UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Subject and topic in Styatyimcets (Lillooet Salish) Roberts, Taylor 1994

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

Item Metadata


831-ubc_1994-0605.pdf [ 2.21MB ]
JSON: 831-1.0087699.json
JSON-LD: 831-1.0087699-ld.json
RDF/XML (Pretty): 831-1.0087699-rdf.xml
RDF/JSON: 831-1.0087699-rdf.json
Turtle: 831-1.0087699-turtle.txt
N-Triples: 831-1.0087699-rdf-ntriples.txt
Original Record: 831-1.0087699-source.json
Full Text

Full Text

SUBJECT AND TOPIC IN ST’AT’IMCETS (LILLOOET SALISH) by TAYLOR ROBERTS B.A., York University, 1992 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS in THE FACULTY OF GRADUATE STUDIES (Department of Linguistics) We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA June 1994 © Taylor Roberts, 1994  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 shall not be allowed without my written permission.  (Signat  Department of  T.ingiiitics  The University of British Columbia Vancouver, Canada Date  DE-6 (2/88)  79  Time  1994  Abstract The goal of this thesis is twofold: first, to describe some of the symmetric and asymmetric behaviours of transitive and intransitive subjects in St’át’imcets, a Northern Interior Salish language spoken in southwest mainland British Columbia; second, to consider how the Principles and Parameters framework (Chomsky 1981; 1982; 1986; 1992; etc.) can explain the asymmetries. Although many Salish languages are known to display ergativity in their third-person subject inflection, the extent to which these languages are syntactically ergative is not well documented—perhaps because their accusativity has been more salient.  The question has not been investigated for  St’át’imcets, and this thesis shows that there is at least one aspect of St’át’imcets syntax—relativization—that appears to be ergative.  Evidence of ergativity in  coreference across conjuncts in St’át’imcets is not as clear, though; rather, coreference is restricted by a rule of one-nominal interpretation (Gerdts 1988) and a constraint on parallelism of discourse functions (Matthewson 1993a), both of which are shown in this thesis to derive from more general constraints on discourse. Unifying the explanations for the various asymmetries is the idea—independently motivated and proposed to account for facts in other languages—that NPs that are topics are structurally higher than NPs that are focused. Salish languages are often presented as though they were radically different from other languages, but with respect to the complex and subtle data examined in this thesis, St’ät’imcets resembles other known linguistic systems.  Most of the thta are  from original fieldwork, and they will be useful in the kind of comparative Northern Interior Salish research begun by Davis et al. (1993), Gardiner et al. (1993), and Matthewson et al. (1993). Syntactic pivots have not been investigated in the other MS languages, and so establishing the ways in which Nla’kapmxcin (Thompson) and Secwepemctsfn (Shuswap) differ from St’át’imcets will ideally help to explain the nature of parametric variation in syntax. 11  Table of Contents Abstract  .  ii  Table of Contents  iii  Symbols and Abbreviations  v  Acknowledgments  vi  Chapter 1 Introduction 1.1. Goals and outline of thesis 1.2. Grammatical relations and ergativity 1.3. Outline of St’át’imcets 1.3.1. Morphosyntax 1.3.2. Regional variation 1.4. Data and methodology  1 1 3 5 5 13 15  Chapter 2 Relativization 2.1. Introduction 2.2. Pronominal inflection of relative clauses 2.2.1. Subject of transitive 2.2.2. Subject of intransitive 2.2.3. Direct object 2.2.4. Possessor 2.2.5. Objectof comparison 2.3. Summary  17 17 21 21 23 27 30 34 35  Chapter 3 Conjunction 3.1. Introduction 3.2. Conjuncts with non-third-person 3.3. Conjuncts with third-person subject and object 3.3.1. Conjoined transitives Parallelism constraint on discourse functions Interaction with the one-nominal interpretation law 3.3.2. Intransitives Intransitive in first conjunct Intransitive in second conjunct 3.4. Summary  38 38 39 42 42 42 47 51 51 53 55  111  Chapter 4 Structural Asymmetries 4.1. Introduction 4.2. The one-nominal interpretation law as focus 4.3. Parallelism 4.3.1. Transitives 4.3.2. Intransitives 4.4. Extraction in relative clauses 4.5. Summary  57 57 62 66 66 83 87 102  Chapter 5 Conclusion  104  Appendix A: Morphology  106  Appendix B: Key to Orthography  111  Appendix C: Consultants  112  Appendix D: Elicitation Sessions  115  References  117  .  iv  Symbols and Abbreviations *  7 (x) (*x) *(x) {xly} -  1, 2, 3 A abs acc adhort  anti  appi comp conj cons def deic det emp emph erg evid F foc fut GF hyp incip md indic interrog intr M MS nom now NP O obj obl  ungrammatical 1. marginally grammatical; 2. grammaticality varies x is optional ungrammatical if x is present ungrammatical if x is absent either x or y, but not both morpheme boundary first, second, third person subject of transitive absolutive accusative adhortative antithesis applicative complementizer 1. conjunction; 2. conjunctive inflection consequential definite deictic determiner emphatic pronoun emphasis ergative evidential Fountain dialect focus remote future, possibility grammatical function hypothetical incipient indirective indicative interrogative intransitive Mount Currie dialect Northern Interior Salish nominalizer demarcation of time noun phrase direct object object oblique  oblig ONO  sg su TO tr  obligation, expectancy one-nominal interpretation law (23, 94, 120) out of control particle: ‘well, but, so’ passive parallelism constraint on discourse functions (87, 128) plural possessive predicate presupposed knowledge progressive quotative relative clause reciprocal redirective 1. subject of intransitive; 2. sentence constituent singular subject (indicative) topical object marker (-tali) transitive  Consultants AA BF DU GN LT RW  Alice Adolph Beverley Frank Dorothy Ursaki (Nla’kapmxcmn) Gertrude Ned Laura Thevarge Rose Whitley  ooc part pass PC p1 poss pred presupp prog quot RC recip red S  V  Acknowledgments The Lillooet Tribal Council gave permission to study St’át’imcets. Thanks are due to the following consultants for their help—and for enduring my ghastly pronunciations and sentences—while I tried to learn their language: Alice Adolph, Beverley Frank, Gertrude Ned, Laura Thevarge, and Rose Whitley. During the fieldwork, Bucky Ned of Ts’k’wáylacw (Pavilion) provided meals and a place to sleep, and he has been unbelievably kind and generous to the linguists who invade his home every two weeks. Kukwstum ‘ckál ‘ap, I nsnek ‘wntk ‘w7a. Thanks to my thesis supervisor, Hamida Demirdache, who has done the work of three people this year, and who has given freely of her time and ideas to me and to the other students who desperately needed them. I’m lucky also to have had the following professors volunteer to be on my committee: Henry Davis, M. Dale Kinkade, and Patricia A. Shaw. Paul Kroeber provided detailed comments on an earlier version of chapter 2 and generally shared his expertise with Salish throughout the year. Lisa Matthewson also read most of the thesis and saved me from making several errors. Dorothy Ursaki provided some Nla’kapmxcin data for chapter 2. My other professors at UBC included Laura Downing, Doug Pulleyblank, and Michael Rochemont. Myles Leitch and Doug Pulleyblank have assisted me at various times in persuading an IBM-compatible computer to do linguistics papers and, finally, this thesis. Carmen de Silva has also been helpful during my programme. The Project on Lexical Interfaces with Phonology and Syntax in North West Coast Languages (SSHRCC grant 410-924629) supported fieldwork on St’át’imcets, and I’ve benefited phonologically and syntactically through collaboration with the project researchers: Susan Blake, Henry Davis, Lisa Matthewson, and Pat Shaw. Henry and Lisa first encouraged me to work on St’át’imcets, and the results presented here, such as they are, were made possible only by Henry and Lisa having done hours of fieldwork on St’át’imcets before me, and then sharing their knowledge of the language with me. Henry, in particular, has done more than anybody to help me with the fieldwork—intellectually, emotionally, financially, and otherwise. This thesis would not have been possible without Henry, and without a heart-toheart tsétstecwts with him several months ago. Some of my other friends and fellow students at UBC include Eleanor Blain, Helmi Braches, Sohee Choi, Elizabeth Currie, Ping Jiang-King, Wen Li, Nike Ola, Yan Feng Qu, Monica Sanchez, Kimary Shahin, Will Thompson, Bill Turkel, Aki Uechi, and Vanessa Valerga. I’m thankful for the courage that several of them had in daring to feed a vegetarian and for their thoughtfulness in sharing their company with me when I needed it. Susan Blake, especially, has helped to make the department more fun that it otherwise would have been. Ruth King and Barry Miller encouraged me to pursue syntax as an undergraduate at York University, and I’ll always be grateful for their help and enthusiasm, which continues to this day. My largest debt is to Erwin, Andrea, and Brent (my family), who—though two thousand miles away for the Last two years—have remained my best friends. I love them.  vi  Chapter 1 Introduction  1.1.  Goals and outline of thesis  The goal of this thesis is twofold: first, to describe some of the symmetric and asymmetric behaviours of transitive and intransitive subjects in St’át’imcets, a Northern Interior Salish language; second, to consider how the Principles and Parameters framework (Chomsky 1981; 1982; 1986; 1992; etc.) can explain the asymmetries.  Although many Salish languages are  known to display ergativity in their third-person subject inflection, the extent to which these languages are syntactically ergative is not well documented—perhaps because their accusativity has been more salient. The question has not been investigated for St’át’imcets, and this thesis shows that there is at least one aspect of St’át’imcets syntax—relativization—that appears to be ergative. Evidence of ergativity in coreference across conjuncts in St’át’imcets is not as clear, though; rather, coreference is restricted by a rule of one-nominal interpretation (Gerdts 1988) and a constraint on parallelism of discourse functions (Matthewson 1993a), both of which are shown in this thesis to derive from more general constraints on discourse. Salish languages are often presented as though they were radically different from other languages, but with respect to the complex and subtle data examined in this thesis, St’át’imcets resembles other known linguistic systems. The results of this study have broader implications.  Understanding the St’át’imcets  pivots for relativization and coordination will shed light on discourse binding (Matthewson 1993; in prep.) and St’át’imcets grammatical relations generally.  Once the pivots for these  language-particular processes have been ascertained, they will help to explain, for example, to what extent the topical object marker -tall alters the grammatical relations of arguments. Given that -tall appears only in certain ergative-centered relative clauses, its function may be to antipassivize the predicate, deriving an S from an underlying A in order to satisfy the ergative pivot for relativization.  Similar questions arise concerning the passive suffix -em.  The effects of these affixes have remained unclear precisely because it has not been determined  1  what syntactic requirements they are feeding; hypotheses concerning their functions may be tested once the pivots in St’át’imcets have been established. Moreover, most of the data in this thesis are from original fieldwork, and they will be useful in the kind of comparative Northern Interior Salish research begun by Davis et al. (1993), Gardiner et al. (1993), and Matthewson et al. (1993). Syntactic pivots have not been investigated in the other MS languages, and so establishing the ways in which Nl&kapmxcin (Thompson) and Secwepemctsfn (Shuswap) differ from St’át’imcets will ideally help to explain the nature of parametric variation in syntax. The thesis is organized as follows. Chapter 1 is the introduction. Chapter 2 examines relativization and concludes that it is morphologically (if not also syntactically) ergative. Coreference across conjuncts, on the other hand—the topic of chapter 3—will be seen to be neither ergative nor accusative; rather, it is most strongly constrained by a rule of one-nominal interpretation and a requirement that arguments of transitive predicates share the same discourse function across conjoined clauses. Chapters 2 and 3 are largely descriptive, and the conclusion of each chapter is that St’át’imcets is not typologically unusual in any respect. Because this point is seldom made, and because much effort has been expended in collecting the facts necessary to prove it, the presentation of data in these chapters is fairly detailed. Chapter 4 proposes structural analyses of the data in the preceding chapters, repeating the generalizations and some of the data that illustrate them. The thesis is therefore organized in such a way that it should be accessible to readers whose interests lie mainly in Salish language data, and also to readers interested primarily in syntactic theory. Readers of the latter category who skip directly to chapter 4 should bear in mind that the data presented there have necessarily been idealized, and that the preceding chapters may be consulted if information about the variation across speakers and across elicitation sessions is desired. Finally, there are four appendices that offer grammatical paradigms, a key to the orthography, biographies of the language consultants, and a table of the elicitation sessions. The next subsection of this introductory chapter briefly introduces the problems that languages displaying ergativity have posed for some theories of syntax, and the following  2  subsection outlines the morphosyntax of St’át’imcets.  The chapter concludes with an  explanation of the data and methodology that were used in preparing the thesis.  1.2.  Grammatical relations and ergativity  The grammatical relations of transitive subject, intransitive subject, and direct object, are the principal concern of this thesis. Following Dixon (1979), they are henceforth abbreviated as A (transitive subject), S (intransitive subject), and 0 (direct object). The abbreviation A for transitive subject is best regarded as mnemonic for ‘agent’, since transitive subjects are typically agentive. The GFs frequently pattern together in systems that are called ‘accusative’ and ‘ergative’; the relevant groupings of these relations are schematized below: (1)  Ergative  Accusative A S 0  nominative  A S 0  accusative  ergative absolutive  Processes that do not systematically group A, 5, and 0 are called ‘neither accusative nor ergative’. Ergative languages have long posed a problem for theories that identify a category ‘subject’ for the grouping {A,S}, one of the major difficulties being how to explain the method of case-assignment. For example, English can be seen to have accusative morphology in its pronoun system (i/me, he/him, etc.) as well as in syntactic properties that group the A and S roles into a single category ‘subject’.  An example of the latter is the so-called that-trace  effect: embedded subjects (i.e., A and S) may not be questioned if the complementizer that is present (2a,b). This restriction does not hold of 0, however, as (2c) indicates. Note that all of the examples in (2) are grammatical if that is absent. (2)  a. b. c.  who do you think (*that) tj likes Fred? (A) 1 1 do you think (*that) tj is dying? (S) who 1 do you think (that) Fred likes tj? (0) who  3  The correlation between morphological case-marking (nominative/accusative) and grammatical relations (subject/object) in languages like English has been taken to support the idea that nominative case is assigned from a unique case-assigner (tensed Infi) to a unique structural position (specifier of IP). Ergative languages pose a problem for this theory of caseassignment.  Anderson (1976), for example, claims that subjecthood is best defined not by  morphological categories like case inflection and verb agreement, but by syntactic properties like control, raising, reflexive, and coordination.  He shows that many languages that are  morphologically ergative are in fact syntactically accusative, since they treat {A,S} in opposition to 0 with respect to these syntactic properties—thus supporting his claim that {A,S} forms a universal category of ‘subject’, whether it is defined as a primitive, as it is in Relational Grammar, or structurally, as it is in the Principles and Parameters framework. The fact that languages can have ergative morphology but accusative syntax challenges theories that assume a structural correlation between case and grammatical functions.  For some recent  approaches to these issues, see Murasugi (1992) and Campana (1992) and the references there. Dixon (1979) refines Anderson’s (1976) observations by arguing that the notion of ‘pivot’ is needed in addition to Anderson’s notion of subject {A,S}.  While agreeing that  certain syntactic constructions like imperatives and jussives universally refer to a ‘deep subject’ that comprises {A,S}, Dixon shows that other processes like coordination and subordination may have either accusative or ergative properties, as determined on a language-particular basis.  Languages like English that are thoroughly accusative happen to have their pivots  defined on the same set as the ‘deep subject’ {A,S}, but other languages select an ergative St’át’imcets may be a language of the latter  {S,0} pivot for some aspects of their syntax.  category. Chapter 2 will show that an ergative pivot is at work in St’át’imcets relativization (as it is in many languages), and this presents a problem for a theory of relativization that assumes a notion of subject that is defined on {A,S}, e.g., Keenan and Comrie (1977: 80 ff.).  4  1.3.  Outline of St’át’imcets  St’át’imcets is spoken in southwest mainland British Columbia, in an area 160-300 kilometers north-northeast of Vancouver. A grammatical sketch of the language is given in § 1.3.1, and some differences between the Mount Currie and Fountain dialects are noted in § 1.3.2.  1.3.1. Morphosyntax This section introduces some morphosyntactic characteristics of St’át’imcets that will be relevant for considering the kind of data to be presented in the thesis. Data are presented in the orthography that is used in van Eijk (1981, 1983) and Peters et al. (1992), with the exception of additional hyphens  (-),  orthography is in appendix B.  Right-aligned below each form cited in the thesis are the  which indicate morpheme boundaries.  A key to the  speakers’ initials and the token number(s) in the database, or other information identifying the source (see § 1.4). Appendix A provides pronominal paradigms and some other grammatical information, but van Eijk (1985) should be consulted for a fuller description of St’át’imcets. For an overview of the Salish language family, see Thompson (1979). St’át’imcets is a so-called radical head-marking language, since subject and object arguments are marked by obligatory pronominal affixes on the predicate (Davis 1993b: §2). Overt NPs are optional, as the following example shows: (3)  tsiin-tsi-lhkan  tell-2sg. obj-lsg. su ‘I told you’  (AA, LT 2296; van Eijk 1985: 174)  When overt nominals appear in addition to pronominal affixes, word order is fairly rigidly predicate-initial; specifically, VOS order (verb-object-subject) is preferred in elicitation by many speakers, while in texts VSO order appears about four times more frequently than VOS  (van Eijk 1985: 268 n. 5). VOS order is exemplified below:  5  (4)  ti sqáycw-a tsuw’-n-ás det man-det kick-tr-3sg. conj ‘the rock kicked the man’ * ‘the man kicked the rock’  ti k’ét’h-a det rock-det (AA 2223; LT 2324)  Some speakers are not as strict as others in their post-predicate word order, although they still prefer  (5)  VOS over VSO order: sqáycw-a ta áts’x-en-as ta man-det det see-tr-3erg det ‘the woman saw the man’ (preferred) ‘the man saw the woman’ (also possible)  smülhats-a woman-det (RW 48; Matthewson 1993a: 2-3)  For another example of this variation in basic word order between speakers, see (111) in chapter 3  (p.  54).  VSO order may be obtained if the predicate has restrictions concerning the animacy of its arguments. Consider the following example: (6)  i 1 tsftcw-a p1. det house-det  (tsicw) áts’x-en-as ti syáqts7-a see-tr-3erg det woman-det go ‘the woman saw the houses’  (AA 2229; LT 2326)  Here, a’ts’x-en ‘to see (tr)’ requires an animate experiencer as subject, and because this sentence has only one such NP (ti syáqts7a ‘the woman’), it may precede the inanimate object. When the sentence has two animate NPs, however, VSO order is not allowed, as the following example illustrates: (7)  *ats’x.en4tas ucwalmIcw-a i see-tr-3p1. su p1. det people-det ‘the people saw the woman’  1  ti syáqts7-a det woman-det (AA 2230)  The initial auxiliary tsicw ‘go’ was given only by LT, not by AA. 6  Nominals may not precede the predicate unless introduced by the focus particle nilh, as shown in (89).2 The asterisk (*) outside of the parentheses surrounding nil/i indicates that the entire sentence is ungrammatical if nil/i is absent: (8)  (9)  *(nilh) ts’um’-qs-án’-as ti syáqts7-a det woman-det lick-nose-tr-3erg foc ‘it was the woman that kissed the man’ *(nilh) ti sqáycw-a det man-det foc ‘the man left’  ti sqáycw-a det man-det (AA 2217)  qwatsáts leave (AA, LT 2214)  Emphatic pronouns, however, may behave predicatively and appear sentence-initially without the focus marker. For examples, see the footnotes pertaining to (46) and (47c) in chapter 2 (p. 24).  This property is apparently common to Salish languages, and Kinkade (1983: 28) has  taken it as evidence against a noun/verb distinction; see van Eijk and Hess (1986) for a different perspective. Nominals usually appear with what may be regarded as a discontinuous determiner, in that the first element of an NP is preceded by an article that encodes such categories as singular and plural, and followed by an enclitic -a. Examples of both singular ta and plural i appear above in (6). The enclitic -a is dropped if the progressive auxiliary wa7 intervenes between the article and the nominal; for examples, compare sentences (52-53) to (51) in chapter 2 (p.  (p.  26), and see also the footnote in chapter 3 giving AA’s version of sentence (86)  43). It can also be dropped in order to express other, as yet elusive, semantic contrasts.  There is phonological variation in the singular article, as can be seen by comparing (4) and (5) above. According to van Eijk (1985: 223), ti and ta are characteristic of the Mount Currie and Fountain dialects respectively, but they seem simply to be in free variation for the consultants of this thesis, who can even use ri and ta as articles for different NPs within a single sentence. 2  Some speakers, notably BF and Desmond Peters, permit SVO order. A possible explanation for this innovation is that these speakers have been in extensive contact with speakers of SecwepemctsIn (Davis 1993b: §3.1 n. 5), a language that allows nominals to precede the predicate (Gardiner 1993; Gardiner et al. 1993). 7  This is exemplified below for two speakers, who volunteered these sentences as translations for the English glosses indicated: (10)  (11)  ta maw-a tsuw’-n-ás áts’x-en-lhkan i when.past kick-tr-3sg.conj det cat-det see-tr-lsg.su ‘I saw when the woman kicked the cat’ áts’x-en-lhkan ti smülhats-a det woman-det see-tr-lsg. su ‘I saw the woman kick the cat’  ti smülhats-a det woman-det (RW 1629)  ta maw-a tsuw’-n-ás kick-tr-sg. conj det cat-det (GN 1630)  The vowel of the singular determiner can also be rounded if it precedes a word having an initial labial consonant; for an example, see (66) in chapter 2 (p. 32). Also see  § 1.3.2 below  for more information about the differences between Mount Currie and Fountain speech. The indicative paradigm (called ‘plain’ by Kroeber 1991), from which the main clause person markers are drawn, is said to be morphologically ‘split-ergative’. The 3sg subject of a transitive predicate in a main clause is marked on the predicate by the ergative suffix -as, as shown in (12). Direct objects and subjects of intransitive predicates, however, do not induce overt agreement on the predicate, as indicated in (12) and (13) by absolutive -Ø (examples (1213, 15-17) are adapted from van Eijk 1985: 172, 174): (12)  (13)  tsün-O-as tell (trans)-3sg. abs-3sg. erg ‘she told him’  (LT 2425)  tsut-O  say (intr)-3sg. abs ‘she said’  (LT 2426)  The null symbol -Ø is omitted from cited examples, as in (4-9) above, unless clarity requires it.  In addition to displaying ergativity in its morphology, St’át’imcets appears to display  syntactic ergativity in relativization, as chapter 2 will show.  8  In main clauses, non-third-person subject markers usually cliticize to the first pre predicate auxiliary if one is present. Compare the position of the lsg subject -lhkan in (3) and (14) in this respect, where tsukw ‘finish’ in (14) is the main predicate, and would otherwise host the subject marker -lhkan if no auxiliary were present: (14)  tsukw ka7lh hüy’ lhkan incip lsg. su for. while finish ‘I am going to quit for a while’  (AA, LT 2297; van Eijk 1985: 265; 1987: 163)  Non-third persons are inflected on a nominative/accusative pattern, since transitive and intransitive subjects are inflected alike, in opposition to direct objects: (15)  (16)  (17)  tsun-ts-ká1’ ap tell-lsg. acc-2p1. nominative ‘you guys told me’  (LT 2427)  tsüt-kal’ap say-2p1. nominative ‘you guys said’  (LT 2428)  tsüt-kan say-lsg. nominative ‘I said’  (LT 2429)  Note that 2pl above is marked by the same suffix (-kal’ap) when it is either a transitive subject (15) or an intransitive subject (16).  Direct objects, however, are marked differently from  subjects of intransitive; the 1 sg direct object in (15) is indicated by the affix -ts, while the 1 sg intransitive subject in (17) is marked by -kan. Given that NPs do not show overt case, when a single overt nominal appears in a sentence with a transitive predicate that bears third-person subject and object affixes, it might be expected to be ambiguous as to whether it is the subject or object. In such cases, however, there is a strong tendency for the overt nominal to be interpreted as the object rather than as the subject, as the glosses for the following sentences indicate: 9  (18)  (19)  wa7 k’al’em-mfn-as ta {smülhats-a/syáqts7-a} wait-appl-3erg det woman-det prog ‘he is waiting for the woman’ *‘the woman is waiting for him’ qvl-mmn-as ti syáqts7-a (wa7) prog bad-appl-3erg det woman-det ‘he doesn’t like the woman’ *‘the woman doesn’t like him’  (AA, GN 1312; LT 2329)  (AA 2233; cf. ON 1313; LT 2331)  In order for a sentence to be interpreted as having an overt NP subject and a null 3sg pronominal object, the sentence is passivized, as shown in (20-2 1).  Another example is the  passive in (22), which was volunteered as a form having the overt NP as subject, and it corresponds to the non-passive in (18) above. (20)  (21)  (22)  áts’x-en-as ta sqáycw-a see-tr-3erg det man-det ‘he saw the man’  (AA, BF, ON, LT, RW 29)  áts’x-en-em 1 ta sqáycw-a see-tr-pass obi det man-det ‘he was seen by the man’  (BF, ON, RW 28)  k’al’-em-mfn-em ti syáqts7-a nilh t’u7 s-e-s foc part nom-prog-3sg.poss wait-intr-appl-pass det woman-det ‘he is being waited for by the woman’  (LT 2330)  The same restriction has been observed in Halkomelem, a Coast Salish language, and it has come to be known as the ‘one-nominal interpretation law’ (Gerdts 1988: 59): (23)  One-Nominal Interpretation Law (ONO) In the absence of marking for other persons, a single 3rd person nominal is interpreted as the absolutive.  10  The rule of one-nominal interpretation will be seen in chapter 3 (pp. 47 ff.) to be a fairly strong constraint in St’át’imcets.  Its effects can also be observed in the two other MS  languages, Nla’kapmxcmn (Thompson and Thompson 1992: 145, 148) and SecwepemctsIn (Gardiner 1993: 214-219, §4.3.1). Pronominal markers in the subjunctive paradigm (called ‘conjunctive’ by Kroeber 1991) are similar in form to the plain clitics, except that they lack the -(lh)k- indicative marker.  Conjunctive inflection is used mainly for adverbial clauses and interrogative  complements, as well as for transitive complements of negation (cw7aoz ‘not’), and in other nominalized environments.  Adverbial clauses are introduced by the complementizers lh  ‘hypothetical’ or i ‘when.past’. Examples follow: (24)  ni7 qmut ] swan’ulh-ás saw-en-tsál-itas [ lh hyp whose-3sg. conj deic hat ask-tr-lsg. obj pl. subj 3 ‘they asked me whose hat that was’ (AA, LT 2299; van Eijk 1985: 272)  (25)  t’fq-as láni7 [ i when.past come-3sg. conj then ‘it is then that he came’  ] (LT 2431; van Eijk 1985: 272)  Relative clauses usually receive conjunctive inflection, although some speakers will accept plain (main clause) inflection in them; see sentence (49) in chapter 2  (p.  25).  Finally, factual inflection (called ‘nominalized’ by Kroeber 1991) appears in non-initial conjuncts introduced by nilh ‘then; so, then’, many examples of which appear in chapter 3. The predicate is preceded by the nominalizer s-, intransitive subjects are marked by possessive affixes, and transitive subjects are marked by conjunctive affixes. Nominalization also appears in complement clauses that are introduced by the complementizer kw, which is glossed  det(’erminer) because of its formal similarity to the indefinite determiners. This general use of the indefinite determiner as a complementizer in St’át’imcets resembles the Coast Salish pattern, and differs from Nla’kapmxcIn and Secwepemctsfn, which choose definite and  11  indefinite determiners as complementizers according to the semantics of the matrix predicate (Kroeber 1991: 135). A complement clause in St’át’imcets is exemplified below: (26)  wa7  prog  lhkan  lsg.su  zewát-en  know-tr  [kw  det  s-t’iq-s  nom-come-3sg.poss  ‘I know that he came’  ]  (AA, LT 2300; van Eijk 1985: 270)  The characteristics of these clause-types are summarized in the table in (27).  See  Kroeber (1991: 165) for a similar chart (but without introductory particles) for the Salish family. (27)  Inflection of clause types in St’át’imcets  [ Introductory particle  Clause type  Inflection  main adverbial, interrogative, negation  plain conjunctive  complement; non-initial conjunct  nominalized  none lii ‘ hypothetical’ 1 ‘when.past’ cw7aoz ‘not’ kw; nilh ‘then; so, then’  The facts outlined above represent the core of St’át’imcets grammar.  Naturally, there  is slight variation across speakers, the regional aspect of which is discussed in the next section. Concerning phonology, stress often shifts as affixes are added, and determiners sometimes appear to trigger metathesis in the roots to which they attach.  There is also an  alternation between -lhk/-k as the indicative marker for non-third-person subjects, which for brevity is usually not glossed as a separate morpheme, but is treated as part of the rest of the subject affix.  This morpheme appears as -lhk usually after vowels and resonants, and as -k  elsewhere; note the alternation for the lsg subject marker in (28-29) below, for example. These alternations are irrelevant to the thesis, and so nothing will be said concerning them; they should be small enough not to be distracting, and van Eijk (1985: 20-24, 32, 169) can be consulted for more information if desired.  12  1.3.2. Regional variation Two dialects of St’át’imcets are recognized by van Eijk (1985): the Fountain dialect, which is associated with the communities surrounding Sat’átqwa7 (Fraser River) near Sat’ (Lillooet), and the Mount Currie dialect, which is associated with the vicinity of Lil’wat7ül (Mount Currie, near Pemberton). The two groups are separated by the Lillooet mountain range, but are connected by about 100 kilometers of waterway and, today, by road.  The dialects are  mutually intelligible, the primary differences being a few lexical items, most of which are Such variation is exemplified by the words  recognized by the consultants of this thesis.  smülhats (F) and syáqtsa7 (M) ‘woman’, as in (18) above.  There is also a phonological  difference in the retraction of vowels that does not bear on the data here (van Eijk 1985: 8; 1987: 5). A small difference—not noted by van Eijk (1987: 212)—has emerged concerning the retraction of consonants: the lexical suffix -ts  [] ‘mouth; language’ in qvlqvl-ts’-mIn’  [qA+-qA+-c’-mmn’J ‘to swear at (tr)’ is retracted and glottalized only for Mount Currie speakers. Fountain speakers do not retract or glottalize the -ts, as example (85) in chapter 3 shows (and other sentences starting at p. 43). Apparently there is a syntactic difference between the Fountain and Mount Currie dialects. Although certain pronominal markers are second-position clitics, as noted on page 9 with respect to (14), Mount Currie speakers allow the person marker to appear sentenceinitially: (28)  xát’-min’ ku kál’wat kan lsg. su want-appi det medicine ‘I want some medicine’ (AA, 17 November 1993, UBC Field Methods, token 150; LT 2432)  These structures are analyzed by van Eijk (1987: 18) as having a pre-predicate auxiliary— progressive wa7—which is dropped (presumably) phonologically. (28) would therefore be:  13  The structure underlying  (29)  xát’-min’ wa7 lhkan prog lsg.su want-appi ‘I want some medicine’  ku kál’wat det medicine (LT 2433)  This analysis may not be correct, however, as one example of a fronted subject marker co occurring with progressive wa7 has been volunteered: (30)  t’u7 wa7 kan prog part lsg. su ‘I am standing’  s-tálh-lec nom-upright-body (AA, LT, 6 April 1994, UBC Field Methods, token 374; LT 2434)  Moreover, clitics may appear pre-predicatively in other Interior Salish languages (e.g., Columbian), suggesting that (28) represents the older pattern, structures like (29) instead being innovations. Another area of variation that deserves further study is the determiner system.  The  singular ti/ta alternation has been claimed by van Eijk (1985: 223)—and challenged above on page 8—to characterize the Mount Currie and Fountain dialects respectively.  Similarly,  proper nouns are almost invariably preceded by the nominalizer s- or the determiner nominalizer sequence kw-s for Fountain speakers, but AA and LT (Mount Currie speakers) almost never use the kw-s combination, preferring the nominalizer s- alone or the determiner  kw alone, often unrounding the latter to k or eliding it completely—although the unrounded sequence k-s has also been recorded. Van Eijk (1985: 228, 229 n. 2) notes that this ‘dropping’ of parts of kw-s is characteristic of younger speakers, but with this group of consultants it looks indeed more like an isogloss, since the full kw-s sequence before proper nouns and elsewhere is generally ungrammatical for Mount Currie speakers: (31)  á7ma k(*ws) Mary ihel s-Jane det-nom Mary than nom-Jane cute ‘Mary is prettier than Jane’ (AA, LT, 9 March 1994, UBC Field Methods, token 294; LT 2436)  14  (32)  áts’x-en-lhkan k(*ws) George i-nátcw-as see-tr-lsg. su det-nom George when.past-day-3sg. conj ‘I saw George yesterday’ (AA, LT, 9 March 1994, UBC Field Methods, token 304; LT 2437)  The ungrammaticality of kw-s in the above sentences makes this look like more of a syntactic phenomenon, rather than one of phonological reduction, but the matter needs further 3 investigation before conclusions can be drawn. The regional differences outlined above are the only ones that have been obvious, and where these differences exist in the data cited, they are mentioned in footnotes.  1.4.  Data and methodology  Data for this thesis were collected from native St’át’imcets speakers during the period November 1993 to May 1994, under the auspices of the Project on Lexical Interfaces with Phonology and Syntax in North West Coast Languages, as part of an ongoing syntactic database of St’át’imcets that was begun by Henry Davis and Lisa Matthewson in October 1992. Most of the original data for this thesis are from Gertrude Ned and Rose Whitley, and thus—despite minor variation between these speakers—can be taken to represent the Fountain dialect. Most (if not all) of the original sentences cited here have been checked and rechecked during sessions subsequent to their initial collection, and the crucial data have additionally been confirmed in independent sessions with Alice Adolph and Laura Thevarge, speakers of the Mount Currie dialect. (See appendix C for biographies and genealogies of the language consultants.)  The original elicitation of a sentence and its subsequent confirmations were  usually recorded in a single database record, and so several speakers are often associated with a single token number.  Sentences confirmed with the Mount Currie consultants may  occasionally differ with respect to lexical items like smálhats/sydqtsa7 ‘woman’, but these differences are usually not noted.  To get a clearer picture of which consultants originally  supplied a particular sentence, consult appendix D for a table of the elicitation sessions, which Fountain speaker RW characterizes sentences like these as typical of Mount Currie speech (Henry Davis, p.c.). 15  indicates which consultants were present at individual sessions.  Note also that different  sentences occasionally have the same token number because variants of a basic sentence have sometimes been recorded in the same database record. Only data as are sufficient to illustrate generalizations are given in the thesis, but other data that confirm these generalizations are in the aforenamed database. Some idealization of the data has been necessary in chapter 3 (where subtle grammaticality judgments are not always consistent across speakers and across elicitation sessions), but it is always stated explicitly. If it seems like undue attention is given to explicating the varying grammaticality judgments in parts of chapters 2 and 3, it is precisely because generalizations cannot be made. Great care has been taken to represent fully and accurately the speech and grammaticality judgments of the consultants, and where differences among speakers and sessions exist, these are noted in the text and footnotes. Textual data are not cited because texts cannot furnish two kinds of information that are necessary to the thesis: first, whether a sentence has more than a single interpretation; and second, whether a sentence would be ungrammatical if it had a slightly different form than the attested sentence. Textual data are useful for showing what syntactic structures are possible, but they do not reveal what structures and interpretations are impossible. This information is crucial for being able to describe a language’s syntactic restrictions, and thus for trying to understand what constitutes a speaker’s knowledge of language.  16  Chapter 2 Relativization  2.1.  Introduction  This chapter examines the structure of non-oblique-centered relative clauses (RC5) in 4 Transitive and intransitive St’át’imcets, filling in some of the gaps of previous descriptions. subjects will be seen not to pattern alike with respect to relativization, thus presenting a problem for a theory of relativization that assumes a unified notion of ‘subject’, e.g., Keenan and Comrie (1977: 80 ff.). Additionally, data are presented that clearly show that absolutive centered relative clauses contain a gap corresponding to the relativized constituent—not covert 3sg agreement, a possibility suggested by Matthewson (1993b). The implication of this is that, although St’át’imcets (like other Salish languages) does not display overt extraction of a relative pronoun as Indo-European languages do, relativization nevertheless involves extraction of the relativized constituent. A fuller analysis of the structure of RCs remains to be done not only for St’át’imcets, but for Salish languages generally. Kroeber (1991) gives an excellent overview of the variety of forms of RCs in Salish, although he does not have as much data on St’át’imcets as on the other Northern Interior languages, Nla’kapmxcmn (Thompson) and SecwepemctsIn (Shuswap).  The data presented in this chapter resolve all of the questions  raised by Kroeber (1991) and Matthewson (1993b) pertaining to pronominal inflection inside RCs. In comparing RCs across languages, Keenan and Comrie (1977: 63) and Kroeber (1991: 175) employ a semantically-based definition of RC, since the syntax of individual languages differs to an extent such that a purely syntactic definition of RC is difficult to maintain universally. The following definition, from Comrie (1981: 136), will suffice for this chapter: Parts of this chapter are from Roberts (1994); thanks to Paul Kroeber for his detailed comments on that paper. Previous descriptions of St’át’imcets RCs include Gardiner et al. (1993: 145-147), Matthewson (1993a: 14-27; 1993b: 14-18), Matthewson et al. (1993: 224226), Kroeber (1991: 258-264, 281-288), and van Eijk (1985: 185-187, 271). 17  restrictive relative clauses are more central to the notion of relative clause than are A relative clause then consists necessarily of a non-restrictives [= appositives]. head in itself has a certain potential range of head and a restricting clause. The referents, but the restricting clause restricts this set by giving a proposition that must be true of the actual referents of the over-all construction. •  .  .  .  .  .  In the English sentence Sally met the man who Fred hired, the NP the man is said to be the head, while who Fred hired is the restricting clause. Note that the restricting clause has a gap where we expect the direct object, and that this gap corefers with the head of the RC, the man. This is therefore an object-centered RC, since the relativized constituent (or ‘target’) bears the grammatical relation of object within the restricting clause (a property of the internal syntax of the RC). Moreover, this English example is said to be head-initial, or to have a post-head RC, since the head of the entire construction precedes the RC itself. The external syntax of the RC refers to the role of the entire NP the man who Fred hired with respect to the main clause— here, direct object of the matrix predicate meet. This chapter is concerned with the internal syntax of RCs, since it is at this level of structure that St’át’imcets RCs differ. Finally, another type of relative is the headless (or ‘free’) relative; an example corresponding to the English sentence above is Sally met who (ever) Fred hired. St’át’imcets has headless, post-head, and pre-head RCs, although the status of the latter construction as a true RC remains unclear. Nothing is said about them here, though it would be worthwhile to examine these structures in light of the findings that are presented in this chapter concerning the other two RC types. Kroeber (1991: 176) notes that appositives  [= non-restrictive RCs] are not widely  attested in Salish, and Comrie (1981: 132) states more generally that the distinction between restrictive and non-restrictive RCs is found only sporadically across the world’s languages. 5 Likewise, attempts to elicit appositives in St’át’imcets have not met with succeSs.  One  straightforward test for the existence of appositives is to determine whether an RC may modify a proper noun.  Unlike a restrictive RC, whose restricting clause serves to narrow the  reference specified by the head, an appositive merely supplies additional information about a  Thanks to Henry Davis for checking these data. 18  fully-identified head. Proper nouns are fully referential, hence they may normally head only  appositives—as in the English sentence John, who was only hired this morning, was fired this 6 named John, with afternoon. Here, there are not several persons in the universe of discourse the RC serving to narrow the reference of the main clause to a single individual. Rather, there is a single person named John in the universe of discourse, and the RC merely supplies additional information about him, namely, that he was only hired this morning. In St’át’imcets, however, proper nouns cannot head an RC, as the following example shows: (33)  *pzán4hkan i-nátcw-as ats’x-en-ácw-a [ta 1] [ s-John nom-John det see-tr-2sg. conj-det when.past-day-3sg. conj meet-tr-lsg. subj ‘I met John, who you saw yesterday’ (GN, RW 711; AA, LT 2303)  Sentences having an appositive interpretation may be expressed instead by coordination, as in the following example: (34)  [ áts’x-en-lhkacw s-John i-nátcw-as  see-tr-2sg. su nom-J. when.past-day-3sg. conj who saw yesterday’ met John, you ‘I  I müta7 [pz-án-lhkan conj  lhkünsa] meet-tr-lsg. subj now (BF 828)  Because a structure like (33) cannot be used for appositives, it can be safely assumed that all of the RC data examined in this chapter—which take this form—represent restrictive RCs. Keenan and Comrie (1977: 66) propose an Accessibility Hierarchy for relativization. The higher a grammatical function (GF) is on this scale, the easier it is to relativize:  (35)  Accessibility Hierarchy Subject > Direct Object > Indirect Object > Oblique > Genitive > Object of Comparison  Calabrese (1990: 12) defines the universe of discourse for an utterance U as ‘the set of referents, properties and presuppositions which the speaker believes the hearer presupposes in the time instant tin which U is uttered.’ 6  19  If a language allows relativization of more than one of these roles, the Accessibility Hierarchy predicts that they will fall contiguously on the scale. For example, if a language can relativize obliques (by which Keenan and Comrie mean arguments, not adverbials), then all of the higher GFs should also be relativizable. In this instance, it would not be expected that only obliques, subjects, and direct objects—but not indirect objects—could be relativized. Distinct strategies of relativization within a language (e.g., changes in word order, gapping, case-marking, etc.) are also predicted to share contiguous elements on the scale. When a language does employ different strategies of relativization, it is further predicted that the less explicit types (e.g., omitting morphology corresponding to the relativized target) will be employed higher on the scale since those positions are easier to relativize, whereas the more explicit RC forms (e.g., retaining target morphology) will be employed for the roles that are lower on the scale, since they are harder to relativize (Comrie 1981: 156). The Accessibility Hierarchy is a good place to begin the examination of St’át’imcets RCs, since Salish languages can challenge and test it in several respects.  Relativization in  St’át’imcets will be seen to distinguish ergative from absolutive with respect to relativization, contra Keenan and Comrie, who regard ‘subject’ as a unified notion in both accusative and ergative languages, and who argue that the notion of ‘subject’ is valid even in an ergative system, where the single argument of an intransitive predicate patterns with the object of a transitive predicate—rather than with the agent, as it does in an accusative language. Recall the chart in (1) in chapter 1 (p. 3). Salish lacks a distinct ‘indirect object’ role, and so this member of the Accessibility Hierarchy may be disregarded (Kroeber 1991: 232).  Kroeber (1991: 233) argues moreover  that genitive (possessor) must be higher on the hierarchy than Keenan and Comrie suggest, and also that the ‘oblique’ category needs to be subdivided into oblique object, instrument, and locative. His revised scale for Salish therefore has the following form: (36)  Accessibility Hierarchy for Salish Subject > Object > Possessor > Oblique Object > Instrument > Locative  20  Kroeber (1991: 232) does not discuss objects of comparison, due to lack of data. Little will be said of them here, either, beyond showing that—regardless of whether they are a distinct grammatical role in Salish (a separate, as yet unaddressed question)—they are difficult if not impossible to relativize. See (73-75) below (p. 34). The only parameter of form that differs among RC types in St’át’imcets is the presence or absence of pronominal inflection corresponding to the target.  In the following section,  presence or absence of pronominal inflection in RCs is examined with respect to subject of transitive, subject of intransitive, direct objects, and possessors.  2.2.  Pronominal inflection of relative clauses  2.2.1. Subject of transitive In A-centered RCs, morphology corresponding to the ergative target is obligatorily retained, regardless of the person of the object. The post-head RC below has a 2sg object: (37)  áts’x-en-lhkan [ ta sqáycw-a see-trans-lsg. su det man-det ‘I saw the man who met you’  [ ta pzántsih*(as)a  det meet-trans-2sg. obj-3sg. conj-det  1]  (AA, GN, LT 1328)  If the object agreement were omitted from (37), it would be interpreted as 3sg, since this person is not overtly marked. In such structures where the object is 3sg and the subject is extracted, it is preferred for some speakers (notably RW) to use either a passive construction or the topical object marker -tall (which is in complementary distribution with the 3sg -as subject suffix, and which only occurs in extraction contexts—see (91-93) on page 46 in chapter 3). The latter strategy is exemplified below in (38). Sentences (39-40) show that it applies with equal force to wh-extraction, and it is preferred especially when coreference between the wh-word and a possessive pronominal—indicated in the following glosses by matching subscript indices—is intended (Matthewson 1993a: 3, 19):  21  (38)  (39)  (40)  s-Bill ]] nom-Bill conj]-det (TO/3sg. hit -t rdet  [ ti  áts’x-en-as [ ti sqáycw-a det man-det see-tr-3erg ‘he saw the man that hit Bill’ ats’x-en-táli swat ku see-tr-TO det who their mother?’ 1 saw 1 ‘who  (GN 141; GN, RW 218; LT 2581) i pl.det  *swat áts’x-en-as 1w conj see-tr-3sg. det who 1 mother?’ 1 saw their ‘who  skicez-f-ha mother-3p1.poss-det (BF, GN, RW 41)  i p1. det  skicez-I-ha nwther-3p1.poss-det (BF, GN, RW 41)  Interestingly, neither -tall nor passive may mark a relativized subject when the object is nonthird-person, as in the following examples corresponding to (37) above: (41)  (42)  *áts’xenlhkan [ ta sqáycw-a see-trans-lsg.su det man-det ‘I saw the man who met you’  [ ta pz-an-tsi-táli-ha  *âts’x.en4hkan [ ta sqáycw-a see-trans-lsg. su det man-det ‘I saw the man who met you’  [ta pz-.n-tsi-m-a det meet-trans-2sg. obj-pass-det  det meet-trans-2sg.obj-TO-det  1]  (AA, GN, LT 1329)  ]] (GN, LT 1329)  Because -tall appears only in third-person subject/third-person object sentences, Matthewson (1993b: 18-2 1) suggests that affixation of -tall is a disambiguation mechanism—rather than a GF-changing operation—since when a 3sg transitive subject is extracted (i.e., focused, questioned, or relativized) in a sentence having a third-person object (marked by -0), there 7 will not otherwise be any indication of which argument has been focused.  This seems  plausible, except that it does not explain why subject morphology is obligatorily retained in an RC like (37).  If the subject morphology were gapped, for example (*ta pzántsiha), the  sentence should still be unambiguously interpretable as ‘the one who met you’, since there would be an overt 3sg nominal to serve as the subject (the head of the RC, ta sqáycwa ‘the Craig (1977: ch. 7) gives the same analysis of a similar effect in Jacaltec. 22  man’).  The fact that ergative morphology must be retained when the object is non-third-  person argues against a ‘disambiguation strategy’ analysis of the retention of morphology, since no ambiguity would arise if it were omitted.  Obviously, some other element of the  grammar of St’át’imcets is implicated in the retention of pronominal morphology here; see §4.4 for a structural analysis.  2.2.2. Subject of intransitive In all Salish languages except those of the Southern Interior, person morphology corresponding to a relativized subject of intransitive is absent (Kroeber 1991: 235), but because 3sg absolutive is regularly marked by -0, the only way to determine whether the morphology is indeed absent in an intransitive-centered RC is to cleft a non-third-person nominal. Assuming that the residue of a cleft is an RC (Kroeber 1991: 184-187), the morphology corresponding to non-third-person targets in such constructions is indeed absent. Compare (43) and (44) in this regard, where the bracketed constituent in (44)—the residue of the cleft—is argued to be a headless RC: (43)  (44)  tsIcw-kan go-lsg. subj ‘I went’ tsukw t’ u7 part only ‘Only I went’  (LT 2438; Kroeber 1991: 262, citing van Eijk 1985: 279) s7ents lsg. emph  [ ti  tsIcw-a det go-det (AA, GN, LT, RW 1200; Kroeber 1991: 262, citing van Eijk 1985: 279)  Note that the lsg suffix -kan seen in the main clause (43) is absent from the RC in (44). However, Matthewson (1993b: 15) suggests that—despite the apparent lack of subject morphology in (44)—there could simply be null (indicative) 3sg agreement with the cleftee in (44), since in clefts where morphology is overt (e.g., when clefting the subject of a transitive), the subject morphology of the residue does not agree with the person of the cleftee:  23  (45)  *njlh ti snüwa 2sg. emph det focus ‘it was you who saw m&  ãts’x-en-ts-acw-a see-tr-lsg. obj-2sg. conj-det (ON 116, 1320, 1330, 1331; AA, ON, RW 1403; AA 2226, 2227; LT 2333)  Rather, there is obligatory 3sg agreement on the predicate: (46)  ti nilh snüwa 2sg. emph det focus ‘it was you who saw me’  áts’xents*(as)aS see-tr-lsg. obj-3sg. conj-det (ON, RW 60; GN 115, 1332; AA, GN, RW 1404; AA 2225; LT 2332)  Nevertheless, there has been variation for one consultant (GN), who at one time accepted (45) as grammatical. These agreement facts obtain even with defied plurals, as exemplified below: (47)  a.  b.  c.  *nilh snülap ats’x-en-ts-ál’ap-a [ i p1. det see-tr-lsg. objfocus 2pl. emph pl. conj-det 2 ‘it was you guys who saw me’ ats’x-en-tsal-itás-a ?nilh snülap [ i p1. det see-tr-lsg. obj focus 2pl. emph pl. conj-det 3 ‘it was you guys who saw me’  ] (AA, GN, LT, RW 1182)  ] (AA, ON, LT, RW 1183)  ]9 ats’x-en-ts-ás-a nilh snülap [ i p1. det see-tr-lsg. obj-3sg. conj-det focus 2pl. emph ‘it was you guys who saw me’ (AA, GN, LT, RW 1182)  Interestingly, (a) is ungrammatical when the RC residue agrees with the clefted 2pi pronoun. Third-person agreement inside the RC is preferred, with singular agreement (c) being preferred to plural agreement (b), despite the fact that the focused pronoun is plural. Another  The emphatic pronoun may appear sentence-initially without the focus marker nilh. In this respect, emphatic pronouns differ from other nominals, which cannot appear pre predicatively without a preceding nilh (as noted in § 1.3. 1). See the previous footnote. 8  24  reason that singular agreement is preferred here is that plural morphology is in complementary distribution with overt nominals and plural determiners (van Eijk 1985: 277-278). Under a 3sg-agreement analysis, then, sentence (44) would be more fully represented as follows: (48)  tsukw t’u7 part only ‘Only I went’  s7ents lsg. emph  [ ti  tsicw-ø-a j det go-3sg. intr. su-det  A possible argument against the 3sg-agreement analysis, though, is that the residue of clefts generally receive conjunctive, not indicative, inflection, and so absence of a 3sg pronominal in an RC would have to indicate a real gap, since conjunctive inflection is overt for 3sg (examine the subject suffix paradigms in appendix A, p. 107).  However, for some speakers,  conjunctive inflection is only a preference for RCs, and indicative inflection is also possible, as the following example shows: (49)  cillel ats’x-en-(lhk)-án-a ta nilh sqáycw-a ta run see-tr-indic-lsg.su-det det man-det det focus ‘it’s the man I saw that ran away’ (ON, RW 349; GN 1319; Matthewson 1993a: 14)  Although the indicative variant of (49) is ungrammatical for AA and LT, suggesting that their S-centered RCs do contain a gap, it would be desirable to be able to show that the other speakers have a gap in these structures as well. Another argument for the existence of a gap in S-centered RCs is needed. The correctness of the gap analysis is suggested by the fact that the intransitive-centered RC (44) is indeed ungrammatical with conjunctive inflection, regardless of whether there is lsg or 3sg agreement: (50)  a.  *tsu t’u7 s7ents ] [ ti tsfcw-an-a part lsg. emph det go-lsg. conj-det only ‘I am the only one who went’  25  (AA, ON, LT, RW 1200)  b.  t’u7 s7ents ] [ ti tsfcw-as-a part lsg. emph det go-3sg. conj-det only only one who went’ the am ‘I  (AA, GN, LT, RW 1201)  The fact that even the default 3sg agreement illustrated in (46) and (47) for subject of  transitives is not possible for intransitive subjects shows that relativization does not refer to a single ‘subject’ role that subsumes both transitive and intransitive subjects. Further evidence against the 3sg-agreement analysis in (48) is the behaviour of plural intransitive subjects when relativized, since plural morphology is overt in both the plain and conjunctive paradigms.  Because plural morphology is in complementary distribution with  overt nominals and plural determiners, as noted above, this can only be checked with headless RCs.  The following examples take this form, and they confirm that the morphology  corresponding to the relativized intransitive subject is obligatorily absent: (51)  qwatsats(*as)a áma-s-kan [i good-tr-lsg. su p1. det leave-3sg. conj-det ‘I like (the ones) who are leaving’  ] (LT 17 May 1994; cf. AA, GN 1324)  (52)  wa7(*as) güy’t ]° [i áma-s-kan good-tr-lsg. su p1. det prog-3sg. conj sleep ‘I like (the ones) who are sleeping’ (LT 17 May 1994; cf. RW 1075; AA, GN 1321)  (53)  wa7(*as) q’flhil ] [i áma-s-kan good-tr-lsg. su p1. det prog-3sg. conj run ‘I like (the ones) who are running’ (LT 17 May 1994; cf. RW 1076; AA, GN 1323)  Note that the presence or absence of subject morphology is not affected by whether the predicate is unaccusative, as in (51), or unergative, as in (52-53). 11 The fact that all of these sentences are ungrammatical with 3sg agreement constitutes strong evidence that the  Pursuant to the discussion in § 1.3.1, the enclitic half of the discontinuous determiner is absent here and in the following sentence because progressive wa7 is present. gay ‘t ‘sleep’ patterns with unergatives according to Davis (1993a). 10  26  grammatical variants of these sentences—and S-centered RCs generally—contain a gap corresponding to the relativized constituent.  2.2.3. Dfrect object Relativization of direct objects mirrors relativization of subject of intransitive, in that the morphology corresponding to the target is absent:  (54)  ti (nilh) snüwa 2sg. emph det focus ‘it was you that I saw’  ats’xen(*tsfn)ána see-tr-2sg. obj-lsg. conj-det (ON 119, 120; AA, ON, LT, RW 1405, 1406)  Both Kroeber (1991: 259-263) and Matthewson (1993b: 16-17) state that object morphology in object-centered RCs may be freely retained or omitted, but this is not certain. ON apparently accepted overt object morphology at one point (token 120), but ON and RW together rejected this variant of (54) during its most recent elicitation; AA also independently rejects it.  If  object morphology in these structures is indeed obligatorily absent, a possible analysis, suggested by Matthewson (1993b: 17), is that the missing object morphology does not represent a gap; instead, there is null 3sg agreement with the cleftee—analogous to the overt, 3sg agreement seen above in (46-47)—since 3sg objects are marked by -0. The structure of (54) would therefore be more fully represented as in (55): (55)  (nilh) snüwa ti 2sg. emph det focus ‘it was you that I saw’  ats’x-en-O-án-a see-tr-3sg. obj-lsg. conj-det  Because 3sg objects are never marked overtly on any predicate, Matthewson (1993b: 17) concludes that it is impossible to distinguish a gap from covert 3sg agreement in 0-centered RCs. A test for a gap is suggested by van Eijk’s (1985: 278-279) statement that conjoined proper noun complements generally require a plural affix on the predicate. This is exemplified below:  27  (56)  12 müta7 s-Bill wi s-John wa7 k’wzüsem{wit/*ø} nom-Bill prog work-intr- (3pl. su/3sg. intr) p1. det nom-John conj ‘John and Bill are working’ (LT 2334; van Eijk 1985: 278)  In (56), the predicate takes the 3p1 subject marker in agreement with the conjoined subject, ‘John and Bill’. Default 3sg agreement with the subject is in fact ungrammatical, unlike the sentences in (46-47) above. Because third-person object agreement is overt when the object is plural and the subject is non-third person (the 3p1 object affix is variously -tani and -wit—see the object suffix paradigm in appendix A, p. 106), it is possible to distinguish a gap from object agreement by clefting a conjoined object that requires plural agreement. For example, the following non-clefted sentence requires 3pl agreement with its object, ‘John and Bill’: (57)  wi see-tr- { pl. obj/3p1. obj/3sg. obj]-lsg. su p1. det 3 ‘I saw John and Bill’  s-John nom-John  mita7 s-Bill nom-Bill conj (LT 2335, 2336, 2337)  When the object is clefted, however, as in (58), the pattern of agreement is exactly opposite:  (58)  s-John müta7 s-Bill nilh wi foc pl.det nom-J conj nom-B ‘it was John and Bill that I saw’  lhkünsa i pl.det see-tr-{-Ø/3p1.obj]-lsg.conj-det now (LT 2338, 2339)  The fact that the 3p1 agreement morphology that was obligatory in (57) must be absent from its corresponding cleft in (58) establishes that there is a gap in the RC residue corresponding to the clefted object.  Note that this test requires a cleft construction and an RC residue; an  ordinary headed RC may not be used, since RCs may not modify proper nouns in St’át’imcets, as already seen in (33) on page 19. The same fact may be demonstrated when an emphatic pronoun and another NP are conjoined as a single object, since they require plural agreement. 12  In (59) below, a 2sg  This sentence was volunteered by LT as a translation for the English gloss indicated. AA prefers not to have the wi determiner before the compounded proper nouns, and also has 3sg and 3pl agreement in free variation. Additional comments on the differences between LT and AA with respect to person/number agreement follow in the text. 28  Unlike the clefting of a single 2sg  emphatic pronoun is conjoined with a proper noun.  emphatic pronoun as the subject of a transitive, as in (45-46), there cannot be 3sg agreement. 3 Rather, agreement is obligatorily second-person:’ (59)  snüwa áts 2sg. emp obj]-lsg. conj obj/3sg. see-tr- {2sg. obj/ pl. 2 ‘I saw you and Mary’  müta7 s-Mary nom-Mary conj (LT 2340, 2341)  When the conjoined object in (59) is clefted, however, the RC residue contains an obligatory gap corresponding to the object: (60)  müta7 s-Mary (nilh) snüwa foc 2sg. emp conj nom-Mary ‘it was you and Mary that I saw’  {ti/i} det/pi. det see-tr- (G/2pl. obj/2sg. obj] -lsg. conj-det (AA 2260, 2261, 2264; LT 2342, 2343)  This comprises additional proof that object-centered RCs have a gap corresponding to the target, not covert 3sg agreement, contra Kroeber’s (1991: 235) claim that object pronominals are ‘probably never obligatorily deleted in object-centered RCs.’ The tests above do not work for all speakers.  AA, for example, requires 3sg  agreement in both (57) and (59), and so the absence of object morphology for her in (60) cannot be taken as evidence for a gap in this structure. LT in fact has an ergative/absolutive asymmetry, requiring person/number agreement for conjoined absolutive arguments, but freely alternating between person/number agreement and default 3sg agreement for conjoined ergative arguments.  Although AA and LT are both Mount Currie speakers, the facts above  confirm that speakers of the same region may nevertheless vary with respect to subtle phenomena like agreement with compounded arguments.  It is important to note that this  LT has alternately preferred 2sg and 2pl agreement. In checking similar data with the Fountain speakers, GN stated that the number varies depending upon the number of the addressee, i.e., whether the referent of the proper noun is present as an addressee; if so, 2pl agreement is preferred. 14 The singular and plural determiners ti and i have been in free variation for both AA and LT in this sentence, although LT preferred plural i during the most recent elicitation. Perhaps the alternation is correlated with the number of the addressee, as suggested by the previous footnote. 13  29  represents true idiolectal variation since each speaker is systematic in her own agreement pattern, LT requiring person and number agreement with compounded absolutive arguments, and AA preferring default 3sg agreement in all cases, analogous to the A-centered RCs in (4547). Fountain speakers GN and RW are yet different with respect to agreement with conjoined arguments.  All conjoined proper noun complements—whether as ergatives or  absolutives—may freely have either 3sg or 3pl on the predicate.  Conjoined arguments  containing an emphatic pronoun are more complicated: there appears to be obligatory person agreement with the emphatic pronoun only with the transitive subjects and direct objects; these conjoined arguments are moreover ungrammatical in intransitive sentences unless focused, regardless of the agreement on the predicate. Data from these speakers are not cited because— although they corroborate the existence of a gap in 0-centered RCs—they have only been elicited once, and need to be confirmed individually with the consultants. The gap found in object-centered RCs for several speakers therefore parallels the gap found in intransitive-subject-centered RCs. Stated more generally, absolutive-centered RCs in St’át’imcets contain a gap corresponding to the target.  2.2.4. Possessor Word order of possessors is very restricted. The sentence in (61) shows that while a possessor and a head may be preposed together in a focus construction, (62) reveals that neither element may be brought into this position alone, since it creates a discontinuous constituent: (61)  s-Mary ta ciilel-a nilh ] [ ta sqáxa7-s-a det run. away-det det dog-3sg.poss-det nom-Mary foc ‘it was Mary’s dog that ran away’ (AA, LT 2304; Gardiner et al. 1993: 144)  (62)  a.  *njlh foc  s-Mary ta ctilel-a ta sqáxa7-s-a det dog-3sg.poss-det det run. away-det nom-Mary (AA, LT 2305; Gardiner et al. 1993: 144)  30  b.  *nilh foc  s-Mary nom-Mary  ta cülel-a det run. away-det  Similarly, a wh-possessor may not be extracted.  ta sqáxa7-s-a det dog-3sg.poss-det (AA, LT 2306; Gardiner et al. 1993: 144) The possessor and head may be preposed  only as a constituent: (63)  a.  b.  [  skicza7 ] swat mother who ‘whose mother left?’  *swat who  ku det  qwatsáts leave  qwatsáts leave (AA, GN, LT, RW 425; Gardiner et al. 1993: 145) ku det  skIcza7-s mother-3sg.poss (AA, GN, LT, RW 425; Gardiner et al. 1993: 145)  In this respect, St’át’imcets thus appears to be more conservative than the two other NIS languages, Nla’kapmxcIn (Thompson) and SecwepemctsIn (Shuswap), which do permit possessor extraction in certain contexts (Gardiner et al. 1993: 139-145). However, there is evidence for possessor extraction of some kind in St’át’imcets, since it is possible to relativize a possessor with unaccusative and unergative predicates; examples are given in (65-67) below.  For the sake of comparison, (64) is a non-relativized structure  corresponding to the relativized construction in (65). Note that the morphology corresponding to the relativized possessor is obligatorily retained, as permitted by Keenan and Comrie’s (1977: 66) Accessibility Hierarchy for relativization (35), since possessors are lower on the hierarchy than direct objects—which were shown in the preceding section to be gapped under relativization. (64)  (65)  zuqw plan ta skiiza7-s-a already die det child-3sg.poss-det ‘the woman’s child has died’  ta smülhats-a det woman-det  wa7 flal ta smüthats-a ta züqw-a det die-det prog cry det woman-det died crying’ is ‘the woman whose child  31  (GN 1335; LT 2440) ta sküza7*(s)a det child-3sg.poss-det (ON 1336; LT 2441)  (66)  (67)  wa7 alkst tu qwits ti sqáycw-a det prog work rich det man-det is rich’ working wife is whose ‘the man  5 ti sem7ám*(s)al det wfe-3sg.poss-det  xzüm-a ta smülhats-a qwits ta det big-det rich det woman-det ‘the woman whose house is big is rich’  ta tsftcw-s-a det house-3sg.poss-det  (GN 1340; LT 2442)  (GN, RW 1354; LT 2443)  These constructions are difficult for speakers to process, though GN, RW, and LT accept them as grammatical.  Interestingly, AA does not find these sentences grammatical, but more  readily accepts relativization of objects of comparison, which GN, RW, and LT do not accept; see (73-75) in the next section. It has not been possible to relativize the possessor of an A or an 0, presumably because possessors are already fairly inaccessible to relativization, and it is yet more difficult if the sentence has more than one overt argument. The asymmetric behaviour of possessors outlined above suggests that in St’át’imcets, the syntax of relativization differs from that of wh-extraction/focus—an important discovery, if correct, since it has not hitherto been obvious.  Kroeber (1991: 187), for example, regards  relative clauses and the residues of cleft constructions as nondistinct in Salish, since they appear identical in form. The data examined above suggest the summary in (68): Possessor-extraction contexts in St’ at’ imcets  (68)  I  Grammatical relativization  I  Ungrammatical focus; wh-questions; ordinary clauses  Finally, note that possessor relativization does not show stage- vs. individual-levelpredicate asymmetries of the kind reported for wh-possessor extraction in Nla’kapmxcmn, where individual-level predicates allow a wh-possessor to extract, but stage-level predicates apparently do not (Gardiner et al. 1993: 140-141). A similar asymmetry originally appeared  The determiner tu is underlyingly the usual ti/ta, the vowel merely having assimilated to the initial labial of the following wa7. 15  32  to exist for possessor relativization in St ‘at ‘imcets, as in (69), which relativizes the possessor of an argument of a stage-level predicate, wáz’arn ‘to bark’: (69)  *(wa7) wáz’-am we7áw’ ta smiMhats-a ta wa7 bark-intr shout prog det woman-det det prog ‘the woman whose dog barked is shouting’  ta sqáxa7-s-a’ 6 det dog-3sg.poss-det (ON, RW 1356)  Possessor relativization is ungrammatical without the progressive auxiliary wa7, but this ill formedness is merely aspectual in nature (Henry Davis, p.c.). The sentence is improved when  wa7 is present in the RC. Perhaps it is possible to reanalyze the Nla’kapmxcIn asymmetry as an aspectual one, namely, that stage-level predicates are better when the progressive auxiliary is present. Note, for example, that all of the Nla’kapmxcIn data cited by Gardiner et a!. (1993) lack 2u2tx or any other auxiliary.  More importantly, however, although the Nla’kapmxcmn consultant for  Gardiner et al. (1993), DU, made the distinctions reported by them on 2 June 1993, during the 7 most recent elicitation of these data she did not. The relevant sentences are the following:’ (70)  zum k jtx”-s swat k unr house-3sg.poss who unr big ‘whose house is big?’  (71)  wác’ama swat k who unr bark ‘whose dog barked?’  (72)  q’iyx swat k who unr leave ‘whose dog left?’  sqaqa?-s k unr dog-3sg.poss sqáqa?-s k unr dog-3sg.poss (DU, Nla’kapmxcfn, 19 April 1994)  The variant of this sentence without wa7 must also have the enclitic portion of the discontinuous determiner (ta wáz’-am-a)—as discussed in § 1.3.1—although this variant is irrelevant, since the sentence without wa7 is in any case ungrammatical (as noted in the text). Consult Thompson and Thompson (1992) for explanation of the grammatical abbreviations in these examples. 16  33  Gardiner et al. (1993: 140-141) report that (70) (their example (7))—having possessor extraction with an individual-level predicate—is grammatical, while (7 1-72) (their examples (5-6))—having possessor extraction with a stage-level predicate—is not. As indicated above, however, DU now regards all of these sentences as grammatical. If the judgments represented in (70-72) remain stable, then Nla’kapmxcfn and St’át’imcets may be more alike with respect to possessor extraction than has hitherto been suspected. Unfortunately, possessor extraction is marginal enough that it will likely be difficult to ascertain consistent judgments for this construction in either language.  2.2.5. Object of comparison Recall from page 21 that Kroeber (1991: 232) did not have any data on relativization of objects of comparison. These are ranked lowest on Keenan and Comrie’s (1977) Accessibility Hierarchy (35), and indeed it is difficult to relativize them in St’át’imcets.  This is not  surprising, given the difficulty in relativizing possessors, which are argued by them to be higher on the scale. (73) is a main clause comparative, and (74-75) attempt to relativize the object of comparison: (73)  (74)  (75)  ta sqáycw-a p’a7cw s-zác-al’qwem’-s nom-long-appear-3sg.poss det man-det more ‘the man is taller than the woman’  lhel than  sqaycw ta p’á7cw-a s-zác-al’qwem’ ?wa7 tayt prog hungry det more-det nom-long-appear man ‘the woman whom the man is taller than is hungry’  ta smülhats-a det woman-det (AA, LT, RW 1051) lhel ta smülhats-a than det woman-det (AA, LT, RW 1052)  ta sqáycw-a wa7 p’á7cw-a zác-al’qwem’-a lhel ta smülhats-a ?wa7 tayt prog hungry det man-det prog more-det long-appear-det than det woman-det ‘the woman whom the man is taller than is hungry’ (AA, LT, RW 1054)  Judgments differ on these sentences. LT and RW consider (73-75) to be ungrammatical. AA, however, marginally accepts them, and moreover expresses a preference for (75) over (74).  34  The consultants therefore can be seen to differ in the ease with which they relativize the lowerranked GFs: LT and RW allow relativization of possessors but not objects of comparison, while AA permits relativization of objects of comparison but not possessors. Of course, these data may be irrelevant to the predictions of the Accessibility Hierarchy if objects of comparison are not a distinct grammatical role in Salish—a separate question that has not been investigated.  2.3.  Summary  Consider the general patterning of morphology that corresponds to non-oblique targets of relativization, as summarized from the preceding discussion in (76): (76)  Morphology with non-oblique targets of relativization  Ergative 18 possessor subject of transitive morphology corresponding to target is obligatorily retained  I  Absolutive direct object subject of intransitive obligatory gap corresponding to target  I  A clear ergative/absolutive pattern has emerged.  Absolutive-centered RCs obligatorily omit  person morphology. Ergative-centered RCs, however, are complicated in two respects: first, subject morphology corresponding to the target is retained, and second, there exists special morphology (the topical object marker -tall) that can occur only in this type of RC. Not only are ergative-centered RCs therefore marked more explicitly than absolutive-centered RCs (by retaining the 3sg subject suffix -as), but they can also employ a distinct strategy of relativization (affixation of -tall).  These are exactly the properties claimed by Keenan and  Comrie (1977) to show that such a relativized constituent is lower on the Accessibility Hierarchy than higher roles, which are relativized more easily. The Accessibility Hierarchy  18  Possessor appears under the ergative column because it behaves the same as A with respect to retention of morphology when relativized. It is not clear whether this fact has the same explanation in both ergative- and possessor-centered RCs, but see Allen (1964) for a survey of languages in which there is a formal correspondence between ergatives and possessors.  35  for Salish (36) should probably undergo revision, as already suggested by Kroeber (1991: 233), with the subject GF split into ergative and absolutive: (77)  Absolutive > Ergative > Possessor > Oblique Object > Instrument > Locative  Keenan and Comrie (1977: 80 ff.) are reluctant to assign distinct positions on the Accessibility Hierarchy to ergative and absolutive (preferring instead to maintain the single role of subject), since distinct positions would predict the existence of relativization strategies that could apply only to absolutives, whereas such strategies are not attested in any language. For this chapter, though, the Accessibility Hierarchy has no theoretical status.  The scale in (77) serves as a  useful description of the facts in St’át’imcets relativization, but a structural explanation for the ranking in (77) is offered in §4.4. An ergative pivot is apparently at work in St’át’imcets relativization, at least morphologically, and this presents a problem for a theory of relativization that assumes a notion of subject that is defined on {A,S}. The status of the morphology in ergative-centered RCs is not clear, though; if -as were not a subject suffix, for example, then the ergative centered RCs would more plainly have a gap, exactly as do the absolutive-centered RCs. The analysis of RCs in §4.4 will assume that ergative-centered RCs do indeed involve extraction of the relativized constituent, and such an analysis is surely desirable, not only on general theoretical grounds, but because there is indirect evidence for ergative extraction in Halkomelem, a Coast Salish language that omits the person morphology corresponding to relativized ergatives (Gerdts 1988: 82-83)—exactly the environment in which St’át’imcets seems to retain such morphology.  Nevertheless, in having a restriction on relativization of  ergatives (whether it is analyzed as morphological or syntactic—and the division is not always clear), St’át’imcets is only one of many such languages, including Tzutujil (Dayley 1985: 2102 15), Dyirbal (Dixon 1979: 127-128), Tongan and other Polynesian languages (Chung 1978: 37-44), Jacaltec (Craig 1977: ch. 6), Chukchee (Comrie 1979: 225-226, 229-230), Coast Tsimshian (Mulder 1988: §3.10; 1989: 426-428), Inuktitut (Creider 1978: 99 ff.; Johns 1992:  36  72), Yup’ik Eskimo (Payne 1982: 85-87), and Greenlandic Eskimo (Woodbury 1975, cited by Keenan and Comrie 1977: 83; cf. Smith 1984). However the facts presented in this chapter are analyzed, it is important to bear in mind that relative clauses in Stát’imcets resemble RCs in other languages, in that they exhibit an obligatory gap. This gap mirrors the extraction of relative pronouns and empty operators in Indo-European languages like English.  37  Chapter 3 Conjunction  3.1.  Introduction  This chapter examines the properties of pronominal coreference across sentential conjuncts in St’át’imcets. As explained in chapter 1, Dixon (1979) shows that languages may vary in their choice of syntactic pivot for processes of conjunction (‘coordination’ in his terminology) and subordination. If a syntactic process treats A and S alike, in opposition to 0, it is said to have an accusative pivot. On the other hand, if S and 0 (but not A) are uniquely identified by a syntactic operation, the process is said to have an ergative pivot. Recall the chart in (1) on page 3. It is also possible for a language to be neither ergative nor accusative with respect to these processes; Coast Tsimshian, for example, has been argued by Mulder (1988: §3; 1989) to lie somewhere in the middle of a continuum between syntactic accusativity and ergativity. Chapter 2 has shown that an ergative pivot is at work in one of the major subordination types in St’ät’imcets, relativization.  This chapter examines conjunction (the other major  process besides subordination claimed by Dixon (1979) to vary cross-linguistically in its pivot) in St’át’imcets in order to determine the behaviour of {A,S,O} with respect to each other in this construction. It will be seen that conjunction, unlike relativization, is neither ergative nor accusative.  Although one speaker sometimes appears to have an accusative pivot, it is  probably best regarded not as a syntactic pivot, but instead as a constraint on the discourse roles of coreferent NPs (Matthewson 1993a). For all speakers, coreference is most forcefully constrained by the one-nominal interpretation law (Gerdts 1988), which compels a unique interpretation in conjuncts that contain a single overt nominal. St’át’imcets has many ways of marking conjunction; for a sampling of subordinating and coordinating conjunctions, see van Eijk (1985: 172-173, 211-213, 217-219, 252, 270273).  In order to restrict the domain of investigation and thus facilitate the comparison of  sentences, all of the sentences presented below have the subordinating conjunction nilh ‘then; so, then’, which is not to be confused with the homophonous focus-marker mentioned in  38  chapter 1 (p. 7) and seen throughout chapter 2 in cleft constructions. The conjunction nil/i is clearly subordinating, rather than coordinating,  as it induces subordinate inflection  (specifically, nominalization) in the non-initial conjunct.  Inflection of clause types in  St’át’imcets was discussed in §1.3.1 and summarized in the chart in (27) on page 12. There is another methodological point that is worth noting. Although sentences having both a subject and an object are rare in discourse (van Eijk 1985: 262)—since St’át’imcets is a radical head-marking language and thus allows arguments to be referenced solely by morphology on the predicate—entirely third-person sentences without at least one overt nominal argument are highly disfavored discourse-initially. For this reason, in eliciting most of the data, a discourse context was provided immediately preceding the sentences of interest; this context was necessary especially before sentences having no overt nominals in the first conjunct.  The context that was used in each case is always cited below.  Even transitive  sentences having a single overt nominal—for example, (18-19) in chapter 1 (p. 10)—are disfavored discourse initially, and are regarded as felicitous only when there is a discourse context that can supply a subject for such sentences.  For elaboration of this point, see the  discussion concerning (124-126) in chapter 4 (pp. 63 ff.).  3.2.  Conjuncts with non-third-person  It is easy to show that there is no single pivot for conjunction that is either ergative or accusative by constructing null-pronominal sentences whose second, transitive conjunct has one pronominal affix (marking either A or 0) that is uniquely non-third-person. Consider, for example, sentences (79-80), which have (78) as their context: (78)  smelhmülhats-a i qwatsáts p1. det women-det leave 1 left’ ‘the women  39  (79)  (80)  19 s-7ats’ x-en-tsál-itas nilh p’ an ‘t-wit obj-3p1. su nom-see-tr-lsg. return-3p1. intr conj 1 returned and they saw me’ ‘they nilh s-7ats’ x-en-tánih-an ‘an’ t-wit nom-see-tr-3p1. obj-lsg. conj return-3pl. intr conj 1 returned and I saw them’ ‘they  (AA, GN, RW 1382)  (AA, GN, RW 1384)  Both (79) and (80) have a 3pi subject in the first conjunct, coreferent with I smelhmiUhatsa ‘the women’. In the second conjunct of each sentence is a single 3p1 affix: it marks the subject pi argument of the second conjunct may in (79), and the object in (80). In each sentence, the 3 corefer with the 3pi subject of the first conjunct, as indicated in the glosses by subscript indices. Note, importantly, that (79) and (80) differ minimally in the person morphology of the second conjunct; the conjoined predicate in neither of these sentences needs grammaticalfunction-changing morphology, e.g., passive -m or the topical object marker -tall.  If  St’át’imcets conjunction had an ergative or an accusative pivot, it might be expected that a process like passive or antipassive would operate in either (79) or (80) in order that such a GF changing operation could feed the syntactic pivot (Dixon 1979: 120-130). There are similarly no asymmetries in sentences having two intransitive conjuncts; a pronominal S of the first conjunct may freely corefer with an S of the second conjunct, irrespective of unaccusative/unergative distinctions.  The subject clitic is obligatory in each  conjunct of the following examples; it may not be deleted under identity with the clitic in the first conjunct:  LT (2445, 2446) prefers that this sentence and the next one begin with the past complementizer I, which also triggers conjunctive -as on the predicate: i p’án’t-wit-as ‘when they returned’. Coreference obtains as indicated in the text by the subscript indices. 19  40  (81)  (82)  (83)  O 2 qwatsats*(kalh) nilh s-p ‘an’ t*(ka1h) conj nom-return-lpl.poss leave-ipi. intr ‘we left and then we returned’  (AA, GN, RW 1389)  skwis*(kalh) nilh mátq-kalh walk-ipi. intr conj nom-fall-lpl.poss ‘we walked and then we fell’ nilh ka-kwfs-kalh-a ooc-fall-lpl. intr-ooc conj ‘we fell and then we slept’  (AA, GN, RW 1391)  l 2 s-ka-güy’ t*(ka1h)a nom-ooc-sleep-lpl.poss-ooc (AA, GN 1509)  (82) has an unergative predicate in its first conjunct, as does the second conjunct of (83); the remaining conjuncts in (8 1-83) have unaccusative predicates.  Because coreference between  conjuncts is free in all of these combinations of different kinds of predicates—and because none of them show signs of GF-changing operations—there is again no evidence for a syntactic pivot. Conjunction in St’át’imcets therefore appears to be neither ergative nor accusative, In this respect, St’át’imcets probably  unlike relativization, which is ergative (chapter 2).  behaves like most of the world’s languages. Dixon (1979: 129) remarks: It may be that some languages cannot be clearly characterized, at the syntactic level, in terms of the ergative/accusative continuum. That is, processes such as coordination [= conjunction] may not operate in terms of well-defmed constraints like those Certainly, some languages have a applicable to Walmatjari and Dyirbal. considerable set of well-defmed syntactic constraints, which facilitate a clear judgment of their position on the ergative/accusative syntactic scale; but others have more fluid conditions that provide slimmer evidence for judgment. For instance, coordination .  .  .  LT (2447, 2448) offers an interesting correction to this sentence and the next, apparently with two conjunctions, as exemplified in (i): (i) qwatsats*(kalh) nilh t’u7 müta7 sp’án’t*(kalh) conj part conj nom-return-lpl.poss leave-ipi. intr then we returned’ left and ‘we 21 LT (2449) prefers the discontinuous out-of-control auxiliary to be absent from kwis ‘fall’ here, as it is in the previous sentence in the text: (i) kwIs-kalh nilh t’u7 skaguy’t*(kalh)a fall-ipi. intr conj part nom-ooc-sleep-lpl.poss-ooc ‘we fell and then we slept’ 20  41  may largely follow semantic, stylistic, or discourse-organization preferences, rather than conforming to any strict syntactic matrix. The last sentence describes conjunction in St’át’imcets well, as will be seen in the next section.  3.3.  Conjuncts with third-person subject and object  Coreference across conjuncts is more restricted in sentences having third-person subjects and objects with various combinations of proper nouns.  Because consultants have the most  consistent judgments for coreference in sentences having two transitive conjuncts, these will be examined first (3.3. 1). Section 3.3.2. then examines sentences in which one of the conjuncts is intransitive; the data here are less firm, and so it is harder to base generalizations on them.  3.3.1. Conjoined transitives Parallelism constraint on discourse functions  Sentences having two transitive conjuncts display striking evidence for a parallelism constraint on discourse functions. This is seen most clearly in sentences having no overt NPs, as in (85) below.  This sentence is preceded by the context in (84), as are the others in this section  (3.3.l).22  (84)  cmán’-twal’-wit. Wa7 wi7 müta7 s-John. Pz-án-twal’ wi s-Bill enerny-recip-3p1.su prog ernph nom-Bill conj norn-John meet-tr-recip p1 1 and Johns met each other. They’re enemies.’ ‘Bill  22  A reciprocal context with same-sex participants (‘Bill and John met each other’) is order to minimize the possibility that one participant will be more prominent in the used in discourse and unduly affect the coreference possibilities in the sentences of interest. Initial attempts to elicit conjoined transitives used asymmetric discourse contexts, with male and female participants and predicates like ts ‘urn ‘qsán’ ‘kiss’, and coreference judgments consequently turned out to be asymmetric in unusual ways that suggested that the consultant was basing them on facts that she thought would be more likely to obtain between men and women (as she later confirmed).  42  (85)  nilh áts’ x-en-as see-tr-3erg conj 1 saw him, and a. ‘he , and 1 b. ‘he saw him  s-qvlqvl-ts-mf& -as 23 nom-bad-mouth-appl-3sg. conj }’ 1 then {he 1 swore at him/*he swore at him / 1 him *he him}’ at swore then {he swore at (AA, GN, RW 1595)  Because the first conjunct of (85) has no overt NPs, it may be ambiguously interpreted as to who is seeing whom. However, given a specific interpretation of the first conjunct, there is a tendency, for all speakers, to interpret the subject and object of the second conjunct as referring to the same subject and object of the first conjunct. A similar phenomenon can be observed in complex clauses. In examining pronominal coreference into relative clauses in St’át’imcets, Matthewson (1993a) independently found that—although coreference into relative clauses is not restricted by Condition C of the binding theory (see Chomsky 1981; 1986; etc.)—it is restricted by a constraint on the discourse functions of the coreferent items. Consider the following paradigm: (86)  ta xwis-ás-a Mary ta sqáycw-a ts’um’-qs-án’-as kw-s det love-3sg. conj-det Mary det man-det lick-nose-tr-3erg det-nom topic) (topic. 1 kissed the man that she loves’ a. ‘Mary *‘the (non-topic. non-topic) ’ 1 1 kissed Mary man who loves her b. 1 kissed the man who loves her non-topic) (topic. ’ 1 c. *‘Mary non-topic) (topic. ’ 1 1 loves kissed Mary d. *‘the man that she Matthewson et al. 1993: 229) (GN, RW 386; Matthewson 1993a: 19; .  .  .  .  .  .  .  .  Kinkade (1989; 1990) shows that topics (‘old’, presupposed information in the discourse) in Salish languages typically occupy the subject position of clauses, whereas non-topics (‘new’ information) appear in object position.  Topics and non-topics may appear in non-canonical  positions (object and subject respectively) if the predicate is passivized or affixed by the topical object marker, though in St’át’imcets the latter strategy may be used only in extraction contexts, as noted above. (For examples of passivization switching discourse topics, see (90) and (107) in this chapter.) Because the predicates in (86) are neither passivized nor affixed by the topical object marker, the subjects in the interpretations (a-d) correspond straightforwardly 23  Qvlqvl-ts-mmn’ ‘to swear at (tr)’ here and throughout represents the Fountain pronunciation; Mount Currie speakers retract and glottalize the lexical suffix -ts ‘mouth; language’, as noted in §1.3.2. 43  to topics, and objects to non-topics. As indicated, the preferred interpretation is the one in which ‘Mary’ and the coreferent null pronominal are the topics of their clauses (a). The other interpretations (b-d) differ in this respect, and are dispreferred. Reading (b), with ‘Mary’ and the coreferent null pronominal as non-topics, is also allowed for some speakers, and was the first translation offered for (86) by AA, presumably in accord with her fairly strict VOS 24 order. Data like those in (85) and (86) suggest that St’át’imcets is subject to the following restriction (Matthewson 1993a: 20-21; cf. Matthewson et al. 1993: 228): (87)  Parallelism Constraint on Discourse Functions (PC) For two items to corefer, they must both fulfill the same discourse function (either In addition, there is a preference for both  topic of the discourse or non-topic).  coreferential elements to fulfill the topic of the discourse function. Constraint (87) can be observed in the two other MS languages, Nla’kapmxcfn and Secwepemctsfn (Matthewson et al. 1993: 228-230), and a similar parallelism restriction obtains in Nuxalk (Bella Coola), another Salish language (Davis and Saunders 1984). It should be noted that some idealization of the data is necessary in order to accept (87). The starred interpretations in sentences like (85), and in the remaining sentences of this section, are ungrammatical only for RW.  , the 25 For the other consultants (AA and GN)  interpretations marked by the asterisk (*) should be taken only as less obvious and mildly dispreferred. When translating the sentences in this section into English, for example, these consultants spontaneously provide glosses in accord with (87), but will concede that the non parallel reading is possible if prompted for it. 24  AA prefers the only the determiner k in front of the proper noun, and also a progressive auxiliary inside the relative clause, which requires the enclitic -a portion of the discontinuous determiner to drop. ta wa7 xwis-as ta sqáycw-a ts’um’-qs-án’-as k Mary (i) det prog love-3sg. conj det man-det lick-nose-tr-3erg det Maiy 25 BF shares the same judgments as AA and GN. Thanks to Henry Davis for checking these data with her. 44  The parallelism constraint is expected to hold across conjuncts if one conjunct has two overt NPs. This is indeed the case, as (88-89) show: 26 Bill s-qvlqvl-ts-mIn’-as kw-s John kw-s nom-bad-mouth-appl-3sg. conj det-nom John det-nom Bill }’ 1 and then {Bill swore at John/*John swore at Bi11 }’ 1 and then {Bill swore at John/John swore at Bill (AA, GN, RW 1597)  (88)  áts’x-en-as nilh see-tr-3erg conj 1 saw him, a. ‘he , 1 b. *‘h saw him  (89)  John kw-s Bill nilh s-qvlqvl-ts-mfn’-as áts’x-en-as kw-s see-tr-3erg det-nom John det-nom Bill conj nom-bad-mouth-appl-3sg. conj 1 swore at him/*he swore at him }’ 1 a. ‘Bill saw Johns, and then {he *‘John /he swore at him}’ him saw Bill , and then {he swore at 1 1 b. (AA, GN, RW 1596)  Recall from chapter 1 that the preferred word order in elicitation is VOS, especially in sentences like these, in which only the word order (not animacy requirements of the predicate) is available to disambiguate the grammatical roles of the overt NPs.  This word-order  preference immediately accounts for the ungrammaticality of interpretation (b) for (88), since it would require VSO order in the second conjunct. Interpretation (a), however, respects the VOS order of the second conjunct (‘Bill swore at John’) and—as expected in accord with (87)—imposes parallelism on both conjuncts, so that the subject of qvlqvl-ts-mIn’ ‘to swear’ (viz., ‘Bill’) must also be the subject of áts’x-en ‘to see’. Sentence (89) differs from (88) only in that the two overt NPs are in the first conjunct, rather than in the second. The same point is illustrated here: interpretation (b) is not allowed, since it would require VSO order; rather, the VOS interpretation of (a) is preferred for the first conjunct (‘Bill saw John’), in which case the interpretation of the second conjunct must  26  Here and throughout this chapter, the cited forms are the ones that originate from GN and RW (Fountain speakers). Although AA and sometimes LT are cited with each form as one of the consultants, the sentences that they examined differed minimally in that proper nouns were preceded by either the determiner k(w) alone or the nominalizer -s alone, but never by both together, as are the proper nouns in this sentence. It is hard to be sure what characterizes this variation between speakers; see § 1.3.2 for further examples and discussion. 45  27 have a null pronominal subject coindexed with the subject of the first conjunct.  An  interpretation in which the object of the first conjunct is the subject of the second conjunct (‘John swore at Bill’) is ungrammatical, although this reading in (89a) can be saved with a passivized predicate in the second conjunct, as exemplified below: (90)  Bill nilh s-qvlqvl-ts-mfn’-em áts’x-en-as kw-s John kw-s see-tr-3erg det-nom John det-nom Bill conj nom-bad-mouth-appi-pass 1 was sworn at by him’ 1 saw Johns, and then he Bill (AA, GN, RW 1596)  Another example of passivization switching discourse topics is in (107) below. Note that the ungrammatical interpretation in (89a) cannot similarly be saved by affixation of the topical object marker -tall; such a sentence is ungrammatical under any reading: (91)  *áts’xenas see-tr-3erg  John kw-s det-nom John  Bill nilh s-qvlqvl-ts-mfn’-tali kw-s det-nom Bill conj nom-bad-mouth-appl-TO (AA 1596)  Although -tali is a cognate of the reflexes of Kinkade’s (1989: 38) Proto-Salish *...wali, the St’át’imcets cognate appears only in extraction contexts (Matthewson 1993a: 4-5). This was noted with respect to examples (38-39) in chapter 2 (p. 22). sentence (22) in chapter 1 (92)  nilh t’u7  s-e-s  (p.  Consider also the passive  10), repeated below as (92): k’al’-em-mIn-em  ti  syáqts7-a  foc part nom-prog-3sg.poss wait-intr-appl-pass det woman-det ‘he is being waited for by the woman’  (LT 2330)  The equivalent sentence with -tali is ungrammatical, precisely because there is no extraction: (93)  *nilh t’u7 s-e-s k’al’-em-mfn-tali foc part nom-prog-3sg.poss wait-intr-appl-TO ‘he is being waited for by the woman’  27  ti syáqts7-a det woman-det (LT 2430)  LT, who often has VSO order in elicitation, allows interpretation (a) like the other , 1 speakers, but additionally permits the reading of (b) which has parallelism, ‘John saw Bill ’. 1 and then he swore at him 46  Interaction with the one-nominal interpretation law  The remaining class of sentences with two transitive conjuncts are those that contain a single NP in one or both of the conjuncts. Recall from chapter 1  10) that St’át’imcets is subject  to a rule of one-nominal interpretation (23), repeated for convenience in (94): (94)  One-Nominal Interpretation Law (ONO) In the absence of marking for other persons, a single 3rd person nominal is interpreted as the absolutive.  That is to say, in sentences having a single NP and only third-person morphology on the predicate, the NP is interpreted as absolutive, i.e., as the only argument (subject) of an intransitive predicate, or as the direct object of a transitive predicate. The one-nominal interpretation law can be seen to apply in the following sentence:  (95)  áts’x-en-as kw-s John nilh s-qvlqvl-ts-mfn’-as see-tr-3erg det-nom John conj nom-bad-mouth-appl-3sg. conj a. ‘he saw Johnj, and then {he swore at him/*he swore at him}’ 1 swore at him}’ , and then {he swore at him/he 1 b. *‘John saw him (AA, GN, LT, RW 1598)  The (b) interpretation is excluded because there is a single NP (‘John’) in the first conjunct, which must be interpreted as the absolutive, i.e., as the direct object. Because (95) can only be grammatical when ‘John’ is the object of the first conjunct, the null pronominal object of the second conjunct in such a case must likewise corefer with the object ‘John’ of the first conjunct, in accord with the parallelism constraint (87). The obvious question that arises with a constraint like parallelism is whether it interacts with the one-nominal interpretation rule, since it is possible for the parallelism constraint and the one-nominal interpretation law to impose conflicting demands in a single sentence. The relevant context is a sentence having a single overt NP in each conjunct, i.e., structures of the following form, where pro marks a null pronominal argument, and NP 1 is not coreferent with NP (the order of elements within brackets is irrelevant): 47  (96)  . . . pro. . . j conj [s. . . preci. . . NP . . . pro. 1 [s. . . pred.. . . NP  .  .]  Sentences having this form might be expected to be ungrammatical, or at least dispreferred, since—no matter what interpretation is assigned to them—a constraint will be violated.  If  {NP,NP} are both direct objects in (96), the one-nominal law will be satisfied, but ,NP} bear different GFs or discourse 1 parallelism will be violated. On the other hand, if {NP functions, parallelism will be satisfied (assuming also the appropriate indexing of pro in each conjunct), but the one-nominal law will be violated. A sentence having the structure of (96) is given in (97); rather than being dispreferred, it does have a unique grammatical interpretation: (97)  kw-s Bill áts’x-en-as kw-s John nilh s-qvlqvl-ts-mfn’-as see-tr-3erg det-nom John conj nom-bad-mouth-appl-3sg. conj det-nom Bill /(ii) *Bil1 swore at him’ 1 1 saw John, and then (i) hej swore at Bi11 a. ‘he 1 swore at him’ /(ii) Bill 1 b. *‘John saw him, and then (i) he swore at Bi11 (AA, GN, RW 1593)  That (97) has a grammatical interpretation at all is striking, since it illustrates that the parallelism constraint (87)—whatever its structural interpretation might be—is violable. Specifically, it may be violated when such violation leads to the satisfaction of the one-nominal interpretation law. The relatively low ranking of the parallelism constraint (87)  expresses two  generalizations about coreference across transitive conjuncts in St’át’imcets, the most obvious of which is that, for all but one consultant, parallelism is merely a preference; speakers permit interpretations that violate it, even in sentences about which the one-nominal interpretation law has nothing to say.  More importantly, however, it captures the difference between the  grammars of different consultants. Given only sentence (85), for example, it is unusual that violations of parallelism should be ungrammatical for a single consultant, but possible for the others. However, sentences like (97)—in which parallelism and the one-nominal law impose conflicting demands—show that PC is violable even for the more restrictive consultant, who would otherwise appear to have a grammar different from other speakers of the same 48  language. (97) therefore reveals that all of these speakers share the dominant ranking of the one-nominal interpretation law. One way of encoding the relationship between these constraints is to employ the formalism of Optimality Theory (Grimshaw 1993; Prince and Smolensky 1993), which hypothesizes that grammatical constraints are in principle violable, and that minimal violation is allowed when it leads to the satisfaction of higher-ranked constraints.  Although the  formalism of this theory is useful for explicating the coreference data of St’át’imcets, it is used merely for expository convenience, and it has nothing of interest to say about the nature of these constraints, whose interpretations are tied to syntactic structures (examined in §4.3). Because the one-nominal constraint has precedence over the parallelism constraint in evaluating whether a given interpretation is grammatical, the former is ranked above the latter, as formalized in (98), where >> is interpreted as ‘dominates’: (98)  ONO >> PC  Each possible interpretation of (97) is a member of the ‘candidate set’, and is evaluated with respect to its satisfaction of each constraint. Following the conventions of Grimshaw (1993), each candidate is listed next to the set of constraints, where left-to-right order reflects their ranking. Each occurrence of an asterisk (*) before a constraint indicates a single violation of that constraint; lack of such a mark indicates that the constraint is satisfied or irrelevant. The optimal candidate—which is the grammatical interpretation—is identified by the dollar sign  ($).  (97) is therefore evaluated as follows:  (99) a. b. c. d.  áts ‘x-en-as kw-s John nilh s-qvlqvl-ts-mmn ‘-as kw-s Bill (= 97) ‘he saw John, and then he swore at Bill ’ 1 1 saw John, and then Bill swore at him’ ‘he ‘John saw him , and then hej swore at Bill 1 ’ 1 ‘John saw him , and then Bill 1 1 swore at him’  (=97a.i) (=97a.ii) (=97b.i) (=97b.ii)  ONO *PC *ONO PC *ONO PC **ONO *p  $  Notice that the total number of constraint violations is not relevant to the evaluation of candidates. The candidates (a-c) each have a single constraint violation, yet only one of them  49  (a) is grammatical. The fact that candidates (b-c) satisfy PC is not sufficient, since it is better to satisfy the higher-ranked ONO, even if it requires that the lower-ranked PC be violated, as it is in (a).  The contrast between (a) and (b-c) illustrates a clash between competing  constraints, and shows that the conflict is indeed resolved in accord with the ranking in (98). A slight complication arises with the sentence in (100).  As indicated in the glosses,  there are two grammatical interpretations of this sentence. However, as the constraint tableau in (101) shows, one reading (c) should be preferred over the other (a): (100) áts’x-en-as see-tr-3erg 1 saw a. ‘he b. ‘he saw  nilh s-qvlqvl-ts-mmn’-as conj nom-bad-mouth-appl-3sg. conf him, and then (i) he swore at Bill/(ii) /(ii) 1 , and then (i) hej swore at Bill 1 him  Bill kw-s det-nom Bill *Bill. swore at him’ *Bill swore at him’ (AA, ON, LT, RW 1594)  (101) áts ‘x-en-as nil/i s-qvlqvl-ts-mmn ‘-as kw-s Bill (=100) a. b. c. d.  ‘he 1 1 ‘he ‘he ‘he  saw saw saw saw  him, him, , 1 him , 1 him  and and and and  then then then then  ’ 1 he swore at Bill 1 swore at him’ Bill ’ 1 he swore at Bill Bill swore at him’  FI 0  (=lOOa.i) (= lOOa.ii) (=lOOb.i) (= lOOb.ii)  ONO *p *ONO PC ONO PC *ONO *PC  $  The checkmark (0) indicates the interpretations that ON and RW allow. Candidates (b, d) are correctly excluded as ungrammatical, since they violate the highly-ranked ONO, but candidate (a) is wrongly excluded, since it violates PC. Candidate (c) does not violate any constraints. It should be determined whether reading (c) is preferred over that of (a); if so, the tableau in (101) correctly predicts a gradation in grammaticality. Note, however, that in checking this sentence with AA and LT, interpretation (c) was uniquely grammatical, and interpretation (a) was not allowed, in accord with the candidate evaluation in (101). Grammaticality judgments for these sentences are subtle, and it is likely that the data concerning this point will continue to vary across speakers and across elicitation sessions. The constraint-ranking motivated above accords with the fact that the one-nominal interpretation law is a salient property of St’át’imcets grammar, applying generally throughout the language, in simple sentences as well as in compound and complex ones. established that this is the case, it remains to be explained. 50  Having  Of course, it will likely remain  unexplained why the parallelism constraint has varying importance across speakers, although it accords with Dixon’s (1979: 129) statement, quoted above on page 41, that such as coordination [Sjome languages.  ..  ‘.  .  .  processes  [= conjunction] may not operate in terms of well-defined constraints. have more fluid conditions... [F]or instance, coOrdination may largely  follow semantic, stylistic, or discourse-organization preferences, rather than conforming to any strict syntactic matrix.’ St’át’imcets conjunction takes on yet more of this character in the next section, which examines intransitive conjuncts. A table summarizing the various combinations of null and overt NPs and their coreference possibilities across transitive conj uncts is given in (152) in chapter 4 (p. 82).  3.3.2. Intransitives This section is divided into two parts: § examines data that have an intransitive predicate in the first conjunct, and § presents data that have an intransitive predicate in the second conjunct.  Intransitive in first conjunct  If subject of intransitive (S) is in the same structural position as subject of transitive (A), both S and A might be expected to pattern alike with regard to the parallelism constraint (87). Consider in this respect the sentence in (103), which is prefaced by the context in (102): (102) wa7 k’1’-em kw-s prog wait-intr det-nom ‘Johns is waiting’  John John  Bill nilh s-7áts’x-en-as (103) p’an’t kw-s return det-nom Bill conj nom-see-tr-3sg. conj ’ 1 1 saw him/(ii) *he saw him 1 returned and (i) he ‘Bill  51  (AA, GN, RW 1586)  The second conjunct of (103) does not have any overt nominals, and so the one-nominal 28 interpret the interpretation law (94) is not implicated here. For this sentence, all speakers second conjunct in accord with parallelism, suggesting that A and S are indeed in the same position. However, consider (104) below, which is complicated by the addition of a single NP in the second conjunct. (104)  John kw-s s-7áts’x-en-as kw-s Bill nilh return det-nom Bill conj nom-see-tr-3sg. conj det-nom John ’ 1 1 saw John/(ii) ?John saw him ‘Bill returned and (i) he (AA, GN, RW 1584)  p’an’t  Reading (i) was spontaneously volunteered by all speakers as a translation of (104), although reading (ii) was unexpectedly accepted by GN and RW after it was suggested to them (in violation of the one-nominal interpretation law). The same judgments are given by BF (Henry Davis, p.c.). For one speaker, AA, reading (ii) is ungrammatical, as expected, given the high ranking of the one-nominal interpretation law demonstrated in § It is not clear why several speakers would accept interpretation (ii) for (104), especially since a sentence having the same structure, but with different names, was on two occasions interpreted by GN and RW as expected. The context for (106) is (105): k’ál’-em (105) wa7 wait-intr prog ‘Mary is waiting’  kw-s det-nom  Mary Mary  Mary kw-s Bill nilh t’u7 s-áts’x-en-as (106) p’an’t kw-s return det-nom Bill conj part nom-see-tr-3sg. conj det-nom Mary 1 returned and (i) he saw Mary/(ii) *Mary saw him’ ‘Bill (AA, GN, RW 1392) In (106), the one-nominal interpretation rule pressures ‘Mary’ in the second conjunct to be the direct object, thus excluding the reading in (ii). This reading can be saved, however, if the predicate is passivized:  Except BF and LT, who accept both interpretations (i) and (ii). Thanks to Henry Davis for checking this point with BF. 28  52  s-Mary Bill nilh t’u7 s-áts’x-en-em (107) p’an’t kw-s nom-Mary nom-see-tr-pass part conj return det-nom Bill 1 was seen by Mary’ ‘Bill returned and he  (AA, GN, RW 1393)  Another example of passivization switching discourse topics is given above in (90); cf. also (2 1-22) in chapter 1 (p. 10). The fact that passivization is needed in (107) in order to permit an interpretation that is excluded by the one-nominal rule suggests that ONO indeed holds strongly throughout the corpus.  It would be worthwhile to recheck the problematic  interpretation in (lO4ii), though.  Intransitive in second conjunct  The final set of sentences to be considered has an intransitive predicate in the second conjunct. These sentences are again prefaced by the context in (84), repeated for convenience here in (108).  The easiest sentences to consider are those in (109-110), which have a single overt  nominal in one conjunct or the other. As expected, their possible interpretations are strongly determined by the one-nominal interpretation law: cmán’-twal’-wit. Wa7 wi7 m.ita7 s-John. (108) Pz-án-twal’ wi s-Bill meet-tr-recip p1 nom-Bill conj nom-John prog emph enemy-recip-3p1.su 1 and John met each other. They’re enemies.’ ‘Bill nilh s-qwatsáts John kw-s (109) qvlqvl-ts-mIn’-as nom-leave conj John bad-mouth-appl-3erg det-nom /(ii) *he. left’ 1 a. ‘he swore at John, and then (i) Bi11 *‘John /he} left’ 1 swore at him , and then {Bill 1 b. nilh John kw-s (110) qvlqvl-ts-mfn’-as bad-mouth-appl-3erg det-nom John conj left’ 1 swore at John, and then a. ‘he *‘John then left’ swore and him, at b.  Bill kw-s det-nom Bill (AA, GN, RW 1589)  s-qwatsáts nom-leave (AA, GN, LT, RW 1590)  The (b) glosses in each of (109-110) are excluded because they violate the rule of one-nominal interpretation, which requires ‘John’ to be the direct object, as it is in the (a) readings. The second conjunct of (109) must have ‘Bill’ as subject of the intransitive predicate, since it is an 53  overt NP. When the overt NP is absent from the second conjunct, though, as it is in (110), coreference for the null subject pronominal is apparently determined by the parallelism constraint, under the assumption that S and A are in the same structural position.  It is  important to note, though, that the judgments in (1 lOa) are those of GN and RW only; parallelism is not as strong a tendency here for AA, who accepts both interpretations in (1 lOa). The sentence in (111) has both arguments of the transitive conjunct realized by overt NPs. Both VOS (a) and VSO (b) interpretations are accepted by GN and RW, while AA— who is more strict in requiring VOS order—regards interpretation (b) as ungrammatical: (111) qvlqvl-ts-mfn’ s-John bad-mouth-appl-3erg nom-John 1 swore at John, and then a. ‘Bill b. ‘John swore at Bill , and then 1  s-Bill nilh nom-Bill conj 9 left’ he, 1 left’ heji*  s-qwatsáts nom-leave (AA, GN, RW 1592)  All consultants accept both coreference possibilities indicated in (a), in violation of the parallelism constraint (if A and S indeed occupy parallel positions in syntactic structure), although the first of these readings, with the null pronominal S coreferent with the overt transitive subject ‘Bill’, was spontaneously offered by RW as a translation for (111). Moreover, during the most recent elicitation, the parallel interpretation was in fact preferred by both GN and RW to the non-parallel one.  On these grounds, then, the former  interpretation can perhaps be taken as preferred to the reading in which the null pronominal S is coreferent with the 0 of the first conjunct. Consider finally a sentence in which there are no overt NPs in either conjunct: (112) qvlqvl-ts-mmn’-as bad-mouth-appl-3erg 1 swore at him, and ‘he ‘he swore at him, and  nilh s-qwatsáts nom-leave conj then he j left’ 1 then hei left’  (AA, GN, RW 1591)  All consultants have said at various times that all readings are possible in a structure like (112), violating parallelism.  Most recently, GN and RW even reject (112) as ‘incomplete’  (not merely ambiguous) and ungrammatical; the perceived incompleteness of (112) is difficult  54  to explain, given a discourse context like (108) having enough overt NPs that could plausibly be coreferent with the null pronominals in (112). The conclusion to be drawn concerning coreference between conjuncts having an intransitive predicate is that the one-nominal interpretation law holds strongly, as it does across transitive conjuncts, but that parallelism is at best a tendency.  3.4,  Summary  Coreference across conjuncts is neither ergative nor accusative in St’át’imcets.  It is most  strongly constrained by the one-nominal interpretation law (94), and the data that are presented in §3.3. 1.2-3.3.2 in support of this point further establish ONO as a salient property of the grammar of St’át’imcets. Parallelism (87) imposes a secondary, but otherwise fairly strong, constraint on coreference across transitive conjuncts. In sentences in which one conjunct is intransitive, though, PC does not hold strongly (if at all), and the grammaticality of the relevant sentences is more difficult to establish firmly and consistently.  This conclusion is  consistent with that of Matthewson (1993a), who found that intransitive subjects patterned with transitive subjects only across discourse, not intrasententially. The data in this chapter involve only intrasentential coreference, and so the failure of parallelism to hold for intransitive subjects is not unexpected. An important question is therefore how to capture the asymmetric behaviour of transitive and intransitive conjuncts with regard to coreference possibilities. Just as in chapter 2, the facts presented here again present a problem for a theory that treats A and S together as members of the category ‘subject’, e.g., Anderson (1976) and Keenan and Comrie (1977: 80 ff.).  However, while chapter 2 showed that S and 0 pattern alike with  respect to relativization (in opposition to A), this chapter establishes that each of {A,S,0} shows distinct behaviour in coreference across conjuncts, and so conjunction cannot be said to have either an ergative or an accusative pivot. It should be stressed that in having a parallelism constraint that applies only to transitive conjuncts, St’át’imcets is not unusual.  In examining a similar range of facts in  English, Oehrle (1981) comes to nearly identical conclusions.  55  Oehrle’s propositional  congruence (his ‘rule R’), though, requires additionally that the predicates in each clause be 29 identical and that the rule apply with a specific intonation.  Oehrle notes that having an  intransitive predicate in one conjunct makes the reference of its subject difficult to establish in isolation (because the different predicate disrupts the congruence across conjuncts), although its reference is easy to establish unambiguously if there is an appropriate discourse context. That these generalizations appear to hold across languages as seemingly disparate as English  and St’át’imcets, in rather complex constructions requiring subtle intuitions about their ambiguity and grammaticality, strongly suggests that the parallelism constraint—whatever its ultimate explanation—is a plausible candidate for Universal Grammar.  29  The intonational properties of sentences in this chapter have not been studied, and so it is possible that some of the seemingly problematic data are due to misunderstood prosodic properties of the utterances.  56  Chapter 4 Structural Asymmetries  4.1.  Introduction  This chapter shows that the various asymmetric behaviours of subjects in St’át’imcets, as described in chapters 2 and 3, may be given a structural explanation within the broader Principles and Parameters framework of Chomsky (1981; 1982; 1986; 1992; etc.). Section 4.2 addresses the one-nominal interpretation law, §4.3 examines the appearance of parallelism on pronominal coreference across conjuncts, and §4.4 derives the restriction that ergatives, unlike absolutives, cannot be directly relativized. Unifying all of these accounts is the idea— independently motivated and proposed to account for facts in other languages—that NPs that are topics are structurally higher than NPs that are focused. 30 The terms ‘topic’ and ‘focus’ have been used with different meanings in the literature, a full discussion of which is beyond the scope of this thesis. Consult the essays in Li (1976) and the more recent works cited throughout this chapter for some background. To give rigid definitions of topic and focus here would not serve well the purpose of this chapter, which crucially assumes only that, however topic and focus are defined, the NP representing the former is higher in constituent structure than the NP representing the latter.  Nevertheless,  before turning to the particular analyses, it will be worthwhile to have some general definitions. Reinhart (1981: 57-58) identifies two majors approaches to topichood. One approach defines the topic as the noun phrase whose referent a particular sentence is about, and the other defines the topic as the old information in the discourse.  Reinhart argues for the former  approach (‘pragmatic aboutnes&), and suggests the following test for topichood (Erteschik-Shir 1993: 1; 22, §4): 30  The idea to apply this structural asymmetry to an analysis of St’át’imcets is due generally, and in each specific case examined below, to Henry Davis and Hamida Demirdache. (The particular structures given below were suggested by Henry Davis.) This chapter could not have been written without their help. 57  (113) Topic test Speaker A: Speaker B:  Tell me about X .  .  .  X.  .  (X  .  =  topic)  In speaker B’s answer to A’s question, X is the topic, since B is telling A about X. The topic is therefore the referent or subject that is presupposed in the discourse. The unmarked topic of the discourse is usually the grammatical subject of a sentence, as has been noted by (among others) Erteschik-Shir (1993: 27, §5; 45, §6.3.1), Kinkade (1989: 1), Matthewson (1993a: 4-5), and Reinhart (1981: 62).  This canonical mapping of  subject to topic can of course be altered by special morphology, phonology, syntax, or by whatever else a language might use to alter the topic-focus relation, and examples from St’át’imcets will be seen below in (150, 181) (pp. 81, 100). Erteschik-Shir (1993: 45, §6.3.1) refers to this mapping as the topic constraint, and it is formalized for the purpose of this thesis 31 as follows: (114) Topic Constraint *  1 [SUBJECT TOP  [vp.  .  .  The topic constraint does not exclude overt (S-structure) topicalization (e.g., 1 Fred I like t,) , nor left-dislocation (e.g., Fred , I like him); rather, the topic constraint merely maps syntactic 1 positions to discourse functions at a level of derivation beyond S-structure, here assumed to be logical form (LF).  Because this chapter will adopt a version of the VP-internal subject  hypothesis, it is necessary to revise the topic constraint as in (115), in order that the subject of a clause may be base-generated in Spec/VP and still be mapped to topic: (115) Topic Constraint (revised) *  1 [SUBJECT [v’. TOP  .  .  NP . 1 ..]]  31  The topic constraint captures a valid cross-linguistic generalization, though it has apparently never been derived. Its explanation may be related to the fact that subjects are always generated higher than objects, though the question is beyond the scope of this thesis. 58  The revised topic constraint (115), but not (114), permits well-formed structures such as (162, 168) below (jp. 89, 92), in which a trace in VP-internal subject position is coindexed with the topic. The topic constraint forbids the object from being the topic, but as a negative constraint it does not specifically compel the subject to be the topic. Consider, however, that the topic is generally regarded as being in an A’-position, hence operator-like; see, for example, the The ban on vacuous quantification (Chomsky  sentences in (140, 146) below (pp. 74, 78).  1982: 11-13) would exclude a sentence like the following, since the topicalized NP does not bind anything in the sentence: (116) *Fred, I like her The ban on vacuous quantification conspires with the topic constraint (115) to ensure that the subject and topic are coindexed. Specifically, if the topic and object are coindexed, the topic constraint (115) is violated. Although the topic constraint will be satisfied if the topic does not bind any NP in the sentence, such quantification would be vacuous. Therefore, a well-formed structure is derived only when the topic and subject are coindexed. The mapping of subject to topic that emerges from the interaction of the topic constraint (115) and the ban on vacuous quantification can be satisfied either by LF movement of the subject to topic position, or by coindexation of a discourse topic with a null pronominal subject.  Both options will be  exemplified in this chapter. The next question to be addressed concerns the syntactic position of the topic, some recent proposals for which are illustrated below: (117)  a. Internal Topic  b. External Topic  c. Topic Phrase  CP  E  Top-P  1 Topic  IP ti...  1 Topic  CP  .  59  .  .  )... 1 (pro  Top’  e Top  CP/IP  Aissen (1992: 47) suggests that Mayan languages have two topic positions, one for an internal topic, which is in Spec/CP (1 17a), and another for an external topic, which is prefixed to an entire clause CP under a node labeled E (11Th). The internal topic position is a landing site for movement of a topic from within the sentence, while the external topic position contains a base-generated topic that may be coindexed with a pronominal within the sentence. The node E in (1 17b) is an abbreviation for Expression, which Aissen (1992: 47) adopts from Banfield (1973: 14 ff.) and Emonds (1985: 316 ff.). The final structure (c) represents the topic phrase structure of Chomsky (1977: 91) in X-bar theoretic terms; it shares with the external topic (b) the property of being the highest element in any sentence. Top is the head of the maximal projection Top-P, and the NP that is the topic—because it is a maximal projection—appears in specifier position. The topic NP may be overt, as in the case of S-structure topicalization, or it may be non-overt, as it is in (c). In this diagram, a null topic e occupies Spec/Top-P, and in such a case its reference would be determined by discourse. 32 It is not clear which structure in (117) is appropriate for St’ at’ imcets. Gardiner (1993: 125-138, §3.1) motivates an external topic position for Secwepemctsmn on the basis of word order, the position of clitics, and the possibility of doubling an argument with a deictic. However, because the data to be examined below require both sentential topics (which are mapped at LF from overt NPs) and discourse topics (which are coreferential with NPs outside of the sentence), neither the internal nor the external topic position in (1 17a, b) seems appropriate. Instead, the topic phrase structure of (1 17c) is adopted here because it is broad enough to subsume both sentential topics and discourse topics.  Other structures for  representing topics could be accommodated by the analysis here, though the only crucial (and uncontroversial) requirement of the representation of topic is that it be structurally higher than 33 the non-topic/focus of a sentence—which is examined next. 32  Another structure that is not considered here, suggested by Authier (1992) and Watanabe (1992), is that the topic appears in specifier position of a recursive CP. The possibility of there being multiple topics in a single discourse is excluded from this analysis. Although there might be more than one topic in the discourse, the notion of topic that is considered here is that of topic of a sentence, i.e., the single NP which the sentence is about or is predicated of (in the sense of Erteschik-Shir 1993 and Reinhart 1981). 60  The focus of a sentence introduces a new referent or proposition into the discourse (Calabrese 1990: 12), and so a common test for a focused constituent is to determine whether it may answer a wh-question.  In the following examples, the capitalized constituent is the  focused constituent corresponding to the wh-word: (118) a. What did John do? John ATE THE APPLE. b. What did John eat? John ate THE APPLE. c. Who ate the apple? JOHN ate the apple.  (Erteschik-Shir 1993: 23, §4)  The focus is often identified phonologically as the stressed element, but a syntactic correlation that has been observed in intonationally unmarked sentences is that a focused NP typically appears within the verb phrase, usually as the object (Calabrese 1990: 4; Diesing 1992: 49-53; Matthewson 1993a: 4). Heim (1988) claims that the logical representation of a sentence with quantified NPs has a tripartite structure that includes the quantifier, restrictive clause, and nuclear scope. The quantifier has scope over variables in the restrictive clause, while unbound variables in the nuclear scope get bound by ‘existential closure’, which is the presence of a covert existential quantifier.  In theories that attempt to derive such logical representations from syntactic  structures, the division between the restrictive clause and the nuclear scope is drawn near the VP (Erteschik-Shir 1993; Diesing 1992: 9-10, 49-53; Partee to appear: 5). Further, as argued by Erteschik-Shir (1993) and Partee (to appear), and as suggested by Diesing (1992), the nuclear scope contains the focus. The topic and non-topic/focus constituents therefore appear in the following structural relation, (1 19a) representing Erteschik-Shir and Partee’ s proposal, and (1 19b) representing Diesing’s: (119) a. b.  [cp topic [vp focus]] [cp topic [vp non-topic]]  The lower node that contains the focus is identified here as VP for the sake of concreteness, following the aforenamed authors, although the split between the restrictive clause and the  61  nuclear scope could be made at some other node without consequences for the present analysis. Aissen (1992: 47) and Uechi (1994), for example, put the domain of focus as high as IP. Common to all approaches, however, is that the topic is higher than the focus or non-topic, and this is similarly the only crucial relation that is needed here.  4.2.  The one-nominal interpretation law as focus  This section gives a structural interpretation of the one-nominal interpretation law (Gerdts 1988: 59), cited in (23) in chapter 1 and (94) in chapter 3, and repeated below: (120) One-Nominal Interpretation Law (ONO) In the absence of marking for other persons, a single 3rd person nominal is interpreted as the absolutive. When the predicate is intransitive, a single 3rd person nominal must be interpreted as absolutive, since the predicate takes a single argument. What (120) leaves unexplained is why the same fact obtains when the predicate is transitive, since the nominal might just as easily be interpreted as the subject.  In the following sentence, for example—repeated from (20) of  chapter 1 (p. 10)—the single overt nominal is uniquely interpreted as the object. (121) áts’x-en-as ta sqaycw-a man-det det see-tr-3erg ‘he saw the man’  (AA, BF, GN, LT, RW 29)  (121) may not be interpreted as ‘the man saw him’.  This restriction may be structurally  encoded given the reasonable assumption that NPs will not appear overtly in a sentence if their referents are presupposed in the discourse. As discussed in §1.3.1, St’át’imcets is a radical head-marking language (Davis 1993b: §2), which entails both subject and object pro-drop (these arguments being marked by pronominal affixes on the predicate that selects them). Because of this rich subject and object agreement, overt NPs are optional. example is repeated from (3) of chapter 1  (p.  5):  62  The following  (122) tsün-tsi-lhkan tell-2sg. obj-lsg. su ‘I told you’  (AA, LT 2296; van Eijk 1985: 174)  (122) is felicitous because the referents specified by the affixes (first- and second-person) are always clear from the discourse, in which there is a speaker (first-person) and an addressee (second-person). However, overt arguments are optional even when they are third-person, as the following example shows (repeated from (12) of chapter 1 (p. 8)): (123) tsün-ø-as tell-3sg. abs-3sg. erg ‘she told him’  (LT 2425)  In sentences like these that do not have overt lexical arguments, the arguments of the predicate are generally assumed to be represented by null pronominal constituents (the empty category  pro).  Each pro is in an argument position, and is licensed by a pronominal affix on the  34 predicate. Note, however, that although sentences like (123) are grammatical, they are dispreferred discourse-initially, since the third-person referents are not identified, and because  pro cannot be used deictically. The inability of pro to refer independently has been taken to be a characteristic of anaphors, as the contrast below suggests: (124) a.  b.  [Somebody in the audience gets up and leaves] Speaker: He is weird Where is the bastard going? Where is John going? I guess his patience ran out [Same situation] Speaker: *J like himself (vs. I like him)  (Thráinsson 1991: 61-62)  Baker (1991) proposes this representation for pronominal arguments in Mohawk. See Jelinek (1984) for a different approach to the structure of radical head-marking languages, in which the pronominal affixes themselves are the arguments of the predicate. -  63  c.  [Same situation] Speaker: *Hielf is weird  The italicized NPs refer to the person who is leaving. Crucially, none of the NPs in (124a) is syntactically bound, since the antecedent comes not from the sentence or even from the discourse, but merely from the speech situation. The same situation in (b,  C),  however, shows  that an anaphor (himself) may not be used deictically—in contrast to the non-anaphoric elements in (a). The pro of St’át’imcets therefore behaves as an anaphor in this respect, since it likewise may not appear unbound in a sentence, and minimally requires a discourse antecedent. Moreover, because pro is a bound variable anaphor, it may not serve as a topic, though pro itself may be bound by a topic. This property is not particular to St’át’imcets; the pro of Romance languages, for example, is subject to the same restriction. Consider the following contrast from Italian: arrossito è be. 3p blush.prtc  (125) a.  vista, ha 1 1’ Quando Carlo Carlo her have. 3p see.prtc when ‘when Carlo saw her, he blushed’  pro 1 pro  b.  vista, ha 1 1’ Quando pro pro her have.3p see.prtc when 1 saw her, Carlo 1 blushed’ ‘when he  arrossito Carlo è 1 Carlo be.3p blush.prtc  (126) a.  b.  *Quando 1’ Carlo, vista ha her have. 3p see.prtc Carlo when 1 blushed’ ‘when Carlo saw her, he *Quando pro vista, ha 1 1’ pro her have. 3p see.prtc when 1 saw her, Carlo blushed’ ‘when he  proj pro  arrossito è be. .3p blush.prtc  1 arrossito Carlo è be. 3p blush.prtc Carlo (Calabrese 1990: 10)  The sentences in (125) contain a preverbal subject, Carlo, in either the adverbial clause (a) or the main clause (b); in each case, the pro subject that is in the other clause may take its reference from the preverbal subject Carlo.  The sentences in (126) correspond to those in  (125), their only difference being that Carlo appears in post-verbal position—which Calabrese (1990) shows is a VP-internal focus position. Coreference between pro and the focused, post  64  verbal subject Carlo is not possible, and because these sentences do not contain a topic, the pro subject lacks an antecedent. A St’át’imcets sentence like (123), then—which has no overt nominals—may be freely used non-initially, since two participants may be tracked throughout a discourse by mapping GFs to discourse roles according to one of the schemata in (119) and by using other methods 35 Once a discourse topic has been established, it need to indicate the switching of these roles. not be expressed subsequently by an overt NP. Bearing this in mind as we return to a sentence like (121)—which has a single overt NP—it is clear that the overt NP ta sqáycwa ‘the man’ would not appear in this sentence if it were the discourse topic, since the topic—the entity or person about which the discourse is concerned, the presupposed information—would normally be referenced solely by a third-person pronominal affix on the predicate. Because this NP is a new referent—and because the topic of a sentence represents a presupposed referent, as the topic test in (113) shows—the most natural interpretation of this sentence is therefore the one in which ta sqáycwa ‘the man’ is the non-topic or focus. Given the mapping in (119), then, it must be inside the VP as the object: (127) [cp [vp áts’x-en-as see-tr-3erg  ta sqáycw-a]] det man-det  Under this approach, the one-nominal interpretation law as stated in (120) is not a rule of Salish grammar, but simply a description of facts that are derived from (i) an independently needed mapping that gives the result of (119), and (ii) the principles that license null NPs as 36 The interaction of these general sub-theories is what compels the single NP in arguments. (121) to be the object rather than the subject.  35  For examples of discourse tracking in Salish and other languages, see Kinkade (1989; 1990). 36 For some proposals concerning (ii), see Jelinek (1984) and Baker (1991). Davis (1993b) shows that neither Jelinek’s nor Baker’s approaches to radical head-marking languages makes entirely correct predictions for St’át’imcets, which exhibits properties of both radical head-marking languages like Mohawk and lexical argument languages like English.  65  4.3.  Parallelism  4.3.1. Transitives Having derived the effects of the one-nominal interpretation law, most of the effects of parallelism may now also be derived without further elaboration. As discussed in chapter 3, coreference across transitive conjuncts is strongly constrained not only by ONO, but by the parallelism constraint on discourse functions (87), repeated below: (128) Parallelism Constraint on Discourse Functions (PC) For two items to corefer, they must both fulfill the same discourse function (either topic of the discourse or non-topic).  In addition, there is a preference for both  coreferential elements to fulfill the topic of the discourse function. The effect of parallelism is illustrated below in (130), which has the context (129).  These  examples are repeated from (84-85) of chapter 3 (p. 42). (129) Pz-án-twal’ wi s-Bill müta7 s-John. Wa7 wi7 cmn’-twa1’-wit. meet-tr-recip p1 nom-Bill conj nom-John prog emph enemy-recip-3p1.su 1 and John met each other. They’re enemies.’ ‘Bill (130) áts ‘x-en-as  nilh  s-qvlqvl-ts-mIn ‘-as see-tr-3erg conj nom-bad-mouth-appl-3sg. conj 1 swore at him/*he swore at him}’ a. ‘he saw him, and then {he 1 swore at him}’ b. ‘he saw him, and then {hej swore at him/*he (AA, GN, RW 1595)  Recall from the previous section that because null pronominals are bound variables (which do not have independent reference), they are disfavored discourse-initially; rather, they must be bound by referring expressions at the relevant level of representation (presumably LF). Neither conjunct in (130) has an overt nominal, and so the topic must come from the discourse (129). Regardless of whether Bill or John is chosen as the topic, it is coindexed with the null pronominal subject in each conjunct, in accord with the parallelism constraint.  The null  pronominal that is assumed to be in subject position is not itself a topic, but instead takes its  66  reference from the discourse topic because of the unmarked correlation noted above between subjects and topics. Before examining the structure of this sentence, it is worth considering whether the effects of parallelism as stated in (128) can be derived from the topic constraint (115). Because pro needs to be bound, and because a topic must bind some element in order not to violate the ban on vacuous quantification, it follows that the pro subject in both conjuncts of a compound sentence like (130) will be bound by the same topic in Spec/Top-P (which is the highest phrase in any sentence), in accord with the topic constraint. Parallelism is therefore not the result of a principle like (128); instead, the effect of parallelism is derived by the interaction of the topic constraint with the requirement that pro be bound, and the fact that the 37 universe of discourse contains only two persons, Bill and John.  Consider how the topic  position and the coindexation between subject and topic account for the parallelism exemplified in (130). This sentence has the structure in (131)—as suggested by Henry Davis (p.c.)—with each conjunct assumed to be in an asymmetrical relation with respect to each other. Subjects are shown in their S-structure position (Spec/IP) for the sake of clarity, though §4.4 will demonstrate that transitive subjects are base-generated in Spec/VP and raised at S-structure in order to get case. Irrelevant details of structure are ignored here and throughout the chapter.  It would be interesting to collect data having three discourse participants—especially double object constructions—in order to determine whether they provide evidence for a parallelism constraint that is independent of the topic constraint. 67  Top-P  (131)  VP  A  \ / áts’x-en-as Proj see-tr-3erg  NP  X  1 pro  nilh  VP  NP prof  s-qvlqvl-ts-mIn’ -as proj nom-bad-mouth-appl-3sg. conj The conjunction nilh is labeled simply as a head X, since its exact category is neither easy to determine nor crucial to this account; perhaps it is a one-place predicate or a complementizer. Recall also from  § 1.3.1 that categorial distinctions in Salish languages are a subject of debate  (Kinkade 1983; van Eijk and Hess 1986).  The second conjunct is the complement of the  . 1 conjunction nil/i, and the XP within which it is contained is adjoined to 1P There are two important reasons for adopting this structure.  First, such an analysis  accommodates coordination structures into the more general X’-schema of phrase structure, which is strictly binary branching.  Second, the conjunction nil/i is clearly subordinating,  rather than coordinating, and this structure accounts for the subordinate inflection (specifically, the nominalizer s-) that occurs in the second conjunct of coordinate structures headed by nilh; recall the discussion of clause inflection in chapter 1 (p. 12) leading up to the chart in (27). In ) is the one that is 1 a structure like (131), the conjunct to receive non-subordinate inflection (1P not a complement of the conjunction nil/i, while the non-initial conjunct that is selected by nilh ) receives subordinate inflection. 2 (1P Returning to the facts of (130) that the structure in (131) is intended to explain, there is a null topic in this structure, represented by e, which is in Spec/Top-P. The referential index of the null topic is chosen from one of the topics in the discourse (129).  68  This coindexation  occurs at LF (Huang 1984: 550), and it resembles covert left-dislocation, corresponding to English S-structure left-dislocation in a sentence like Sam , he 1 1 swore at Fred. Regardless of whether the null topic in (131) is coindexed with Bill or John from the discourse (129), note that the subject in Spec/IP of each conjunct is coindexed with e in accord with the topic constraint (115).  Huang (1982: 359-360, §5.4.1, 444 n. 14; 1984: 542-543) identifies the  same topic-bound empty category in Chinese as a variable—following Chomsky’s (1981: 185, 330) functional determination of empty categories—since it is in an A-position and is locally A’-bound by the null topic. The same analysis is adopted here. Huang assumes moreover that the variable is the trace of the moved topic.  However, the null topic is represented here  simply as a base-generated empty category that must be coindexed with a pro variable in order not to violate the ban on vacuous quantification (Chomsky 1982: 11-13). This coindexation is compelled because pronominals are not deictic, and hence must be topic-bound. Even if the topic were a null operator that had been moved from an A-position, it would still need to find an antecedent from the discourse. 38 In a sentence in which there is an overt subject and object, the topic is not taken from the discourse, since the sentence already has a subject that may move to the topic position at LF. Consider the following sentence, which is repeated from (88) of chapter 3 (p. 45). (132) áts’x-en-as nilh see-tr-3erg conj 1 saw him, a. ‘he *‘h b. saw him,  s-qvlqvl-ts-mfn’-as kw-s John kw-s Bill nom-bad-mouth-appl-3sg. conj det-nom John det-nom Bill and then {Bill swore at John/*John swore at Bill )’ 1 and then {Bill swore at John/John swore at Bill }’ 1 (AA, GN, RW 1597)  Interpretation (a) is preferred because it respects the VOS order of the second conjunct, whereas the ungrammatical interpretation in (b) is excluded because it would require VSO order.  The subject Bill in the second conjunct will therefore control the reference of the  subject of the first conjunct.  38  See Cinque (1990: §3) for motivation for distinguishing pronominal variables from pure variables (i.e., traces of moved operators).  69  Before examining the structure of this sentence, it should be noted that the site of overt nominals in radical head-marking languages is a subject of debate: it has been claimed that overt nominals appear in adjoined positions and are either coindexed with pronominal affixes (Jelinek 1984) or with empty categories in argument positions (Baker 1991). Neither approach is correct for St’át’imcets, since Davis (1993b), Matthewson (1993a), and Matthewson et al. (1993) have shown that Condition C effects obtain with overt possessor NPs in St’át’imcets— unlike in Mohawk (Baker 1991), where Condition C does not appear to apply to overt possessor NPs.  This fact can only be explained if overt nominals occupy argument  39 Davis’s and Matthewson’s structure is assumed here, and so the overt nominals of positions. (132) appear in argument positions in the following structure: (133) xP  Ipi VP  2 1P  NP  X  pro  nilh  v  NP  áts’x-en-as pro  see-tr-3erg  Bill s-qvlqvl-ts-mmn’-as kw-s John  nom-bad-mouth-appl-3sg. conj det-nom John  2 moves to Spec/Top-P, and it is then coindexed with the pro At LF, the subject Bill from 1P 1 in accord with the topic constraint (115). subject in 1P  This coindexation is essentially LF  , Sally 1 topicalization, resembling English S-structure topicalization in a sentence like Fred  believes  t  to be a fool. The LF representation is as follows:  Gardiner (1993: Secwepemctsfn.  §5) reaches the same conclusion for overt nominals in  70  Top-P  (134) Bi11  IP  Ipi VP  xP NP  X  prof  nilh  2 1P  áts’x-en-aspro  VP  NP  see-tr-3erg  s-qvlqvl-ts-mfn’-as kw-s Johns nom-bad-mouth-appl-3sg. conj det-nom John The objects are free to corefer with each other, as indicated. (89) of chapter 3 (p. 45), which differs from (132)  (= 88)  The same analysis extends to  only in having the overt subject and  object in the first conjunct rather than in the second. A question concerning (134)—given that XP is in an adjoined position—is whether LF extraction from XP violates the Condition on Extraction Domains (Huang 1982: §6.4), which forbids extraction from a domain that is not properly governed. The resolution of this question is only a technical matter, since it will be seen below in (135) and (142) (pp. 72, 76) that extraction is indeed permissible only from the non-initial conjunct.  See also the discussion  pertaining to (146) below (p. 78) for one manner of addressing this issue. Note moreover that extraction of Bill in (134) cannot be regarded as an across-the-board violation, since at LF, the topic Bill obligatorily binds an empty NP in each conjunct: pro in 1P , and trace in 1P 1 . The 2 binding relation in each case is an A’ relation. Parallelism obtains in sentences that have only null NPs, and in sentences in which there are two overt NPs in a single conjunct. In all of the cases examined above, the subject is mapped to topic, and because there is only one binder (the topic in Spec/Top-P), it must bind two elements, in accordance with the topic constraint.  The sentences that remain to be  explained are those that do not exhibit parallelism; they differ from the former sentences in that one of the conjuncts contains a single overt NP. Recall from chapter 3 (pp. 48 ff.) that  71  sentences having a single overt nominal in each conjunct provided evidence for constraint ranking, since the one-nominal interpretation law was respected at the expense of violating the lower-ranked constraint on parallelism.  The relevant sentence was (97), repeated here as  (135):  (135) áts’x-en-as kw-s John nilh s-qvlqvl-ts-mIn’-as kw-s Bill see-tr-3erg det-nom John conj nom-bad-mouth-appl-3sg. conj det-nom Bill 1 saw John, and then (i) hej swore at Bi11 a. ‘he /(ii) *Bill swore at him’ 1 *‘John b. saw him, and then (i) he swore at Bill/(ii) Bill swore at him’ (AA, GN, RW 1593) Interpretations (a.ii) and (b) are excluded because they would require the single NP in one of the conjuncts to be outside of VP, an interpretation that was shown in §4.2 to be disallowed. The S-structure of (135) is therefore as follows, with each overt NP within the VP of its own conjunct: (136) Ipi VP  xP  NP  X  pro  nilh  2 1P  áts’x-en-as kw-s John see-tr-3erg det-nom John  NP pro  s-qvlqvl-ts-mmn’-as kw-s Bill nom-bad-mouth-appl-3sg. conj det-nom Bill Because there are only two persons in the universe of discourse, it is not possible to leave the overt NPs in situ at LF by coindexing the topic with some preceding NP that might have had a distinct reference from John and Bill. movement to Spec/Top-P.  Nor is there an overt subject that is available for  Null pronominals are non-referring, as discussed above, and so  neither of the pro subjects may raise to topic. Moreover, coindexation of a null topic with one of the overt NPs, as below, is also not possible:  72  *  (137)  Top-P IP xP  Ipi VP  2 1P  NP  X  pro  nilh  NP  áts’x-en-as kw-s John see-tr-3erg det-nom John  pro 1 s-qvlqvl-ts-mfn’-as kw-s Bill nom-bad-mouth-appl-3sg. conj det-nom Bill  •  Whatever index is borne by the null topic in (137), a Condition C violation will result in one of the conjuncts, since either John or Bill will not be A-free after the topic constraint (115) compels coindexation with the pro subjects. The only way to rescue this structure is to move one of the overt object NPs to topic position, violating the topic constraint. In the following 2 moves: structure, the object Bill from 1P *  (138)  Top-P  1 Bill  IP xP  1 IP VP  2 1P  NP  X  1 pro  nilh  áts’x-en-as kw-s John see-tr-3erg det-nvm John  VP  NP 1 pro  s-qvlqvl-ts-mfn’ -as t 1 nom-bad-mouth-appl-3sg. conj 1 is coindexed with the topic, Bill. Under the topic constraint (115), the The subject pro of 1P subject of the other conjunct is similarly indexed with the topic, exactly as in (131, 134)  73  , since the 2 above. However, this coindexation results in a strong crossover configuration in 1P trace of Bill is a variable, and the closest binder is not its antecedent Bill in topic position, but . Condition C therefore excludes (138), since the trace of 2 rather the A-binder pro in SpecIIP Bill is not A-free.  2 (135) must instead have the following structure, with pro in Spec/1P  disjoint in reference from the trace that it c-commands: Top-P  (139) Bill  IP ------  xP  Ipi VP  2 1P  NP  X  1 pro  nilh  VP  áts’x-en-as kw-s John see-tr-3erg det-nom John  NP prod  1 s-qvlqvl-ts-mf& -as t nom-bad-mouth-appl-3sg. conj The LF representation in (139) corresponds to the only grammatical reading for this sentence,  (135a.i).  Note that the reference of the subjects is crucially not parallel—a fact that was  described in chapter 3 by ranking the parallelism constraint below the one-nominal interpretation law; recall tableau (99) on page 49.  In the analysis here, though, the non  parallel reading is derived by a conspiracy between the one-nominal interpretation law, which forces the single overt NP in each conjunct to remain within VP, and Condition C, which requires that the trace of the object that is moved to topic position be disjoint from the subject of its clause. A strong crossover configuration is similarly reproducible in English by topicalization (Lasnik and Uriagereka 1988: 154): (140) a. b.  who does heji* like t John, heji* likes t  74  Both the wh-trace in (a) and the trace of the topicalized NP in (b) are variables, hence must be A-free. English S-structure topicalization therefore mirrors St’át’imcets LF topicalization—as was already noted with respect to (134), which had an overt subject and object in the second conjunct. Apparently, the single overt NP John in the first conjunct of (135) may instead raise to topic position at LF, with the single overt NP Bill in the second conjunct remaining in situ. The structure of the sentence in this case is as follows: Top-P  (141)  IP  John.  xP  Ipi  VP  A  /  NP  X  1 pro  nilh  2 1P VP  NP  áts’x-en-as t  see-tr-3erg  proj s-qvlqvl-ts-mIn’-as kw-s Bill  nom-bad-mouth-appl-3sg. conj det-nom Bill  It thus appears that there are two well-formed LF derivations for (135): one with Bill as topic  (= 139),  and another with John as topic  (= 141).  Regardless of which object in (135) is raised  to Spec/Top-P, the single grammatical interpretation in (135a.i) is obtained. As there is a single interpretation for (135), it would be preferable if there were a ° 4 single corresponding LF representation.  In fact, LF-raising of an object is possible only  , although this can be ascertained only by examining a 2 from the non-initial conjunct 1P sentence having a single overt NP in the first conjunct, but no overt NP in the second conjunct. An example is (142), repeated from (95) of chapter 3 (p. 47):  40 Chomsky (1992) argues that there should be only one ‘converging’ (i.e., well formed) derivation for any given sentence. 75  John nilh s-qvlqvl-ts-mIn’-as (142) áts’x-en-as kw-s see-tr-3erg det-nom John conj nom-bad-mouth-appl-3sg. conj ’ 1 1 swore at him/(ii) *hj swore at him 1 saw Johns, and then (i) he a. ‘he *‘John /he swore at him}’ 1 saw him , and then {hej swore at him 1 b. (AA, GN, RW 1598) In (142), the single overt nominal John in the first conjunct is interpreted as inside the VP, pursuant to the discussion in §4.2, and so interpretation (b)—with John as subject—is immediately excluded. The non-parallel interpretation (142a.ii) is excluded because it would require the second conjunct to violate the topic constraint (115). Since there are no other overt NPs in this sentence that can move to Spec/Top-P (John is excluded, as will be seen presently), the topic is taken from the discourse, and so the empty topic is coreferent with Bill from the context (129). This sole grammatical reading (142a.i) therefore has the following representation: Top-P  (143)  IP  1 e Ipi VP  xP  2 1P  NP  X  1 pro  nilh  áts’x-en-as kw-s John see-tr-3erg det-nom John  VP  NP 1 pro  s-qvlqvl-ts-mfn’ -as pros nom-bad-mouth-appl-3sg. conj Crucially, the NP John in the first conjunct may not map to topic position. Such a structure, , and corresponds to the ungrammatical 1 shown below, violates the topic constraint (115) in 1P reading in (142a.ii):  76  *  (144)  Top-P IP  John.  NP  X  1 pro  nilh  VP  A  /  \  2 1P NP  VP  áts’x-en—as t  see-tr-3erg  prod 1 s-qvlqvl-ts-mmn’ -as pro  nom-bad-mouth-appl-3sg. conj  The generalization illustrated by the contrast between (135) and (142) is that an object NP may  be extracted to topic position—violating the topic constraint (115)—only if it is at the periphery of the sentence.  (Stated differently, the VP-internal overt nominal may escape the nuclear  scope and raise to topic only if it originates in the non-initial conjunct.) The validity of this restriction is further supported by the data below in (148), and it is also motivated by the cross-linguistic observation that topics typically appear at clause-peripheries, rather than in embedded positions (e.g., As for Fred, I like him).  The correlation seems natural, in that  subjects in canonical VOS main clauses in St’át’imcets are likewise the peripheral element. Moreover, both subjects in simple clauses and objects in non-initial conjuncts (as in (149) below) occupy the highest S-structure position—the latter because the second conjunct is . 1 adjoined to 1P  Consider the following sentence, in which parallelism obtains in an adjunct  clause only for the subject:  (145) ts’um’-qs-án’-as lick-nose-tr-3erg  lh  hyp  áts’x-en-as  see-tr-3erg  ta  sqáycw-a  det man-det  1 sees the man’ 1 will kiss him if Mary ‘she a. *‘she 1 ’ 1 will kiss him if the man sees Mary b. c. * ‘he will kiss her if Mary sees the man’ ’ 1 d. *‘he will kiss her if the man sees Mary RW 505; Matthewson (GN,  77  s-Mary  nom-Mary  topic) (topic. (topic. non-topic) topic) (non-topic. non-topic) (non-topic. 1993a: 29; Matthewson et al. 1993: 230; cf. BF, RW 567) .  .  .  .  .  .  .  .  Readings (b, c) are not parallel.  The interesting contrast is between (a) and (d); both are  parallel, but only the former has the peripheral NP Mary acting as both the subject and the topic, while the latter has Mary acting—ungrammatically—as the non-topic. (Or, equivalently, the non-peripheral NP ta sqáycwa ‘the man’ cannot be the topic.) Because these speakers are not especially strict in requiring VOS order, the contrast between (a) and (d) provides further evidence that the topic is always the peripheral element in the structure. With this description in hand, it remains to be explained how an appropriate LF representation is derived for such sentences. Reconsider (135), which had a single overt NP in each conjunct, but apparently two well-formed LF representations (139, 141). It is now clear from the discussion of (142) that (141)—with the object raising from the initial conjunct—is not a legitimate structure for (135).  The only well-formed representation underlying (135)  must be the one in which the object from the non-initial conjunct is raised to topic position (139). One way of deriving this representation, suggested by Henry Davis (p.c.), is to regard the subject Bill as having scrambled at S-structure to an A’-position (perhaps adjoining to XP), from which it then moves at LF to Spec/Top-P.  There are two reasons for regarding this  scrambling as A’-movement. First, reconstruction is always from an A’-position, and there is evidence for such reconstruction of scrambled NPs in St’át’imcets (Henry Davis, p.c.). Second, if the Coordinate Structure Constraint holds at LF (Davis et al. 1993: 88 show that it clearly holds at S-structure), it would be puzzling why LF movement of an object NP to topic position (from a single conjunct) is grammatical; if the object has first scrambled to an A’position, however, the well-formeciness of its further extraction to topic position can be explained by the generalization that a phrase that has been moved to an A’ -position does not constitute as strong a barrier for extraction as does a phrase in an A-position. As evidence for this claim, consider the following contrast (from Fiengo et al. 1988: 89, §3): (146) a. b.  *[vowel harmony] , I think that articles about tj have been published 1 ?[vowel harmony], I think that [articles about tj], you should read tj carefully  The contrast between (a) and (b) shows that the effects of the Condition on Extraction Domains (CED) are weakened if extraction occurs from a topicalized constituent (which is in 78  an A’-position).  In (a), extraction of vowel harmony directly from the embedded subject  articles about vowel harmony violates the CED, since subject position is not properly governed. In (b), though, articles about vowel harmony is first topicalized (moved to an A’position from the position of tj), and from this position vowel harmony may be further topicalized. The generalization is stated by Fiengo et al. (1988) as in (147): (147) x is a barrier only if it is not an A’-binder Because articles about vowel harmony is an A’-binder in (146b) (it A’-binds tj), it is not a barrier, and so vowel harmony may be further extracted.  See Fiengo et al. (1988) for an  attempt to derive the effect of (147) within the framework of Chomsky (1986). This analysis seems promising, in that the extraction in St’át’imcets would mirror the extraction in the English sentences (146). The non-initial conjunct in St’át’imcets is probably best regarded as an ungoverned domain, since—like an adjunct—it is in an adjoined position . 1P ) (specifically, the XP in which it is contained is adjoined to the initial conjunct 1 subject position is English is similarly ungoverned.  The  In both languages, A’-movement  (topicalization in English and scrambling in St’át’imcets) therefore facilitates further extraction, which would not otherwise have been allowed if it had occurred from an in situ 41 ungoverned domain. The structural approach to coreference across conjuncts has a particular advantage over the Optimality Theoretic approach outlined in § when one considers a sentence like (100)—repeated here as (148)—since the peripheral location of the overt nominal Bill permits it to simultaneously be the topic of the sentence while showing the usual one-nominal interpretation effect within its own conjunct.  41  A test that could help to confirm whether a scrambled NP had indeed become the discourse topic (suggested by Henry Davis, p.c.) would be to include a ‘post-context’ having a pro subject after the sentence of interest. More research on the properties of scrambling in St’át’imcets that might confirm or refute this analysis remains to be done. 79  (148)  áts ‘x-en-as  kw-s  s-qvlqvl-ts-mIn’ -as  nilh  Bill  conj nom-bad-mouth-appl-3sg. conj det-nom Bill /(ii) *Bjll. swore at him’ 1 1 saw him, and then (i) he swore at Bill a. ‘he /(ii) *Bill swore at him’ 1 b. ‘he saw him, and then (i) hej swore at Bill  see-tr-3erg  (AA, GN, RW 1594)  The interpretations in (a.ii) and (b.ii) are not allowed, because they would require that the single overt NP Bill be the subject of the second conjunct, violating the one-nominal interpretation effect discussed above in §4.2. The expected reading is (b.i). In this structure, shown below in (149), Bill remains within the VP of the second conjunct and is interpreted as the object.  Because there is no disjoint, overt nominal elsewhere in the sentence that can be  mapped to topic, the null topic is coindexed with John from the context (129).  This topic is  then coindexed with the pro subject of each conjunct, satisfying the topic constraint (115). Top-P  (149) ej  IP  xP  Ipi  VP  /  NP  X  pros  nilh  2 1P VP  NP  1 áts’x-en-as pro  see-tr-3erg  pros 1 s-qvlqvl-ts-mIn’ -as kw-s Bill  nom-bad-mouth-appl-3sg. conj det-nom Bill  In a sentence like (135), which has two overt NPs—one in each conjunct—there is a single interpretation for the sentence. There is also only one possible reading in a sentence in which there is a single overt NP in the initial conjunct, as in (142).  However, in a sentence like  (148), which has a single overt NP in the non-initial conjunct, there are two possible readings, since the discourse may supply the topic, as shown in (149), or the single, peripheral NP may serve as the topic.  The latter interpretation is (148a.i), in which Bill is the topic of the  sentence at the same time that it is the object of the second conjunct.  80  After Bill moves to  Spec/Top-P (perhaps after having scrambled to an A’-position, as suggested above), it is , and the object of this conjunct is free to 1 coindexed with the subject of 1P of  . 2 John from the context (129), as must the subject of 1P  take the reference  The well-formed structure  corresponding to (148a.i) is therefore as follows: Top-P  (150) 1 Bill  IP  xP  Ipi  VP  /  NP  X  prof  nilh  2 1P  NP  VP  áts’x-en-as pros  see-tr-3erg  proj 1 s-qvlqvl-ts-mfn’ -as t  nom-bad-mouth-appl-3sg. conj  2 may not corefer with Crucially, the subject of 1P : the trace of 2 crossover configuration in 1P  Bill, since this would create a strong  Bill would be locally A-bound by the pro subject,  resulting in a Condition C violation. *  (151)  Top-P  1 Bill  IP  xP  Ipi  VP  /  NP  X  1 pro  nilh  2 1P  VP  NP  áts’x-en-aspro  see-tr-3erg  1 pro s-qvlqvl-ts-mIn’ -as t 1  nom-bad-mouth-appl-3sg. conj  81  Whatever principle allows the overt NP in the second conjunct of (135)—which has a single overt NP in each conjunct—to be raised to topic position applies with equal force to (148). The intuition that underlies this account is that such NPs appear at the periphery of the clause and/or occupy the highest position (exactly as does the subject/topic in ordinary VOS clauses). As noted above, following Henry Davis (p.c.), this might be encoded structurally as scrambling of the peripheral object to an A’-position (e.g., adjoined to XP) before mapping it to topic position (in the case of objects). Other approaches are conceivable, but it is clear that a structural account of these asymmetries is preferable to an Optimality Theoretic account. This is especially obvious when trying to capture the ambiguity of a sentence like (148). As was noted in § (p. 50), the tableau for this sentence in (101) uniquely identified as grammatical the reading with the overt object as non-topic, incorrectly dispreferring the reading in which the object is topic. The structural account proposed in this chapter, on the other hand, predicts exactly the attested ambiguity. The interpretations that are allowed across transitive conjuncts having various combinations of overt nominals are summarized in the chart below, in which pro represents a null pronominal, and NP represents an overt R-expression. 1 1P  Topic  (152)  Constraint  (115’) (130) (132) (89) (135) (142) (148a.i) (148b.i)  *  *  object prod prod NP NP NP prod 1 pro  subject 1 pro 1 pro 1 NP 1 pro 1 pro 1 pro 1 pro  subject 1 pro 1 NP 1 pro prod 1 pro prod proj  1  object prod proj 1 NP proj NP, NP  The structures that violate the topic constraint are (135) and (148a.i). The common property . Note (1P ) of these structures is that they have a single overt NP in the non-initial conjunct 2 that the topic constraint is only violable in this non-initial conjunct. Sentence (89), which has two overt NPs in the first conjunct, obeys the topic constraint in all conjuncts; note moreover  82  that LF extraction of the subject to Spec/Top-P must be allowed from the initial conjunct, and that this is presumably allowed because the overt NP subject is the highest element in that clause. For completeness, a table of the excluded interpretations for the same set of sentences is given below: 1 1P  Topic  (153)  Constraint  (115) (130) (132) (89) (135) (142) (148)  * * *  *  object pros prod NP 3 NP NP 1 pro  subject pro, 1 pro NP, 1 pro 1 pro 1 pro  subject prod NP pros pro, prod 1 pro  object 1 pro NP, 1 pro NP 1 pro NP  Parallel interpretations for sentences having a single overt NP in the second conjunct are ungrammatical, despite satisfying the topic constraint in that conjunct.  4.3.2. Intransitives Section 3.3.2 showed that parallelism—the preference that subjects corefer—does not hold 42 One case where it does hold for strongly across conjuncts if one of them is intransitive. most speakers, though, was (103), repeated here as (155), and which has the context (154): (154) wa7 k’ál’-em kw-s prog wait-intr det-nom ‘John is waiting’  John John  Bill nilh s-7áts’x-en-as (155) p’an’t kw-s return det-nom Bill conj nom-see-tr-3sg. conj 1 returned and (i) he saw him/(ii) *he. saw him’ ‘Bill  (AA, ON, RW 1586)  The effect of the semantics of intransitive predicates in determining coreference has not been investigated, though it should be. In an examination of Italian, for example, Calabrese (1990: §6) shows that unaccusative verbs do not have consistent focus characteristics, these properties apparently being determined by whether or not the verb is stative. 42  83  Note that the subject of the first conjunct is preferably coreferent with the subject of the second If the overt nominal subject of the intransitive first  conjunct (the interpretation in (i)).  , coreference between the subjects in each conjunct follows 1 conjunct occupies Spec/1P straightforwardly. The overt nominal Bill moves to the topic position, which is then coindexed , as schematized below: 2 with the pro subject of 1P Top-P  (156)  IP  Bill  xP  Ipi VP  NP  X  / \  1 t  nilh  A  2 1P NP  VP  p’an’t return  1 pro s-áts x-en-as prod nom-see-tr-3sg. conj  Like sentences having two transitive conjuncts (4.3. 1), sentences having at least one intransitive conjunct allow LF movement of an NP from the initial conjunct to Spec/Top-P as long as the overt NP being moved is the subject of that conjunct. A structure in which transitive and intransitive subjects occupy the same position at 5structure (Spec/IP) gives the correct result for (155), but not always for (111), repeated for convenience as (157): s-John (157) qvlqvl-ts-mfn ‘-as bad-mouth-appl-3erg nom-John 1 swore at John, and then a. ‘Bill b. ‘John swore at Bill, and then  s-Bill nom-Bill pj left’ 1 he hej,* left’  nilh conj  s-qwatsáts nom-leave (AA, GN, RW 1592)  The VOS interpretation of (a) is preferred by most speakers, but the VSO interpretation of (b) is also allowed by speakers who have fairly free word order; note that in both cases, the  84  subject is mapped to topic, and this topic corefers with the subject of the intransitive in the second conjunct.  In interpretation (b), the NP John would have to be scrambled and/or  topicalized in order that VSO order could be derived in the first conjunct, and this representation then apparently requires that the derived subject John corefer with the pro subject of the intransitive clause. Of particular interest, though, is the reading in (a) marked by ?, which is meant to indicate not that this reading—which is not parallel—is marginal, but that its grammaticality has varied across consultants and across elicitation sessions, and has been difficult to settle with certainty.  This variability is not surprising, since St’át’imcets  exhibits ergativity, which entails that intransitive subjects pattern with direct objects with respect to certain processes. The variability seen with sentences like (157) suggests that the intransitive subject functions sometimes as the topic, but at other times as the non-topic.  A  simple way to explain the variability is to base-generate intransitive subjects within VP. Note that intransitive subjects get the case of an object (absolutive), hence it need not raise at S structure in order to get ergative case in Spec/IP, as must a transitive subject.  Intransitive  subjects may remain in situ at LF when functioning as non-topics, and so the topic constraint will not exclude such representations.  When an intransitive subject functions as a topic,  however, it is because it has raised to Spec/IP at or before LF, hence is subject to the topic 43 constraint and consequently must corefer with the subject of the initial conjunct. Returning to (157a), the parallel interpretation has the following structure, with the 1 moving to topic position at LF and then being coindexed overt transitive subject Bill from 1P . 2 with the intransitive pro subject in 1P  It is uncontroversial for the object to receive case within VP, the problematic GF of course being the intransitive subject. No specific claim will be made concerning whether intransitive subjects get case in Spec/VP, as sisters of the verb, or perhaps through a relation with some other functional projection outside of the VP; note that the last possibility seems unlikely, since absolutive agreement (unlike ergative agreement) is not overtly marked on the predicate. It would be desirable if transitive and intransitive subjects shared the same position at D-structure, since some aspects of the syntax of St’át’imcets—such as the formation of imperatives and jussive complements—are clearly accusative. ‘  85  Top-P  (158) 1 Bill  IP xP  Ipi VP  2 1P  NP  X  t  nilh  NP  VP  qvlqvl-ts-mI&-as s-John bad-mouth-appl-3erg nom-John  1 pro s-qwatsáts nom-leave  , and so the structure of 2 For this particular reading, then, the intransitive subject is in Spec/1P (158) is accusative, since the transitive subject in the first conjunct is also in Spec/IP. In order to derive the non-parallel reading in (157), however, the intransitive subject must remain in Spec/VP (but see the previous footnote), because in this lower position it need not be coreferential with the topic Bill, as the following structure shows: Top-P  (159) 1 Bill  IP xP  Ipi VP  NP  X  t.  nith  2 1P VP  qvlqvl-ts-mfn’-as s-John bad-mouth-appl-3erg nom-John s-qwatsáts prod nom-leave 2 that needs to be coindexed with the topic, and the topic There is no VP-external subject in 1P constraint (115) is also satisfied, since the pro inside VP in the second conjunct does not corefer with the topic Bill; instead, it is free to corefer with the other non-topic, John. 86  In  summary, the two readings of (157a) therefore share LF movement of the subject Bill from the first conjunct to topic position. Their different interpretations are derived from the position of the intransitive subject, which in a split-ergative language like St’át’imcets may appear in Spec/IP when behaving accusatively, and within VP when behaving ergatively.  As noted  above, perhaps the semantics of the intransitive predicate can account for the variation, though this remains to be investigated. In conclusion, the facts concerning coreference across conjuncts in St’át’imcets may be explained by mapping grammatical functions to discourse functions at LF.  Specifically, the  grammatical subject is canonically mapped to topic position (Spec/Top-P) in accord with the topic constraint (115) (although the topic constraint is violable in the second conjunct of sentences that have a single overt, peripheral NP). In so far as coreference across conjuncts is determined by this mapping of grammatical functions to discourse functions (Reinhart 1981 and Erteschik-Shir 1993), coreference in St’át’imcets may be regarded as having an ergative pivot, since it has been seen that the topic constraint is a constraint on the mapping of ergative NPs (transitive subjects) to topic position. Additionally, St’át’imcets displays a split between transitive subjects (ergative) and intransitive subjects (absolutive); specifically, absolutive subjects display variability in the way they satisfy the topic constraint.  Note that in an  accusative language like Italian (Calabrese 1990), there is a similar split between unaccusative subjects on the one hand, and transitive and unergative subjects on the other.  This is  interesting, since in St’ at’ imcets, all intransitive subjects are unaccusative, in the sense that they get the case of D-structure objects. It remains a topic for investigation to determine in what further respects intransitive subjects behave alike in both types of languages. The structural asymmetries proposed above may be fruitfully extended to explain the ergative/absolutive extraction asymmetries in relative clauses, the subject of the next section.  4.4.  Extraction in relative clauses  Chapter 2 established that there is an ergative/absolutive asymmetry in relative clause formation in St’át’imcets.  Specifically, pronominal morphology corresponding to the  87  relativized constituent is obligatorily retained in ergative-centered RCs, but obligatorily gapped in absolutive-centered RCs. Consider the 0-centered RC in (58), repeated here as (160), and the S-centered RC data of (44, 50), consolidated below as (161): s-John müta7 s-Bill (160) nilh wi conj nom-B nom-J pl.det foc ‘it was John and Bill that I saw  lhkiinsa i pl.det see-tr-(-Ø/3p1.obj]-lsg.conj-det now (LT 2338, 2339)  (161) tsukw t’u7 s7ents [ ti tsIcw{ø/*_an/*_as}a] det go- {-ø/lsg. conj/3sg. conj)-det part lsg. emph only ‘I am the only one who went’ (AA, GN, LT, RW 1200, 1201) The obligatory absence of pronominal morphology on the RC predicate in each of these structures suggests that RCs in St’át’imcets do indeed involve extraction of a constituent from the position that is relativized. Because this extraction is not seen in the form of an overt relative pronoun, the extracted element is presumably a null operator that moves to Spec/CP; from this position it binds its trace, which is a variable. The structure of the cleft construction in (160) is in all relevant respects as in (162). The focus marker nilh selects a DP complement (John and Bill; note that the St’át’imcets sentence has an overt, plural determiner wi). The 0-centered relative clause forms the residue of the cleft construction; it is also a complete DP because of the discontinuous, plural determiner i.  .  .  ) is empty 1 -a. The head of the NP that is selected by this determiner (NP  (though coreferent with John and Bill) and omitted from the diagram for conciseness, as indicated by the dashed line between DP and CP.  The cleft residue is therefore more  specifically a headless RC, and so (160) might be more literally translated as ‘it was John and Bill, the ones that I saw’. Further discussion follows the diagram:  88  ‘P  (162)  DP  XP D  DP.  X  CP  nil,,Z  prof  IP  wi s-John mita7 s-Bill p1. det nom-J conj nom-B pl.det...-det VP I  lsg.conj  prod  1 ats’x-en t see-tr  Because the target of relativization is absolutive, the trace of the null operator  absolutive case from the predicate inside VP.  (tj)  receives  The operator itself is assumed to be a null  pronominal pro in Spec/CP. Because pro is non-referring and never used deictically, it needs an antecedent; following Browning (1987) and Cinque (1990: 160), the reference of a pro operator is identified via agreement between CP and the head of the relative clause (here, the null head, not shown, which is coreferent with John and Bill). The pro subject in (162) (pro) is base-generated in Spec/VP, but raises to Spec/IP in order to be identified by the lsg subject morpheme -an in I(nfl). As indicated by the arrows, the predicate raises to I (in order to take the lsg subject suffix -an) and finally to D, where it appears with the discontinuous determiner.  Intermediate steps in the raising of V as are necessary to satisfy the Head  Movement Constraint (HMC) (Chomsky 1986: 71) are also assumed, but omitted from the diagram for clarity. A similar derivation applies to the S-centered RC in (161). The constituent tsukw t’u7 is a focus marker analogous to nilh (van Eijk 1985: 279, §38.5; 1987: 72), and so it appears under X, selecting the lsg emphatic pronoun s7ents.  89  The intransitive subject pro is base  generated within VP, per the discussion in the previous section. After it moves to Spec/CP (it is an operator), absolutive case is assigned inside VP to the trace of the operator  (tj):  (163) XPDP X  A  / \  tsukw t’ u7 only part  DP  A  / \  s7ents lsg. emp  CP  D ti...-a det -det  IP  prof  ...  VP  A  tsicw t go The predicate tsicw ‘go raises to I, then to C, deriving a complete S-structure. Intermediate raising from V to I is needed to satisfy the HMC, but it is omitted for clarity in the diagram, since there is no affix under I in this sentence: (164)  XDPDCP  tsukw t’u7 only part  s7ents lsg.emp  IP  1 pro  ti tsfcw -a det go-det  A ti  ti  The topic node has also been omitted from the above structures, since it is irrelevant.  90  The structures in (162-164) give an idea of how absolutive-centered RCs are derived. Ergative-centered RCs differ from absolutive-centered RCs in that the 3sg subject morphology corresponding to the relativized constituent must be retained, or else the topical object marker 44 Recall (38) from chapter 2 (p. 22), repeated here: -tall must appear on the predicate. (165) áts’x-en-as [ ti sqáycw-a see-tr-3erg det man-det ‘he saw the man that hit Bill’  tup*unh..*{_tali/as}(h)a s-Bill ]j det hit-tr- (TO/3sg. conj]-det nom-Blil  [ ti  (ON 141; ON, RW 218; LT 2581)  45 require either -tall or -as in ergative-centered RCs, and so the appearance of All consultants both morphemes needs to be explained. The conclusion of chapter 2 (2.3) emphasized that this sort of restriction on relativization of ergatives is common in many languages outside of Salish, and so it should be possible to give a principled account of this ergative/absolutive asymmetry. An explanation is indeed possible by adopting a version of the VP-internal subject hypothesis (Koopman and Sportiche 1991 and the references cited there). It has already been Suppose that ergatives likewise  assumed that intransitive subjects are generated within VP.  originate in VP, specifically in Spec/VP (for the sake of concreteness, though they could just as easily be in some other projection selected by I, depending on how one analyzes the transitive affix), and that they move to Spec/IP in order to get ergative case from the ergative marker -as in I—the absolutive case that is available within VP being needed by the object. A sentence like (5), repeated here as (166), therefore has a structure like (167) after caseassignment to the subject ta smiflhatsa ‘the woman’ that has raised to Spec/IP: ta sqáycw-a (166) áts’x-en-as det man-det see-tr-3erg ‘the woman saw the man’  ta det  smülhats-a woman-det (RW 48; Matthewson 1993a: 2-3)  The passive morpheme -m is sometimes, but not always, interchangeable with the topical object marker -tall. The former appears to occur in more restricted environments than the latter, but the reasons for the variation remain unclear. Except ON (Henry Davis, p.c.). ‘  ‘  91  ‘P  (167)  NP  I’  -as  smülhats-a ta 1 det woman-det  VP  I  NP  V’  3e>/\ ats’x-en ta sqáycw-a see-Ir det man-det moving to V-to-I raising derives the S-structure, and the subject is mapped to topic at LF by Spec/Top-P: Top-P  (168)  IP  NP  NP  I’  ta smülhats-a det woman-det I  VP  ts’x-en -as V’ 3e, /”\ -tr,,,/ see / 5  t NP  t ta sqáycw-a det man-det Returning to the variant of the ergative-centered RC in (165) with the 3sg subject in marker -as, the representation of the complex NP after operator movement would be as ti), (169). The null operator originates in the VP-internal subject position (marked below by te’s from which it moves to Spec/IP in order to receive case from -as in I (the predica absolutive case having been assigned to the object, Bill).  The operator ultimately lands in  Spec/CP, the A’-position from which it binds the variable in Spec/IP (the case-marked t2):  92  (169)  A  CP  D  ii sqáycw-a det man-det ti det  ...  ...  -a -det  C’  pros  IP  C  NP  I’ VP  I -as 3sg.conj  1 t2 NP ti i  tup-un’ s-Bill hit-tr nom-Bill There is thus an explanation for why relativization of ergative requires that the person morphology corresponding to the relativized constituent be retained: if the 3sg subject marker  -as were not present under I, neither ti nor t2 would be in a case-marked position, and so neither could function as a variable. The pro in Spec/CP in such a structure would violate the ban on vacuous quantification.  In the absolutive-centered RCs diagrammed in (162-164),  however, there is a single argument that originates within V’—the null operator pro—and because its trace at S-structure receives absolutive case from the predicate, a well-formed operator-variable chain is created.  In this analysis, then, both ergative- and absolutive  centered RCs have extraction of an operator. Note, incidentally, that Jelinek’s (1984) analysis of pronominal morphology is incompatible with the present analysis of RCs, since for Jelinek, pronominal affixes in radical head-marking languages bear theta-roles (i.e., they are the actual arguments).  In this account, however, which does not regard relativization as having a  syntactically ergative pivot (since all RCs have extraction of an operator), the ergative  93  pronominal affix is not itself the argument, but instead serves to case-mark an argument trace (thus licensing an operator-variable chain). Further evidence that variables (wh-traces) need case is found in English. Consider the following contrast, in which a wh-word is extracted from the subject position of an embedded clause (Lasnik and Uriagereka 1988: 90: (170) a. b.  1 [is it likely who  [tj  will win the race?]]  1 [is it likely [tj to win the race?]] *who  The embedded subject position is case-marked in (a)—because of the finite Infi in the embedded clause—but not in (b), which has an infinitival embedded clause. Both sentences are otherwise identical, and so the ill-formedness of (b) must be due to the fact that the wh trace lacks case. The structure in (169) does not yet have a topic. The head of the RC, ti sqáycwa ‘the man’, may not lower to Spec/Top-P, downward movement being generally prohibited. Nor may the null operator in Spec/CP raise to Spec/Top-P, since pro—being non-referential—is excluded from topic position. (The null pronominal pro is as a bound variable, hence itself is always topic-bound, a point that was illustrated above (p. 66) for both St’át’imcets (130) and Italian (125-126).) An additional reason that the relativized ergative is excluded from serving as the topic is that the head of the RC is the focus of a subordinate predication (Erteschik-Shir 1993: 35-39, §6.2.2); since the RC head ti sqáycwa ‘the man’ is coreferential with the ergative operator—the person understood to be hitting Bill in (165) is the man—moving the operator into topic position would illegitimately entail that in the subordinate predication the focus and the topic would have the same referent. Another option is that the object, Bill, becomes the topic in (169); this possibility is forbidden by the topic constraint (115), however, and in any case it will be reserved for the variant of (165) that contains the topical object marker -tall rather than the 3sg subject marker -as.  (As will be seen below, sentences with -tall signal  focus of the subject rather than the object.) Instead, it will be assumed (following Erteschik Shir 1993) that the entire clausal portion of the relative clause in (169) denotes the subordinate 94  topic.  The only remaining step in this derivation is therefore V-to-I, I-to-C, and C-to-D  raising, as indicated below by the arrows: DP  (171)  A  CP  D  ti sqáycw-a det man-det  ti...-a -det det  C’  1 pro  ...  IP  C  NP  I’  v  i -as  1 t2 NP  3sg. conj  •  tup-un’ s-Bill hit-tr nom-Bill The restriction on extraction/relativization of topics is not unique to St’át’imcets. The same restriction is particularly salient in Japanese, for example, since topics are overtly marked by the morpheme  -Wa,  an example of which is given in (172).  A homophonous  morpheme indicates contrastive focus, however—as shown in (173), in which the first NP bears the topic  -Wa,  and the second NP bears the contrastive-focus  Japanese sentences are from Uechi (1994). ooi (172) Nagano-wa hito-ga Nagano-top people-nom many ‘speaking of Nagano, there are many people there’  95  -Wa.  These and subsequent  ooi (173) Nagano-wa hito-wa many Nagano-top people-contr ‘speaking of Nagano, it is people, not others, that are many’ Recall from (118) above (p. 61) that focus is associated with wh-questions. In particular, the constituent that answers a wh-question is the focus of the sentence.  The reason for this, of  course, is that a wh-question has the communicative goal of seeking new information, and so the answer to a wh-question comprises non-topical, or new, information. In contrast, the topic of a sentence typically refers to something that is presupposed and familiar from the discourse. It is reasonable, then, that non-topics—but not topics—would be questioned, and this intuition is syntactically realized in Japanese.  Both RC heads and wh-words may be marked by the  focus morpheme -ga, as shown in the (a) examples of (174) and (175) respectively. Crucially, however, neither a relative clause nor a wh-question may be formed with the topic marker -Wa,  as the corresponding (b) sentences show.  Rather, only the contrastive-focus  interpretation is possible for the -wa-marked NPs (174b, 175c): (174) a.  b.  (175) a.  b.  c.  Tamako-ga Tamako-nom (i) ‘the peach (ii) ‘the peach  sukina momo peach like Tamako likes’ (wide focus) TAMAKO likes’ (narrow focus)  Tamako-wa sukina momo peach Tamako-nom like ‘the peach which TAMAKO, not others, likes’ (contrastive focus only) kimasu ka? Kyoo-wa dare-ga today-top who-nom come interrog ‘as for today, who is coming?’ *Kyoowa dare-wa kimasu ka? today-top who-top come interrog ‘as for today, speaking of who, is s/he coming?’ kimas-en ka? kuru kedo, dare-wa Kyoo-wa dare-wa who-contr come-not interrog today-top who-contr come but ‘as for today, who is coming, but who is not coming?’  96  A similar paradigm can be constructed in English. The as for x construction carries topical information; examples are the following: (176) a. b.  As for Sally, she writes poems and short stories As for donuts, Fred eats them for breakfast  A topic may appear in a wh-question if the topic and the wh-word have distinct referents, and as long as the topic appears higher than the wh-word: (177) a. b. c. d.  As for Sally, what does she write? *What, as for Sally, does she write? As for donuts, who eats them? *Who, as for donuts, eats them?  However, an as for x construction may not be used to ask a wh-question directly, since the wh word may not simultaneously be the topic and the focus of a sentence.  The following  sentences are parallel to the Japanese sentence in (175b): (178) a. b. c. d.  *As *As *As *As  for for for for  who, (she) writes poems and short stories? who, does she write poems and short stories? what, Fred eats (them) for breakfast? what, does Fred eat (them) for breakfast?  These data are reminiscent of the topic/focus structure subordinate predication in the RC diagrammed in (169). Just as the focused head ti sqdycwa ‘the man’ could not lower to topic position (since the topic and focus would illegitimately have the same referent), neither may the wh-word in (178) be both the topic and the focus of the sentence. To summarize, the obligatory retention of pronominal morphology corresponding to relativized ergatives in St’át’imcets can be explained by Case theory and by a structure in which ergative NPs occupy a higher position at S-structure than do absolutive NPs. The lack of a topic in these relative clauses probably represents a universal phenomenon, as suggested by the Japanese and English data cited above. See also Schachter (1973) for instances of other languages in which RCs and focus/cleft constructions resemble each other.  97  The final task is to explain the difference between an ergative-centered RC with -as as opposed to one with the topical object marker -tall.  The discourse facts that need to be  captured in the syntax are described thus by Matthewson (1993a: 4-5): [The] parallelism between the object of an ordinary transitive clause and the subject of a -tall clause extends to their discourse function; these are the slots where entities new to the discourse are introduced. Conversely, that entity which is already ‘under discussion’ and is the ‘topic of the discourse’ tends to occur in the subject position of an ordinary transitive clause or the object position of a -tall clause. The structure in (171) gives exactly the desired consequence for an ordinary transitive clause, since the object has not been mapped to topic; instead, it remains within VP, where it is associated with the focus, pursuant to Diesing’s (1992: 9-10) Mapping Hypothesis or some similar theory that relates syntactic structures to logical representations (as discussed above in §4.1).  Given the different discourse function of NPs in a -tall clause, though, a different  derivation is needed for those structures.  The most sensible way to facilitate the required  derivation is to base-generate -tall within VP, and to suppose that—like 3sg -as—it has its own case to assign. The D-structure for the variant of (165) that contains -tall would therefore be as follows: (179)  A  ti sqáycw-a det man-det  IP  D ti...-a -det det ...  VP  I  NP  tup-un’-táli s-Bill hit-tr-TO nom-Bill  98  The morpheme -tall may occupy a separate projection within VP—perhaps Agr-Object. For the sake of clarity, however, it is shown here as affixed to the predicate within VP, since this account requires only that -tall occupy a projection lower than I.  Within VP, there are  therefore two cases available: one from the predicate (which is received by the object, Bill, as usual), and the other from -tall, which is free to be assigned to Spec/VP. Because Spec/VP is a case-marked position, the operator pro can move directly from Spec/VP to Spec/CP. The predicate raises and ultimately lands in D, and the S-structure in (180) is derived: (180)  A  ti sqäycw-a det man-det  CP  D  ti tup-un’-táli-ha det hit-tr-TO-det  1 pro  C’ IP  C I’  VP  I  NP  V’ 3 s—Bill t nom-Bill  tj  The operator pro may not move to Spec/Top-P for the same reasons that it could not do so in the relative clause with -as in (169). However, because the object of a -tall clause is the slot in which presupposed information appears, the object Bill must raise at LF, landing in Spec/Top-P.  From there, it A’-binds its case-marked trace.  therefore be as follows:  99  The LF representation would  DP  (181)  A  ti sqáycw-a det man-det  D  ti tup-un’-tá1i-ha det hit-tr-TO-det  cP  CP  NP s Bulk nom-Bill  1 pro  C’ IP  C II  I  VP 1 V  tj  tk  NP ti  Such a derivation violates the topic constraint (115), since an NP under V’ (i.e., the trace of the object tk) is coreferential with the topic. However, this violation alone does not preclude the analysis outlined here, since the topic constraint (115) is intended to represent only the canonical mapping of subject to topic.  Another violation of the topic constraint was seen  above in (150) (p. 81), in which LF topicalization of an object was licensed by its peripheral position in the clause. It remains to be explained why -tall forces the object to raise to topic.  When the  subject is focused, as it is in -tall clauses (which include wh-questions, clefts, and RCs), it must also be extracted. Perhaps because the VP in a -tall clause already contains a focused constituent (i.e., the trace in subject position), the object must move to Spec/Top-P at LF in order that there not be two focused constituents in the same domain. The topic constraint can be violated also in English, as an intonationally prominent subject is focused. This is exemplified below: 100  (182) a. b.  Fred hates Bill FRED hates Bill  Sentence (a) with normal intonation has the unmarked mapping of subject to topic, in which Fred is the topic. In (b), though, with stress on the subject, Fred, this NP is interpreted as the focus, not the topic. Clearly, a language may have various strategies for altering the subject-to-topic mapping, and -tail represents another such strategy that is available in St’át’imcets. Because there are two cases available to be assigned within VP in a -tali clause, the relative clause examined here could conceivably have the same representation at S-structure as at LF. Such a structure, shown above in (180), would resemble the one with -as in (169), in that the topic of the subordinate predication in (180) would be the clausal portion of the RC (following Erteschik-Shir 1993).  In a structure like (181), though, the function of the topical object  marker -tall is precisely to allow the object of a transitive clause to function as the topic within the RC itself, which has its own topic/focus structure independent of the matrix predication. A benefit that accrues from this analysis is that the subject of a -tall clause remains in Spec/VP, within the domain that is mapped to the nuclear scope. Because the nuclear scope is associated with the focus, it is possible to explain why some speakers—notably RW (and probably also LT, based on the limited elicitation with her regarding these data)—prefer and usually require -tall in sentences in which the subject has been focused (whether relative clauses, clefts, or wh-questions).  The head of an RC is the focus of the subordinate  predication, according to Erteschik-Shir (1993), and because this head corefers with the extracted operator, it would naturally be preferable if the variable was within the domain of focus (VP)—as it is in the LF representation for the -tall clause in (181). In an RC with -as, however, the ergative operator must raise out of VP in order to get case in Spec/IP. In the structure in (169), then, the variable t2 corefers with the focused RC head ti sqáycwa ‘the man’, but the variable itself will not be in the domain of focus, since it has escaped VP. This fact may be the reason why ergative extraction with -as instead of -tall is ungrammatical for 101  The more general question concerns why focusing the subject with -tali  some speakers.  requires it to be extracted; it is likely related to the topic constraint, though this question is left to future research. This analysis explains the retention of ergative morphology and topic-focus structure only with respect to relative clauses, and more specifically with regard to a subset of data. It remains to be explained why some consultants (GN is one) can extract an ergative NP with neither -tali nor -as on the predicate. In the meantime, the present analysis offers a framework within which such additional data may be addressed.  Further fieldwork will determine  whether this system is adequate to explain the additional facts while remaining a restrictive and coherent theory.  4.5.  Summary  This chapter has shown that the asymmetric behaviours of subjects and topics in St’át’imcets may be explained within the Principles and Parameters model of syntax (Chomsky 1981; 1982; 1986; 1992; etc.). Radical head-marking languages and languages displaying ergativity have been integrated into this approach to syntax with limited success, since the model has been built mainly upon well described, accusative Indo-European languages. However, recent work on topic-focus structure (Reinhart 1981; Calabrese 1990; Diesing 1992; Erteschik-Shir 1993; Partee to appear) and the position of subjects in hierarchical structure (Koopman and Sportiche  1991 and the references cited there) is beginning to reveal how typologically diverse languages converge structurally. Section 4.2 demonstrated that the effects of the one-nominal interpretation law may be derived by identifying the nuclear scope of a sentence as the structure below VP. Section 4.3 showed that parallelism in coreference across conjuncts is the result of an LF principle—the topic constraint (115)—that maps the subject of a sentence to a structural topic position (Spec/Top-P).  Finally, §4.4 accounted for extraction asymmetries in relative clauses by  employing Case theory to ensure that ergative NPs are higher than absolutive NPs at S structure. The topical object marker -tali appears when the subject is focused, and it moreover  102  requires that the subject be extracted.  Most of the analyses in this chapter assume that  grammatical functions are mapped to discourse functions at LF, and that this mapping is constraint by the topic constraint. All of the constraints and asymmetries outlined in this chapter have appeared in the literature in various forms. Although very specific, all appear to underlie a broader linguistic system, since they manifest themselves universally in discourse. It would be surprising to find a language, for example, that regularly marked a topic by repeating it throughout the discourse as an overt NP, or that had a device for questioning given information in the discourse. All of these discourse-based notions are amenable to a structural analysis. That this analysis applies with equal validity to languages as genetically and typologically diverse as St’át’imcets, Italian, Japanese, and English—and others sharing their characteristics—strongly suggests that it is capturing not merely language-specific features of sentence structure, but properties of Universal Grammar.  103  Chapter 5 Conclusion  The principal contribution of this thesis has been to describe some hitherto uninvestigated syntactic properties of St’át’imcets, with a view to determining specifically to what extent the language is syntactically ergative. Chapter 1 outlines some morphosyntactic characteristics of St’át’imcets, showing how it shares with other Salish languages the property of morphological ergativity in its pronominal inflection. Dixon (1979) identifies relativization and conjunction as two major processes in which languages manifest either syntactic ergativity or syntactic accusativity, and so chapter 2 explores relativization and chapter 3 investigates coreference in conjoined clauses. Both processes appear to be ergative in certain respects. Relativization in St’át’imcets has the appearance of syntactic ergativity, since ergative NPs are relativized by processes that are not employed in relativizing absolutive NPs. Specifically, certain speakers require that relativized ergatives have the topical object-marker  -tall on the relative clause (RC) predicate, while other speakers alternately allow the usual pronominal morphology (3sg -as) corresponding to the relativized ergative NP to be retained. Both sets of speakers have in common that they mark ergative RCs more explicitly than absolutive RCs.  In contrast, all speakers have an obligatory gap corresponding to the  relativized constituent in absolutive-centered RCs. This fact has not been previously noted, and it strongly suggests that St’át’imcets RCs involve extraction of the relativized constituent. Conjunction in St’át’imcets is constrained most strongly by the one-nominal interpretation law (Gerdts 1988) and by the topic constraint (Erteschik-Shir 1993), which gives the effect of a parallelism constraint on discourse functions (Matthewson 1993a).  When  coreference is examined in terms of a mapping of grammatical functions to discourse functions, it might be regarded as behaving ergatively, since the topic constraint applies specifically to transitive subjects.  St’át’imcets, like other radical head-marking languages,  may alternately omit or specify lexical NPs according to rules that are not well understood, but  104  chapter 3 supports the conclusions of Kinkade (1989; 1990) and Matthewson (1993a) that the principles governing pro-drop and coreference are heavily discourse-based. Chapter 4 accounts for the asymmetries outlined in the previous chapters within the Principles and Parameters syntactic framework.  The analyses there are tentative, and will  surely need to be refined as additional aspects of Stát’imcets are documented.  Data from  St’át’imcets have only recently been brought to bear on this syntactic framework (principally by Davis et al. 1993, Gardiner et al. 1993, Matthewson et al. 1993, and in other references by these authors), and the importance of fieldwork on this language cannot be emphasized strongly enough.  The relative clause and conjunction data in this thesis represent a small  contribution in this regard. Each chapter emphasizes the similarity of the St’át’imcets facts to those found in better described languages outside of the Salish family.  St’át’imcets poses several problems for  syntactic theories, but these problems have many precedents, and this fact must only reinforce the view that underlying the surface dissimilarity of genetically and typologically diverse languages is a unique, shared linguistic system.  105  Appendix A Morphology The words and morphemes listed here are given in orthography, a key for which is in the next appendix. An underline indicates that the following morpheme is an enclitic; if the underline follows the morpheme, then it is a proclitic. Page references are to van Eijk (1985), which should be consulted for detailed information on St’át’imcets. —  Personal affixation Possessive affixes (p. 170) sg  p1  1  n-  -kalh  2  -tsu (after lh, s)  -lap  -su_(elsewhere) 3  -i  -ts (after lh, s) -s_(elsewhere)  Object suffixes (p. 171) sg subject  transitivizer  1 p  suffix  I 1 2 3  ll  3pl  -tumc-al  -ts-al  -tumul  3sg, 2sg/pl  -tumc  -ts  -tumulh  lsg  -tumi(n)  -tsi(n)  -tumulh  3sg/pl  -tumi  -tsi  -tam-al’ap -tan-i  1 sg lsg, 2sgIpl  -wit  -Ø  -ø  3sg/pl  106  Subject suffixes (p. 171) a. Indicative (plain) p1  sg 1  -(lh)k-an  2  -(lh)k-acw  I  -as  3  trans.  mtr.  see §22.3.4 (p. 176), and ‘Passive’, below  -(lh)kalh  intr.  trans.  -(lh)k-al’ap  -ø  -it-as lsg/pl  -as-wit  -twit-as  -it-as  I  II  -wit  3sg/pl  2sg/pl object suffixes  b. Subjunctive (conjunctive) sg  p1  I  trans.  trans.  mtr.  see §22.3.4 (p. 176), and ‘Passive’, below  -at  intr.  1  -an  2  -acw  3  -as  -al’ap -it-as  -as-wit  -twit-as  I  -it-as  -wit-as  1111 lsg/pl  3sg/pl  2sg/pl object suffixes  Passive (p. 176) The theme argument must appear closer to the predicate than the agent. Passivized predicates are interpreted as having a ipi subject unless there is an additional, overt subject NP. Subjunctive forms are made by adding -as (3sg.conj) to the indicative forms, which are listed below with the object suffixes that they select: object  ( = theme) 1  2 3  [  trans I II I II I II  p1 -tumul-em  sg -tumc-al-em -ts-al-em -tumi-m -tsi-rn -turn -em  -tam-lhk-al’ap -tan-em-wit  107  (In)transitivizers (pp. 130 ff., 150-152) intransitivizers  transitivizers  I -s (ts after s or lh) -en-s  II -Vn, -Vn’ -nun/-nun’ -mm/-mm’ -cit -min-cit/-min’-cit  -cal -em, em’ -ø  Personal (emphatic) pronouns (p. 191)  1 2 3  sg s-7ents s-nüwa s-nilh  p1 s-nfmulh s-nuláp wi s-nilh  Wh-words (Davis et al. 1993) who what where when why how  swat ku s-tam’ ku n-ka7 lh (i)-kanm-ás-(as) lh kánem s-kás-ts-as lh  induces induces induces induces  conjunctive inflection conjunctive inflection nominalization conjunctive inflection  Demonstrative pronouns (deictics) (p. 198)  visible  invisible  this, these that, those that, those (farther) this, these that, those that, those (farther)  sg ts7a ti7 t7u kw7a ni7 ku7  108  p1 izá iz’ izü kwelha nelh kwelh  Determiners (p. 223) variable words  sg  I  individual  ki  ta /(t)i  na  mi  nelh  i  p1  absent  present  absent  present collective  proper  unknown  known  ku  kw  kwelh  wi  I  with a  without a  Encitics (pp. 231 ff.) a an’ hem’ ha k’a kelh ka ku7 malh ga7 tu7 t’th wen wi7 cwilh  cwilh, kaj reinforcement (with evidential antithesis interrogative possibility, surmise remote future, possibility obligation, expectancy guotative adhortative presupposed knowledge definite past demarcation of time well, but, so emphasis emphasis after all, it turned out to be  det evid  anti  interrog evid fut oblig guot adhort presupp def.past now part emph emph cons  Procitics (pp. 252-253) (7e)lh ku  ‘before’; links independent (non-subordinate) sentences links attributes to their objects; can often be dropped  109  nouns  239 232 237 237 234 233 233 234 231 236 231 232 232 238 238 235  Multi-clausal sentences (pp. 270 ff.) nilh kw (ku] s t complementizers; normally trigger conjunctive inflection  conj (subordinating conjunction) det; goes with s- nominalizer nominalizer; induces possessive det lh hyp i when.past  Primary prepositions (pp. 254-256) Deictics and articles starting with 7 change the 7 to k when following these prepositions: in, on, at, with, among (oblique) 1. toward, along; 2. by (with agent in passive constructions); (this preposition is often dropped) around, via 1. from, out of; 2. than (in comparisons)  1 e_ ken ihel Focus  nilh -tali  foc TO  tsükw t’u7 17wa7  only even  focus marker topical object: follows transitivizer in A-centered RCs with 3rd person object; subject affix is omitted predicate predicate  Auxiliaries wa7 ka  ...  a  progressive out-of-control  prog ooc  110  Appendix B Key to Orthography Data in this thesis are presented in the orthography that is used in van Eijk (1981, 1983) and Peters et al. (1992), with the exception of additional hyphens (-), which indicate morpheme boundaries. This orthography is the one that is currently the most widely used in St’át’imcets language courses. For a key to the earlier Bouchard and Powell alphabets, consult Peters et al. (1992: appendix B).  a  Americanist Phonetic Representation e,a  e  ,I,i  i  i  0  0  u ao v  U  p t ts k kw  p t  Orthography  Orthography qw x xw g gw r 7  a A  1 p  y  h c cw lh q  V  çw  y p’  ti  c’,ts’ k’ k” q’  X  ts’ k’ k’w q’ q’w z’ 1’ m’ n’ w’ y’ r’  X”  g’  + q  g’w ii  c,’”  k k’’  5  z 1 m n w  Americanist Phonetic Representation q’’  1 m II  w y  h  111  qW  1’ m’ w’ y’  e,I  Appendix C Consultants This appendix lists the principal language consultants for the Project on Lexical Interfaces with Phonology and Syntax in North West Coast Languages. The short biographies and genealogies were prepared from information offered by each individual, and they are given here with their permission. All of the consultants are presently participating in linguistics courses in St’át’imcets and Nla’kapmxcmn at the University of British Columbia and Simon Fraser University. Alice Adolph  Kawá7tu. Born in Léqem’ts (near Mount Currie) to Francis Felix Leo and Evelyn Sam on 13 March 1939, and raised in nearby Xit’lólacw;  attended St. Mary’s Residential School in Mission, B.C., from ages six to seventeen, during which period she also lived in Tsal’álh (Shalalth/Seton Lake) with her mother and stepfather, Solomon Peters; married Sam Adolph (Tsáqwemlha7) of Cáclep (Fountain); now living in Vancouver, but still a member of the Cáclep Band. AA was raised by her paternal grandparents, Felix Leo (Lha7q) and Susan Felix Leo (Xánaq’a7) until she was six, and she spoke only St’át’imcets until she Her maternal grandparents were Sam Jim (Sitii) and Caroline (Maggie) Jim of Sqátin (Skookumchuck); Sam Jim’s mother was Kawá7tu, and Maggie Jim’s parents were Peter Williams (an  attended residential school.  Beverley Frank  hereditary chief) and Lucy Williams (Mamahisi), all of Sqátin. Cdzll’. Born in Sék’wel’was (Cayoose Creek) to Baptiste (Han) and Catherine (KatlIin) Frank, where she still lives; attended Lillooet Public School for eight years before transferring to Kamloops Indian Residential School, from which she graduated in 1959; attended the University of British Columbia, from which she received a Bachelor of Education degree in Elementary Education. Until leaving for residential school in Kamloops, BF spoke St’át’imcets daily with her parents and community Her father was from Sék’wel’was, and her mother was from Nxwéysten (Bridge River); her maternal grandfather was from  elders.  Nqwátqwa7 Nxwéysten. Gertrude Ned  (Darcy),  and  her  grandmother,  Cázil’,  was  from  Born in T’ft’q’et (Lillooet) to Bernice Adolph on 1 February 1930, and raised by her maternal grandparents, Jack James of Nxwéysten (Bridge River) and Susan James of T’ft’q’et; attended St. Mary’s Residential School in Mission, B.C.—where she and RW were classmates—from age 112  seven until the seventh grade; thereafter attended high school in Kamloops; left school in 1951, married, and moved to Cáclep (Fountain), where she still lives. Susan James’ father was Ngay’tasq’et, hereditary chief of T’ft’q’et. GN was a nurse for seven years at Mountain View Lodge, where she specialized in long-term care aid, and for several years she has taught St’át’imcets at both the high school level and at Stsmál’tsa i Sqwéma (Children of the Mountain) Public School. Laura Thevarge  Born in Lil’wat7ül (Mount Currie) to Harry Dan and Placida Pascal on 1 June 1931; attended St. Mary’s Residential School in Mission, B.C., from ages eleven through sixteen; returned to Lil’wat7ül until she was twenty-four, when she was married; now living in Vancouver. Harry Dan was from Seabird Island and Placida Pascal was from Lil’wat7ül. LT was raised by her parents and maternal grandparents, Joseph Pascal Her paternal (hereditary chief of Lil’wat7ül) and Louise Pierre. grandparents were Dan and Rosemarie Michel of Seabird Island and Sqátin (Skookumchuck) respectively. LT spoke only St’át’imcets until she attended residential school.  Dorothy D. Ursaki  Born in Lytton to Charles Walkem and Christina Paul in October 1907, and raised in Spences Bridge; attended residential school at Yale, B.C., from ages seven through sixteen; has lived her adult life in Vancouver, and is now living in New Westminster. Christina Paul was born in Lytton, and Charles Walkem—who was the leader of the Cook’s Ferry Band—was born in Spences Bridge. His uncle, George Walkem, was at one time the premier of British Columbia. DU’s paternal grandfather, Hugh Blake Walkem, was born in Montreal and was a surveyor for the Canadian Pacific Railway; Hugh Walkem’s father, Charles Walkem, DU spoke only See Smith (1989: 102). was born in Ireland. Nla’kapmxcfn until she attended residential school.  Rose A. Whitley  Born in Cáclep (Fountain) to Sebastian (Nk’yáp) and Adeline Peters on 18 May 1930; attended St. Mary’s Residential School in Mission, B.C., from 1938 to 1946; spent her summer holidays in  K’wstátqwa7.  Cáclep and in the United States picking berries. RW’s paternal greatgrandparents were Peter Qwá7na and Pauline (Mutátkwa), Joseph Tsil.hüsalts (Cáclep’s first chief) and K’wswapáw’s, and Eustache Peter (Qatsk) and Rosalee (CaIts’a7). Her maternal great-grandparents were Billy Fountain (Yawá7tulh) and Seraphine (TIcnek), Joe Joseph (Lil’wat7ül) and Josephine (K’wstátqwa7). RW’s family ancestry is mainly ScwepmectsIn and Lil’wat7ülmec. Her parents also attended St. 113  Mary’s Residential School, and so RW grew up speaking both St’át’imcets and Sám7ats (English). She married Edward Napoleon of the Lillooet Band (T’ft’q’et), where she still lives. RW has numerous grandchildren and one great-grandson, and she holds a Bachelor of Arts degree in Native Studies and Social Work Practice from Evergreen State College (1987).  114  Appendix D Elicitation Sessions This appendix is an index to the St’át’imcets syntax database for the Project on Lexical Interfaces with Phonology and Syntax in North West Coast Languages, from which the data for this thesis are drawn. Session 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34  Tokens 1-27 28-42 43-59 60-78 79-114 115-149 150-167 168-191 192-230 231-256 257-334 335-372 373 374-375 376-434 435-5 18 5 19-536 537-601 601-669 670-743 745-796 797-835 836-918 9 19-954 955-976 977-1035 1036-1094 1095-1224 1225-1298 1299-1353 1354-1445 1446-1502 1503-1528 1529-1583 1584-1643 1644-1712 1713-1808 1809-1834  Consultants BF, ON, RW BF, ON, RW RW GN GN, RW GN BF,GN,RW BF (168-184), ON (all but 184) GN, RW GN, RW GN, RW GN, RW BF BF GN, RW GN, RW BF BF, RW ON, RW ON (675-743), RW ON, RW BF ON (836-875), RW BF ON, RW BF RW ON, RW ON ON ON, RW RW ON AA ON, RW BF, ON, RW (variously) ON, RW AA 115  Date 8 October 1992 29 October 1992 12 November 1992 26 November 1992 3 December 1992 8 December 1992 7January 1993 21 January 1993 4 February 1993 18 February 1993 4 March 1993 18 March 1993 25 March 1993 1 April 1993 21 April 1993 22 April 1993 6 May 1993 20 May 1993 16 September 1993 30 September 1993 13 October 1993 14 October 1993 27 October 1993 28 October 1993 4 November 1993 25 November 1993 9 December 1993 6 January 1994 20 January 1994 3 February 1994 17 February 1994 2 March 1994 3 March 1994 8 March 1994  35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53  1835-1870 1871-2034 2035-2064 2065-2083 2084-2191 2192-2212 22 13-2222 2223-2279 2280-2306 2307-2322 2323-2415 2416-2461 2462-2557 2558-2641 2642-2680 268 1-2695 2696-2737 2738-2746 2747-2759 2760-2806  AA BF, GN, RW (variously) GN, RW AA GN, RW (variously) AA AA AA AA AA LT LT GN, RW LT AA LT LT AA BF, RW (variously) AA  116  16 March 1994 17 March 1994 24 March 1994 30 March 1994 31 March 1994 6 April 1994 7 April 1994 13 April 1994 18 April 1994 20 April 1994 22 April 1994 27 April 1994 28 April 1994 3 May 1994 4 May 1994 5 May 1994 10 May 1994 11 May 1994 12 May 1994  References Aissen, Judith L. 1992. Topic and Focus in Mayan. Language 68(1): 43-80. Allen, W. Sidney. 1964. Transitivity and Possession. Language 40(3): 337-343. Anderson, Stephen. 1976. On the Notion of Subject in Ergative Languages. Pp. 1-23 in Li, ed. (1976). Authier, J. -Marc. 1992. 23(2): 329-336.  Iterated CPs and Embedded Topicalization.  Linguistic Inguiiy  Baker, Mark C. 1991. On Some Subject/Object Non-Asymmetries in Mohawk. Language and Linguistic Theory 9: 537-576.  Natural  Banfield, Ann. 1973. Narrative Style and the Grammar of Direct and Indirect Speech. Foundations ofLanguage 10: 1-39. Browning, Marguerite A. 1987. Institute of Technology.  Null Operator Constructions.  PhD thesis, Massachusetts  Calabrese, Andrea. 1990. Some Informal Remarks on Focus and Logical Structures in Italian. Ms., Harvard University. Campana, Mark. 1992. A Movement Theory of Ergativity. PhD thesis, McGill University. Chomsky, Noam. 1977. On Wh-Movement. Pp. 71-132 in Formal Syntax. Peter W. Culicover, Thomas Wasow, and Adrian Akmajian, eds. New York: Academic Press. —.  —.  —.  —.  1981. Lectures on Government and Binding: The Pisa Lectures. Publications.  Dordrecht: Foris  1982. Some Concepts and Consequences of the Theory of Government and Binding. Cambridge: MIT Press. 1986. Barriers. Cambridge: MIT Press. 1992. A Minimalist Program for Linguistic Theory. Linguistics 1.  MIT Occasional Papers in  Chung, Sandra. 1978. Case Marking and Grammatical Relations in Polynesian. University of Texas Press. Cinque, Guglielmo. 1990. Types ofA’ -Dependencies. Cambridge: MIT Press.  117  Austin:  Comrie, Bernard. 1979. Degrees of Ergativity: Some Chukchee Evidence. Pp. 2 19-240 in Ergativity: Towards a Theory of Grammatical Relations. Frans Plank, ed. New York: Academic Press. —.  1981. Language Universals and Linguistic Typology: Syntax and Morphology. Chicago: University of Chicago Press.  Cook, Eung-Do and Donna B. Gerdts, eds. 1984. The Syntax of Native American Languages (Syntax and Semantics, vol. 16). Orlando: Academic Press. Craig, Colette Grinevald. Press.  1977.  The Structure of Jacaltec.  Austin: University of Texas  Creider, Chet. 1978. The Syntax of Relative Clauses in Inuktitut. Inuit Studies 2(1): 95-110. Davis, Henry. 1993a. Anti-Transitives in St’át’imcets (Lillooet Salish). Paper presented at the University of British Columbia, 19 November. —.  —,  1993b. A Configurational Pronominal Argument Language. Paper presented at WECOL 23. Seattle: University of Washington, 22-24 October. Dwight Gardiner, and Lisa Matthewson. 1993. A Comparative Look at WH-Questions in Northern Interior Salish. Pp. 79-95 in Papers for the 28th International Conference on Salish and Neighboring Languages. Seattle: University of Washington.  Davis, Philip W. and Ross Saunders. 1984. An Expression of Coreference in Bella Coola. Pp. 149-167 in Cook and Gerdts, eds. (1984). Dayley, Jon P. 1985. Voice in Tzutujil. Pp. 192-226 in Grammar Inside and Outside the Clause: Some Approaches to Theory from the Field. Johanna Nichols and Anthony C. Woodbury, eds. Cambridge: Cambridge University Press. Diesing, Molly. 1992. Indefinites. Cambridge: MIT Press. Dixon, R.M.W. 1979. Ergativity. Language 55(1): 59-138. Emonds, Joseph E. Publications.  1985.  A Unified Theory of Syntactic Categories.  Dordrecht: Foris  Erteschik-Shir, Nomi. 1993. The Dynamics of Focus Structure. Ms., Ben Gurion University of the Negev. Fiengo, Robert, C.-T. James Huang, Howard Lasnik, and Tanya Reinhart. 1988. The Syntax of Wh-in-Situ. Pp. 8 1-98 in Proceedings of the Seventh West Coast Conference on Formal Linguistics. Hagit Borer, ed. Stanford: Stanford Linguistics Association.  118  Gardiner, Dwight. 1993. Structural Asymmetries and Preverbal Positions in Shuswap. PhD thesis, Simon Fraser University. To be published by the University of British Columbia Press. —,  Lisa Matthewson, and Henry Davis. 1993. A Preliminary Report on Word Order in Northern Interior Salish. Pp. 139-157 in Papers for the 28th International Conference on Salish and Neighboring Languages. Seattle: University of Washington.  Gerdts, Donna B. 1988. Object and Absolutive in Halkomelem Salish. New York: Garland Publishing. Grimshaw, Jane. 1993. Minimal Projection, Heads and Optimality. Technical Report #4 of the Rutgers Center for Cognitive Science, Rutgers University. Heim, Irene. 1988. The Semantics of Definite and indefinite Noun Phrases. Garland Publishing.  New York:  Huang, Cheng-Teh James. 1982. Logical Relations in Chinese and the Theory of Grammar. PhD thesis, Massachusetts Institute of Technology. —.  1984. On the Distribution and Reference of Empty Pronouns. Linguistic Inquiry 15(4): 531-574.  Jelinek, Eloise. 1984. Empty Categories, Case, and Configurationality. Natural Language and Linguistic Theory 2: 39-76. Johns, Alana. 1992. Deriving Ergativity. Linguistic Inquiry 23(1): 57-87. Keenan, Edward L. 1976. Towards a Universal Definition of ‘Subject’. Pp. 303-333 in Li, ed. (1976). —  Noun Phrase Accessibility and Universal Grammar. 1977. and Bernard Comrie. Linguistic Inquiry 8(1): 63-99.  Kinkade, M. Dale. 1983. Salish Evidence Against the Universality of ‘Noun’ and ‘Verb’. Lingua 60(1): 25-39. —.  —.  1989. When Patients are Topics: Topic Maintenance in North American Indian Languages. Paper presented at the 24th International Conference on Salish and Neighboring Languages, Steilacoom, Washington, August 16-19. 1990. Sorting Out Third Persons in Salishan Discourse, American Linguistics 56(3): 34 1-360.  Koopman, Hilda, and Dominique Sportiche. 211-258.  1991.  119  international Journal of  The Position of Subjects. Lingua 85:  1991. Comparative Syntax of Subordination in Salish. Kroeber, Paul. University of Chicago.  PhD thesis,  Lasnik, Howard, and Juan Uriagereka. 1988. A Course in GB Syntax: Lectures on Binding and Empty Categories. Cambridge: MIT Press. Li, Charles N., ed. 1976. Subject and Topic. New York: Academic Press. Matthewson, Lisa. 1993a. Binding and Configurationality in St’át’imcets. Ms., University of British Columbia. —.  —.  —,  1993b. A Preliminary Investigation of z-Questions in Northern Interior Salish. Ms., University of British Columbia. In preparation. PhD thesis, University of British Columbia. Henry Davis, and Dwight Gardiner. 1993. Coreference in Northern Interior Salish. Pp. 217-232 in Papers for the 28th International Conference on Salish and Neighboring Languages. Seattle: University of Washington.  Mulder, Jean Gail. 1988. Ergativity in Coast Tsimshian (Sm’algyax). PhD thesis, University of California, Los Angeles. —.  1989. Syntactic Ergativity in Coast Tsimshian (Sm’algyax). Studies in Language 13(2): 405-435.  Murasugi, Kumiko G. 1992. Crossing and Nested Paths: NP Movement in Accusative and Ergative Languages. PhD thesis, Massachusetts Institute of Technology. Oehrle, Richard T. 1981. Common Problems in the Theory of Anaphora and the Theory of Discourse. Pp. 509-530 in Possibilities and Limitations of Pragmatics: Proceedings of the Conference on Pragmatics, Urbino, July 8-14, 1979. Herman Parret, Marina Sbisà, and Jef Verschueren, eds. Amsterdam: John Benjamins B.V. Partee, Barbara H. To appear. Topic, Focus and Quantification. Proceedings of the 1991 SALT Conference. Steve Moore and Adam Wyner, eds. Ithaca: Department of Modern Languages and Literatures, Cornell University. Payne, Thomas E. 1982. Role and Reference Related Subject Properties and Ergativity in Yup’ik Eskimo and Tagalog. Studies in Language 6(1): 75-106. Peters, Desmond, Sr., et al. [1992]. Introducing St’át’imcets (Fraser River Dialect): A Primer. Kamloops: Secwepemc Cultural Education Society. Prince, Alan, and Paul Smolensky. 1993. Optimality Theory: Constraint Interaction in Generative Grammar. Technical Report #2 of the Rutgers Center for Cognitive Science, Rutgers University. 120  Reinhart, Tanya. 1981. Pragmatics and Linguistics: An Analysis of Sentence Topics. Philosophica 27(1): 53-93. Roberts, Taylor. 1994. Relative Clauses in St’át’imcets (Lillooet Salish). Ms., University of British Columbia. Ross, John Robert. 1967. Constraints on Variables in Syntax. PhD thesis, Massachusetts Institute of Technology. Distributed by Indiana University Linguistics Club. Page references here are to the edition published as Infinite Syntax! (Norwood, NJ: Ablex Publishing, 1986). Schachter, Paul. 1973. Focus and Relativization. Language 49(1): 19-46. Smith, Jessie Ann. 1989. Widow Smith of Spence ‘s Bridge. As told to J. Meryl Campbell and Audrey Ward. Murphy Orlando Shewchuk, ed. Merritt: Sonotek Publishing. Smith, Lawrence R. 1984. On the Nonergativity and Intransitivity of Relative Clauses in Labrador Inuttut. Pp. 289-3 15 in Cook and Gerdts, eds. (1984). Thompson, Laurence C. 1979. Salishan and the Northwest. Pp. 692-765 in The Languages of Native America: Historical and Comparative Assessment. Lyle Campbell and Marianne Mithun, eds. Austin: University of Texas Press. —  and M. Terry Thompson. 1992. The Thompson Language (University of Montana Occasional Papers in Linguistics, no. 8). Missoula: University of Montana.  Thráinsson, Höskuldur. 1991. Long-Distance Reflexives and the Typology of NPs. Pp. 4975 in Long-Distance Anaphora. Jan Koster and Eric Reuland, eds. Cambridge: Cambridge University Press. Uechi, Aki. 1994. Focus and Quantification: A Case Study of Japanese Subjects. University of British Columbia.  Ms.,  van Eijk, Jan P. 1981. CuystwI Malh UcwalmIcwts: Teach Yourself Lillooet: UcwalmIcwts Curriculum for Advanced Learners. Mount Currie: Ts’zil Publishing House. —.  —.  —.  —  A Lillooet-English Dictionary. 1983. Development.  Ms.,  Mount Currie Cultural/Curriculum  1985. The Lillooet Language: Phonology, Morphology, Syntax. PhD thesis, Universiteit van Amsterdam. To be published by the University of British Columbia Press. 1987. Dictionary of the Lillooet Language. Ms., University of Victoria. and Thom Hess. 1986. Noun and Verb in Salish. Lingua 69(4): 319-331.  121  Watanabe, Akira. 1992. Larsonian CP Recursion, Factive Complements, and Selection. Pp. 523-537 in Proceedings of the North East Linguistic Society 23. University of Ottawa. Amy 3. Schafer, ed. Amherst: Dept. of Linguistics, University of Massachusetts. Woodbury, Anthony C. 1975. Ergativity of Grammatical Processes: A Study of Greenlandic Eskimo. MA thesis, University of Chicago.  122  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items