantifiers adjoin to DP 251 2.2. Not all quantification is adverbial in Salish 261 3. Evidence against a two-way split: missing A-quantifiers 265 4. Revision of the D- vs. A-quantification division 267 5. How do Salish and English differ? 268 6. Conclusions 271 C H A P T E R 6 G E N E R A L I Z E D QUANTIFIERS IN SALISH A N D THE S T R O N G - W E A K DISTINCTION 272 0. Introduction 272 0.1. Generalized quantifiers in Salish 272 0.2. The strong-weak quantifier distinction 273 1. DP-generalized quantifiers in Salish 275 1.1. What are DP-generalized quantifiers? 275 1.2. The NP-Quantifier Universal 276 1.3. Syntactic evidence for generalized quantifiers in Salish 277 1.4. Semantic evidence for generalized quantifiers in St'at'imcets 279 1.4.1. Quantificational DPs in St'aYimcets denote families of sets 279 1.4.2. Quantifiers in St'at'imcets are conservative 281 1.4.3. Quantifiers in St'dt'imcets are monotone 283 1.4.4. Quantificational DPs in St'at'imcets form tripartite structures.... 285 1.4.5. Quantifiers in St'at'imcets take their DP-internal range as the restriction 287 The lack of unselective binding outside the DP 288 Donkey sentences 289 1.5. Conclusions - the NP-Quantifier Universal is upheld 290 1.6. Are all DPs in Salish generalized quantifiers? 291 2. The strong/weak distinction in Salish 293 2.1. Syntactic differences between strong and weak quantifiers in Salish 294 2.1.1. Only weak quantifiers are main predicates 294 2.1.2. Only weak quantifiers appear in existential sentences 295 2.1.3. Only strong quantifiers bind w/i-polarity items 296 2.1.4. Only strong quantifiers allow null ranges 296 2.1.5. Only strong quantifiers can strand their range 297 2.2. Problems with the strong/weak division 297 3. DP-internal weak quantifiers in Salish 301 3.1. The syntax of DP-internal weak quantifiers 301 3.1.1. Weak quantifiers adjoin to NP 302 3.1.2. Weak quantifiers adjoin to DP 304 3.2. The semantics of DP-internal weak quantifiers 310 3.2.1. The semantics of weak quantifiers in English 310 3.2.2. Predictions for weak quantifiers in Salish 316 3.2.3. Readings for St'at'imcets DP-internal weak quantifiers 318 The failure of symmetry 318 Only the proportional reading is possible 320 3.3. DP-internal weak quantifiers set up triprartite structures 324 3.3.1. DP, not NP, forms the restriction of a weak quantifier 327 4. Predicative weak quantifiers in Salish 328 4.1. The cardinal reading is available 328 5. Explanations and questions 330 5.1. Why is the cardinal reading disallowed inside arguments? 330 5.2. Salish evidence on the strong/weak distinction 334 6. Conclusions 337 C H A P T E R 7 QUANTIFICATION LN SALISH: FURTHER IMPLICATIONS 339 1. Gaps in the inventory of quantifiers in Salish 340 1.1. Salish languages do not possess a quantifier with the function of every.... 341 1.2. The absence of generic all 351 1.3. Salish lacks most 353 1.3.1 Test la : Unexpected proportions 354 1.3.2. Test lb : Most is greater than half. 355 1.3.3. Test lc : Exhaustive partitioning 356 1.3.4. Test 2: Pure proportionality 357 1.4. Some does not exist in St'at'imcets 359 1.5. DP-internal no in Salish 362 1.6. The syntax / semantics mapping 364 2. On the nature of partitivity 365 3. Conclusions 369 3.1. How much of the typological split has been explained? 370 Abbreviations 372 REFERENCES 373 vi ACKNOWLEDGEMENTS There is a University regulation which only allows me 1 page to thank everyone, so please believe that I wanted to say more. M y greatest thanks go to the St'at'imcets speakers who have shared their knowledge with me, since I literally could not have done it without them: Beverley Frank, Gertrude Ned, Rose Whitley, Alice Adolph and Laura Thevarge. Spending time with these women is a great pleasure; kukwstum'ckaTap, nsnek'wnuk'wa7! The members of my thesis committee have been tirelessly acute, supportive and demanding. Rose-Marie Dechaine has read many drafts of each chapter, and provided many insightful suggestions, as well as optimistic encouragement at every stage. Any clarity of exposition which this thesis possesses is largely due to her. Dr. Dale Kinkade contributed to my becoming a Salishanist by making my first field methods class a lot of fun, and has been very supportive of my attempts to explain the data. I appreciate his persistence in providing counter-examples to my grandiose claims about Salish. To Hamida Demirdache, thanks for staying interested in my thesis even when I was temporarily fed up with it, keeping an eye on where the large picture was heading, giving constant encouragement that my work was worthwhile, and commenting on many drafts at short notice. Thanks to Michael Rochemont for sharing his knowledge of presupposition and related issues, noticing imprecisions in my discussion, and offering perspicuous and thought-provoking comments on each chapter. Thanks to the following people for discussion of portions of an earlier draft: Irene Heim, Barbara Partee, Ken Hale, and Kai von Fintel. Irene Heim gave extremely detailed and valuable comments on the entire draft. Henry Davis has helped me at every stage of writing this thesis, from the initial elicitation sessions to the final product. He is also the main reason I have stayed relatively happy throughout these last few months. It would have been a lot less bearable without you, niw; wa71hkan xwistumin. Many thanks to my family for support throughout my entire life. Thank you, Claire, for having done a thesis and reminding me that the most difficult chapters sometimes turn out the best. Bucky Ned has provided friendship and endless hospitality to me over the past 4 years. Thanks, nsnuk'wa7. Thanks to the students and teachers of St'at'imc classes, in particular Lemya7 (Neawana John). Thanks to the Lillooet Tribal Council, the Upper St'at'imc Language, Culture & Education Society, the Secwepemc Culture & Education Society, & Simon Fraser University. Thanks to Marianne Ignace for being a driving force behind language programs in the Interior. Thanks to Patricia Shaw, for employing me on a research grant which started my St'aTimcets fieldwork. Thanks to Eloise Jelinek for providing thought-provoking analyses of Salish. Many thanks to my friends Signy Evangeline, Quinn Anderson and Dale Benson for keeping me happy and reminding me that there is life outside the linguistics department. Thanks to Allan Philipson for his friendship through the 1st difficult year in Vancouver. Carmen de Silva is the most important person in the linguistics department; thanks for everything, Carmen. My fellow graduate students have been a necessary part of what sanity I still possess: Ping Jiang-King, Kimary Shahin, Myles Leitch, Yanfeng Qu, Nike Ola, Susan Blake, Taylor Roberts, Takeru Suzuki, In-Que Lee, Wen Li, Carl Alphonce, Bill Turkel, Will Thompson, Vanessa Valerga, Aki Uechi, Eric Rosen & Mandy Jimmie, and in particular to Eleanor Blain, Monica Sanchez, Elizabeth Currie & Helmi Braches. Thanks to my early linguistics teachers Ray Harlow, Laurie Bauer, Liz Pearce, Janet Holmes, Chris Lane & Dorothy McMillan. Thanks to Graham McGregor, my Scottish chum and fellow M & Ms fan, for many joyful hours. I have appreciated being part of the Salishanist community. Thanks to Ewa Czaykowska-Higgins, Leslie Saxon and the crew at the University of Victoria for arranging the Salish morphosyntax workshops. Dwight Gardiner has been a friend as well as a colleague for four years. I have enjoyed working with Peter Jacobs & learning about Squamish. Jan van Eijk's excellent research on St'at'imcets gave me a 20-years' head start; thanks also to Jan for being my most outrageous correspondent. It was a pleasure having Paul Kroeber at U.B.C. for 2 years. Taylor Roberts was a great companion during lots of fieldwork trips. Thanks to David Beck for a long discussion of deixis in Salish. Thanks to Strang Burton for helpful feedback and craziness. Thanks to the 'colloquium crowd' at U.B.C. Thanks to Dorothy Ursaki, the first Salish consultant I ever worked with. Thanks to Nicola Bessell, Jila Ghomeshi, Sharon Peperkamp, Myriam Uribe-Echevarria, Jonathon Bobalijk, Suzanne Urbanczyk, and Uli Sauerland. Thanks to SSHRCC grants #410-92-1629 & #410-95-1519. M y Ph.D. studies were financially supported by a Canadian Commonwealth scholarship. Paptas t'u7 wa7 mawal' ti ucwalmicwtsa - May the language always live. vi i This thesis is dedicated to Dobie. viii INTRODUCTION [I]n order to provide a grammatical analysis of the morphosyntactic marking of (in)definiteness in English, it is necessary to investigate not only English but also Japanese and other languages, thereby constructing a Universal Grammar within which the marking of (in)definiteness in English wil l emerge as one of several consequences of a particular assignment of values to a certain parameter. (Gil 1987:269) 1. Goals of the investigation This dissertation has three main goals, summarized in (l)-(3). 1. To provide an analysis of the syntax and semantics of Salish determiners and quantifiers. 2. To provide an account of differences in the determiner and quantification systems of Salish and English which reduces cross-linguistic variation to a minimum, in line with a restrictive theory of Universal Grammar. 3. To assess the theoretical consequences of the analysis of Salish, including implications for the range of possible cross-linguistic variation in determiner and quantification systems, and the nature of the relationship between syntactic structure and interpretation. The impetus for this investigation arose out of the observation that one common method of expressing quantificational notions in English is absent in Salish. As illustrated in (4), English readily allows quantifiers to appear preceding a common noun phrase. These quantifiers are in complementary distribution with definite or indefinite determiners, and plausibly occupy the syntactic position of the determiner (cf. Jackendoff 1977). 4. a. [Every woman] picked berries. b. [No woman] picked berries. c. [Most women] picked berries. Salish languages do not permit quantificational elements to occupy determiner position (see also Jelinek 1995, Demirdache et al. 1994). Constructions of the form [Quantifier NP] are systematically impossible: 1 5. a. * d^alaw-sm [takem q'welaw'-em [takem pick.berries-inlT [all 'Every woman picked berries.' §m94md+aC] smelhmuihats] woman(redup)] (St'at'imcets) b. * * a;w8law-em [xw?az q'welaw'-em [cw7aoz pick.berries-intr [neg smelhmuihats] woman(redup)] 'No woman picked berries.' (St'at'imcets) Although quantifiers which replace the determiner within the noun phrase are ruled out in Salish, quantifiers which occupy other positions within the noun phrase are possible. Noun phrase-internal quantifiers always co-occur with a determiner, as shown in (6). The semantic literature on quantification provides no obvious way of deriving the contrast between (5) and (6), since syntactic distinctions inside noun phrases are not usually taken to be relevant.1 It is not clear how to rule out quantifiers which occupy the position of the determiner, as in (5), while allowing quantifiers which appear in other positions within the noun phrase, as in In this dissertation I will argue that the ungrammaticality of (5) has its explanation in the nature of Salish determiners themselves, which differ in several respects from English determiners. The absence of quantificational determiners in Salish is just one reflex of a more general difference between Salish and English determiner systems. Before outlining the main proposals of the dissertation, I provide some necessary background information. §2 contains an introduction to the typological split between Salish languages and English. It is this typological split which makes comparison between Salish and English useful 6. a ,w9law-9m [takem q'welaw'-em [takem pick.berries-intr [all ' A l l the women picked berries.' ?1 Sme+mti+ac-a] i smelhmulhats-a] pl.det woman(redup)-det] (St'at'imcets) (6). 1 See Barwise and Cooper (1981), van Benthem (1983), Keenan and Moss (1985), Keenan and Stavi (1986), Partee (1991,1995), Keenan (1996), among others. 2 for the study of cross-linguistic variation, and which can ultimately shed light on the properties of Universal Grammar. In §3, I briefly outline the theoretical assumptions which are made about determiners and quantifiers. §4 contains an overview of the structure of the dissertation and the main claims which w i l l be made, and §5 concludes with famil ial information about Salish and the subset of Salish languages investigated here. 2. Why study Salish? Sal ish languages differ from Engl ish in many respects. 2 For example, Sal ish languages are morphological ly r ich, with obligatory pronominal agreement markers appearing on main predicates. Nu l l arguments are common. Overt D P arguments display relatively free word order. Morphological tense marking is usually absent. Any open-class lexical item may function as a main predicate, and it has been claimed that Salish lacks a distinction between nouns and verbs (see e.g. Kinkade 1983, Jelinek 1993c, 1995). The deep typological split between Salish and English provides fertile ground for the study of Universal Grammar. We must first ascertain the exact ways in which Salish differs from English, and then attempt to reduce the differences to a small number of learnable parameter settings. English is used as a comparison language for the simple reason that a large amount of theoretical work has concentrated on English. Ultimately, of course, analyses proposed should account for al l natural languages. Theoretical work on Salish has already provided some intriguing proposals about the source of cross-linguistic variation. One general approach, represented recently by the work of Jelinek (1984, 1993c, 1995), views Salish as differing at a fundamental level f rom Engl ish-type 2 A l l the properties of Salish listed here wi l l be exemplified later in this section. 3 languages. 3 According to Jelinek's Pronominal Argument Hypothesis, the Salish lexicon does not differentiate nouns from verbs, but rather contains one open-class category of inflected predicates, with pronominal agreement affixes already attached. These pronominal agreement morphemes occupy argument positions. A s a consequence, overt DPs are optional, and when they do appear are necessarily adjoined to the clause (cf. also Baker 1991, 1996). The theory of Salish developed in this dissertation differs in many respects from Jelinek's analysis. Nevertheless, the aim is to follow Jelinek's example in at least, two respects. Jelinek's work serves firstly to bring a group of under-studied and interesting languages to the attention of the theoretical literature. Secondly, and more importantly, it consistently forces re-examination of theoretical assumptions based on English or Indo-European. Research of this nature highlights the importance of the questions raised by Salish for our understanding of Universal Grammar. 4 In the rest of this section, morphological and syntactic characteristics which illustrate the typological split between Salish and English are outlined. Familiarity with these basic features of Salish w i l l facilitate understanding of the proposals made in later chapters. 2.1. Morphology Salish languages are highly polysynthetic, and can be classified as radical head-marking languages. Head-marking languages are those in which syntactic relationships are indicated primarily by agreement morphology on the head of a construction (such as the verb, rather than its arguments; see Nichols 1986, Baker 1996).5 Pronominal agreement (in the form of affixes or clitics) appears on predicates in Salish, and possessive marking appears on the head noun. 6 3 The view that Amerindian languages differ fundamentally from English can also be found in early works such as Swadesh (1936), Whorf (1950), Bloomfield (1933). 4 See also G i l (1987) on the imporance of cross-linguistic research for theoretical linguistics. 5 Head-marking languages contrast with dependent-marking languages, which utilize Case rather than agreement morphology on the verb. 6 Examples taken from written works use the orthography of the original work, unless otherwise stated. 4 7. yawat-tsi-chen skwa wake-2sg.obj-lsg.subj future I ' l l wake you.' (Sechelt; Beaumont 1985:86) te snexwilh-s te stumish det canoe-his det man 'the man's canoe' (Sechelt; Beaumont 1985:28) Salish languages display split-ergativity in their pronominal morphology, with splits along person lines (1st and 2nd vs. 3rd person), main clause/subjunctive clause lines and/or aspectual lines. (8) illustrates the person split in Squamish. First person arguments distinguish nominative (8a,d) from accusative (8c). Third person arguments distinguish ergative (8c) from absolutive (8b,d)-7 8. C-n lsg.nomin 'I am a man.' sw1?qa man (Squamish; Kuipers 1967:89)8 sw1?qa-0 man-3abs 'He is a man. / They are men.' (Squamish; Kuipers 1967:89) na rjaw-at-umu+-as rl help-tr-lsg.acc-3erg 'S/he helped me.' (Squamish; Kuipers 1967:90) C-n £aw-at-0 lsg.nomin help-tr-3abs 'I helped him/her.' (Squamish; Kuipers 1967:90) Third person absolutive is marked zero in every Salish language except for Bella Coola (Nater 1984), and Upper Chehalis and Cowlitz in the imperfective aspect (Kinkade 1991, p.c). The absence of overt absolutive marking contributes to the debate about whether nouns and intransitive verbs belong to different categories in Salish. Any lexical item, including one with 7 See Newman (1977), (1979a,b), (1980) for discussion of pan-Salish pronominal systems and their historical development. 8 Kuipers' (1967) orthography has been translated into a phonemic script. 5 nominal semantics, can be hypothesized to contain a null 3rd person absolutive agreement morpheme, and thus to constitute a complete clause. See §2.1.1 for further discussion. The Salish word is internally very complex. In addition to (in)transitivizing morphemes and pronominal agreement morphology, the word contains aspectual affixes, derivational affixes, and lexical suffixes.9 The structure of the word in one Salish language, St'at'imcets, is given in (9) (Davis to appear; see also van Eijk 1985). 9. [[procl[nom[[[[[loc[nom/sta[ROOT] asp] lex] in/trans ] abst] obj] erg]] subj] encl] 1 2 3 4 Four word-internal domains can be distinguished, based on evidence from both prosodic and morphological criteria. The innermost, (1), contains the root, the only element which is obligatory in all predicates. The stem-level domain, (2), contains a variety of aspectual and other affixes, including transivitizers and intransitivizers, but excluding pronominal affixes. The latter occupy (3), the outermost affixal domain, which is equivalent to the level of the morphological word. Domain (4), which contains various pro- and en-clitics, is the maximal domain of word-level stress assignment and corresponds to the prosodic word (Davis to appear). The details of this analysis of the St'at'imcets word do not extend to all Salish languages, although all the languages display similar complexity at this level of structure.10 2.1.1. Lexical categories Superficial syntactic evidence does not distinguish between lexical categories in Salish, but only between main predicates on the one hand, and closed class items (such as deictics) on the other. 9 Lexical suffixes are affixes with lexical content (often somatic, as in (i)). See Hinkson (in prep) and references cited therein on lexical suffixes in Salish. i . ?u-bakw4-ac'l? Ced perf-hurt-hand lsg.subj T hurt my hand.' (Lushootseed; Bates et al 1994:32) 1 0 The structure in (9) cannot be extended to Upper Chehalis, if Kinkade (1967) is correct in claiming that the forms in (i) lack roots altogether, being composed of prefix-suffix combinations. i . a. ?ac-a"w* b. ?ac-nowt stat-canoe stat-mind 'be in a canoe' 'thought, mind, something in the mind' (Upper Chehalis; Kinkade 1967:1) See Mattina (1996) for an analysis of word structure in Okanagan (Southern Interior) which differs from that given in (9). 6 A l l open-class elements (those which correspond to nouns, verbs and adjectives in English) can function as the main predicate of a sentence, as shown in (10). 10. qwaC^C-kaxw qwatsdts-kacw leave-2sg.subj 'You left/you leave.' (St'dt'imcets) xztjmH-kaxw xzum-lhkacw big-2sg.subj 'You are big.' (St'at'imcets) §mtj+ae-kaxw smulhats-kacw woman-2sg.subj 'You are a woman.' (St'at'imcets) Projections of any open-class lexical item can combine with a determiner to form a Determiner Phrase, as iluustrated in (11). 11. a. caqw-anH-kan h i ts'aqw-an'-lhkan [ni eat-tr-lsg.subi [det 'I ate the one I caught.' k waYi-an-a] kwan-an-a] catch(tr)-1 sg.conj-det] (St'at'imcets) caqw-anH-kan [ni xztim-a] ts'aqw-an'-lhkan [ni xzum-a] eat-dir-lsg.subj [det big-det] 'I ate the big one.' (St'at'imcets) ca*qw-anH-kan [ni ts'aqw-an'-lhkan [ni eat-dir-lsg.subj [det 'I ate the fish.' cuqwaz'-a] ts'uqwaz'-a] fish-det] (St'at'imcets) The DP in (11c) contains no overt inflectional morphology. However, the null status of 3rd person absolutive marking makes it possible to argue that (11c) contains a null-headed relative clause ('the one who is a fish'). Precisely this claim is made by those who deny the existence of a noun/verb distinction in Salish (e.g. Kinkade 1983, Jelinek 1993c, 1995).11 1 1 See Dernirdache and Matthewson (1995a), Matthewson and Davis (1995) for arguments that (lla,b) contain null-headed relative clauses, while (11c) contains a [D NP] constituent. 7 The literature generated by the Salish categorial debate is very large. Kinkade (1983), Jelinek and Demers (1982) and Jelinek (1993b,c, 1995) propose that there are no distinctions between lexical categories in Salish; see also Kuipers (1968), Bloomfield (1933), and Thompson and Thompson (1980). On the other hand, categorial distinctions have been argued for by Davis and Saunders (1974), van Eijk and Hess (1986), Mattina (1994), Beck (1995a,b), Davis and Matthewson (1995), Demirdache and Matthewson (1995a), Kroeber (1991), Davis et al. (in prep), among others (see also Birch 1993).12 Many grammars of Salish languages assume categorial distinctions; see for example Nater (1984). The analysis of Salish determiner and quantification systems to be presented in this dissertation will not rely crucially on any assumptions about lexical categories, and I leave the interested reader to pursue the references cited if desired. 2.2. S y n t a x A Salish clause obligatorily contains a predicate; overt DP arguments are optional. 12. a. s-xlp-w-n contin-dry-contin.intr-contin.3rd.subj 'He's drying it' (Upper Chehalis; Kinkade 1983:27) b. cey-0 work-3abs 'He/she/they worked.' (Lummi; Jelinek 1993a:4) Clauses are typically predicate-initial; languages differ in the extent to which they allow arguments of the main predicate to occur before the predicate. Various clefting strategies are available which give argument-predicate order, as shown for example in (13c). 13. a. nahuya7 [ta slhanay'] rl leave [det woman] 'The woman left.' (Squamish; Demirdache et al. 1994) 1 2 See also Bach (1992) on the categorial issue. 8 b. [ta slhanay'] na huya7 [det woman] rl leave 'The woman, she left.' (Squamish; Demirdache et al. 1994) nilh [ta swi7ka] na foe [det man] rl 'It's the man that is sleeping.' wa i7tut prog sleeping (Squamish; Demirdache et al. 1994) Post-predicate word order is relatively free, with some languages favouring V S O as a basic order (e.g. Halkomelem; Hukari 1996) and some favouring VOS (e.g. St'at'imcets; Gardiner et al. 1993, van Eijk 1995). Clauses containing two overt DP arguments are rare in discourse or narration, since subjects are usually represented by null pronouns. Some Salish languages even disallow two overt arguments: The type of English transitive sentence in which both agent and patient are indicated by noun phrases (e.g. Bill killed the bear) seems atypical of at least many Salish languages, and is actually impossible in Lushootseed, where only the patient can be so specified (Thompson 1979:740). W7i-questions in Salish contain a clause-initial w/i-word, followed by a subordinate clause. Wh-questions have been analyzed as cleft-like constructions by Kroeber (1991), Davis et al. (1993), and Gerdts (1988). 14. st£? e [s-cex/-xft-xw] what det [nom-clean-ditr-2sg.subj] 'What did you clean for him?' (N+e?kepmxcfn; Kroeber 1991:229) b. wit kw9 who det 'Who walked?' [ni ?fme§] [aux walk] (Halkomelem; Gerdts 1988:67) Wh-v/ords may appear in non-sentence-initial position, in which case they function as polarity items, which must fall within the scope of either a strong quantifier or negation (cf. Cheng 1991, Nishigauchi 1986,1990). 15. wl q^ x,+ stfx w n-sawm§ Xw^qwu teim and many kill-3pl.poss all what 'and much is their kil l of all kinds/of everything.' (Upper Chehalis; M . D . Kinkade, p.c.) 9 b. E"Wl? ?al ^ w a - n a ? s-t lr j s-Hen-s act+hungry limit not-exist s-what s-eat-3poss 'They were hungry but they didn't have any food.' (Saanich; Montler 1986:242) Syntactic extraction is often marked by special morphology; see Kroeber (1991), Hukari (1993, 1995), Gerdts (1988), Davis et al. (1993), among others. Special extraction morphology is often useful for ascertaining that a certain string has undergone movement, and thus is acting as a syntactic constituent (see Chapter 5). Extraction morphology is illustrated in (16). In St'at'imcets, conjunctive morphology replaces ordinary transitive subject morphology in relative clauses, focus constructions and w/i-questions. Compare the subject morphology in (16a), where no movement has taken place, with that in (16b-d), where movement (of either an overt DP or a null operator) has taken place. 16. ?a"cx-9nH-kaxw Ct 1 ats'x-en-Ihkacw [ti see-tr-2sg.subj [det 'You saw the chief.' kwdkwp1?-al kukwpi7-a] chief-det] (in situ) (St'at'imcets) wa? wa7 prog latl? [tl lati7 [ti deic [det 'There's the chief you saw.' acx-an-aV-a ats'x-en-acw-a see-tr-2sg.conj-det kwdkwp1?] kukwpi7] chief] (relative clause) (St'at'imcets) kwdkwp1?-a] ntt [tl nilh [ti kukwpi7-a] foe [det chief-det] Tt was the chief you saw.' ?a'cx-9n-axw ats'x-en-acw see-tr-2sg.conj (focus) (St'at'imcets) d. Swat k w u ?^cx-an-axw swat ku dts'x-en-acw who det see-tr-2sg.conj 'Who did you see?' (wh-question) (St'at'imcets) 10 2.2.1. Phrasal categories Lexical phrasal categories: NP and VP The category-neutral view of Salish, introduced briefly above and instantiated by Jelinek (1995), holds that there is no distinction between NP and V P , and hence that it wi l l be impossible for syntactic combinations to show restrictions according to category. However, headed relative clauses in St'at'imcets provide evidence that a syntactic category of NP must be distinguished in that language, as argued by Demirdache and Matthewson (1995a), Matthewson and Davis (1995), and Matthewson and Demirdache (1995). NP is the only syntactic category which can project to the head of a relative clause. The headed relative clause in (17) contains two identical determiners (i.e. two instances of the discontinuous determiner ti.. .a). 17. wa? la'tl? [tl smd+aC-a tl ^acx-an-aVa] wa7 l£ti7 [ti smulhats-a ti ats'x-en-an-a] aux deic [det woman-det det see-tr-lsg.conj-det] 'There's the woman I saw.' (St'at'imcets) The only possible configuration for a two-determiner relative contains an NP, followed by a clause (designated as 'S' in (18)). A clause followed by an NP is impossible (18b), as is either two NPs (18c) or two clauses (18d). Matthewson and Davis (1995) identify this construction as a head-initial relative clause. 18. a. det NP det S b. * det S det NP c. * det NP det NP d. * det S det S Examples of the possible and impossible combinations are given in (19). 11 19. a. det NP det S: puViH-kan [tl cqax'-a tl xwdlal-a] pun-lhkan [ti ts'qax7-a ti culel-a] find(tr)-lsg.subj [det horse-det det run.away-det] 'I found the horse which ran away.' (St'at'imcets) b. * det S det NP: * ptin-rkan [tl xwtil9l-a tl rjqaV-a] * pun-lhkan [ti culel-a ti ts'qax7-a] find(tr)-lsg.subj [det run.away-det det horse-det] 'I found the one which ran away which was a horse.' (St'at'imcets) c. * det NP det NP: * zwa"t-an-4kan [tl ?xwalmfxw-a tl naplft-a] * zwat-en-lhkan [ti ucwalmfcw-a ti naplit-a] know-tr-lsg.subj [det Indian-det det priest-det] 'I know an Indian who is a priest.' (St'at'imcets) d. * det S det S: * pdn-4kan [tl xwtil9l-a tl xwll-en-an-a] * pun-lhkan [ti culel-a ti cwil'-en-an-a] find(tr)-lsg.subj [det run.away-det det look.for-tr-lsg.conj-det] 'I found the one who ran away who I was looking for.' (St'at'imcets) A category-neutral analysis is unable to account for the facts in (19), incorrectly overgenerating the ungrammatical combinations in (19b-d). - Functional phrasal categories It is not clear at this stage of research which functional categories are distinguished in Salish languages. For example, it is not clear whether one should posit a Tense Phrase, as is often assumed for English and other languages: 12 20. TP (Spec) T V (D6chaine 1993:42) Tense is not an obligatory morphological category in Salish (M. Dale Kinkade, p.c , Thompson and Kinkade 1990:33; see also Demirdache 1996a,b).13 Clauses which lack morphological tense marking are shown in (21). 21. a. ?f ial [tl Sqayxw-a] flal [ti sqaycw-a] cry [det man-det] 'The man cried.' / 'The man is crying.' b. nwal-p-efn [+a n - t A q - c l n - t n l open-inch-door [det loc-cover-door-instr] 'The door opened.' / ' The door is opening.' (St'at'imcets; Demirdache 1996b) (N+e?kepmxcfn; DU) Temporal notions are often encoded on deictic elements or on determiners in Salish, and there is some neutralization of the distinction between space and time (Davis and Saunders 1975, Demirdache 1996a,b). It is not clear whether the structure in (20), which contains a clausal-level Tense Phrase, accurately represents Salish. On the other hand, Dechaine (1993:17) claims that a Tense Phrase is universally present in the syntax, even if it is not morphologically filled. Further research is required into these issues. One functional category which I crucially assume to exist in Salish is the Determiner Phrase. See §3.1 below. 1 3 Aspect marking is much more pervasive than tense marking in Salish (M. Dale Kinkade, p.c, Thompson and Kinkade 1990). 13 2.2.2. Overt DPs appear in argument position One Salish language, namely Straits, has been analyzed as a 'pronominal argument' language, in which all overt DPs are adjoined to the clause (as in Jelinek 1984, 1995, Jelinek and Demers 1994; see also Baker 1991, 1996). However, detailed syntactic investigation reveals that overt DPs occupy argument positions in at least some Salish languages (Matthewson 1993, Matthewson et al. 1993, Davis 1993, 1994b). One type of evidence for configurational structure comes from subject-object asymmetries in determiner distribution. I wil l argue in Chapter 4 that the determiner ku in St'&t'imcets requires a c-commanding licenser, usually an intensional operator. (22) shows that ku is licensed on objects, but not subjects, of intensional verbs. 22. a. xaVmlnH-kan tkwu £tiqwaz'] xat'-min'-lhkan [ku ts'uqwaz'] hard-appl-lsg.subj [det fish] 'I want some fish.' (transitive object) (St'at'imcets; RW,GN) b. * xafc-mln-aS tkwu Smanxl [kwu §a"ma?] * xdt'-min'-as [ku sman'c] [ku sama7] hard-appl-3erg [det tobacco] [det white.person] 'A white guy wants some tobacco.' (transitive subject) (St'at'imcets; RW) Under a pronominal argument analysis of the language, no such subject-object asymmetries are predicted to exist. The asymmetries accord with a configurational structure as in (23), where the subject occupies a higher position than the object.14 23. V P V DP-subject fJP-object 1 4 The structure in (23) violates Kayne's (1995) proposals that Specifiers always appear on the left. I am not concerned here with details of the clausal structure, but only with the necessity for a hierarchical structure of some kind. 14 Further evidence for configurational structure in at least some Salish languages will be given in Chapters 4 and 5. 2.3. The status of the typological split Many of the properties of Salish languages are reminiscent of so-called 'non-configurational' languages (see Hale 1983, Speas 1990, Baker 1996 and references cited therein for discussion). Richness of agreement morphology, the presence of null arguments, and relative freedom of word order have all been linked with a non-configurational clause structure. On the other hand, I have suggested above that at least some Salish languages have a hierarchical clause structure, very similar to that of English. M y view on the Salish-English split, as will become clear in the chapters which follow, is that the manifold differences between the languages cannot all be reduced to a single 'macro-parameter' (such as that proposed by Baker 1996). The term 'typological split', therefore, is used as a descriptive term, and indicates only that Salish and English differ from each other in a number of respects. The correct analysis of the differences is the topic not only of this dissertation, but of much other current and future research. The major focus of this dissertation is the Determiner Phrase (DP). I assume the basic X-bar structure for DP in (24), following Abney (1987). 3. Theoretical Assumptions 3.1. Determiners and Determiner Phrases 24. DP Specifier D' D Complement 15 According to the DP analysis in (24), the determiner is the head of the DP and takes NP as its complement. This contrasts with earlier versions of the internal structure of noun phrases (see e.g. Jackendoff 1977, Chomsky 1981), where the noun is the head of the phrase, and the determiner occupies the Specifier position: 25. NP Specifier N ' I I D N According to the DP analysis, D is a head (X° category), rather than a phrase (XP category). D is moreover a functional head, which selects a lexical projection as its complement. The lexical/functional split is summarized in (26): 26. If X ° e {V, N , P, A} , then X° is a Lexical head (open-class element). If X ° e {Tense, Det, Comp, Kase}, then X° is a Functional head (closed-class element). (Dechaine 1993:2) The complement of D is usually assumed to be NP. Another way to say this is that the extended projection of N (i.e. the potential set of functional heads which dominate N) includes D (D6chaine 1993; see Grimshaw 1991, Davis 1987). A major motivation for the DP-analysis of noun phrases comes from the many parallels between clauses and noun phrases. For example, Abney (1987) notes that many languages contain agreement within noun phrases which parallels agreement at the clausal level. In Yup'ik, subjects of noun phrases take ergative case, just like subjects of transitive verbs: 27. a. angute-m kiputa-a-0 man-erg buy-obj-subj 'The man bought it.' (Yup'ik; Abney 1987:39) b. angute-m kuga-0 man-erg river- subj 'The man's river.' (Yup'ik; Abney 1987:39) 16 The presence of noun phrase-internal agreement morphology suggests the presence of an inflectional functional element within the noun phrase, which parallels inflectional functional elements within clauses. Abney proposes that the noun phrase-internal functional element is the determiner itself. Nominal gerunds in English also display many clause-like properties. Gerunds contain VPs, and a gerundive verb can assign Case, unlike a deverbal noun: 28. a. John's [destroying the spaceship]vp b. * John's [destruction the spaceship] (Abney 1987:16) The presence of a V P inside the noun phrase does not accord with the structure in (25), according to which N is the head of the phrase. It can be accomodated within the DP-analysis, however, as shown in (29). 29. DP John's D Gen. Casej D' -ing NP V P DP destroy the spaceship (cf. Abney 1987:223) Since Abney (1987), many authors have adopted the DP-analysis and/or argued for extensions of it; see for example Tellier (1991), Szabolsci (1983, 1987), Ritter (1991, 1993), Valois (1991), Longobardi (1994). Some of the literature cited in this dissertation uses the terms 'NP' or 'noun phrase' to refer to phrases which according to the DP analysis are Determiner Phrases. In many cases, this is because the works involved were written before the DP analysis was proposed. I wi l l consistently use the term 'DP' throughout the discussion, where it is clear that this does not 17 change the intent of the original work. I will reserve the designation 'NP' for the lexical phrasal category which appears as the complement to a determiner. . 3.2. Quantifiers Chierchia and McConnell-Ginet (1990:91) give the following definition of quantification (emphasis original). quantificational expressions ... introduce the power to express generalizations into language, that is, the power to move beyond talk about properties of named individuals to saying what quantity of the individuals in a given domain have a given property. Natural language quantifiers include the familiar English examples in (30). 30. English quantifiers: all, every, each, some, most, many, few, a few, both, half, three, no, neither,... The quantifiers in (30) all range over (sets of) individuals, and may all appear DP-internally. In (31), the range of each quantifier is provided by an NP. 31. a. [Every girl] forgot her pencil. b. [Most men] shave their legs. c. [Three hockey players] crashed my party. With regard to the interpretation of quantifiers, Barwise and Cooper (1981:163-164) claim that 'Quantifiers are used to assert that a set has some property ... quantifiers are taken to denote the family of sets for which they yield the value "true".' According to May (1985), a quantifier Q is interpreted as a function from subsets of the domain onto {1,0} (truth and falsity). Examples of the functions represented by quantifiers are given in (32), with prose translations in (33). 32. a. N o ( X , Y ) = l i f f X n Y = 0 = 0 otherwise b. Every (X,Y) = l i f f X = X n Y = 0 otherwise (May 1985:8) 18 33. a. No (X,Y) is true if and only if the intersection of sets X and Y is empty. b. Every (X,Y) is true if and only if the set X is equivalent to the intersection of sets X and Y . Applying the definitions in (32), the sentences in (34a,b) have the informal truth conditions given. 34. a. No girl forgot her pencil. = true iff the set of individuals who are girls and who forgot their pencil is empty. b. Every girl forgot her pencil. = true iff the set of individuals who are girls is identical to the set of individuals who are girls and who forgot their pencils. For reasons which will become clear in Chapters 5 and 6,1 adopt the assumption that quantified DPs appear adjoined to the clause at the level of Logical Form (LF) (see Huang 1982, May 1985 and many others). (35) represents the LF of the sentence Sophie broke every pencil. 35. IP DPi / \ every pencil DP IP Sophie I V P broke The structure in (35) does not take into account a large amount of recent work on the internal structure of the clause (see for example Pollock 1989, Chomsky 1991, among others). For the most part, the analysis to be presented below is independent of particular instantiations of clause structure, such as the number of functional projections required. Where a particular theoretical assumption is necessary to the analysis, it is explicitly pointed out at the relevant moment. 19 Quantifiers may also range over events (or cases, times, or situations), as shown in (36) and (37). 36. English quantifiers: always, sometimes, often, usually, seldom, three times, never,... 37. a. Sophie seldom forgets her pencil, b. M y husband usually bikes to work. Since the focus of this dissertation is the Salish Determiner Phrase, I wil l be most interested in the ways in which quantificational notions are expressed DP-internally. I w i l l not be concentrating on the adverbial quantifiers in (36-37). 4. Outline of the dissertation As indicated above, the absence of quantificational determiners in Salish will be analyzed as one reflex of a more general difference between determiners in Salish and in English. To show that this is so, I present a detailed examination of the syntax and semantics of Salish determiners in Chapter 1.1 propose that Salish determiners display the properties in (38) (cf. also Matthewson to appear). 38. a. Salish determiners do not encode definiteness. b. Salish determiners do not encode specificity. c. There are no quantificational determiners in Salish (see also Jelinek 1995). d. Salish determiners encode 'assertion of existence'. The claims in (38) are more than superficial statements about morphological encoding. For example, with regard to definiteness, I show not only that determiners do not overtly encode a definiteness distinction in Salish, but also that determiners cannot be analyzed as homophonous between definites and indefinites.15 Chapter 2 aims to account for the properties in (38a-c). I propose a parametric account of the Salish-English split, whereby the ability to encode distinctions which rely on presuppositional 1 5 The suggestion that definite and indefinite determiners in Salish could be simply homophonous with each other was made by Robert May (p.c). See Chapters 1 and 2 and Demirdache (1996a,b) for arguments that such a proposal cannot account for the Salish facts. 20 notions is absent in Salish determiners. I argue that all the distinctions which Salish determiners lack involve presuppositions of existence. Since presupposition crucially relies on information contained in the common ground of the discourse (see Stalnaker 1974, Heim 1982, among many others), we can rule out all the impossible determiner types by means of a single Common Ground Parameter, given in (39). 39. Common Ground Parameter Determiners may access the common ground of the discourse Yes: {English,... } No: {Salish,...} The parameter in (39) derives the absence of a definiteness distinction in Salish, the absence of a specificity distinction, and the absence of quantificational determiners. The derivation of the last result relies on the claim that all quantificational determiners, both strong and weak, induce presuppositions of existence (pace Diesing 1992). The parametric approach has several desirable features, as outlined in (40). 40. a. A wide range of characteristics are explained by means of a single locus of variation. b. Aparently major differences in quantificational systems are accounted for without having to postulate differences at the level of the semantics. c. The parameter is stateable at the level of the lexicon, and its setting relies on learnable cues. With regard to (40b), notice that we do not have to postulate that certain semantic operations are missing from Salish languages. Presupposition is not absent from the entire grammar of Salish, but is merely unavailable for a well-defined class of lexical items: determiners. Similarly, quantification which ranges over individuals exists in Salish, just as in English, and no semantic variation is required. Salish languages differ from English only in the particular subset of (universally available) determiner distinctions which is selected. 21 Chapter 3 investigates the assertion of existence distinction (38d). I show that the assertion of existence distinction cannot be accounted for by currently available theories. I then propose a modification of Discourse Representation Theory (Kamp 1981, Heim 1982) which enables us to account for Salish. I also argue that the presence of an assertion of existence distinction in Salish determiner systems is compatible with the parameter in (39). Chapter 4 contains a detailed examination of the determiner system of one language, St'at'imcets. The analysis of the assertion of existence distinction is fine-tuned, and evidence is provided which suggests that a hierarchical clausal structure is necessary in St'at'imcets, with an overt subject generated higher than an overt object (pace what has been claimed for Straits Salish by Jelinek 1993c, 1995). The last three chapters investigate Salish quantification. In Chapter 5,1 provide syntactic and semantic evidence that Salish possesses a robust system of DP-internal quantification. I argue that quantifiers which syntactically appear inside DP do not behave like adverbial quantifiers, and that we thus have true D-quantification in Salish (pace Partee 1990, Jelinek 1995). The analysis suggests redefinition of the distinction between D-quantification and A-quantification, in favour of a three-way split between A-quantifiers, DP-quantifiers and D°-quantifiers. Salish possesses DP-quantifiers (quantifiers which appear inside DP), but lacks D°-quantifiers (quantifiers which occupy the position of the D° head), for reasons already discussed above. Chapter 6 argues that Salish possesses DPs which correspond to generalized quantifiers (Barwise and Cooper 1981; cf. opposing claims made in Jelinek 1995). The evidence provided is both syntactic and semantic. Semantically, DPs containing quantifiers in Salish demonstrate the same properties as English generalized quantifiers. As expected, they obey Barwise and Cooper's universal claims on the nature of generalized quantifiers. For example, generalized quantifiers in Salish obey conservativity and monotonicity. 22 Chapter 6 also addresses the nature of the strong-weak quantifier distinction (Milsark 1974). I show that weak quantifiers in St'at'imcets do not display the ambiguity commonly assumed for English weak quantifiers. Rather, the interpretation of St'at'imcets weak quantifiers is straightforwardly correlated with syntactic position. Inside DP, weak quantifiers have only a proportional reading, while predicative weak quantifiers allow a cardinal reading. This can be accounted for by means of an isomorphism constraint on the relationship between the overt syntax and the logical representation. Chapter 7 discusses outstanding differences between quantification in Salish and English. Although the differences can largely be derived from the independently-required parameter on determiners (see (39)), certain auxiliary assumptions are required regarding the mapping between syntax and semantics. In particular, I speculate that there is a necessary relationship between the syntactic position of the determiner (D°) and the introduction of a resource domain variable (i.e. the implicit limitation of the range of a quantifier; von Fintel 1994). I suggest that only quantifiers which occupy the D° position may introduce a resource domain variable. Since D°-quantifiers do not exist in Salish, the range of a quantifier must always be explicitly rather than implicitly limited. In turn, this correctly predicts the absence of a generic universal quantifier in Salish languages. 5. The Salish family 5.1. Subgrouping A complete list of Salish languages is given in (41) (cf. van Eijk 1987:ix-x, Thompson and Kinkade 1990:34-35, M . Dale Kinkade, p.c). Some of the finer dialect divisions have been omitted, where the distinctions are not relevant for the material presented in this dissertation. Languages marked with * are extinct. 23 41. The Salish language family Branch | Language | Dialects Bella Coola Central Salish Comox Sliammon, Klahoose, Homalko, Island Comox Pentlatch * Sechelt Squamish Halkomelem Chilliwack/Upriver Halkomelem, Musqueam, Nanaimo/Cowichan Nooksack * Northern Straits Semiahmoo, Saanich, Lummi , Songish, Samish, Sooke Klallam Lushootseed | Northern, Southern Twana * Tillamook * Tsamosan Upper Chehalis | Satsop, Oakville, Tenino Cowlitz * Lower Chehalis Quinault * Interior Northern Lillooet (St'at'imcets) Mount Currie/Lower Lillooet, Fountain/Upper Lillooet Thompson (N+e?kepmxcfn) Shuswap (Secwepemctsfh) Eastern, Western Southern Okanagan Northern, Southern/Colville Columbian Kalispel Spokane, Kalispel, Flathead (Sel1§) Coeur d'Alene 5.2. A subset of Salish languages The language from which the largest body of data is drawn for this study is St'at'imcets (Northern Interior). Most of the St'at'imcets data comes from my original fieldwork; initials following an utterance identify the speaker concerned. For much of the discussion, particularly those parts where semantic judgements are required, St'at'imcets is the only language from which data can be drawn. Native speakers can give judgements on semantic ambiguities, as well as negative data; such information is unobtainable from textual study and largely unavailable in current published works. 24 There are two main dialects of St'at'imcets, and data are drawn from both dialects. Readers may notice lexical differences between examples provided by different speakers; however, any syntactic or semantic differences between speakers are explicitly noted in the text. A l l St'at'imcets examples are given both in a phonemic script, and in the practical orthography of the language (devised by Jan van Eijk; see van Eijk and Williams 1981). This is to faciliate access by speakers of the language, who are often familiar only with the practical orthography. Wherever possible, claims made on the basis of St'at'imcets have been checked in other languages. Fieldwork data is cited for Secwepemctsin (provided by Dwight Gardiner), N+e?kepmxcfn (collected by the author), Upper Chehalis (provided by M . Dale Kinkade), and Squamish (provided by Peter Jacobs). The complete list of languages discussed in this dissertation is given in (42). 42. Languages investigated Language Branch Bella Coola — Sechelt Central Squamish Central Halkomelem Central Northern Straits Central Lushootseed Central Upper Chehalis Tsamosan St'at'imcets Northern Interior N+e?kepmxcfn Northern Interior Secwepemctsin Northern Interior The choice of the languages in (42) is motivated exclusively by availability of materials. Languages omitted from (42) do not to my knowledge reflect systematic counter-examples to the claims made below. On the basis of the subset of ten languages in (42), I generalize to the entire family, often making claims about Salish as a whole. Obviously, such claims are falsifiable by evidence which might turn up from other languages. It should also be noted that there is variation within the family on a number of points; I shall discuss these as they arise. 25 C H A P T E R 1 T H E S Y N T A X A N D S E M A N T I C S O F S A L I S H D E T E R M I N E R S 0. Introduction A satisfying theory of the semantics of determiners should account for the range of distinctions encoded by natural language determiners. Some examples of attested distinctions are given in (1). 1. a. b. c. d. e. f. definiteness specificity visibility proximity gender number Case English,. . . Turkish (En? 1991), Polynesian (Chung 1978), Bella Coola (Davis and Saunders 1975),... St'at'imcets (van Eijk 1985),... German,... German,... German,... Our theory should also account for cross-linguistic variation in the distinctions which determiners encode. For example, English determiners encode definiteness, but not specificity; Polynesian languages encode specificity, but not definiteness, according to Chung (1978). Although determiners vary cross-linguistically, it would not be an appealing hypothesis to say that the semantics of detenniners varies randomly from language to language. In order to obtain a restricted theory of Universal Grammar, our null hypothesis should be that the semantics of determiners is universally provided and universally invariant.1 In this chapter, I wi l l argue based on Salish languages that the null hypothesis is untenable. Cross-linguistic variation in determiner systems is more than superficial; the range of semantic distinctions available for detenniners must be parameterized. 1 For attempts to make universally valid hypotheses about determiner semantics, see Barwise and Cooper (1981), Keenan and Moss (1982), van Benthem (1983), Zwarts (1983), Keenan and Stavi (1986), among others. 26 The first goal of the chapter is to establish which distinctions are encoded by Salish determiners. In §1,1 argue for four major proposals about Salish, summarized in (2). 2. a. Salish determiners do not encode definiteness. b. Salish determiners do not encode specificity. c. There are no quantificational determiners in Salish (see also Jelinek 1995). d. Salish determiners encode 'assertion of existence'. To my knowledge, all the claims in (2) except (2c) are novel to this dissertation. There has been no previous discussion of either specificity or assertion of existence in Salish. The absence of quantificational determiners is proposed by Jelinek (1995), as well as by Demirdache and Matthewson (1995b).2 The absence of a definiteness distinction which exactly parallels the English definiteness distinction is noted by Jelinek (1995) for Straits. However, my claims about the status of definiteness in Salish differ from Jelinek's (see §1.2.2). After arguing for the proposals in (2), I wil l demonstrate that the differences between Salish languages and English-type languages cannot be reduced to superficial matters of morphological encoding. The argument proceeds as follows, taking the definiteness distinction as an example. In languages with no overt distinction between definite and indefinite determiners, there are a priori two logical possibilities, given in (3). The first possibility entails a relatively trivial difference between languages; the second possibility suggests a more fundamental difference.3 3. a. A definiteness distinction is not encoded on the determiners, but is still present in the grammar of the language (i.e. definite and indefinite determiners are homophonous). b. No definiteness distinction is present in the grammar of the language. (3a) accords with the null hypothesis that the range of distinctions accessed by determiner systems is universally provided, and therefore that all languages possess the definite/indefinite distinction, even though it may not be overtly encoded on the determiners. 2 Demirdache and Matthewson (1995b) claim that Salish lacks all quantificational determiners, including a determiner corresponding to English the (since the has been analyzed as a universal quantifier by Milsark 1974 and others). Their arguments are based not on examination of the semantics of determiners (as here), but on relative scope phenomena. Their analysis is discussed in § below, and in Chapter 2, §4.1. 3 Thanks to Robert May (p.c.) for pointing out the first possibility. 27 I wi l l show in this chapter that (3a) is an impossible analysis of Salish languages. Salish determiners not only do not encode definiteness, but also cannot be analyzed as homophonous between definites and indefinites. This is because the distinctions which are encoded in Salish cross-cut the definite/indefinite distinction. The semantic 'pie' is cut up differently in Salish from in English, in ways to be made precise below. Furthermore, DPs in Salish do not display several properties associated with definite DPs, a result which is unexpected if they are ambiguous between definite and indefinite descriptions (see Demirdache 1996a,b,c, §1.2.3 below). The second part of this chapter (§2) discusses the syntactic distribution of determiners in Salish. It is shown that determiners in Salish are generally obligatory on arguments, as predicted by a theory such as that of Higginbotham (1985). Determiners are absent on main predicates, and optional on quantified temporal adjuncts (such as every day). I conclude that determiners in Salish languages function to saturate NPs, enabling them to function as arguments. 1. Distinctions encoded in Salishan determiner systems 1.1. Visibility, proximity, gender and number This subsection will serve as a general introduction to determiner systems in Salish. I wil l show that Salish determiners encode (various subsets of) the distinctions in (4). 4. a. visibility b. proximity (to the speaker) c. gender d. number While all Salish languages encode deictic features (visibility and/or proximity), only a subset of Salish languages encode pronominal features (gender and/or number). The deictic nature of Salish determiners will become relevant in later chapters, as wil l the fact that the visibility and 28 proximity distinctions are always speaker-oriented (i.e. distance from the speaker is what is relevant). See Chapters 2 and 3 for discussion.4 The languages examined in this section are Bella Coola, Upper Chehalis (Tsamosan), Straits (Lummi dialect, Central), Sechelt (Central) and Secwepemctsin (Northern Interior). The systems presented in this subsection will be partially reanalyzed in §1.5, where I introduce the notion of 'assertion of existence'. For a detailed case study of one language (St'at'imcets, Northern Interior), see Chapter 4. Bella Coola. The Bella Coola determiner system is represented in (5). A n over-arching distinction between 'proximal' and 'non-proximal' further divides into six proximity distinctions (labelled I - VI). The proximity dimension encodes both spatial and temporal proximity; see Davis and Saunders (1975) for detailed discussion.5 Gender and number are also encoded. 5. Bella Coola determiners (Davis and Saunders 1975:14): proximal proximal space, present time non-proximal middle space, near past/present distal space, distant past I n ni IV V VI -plural +female C1...CX c1...cayx +a-?1+a?H 4a...+ +a...?H +a...?1+ -female t1...tx t l ...tayx ta-tax. ta...+ ta...tx ta...tax +plural wa...c wa...?ac ta-tax w ta...+ ta...txw ta...tux 4 There are interesting topics for future research involving the distinction between features which are typically encoded on determiners (such as deixis) and features which are often encoded on nouns (such as gender and possibly number). See Keenan (1996), and see Ritter (1991) for proposals that features like number may have their own functional projections inside DP. 5 Bella Coola is not unusual within Salish in allowing temporal notions to be encoded on determiners. M . Dale Kinkade informs me (p.c.) that similar effects are found in Sechelt, Halkomelem, Upper Chehalis, Cowlitz, and possibly in Kalispel and Tillamook. See also Chapter 4 below and Demirdache (1996a,b). 29 Upper Chehalis. Kinkade's (1964) classification of Upper Chehalis, given in (6), emphasizes that the proximity distinctions are speaker-oriented. Speaker-orientation in the determiner system is a reflex of a more general Salish tendency; see Chapter 3. 6. Upper Chehalis determiners (adapted from Kinkade 1964): indefinite by speaker near speaker not near speaker +female t ic , clc ?1c tac, cac c -female t i t ?1t tat t Lummi . Like Upper Chehalis, the Lummi dialect of Straits does not encode number distinctions. Four levels of proximity/visibility are encoded, as well as a gender distinction.6 7. Lummi determiners (adapted from Jelinek and Demers 1994:717): proximate, visible neutral distal/out of sight remote +female Sl 'a sa k w a . ' k w sa general t r a C9 kwca Sechelt. In Sechelt, visibility, number and gender are encoded, but there is a certain amount of neutralization, as shown in (8). The determiner te is ambiguous with respect to visibility, as is the. The determiners tse and che are unambiguously visible and invisible respectively. 8. Sechelt determiners (adapted from Beaumont 1985): visible unspecified or abstract invisible +female -plural . tse, lhe lhe she +plural te te, che -female -plural +plural 6 Montler (1986) assigns internal structure to the determiners in Saanich, another Straits dialect. For example, the formative k w - in that dialect is glossed 'invisible, remote'. It is possible that the determiners in (7) also have internal structure. 30 The 'unspecified or abstract' category will be reanalyzed in §1.5. Secwepemctsin. Secwepemctsin determiners encode a case distinction between 'absolutive' (on objects and subjects of both transitive and intransitive predicates) and 'relative' (on oblique arguments or adjuncts).7 A deictic distinction is encoded between 'present' and 'absent'. For a reanalysis of the division into 'actual-determinate' vs. 'hypothetical-indeterminate', see §1.5 below. 9. Secwepemctsin determiners (adapted from Kuipers 1974:57): actual-determinate hypotheticd-indeterminate absent absolutive 7 1 1 k relative t / t tk/ tk (seldom tke?) 8 The examples presented in this section show that Salish determiner systems encode a range of deictic and pronominal features. Proximity distinctions are robust in all the languages. A subset of Salish determiner systems encode gender and/or number, and Case distinctions are rare. A summary of the discussion so far is provided in (10).9 10. proximity gender number Case Bella Coola i -Upper Chehalis, Lummi V - -Sechelt -Secwepemctsin - - V (subj/obj vs. oblique) 7 Case distinctions are fairly common in Salish, often distinguishing 'direct' arguments (subject and object) from 'obliques' (oblique arguments and adjuncts). The distinctions are usually marked by means of oblique particles, rather than by opposing determiner sets. See for example Hess (1995:82-85) for Lushootseed (Central). 8 The variation in the relative detenniners is dialectal (Dwight Gardiner, p.c). 9 While the subset of Salish languages examined here is large enough to suggest an identifiable 'Salish pattern', there may be counter-examples in other languages of the family. 31 1.2. Definiteness Definiteness is marked on the determiners of many languages, English being a famous example. The purpose of this section is to establish that Salish determiners do not encode definiteness. I wil l also argue that Salish determiners are not amenable to an 'ambiguity' hypothesis, whereby determiners are simply ambiguous or homophonous between definite and indefinite. 1.2.1. Defining definiteness Following Heim (1982) and others, I take the major distinction between definite and indefinite determiners to be a familiar - novel distinction. Definites are familiar to the common ground of the discourse, while indefinites are novel to the common ground of the discourse. This is illustrated in (11-13) for English. When a DP has no discourse antecedent and is unfamiliar to the common ground of the discourse participants, an indefinite is the only possible choice: 11. Novel context: a. I met [a man] today. b. * I met [the man] today. If, on the other hand, the individual corresponding to a DP is already familiar to the common ground of the discourse (for example by being previously mentioned), a definite is the only possible choice. Identity of indices marks coreference: 12. A . I met [a man] i today. (novel) B. What did [the man]j look like? (familiar) 13. A . I met [a man]j today. (novel) B. * What did [a man], look like? (familiar) The notion of 'common ground' includes, but is not restricted to, information introduced overtly into prior discourse; see Heim (1982), Chierchia and McConnell-Ginet (1990), among others. For example, i f a dog runs up to the participants during a conversation and is noticed by all participants, the dog enters the common ground and may be referred to by a definite DP. 32 While familiarity is the main difference between definite and indefinite DPs, there are exceptions to the claim that definite descriptions must always be familiar to the common ground of the discourse, as shown in (14). In (14a), the definite DP introduces an individual who does not exist yet, and may not previously have been under discussion. Similarly, (14b) can be uttered felicitously even in a situation where there was no previous mention of a dog and there is no dog in sight (Heim 1982:371; see also Hawkins 1978). 14. a. I will meet the first baby to be born in the year 2010. b. Watch out, the dog will bite you. Heim (1982) claims that novel definites are rendered felicitous by ACCOMMODATION (see Lewis 1979), a process which adjusts the common ground in the face of a violation of a felicity condition. Lewis's (1979:172) definition of accommodation is as follows: 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. See Heim (1982) and references cited therein for discussion of the conditions under which accommodation is possible. 1.2.2. Salish determiners do not encode definiteness The familiar-novel distinction, so pervasive for determiner choice in English, does not affect determiner choice in Salish. The evidence presented in this section consists of pairs of coreferential DPs, one of which is used in a novel context, and one of which is used in a familiar context. The same determiner is used in both novel and familiar instances, showing that familiarity is not encoded in these languages. 33 Data for this section comes from Sechelt (Central), Lushootseed (Central), St'at'imcets (Northern Interior), Secwepemctsin (Northern Interior), Straits (Central), Bella Coola and Upper Chehalis (Tsamosan). Choice of languages is based on availability of relevant data. Sechelt. The following example from Sechelt illustrates the absence of a familiar-novel distinction. (15a) is the first mention of a snake woman in the text; (15b) contains a subsequent mention of the same creature. In both cases, the same determiner (the) is used. 1 0 15. a. t'i suxwt-as [lhe 7ulhka7 slhanay]i... fact saw-he [det snake woman] ... 'He saw [a snake-woman],..." (novel) b. t'i tl'um s-kwal-s [lhe slhanay]*:... fact then nom-speak-her [det woman]: ... 'Then [the woman]* said: ..." (familiar) (Sechelt; Beaumont 1985:188) In accordance with the lack of a definiteness distinction, all the Sechelt determiners are glossed as 'the, a' by Beaumont (1985). Lushootseed. Lushootseed determiners do not encode definiteness; according to Hess (1995:9) 'The distinction in English between the and a does not exist in Lushootseed.' (16) shows that the same determiner may be used, regardless of whether the individual concerned is novel or familiar to the discourse situation. (16a) is the first mention of a whale in the text; (16b) occurs two lines later and refers to the same whale. 1 1 16. a. huy, Sudxw9xw [t1?H c"xw9lu?] complete see [det whale] 'They saw [a whale]i.' (novel) b. bapad9x w 9 lg w 9? ft1?1+ c"xw9lu?] pester pl [det whale] 'They pestered [the whale]i.' (familiar) (Lushootseed; Hess 1995:140) 1 0 In Sechelt examples I substitute a 7 for Beaumont's ?, since this eliminates a number of otherwise unnecessary font-changes. 1 111 ?1+is the distal, non-feminine determiner (Hess 1995:77). 34 St'at'imcets. The absence of a familiar-novel distinction on St'at'imcets determiners is illustrated in (17). The same determiner is used when the individual is novel and when she is familiar. (The relevant determiner is discontinuous, containing both a proclitic and an enclitic portion. See Chapter 4 for detailed discussion of St'at'imcets). 17. a. huyH-kan ptakw+, p t a W m l n lC?a [tl Smlrri+arj-a]... huy'-lhkan ptakwlh, ptakwlh-min lts7a [ti sm6m'lhats-a] ... going.to-lsg.subj tell.story tell.story-appl here [det woman(redup)-det] 'I am going to tell a legend, a legend about [a girl ] i . . . ' (novel) b. wa? kwu? ?flal la'tl? [tl Sm^rMaC-al wa7 ku7 flal l£ti7 [ti smem'lhats-a] prog quot cry deic [det woman(redup)-det] '[The girl]j was crying there.' (familiar) (St'at'imcets; van Eijk and Williams 1981:19) Secwepemctsin. The lack of a definiteness distinction in Secwepemctsin is illustrated in (18), where (18a) is the first sentence in the narrative, and (18b) appears a few lines later. The distinction between novel and familiar entities is not marked on the determiner. 18. a. w ? l x - 9 k w e [y twwfwt], n'eYi n-sqwmcfns [y Fraser R i v e r ] . . . be-3 [det boy] deic prep-shore-3poss [det Fraser River] 'There was [a boyh, who lived over on the bank of the Fraser River ..." (novel) b. w ? l x [y twwfwt] wlex-mn-s ?s - t fq w -n -s be [det boy] think.pity-tr-3erg to-kill-tr-3erg yyf -y?e ' n e [y sqaxqxe ?lye] deic [det puppy] '[The boyh thought it a pity to kil l those puppies.' (familiar) (Secwepemctsin; Kuipers 1974:103) Straits (Lummi, Saanich dialects). Lummi determiners also do not encode definiteness, according to Jelinek (1995:512) (see also Jelinek 1993a). Jelinek claims that absolutive patients 'may be freely interpreted as definite or indefinite' (1995:528): 19. ler j - t-0-s tea smayes] see-tr-3abs-3erg [det deer] 'He saw the / a deer.' (Lummi; Jelinek 1995:528) 35 The claim that Straits lacks a definiteness distinction is supported by data from the Saanich dialect. In (20), novel and familiar entities receive the same determiner: 20. a. ?1? xwa"lkwt-9n [tsa esq qw-r-e^] accom drift-cmdl [det big log-wood] "There was [a big log]i floating in the water.' (novel) (Saanich; Montler 1986:245) b. ?aa+ ?9 [tsg qw+-^y] . . . go.aboard obi [det log-wood] ... 'Go aboard [the log]j.. . ' (familiar) (Saanich; Montler 1986:246) Jelinek's (1995) claims about the status of definiteness in Straits differ from my proposal that the distinction between familiar and novel usages is simply irrelevant in Salish. According to Jelinek, certain contexts force only definite readings. Most subject DPs are either definite or generic; 'intransitive subjects are presuppositional' (i.e. familiar). 1 2 The expression of an ' indefinite generally requires a syntactic paraphrase, such as using the nominal concerned as a predicate rather than an argument, as in (21). 21. c"-s4enly-sxw rel-female-2sg.nom 'You have a wife/woman.' (Lummi; Jelinek 1995:529) Throughout the discussion of definiteness so far, I have not made reference to grammatical functions, assuming that the absence of a familiar/novel distinction extends to any argument DP. There may be a tendency for subjects to refer to familiar individuals; this follows independently, however, from the mapping of subject to discourse topic in Salish (see Kinkade 1989, 1990, Davis 1994c, Roberts 1994, Demirdache and Matthewson 1995b). Topics of the discourse are by definition familiar to the discourse participants and present in the common ground; as such they wi l l be interpreted as definite. However, just as the mapping from subject to topic is not absolute, the correlation of subject with a definite interpretation is not absolute, in at least some languages. The St'at'imcets examples in (22) show that an intransitive subject can easily be used 1 2 Exceptions arise with explicit existential or locative contexts, which allow an indefinite reading for a subject (Jelinek 1995:528). 36 to refer to a novel individual (i.e. in an indefinite context). The DPs in both (22a) and (22b) introduce previously unknown individuals. 22. a. nH kwu? ?ay+ S-Cut-§ [tl qXl-a ?ux w alm1x w ] . . . nilh ku7 aylh s-tsut-s [ti qvl-a ucwalmicw] ... foe quot then nom-say-3sg.poss [det bad-det person] 'And then a bad person said ..." (novel) (St'aYimcets; van Eijk and Williams 1981:32) b. *ak kwu? ka"t1? t?1 nkya'p-a], ?^?9nwa§ t'ak ku7 k£ti7 [i nk'yap-a], a7en'was go quot deic [pl.det coyote-det] two(animal) 'Two coyotes were going along.' (novel) (St'aYimcets; van Eijk and Williams 1981:32) Bella Coola. Bella Coola's complex determiner system is discussed in detail in §1.5 below, but a few remarks can be made here. Nater (1984:41) (who does not propose a definiteness distinction) notes that when a prefixal determiner is used without an accompanying deictic suffix, the usual translation is English a(n) (or 0 for plurals). Hence, the most likely candidate for an indefinite determiner would be these lone prefixal determiners. However, (23) shows that a novel individual may also be represented by a prefix-suffix combination. This suggests that determiner choice in Bella Coola cross-cuts the familiar-novel distinction. 23. s-kwn-tim-tuts' c [ti qw'xwmtimut t'ayc] nom-take-3pl.pass-again prep [det car det] 'Somebody picks them up again in a car.' (novel) (Bella Coola, Nater 1984:140) Upper Chehalis. At first glance, Upper Chehalis presents a problem for the generalization that Salish lacks definiteness marking. As seen in (6) above, Kinkade classifies the articles t and c as indefinite. 1 3 However, the Upper Chehalis 'indefinite' articles t and c do not correspond to the English indefinite article a(n). For example, when a speaker refers to an individual who is known to the speaker, but who has not been previously mentioned in the discourse, we would tend to expect an indefinite in English. However, in such contexts, the article tit, rather than t or c, is likely to be used (M. Dale Kinkade, p.c.). 1 4 1 3 Cowlitz (Tsamosan) also shows definiteness marking, according to M . D . Kinkade (p.c). 1 4 To be precise, an indefinite is expected whenever the discourse referent is not part of the common ground; being previously mentioned is just one way of being part of the common 37 In Upper Chehalis texts, there are also puzzling instances of t, the supposedly indefinite determiner, in familiar contexts. For example, the women mentioned in (24) (the daughters of fire) are the topic of the story and have been mentioned several times already, yet / is used: 24. wl ?9ya"llwn tt Eawa+umSl and good-appear [det women] 'and the women were good-looking' (Upper Chehalis: M D K , p.c.) Such examples suggest that the Upper Chehalis distinction does not correspond to the familiar-novel (definite-indefinite) distinction; nor, however, does the distinction appear to fit in with patterns found in the rest of the Salish family. Determiners in Upper Chehalis are frankly puzzling, and the lack of native speakers means that fieldwork cannot help us sort out the puzzle. 1.2.3. Evidence against a homophony analysis In the previous section, it was shown that Salish determiners do not overtly distinguish between entities which are novel (not present in the common ground) and familiar (present in the common ground). It is still possible, on the basis of the data presented thus far, to claim that Salish determiners simply display homophony between definite and indefinite determiners. That is, the distinction might still be present in the grammar, although it is not overtly encoded. If DPs in Salish were ambiguous between definites and indefinites, then they should be able to display the properties of either definites or indefinites. However, Demirdache (1996a,b,c) shows that DPs in St'at'imcets fail to display the properties of definite descriptions. This provides evidence against the homophony analysis. ground, as noted above. Therefore, to rule out the definiteness analysis conclusively for Upper Chehalis, we would need proof that tit can be used when describing discourse referents which are not part of the common ground for any reason. Such information is not available at the present time. 38 One way in which DPs in St'at'imcets do not act like definite DPs is that they do not have an Individual Concept Reading. The Individual Concept Reading is illustrated in (25). 25. The president of the United States is powerful. for any time t, whoever is president at t is powerful at t (Demirdache 1996c: 8) The definite DP the president of the United States has a reading where it does not describe a particular individual, but indicates any past, present or future president. The corresponding DP in St'at'imcets (shown in (26)) can be used either to refer to a particular individual that the speaker has in mind, or it can be used attributively, to refer to whoever is the current president. It cannot be used to describe any past, present or future president.15 26. vaxa? [tl kel 'aqStsn-S-a t l Unlted.States-al a7xa7 [ti keI7aqsten-s-a ti United. States-a] powerful [det chief-3sg.poss-det det United. States-det] 'The president of the United States is powerful.' * for any time t, whoever is president at t is powerful at t. (Demirdache 1996c) This shows that DPs containing the determiner ti.. .a in St'at'imcets do not display a property associated with definite DPs, namely allowing an Individual Concept Reading. This in turn indicates that the determiner ti.. .a is not homophonous between a definite and an indefinite determiner. If the definite and indefinite uses of ti.. .a were merely homophonous, then ti.. .a should be able to be interpreted as a definite description. In that case, we would expect the Individual Concept Reading (one of the possible readings of definite DPs) to be possible.1 6 1 5 If the speaker wishes to denote any past, present or future president, the adverb papt 'always', in combination with the determiner ku, must be used (see §1.5 below, Chapter 4 for discussion of ku): i . papt ?axa? [kwu kal?^q§ten-§ t l United.States-a] papt a7xa7 [ku kel7aqsten-s ti United. States-a] always powerful [det chief-3sg.poss det United. States-det] 'The president of the United States is always powerful.' (St'at'imcets; LT) 1 6 Irene Heim suggests (p.c.) that there is no necessary connection between the Individual Concept Reading and definiteness. At least under a quantificational analysis of the definite determiner (as in Russell's work), we would expect the Individual Concept Reading to be possible for all definites, since it is the reading which gives rise to a universally quantified interpretation. Further research is required into the connection between the two notions. 39 §1.5 below provides further evidence against the homophony analysis of Salish determiners, based on the fact that overt determiner distinctions in Salish cross-cut the definite-indefinite distinction. 1 7 ' 1 8 1.3. Specificity This section wil l demonstrate that specificity is not encoded in the St'aYimcets determiner system. From that basis, I wil l generalize to make the strong prediction that specificity is not encoded in Salishan determiner systems. 1.3.1. Defining specificity A n intuitive definition of specificity says that a DP is specific when the speaker has a specific individual in mind (cf. for example Kamp and Reyle 1993:289, Ioup 1977; see also discussion in Hellan 1981). A more rigorous definition of specificity is required when we are testing languages like Salish, for which most linguists do not have native speaker intuitions. I wil l utilize two definitions of specificity, namely those of Ludlow and Neale (1991) and Enc (1991). 1 7 C.-T. James Huang notes (p.c.) that there are languages which lack determiners altogether, but which display definiteness effects elsewhere in the grammar. If definiteness effects were found elsewhere in Salish, this might constitute evidence for an underlying definiteness distinction in the determiners which is neutralized on the surface (although the empirical problems noted in the text would still remain). However, I have found no indication, either in my own fieldwork or in published materials, that the distinction between familiar and novel individuals is ever grammatically relevant in Salish. 1 8 One final note is in order regarding homophony. According to Heim (1982), a definite determiner instructs the hearer to locate the representation for a familiar individual, while an indefinite determiner instructs the hearer to create a representation for a novel individual. If definite and indefinite determiners are homophonous, there can be no such instructions to the hearer, in effect removing the entire content of the definite-indefinite distinction. One could of course argue that for reasons of Universal Grammar, it is still preferable in this situation to postulate homophony rather than cross-linguistic parameterization in definiteness marking. Note, however, that homophony on such a large scale does not come cheap. Just in the subset of Salish languages investigated here, there are over 50 different determiners, which would have to correspond to over 100 determiners in the underlying representation, with every definite-indefinite pair being homophonous. 40 Ludlow and Neale's (1991) definition of specificity utilizes the following three concepts: 27. a. Speaker's Grounds: the proposition that is the object of the most relevant belief furnishing the grounds for an utterance b. Propositions Meant: the proposition(s) a speaker intends to communicate c. Proposition Expressed: the proposition expressed by the utterance (Ludlow and Neale 1991:176) The Speaker's Grounds, Proposition Meant and Proposition Expressed need not be identical for any one utterance, as will be seen immediately below. Specificity relies on a mismatch between the Speaker's Grounds and the Proposition Meant, according to Ludlow and Neale. In particular, these must differ in whether they are SINGULAR PROPOSITIONS or GENERAL PROPOSITIONS. Singular propositions are those which contain only directly referring expressions (such as proper names), and which are therefore 'about' particular individuals. A n example is given in (28). 28. Moana quit her job. (singular proposition) General propositions, on the other hand, contain only definite or indefinite descriptions, as in (29). It is possible to understand a general proposition without being acquainted with any particular individual who satisfies the description (Ludlow and Neale (1991:173). See also Loar (1976). 29. The woman who won a million dollars yesterday quit her job. (general proposition) Specificity arises when there is a mismatch between the Speaker's Grounds and the Proposition Meant, such that the Speaker's Grounds is a singular proposition, while the Proposition Meant is a general proposition:1 9 1 9 cf. also Donnellan (1966), Kripke (1977) on the distinction between the speaker's grounds and the speaker's intentions in the use of definite descriptions. 41 When the speaker has singular grounds for an assertion of the form 'An F is G 1 but no intention of communicating a singular proposition, let us say that the indefinite description 'an F is used specifically (Ludlow and Neale 1991:181; emphasis original). A n example wi l l make things clearer (adapted from Ludlow and Neale 1991). Suppose that Mary is looking gloomy because a tax auditor she knows, M r Beastly, is coming to audit her today. If asked why she is looking gloomy, she could respond with (30). 30. A n auditor is coming to see me today. Suppose also that Mary has no reason to expect that the hearer of (30) knows Mr Beastly, or knows that he is the particular auditor who is coming to see her. In that case, she intended to communicate only a general proposition. However, the grounds for her utterance were a singular proposition, as shown in (31). Hence, the DP an auditor has been used specifically. 2 0 31. Speaker's Grounds: M r Beastly is coming to see me today. Proposition Meant: An auditor is coming to see me today. Proposition Expressed: A n auditor is coming to see me today. The second test for specificity I will use when investigating Salish comes from Enc (1991). Enc. argues, that if the object DP in the second sentence in (32) is specific, it wil l pick out two of the children already under discussion. A non-specific object DP will pick out two separate girls, not already under discussion. 32. A lot of children came in. I knew two girls. There are languages which overtly encode the distinction between these two readings of the object DP (e.g. Turkish). Salish languages do not, as will be shown in the next subsection. 2 0 Ludlow and Neale do not discuss situations where there is a mismatch between the Speaker's Grounds and the Proposition Expressed, or the Proposition Meant and the Proposition Expressed. These are peripheral to the current concerns in any case. 42 1.3.2. Salish determiners do not encode specificity Evidence about specificity cannot easily be extracted from descriptive grammars, since we need to know which interpretations particular utterances may and may not have. Such matters are not usually overtly addressed in the available materials. Hence, the evidence in this section comes mainly from St'at'imcets, with some indirect evidence drawn from Beaumont's (1985) description of Sechelt. St'at'imcets. In St'at'imcets, mismatches between Speaker's Grounds and Proposition Meant are not relevant for determiner choice. In (33), there is a mismatch between the Speaker's Grounds and the Proposition Meant. The speaker has as her grounds for the utterance a singular proposition, but all that is meant to be expressed to the hearer is a general proposition. The determiner ti.. .a is used. 2 1 33. Context: (specific) The speaker has just heard on the telephone that a teacher she knows named Leo is coming. She reports this information to a colleague. Speaker's Grounds: Leo is coming. Proposition Meant: A teacher is coming. Proposition Expressed: A teacher is coming. In (34), we have a non-specific DP. Both the Speaker's Grounds and the Proposition Meant are general propositions. A n individual who is described only as 'a teacher' is coming to see the speaker. It is possible that the speaker herself does not know who the teacher will be. Again, the determiner ti.. .a is used. cuz' ku7 going.to quo A teacher is coming. (St'at'imcets; LT) 2 1 The enclitic portion of the determiner ti...a is phonologically deleted following the auxiliary wa7 in (33). 43 34. Context: (non-specific) The speaker has just heard on the telephone that a teacher is coming. She reports this information to a colleague. Speaker's Grounds: A teacher is coming. Proposition Meant: A teacher is coming. Proposition Expressed: A teacher is coming. xwuz" kwu? C?a§ [tl wa? c'una'rri-xal] cuz' ku7 ts7as [ti wa7 tsunam'-cal] going.to quot come [det prog teach-intr] A teacher is coming.' (St'aYimcets; LT) A third and final possibility is illustrated in (35). Here, all propositions are general, just as they were in (34). This time, however, the speaker uses the determiner ku, rather than ti.. .a. For the distinction in meaning between ku and ti...a, see §1.5. For now, all that is relevant is that non-specific DPs may contain either determiner. 35. Context: (non-specific) The speaker has just heard on the telephone that a teacher is coming. She reports this information to a colleague. Speaker's Grounds: A teacher is coming. Proposition Meant: A teacher is coming. Proposition Expressed: A teacher is coming. xwuz" kwu? c"?a§ [kwu wa? c'una'rri-xal] cuz' ku7 ts7as [ku wa7 tsunam'-cal] going.to quot come [det prog teach-intr] A teacher is coming.' (St'aYimcets; LT) The data in (33-35) show that St'aYimcets does not encode the difference between specific and non-specific. The reasoning follows the same logic as in the preceding section on definiteness. When we compare a context in which there is a mismatch between the Speaker's Grounds and the Proposition Meant (33) with the two contexts where there is no mismatch (34-35), we see that the difference between a mismatch and no mismatch is not reflected in determiner choice. The distinctions which are encoded in St'aYimcets 'slice up' the available semantic space differently than a specific/non-specific distinction would. On the one hand, the same determiner can be used in both specific and non-specific contexts ((33) vs. (34)), and on the other hand, the 44 whole set of non-specifics is not uniformly marked (as seen in the contrast between (34) and (35)). The determiner ku can only be non-specific, but the determiner ti.. .a may be either specific or non-specific. It is therefore impossible to analyze St'at'imcets determiners as simply homophonous between specific and non-specific readings.22 Applying Enc/s test for specificity similarly shows that specificity is not encoded in St'at'imcets. Recall that according to En9, if the object DP in the second sentence in (36) is specific, it picks out two of the children already under discussion. A non-specific object DP picks out two separate girls, not already under discussion. 36. A lot of children came in. I knew two girls. The St'at'imcets version of (36) is given in (37). 37. txw?1t ?1 Skwamkwukwm1?t-a] ?u+xw [cw7it i sk'wemk'uk'wmi7t-a] ulhcw [many pl.det child(redup)-det] go.in 'A lot of children came in.1 zwat-sn-i-kan [?1 n?anwa§-a Sme+ma'm+ac'] zwat-en-lhkan [i n7an'was-a smelhme'm'lhats] know-tr-lsg.subj [pl.det two(human)-det girl(redup)] T knew two girls.' (St'at'imcets; BF) In (37), both the 'under discussion' and the 'not under discussion' readings are possible for i n7dn'wasa smelhme'm'lhats 'two girls'. This DP can be either specific or non-specific. Sechelt. Beaumont (1985) divides determiners in Sechelt into two groups: one 'unspecified or abstract' determiner, she, vs. all the rest. If Sechelt determiners encoded specificity, then all the non-she determiners would constitute the specific set. However, Beaumont claims that the non-she determiners are used whenever the speaker either can or could point out the individual concerned. In other words, it is not necessary that the speaker actually be able to identify the 2 2 Irene Heim (p.c.) expresses doubt that there are any languages at all which encode specificity in the sense of Ludlow and Neale (1991). If this is correct, then the absence of such encoding in Salish wil l not need to be specially derived. See also discussion in Chapter 2, §5.4. 45 individual, but only that it is potentially possible for the individual to be identified. This could be interpreted as meaning that the distinction between a singular and a general proposition in the Speaker's Grounds is not relevant for determiner choice in Sechelt. Further evidence is given in §1.5 for the claim that specificity is not encoded in Salish. In that section, I argue that the only distinction encoded in Salish other than the deictic and pronominal distinctions is one of assertion of existence, a notion which differs from specificity in ways to be made precise. There has been much debate about whether specificity is related to or reducible to wide scope. There is also debate about whether specific indefinites are REFERRING TERMS, i.e. terms that are used to refer directly to individuals. Fodor and Sag (1982), Kamp and Reyle (1993:290) among others claim that specifics are referring terms; Enc (1991), Ludlow and Neale (1991) make the opposite claim. See Heim (1989) for an overview of the issues involved. While I suspect that the specific - non-specific distinction is independent of scopal behaviour, and that specific indefinites are not directly referring terms, these issues are not crucial here. It suffices to show that Salish languages do not encode specificity on their determiners. 1.4. Quantificational determiners This section will show that quantificational determiners are absent from Salish languages. 1.4.1. Defining quantificational determiners Quantificational determiners in English are illustrated in (38). 38. a. [Every man] loves hockey. b. [No man] loves hockey. c. [Most men] love hockey. 46 The lexical items every, no, and most are in complementary distribution with definite or indefinite determiners. 39. a. [(*the) every (*the) man] loves hockey. b. [(*the) no (*the) man] loves hockey. c. [(*the) most (*the) men] love hockey. This follows under the common analysis whereby the quantifiers occupy the D° position (since there may only be one D head in each Determiner Phrase).23 The constructions in (38) will be shown to be systematically absent in Salish. I wil l first show that elements with quantificational force do not occupy the D position, and then argue that elements which are clearly determiners do not have quantificational force. The discussion of quantifiers is organized according to the strong/weak quantifier division. 2 4 Weak quantifiers, following Milsark (1974, 1977) and others, are those that are legitimate in the environment in (40) (a fftere-insertion context). 40. There are [ New Zealanders] in the garden. The determiners in (41) are strong, while those in (42) are weak (for the analysis of the as a strong quantifier, see Milsark 1974). 41. Strong quantifiers: a. * There is every New Zealander in the garden. b. * There are all New Zealanders in the garden. c. * There are most New Zealanders in the garden. d. * There are the New Zealanders in the garden. 2 3 This is not unanimously accepted; see for example Stowell (1993), who proposes a separate phrase QP, of which the quantifiers in (38) presumably occupy the head position. 2 4 The discussion in this section is organized according to the strong/weak division in English. I wil l provide Salish-internal evidence for a strong/weak distinction in Chapter 6. 47 42. Weak quantifiers: a. There are some New Zealanders in the garden. b. There are many New Zealanders in the garden. c. There are three New Zealanders in the garden. d. There are no New Zealanders in the garden. 1.4.2. There are no quantificational determiners in Salish The absence of strongly quantificational determiners Based on the there-insertion diagnostic, quantifiers such as every, all, most and the are classified as strong quantifiers in English. I have already established that Salish does not have the equivalent of definite the. In this section I will show that Salish lacks determiners corresponding to universal quantifiers or most. The languages discussed are St'at'imcets (Northern Interior), Secwepemctsin (Northern Interior), Squamish (Central), Upper Chehalis (Tsamosan), Halkomelem (Cowichan dialect; Central), and Lushootseed (Central). DP arguments may contain universal quantifiers in Salish languages; see Demirdache et al. (1994), Matthewson (1994a,b), Chapter 5 below. However, the quantifiers are not in complementary distribution with determiners, but on the contrary must co-occur with them, as in (43).25,26 43. a. qw8la 'w-8m [takem q'welaw'-em [takem pick.berries-intr [all ' A l l the women picked berries.' ?1 §yaqc?-a] i syaqts7-a] pl.det woman-det] (St'at'imcets; LT) qwetsets [xwexweyt re leave [all det ' A l l the men left.' sqelemc] man] (Secwepemctsin; Demirdache et al. 1994) 2 5 Languages which are not included in (43) are omitted because of lack of available data. For example, I can only find one example of a universal quantifier in Beaumont's (1985) description of Sechelt, and only a couple in Nater's (1984) description of Bella Coola. Neither author provides discussion which would enable generalizations about the syntax of universally quantified phrases to be made. 2 6 The glossing of the quantifiers in (43) and elsewhere as 'all' relies on detailed semantic argumentation to be provided in Chapter 6. Jelinek (1995) has argued that Straits Salish possesses only an adverbial universal quantifier, with the meaning of 'always'. 48 c. na ch'aw-at-as [i7x,w ta siw'i7ka] [ta slhenlhanay1] rel help-tr-3erg [all det men] [det women] ' A l l the men helped the women.' (Squamish; Demirdache et al. 1994) d. ?axwa"-w-n [xwaqwu t ?^11s-umSl run-intr-3subj [all det chief-people] ' A l l the upper-class people run.' (Upper Chehalis; M . D . Kinakade, p.c.) e. ni x^lsnrjansm [makw kwGa s*al?fqa+] aux run(pl) [all det children] 'AH the children ran.' (Cowichan; Gerdts 1988:79) f. xw1? g^-ad-s-as-tala'but tb9kw t(1) ads-g '9d .gwat9dl neg if-my-nom-stat-understand [all det your-language] 'I don't understand all your language.' (Lushootseed; Hess 1976:26; Bates et al 1995) In the languages for which I have negative evidence (St'aYimcets, Squamish and Secwepemctsin), deletion of the determiner in constructions such as (43) is not possible. For example, (44a) corresponds to (43a) with the determiner omitted, and (44b) likewise shows the impossibility of deleting the determiner. 44. a. * q w 9 l a w - 9 m [teikam §m9+mu+ac1 * q'welaw'-em [uikem smelhmulhats] pick.berries-intr [all woman(redup)] ' A l l the women picked berries.' (St'at'imcets) b. wa? ?ama-mfn-1ta§ k-wa pfx-am [tak9m twlww9t-*(a)] wa7 ama-min-itas k-wa pix-em' [tak^m *(i) twew'w'et-*(a)] prog good-appl-3pl.erg det-prog hunt-intr [all *(pl.det) boy(redup)-(*det)] ' A l l boys love hunting.' (St'aYimcets; LT, RW) The constructions exemplified in (43) are not the only ones in which the universal quantifiers appear; see Demirdache et al. (1994), Kroeber (1994c), Jelinek (1995), §2.2.3 below. The generalization holds that when universal quantifiers appear inside an argument DP, they must always co-occur with a determiner.27 2 7 Lummi (Northern Straits; Central) does not allow a quantifier to attach prior to a determiner inside a DP as in (43) (Jelinek 1995). 49 Turning to the strong quantifier most, I have not found an element corresponding to most in the materials available for any Salish language. The languages which appear to lack such a determiner include Sechelt (Central; Beaumont 1985), Lushootseed (Central; Bates et al. 1994), Straits (Central; Jelinek 1995), Columbian (Southern Interior; M.D. Kinkade, p.c.) and Upper Chehalis (Tsamosan; Kinkade 1991). St'at'imcets definitely lacks most; for a detailed demonstration of this, see Chapter 7. In this language, the meaning of most is paraphrased either by 'almost all', as in (45a), or by an overt partitive 'many of, as in (43b). 45. a. ttqtt * u ? taksm ?1 Sma+mu+ac'-a] &1q [tqilh t'u7 tikem i smelhmulhats-a] t'iq [almost just all pl.det woman(redup)-det] arrive 'Most of the women arrived.' ('Almost all of the women arrived.')(St'at'imcets; RW) b. [xw?1t +^ 1-k1 n-Snakwnukw?-a] ?acx-an-an ?1 natxw-aS [cw7it lheT-ki n-snek'wnuk'w7-a] £ts'x-en-an i n&tcw-as [many from-pl.det lsg.poss-friends-det] see-tr-lsg.conj when. day-3sg.conj T saw many of/most of my friends yesterday.' (St'at'imcets; RW) We have seen in this section that strong quantifiers in Salish either do not exist as single lexical items (as is the case with most), or must co-occur with an overt determiner (as with the universals). In no instance does a strongly quantificational element occupy the D° position, heading the DP. The absence of weakly quantificational determiners Based on the r/iere-insertion diagnostic, quantifiers such as some, many, the numerals and no are classified as weak quantifiers in English (see (42) above). This section will show that the corresponding elements in Salish do not have the status of determiners. 50 Weak quantifiers in Salish languages do not occupy the determiner position, as shown in (46) for St'aYimcets.28 46. * ?lcx-9n-+kan [n-^lftwaS §mu+ac'] * dts'x-en-lhkan [n-7an'was smulhats] see-tr-lsg.subj [two(human) woman] 'I saw two women.' (St'aYimcets; G N , RW) Instead, weak quantifiers in Salish must co-occur with a determiner when they appear inside DP. Examples are given in (47) containing numerals and many; see Chapter 6 for further discussion. 47. a. ?1 x w ?£z -a§ kw-S xln-S, i cw7aoz-as kw-s cin'-s, when.past neg-3sg.conj det-nom long.time-3sg.poss zuq w -a§ [?1 n^aTiwaS-a ?uxwalm1xw wa? zwaY-an-an] zuqw-as [i n7an'was-a ucwalmicw wa7 zwaYen-an] die-3sg.conj [pl.det two(human)-det person prog know-tr-lsg.conj] 'Not long ago two people that I knew died.' (St'aYimcets; BF) b. Sfs-n It q&+ rjawa+tclmsl come-3subj [det many girl[dimin]] 'Many girls come.' (Upper Chehalis; M . D . Kinkade, p.c.) c. s-1? k wan-nax w-s ttsa rjafi §a-§k w ^ml nom-accom see-cont.tr-3poss [demon many actual-swim] 'and he did see a bunch of swimmers' (Saanich; Montler 1986:251) The strong/weak classification in English groups the negative no with the weak quantifiers (see (42) above). In Salish, there is syntactic dissimilarity between negation on the one hand, and quantifiers such as many, (a) few and the cardinal numbers on the other. It is not clear that the English strong/weak grouping carries over to Salish when it comes to negation. However, for completeness I will show here that negation also does not occupy determiner position in Salish (see Chapter 7 for further discussion). 2 8 Numerals in Upper Chehalis have been recorded in a construction parallel to (46) (M.D. Kinkade, p.c). The analysis of such DPs in Upper Chehalis is a topic for future research. 51 Salish languages do not possess negative determiners. The negative element in Salish typically functions as a main predicate, as shown in (48a-g). In each case the negation appears clause-initially and is followed by a nominal or a subordinate clause. 48. a. t8ta"?a [k-s slk-t-anal neg [det-nom hit-tr-lsg.subj] T never hit him. /1 didn't hit him.' (N+e?kepmxcfn; DU) mf+ta tt neg [det T won't sing.' n-qa-f my-modal s-'flaft] impf-sing] (Upper Chehalis; Kinkade 1976:19) xwa-kwat syaqcu-s neg-thing wife-his 'He had no wife.' ('His wife didn't exist.1) (Sechelt; Beaumont 1985:187) d. f. k w £-k w 1? 7 a ] ?£wa-na? s-teV) act+hungry limit not-exist s-what "They were hungry but they didn't have any food.' xw1? gw-ad-yfqus neg might-my-basket T don't have a basket.' s-?ttan-s s-eat-3poss (Saanich; Montler 1986:242) (Lushootseed; Hess 1976:567, Bates et al. 1994) • we-te kwa la qw8ls0a.m neg-thing det subj? medicine 'Isn't there any medicine?' (Chilliwack: Galloway 1977:285) g. 7axw ti ka lhalas 7ala 7ats neg det hyp boat prep here 'There is no boat here.' (Bella Coola; Nater 1984:123) Bella Coola allows negation inside an argument DP, but the negative element is not a determiner. Nater notes that when the negative occurs attributively, it results in a meaning such who do(es) not X' , as shown in (49). This corresponds to predicate negation. as 'the one(s) 49. ti 7axw t'ayc ksnmak det neg det work 'this one who is not working' (Bella Coola, Nater 1984:122) 3 x, - i do it (x) 52 Compare this with an English D P like no woman, which does not mean 'the one who is not a woman', as shown in (50). 50. [No woman] loves hockey. a. - i 3 x (woman (x) and loves hockey (x)) b. * 3 x, - i woman (x), loves hockey (x) Deteminers do not have quantificational force So far we have seen that quantifiers corresponding to all, most, many, the cardinal numbers, and negation never occupy the determiner (D°) position inside an argument D P in Salish. To prove that Salish lacks quantificational determiners, I must also demonstrate that the determiners which do exist do not have quantificational force. One way of doing this is to show that DPs of the form [D NP] do not have quantificational force, unlike DPs which contain overt quantifiers. Demirdache and Matthewson (1995b) provide support for this c la im by showing that DPs do not undergo Quantifier Raising in Salish languages, unless an overt quantifier (such as the D P -adjoined universal quantifiers in (43)) is present. Evidence that determiners do not induce Quantifier Raising in St'at'imcets comes firstly from word order. Reca l l that the underlying word order in Salish languages is predicate ini t ial . Wh i le DPs containing adjoined quantifiers may raise to the front of the sentence by S-Structure, DPs without adjoined quantifiers may not. Demirdache et al. (1994) analyze the movement in (51a) as Quantifier Raising. (51b) shows that DPs may not undergo this movement unless an overt universal quantifier is present. 2 9 51. a. [takam ?1 ?ux w a lmfx w -a l i wa? ?a lk§ t t i [ t ikem i ucwalmfcw-a]i wa7 alkst ti [all pl.det person-defJi prog work ti ' A l l the people are working.' (St'at'imcets; B F ) 2 9 Speakers of the Lower dialect of St'at'imcets allow DPs to appear sentence-initially with more freedom than speakers of the Upper dialect, whose judgements are reflected in (51). 53 b. * [?1 ?uxwalmfxw-a]i wa? valkSt t i * [i ucwalrnicw-a]i wa7 alkst ti [pl.det person-det]i prog work tj "The people are working.' (St'at'imcets; G N , RW) Further evidence that determiners in St'at'imcets do not induce Quantifier Raising comes from discourse properties. Demirdache and Matthewson (1995b) argue that overt DPs in St'at'imcets (and in Salish more generally) never refer to the topic of the discourse, or to 'old' information, unless they are accompanied by an adjoined quantifier (see also Davis 1994c). Assuming (with Diesing 1992) that it is presuppositional (topical) elements which escape the V P by Logical Form, it follows that overt DPs do not escape the V P at Logical Form in St'at'imcets, unless they are accompanied by an adjoined quantifier. Independent evidence that ordinary DPs in St'at'imcets do not have quantificational force is provided by Demirdache (1996a,b). Demirdache argues that 'DPs in St'aYimcets do not have the range of temporal interpretations that are characteristic of presuppositional (quantificational) DPs' (1996b:9). In particular, quantificational DPs allow temporally free readings, as shown in (52) (see also Musan 1995). In (52), the time at which the individuals were/are homeless is independent of the time at which the rally took place: 52. The homeless people were at the rally. (cf. Musan 1995) a. true if the individuals who were homeless at the time of the rally were at the rally b. true if the individuals who are homeless now were at the rally (i.e. they were not necessarily homeless at the time of the rally) Demirdache (1996a,b,c) demonstrates that DPs in St'aYimcets are temporally bound. In (53), the DP cannot have a temporal interpretation which is independent of the temporal interpretation of the predicate (which is this case is 'past', due to the presence of the completive marker tu7). 54 53. ?axav tu? [tl kal?lqSten-S-a t l Unlted.States-a] a7xa7 tu7 [ti kel7aqsten-s-a ti United.States-a] powerful compl [det chief-3sg.poss-exis det United.States-exis] "The president of the United States was powerful.' (St'at'imcets; Demirdache 1996c) a. true if the individual who was the president at some past time was powerful at that past time b. true if the individual who is the president now was powerful at some past time (while he was president) which overlaps with the present time c. false if the individual who is the president now was powerful at some distinct past time (before he was president) Notice that the definite DP in the English gloss for (53) can have a temporally free reading, being true in the context in (53c). The English version of (53) can thus be used to assert that the current president (i.e. Clinton) was powerful at some time before he was president (e.g. when he was governor of Arkansas). Simple DPs in St'aYimcets lack the characteristics of quantificational DPs, in at least three respects: they do not undergo overt Quantifier Raising to adjoin to the clause, they do not refer to 'old' (presuppositional) information, and they do not allow temporally free readings. I therefore conclude that determiners in this language do not carry quantificational force. 1.4.3. Summary This section has argued that quantifiers do not appear in the determiner position of argument DPs in Salish. A summary of the behaviour of quantificational elements inside argument DPs is given in (54). For evidence that the universal quantifiers adjoin to DP, see Chapters 4 and 5. Quantifier Syntactic realization •air adjoined to D P 'most' * weak quantifiers [ D N P J D P negative main predicate It wil l be necessary throughout this dissertation to bear in mind that quantifiers may exist within DP in positions other than the syntactic position of the determiner. The term 'quantificational determiner' refers to quantificational elements which occupy the determiner position; this set 55 does not include quantifiers which appear elsewhere within DP. The relevance of the syntactic position of the determiner will become more obvious as the discussion proceeds. 1.5. Salish determiners encode assertion of existence We have so far examined three potential determiner contrasts, and seen that all three are missing in Salish. The results obtained are summarized in (55). 55. a. Salish determiners do not encode definiteness. b. Salish determiners do not encode specificity. c. There are no quantificational determiners in Salish. This section addresses the question of what distinctions are made by Salishan determiner systems (other than the deictic and pronominal features discussed in §1.1). The major proposal is given in (56). 56. Salish determiners encode 'assertion of existence'. It wi l l be argued in Chapters 2 and 3 that the assertion of existence distinction which is encoded on Salish determiners is not capturable by currently available theories of the semantics of DPs. Before we define and defend (56), let us look at some data from St'at'imcets. (57) contains the discontinuous determiner ti...a. The DP ti pukwa can be used to describe either a novel or a familiar entity. What both interpretations have in common is the presence of existential force, as indicated semi-formally in (57c). 57. t8xwp-mfn-+kan tecwp-mfh-lhkan buy-appl-lsg.subj [tl pukw-a] [ti pukw-a] [det book-det] +kun§a lhkunsa today b. c. a. 1 bought [a book] today.' 1 bought [the book] today.' 3 x, book (x), I bought x today. (novel) (familiar) (St'at'imcets) 56 The same determiner appears in (58), this time under the scope of the intensional operator kelh 'might'. A s before, the D P ti piikwa may represent either a novel or a familiar entity, but in each case, existential force is involved. 58. t8xwp-min-+kan ka+ [tl pukw-a] natxw tecwp-min-lhkan kelh [ti pukw-a] natcw buy-appl-lsg.subj might [det book-det] tomorrow a. 'I might buy [a book] tomorrow.' (novel) b. 'I might buy [the book] tomorrow.' (familiar) c. 3 x, book (x), I might buy x tomorrow. (St'at'imcets) There are clearly environments where one does not wish to assert the existence of an individual or entity. In these environments, a different determiner (ku) is used, as in (59). This time, the existence of a book is not asserted. The sentence is translatable into Engl ish only with an indefinite determiner, since definites in English usually entail existence (see Chapter 3, H e i m 1982, among others). 3 0 59. t9xwp-mfn-+kan ka+ [kwu pukw] natxw tecwp-min-lhkan kelh [ k u pukw] natcw buy-appl-lsg.subj might [det book] tomorrow 'I might buy [a book] tomorrow.' The determiner ku is restricted in its syntactic distribution. When it appears on argument DPs, it must fal l within the scope of a non-factual operator, such as negation, a yes-no question marker or the modal kelh 'might ' . 3 1 Thus, (60) is ungrammatical (cf. (57)), since the determiner ku cannot be used in a context which induces existential force, such as an ordinary declarative sentence. 60. * taxwp-mfn-+kan tkwu pukw] 4kunSa * tecwp-min-lhkan [ k u pukw] lhkunsa buy-appl-lsg.subj [det book] today 'I bought [a book] today.' 3 0 There are exceptions to the claim that definites entail existence; see the discussion in §1.2.1. 3 1 Ku may also appear on the morphologically unlicensed 'object' of an intransitive verb (i.e. in a D P representing the theme argument, which does not induce pronominal agreement on the predicate). See Chapter 4 for details. 57 The reading represented by (60) is also impossible in English; its interpretation can be paraphrased as T bought a book today, but I do not assert that a book exists that I bought.' A n informal definition of the distinction being encoded here is given in (61). For a formal definition in terms of Discourse Representation Theory (Kamp 1981, Kamp and Reyle 1993), see Chapter 3. 61. Assertion of existence (informal definition):3 2 'the speaker's 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)' (Givdn 1978:293-4). For further illustration, see (62). In the sentences in the left-hand column, the DPs ti sts'uqwaz'a and a fish have existential force; the sentences assert the existence of a fish which Sofie bought. In the sentences in the right-hand column, there is no assertion of existence; the sentences could be true in a world in which fish did not even exist. The difference between the left and right-hand columns is precisely what is encoded by determiner choice in St'at'imcets. 62. Existential force No existential force Az'-en-as [ti sts'uqwaz'-a] kw-s Sophie. buy-tr-3erg [det fish-det] det-nom Sophie Sofie bought [a fish]. 3x, x a fish, Sofie bought x. Cw7aoz kw-s az'-en-as [ti sts'uqwaz'-a] kw-s Sophie neg det-nom buy-tr-3erg [det fish-det] det-nom S. Sofie didn't buy [a fish]. 3x, x a fish, - i Sofie bought x. Cw7aoz kw-s az'-en-as [ku sts'uqwaz'] kw-s Sophie neg det-nom buy-tr-3erg [det fish-det] det-nom S. Sofie didn't buy [a fish], - i 3x, x a fish, Sofie bought x. A DP with an assertion of existence determiner is a description; it does not directly pick out a referent in the real world. I will argue in Chapter 3 that the assertion of existence distinction is not capturable by any currently available theories of the representation of DPs. For example, assertion of existence DPs can be analyzed neither as introducing an existential quantifier into 3 2 (61) is Givon's (1978) definition of 'referentiality'. Giv6n's definition is based on Bemba (Bantu), whose determiner system shows similarity with Salish systems. I do not adopt the term 'referential', in order to avoid confusion with contradictory uses of this term in the semantics literature (see for example Loar 1976, Ludlow and Neale 1991). The phrase 'within a particular universe of discourse' in (61) is somewhat vague; see Chapter 3 for refinement of the definition. 58 logical form (cf. Russell 1905,1919), nor as introducing a variable which is bound by existential closure (cf. Heim 1982). The different ways in which determiner distinctions divide up the possible semantic space in English and in St'at'imcets are summarized in (63-64). English uses the same determiner for all indefinites, whether they receive an existential interpretation or not. St'at'imcets, on the other hand, uses the same set of determiners (those containing an enclitic .. .a)33 for all nominals which induce an assertion of existence, whether novel or familiar. English: novel familiar existential interpretation a the non-exis. interpretation a 64. St'at'imcets: novel familiar assertion of existence X . . . a X . . . a non-assertion of existence ku Some comments are in order regarding (63-64). The shaded areas represent an impossible combination; I assume that an individual which is familiar must be agreed to exist. This is shown for St'at'imcets in (65); the non-assertion of existence determiner cannot be used when describing a familiar individual. 65. texwp k w Mary [tl pukw-a]j tecwp kw Mary [ti pukw-a]i buy det • Mary [det book-exis] 'Mary bought [a book],.' ?ay &u? kw 'ama-S-aS ay t'u7 kw ama-s-as neg just det good-caus-3erg 'She doesn't like books.' 'She doesn't like [the book]*.' [kwu pukw] [ku pukw] [non.exis.det book] (St'at'imcets; LT) 3 3 The enclitic ...a is present on all and only the assertion of existence determiners in St'at'imcets, and is henceforth glossed as 'exis'. 59 For a coreferential reading in the second sentence of (65), an assertion of existence determiner (ti.. .a) must be used.34 Although familiar individuals must be assumed to exist, recall that in English it is possible (in a restricted set of circumstances) to use a definite DP with a non-existential interpretationis uncommon. An example is given in (66). The individual described by the definite DP does not exist yet. 66. I will meet [the first baby to be born in the year 2010]. Examples similar to (66) must be rendered with a non-assertion of existence DP in St'at'imcets, in accordance with the non-existence of the individual in present time. 67. xwuz-+kan malyf-S [kwu xwuz" k w uk w pir l a W Fount.] cuz'-lhkan mely'i-s [ku cuz' kukwpi7 l£ku7 Fountain] going.to- lsg. sub marry-caus [non.exis.det going.to chief deic Fountain] 'I will marry the next chief of Fountain.' (whoever it is) (St'aYimcets; LT) We see that although definites in English usually have an existential interpretation, the requirement for assertion of existence DPs in St'at'imcets is stronger. Assertion of existence DPs assert existence, while definite DPs presuppose existence (and are subject to accommodation). Hence, definites in English can be used in some contexts in which an assertion of existence DP is inappropriate. See Chapters 2 and 3 for further discussion. While Salishanists have not previously used the term 'assertion of existence' in their descriptions of determiner systems, similar facts appear to hold in a number of other languages, including Sechelt (Central; Beaumont 1985), Bella Coola (Davis and Saunders 1975), Secwepemctsin (Northern Interior; Kuipers 1974) and Halkomelem (Central; Galloway 1993). The restriction of non-assertion of existence determiners to the environment of a non-factual operator also holds in these languages. In the remainder of this section I will provide evidence that assertion of 3 4 For one principled exception to the claim that ku-DPs may not corefer with subsequent DPs, see the discussion of modal subordination in Chapter 3, §2.3. 60 existence is relevant for determiner choice in these languages, and also discuss a language where assertion of existence does not appear to be present, namely Straits. 1.5.1. Assertion of existence in Sechelt The determiner she in Sechelt is glossed as 'unspecified, abstract' by Beaumont (1985) (and as 'non-actual' by Beaumont 1980). The distinction between she and the rest of the determiners is characterized as follows (Beaumont 1985:53, emphasis original). When the speaker of Sechelt talks about something he first determines whether or not it is something 'real' (or 'actual'). If it is an actual thing, being (or even an action) that he can or could identify and point out, he will use [determiners other than she]. ... If the speaker is referring to something that is not 'real' ('non-actual'), that is, something that he can or could not (or doesn't want to) identify specifically in a physical sense, he uses she. A n analysis of the split between all other determiners and she as an assertion of existence distinction accords with the facts. The example in (68a) is taken from a story where the snake-woman has not previously been mentioned, and thus is novel (indefinite). However, she is known to the speaker, and is asserted by the speaker to exist. Consequently, the assertion of existence feminine determiner lhe is used. (68b) and (68c) are contiguous utterances from the same story; here, the first mention of a cloud uses the non-assertion of existence determiner, since no cloud has materialized yet. The second reference to the cloud (68c) uses the assertion of existence determiner. 68. a. t'i suxwt-as [lhe 7tilhka7 slhanay] fact saw-he [exis.det snake woman] 'He saw a snake-woman.' (Sechelt; Beaumont 1985:188) b. t'i tl'um s-7ut-s [she ts'amkwelh] fact then calling-her [non.exis.det cloud] 'Then she called a cloud.' (Sechelt; Beaumont 1985:191) c. kern t'i tl'um s-kw&l'-s [te ts'amkwelh] and fact then coming-its [exis.det cloud] 'and then the cloud came' (Sechelt; Beaumont 1985:191) 61 The presence of she in (68b) highlights an important feature of the non-assertion of existence determiners. Such determiners fail to assert that an individual exists, but they do not assert that an individual does not exist. For example, (68b) does not entail that no cloud exists. Rather, the non-assertion of existence determiner is used because no cloud has yet been witnessed, and therefore the existence of a cloud cannot definitively be asserted. The fact that a cloud appears in the next sentence, and therefore clearly exists, does not at all contradict the use of she in (68b). Non-assertion of existence she is further illustrated in (69), along with Beaumont's explanation of the determiner choice. 69. a. ne sxaYi' [she xeyek'] my wish [non.exis.det crab(meaf)] 'I want some crab (meat).' (Sechelt; Beaumont 1985:147) b. 7e sxatl'-d [she s7ulhku] your wish-int [non.exis.det clams] 'Do you want some clams?' (Sechelt; Beaumont 1985:147) c. stam [she skwalish-it] what [non.exis.det names-their] 'What are their names?' (Sechelt; Beaumont 1985:147) Note the use of she 'a, some' in [69a,b]. This 'non-actual' form ... is used because the crabmeat and clams have not yet materialized. That is, they are not yet 'real' for the speaker. Similarly, she is used in [69c] because the names are not known to the speaker (Beaumont 1985:147). Again, the speaker is not asserting that crab meat, clams, or the people's names do not exist, but only that the speaker does not have personal knowledge that they exist. In other words, for a speaker to be able to assert the existence of an entity in a Salish language, the speaker must have personal knowledge of that entity. The speaker-oriented nature of assertion of existence will become relevant in Chapter 3, where I argue that Salish determiners, unlike English determiners, can access only speaker-oriented distinctions. See also the discussion of Bella Coola determiners immediately below. 62 Notice that the environments in which she appears in Sechelt include under the intensional predicate 'want' (69a), in a yes-no question (69b), and in a w/i-question (69c). These environments correspond to the restricted range of environments permitted by non-assertion of existence DPs in St'at'imcets.35 If she is a non-assertion of existence determiner corresponding to St'aYimcets ku, this yields the following reanalysis of the Sechelt system. 70. Sechelt determiners (cf. Beaumont 1985): assertion of existence non-assertion of existence visible invisible +female -plural tse, lhe lhe she +plural te te, che -female -plural -i-plural 1.5.2. Assertion of existence in Bella Coola Bella Coola has a complex system of prefix-suffix combinations, repeated in (71) (see also Nater 1984:41-8).36 3 5 Beaumont (1980) provides some instances of she without a non-factual operator present. For some reason, the body parts in (i) receive she. I do not have an explanation for this phenomenon. i . a. X9+ §9-n qanxw sore det-lsg.poss throat 'My throat is sore.' (Sechelt; Beaumont 1980) b. se'ssxsa'x §a-n wtynwas creep det-lsg.poss flesh 'My flesh creeps.' (Sechelt; Beaumont 1980) 3 6 Nater (1984) characterizes the determiner elements in (71) as proclitics and enclitics, rather than prefixes and suffixes. 63 71. Bella Coola determiners (Davis and Saunders 1975:14): proximal non-proximal proximal space, middle space, distal space, present time near past/present distant past I II i n IV V VI -plural +female C1...CX c1...cayx 4a-?Ha?1+ 4a...+ +a...?H +a...?U--female t1...tx t1...tayx ta-tax* ta...+ ta...tx ta...tax +plural wa...c wa...?ac ta-tax w ta...+ ta...txw ta...tux The prefix-suffix pairs may be used in combination (bracketing a noun), or either the prefixes or the suffixes may appear alone. An example of a suffix standing alone as demonstrative pronoun is given in (72).37 72. Kap tayx go II "This one is going.' (Bella Coola, Davis and Saunders 1975:23) The proximity distinctions are as follows (Davis and Saunders 1975:15). 73. I visible, near conversation II visible, closest to conversation i n as far as vision extends IV invisible (but may be quite near) V, VI invisible, beyond III and IV The prefixes and suffixes 'place the referents in relative space with respect to the space of the speech act' (Davis and Saunders 1975:15). Temporal as well as spatial proximity may be entailed. For example, if a speaker has witnessed an event in invisible space (i.e. a space which is invisible to the current speech act), he or she may use an affix of proximity V or VI. Since the event took place at a location distant from the speech act, the speaker must have travelled before making the utterance. Therefore, the event must have taken place a while ago. Hence, the deictic features of the determiners encode 'distant past' as well as 'distal space'. 3 7 A l l the suffixes can also attach directly to predicates, with deictic spatio-temporal meanings. I do not discuss this phenomenon here; see Davis and Saunders (1975:18ff). 64 A n example with two argument DPs is given in (74). 74. kx - l s [t1-?1mlk-tx] [+a-xnas-?H] see-3sg [I-man-I] [V-woman-V] "The man [visible] saw the woman [invisible].' (Bella Coola; Davis and Saunders 1975:17) (74) can be used when the man is presently visible (i.e. in the space of I), and the woman is in the space of V (i.e. invisible). Since the man and the woman must have been in the same area at the time the event of seeing took place, time must have elapsed between the event of seeing and the speech act (time during which the man and the woman moved apart from each other). Hence, the sentence is interpreted as 'past'.38 A n important feature of Bella Coola is summarized in (75). 75. Any declarative sentence implies that the SPEAKER WITNESSED THE EVENT. This requirement ensures that (74) cannot be interpreted as 'distant future'. The sentence is declarative, therefore the speaker witnessed the event, therefore the event must have already taken place. If a declarative sentence implies that the speaker witnessed the event described, it follows that the participants in a declarative sentence must all have existential force (since for the speaker to have seen them, they must exist). We can therefore predict that any determiners in Bella Coola which are only possible in non-declarative sentences will be good candidates for non-assertion of existence determiners. 3 8 (74) has other possible interpretations, according to contextual factors which can 'neutralize' either the distal space or the distant past components of the V affixes. The brief summary of the system offered here cannot do justice to its complexity. See Davis and Saunders (1975) for details of how the spatial and temporal information interact with each other in Bella Coola determiners. 65 It turns out that there is such a determiner in Bella Coola. Notice in (71) that the prefixes (unlike the suffixes) encode only a two-way proximity distinction (I-II vs. Il l-VI). When the prefixes appear in combination with a suffix, this two-way distinction can be called 'proximal' vs. 'non-proximal'. However, when the prefixes appear alone without the suffixes, the distinction looks more like non-assertion of existence vs. assertion of existence. Davis and Saunders' description is as follows. The function of Proximal deictic prefixation is to mark a referent the speaker has not witnessed. Recalling that deictic suffixation specifies the relative spatio-temporal cirucmstance of that witnessing, the absence of deictic suffixation coincides with the absence of witnessing. Hence [76]. * ksnmak t1-?1mlk [work I-man] is unacceptable because declarative utterances presuppose ... speaker witness; but this contradicts t l ?1mlk, that expresses the claim the speaker has never seen the man (Davis and Saunders 1975:31). The 'proximal' prefixes when used alone are ungrammatical in any ordinary declarative sentence, only being acceptable in questions, negatives and in the presence of certain clitics (denoting that the speaker obtained the information from a third person, is making a conjecture, or expressing an inference). In contrast, a 'non-proximal' (III-VI) prefix when used without a suffix does not imply that the speaker has never seen the referent, but only indicates the absence of the referent (Davis and Saunders 1975:32). Predictably, declarative sentences containing these prefixes alone are possible. This situation is consistent with the following hypothesis. The 'non-proximal' prefixes (along with all the deictic suffixes) encode assertion of existence. The 'proximal' prefixes, when used alone, mark non-assertion of existence (and are subject to parallel restrictions to those on non-assertion of existence determiners in St'at'imcets and Sechelt). This exactly corresponds with Davis and Saunders' description of them as marking lack of speaker witness. 66 Further evidence for this analysis conies from the minimal pair in (77). The question environment licenses both the assertion of existence prefix-suffix combination in (77) and the non-assertion of existence prefix used alone in (77b). 77. a. waks tl-ftap-tx who I-go-I 'Who is the one going?' literally: 'Who is the one (male) observed now in Proximal space [who is] going?' (Bella Coola, Davis and Saunders 1975:30) b. waks tl-ftap who I-go 'Who is going?' literally: 'Who is it (male), whom I have not seen [who is] going?' (Bella Coola, Davis and Saunders 1975:30) The use of a proximal prefix without an accompanying suffix, as in (77b), indicates that the speaker does not know the identity of the individual, and has never seen that individual. 1.5.3. Assertion of existence in Secwepemctsin The Secwepemctsin system contains a contrast between 'actual-determinate' vs. 'hypothetical-indeterminate', according to Kuipers (1974) (see (9) above). Kuipers' description of the facts is as follows (1974:57): The distinction actual-determinate versus hypothetical-indeterminate hinges on whether the thing meant is uniquely determined for the speaker - in the case of a physical object: whether he could conceivably point it out. ... Hypothetical articles are found especially in interrogative, imperative and conditional sentences, in negative sentences, and in sentences referring to the future. Examples of typical usages are given in (78). 78. a. w l k - t - s [y qwanfrrt8q+] see-tr-3erg [actual.det mosquito] 'He saw the mosquito.' (Secwepemctsin; Kuipers 1967:98) b. ta'? tk qwanfmaq+] neg [hyp.det mosquito] 'There are no mosquitos.' (Secwepemctsin; Kuipers 1967:97) 67 These facts parallel those of St'at'imcets, Sechelt and Bella Coola, and I therefore reinterpret the actual-deteterminate / hypothetical-indeterminate distinction as encoding assertion of existence / non-assertion of existence. 79. Secwepemctsin determiners (adapted from Kuipers 1974:57): assertion of existence non-assertion of existence present absent absolutive Y 1 k relative t / t tk/ tk (seldom tke?) 1.5.4. Assertion of existence in Halkomelem (Chilliwack dialect) Chilliwack encodes gender, number and proximity distinctions (Galloway (1993:386). The determiner which interests us in this section is k w a, which is glossed as 'remote, distant (and not visible), abstract (or hypothetical), indefinite, generic, or past'. What is interesting is that k w a appears in the exact contexts which support non-assertion of existence determiners in the other languages looked at above. Galloway's examples of kw8 involve the following sentence-types.39 80. Sentence-types which contain kwa: a. yes-no questions: if (ya) q & [kwa s1ya+] if kwe l l l a m prep (pl) many [det wood] prep det house 'Is there a lot of wood at your house?' (Chilliwack; Galloway 1993:388) b. w/i-questions: s t l m [kwe s*l1 what [det want] 'What do you want?' (Chilliwack; Galloway 1993:389) 3 9 Strang Burton informs me (p.c.) that the determiner kw9 is not used by his Chilliwack consultant. Further research is necessary into the Chilliwack determiner system. 68 c. intensional verbs: 1 s*f tkw9 qa":] lsg.poss want [det water] 'I want (some) water.' (Chilliwack; Galloway 1993:389) Galloway's description of the semantics of k w a, and the sentence-types where it appears, appear to identify it as the non-assertion of existence determiner (although see footnote 37). 1.5.5. The absence of assertion of existence in Straits Determiners in Straits do not seem to encode assertion of existence. The Lummi dialect has the determiner system in (81), according to Jelinek and Demers (1994) (repeated from (7) above). 81. Lummi determiners (Jelinek and Demers 1994:717): proximate, visible neutral distal/out of sight remote +female sl'a S9 kwe kWS9 general t l 'a C9 kW9 kWC9 It is not clear which column in (81) would correspond to non-assertion of existence, and Jelinek and Demers give no indication that determiners in any one column have a restricted distribution. Eloise Jelinek (p.c.) confirms that all the determiners have unrestricted distribution, and Timothy Montler (p.c.) also confirms that there are no determiners in Straits which behave like non-assertion of existence determiners in the other Salish languages. Montler observes that throughout the Northern Straits dialects, invisible determiners contrast with determiners which are unspecified for visibility. The invisible determiners contain kw. The invisible determiners can replace any visible determiners, as shown by pairs such as in (82). 82. a. k w9nn9X w S9n ttS9 swgyqg?] see I [visible.det man] 'I see a / the man.' (Saanich; T. Montler, p.c.) 69 b. kwennexw sen [kwse sweyqe?] see I [invisible.det man] 'I saw a / the man.' (Saanich; T. Montler, p.c.) If kwse were a non-assertion of existence determiner, its use in a declarative sentence like (82b) would be impossible. Since (82b) asserts that an event of seeing took place, the discourse referent corresponding to the man who was seen must be asserted to exist. The absence of an assertion of existence distinction in Straits could be related to an independent difference between Straits determiners and determiners in other languages of the family. Straits determiners may stand alone as demonstratives, without an accompanying nominal, as shown in (83). 83. lerj-t-0 sen kwse see-tr-3abs lsg.nomin det 1 saw her, that one.' (Lummi; Jelinek and Demers 1994:717) This is impossible in languages like St'at'imcets, where the determiners are clitics, and must co-occur with their complements, as shown in (84).40 84. * ?a£x-en-+kan t1(...a) * £ts'x-en-lhkan ti(...a) see-tr-lsg.subj det(...exis) T saw him/her.' (St'at'imcets; LT) The ability of Straits determiners to stand alone leads Jelinek and Demers to call them 'determiner/demonstratives'. The Straits determiners are clearly all strongly deictic in nature. This is related to the absence of an assertion of existence distinction in the following way. In the other Salish languages discussed above, it is the assertion of existence determiners which distinguish deictic notions such as proximity. Within the non-assertion of existence set, deictic features are not distinguished. For example, St'at'imcets possesses just one non-assertion of 4 0 According to Montler (1986:224), there is a subset of determiners in the Saanich dialect of Straits which may not serve as demonstrative pronouns. Saanich is still differentiated from languages like St'at'imcets, where none of the determiners may function as demonstratives. 70 existence determiner, ku, and Sechelt also possesses just one, namely she. Since all determiners in Straits are deictic, it makes sense to assume that in Straits, all the determiners assert existence. This claim accords with Dernirdache's (1996a,b) analysis of assertion of existence determiners as introducing 'stages' of an individual (in the sense of Carlson 1977). According to Demirdache, the deictic features on the determiners locate an entity in time and space, thus supplying it with SPATIO-TEMPORAL BOUNDARIES. Once there is a space/time location, the individual must exist. This follows directly from Carlson's (1977) equivalence between 'existence' and 'having a stage', where a 'stage' means a spatio-temporal realization of an individual. For Carlson, to exist means to be located in space and/or time; hence the correlation between deixis and assertion of existence in Salish. Conversely, a non-assertion of existence determiner does not provide an entity with spatio-temporal boundaries (Demirdache 1996a,b). Dernirdache's analysis predicts that a non-assertion of existence determiner cannot contain deictic features. If Straits determiners are really deictic demonstratives, the absence of an assertion of existence distinction in this language will follow straightforwardly. Having discussed definiteness, specificity and assertion of existence in some detail, we can now elucidate how the three notions interact with each other. A rough guide to the combinatorial possibilities is given in (85). 1.6. Conclusions 85. + Assertion of existence - Assertion of existence + Specific - Specific - Specific + Definite - Definite - Definite - Definite 71 One of the important details glossed over in (85) is the relation between definites and assertion of existence DPs. Although (85) suggests that the class of definites is a subset of the class of assertion of existence DPs, this is not strictly correct, as discussed above with reference to individuals which do not yet exist, but can be described by definite DPs in English (see (66) above). (85) predicts that assertion of existence DPs can be either specific or non-specific. Assertion of existence DPs which are non-specific occur when the speaker believes in the existence of a unique individual, but may not be able to actually identify that individual, i.e. 'he may not know who the person actually is' (Givon 1978:313). Example (34) in §1.3 above showed that non-specific assertion of existence DPs do exist in St'at'imcets. Another example is given in (86). 86. a. Context: Beverley sees John swearing at two white men in the street. He hits one of them, but Beverley doesn't see which one. She tells someone: b. t u p - u n - a S S - J o h n [ t l Sam?-a] tup-un'-as s-John [ti sam7-a] punch-tr-3erg nom-John [det white.person-exis] 'John hit a white man.' (St'at'imcets; BF) Beverley knows that the white man exists, because she saw him. She uses an assertion of existence determiner to indicate this. However, she does not know which particular white man John hit, and therefore means to express only a general proposition (i.e. a proposition which does not involve a directly referring expression). This is schematized according to Ludlow and Neale's terminology in (87). 87. Speaker's Grounds: John hit a white man. Proposition Meant: John hit a white man. Proposition Expressed: John hit a white man. There is no mismatch between the Speaker's Grounds and the Proposition Meant (which is the diagnostic for a specific use). An assertion of existence non-specific can appear in contexts (such as (86)) which lack a non-factual operator, as is predicted, since it is only non-assertion of existence DPs which are disallowed in such contexts. The patterning of all assertion of existence 72 DPs together in the absence of a licensing requirement is further evidence that the relevant distinction in Salish is one of assertion of existence rather than specificity. So far we have seen that none of the languages looked at marks a definiteness contrast in their determiner system. A l l except Straits and possibly Upper Chehalis mark an assertion of existence contrast. The results are summarized in (88). 88. deixis definiteness specificity assertion of existence Sechelt V * * V St'at'imcets * * Secwepemctsin * * Bella Coola * * Straits V * * * Upper Chehalis V * ? ? Non-assertion of existence determiners are differentiated from assertion of existence determiners syntactically, in being subject to a licensing restriction such that they can only appear in the environment of a non-factual operator. In addition, we have seen that non-assertion of existence determiners may not encode deictic features. The syntactic restriction on the distribution of non-assertion of existence DPs is not limited to Salish. Giv6n makes the cross-linguistic claim that nominals falling under the scope of a possible modality or negative modality may receive either a referential (i.e. assertion of existence) or non-referential interpretation; otherwise all nominals are interpreted referentially (1978:294). This is the case in Bemba as well as in Salish. 2. The syntactic distribution of determiners This section turns from the semantics to the syntax of determiners in Salish. Before investigating the Salish data, I briefly outline a theoretical proposal about the function of determiners which makes predictions about their syntactic distribution. 73 2.1. Determiners are operators which bind a variable Higginbotham (1983, 1985, 1987) claims that determiners function to bind the external theta-role of NPs, saturating the NPs so that theta-assignment can take place. Stowell (1989, 1993) also claims that determiners bind the external arguments of NPs, closing off the nominal predicates. He states that 'Whenever a [common noun phrase] functions referentially, it must have a determiner, since NP cannot iself refer' (1989:259). Longobardi (1994) argues in a similar vein that determiners are Operators which bind a variable inside NP. According to Longobardi, the range of the variable is the extension of the kind referred to by the head noun (1994:633). The semantic analysis of nominals runs as follows (Longobardi 1994:648): the N position is interpreted as referring to universal concepts, that is, to kinds; the D position, instead, determines the particular designation of the whole DP, either directly, by being assigned reference to a single individual object [in the case of proper nouns or pronouns (LM)], or indirectly, by hosting the operator of a denotational (operator-variable) structure. The specific readings of common nouns are all obtained by letting the variable bound by the operator (whether lexically present or understood) in the D position range over the extension of the kind referred to by the N position. Determiners are crucial for argumenthood: 89. a. A 'nominal expression' is an argument only if it is introduced by a category D (Longobardi 1994:620) b. NP can only be licensed through a predicative interpretation (Longobardi 1994:628) Evidence for the argument-creating function of determiners comes from Romance and Germanic (including English). In these languages, articles are not obligatory on non-arguments (predicates, vocatives, and exclamations), but are obligatory on arguments except in a restricted set of cases, which Longobardi analyzes as containing a null determiner (hence, the D can be either lexically present or simply 'understood'). The saturation effect of a determiner is illustrated in (90) for 74 English. The NP predicate woman cannot function as an argument by itself; it requires a determiner before theta-assignment can take place. 90. a. * [Woman] laughed. b. [The woman] laughed. The empty D in Western Romance is illustrated in (91); it instantiates an existential Operator (Longobardi 1994:617). 91. Mangio patate I eat/am eating potatoes 1 eat / am eating potatoes.' (Italian; Longobardi 1994:613) 3 x, potato (x), I eat (x) This analysis of Western Romance is extended to the universal claim in (92). 92. [D e] = default existential interpretation (Longobardi 1994:641) Longobardi's analysis means that the DP structure summarized in (93) is obligatory for arguments. He notes that 'Once we adopt [93], the natural way of reformulating the content of principle [89a] becomes the following ... DP can be an argument, NP cannot' (1994:628). 93. [DP [D1 D NP ]] (Longobardi 1994:610) Longobardi explicitly argues for the correctness of (93) only for Romance and Germanic. We wil l see evidence in the remainder of this section that determiners in many Salish languages are consistent with a Longobardi-style analysis, whereby determiners are operators which bind a variable, creating an argument DP. 2.2. The syntactic distribution of determiners in Salish Consistent with approaches in which the presence of a determiner correlates with argumenthood (Higginbotham 1985, Stowell 1989, Longobardi 1994), the distribution of Salish determiners is sensitive to syntactic position. Salish determiners show different behaviour according to whether 75 they are associated with nominal projections in argument position, in main predicate position, or in adjunct position. These three syntactic environments will now be discussed in turn. 2.2.1. The distribution of determiners in argument positions This section examines DPs which function as arguments of the main predicate (e.g. subjects and objects, including both nominal and clausal arguments), arguments which appear inside noun phrases (i.e. possessor DPs), and argument DPs contained within adjunct phrases. St'at'imcets is examined in the most detail, followed by brief remarks about Sechelt, Straits, Halkomelem, Bella Coola and Upper Chehalis. We will find evidence to support Kroeber's (1991:27) claims about the distribution of determiners on argument phrases: Noun phrases (NPs) throughout the family are normally introduced by a determiner (article or demonstrative). In some but not all languages proper nouns do not need to be preceded by a determiner, and certain other exceptions to the generalization occur in Southern Interior languages ... but in general NPs are overtly delimited constituents. St'at'imcets. Determiners are always obligatory on arguments in St'at'imcets, as shown in (94) and (95) for subjects and objects respectively. I have used examples which contain bare nouns in English, to bring out the contrast between the languages. The requirement for a determiner holds for all noun-types (proper, common, count, mass) in argument position in St'at'imcets. 94. a. wa? £aqw-an-ftas" [?1 *aVa] [?1 mfxa+-a] wa7 ts'aqw-an'-ftas [i t'ec-a] [i mfxalh-a] be eat-tr-3pl.erg [pl.det sweet-exis] [pl.det bear-exis] 'Bears eat honey.' (subject) (St'at'imcets; GN) b. * wa? caq w -an-f ta§ [?1 fcaVal [mfxa+] * wa7 ts'aqw-an'-ftas [i t'ec-a] [mixalh] be eat-tr-3pl.erg [pl.det sweet-exis] [bear] 'Bears eat honey.' (subject) (St'at'imcets; LT) 95. a. qw9n-an-+kan [kwu Sqlaw] qwen-an-lhkan [ku sqlaw'] need-tr-lsg.subj [non.exis.det money] 1 need money.' (object) (St'at'imcets) 76 b. * qwan-ln-+kan [Sqlaw] * qwen-ah-lhkan [sqlaw'] need-tr-lsg.subj [money] 'I need money.' (object) (St'aYimcets) Consistent with this, a determiner is obligatory with proper names when they appear in argument position. In (96a-b), the name forms a subject and object respectively. 96. a. ?acx -9n -C -a§ aYs'x-en-ts-as see-tr-1 sg.obj-3erg 'Rose saw me.' [k w -§ Rose] [kw-s Rose] [det-nom Rose] (subject) (St'at'imcets) ?acx-9n-+kan ats'x-en-lhkan see-tr- lsg.subj 'I saw Rose.' [kw-S Rose] [kw-s Rose] [det-nom Rose] (object) (St'at'imcets) Clausal arguments also obligatorily require a determiner, as shown in (97). 97. a. k a t*(t) S-*1q-Su-*(a)l ama [*(t) s-t'iq-su-*(a)] good [*(det) nom-arrive-2sg.poss-*(det)] 'It's good that you came.' ("That you came is good.') (St'aYimcets; LT) Determiners are obligatory on a possessed nominal in argument position, as shown in (98): 98. *1q ttl kwukwp1?-a t'iq [ti kukwpi7-a arrive [det chief-exis 'The chiefs mother arrived.' t l Skfxza?-S-a] ti skicza7-s-a] det mother-3sg.poss-exis] (St'aYimcets) b. * * lq [tl kwukwp1?-a * t'iq [ti kukwpi7-a arrive [det chief-exis 'The chiefs mother arrived.' Skfxza?-s1 skicza7-s] mother-3sg.poss] (St'dt'imcets) c. ^1q Ikwukwp1? t l [kukwpi7 ti [chief det The chiefs mother arrived.' * * t'iq arrive Skfxza?-S-a] skicza7-s-a] mother-3sg.poss-exis] (St'at'imcets) 77 St'at'imcets contrasts with English in this respect; in English, a possessed nominal may not have its own determiner.41 s 99. a. * [The chiefs the mother] arrived, b. [The chiefs mother] arrived. Aguments inside manner or location adjuncts also require a determiner, as shown in (100). 100. a. wa? q w 9Z - f l x [tl §mlm+ac-a] latl? tcfla *(k wu) Sama?] wa7 q'wez-flc [ti smem'lhats-a] lati7 [ts'fla *(ku) sama7] prog dance-body [det woman(redup)-exis] deic [like *(det) white] "That girl is dancing like a white person.' (St'at'imcets; RW) b. wa^H-kan xa^-mln ka ' Icx-am-a +-a§ SltSt wa7-lhkan xat'-min' ka dts'x-em-a lh-as sitst, prog-lsg.subj hard-appl ooc see-intr- ooc when-3sg.conj night [xjl-sm *(k wu) Skalula?] [xfl-em *(ku) skalula7] [do-intr *(det) owl] T would like to be able to see in the dark/at night, like an owl.' (St'at'imcets; RW) 101. c1xw-kan ?acx -9n [?1 wa? cunam-xall tsicw-kan ats'x-en [i wa7 tsunam'-cal] go-lsg.subj see-tr [pl.det prog teach-intr] [l-'*(kD takem-Ma) Skull [l-*(ki) takem-*(a) skul] [in-*(pl.det) all-*(exis) school] T visited teachers in every school.' (St'at'imcets; RW) Sechelt. Beaumont (1985) does not state a generalization about the obligatoriness or otherwise of determiners on arguments in Sechelt. However, perusal of the texts and sentences he provides reveals no instance of a missing determiner. As shown in (102), a determiner is necessary in Sechelt where English allows a bare plural: 102. 7ut-chexw kumut [she xeyxeyeT], we pepe-iwan-axw,... if-you eat [det crabs], when growing-belly-you,... 'If you eat crabs when you are pregnant,...' (Sechelt, Beaumont 1985:201) 4 1 Some consequences of this difference for the structure of DP in St'at'imcets are discussed in Matthewson and Davis (1995). See also Chapter 4, §4.1 on possessors in St'at'imcets. 78 Straits. Determiners are obligatory on arguments in Straits; the language 'has no nominals that are not under the scope of one of the demonstratives. Demonstratives are not optional constituents of nominals in Straits Salish' (Jelinek and Demers 1994:718).42 103. rjey led swayqa'] work [det man] "The man works.' (Straits; Jelinek and Demers 1994:718) Halkomelem. Galloway (1993:386) states that in Chilliwack, determiners are 'obligatory before nominals'. Some examples are given in (104). 104. a. m6yea'-xy-as [K Bi l l ] help-me-3erg [det Bill] 'Bi l l helped me.' 1 s*f Ikw8 q&] lsg.poss want [det water] 'I want (some) water.' (Chilliwack; Galloway 1993:390) (Chilliwack; Galloway 1993:389) Bella Coola. In Bella Coola, overt determiners can be missing in certain restricted environments. There is a zero variant of the plural proximate article wa (see (5) above). This zero variant appears before the hypothetical proclitic ka.43 105. 7alhi-a 7ala-7awcwa 0-ka-tsaatsaws is.there here 0-hyp-church 'Is there a church here?' (Bella Coola, Nater 1984:47) The 'zero variant' of the plural proximate determiner in Bella Coola is the one systematic exception to the obligatoriness of determiners inside argument DPs. Since the zero form paradigmatically contrasts with overt determiners, I analyze it as Nater (1984) does, namely as a null variant, rather than as the absence of a determiner. 4 2 Recall that determiners are equivalent to demonstratives in Straits (§1.5.5 above). 4 3 Nater also notes that 'Proper names and geographical names are often found without an article (due to English influence?)', for example nuxalk 'Bella Coola' (Nater 1984:42). He does not give an example of a proper name appearing in argument position in a sentence without an article. 79 Upper Chehalis. Determiners are occasionally missing from arguments in Upper Chehalis, but are otherwise so pervasive that in the instances in which they are missing, it may be due either to transcription error by Boas, or to a phonological deletion process during fast speech (M.D. Kinkade, p.c.). 4 4 In summary, in all the Salish languages discussed here, determiners are obligatory on arguments. Overt determiners appear on all arguments in St'at'imcets, Sechelt, Straits, and Halkomelem. In Bella Coola, there is a paradigmatically contrasting zero determiner, and in Upper Chehalis, phonological deletion processes obscure the grammatical requirement for determiners on argument DPs . 4 5 2.2.2. The lack of determiners in (main) predicate position Determiners are not required on main predicates in most Salish languages. (106) shows that no determiner is present on nominal predicates in St'aYimcets or Secwepemctsin. 106. a. Smu+ac-kan smulhats-kan woman-lsg.subj 'I am a woman.' (St'aYimcets) b. qlmux w -kn Indian-lsg.subj 'I am an Indian.' (Secwepemctsin, Kuipers 1974:79) In (107a), a common noun is the main predicate, while in (107b,c), a proper name forms the main predicate. 4 4 This is true also in other languages such as N + e ' k e p m x c f n , where the absence of determiners on arguments is due to phonetic deletion processes, and determiners should be regarded as syntactically present (Kroeber 1994b). 4 5 As Kroeber (1991) notes, there may be exceptions to the obligatoriness of determiners in the Southern Interior languages. M . D . Kinkade (p.c.) confirms this for Columbian (Southern Interior). 80 107. a. kwuk wp1? kukwpi7 [tl Sqayxw-a] [ti sqaycw-a] chief [det man-exis] "The man is a chief.' (St'at'imcets; A A , LT) John [ta kwukwp1?-al John [ta kukwpi7-a] John [det chief-exis] 'The chief is John.' (St'at'imcets; LT) c. Jimmy [che-n skwish] Jimmy [det-my name] 'Jimmy is my name.' ('My name is Jimmy.') (Sechelt; Beaumont 1985:15) (108) shows that complex noun phrases can also function as predicates, again without determiners present. (108a) contains a possessed noun as the main predicate, and (108b) a modified noun. 4 6 108. a. [7alesh-s te skw'etu7] [lhe 7£sxw] [sister-his det raven] [det seal] 'Raven's sister was Seal.' (The seal was Raven's sister.') (Sechelt; Beaumont 1985:181) b. [may stumish] [te skw'etu7] [bad man] [det raven] 'Raven was a bad man.' (Sechelt; Beaumont 1985:185) Equative constructions, which involve determiners on predicates in English, are not possible in St'at'imcets. Nominals are always determinerless when used as predicates. 109. a. * t l kwukwp1?-a * ti kukwpi7-a det chief-exis "That man is the chief.' [tl? t l 3qayxw-a] [ti7 ti sqaycw-a] [deic det man-exis] (St'at'imcets; LT) Jelinek (1993a) also argues that equative constructions are impossible in Lummi (Straits, Central). This is shown in (110). 110. a. s l ' e m - s x " chief-2sg.subj 'You are a chief.' (Lummi; Jelinek 1993a:5) Compare (108a) with (98a) above, which contains a possessed nominal in argument position. 81 b. * ca s1'em-sxw det chief-2sg.subj 'You are the chief.' (Lummi; Jelinek 1993a:5) Upper Chehalis and Cowlitz provide exceptions to the claim that main predicates do not take determiners. In the perfective aspect, determiners appear on predicates in these languages. The perfective marker in (111) is homophonous with the determiner tit. 111. . . . w l t i t w u c l - t - m ta(t) s -x w ay-s [tit qa':?]... ... and perf wring.out-tr-pass det nom-urinate-3poss [det water] "... (and when Moon was kidnapped,) the water was wrung out of his diaper,... ' (Upper Chehalis; M . D . Kinkade, p.c.) I do not have an explanation for the use of determiners to mark perfective aspect in these languages. However, there is independent evidence that determiners in Salish take over part of the function which is performed by verbal functional projections in other languages. Demirdache (1996a,b) argues that in St'aYimcets, determiners perform part of the function which belongs to Tense in English; see also Davis and Matthewson (1996a,b), §2.2.3 below. The Upper Chehalis/Cowlitz perfectives may represent a similar phenomenon. Apart from in the perfective aspect, the Upper Chehalis/Cowlitz system follows the general Salish pattern. In particular, nominals which function as main predicates do not require determiners, as shown in (112). The nouns spata'ln 'rock' and s+a'nay 'woman' are the predicates of their respective clauses. 112. spata'ln t1tx.tf. wl huy c£n1 ukwa wl s+a'nay rock this and then (s)he I.guess cop woman "This is a rock. But she is a woman.' (Upper Chehalis; M . D . Kinkade, p.c.) 82 2.2.3. Determiners with temporal nouns in adjunct positions There is one environment (apart from main predicates) where determiners are systematically absent on nominal phrases in Salish. This case involves quantified temporal adjuncts, of the form 'every day' (or 'all days'). Examples are given in (113) from several different languages.47 113. a. [zi^zeS' sqlt l k w - s ?a*cx -an-an [zi7zeg' sq'it] kw-s dts'x-en-an [each day] det-nom see-tr-lsg.conj T saw that person every day.' (St'at'imcets; RW) b. [takam S-lap] kw -n-S-wa mamtaq [takem s-gap] kw-en-s-wa mam'teq [all nom-evening] det-lg.poss-nom-prog walk(redup) T go for a walk every evening.' (St'at'imcets; RW) c. [i7xw skwayel] kwis ne-s wa7ew wa mfkw'entsut [all day] det-nom rl-3poss continue prog bathe 'He bathed every day ...' (Squamish; Demirdache et al. 1994) d. [mekw kwscf]] ?1 aw kwawaykw ttsa s - p a a l ] [all day] accom contem fishingfacf] dem s-Raven 'Every morning he'd go out fishing, the Raven.' (Saanich; Montler 1986:242) These [Quantifier NP] constituents are analyzable as in (114). The universal quantifier is adjoined to a phonetically empty DP range, and the temporal noun phrase functions as an adjunct (i.e. 'every one of them, for days'). 4 7 The enclitic ...a is missing in the St'at'imcets temporal adjuncts in (113). In argument position, the proclitic portion of the determiner may phonetically delete in fast speech, but the enclitic portion never deletes. This is evidence that there is truly no determiner (other than possibly the quantifier) in (113a,b). 83 The ability of a universal quantifier to adjoin to a null DP is attested elsewhere in the languages, as discussed in Chapter 5 and as shown briefly in (115). 115. a. [talom pro]j ' a c x - a n - t a l l [Uikem pro]j ats'x-en-uili [all pro]j see-tr-erg.extr 'Everyone saw the people.' b. xrJaVcln-anS [x.waqwu give-2sg.obj-lsg.subj [all 'I give you all of it.' [?l ?ux w almfx w -a] tj [i ucwalmicw-a] tj [pl.det person-exis] tj (St'aYimcets; LT) pro] pro] (Upper Chehalis; M . D . Kinkade, p.c.) c. [m^kw pro] qay [all pro] die 'They are all dead.' (Chilliwack; Galloway 1977:453) The structure in (114) avoids having to postulate two different DP-internal positions for universal quantifiers; whether inside argument DPs or inside adjuncts, they occupy the adjoined position (see Chapter 5). The structure in (114) also enables us to capture the fact that the quantified temporal adjuncts optionally allow a determiner to be present as well as the quantifier, at least in St'aYimcets. The determiner ku may appear between the quantifier and the temporal noun, as shown in (116). 116. [takem (kwu) Sqlt] k w -an-3-wa * $ x w - a n t l n-xj f lap-a [utkem (ku) sq'it] kw-en-s-w£ t'ecw-en ti n-xlflap-a [all (det) day] det-lsg.poss-nom-prog sweep-tr det lsg.poss-floor-exis 'Every day I sweep the floor.' (St'at'imcets; RW, GN) The determiner ku was analyzed above as encoding non-assertion of existence. However, the quantifier tdkem must quantify over a range which is non-empty, and therefore the quantifier is semanticaily incompatible with the non-assertion of existence determiner ku, since ku fails to assert the existence of a range. This accounts for the fact that inside an argument DP, tdkem and ku are incompatible, as in (117). 84 117. a. * caqw-an-+kan tu? [takem kwu cuqwaz"] * ts'aqw-an'-lhkan tu7 [tikem ku ts'uqwaz'] eat-tr-lsg.subj compl [all det fish] 'I ate all the fish.' (St'at'imcets; RW) How are we to explain the fact that in the temporal adjuncts, takem and ku co-occur? Suppose that ku receives a different interpretation (or performs a different function) according to syntactic position. In an adjoined environment (as in (114)), ku does not encode the non-existence of its NP. This is supported by several other instances of adjoined ku, to be discussed in Chapter 4. We w i l l also see there that the structure in (114) is not unprecedented in the language; see Demirdache and Matthewson (1995a), Matthewson and Davis (1995), where a similar structure is proposed for head-final relative clauses and for demonstrative constructions.48 2.2.4. Accounting for the distribution of determiners The distribution of determiners in Salish is regulated as follows. Arguments always require a determiner, following Higginbotham (1985), Longobardi (1994); a determiner is necessary to bind a variable inside the NP and enable it to function as an argument. Hence, any phrase from which a determiner is syntactically missing must either be in predicate position or in adjunct position 4 9 The absence of determiners on predicates in Salish falls out straightforwardly from this analysis, as does the overwhelming tendency for arguments to contain determiners. Similarly, the ability for determiners to be missing in adjuncts follows from their non-argumental status. 4 8 Support for the claim that the £«-phrases, when they appear, are adjoined and give rise to a meaning like 'every one of them, for days' rather than 'every day' comes from a speaker's attempt to make sense of (i). i . * xek-an-4kan ftu? [takem kwu maw] * xek-en-lhkan t'u7 [takem ku maw] count-tr-lsg.subj just [all det cat] ! T counted them all and they were all cats.' (St'at'imcets; RW) 4 9 Recall that some languages in the Southern Interior branch allow overt determiners to be absent in argument DPs. Further research is required into whether these are instances of null (phonetically empty) determiners (cf. Longobardi 1994), or whether they constitute counter-evidence to the claims being made here. 85 There is one potentially interesting question raised by the adjuncts discussed in this section. Why should it be that the only instances of adjuncts which allow missing determiners are temporal adjuncts (at least in St'at'imcets)? An answer to this question is beyond the scope of the current study; however, I can sketch the direction an answer might take. The deictic component present on all determiners in Salish serves to locate the referent, not only in space, but also in time. The proximity distinctions encoded on determiners have a temporal as well as a spatial component. This is clearly true in Bella Coola (Davis and Saunders 1975), and also in St'at'imcets, where the 'absent' determiner ni.. .a carries a past interpretation (see van Eijk 1985:225, Demirdache 1996a,b). The temporal component of the deictic determiners may even be linked to the lack of systematic tense marking on the predicate in Salish languages (see the Introduction). Intuitively speaking, the determiners take over part of the function which would otherwise be performed by Tense. Demirdache (1996b) states it thus (emphasis original): The locus of parametric variation [between English and St'aYimcets] is ultimately the presence vs. absence of tense as a grammatical category: whereas in English morphological tense partly locates the temporal reference of a clause, in [St'at'imcets] determiners partly locate the temporal reference of a clause. If determiners in Salish have as one of their functions that of situating an event in time, there is an intuitive reason why it is the temporal adjuncts which allow determiners to be missing. If temporal nouns such as day have an intrinsic temporal component, then the temporal adjuncts may already be licensed (or 'situated') without the need for a determiner to perform this function. Of course, this freedom to do without a determiner for semantic reasons is only possible in adjuncts. When a determiner is needed for syntactic reasons to create an argument, even temporal phrases wi l l require a D. This analysis predicts that a determiner wi l l become 86 obligatory when a temporal phrase functions as an argument of a predicate. The examples in (118) show that this prediction is correct.50 In (118a), the presence of a temporal phrase without a determiner in subject position results in ungrammaticality. In (118b,c), the temporal phrase, which again lacks a determiner, cannot be interpreted as the object of the predicate; rather, the sentences are construed with null objects and temporal adjuncts. The marginality of (118b) results from the pragmatic oddness of the only possible reading; it is unusual to claim that one 'likes something every day'. 118. a. * wa? zaxt [zf^zeS' Sqlt] +-a§ * wa7 zact [zOzeg' sq'it] lh-as prog long [each day] when-3sg.conj 'Every day is long in summer.' plpa'nc'ek pipantsek summer (St'at'imcets; RW) b. ? ?ama-S-kan [takem Sqltl ? ama-s-kan [uikem sq'it] good-caus-lsg.subj [all day] T like something every day.' (# T like every day.') (St'at'imcets; RW) c. xeVen-rkan [zf?Z8<J' Sqlt] xek-en-lhkan [zf/zeg' sq'it] count-tr-lsg.subj [each day] 1 count something every day.' (# 1 count every day.') (St'at'imcets; RW) If determiners are inserted into the temporal phrases in (119), they can function as arguments.51 119. a. wa? zaxt [zf 'zaS 1 1 Sqft-a] wa7 zact [zf7zeg' i sq'it-a] prog long [each pl.det day-exis] 'Every day is long in summer.' +-a§ plpa'ncak lh-as pipantsek when-3sg.conj summer (St'at'imcets, RW) ?ama-S-kan [takem ama-s-kan [talcem good-caus-lsg.subj [all T like all the days.' ?1 §qft-al i sq'it-a] pl.det day-exis] (St'at'imcets; GN) 5 0 Sporadic instances have been recorded of temporal phrases in argument position with a missing determiner; I can find no systematicity in where this occurs. 5 1 It is difficult to find a pragmatically possible sentence where a temporal phrase functions as a transitive subject, hence the absence of such an example in (119). 87 c. x.a'k-anH-kan [zf?Z8S' [zf7zeg' [each ?1 Sqft-al i sq'it-a] pl.det day-exis] x6k-en-lhkan count-tr-lsg.subj 'I count every day.' (St'aYimcets; RW) Temporal adjuncts may only omit their determiner when a quantifier is present: 120. a. [takam S-Sap] k w -n-S-wa marhtaq mam'teq walk(redup) [uucem s-gap] kw-en-s-wa [all nom-evening] det-lg.poss-nom-prog 'I go for a walk every evening.' (St'aYimcets; LT) b. * [S-Sapl k w -n -S -wa * [s-gap] kw-en-s-wd [nom-evening] det-lg.poss-nom-prog 'I went for a walk in the evening.' mamtaq mam'teq walk(redup) (St'aYimcets; LT) Under the assumption that the temporal phrase in the quantified temporal adjuncts adjoins to DP as in (114) above, the question of why (120b) is impossible reduces to the question of why pro is not licensed in this environment. We can say that in (120a) the pro is a variable which is ranged over by the quantifier. In (120b), there is no quantifier to range over the variable, and an unbound variable results. To summarize, in argument positions (which require a determiner according to Longobardi's 1994 analysis), determiners are obligatory in most Salish languages. In non-argument positions, determiners are either absent (e.g. main predicate position) or optional (temporal adjuncts). Furthermore, a determiner in adjunct position is dropped only if there is sufficient information provided by the inherent lexical content of the temporal noun to situate the phrase in space and/or time. This recoverability condition is satisfied with nouns such as St'at'imcets sq'it 'day'. We saw in the previous subsections that Salish determiners are compatible with a Higginbotham/Longobardi-style analysis. Determiners are required before theta-assignment can take place. Given the independent evidence that Salish lacks quantificational determiners (§1.4), 2.2.5. Salish determiners accord with the variable-binder analysis 88 Salish therefore provides evidence that the variable-binding function of determiners is separate from any quantificational force that determiners might have. This separation of the two functions of determiners is not immediately obvious when one looks at English, since the class of elements which bind variables is commonly assumed to include the class of quantificational elements. Hence, one possible (but incorrect) interpretation of the Higginbotham theory would be that determiners will always be able to have quantificational force, as they do in English and Italian. Another interesting consequence of the facts outlined in this section is that if Salish determiners do function as operators which bind variables, then they behave the same as English determiners in this respect. In both language-types, determiners are a necessary prerequisite for argumenthood. Hence, the Operator-variable approach does not allow for parameterization which wil l capture the differences between Salish and English. The lack of quantificational determiners in Salish is a separate phenomenon from the argument-creating function of determiners. 3. Conclusions This chapter has made the following four proposals about determiner systems in Salish languages. 121. a. Salish determiners do not encode definiteness. b. Salish determiners do not encode specificity. c. There are no quantificational determiners in Salish. d. Salish determiners encode 'assertion of existence'. I have shown not only that definiteness and specificity are not morphologically encoded, but that Salish determiners cannot be analyzed as ambiguous between definite and indefinite, or specific and non-specific. This is because the distinctions which are marked in Salish cross-cut the definiteness and specificity distinctions, and also because the ambiguity hypothesis makes 89 incorrect predictions about the properties of Salish DPs (falsely predicting, for example, that DPs in Salish should allow Individual Concept Readings). The proposals in (121) form the basis for the theoretical discussions in Chapters 2 and 3. In those chapters, I wil l derive (121a-d) from a single underlying parameter which differentiates English-type languages from Salish-type languages. I will also show that while the assertion of existence distinction is not capturable by current semantic theories as they stand, it can be accounted for by a modification of Discourse Representation Theory (Kamp 1981; see also Heim 1982). With regard to the syntax of Salish determiners, evidence was provided that determiners are necessary for argumenthood in Salish languages. This accords with an analysis whereby determiners bind variables within NP, saturating the phrase and enabling it to receive a theta-role. 90 C H A P T E R 2 P R E S U P P O S I T I O N A L D E T E R M I N E R S A N D T H E C O M M O N G R O U N D P A R A M E T E R 0. Introduction In Chapter 1, four proposals were made about the semantics of Salish determiners. The aim of this chapter is to provide a unified account of the first three of these proposals, repeated in (1). 1. a. Salish determiners do not encode definiteness. b. Salish determiners do not encode specificity. c. There are no quantificational determiners in Salish. The chapter is organized as follows. Sections 1 and 2 introduce theoretical assumptions which wil l be necessary for the discussion. In §1, the notion of presupposition is discussed, and in §2, the dynamic semantic theories of Discourse Representation Theory (Kamp 1981) and File Change Semantics (Heim 1982) are briefly introduced. In §3,1 argue that the determiner types which are missing in Salish are precisely those which induce a presupposition of existence on the set ranged over by their common noun. Definite determiners are presuppositional (§3.1), specific determiners are presuppositional (§3.2), and quantificational determiners are presuppositional (§3.3). With respect to the last point I argue (contra Diesing 1992) that all quantificational determiners (both weak and strong) always presuppose the existence of their range, both in English and in Salish. In §4,1 argue that the deeper generalization which underlies (la-c) is the single claim in (2). 2. There are no presuppositional determiners in Salish. If (2) is the common feature which connects (la-c), then the semantics of determiners must be cross-linguistically parameterized. In §5,1 argue that the lack of presuppositional determiners in 91 Salish is derived from a negative setting of a Common Ground Parameter, which states that determiners in Salish may not access the common ground of the discourse. Since presupposition crucially relies on the common ground, presuppositional determiners are automatically ruled out. I show that the parameter is stateable at the level of the lexicon, is learnable, and sets up a subset-superset relation between Salish and English. In §6, I demonstrate that Discourse Representation Theory (Kamp 1981) and File Change Semantics (Heim 1982) can be adapted, using the Common Ground Parameter, to provide an insightful analysis of the differences between English and Salish. 1. Presupposition Soames (1989) examines three different definitions of 'presupposition' which are available in the literature: logical presupposition, expressive presupposition, and pragmatic presupposition. Pragmatic presupposition is the one I will adopt here; however, it is useful briefly to examine the other two definitions. LOGICAL PRESUPPOSITION can be regarded as a necessary semantic relation between propositions. The definition is given in (3) (which derives from Frege, and is taken from Soames 1989:556). 3. A proposition P logically presupposes a proposition Q iff the truth of Q is a necessary condition for P to be true or false. In (4), the proposition Q must be true in order for P to be either true or false. If Q is false, P cannot be assigned a truth value. 4. P: The queen of England is popular. Q: England has a (unique) queen. (Soames 1989:557) 92 EXPRESSIVE PRESUPPOSITION is a relation beween a sentence S and a proposition Q, defined as follows (deriving from Strawson, again taken from Soames 1989:562). 5. A sentence S expressively presupposes a proposition [Q] relative to a context C iff the truth of [Q] is necessary for S to semantically express a proposition in C. If Q in (6) is false, S not only fails to have a truth value, but fails even more fundamentally: it fails to express a proposition. 6. S: She is popular. Q: There is a contextually salient female under discussion, (cf. Soames 1989:560-562) Soames characterizes the difference between logical and expressive presupposition as corresponding to 'two different stages in the semantic evaluation of a sentence' (1989:562). The first stage is association of the sentence with the proposition which it expresses in the given context. It is at this stage that expressive presuppositions arise. The second stage is evaluation of the truth or falsity of the proposition; this is where logical presuppositions arise. In both cases, there is some proposition Q which must be true in order for the relevant sentence to be semantically felicitous. A more widespread view in recent literature is that presupposition is a relation between a proposition expressed and the COMMON GROUND of the participants in the conversation (cf. Stalnaker 1974:473). This is called PRAGMATIC PRESUPPOSITION (Stalnaker 1974). 'Common ground* is defined in (7). 7. The common ground: The set of propositions that both the speaker and the addressee believe (Chierchia and McConnell-Ginet 1990:290) According to the theory of pragmatic presupposition, a sentence imposes certain requirements on common background assumptions (i.e. the things that are taken for granted in a conversation); these background assumptions are the presuppositions. 93 Logical and expressive presuppositions, as well as other pragmatic factors such as conversational implicatures, can all give rise to pragmatic presuppositions. For example, participants in a conversation usually agree that a sentence will be pragmatically infelicitous if it does not have a truth value. Hence, any logical presupposition of a particular sentence (i.e. any proposition which is required for the sentence to have a truth value) will be presumed to be in the common ground (i.e. will be pragmatically presupposed; see Stalnaker 1973:452). This last view of presupposition, that it is a pragmatic effect, is explicitly or implicitly (usually the latter) adhered to by most recent literature, and it is the one I also adopt. The definition in (8) highlights the discourse-related nature of presuppositions (see also Heim 1982, Soames 1989).1 8. the hallmark of a presupposition is that it is taken for granted in the sense that its assumed truth is a precondition for felicitous utterance of the sentence and places a kind of constraint on discourse contexts that admit the sentence for interpretation (Chierchia and McConnell-Ginet 1990:283). To give a simple example, the utterance in (9S) pragmatically presupposes the information in (9Q), and wil l usually only be considered felicitous if Q is part of the common ground at the time of utterance.2 9. S: It was Joan who emigrated to New Zealand. Q: Someone emigrated to New Zealand. 1 In Heim's definition of presupposition, given in (i), the file F represents the common ground, i . 'A presupposes P means A is felicitous with respect to a file F only if F already contains the information expressed by P' (Heim 1982:366). 2 Presuppositions which are not already present in the common ground can be accommodated under certain circumstances; see Lewis (1979), Heim (1982), Stalnaker (1974), among others, and discussion in Chapter 1 above. 94 2. Discourse representation theory and File change semantics This section gives a brief introduction to two dynamic semantic theories which will be utilized throughout the remainder of the chapter: Kamp's (1981) Discourse Representation Theory and Heim's (1982) File Change Semantics.3 According to both Kamp and Heim, the meaning of a sentence is not determined in isolation, but rather with respect to the preceding discourse. Informally, each proposition which is added to a discourse narrows down the range of possible worlds with which the discourse is compatible. The meaning of a sentence thus consists of the change it makes to the set of possible worlds defined by the discourse. Rules for the construction of representations of meaning must make reference to the discourse representation already present (see e.g. Kamp and Reyle 1993:23-4). One motivation for such an approach comes from cross-sentential anaphora.4 In (10), the felicity of and interpretation of the anaphoric noun phrase it can only be determined with respect to preceding discourse. 10. Maggie owns a pianoi. Iti is blue. Similarly, felicity conditions for definite and indefinite noun phrases crucially relate to preceding discourse, as was shown in Chapter 1. If the individual corresponding to a DP argument is novel to the common ground, an indefinite must be used, as in (11); if the individual is familiar, a definite must be used, as shown in (12,13) (cf. Heim 1982 and references cited therein). 11. Novel context: a. I met [a man] today. b. * I met [the man] today. 3 Due to similarities in the two theories, both are sometimes grouped together under the term Discourse Representation Theory'; see e.g. de Hoop (1992), Chierchia (1995). 4 A major motivation for both theories is the search for an adequate treatment of donkey sentences as in (i), which are not directly relevant to the current discussion. i . Every man who owns a donkey beats it. 95 12. A . I met [a man]i today. B. What did [the man]i: look like? (novel) (familiar) 13. A . I met [a marVJi today. B. * What did [a man]i look like? (novel) (familiar) A second claim which is common to both Kamp and Heim is that indefinite noun phrases are not existentially quantified (contra Russell 1919). Kamp claims that indefinite descriptions are not existential quantifiers. When an indefinite has existential force it has that force in virtue of the particular role played by the clause containing it within the sentence or discourse of which it is part (Kamp 1981:5). For Heim, indefinite noun phrases are variables, which may come to receive existential force in certain syntactic contexts, but are not inherently existentially quantified.5 The formalisms used by the respective theories are as follows, briefly summarized. According to Heim, individuals indicated by DPs are each represented by a FILE CARD.6 Each file card contains all the information about a particular individual which is in the common ground of the discourse participants. As the conversation progresses, information is added to previously-existing file cards and/or new file cards are created. Within Discourse Representation Theory, the meaning of a discourse is represented by a Discourse Representation Structure (DRS). Each DRS consists of a set of DISCOURSE REFERENTS, which make up the universe of the discourse, and a set of conditions on those discourse referents. Discourse referents are formal representatives for the individuals indicated by noun phrases (cf. Karttunen 1976). A simple DRS is given in (14); the discourse referents are 5 Definites are variables also, according to Heim. Unlike indefinites, they are obligatorily unbound by an operator, and as such must refer to a contextually determined individual (1982:230,246). See the discussion of the Extended Novelty-Familiarity Condition below. 6 File cards can be compared to Karttunen's (1976) 'discourse referents'; see Heim (1982:249-263, 281). See also immediately below in the text for discussion of discourse referents within Discourse Representation Theory. 96 x and y. The conditions on the discourse referents are given in the lower half of the structure (cf. Kamp and Reyle 1993:63). 14. Maggie owns a piano. x y Maggie (x) piano (y) x owns y In (15), the use of the N P it introduces a new discourse referent z into the universe. Since it is a pronoun, it must be preferential with a previous discourse referent (in this case, y). 15. Maggie owns a piano. It is blue. x y z Maggie (x) piano (y) x owns y z = y blue (z) DPs may introduce discourse referents even when they have no referent in the real world. A n example is given in (16), involving an indefinite under the scope of negation.7 16. Maggie doesn't own [a clarinet]. Even though the DP [a clarinet] has no real-world referent, it still introduces a discourse referent into the DRS. The negation in (16) introduces a subordinate DRS, as shown in (17) (cf. Kamp and Reyle 1993:102). 7 See also Heim's discussion (1982:253) of the inappropriateness of the term 'discourse referent'. 97 17. Maggie doesn't own [a clarinet]. x Maggie (x) y clarinet(y) x owns y The positioning of a discourse referent within a DRS has consequences for coreference possibilities: a discourse referent which is inside a subordinate DRS may not corefer with a subsequent DP, as shown in (18). 18. Maggie doesn't own [a clarinetJi. * I saw [it]j yesterday. In the remainder of this and the following chapter, the theories of File Change Semantics and Discourse Representation Theory will be used and adapted to account for the Salish determiner system. 3. Presuppositions induced by determiners In this section, I wil l show that the determiner types which are missing from Salish are precisely those which induce presuppositions of existence on the set ranged over by their common nouns. 3.1. Definite determiners presuppose existence The individual corresponding to a definite DP such as the man in English must be familiar to the discourse participants. If an individual is familiar to the discourse participants, then it must be part of the common ground of those participants that the individual exists. Karttunen (1976:365) states it thus: 'definite descriptions ... carry an existential presupposition: to call something "the..." presupposes that there be some such thing.' 98 This result is derived formally by Heim (1982). Recall that Heim's system involves file cards which contain all the information about a particular individual which is in the common ground of the discourse participants. This approach provides a simple way of distinguishing definite from indefinite DPs, as shown in (19). 19. a. For every indefinite, start a new file card (indefinites are novel with respect to the file) b. For every definite, update a suitable old fde card (definites are familiar with respect to the file) (Heim 1982) The use of a definite DP indicates that the speaker presupposes the content of the DP; the descriptive content of the DP has necessarily been entered into the common ground of speaker and hearer (the file) prior to that utterance.8'9 The formal conditions on definites and indefinites are stated in (20), where d) is a logical form, F = the file for the entire discourse, and condition (ii) represents the presupposition of existence induced by definites.10 20. Extended Novelty-Familiarity Condition (Heim 1982:369-70): For <|> to be felicitous w.r.t. F it is required for every NPj in § that: (i) if NPj is [-definite], then i i Dom(F); (ii) if NPi is [-(-definite], then (a) i e Dom(F), and 0>) if NPi is a formula, F entails NPi 8 The identification of the 'common ground' with the 'file' (the set of file cards) comes from Heim (1982:286): T propose that the common ground of a context be identified with what I have been calling the "file" of that context.' 9 Existence within the file (the common ground of the speaker and hearer) must be differentiated from existence within the real world. For example, a DP may introduce a file card even if it has no referent in the real world. In (i), the indefinite under the scope of negation does not (under the preferred reading) correspond to an entity in the real world; it still introduces a file card, however. i . Sophie didn't buy [a fish]. While the indefinite in (i) introduces a temporary file card which lasts only under the scope of the negation, definite DPs always correspond to 'permanent' file cards, and thus induce a presupposition of existence. See Heim (1982) for discussion. 1 0 Since Heim's work was written before the DP-hypothesis was introduced, we can take the 'NP's in the definition to correspond to present-day 'DP's. 99 This condition states that in order for a logical form to be felicitous with respect to a file, no indefinite DP must correspond to a pre-existing file card in the domain of the file at the time of utterance (i). Conversely, every definite DP must correspond to a file card in the domain of the file at the time of utterance (iia), and the descriptive content of the definite must be already established within the file (iib). 1 1 This section has shown that a satisfactory account of definiteness entails that definites presuppose the existence of their discourse referents/file cards.1 2 3.2. Specific determiners presuppose existence1 3 Like definiteness, specificity has been linked to presuppositionality. Diesing (1992:80), for example, claims that 'the essential semantic contribution of 'specificity' [is] in fact presuppositionality.' As in Chapter 1, we must deal separately with the different definitions of specificity which are available. The exact type of presupposition which is induced by specific DPs must also be made 1 1 Some exceptions to the claim that definites always correspond to pre-existing file cards were noted in Chapter 1; an example is repeated in (i). (i) can be uttered when there has been no previous mention of the first baby to be born in the year 2010. The baby in question does not exist yet, so there is also no possibility that a file card for it has been entered into the file by virtue of the conversational participants both knowing the baby. i . I wil l meet the first baby to be born in the year 2010. (i) was claimed in Chapter (1) to involve presupposition accommodation. See also §4.1 below. 1 2 I have not discussed how Discourse Representation Theory (Kamp 1981) deals with the definite/indefinite distinction, mainly because no satisfactory account is available within that theory. Kamp and Reyle (1993:336) suggest that definites and indefinites be distinguished by relative scope: the discourse referent of a definite NP goes into the main DRS (like that of proper names such as Maggie in (17) above), while the discourse referent of an indefinite is inserted at the level at which it is processed (i.e. possibly within a subordinate DRS, as with a clarinet in (17)). Apart from the fact that this characterization of the definite-indefinite distinction ignores the familarity effects discussed by Heim and others, Kamp and Reyle themselves admit that it fails to capture the scope facts, since indefinites may have wide scope, and definites may have narrow scope. See §3.2, §6.2 for more discussion of the main/subordinate DRS distinction. 1 3 Thanks to Irene Heim (p.c.) for discussion of the issues presented in this section. She does not necessarily agree with the views presented here. 100 more precise. Given a sentence of the form in (21) (where 'an F is used specifically), there are three possible existence presuppositions: 21. A n F i s G . a. there is a non-empty set of Fs b. there is a non-empty and contextually salient set of Fs c. there is a non-empty set of Fs that are G Under Enc's (1991) definition of specificity, the presupposition induced by a specific is as in (21b), i.e. that there is a non-empty and contextually salient set of Fs. Enc claims (1991:9) that 'specifics require that their discourse referents be linked to previously established discourse referents.' This is clearly more than just a requirement that a particular set be non-empty, as in (21a). Similarly, Diesing (1992:87) characterizes the presupposition induced by Turkish specific objects as follows (emphasis added): the "specific" reading once again involves the notion of presupposition in that the "specificity" signaled by the accusative case marking corresponds directly to the formation of a restrictive clause that represents the set introduced in the preceding discourse. Under Ludlow and Neale's (1991) definition of specificity, it is not a requirement that a particular set be contextually salient or introduced in preceding discourse. Recall our example from Chapter 1, where Mary is explaining her glum demeanour to a friend. She utters the sentence in (22), and the mismatch between the Speaker's Grounds and the Proposition Meant result in a specific reading for the indefinite DP. 22. A n auditor is coming to see me today. Speaker's Grounds: Mr Beastly is coming to see me today. Proposition Meant: An auditor is coming to see me today. Proposition Expressed: An auditor is coming to see me today. Although the indefinite DP an auditor in (22) is specific according to Ludlow and Neale, the individual it corresponds to is not part of the common ground prior to the utterance in (22). The presupposition which is induced in (22) is crucially not that a discourse referent or a file card for the specific auditor M r Beastly exists in the common ground or is contextually salient. (If it did, 101 then an indefinite DP could not be used to corefer with that previous discourse referent; a definite would be required). However, I claim that the specific DP an auditor induces a presupposition that the set of auditors is non-empty. One way in which we can tell that specific indefinites induce such presuppositions of existence is that specific indefinites preserve their existential force under negation, a phenomenon common to presuppositional items, but differing from ordinary indefinites. Consider (23). 23. Sophie didn't buy [a book I recommended]. The most salient (or perhaps the only possible) reading of (23) is the reading where the indefinite object DP is specific. 1 4 The presupposition of existence induced by the specific DP is given in (24). 24. 3 x [book (x) and I recommended (x)] The existential force of the specific indefinite a book I recommended (i.e. the requirement that the set of books I recommended be non-empty) is preserved even under negation, as shown in the paraphrase of (23) given in (25): 25. There are books I recommended, and Sophie didn't buy any of them. (specific) This follows if the specific indefinite presupposes existence, since presuppositions are upheld under negation, as shown in (26) (see Morgan 1969, Langendoen and Savin 1971, Gazdar 1979, Karttunen and Peters 1979, Soames 1982, Heim 1988, among others). The presupposition P induced by the clefted structure is retained, even though the entire clause containing the cleft is placed under negation: 1 4 For a list of the factors which tend to favour a specific or a non-specific reading for indefinites, see Fodor and Sag (1982). The specific reading of (23) is facilitated by the placing of stress on book.. 102 26. S: It is not true that it was Joan who emigrated to New Zealand. P: Someone emigrated to New Zealand. Contrast the ability of specific indefinites to retain their existential force under negation with non-specific indefinites, which receive an existential interpretation if and only if they are not under the scope of an operator such as negation. 27. Sophie didn't buy [a unicorn]. The non-specific indefinite in (27) does not preserve its existential interpretation under negation. (27) can be felicitous, and receive the truth value 'true', in a world in which unicorns do not even exist. This is not true of the specific indefinite in (23); (23) is pragmatically infelicitous if there does not exist a book which the speaker recommended to Sophie. The claim that specific indefinites presuppose existence is particularly easy to demonstrate for the subset of specifics which Ludlow and Neale (1991) call 'strong' specifics. Consider (28), uttered in the context of driving past a smashed store window and wondering who smashed it. 28. A colleague I had coffee with last night did it. (Ludlow and Neale 1991:181) The speaker of (28) has singular grounds for the assertion (i.e. the speaker knows precisely who he or she had coffee with last night). The speaker does not expect the hearer to know which particular colleague is involved (i.e. intends to express only a general proposition). This much confirms that the indefinite is being used specifically. Furthermore, the speaker of (28) also knows that the hearer wil l realize that a singular proposition forms the Speaker's Grounds (Ludlow and Neale 1991:181). It is this last feature which makes the DP in (28) a 'strong' specific. So-called 'strong' specifics like those in (28) involve presuposition. There are two possible scenarios, depending on the extent of the hearer's prior knowledge. 103 29. a. The hearer already knows that the speaker had coffee with someone last night. b. The hearer does not already know that the speaker had coffee with someone last night. In the case of (29a), the hearer already knows that a coffee-partner exists, and the presupposition of existence of the coffee-partner is satisfied. In the case of (29b), the hearer does not already know that a coffee-partner exists, and there is presupposition failure. However, when the hearer hears (28), he or she wil l realize that the speaker had singular grounds for the assertion, and hence that a particular coffee-partner must exist. He or she will therefore accommodate the proposition that a coffee-partner exists. We have seen in this section that under Enc's definition of specificity, specifics induce a presupposition that there is a non-empty and contextually salient set (as in (30a)). Under Ludlow and Neale's definition, specifics induce the weaker presupposition that there is a non-empty set (as in (30b)). 30. An F is G . a. there is a non-empty set of Fs b. there is a non-empty and contextually salient set of Fs The type of 'existence presupposition' which I wi l l argue throughout this chapter is never induced by Salish determiners is type (30a). Salish determiners may not induce a presupposition that a particular set is non-empty. However, notice that (30b) entails (30a). As was argued in §3.1, a previously established discourse referent is necessarily understood by conversational participants to exist. Hence, under either Enc's or Ludlow and Neale's approach, specific indefinites induce the weaker presupposition in (30a), namely that a particular set is non-empty. So far I have argued that specific indefinites induce presuppositions of existence. In the remainder of this section, let us briefly examine how specifics are dealt with by the theories of File Change Semantics and Discourse Representation Theory. 104 Heim (1982) suggests two possibilities for dealing with the fact that specific indefinites escape being bound by higher operators such as negation, while non-specific indefinites do not. The first possibility is to say that specific indefinites are actually [+ definite]. This would explain why they escape being bound by an operator, since definites are not subject to the construal rule which coindexes them with a higher operator. However, specific indefinites act like indefinites in other ways (such as being used when the discourse referent is novel to the discourse) (Heim 1982:224). Heim therefore rejects this option. The second possibility is to add a construal rule that raises an indefinite into a position dominated immediately by the T (text) node. This raises the specific out of the range of other operators, and at the same time captures the fact that a specific indefinite can violate ordinary scope constraints only in having widest scope, not intermediate scope (see Fodor and Sag 1982).1 5 Heim notes that though the approach would work, it is ad hoc (Heim 1982:225). She leaves the issue unresolved.16 Heim's second option, which moves specific indefinites to a position of widest scope, correctly ensures that they receive an existential interpretation, even when they are c-commanded by negation at S-Structure (as in (23)). Again, we see the necessity for an existential interpretation, which I have claimed above derives from the presupposition of existence induced by specific indefinites. Within Discourse Representation Theory, the specific/non-specific distinction correlates with a difference in the relative positions of discourse referents within the Discourse Representation Structure (DRS). Discourse referents of specific DPs are placed in the universe of the main DRS, while non-specifics are placed inside a subordinate DRS (Kamp and Reyle 1993; this is a 1 5 See Ludlow and Neale (1991) for the claim that specific indefinites can have intermediate scope. 1 6 See Heim (1989,1991) for later discussions of specificity. 105 revision of their earlier suggestion that the main/subordinate DRS distinction correlates with definiteness; see footnote 12). The example in (31) illustrates the procedure. (31) has two readings, depending on whether the indefinite DP is specific or non-specific. a. non-specific reading: Vx, a boy in Mary's class (x), 3y, a girl who Mary doesn't know (y), x fancies y. b. specific reading: 3y, a girl who Mary doesn't know (y), Vx, a boy in Mary's class (x), x fancies y. Only under the specific reading is coreference possible with a pronoun in a subsequent sentence (cf. Kamp and Reyle 1993:288-289). 32. a. non-specific reading: Every boy in Mary's class fancies [a girl who Mary doesn't know]i. * [She]i is tall. b. specific reading: Every boy in Mary's class fancies [a girl who Mary doesn't know]i. [She]i is tall. The DRSs for each reading are given in (33, 34) respectively. In (33), the discourse referent for the DP a girl who Mary doesn't know is introduced inside a subordinate DRS (the one corresponding to the consequent of the conditional set up by universal quantification; see Kamp and Reyle 1993). As such, it cannot license a coreferential pronoun in a subsequent sentence. 33. non-specific reading: 31. Every boy in Mary's class fancies [a girl who Mary doesn't know]. z u v Mary (z) z's class (u) Mary (v) x boy (x) => x in u y ghl (y) v knows y x fancies y 106 In (34), the discourse referent for a girl who Mary doesn't know is introduced into the main DRS. It is therefore accessible to a coreferential pronoun in a subsequent sentence. 34. specific reading: z u v y Mary (z) z's class (u) girl (y) Mary (v) i j v knows y x boy (x) x in u => x fancies y I wi l l argue in Chapter 3 for a re-interpretation of the main/subordinate DRS distinction. Based on data from Salish and English, I wil l propose that the main/subordinate DRS distinction actually captures assertion of existence, rather than specificity. See also §6.2 below. So far we have seen that definites presuppose existence, and that specifics presuppose existence. The final determiner-type which Salish obligatorily lacks is discussed in the following section. 3 . 3 . Quantificational determiners presuppose existence It is often claimed that quantifiers induce a presupposition of existence on the set ranged over by their common noun (see Strawson 1952, Milsark 1974, Soames 1989, Enc 1991, among others). For example, the quantifier every in (35) induces a presupposition that unicorns exist. 35. Every unicorn likes bananas. It is extremely difficult to assign a truth value to (35), if it is not assumed that unicorns exist. Since sentences without truth values are pragmatically infelicitous, (35) ends up sounding odd in a context where the discourse participants do not agree that unicorns exist. This pragmatic 107 'oddness' results precisely from the failure of the presupposition of existence induced by the quantifier.17 Strawson (1952:172f) describes the presuppositions induced by quantifiers as follows. There are many ordinary sentences beginning with such phrases as ' A l l ...', ' A l l the ...', 'No ...', 'None of the ...', 'Some ...', 'Some of the ...', 'At least one ...', 'At least one of the ...' which exhibit, in their standard employment, parallel characteristics to those I have just described in the case of a representative ' A l l . . . ' sentence. That is to say, the existence of members of the subject-class is to be regarded as presupposed (in the special sense described) by statements made by the use of these sentences; to be regarded as a necessary condition, not of the truth simply, but of the truth or falsity, of such statements. A point which must be clarified, as far as possible, is the relevant notion of 'existence' which is presupposed by quantifiers. This is a complex issue. At a trivial level which we can disregard, the use of any common noun already entails that the concept of its potential referents exists in some universe. At the other extreme, we do not want to say that quantifiers presuppose existence 'in the real world', since then abstract entities (including mythical creatures) could never be quantified over without presupposition failure. The presupposition induced by quantifiers is also not equivalent to a requirement for 'familiarity within the current discourse'. As noted in the definitions of presupposition given above, there are more ways to be in the common ground than to have been previously mentioned in discourse. Hence, the presupposition in (35) (that unicorns exist) does not require that unicorns have been explicitly mentioned in prior discourse. On the contrary, the speaker presupposes that it is part of the hearer's general knowledge or belief system that unicorns exist. Thus, the only universes under discussion in a context in which (35) is uttered will be assumed to be those in which unicorns exist. 1 7 The relationship between presupposition of existence and the impossibility of the range of the quantifier being empty is highlighted by de Hoop (1992:213), who claims that in Dutch, 'sommige N presupposes existence, in the sense that it remains undefined if the set denoted by N is empty (cf. Strawson, 1950).' See Chapter 7 for more on sommige 'some' and presuppositionality in Dutch. 108 The relevant type of existence that we need seems to approximate to 'existence in one of the universes under consideration in the discourse'. This usage is intended to be compatible with the notion of 'existence' used by Heim (1982) in her analysis of indefinite DPs in English as receiving existential force. Let us examine some examples. In (36), the indefinite receives existential force. 1 8 A n assertion is made that a ghost exists in a certain universe, and that I saw that ghost. There is no presupposition that ghosts exist. 36. a. I saw a ghost. b. 3 x, ghost (x), I saw x. Once (36) is uttered, the only universes under consideration are those in which the set of ghosts is not empty. (37) differs from (36) in containing a weak quantifier (the numeral three). Unlike (36), (37) cannot be used to assert that ghosts exist in a certain universe. 37. a. I saw three ghosts. b. * 3 x, ghost (x), I saw three x. Rather, (37) presupposes that ghosts exist, and asserts that I saw three of them. The claim that weak quantifiers presuppose existence is not uncontroversial. In the next subsection I wil l examine this issue in more detail. If it is true that quantifiers (both weak and strong) always presuppose the existence of their range, then we will be able to claim that all the determiner types which are ruled out in Salish have one feature in common: they all involve presuppositions of existence. See Chapter 3 for discussion of the exact mechanisms which derive existential force in (36). 109 3.3.1. Do weak quantifiers presuppose existence? Diesing (1992) claims that only strong quantifiers (such as every, most) always 'presuppose the existence of the entities they are applied to.' Weak quantifiers (such as many, some) 'are ambiguous between a presuppositional and a non-presuppositional reading in which they merely assert the existence of whatever entities they are applied to.1 (Diesing 1992:59). The distinction is shown in (38), where (38b) is non-presuppositional, according to Diesing. 1 9 38. a. THREE ghosts are in the pantry; the others are in the attic. (stressed three, PRESUPPOSES the existence of ghosts) b. There are three ghosts in my house. (unstressed three, ASSERTS existence of ghosts) It is certainly true that weak quantifiers are ambiguous. The so-called 'strong', or 'quantificational' reading in (38a) is usually paraphrasable as a partitive (i.e. is felicitous only when a set of ghosts is already under discussion; Milsark 1974:240). The 'weak' or 'cardinal' reading in (38b) is non-partitive.2 0 Only the cardinal reading is available in r/iere-insertion contexts, as shown in (39). 39. a. * There are THREE ghosts in my house. (partitive reading) b. * There are three of the ghosts in my house. (partitive reading) c. There are three GHOSTS in my house. (cardinal reading) However, although weak quantifiers are ambiguous, the difference is not reducible to presupposition. On the contrary, I argue that both readings of weak quantifiers induce a presupposition of existence on the set ranged over by their common noun, for the following reasons. 1 9 Diesing's examples use the weak quantifier some; the same results should apply to all weak quantifiers, according to her theory. 2 0 Focal stress on the noun helps to force the cardinal reading in (39c) (Michael Rochemont, p.c). Michael Rochemont also notes that while the stress pattern in (i) unambiguously induces the 'weak' reading, focal stress on the quantifier, as in (39a), does not unambiguously produce the 'strong' reading. While interesting, these issues are orthogonal to the main points here. 110 In arguing that examples like (38b) do not induce a presupposition of existence, Diesing fails to distinguish between a presupposition of existence of the entire DP, and a presupposition of existence of the set ranged over by the common noun. Thus, (38b) asserts the existence of three ghosts in my house, and the entire DP three ghosts is therefore non-presuppositional. However, it is not true that (38b) asserts the existence of a set of ghosts. Rather, (38b) presupposes the existence of ghosts; the set of ghosts is presupposed to be non-empty within the universe of discourse.21 The necessarily presuppositional nature of weak quantifiers becomes clearer if we avoid the existential there-insertion context of (38b). The sentence in (40) should also have a non-presuppositional reading, according to Diesing. 40. Three ghosts are sitting in my kitchen. (40) is ambiguous between a partitive reading (where ghosts are already under discussion, and where focal stress will appear on three) and a non-partitive reading (with neutral stress, or with stress on ghosts). However, even the non-partitive reading (with unstressed three) sounds pragmatically odd in a context where ghosts are assumed not to exist. If the hearer of (40) is unwilling to accommodate the presupposition that ghosts exist, s/he could deny the presupposition, as in the discourse in (41). 41. A : Three ghosts are sitting in my kitchen. B: Are you crazy? Ghosts don't even exist! Carden (1973:38-9) provides support for the claim that weak quantifiers presuppose existence. His test involves the verb deny, which allows presuppositions to project from a subordinate clause, as shown in (42).22 (42a) and (42b) both presuppose (42P). 2 1 As indicated above, there are some indefinites that do not involve presuppositions of existence of the set ranged over by their common noun, namely those containing the indefinite determiner a in examples such as (36). 2 2 On the projection of presuppositions, see Langendoen and Savin (1971) and much subsequent work (cited in §3.2) above. I l l 42. a. It was Joan who emigrated to New Zealand. b. Sophie denies that it was Joan who emigrated to New Zealand. P. Someone emigrated to New Zealand. Bearing this property of deny in mind, consider the following judgements provided by Carden (1973): 43. a. John denies the Whig candidates won. Assumes there were Whig candidates. Denies that they won. b. John denies that the many candidates won. Assumes that there were many candidates. Denies that they won. c. John denies that many candidates won. Assumes there were candidates who won. Denies that they were many. (Carden 1973:38-39) In (43a), the definite determiner in the Whig candidates predictably creates a presupposition of existence for Whig candidates. The same is true in (43b), where the definite determiner induces a presupposition that many candidates exist. In (43c), the weak quantifier many creates a presupposition that the set of candidates is non-empty. In this respect, it is behaving exactly like the definite determiner the; both induce presuppositions of existence on their range. Another example of the same contrast is given in (44), using an interrogative environment, which also allows presuppositions to project. 44. a. Did the cholera patients survive? Assumes that there were cholera patients, and questions 'survive1. b. Did the many patients survive? Assumes that there were many patients, and questions 'survive'. c. Did many patients survive? Assumes that patients survived, and asks whether they were many. (Carden 1973:43) 112 I am claiming, then, that all weak quantifiers in English, on both their readings, presuppose existence of the set ranged over by their common noun. 2 3 Results from quantifiers other than numerals and many seem to confirm the hypothesis. For example, Michael Rochemont observes (p.c.) that even the weak quantifier some presupposes existence, as in the sentence in (45), which presupposes that ghosts exist. 45. There are some ghosts under the stairs. The weak quantifier no also induces a presupposition of existence. Thus, (46), which clearly involves the cardinal reading of no since it is a f/iere-insertion context, presupposes that the set of ghosts is not empty. 46. There are no ghosts in my house. Examples with no highlight the necessity for separating the presupposition of existence of the set represented by the common noun from the existence of that represented by the whole DP. (46) obviously does not presuppose the existence of any ghosts that are in my house, but it does presuppose that the set of ghosts is non-empty. It therefore ends up being a tautology, and hence pragmatically odd, if it is assumed that ghosts do not exist. One apparent exception to the claim that the weak quantifier no presupposes existence was pointed out to me by Strang Burton, Martina Wiltschko and Michael Rochemont (p.c). In (47), the sentence as a whole explicitly denies that unicorns exist, which seems to be incompatible with a presupposition of the existence of unicorns. 2 3 Eloise Jelinek asks (p.c.) whether the sentence in (i) provides a counter-example to the claim that weak quantifiers always induce presuppositions of existence, i . kan x w f l-am Ikwu ka+l+S mf*a+] kan cwfl'-em [ku kalh€lhs mixalh] lsg.subj look.for-intr [non.exis.det three(animal) bear] Tm looking for three bears.' (St'at'imcets; LT) Jelinek suggests that the co-occurence of ku, which does not assert existence, and kalhelhs 'three', which according to my analysis presupposes existence, is problematic. However, the weak quantifier kalhilhs merely presupposes the existence of a set of bears. The determiner ku fits with the fact that there is no identifiable set of three bears which the speaker knows to exist and which the speaker is looking for. As long as the presupposition of existence of the common noun is separated from the presupposition of existence of the entire DP, there is no clash. 113 47. No unicorns exist. The problem posed by (47) is broader than the issue being discussed here. Chierchia (1995:236-238) discusses similar examples, and notes that sentences which deny existence pose problems for any familiarity-based theory of definites. If definites presuppose existence, as argued by Heim (1982) and many others, then (48) is anomalous in the same way as (47): 48. The king of France doesn't exist. Chierchia does not offer a solution to the problem raised by (48), but notes that (48) requires a very special discourse context; it is used only to deny somebody else's presupposition that the king of France exists. A possible mini-discourse is given in (49): 49. A : I just saw the king of France in the bakery. B : You can't have; the king of France doesn't exist. Presuppositions induced by lexical items or constructions can be explicitly denied or supressed; this is the case in (49B), and, by extension, in (47). Under normal circumstances, the phrase no unicorns results in a presupposition of existence of unicorns. If the context involves a specific denial that unicorns exist, the presupposition is over-ridden.24 If, as I have claimed, the partitive and cardinal readings of weak quantifiers do not differ in terms of the presupposition of existence they induce on their common noun, then the question of the true difference between the two readings arises. This has been the topic of much debate; see the discussion in Chapter 6. 2 4 A related but opposite potential problem is raised by verbs of creation, which assert the existence of their objects (thanks to Strang Burton and Martina Wiltschko for pointing this out). In (i), the verb invent itself asserts the existence of the cures; this implies that the existence of the cures is not presupposed. i . Some cures for that disease have been invented. However, (ii) shows that definites or strong quantifiers, which uncontroversially presuppose existence, are also possible in such sentences. i i . a. The cure for that disease was invented last year. b. Every possible cure for that disease has been invented, and people still die from it. 114 As a final note regarding the presuppositional status of quantifiers, Lappin and Reinhart (1988) argue that presupposition is not semantically a part of either strong or weak quantifiers. They argue that it is impossible for presupposition to be part of the meaning of weak quantifiers, since if one builds the existential requirement into these quantifiers, they are provably no longer intersective or symmetric (1988:1027).25 Their solution is to allow the pragmatics to take care of presupposition, following Strawson (1974). This account is compatible with the pragmatic approach to presupposition taken in this thesis (see §1 above). Based on these considerations, I assume that all quantificational determiners, both weak and strong, necessarily induce a presupposition of existence on the set ranged over by their common noun. The claim that weak quantifiers always presuppose existence becomes crucial in §4, where I derive the missing determiner-types in Salish from a single underlying generalization against presuppositional D°s. However, as pointed out by Irene Heim (p.c), it is possible that weak quantifiers do not occupy Deposition even in English. It could be the case that weak quantifiers do have a non-presuppositional reading (as argued by Diesing), but that for some independent reason, no language allows weak quantifiers to occupy the head of DP. There would then be no need to derive the absence of weakly quantificational D°s in Salish from the parameter to be proposed below, and a Diesing-type analysis of weak quantifiers would cease to conflict with my explanation of Salish. Another way in which the conclusions drawn in this section may be independent of my main thesis is as follows. I shall argue in Chapter 6 that Salish weak quantifiers always have a proportional reading when they appear inside DP. The constraint against presuppositional D°s in Salish would then rule out the weak quantifiers from D° position, even under Diesing's approach 2 5 Lappin and Reinhart admit that the weak quantifiers many and few are non-symmetric anyway, even on their cardinal readings. They therefore exclude many and/<?w from their account. For a definition and discussion of intersection / symmetry, see Chapter 6, §3.2. 115 whereby only the strong readings are presuppositional (thanks to Kai von Fintel for discussion of this point). I leave the issues raised by these suggestions for future research. 4. Salish lacks presuppositional determiners The preceding subsections have argued that definite determiners, specific determiners, and quantificational determiners all induce presuppositions of existence. A l l these determiner types can therefore be ruled out in Salish by means of the single claim in (50). 50. There are no presuppositional determiners in Salish. However, there are two respects in which the evidence provided so far falls short of demonstrating (50). First, I have yet to show that the distinctions which do show up in Salish determiner systems do not involve presuppositionality. This task forms part of Chapter 3, where I argue that the assertion of existence distinction encoded in Salish does not involve presupposition. Second, I argued in Chapter 1 only for the lack of a definiteness or specificity distinction in Salish, rather than for the lack of definite or specific determiners themselves. Hence, it is necessary to clarify exactly how the Salish facts fall out from (50). I take definiteness as an example; the same reasoning applies to specificity. The claim that a language lacks presuppositional determiners entails that the language lacks definite determiners, since definite determiners necessarily presuppose existence (see §3.1). Therefore, (50) must rule out both possible scenarios for a language which possesses definite determiners, given in (51). 51. Possible determiner systems for a language with definite determiners: a. Determiners encode a definite/indefinite contrast. b. A l l determiners are definite. 116 It was demonstrated in Chapter 1 that Salish languages do not encode a definite/indefinite contrast. However, I have not yet explicitly argued that (51b) does not hold in Salish. It could be the case that all determiners are definite in Salish, in which case (50) would not derive the Salish facts. There are good reasons why (51b) cannot be true of Salish languages. First, DPs may appear in clearly indefinite contexts (where the individual concerned is novel to the common ground of the discourse; see §1.2.2 in Chapter 1). Hence, the distinctive feature of definite determiners, their restriction to familiar contexts, is missing. This claim is supported by Jelinek (1995:526-7), who claims that 'Determiner Phrases in Straits permit both definite and indefinite readings.'26 Second, there is independent evidence that DPs in St'at'imcets pattern with English indefinites rather than definites when it comes to such phenomena as freedom of temporal reference (see Demirdache 1996a,b). This was already demonstrated in Chapter l ; 2 7 I briefly review the evidence here. DPs in St'at'imcets lack the Individual Concept Reading, which is allowed for English definites. Compare (52a) with (52b). 52. a. The president of the United States is powerful. for any time t, whoever is president at t is powerful at t (Demirdache 1996c:8) b. ?a*xa? [tl k9 l?aqSt9n -S-a tl United.States-al a7xa7 [ti kel7aqsten-s-a ti United.States-a] powerful [det chief-exis det United.States-exis] 'The president of the United States is powerful.' * for any time t, whoever is president at t is powerful at t. (Demirdache 1996c:8-9) Secondly, DPs in St'at'imcets do not have the range of temporal interpretations that definites have in English. DPs in St'at'imcets are temporally bound, as argued by Demirdache (1996a,b) 2 6 The same reasoning holds for the specific/non-specific distinction; it was shown in Chapter 1 that DPs in Salish may have clearly non-specific uses. This means that they lack the defining properties of specifics. 2 7 See §1.2.3 and § 117 and shown in (53). The DP in (53) cannot have a temporal interpretation which is independent of the temporal interpretation of the predicate. 53. ?a?xa? tu? [tl k9l?aq§t9n-§-a tl Unlted.States-a] a7xa7 tu7 [ti kel7aqsten-s-a ti United.States-a] powerful compl [det chief-3sg.poss-exis det United.States-exis] "The president of the United States was powerful.' (St'at'imcets; Demirdache 1996c) a. true if the individual who was the president at some past time was powerful at that past time b. true if the individual who is the president now was powerful at some past time (while he was president) which overlaps with the present time c. false if the individual who is the president now was powerful at some distinct past time (before he was president) In contrast, the definite DP in the English gloss for (53) can have a temporally free reading, being true in the context in (53c). 4.1. Independent evidence for the lack of presuppositional Ds The lack of presuppositional determiners in Salish has been independently proposed, for theory-internal reasons, by Demirdache and Matthewson (1995b), (Demirdache 1996c). I briefly review their argumentation here. Demirdache and Matthewson (1995b) discuss a well-known restriction which holds across the Salish family, known as One-Nominal Interpretation. Gerdts' (1998:59) formulation for Halkomelem is given in (54), and an illustration from St'at'imcets is given in (55) (see also Gardiner in press, to appear for Secwepemctsin). 54. One-Nominal Interpretation: In the absence of marking for other persons, a single 3rd person nominal is interpreted as the absolutive. 118 55. ^ c x ^ n - a S [tl 5qayxw-a] ats'x-en-as [ti sqaycw-a] see-tr-3erg [det man-exis] a. 'S/he saw the man.' (= pro saw the man) b. * "The man saw him/her.' (= the man saw pro) Demirdache and Matthewson claim that the restriction on the interpretation of (55) has to do with topic-focus structure (following also Roberts 1994, Davis 1994c). The null argument pro, which is anaphoric, is coindexed with the topic of the discourse and represents the topic of the sentence (i.e. a mapping between discourse topic and sentence topic is assumed). The well-known cross-linguistic correlation between topichood and subject position (for which there is independent evidence in Salish) helps to rule in (55a), where the null pronominal is the subject, and rule out (55b), where the null pronominal is the object. Notice, however, that (56) is grammatical. The contrast between (56) and (55b) shows that the ungrammaticality of (55b) does not simply have to do with the fact that pro occupies object position. Rather, the overt nominal in subject position in (55b) contributes to its ungrammaticality. 56. ?acx-an-a3 dts'x-en-as see-tr-3erg S/he saw him/her.' (= pro saw pro) Demirdache and Matthewson claim that overt nominals are never topics in Salish (unless they contain an overt quantifier). 2 8 ' 2 9 The inability of an overt DP to be the topic accounts for the 2 8 The claim requires further refinement, since the patient in a passive sentence may be overt, as in (i). The patient of a passive sentence is the topic in Salish (see e.g. Kinkade 1989, 1990, Matthewson 1993, Thomason and Everett 1993 and many others). i . Ta'cx.-en-am [ta §qayxw-a] tta Smu+ac'-a] dtsx-en-em [ta sqaycw-a] [ta smulhats-a] see-tr-pass [det man-exis] [det woman-exis] 'The man was seen by the woman.' (St'at'imcets; RW, GN) Note that the passive is morpohologically intransitive; subjects of intransitives may freely be overt in Salish, and stand outside the normal topic-tracking processes, according to Davis (1994c). Further research is required on the relation between sentence topic and discourse topic; see Demirdache (1996c) on this issue. 2 9 This leads to the claim that in sentences with two overt nominals, there is no topic. Demirdache and Matthewson suggest that this might account for the rarity of such sentences in Salish languages (see the Introduction). 119 mapping of the overt DP in (55) to the object position (within the canonical domain of focus, VP), rather than the subject position (canonically the topic position). The inability of overt DPs to be topical in Salish correlates with the inability of overt DPs to be presuppositional, in the following manner (see Demirdache 1996c). Syntactic topics have been argued to carry existential presuppositions (Reinhart 1982, Valduvi 1995). If overt DPs cannot induce presuppositions, then it will follow that they cannot be topics. Demirdache (1996c) notes that the ability of overt DPs in Salish to describe either novel or familiar discourse referents is non-problematic, since there is not a strict correlation between 'focus' and 'new information'. Focussed DPs can be either novel or familiar. For further independent arguments against presuppositional DPs in Salish, see Demirdache (1995,1996c). A final piece of evidence against presuppositional determiners in Salish concerns presupposition cancelling. Assertion of existence DPs are impossible in cases where in English, the presupposition of existence of definites can be cancelled. Consider (57). 57. A : I heard you guys have a good-looking chief. B: We don't H A V E a chief, so the chief can't be good-looking! The definite noun phrase the chief normally presupposes existence of a (unique) chief. However, if the presupposition is explicitly denied, as in (57B), the chief becomes felicitous even though no chief exists. We expect similar examples to be impossible in Salish if the function of an assertion of existence determiner is not to presuppose existence, but outright to assert it. Assertion of existence should always be incompatible with denial of existence. This is upheld, as shown in (58). 58. A : qanfm-rkan k w -S ?amh-alqwam [tl kwukwp1?-lap-a] qan'im-lhkan kw-s emh-aTqwem' [ti kukwpi7-lap-a] hear-lg.subj det-nom good-appear [det chief-2pl.poss-exis] 1 heard you guys have a good-looking chief.' 120 B: wi?-H<a+ ?fxwa? §-k w uk wp1? wa7-lhkalh icwa7 s-kukwpi7 prog-lpl.subj without nom-chief 'We don't have a chief.' * n H §-x w ?az k w -§ 'amh-a'lq^rh [tl kwukwp1?-a] * nilh s-cw7aoz kw-s emh-al'qwem' . [ti kukwpi7-a] foe nom-neg det-nom good-appear [det chief-exis] 'So the chief can't be good-looking.' (St'aYimcets; B F ) The assertion of existence determiner in B's second sentence was corrected by the consultant to a non-assertion of existence determiner (because the chief doesn't exist). A similar type of example is given in (59), repeated from Chapter 1. The next chief of Fountain does not exist at the time of utterance (cf. the baby example in footnote 11 of this chapter). In Engl ish, a definite determiner is possible, presumably inducing presupposition accommodation (i.e. altering of the common ground to include a file card for the next chief). In St'aYimcets, a non-assertion of existence determiner is required. 59. xuz'-+kan malyf-S [kwu xwuz' kwukwp1?] laWxa'xlap cuz'-lhkan mely'i-s [ku cuz' kukwpi7] l£ku7 caclep going.to- lsg . sub marry-caus [non.exis.det going.to chief] deic Fount'n 'I w i l l marry the next chief of Fountain (whoever it is).' (St'aYimcets; L T ) This accords with the fact that while presuppositions can be accommodated, the assertion of existence determiners in St'aYimcets do not presuppose existence, but actually assert it. Hence, unlike a definite determiner in Engl ish, they are incompatible with any individual or entity which has not yet come into existence. 5. The Common Ground Parameter Determiners in Salish do not induce existential presuppositions on the set ranged over by their common nouns. Salish determiners crucially cannot be analyzed as homophonous between non-presupposi t ional and presupposit ional uses, but must str ict ly be analyzed as non-121 presuppositional. They differ fundamentally in this from English determiners, which access presuppositional distinctions in a number of ways. In this section, I wil l propose a parametric account of the difference between English and Salish. A Common Ground Parameter will be proposed, which rules out all of the required determiner-types from Salish. The parameter enables the difference between Salish and English to be stated in one simple, learnable statement. It sets up a subset-superset relation between Salish and English, correctly predicting that while English may access every distinction which Salish accesses, the reverse does not hold. §5.1 introduces the parameter and shows how it derives the absence of presuppositional determiners. §5.2 discusses a possible strengthening of the parameter. In §5.3, I address theoretical issues related to language typology and parameter setting, and in §5.4 I discuss the cross-linguistic predictions entailed by the parameter. In §6 I wil l show how the parametric variation is formally instantiated within Discourse Representation Theory and File Change Semantics. 5.1. Presupposition relies on the common ground Presupposition crucially relies on the notion of COMMON GROUND. The common ground encompasses the beliefs of both the speaker and the hearer of any utterance. The relevance of the hearer's beliefs is highlighted by Stalnaker (1974:473; see also Soames 1982), who claims that A proposition P is a pragmatic presupposition of a speaker in a given context just in case the speaker assumes or believes that P, assumes or believes that his addressee assumes or believes that P, and assumes or believes that his addressee recognizes that he is making these assumptions, or has these beliefs.3 0 3 0 Chierchia and McConnell-Ginet's (1990:290) definition of 'common ground' also explicitly makes note of the relevance of the hearer's beliefs. See also Heim (1982). 122 According to this definition, presupposition relies on three separate sorts of assumptions or beliefs, as shown in (60). Two of the three required components involve the speaker believing something about the hearer's state of knowledge. 60. Assumptions or beliefs involved in a pragmatic presupposition P (adapted from Stalnaker 1974): a. The speaker's assumption or belief that P b. The speaker's assumption or belief that the hearer assumes or believes P c. The speaker's assumption or belief that the hearer recognizes that the speaker assumes or believes P The ability to access or refer to hearer assumptions or beliefs is missing in the determiner systems of Salish languages. Only (60a), and never (60b) or (60c), can be explicitly encoded in the determiner system. For example, the presupposition of existence induced by a definite determiner is missing in Salish. In (61), determiner choice indicates only that the speaker assumes or believes that the policeman exists. The determiner ti.. .a crucially does not indicate anything about the hearer's assumptions or beliefs. Thus, (61) is felicitous in instances where the hearer has no knowledge of a policeman, as well as in contexts where the hearer is already familiar with the policeman. 61. tup-un-as" [tl plfSman-a] [kw-§ John] tup-un'-as [ti plfsmen-a] [kw-s John] hit-tr-3erg [det policeman-exis] [det-nom John] 'John hit a / the policeman.' (St'at'imcets) V Speaker assumes or believes that the policeman exists. x Speaker assumes or believes that the hearer assumes or believes that the policeman exists. x Speaker assumes or believes that the hearer recognizes that the speaker assumes or believes that the policeman exists. The claim that Salish determiners ignore hearer knowledge is supported by Kuipers' (1967:137) description of the choice between so-called 'definite' and 'indefinite' determiners in Squamish: The definite forms are used for objects which are individually identified for the speaker in an independent way. The Squamish definite catgory therefore differs from the English one in two ways. In the first place, in English the object must be definite for the hearer also: one begins a report with T met a man ...' (the man is identified for me but not yet for the hearer); in this case Squamish will use a definite form. On 123 the other hand, in English one uses the definite article in cases where the object is defined for the speaker in a way which depends on the description exclusively: 'who is the man you met?' (the man is identified for me merely by the description '(which) you met', but not independently -1 could not point him out); in such cases Squamish uses an indefinite form. The fact that Salish determiners access only speaker's beliefs, not hearer's beliefs, means that Salish determiners are unable to access the common ground. The common ground comprises the beliefs of both the speaker and the hearer; Salish determiners do not refer to this construct, merely taking notice of what the speaker knows. This state of affairs leads me to propose the following parameter. 62. Common Ground Parameter Determiners may access the common ground: Yes: { English, . . . } No: {Salish, . . . } According to the Common Ground Parameter, English determiners can access hearer knowledge, while Salish determiners cannot. However, both Salish and English access and encode speaker knowledge. The locus of the difference between the two language types is in the lexicon (i.e. the languages differ in the properties of individual lexical items, namely determiners). This is consistent with proposals that parametric differences may be situated only in the lexicon (see e.g. Borer 1983, Manzini and Wexler 1987, Chomsky 1993). Access to speaker knowledge must be a language universal, since speech without the expression of speaker beliefs would result in a marked lack of declarative sentences. We can therefore predict the following typology of language types. A language which accesses both speaker and hearer knowledge is English; by definition it accesses the common ground (the combination of speaker and hearer beliefs). A language which accesses only speaker knowledge lacks access to the common ground (Salish). Languages which do not access speaker knowledge are predicted not to exist. 124 63. Speaker knowledge is accessible: Hearer knowledge is accessible: The common ground is accessible: English Salish * * + + - -+ - + -+ - - -Since Salish does not allow access to hearer knowedge, but (by universal necessity) allows access to speaker knowledge, another way of stating the difference between Salish and English is to say that Salish determiners follow a SPEAKER-ORIENTED system, while English determiners follow a SPEAKER-HEARER-ORIENTED system. 5.2. Can Salish access hearer knowledge at all? It would not be true to say that Salish languages can never access hearer knowledge. Presupposition is induced in Salish by syntactic constructions such as clefting. For example, (64a) induces the presupposition in (64b). 64. a. ntt S-Henry tl qaxw9Xw-§-ta"l1-ha tl qflq-a nilh s-Henry ti qacwecw-s-tali-ha ti q'fl'q-a foe nom-Henry det break(redup)-caus-erg.extr-exis det chair-exis Tt was Henry who broke the chair.' (St'at'imcets; LT) b. Someone broke the chair. The presupposition in (64) is induced by a specific syntactic structure, namely clefting. Another syntactic structure which induces a presupposition is syntactic nominalization, illustrated in (65), with the presupposition in (65b).31 65. a. ?ama [tl S -Mq -S-a S-Gertlel ama [ti s-t'iq-s-a s-Gertie] good [det nom-arrive-3sg.poss-exis nom-Gertie] Tt is good that Gertie came.' Gertie came. (St'at'imcets; LT) 3 1 For discussion of the function and phrase structure position of syntactic nominalization in St'at'imcets, see Davis and Matthewson (1996b). 125 The only examples we have seen so far of presuppositions in Salish contain special syntactic constructions. If presuppositions were only induced by syntactic structures in Salish, we could restate the Common Ground Parameter as in (66). 66. Common Ground Parameter (strong version) The common ground may be accessed: a. By lexical items: Yes: {English,. . . } No: {Salish,...} b. By syntactic structures: Yes: { English, Salish, ... } No: 0? The formulation in (66) is a stronger formulation than the version given in (62), which rules out access to the common ground only by a subset of lexical items, namely determiners. Since, by hypothesis, parameters may only refer to lexical items, not to syntactic structures, we could propose that the ability of particular syntactic structures to induce presuppositions wil l be universal. What may vary is only whether particular lexical items (such as determiners) may induce presuppositions. In order to show that the strong formulation of the Common Ground Parameter holds, we would have to show that there are absolutely no lexical items in Salish languages that induce presuppositions. While a full investigation of the implications of the strong formulation goes beyond the scope of this dissertation, there is some evidence that the strong formulation as it stands is too strong. Soames (1982) lists (among others) the following constructions and lexical items which give rise to presuppositions: 67. a. B I L L Y is guilty, too. P: Someone other than Billy is guilty. (too) 126 b. Ivan has stopped beating his wife. P: Ivan has beaten his wife. (aspectual) c. B i l l regrets lying to his parents. P: B i l l has lied to his parents. (factive) St'at'imcets possesses at least the types in (67a,b). (68) shows that the lexical item corresponding to English too induces presuppositions.32 68. x w f l - 9 m * l t Ik Mary] tkwu Sx.uSum] cwil'-em t'it [k Mary] [ku sxusum] look.for-intr also [det Mary] [non.exis.det soapberry] 'Mary also looked for soapberries.' P: Someone other than Mary looked for soapberries. (St'aYimcets; LT) The aspectual type of presupposition in (67b) is rendered as in (69) in St'aYimcets. It is the entire subordinate clause which is presupposed. 69. c"ukw ?ay+ [k Mary] kw-§ x w f l - 9 m fkwu SxdSum] tsukw aylh [k Mary] kw-s cwfl'-em [ku sxusum] stop then [det Mary] det-nom look.for-intr [non.exis.det soapberry] 'Mary stopped looking for soapberries.' P: Mary has looked for soapberries. (St'at'imcets; LT) Factive verbs as in (67c) are rare in St'at'imcets; items corresponding to regret, discover and so on are generally conspicuously absent. A verb know does exist, however, and induces presuppositions, as shown in (70): 70. zwaVan-a§ [kw-S qax w-axw -S-a§ t l qflq-a k w -S Henry] zw£t-en-as [kw-s qacwecw-s-as ti qfl'q-a kw-s Henry] know-tr-3erg [det-nom break-caus-3erg det chair-exis det-nom Henry] 'She knows that Henry broke the chair.' P: Henry broke the chair. (St'at'imcets; LT) The presupposition induced by zwat 'know' in (70) consists of the entire subordinate clause. 3 2 Focal stress as in (67a) does not appear to exist in St'aYimcets (see Chapter 6). 127 The verb zwat 'know' does not by itself induce a presupposition of existence of an entity. The complement of zwat can contain the non-assertion of existence determiner, as shown in (71a,b). If zwat induced a presupposition of existence, this would clash with the failure of ku to assert that any entity exists.33 71. zwat-an-+kaxw ha zwat-en-lhkacw ha know-tr-2sg.subj ynq 'Do you know a chief?' [kwu kwukwp1?] [ku kukwpi7] [non.exis.det chief] (St'at'imcets) xw?ay ku7 kw-an-§ zwat-an cw7ay t'u7 kw-en-s zwat-en neg part det-lsg.poss-nom know-tr T don't know a chief.' tkwu kwukwp1?] [ku kukwpi7] [non.exis.det chief] (St'at'imcets) A yes-no question and a negation environment are used in (71) bec