UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

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

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

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)byTAYLOR ROBERTSB.A., York University, 1992A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF ARTSinTHE FACULTY OF GRADUATE STUDIES(Department of Linguistics)We accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAJune 1994© Taylor Roberts, 1994In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(SignatDepartment of T.ingiiiticsThe University of British ColumbiaVancouver, CanadaDate 79 Time 1994DE-6 (2/88)AbstractThe goal of this thesis is twofold: first, to describe some of the symmetric andasymmetric behaviours of transitive and intransitive subjects in St’át’imcets, a NorthernInterior Salish language spoken in southwest mainland British Columbia; second, toconsider how the Principles and Parameters framework (Chomsky 1981; 1982; 1986;1992; etc.) can explain the asymmetries. Although many Salish languages are knownto display ergativity in their third-person subject inflection, the extent to which theselanguages are syntactically ergative is not well documented—perhaps because theiraccusativity has been more salient. The question has not been investigated forSt’át’imcets, and this thesis shows that there is at least one aspect of St’át’imcetssyntax—relativization—that appears to be ergative. Evidence of ergativity incoreference across conjuncts in St’át’imcets is not as clear, though; rather, coreferenceis restricted by a rule of one-nominal interpretation (Gerdts 1988) and a constraint onparallelism of discourse functions (Matthewson 1993a), both of which are shown in thisthesis to derive from more general constraints on discourse. Unifying the explanationsfor the various asymmetries is the idea—independently motivated and proposed toaccount for facts in other languages—that NPs that are topics are structurally higherthan NPs that are focused.Salish languages are often presented as though they were radically differentfrom other languages, but with respect to the complex and subtle data examined in thisthesis, St’ät’imcets resembles other known linguistic systems. Most of the thta arefrom original fieldwork, and they will be useful in the kind of comparative NorthernInterior Salish research begun by Davis et al. (1993), Gardiner et al. (1993), andMatthewson et al. (1993). Syntactic pivots have not been investigated in the other MSlanguages, and so establishing the ways in which Nla’kapmxcin (Thompson) andSecwepemctsfn (Shuswap) differ from St’át’imcets will ideally help to explain thenature of parametric variation in syntax.11Table of ContentsAbstract.iiTable of Contents iiiSymbols and Abbreviations vAcknowledgments viChapter 1 Introduction 11.1. Goals and outline of thesis 11.2. Grammatical relations and ergativity 31.3. Outline of St’át’imcets 51.3.1. Morphosyntax 51.3.2. Regional variation 131.4. Data and methodology 15Chapter 2 Relativization 172.1. Introduction 172.2. Pronominal inflection of relative clauses 212.2.1. Subject of transitive 212.2.2. Subject of intransitive 232.2.3. Direct object 272.2.4. Possessor 302.2.5. Objectof comparison 342.3. Summary 35Chapter 3 Conjunction 383.1. Introduction 383.2. Conjuncts with non-third-person 393.3. Conjuncts with third-person subject and object 423.3.1. Conjoined transitives 423.3.1.1. Parallelism constraint on discourse functions 423.3.1.2. Interaction with the one-nominal interpretation law 473.3.2. Intransitives 513.3.2.1. Intransitive in first conjunct 513.3.2.2. Intransitive in second conjunct 533.4. Summary 55111Chapter 4 Structural Asymmetries.574.1. Introduction 574.2. The one-nominal interpretation law as focus 624.3. Parallelism 664.3.1. Transitives 664.3.2. Intransitives 834.4. Extraction in relative clauses 874.5. Summary 102Chapter 5 Conclusion 104Appendix A: Morphology 106Appendix B: Key to Orthography 111Appendix C: Consultants 112Appendix D: Elicitation Sessions 115References 117ivSymbols and Abbreviations* ungrammatical oblig obligation, expectancy7 1. marginally grammatical; ONO one-nominal interpretation law2. grammaticality varies (23, 94, 120)(x) x is optional ooc out of control(*x) ungrammatical if x is present part particle: ‘well, but, so’*(x) ungrammatical if x is absent pass passive{xly} either x or y, but not both PC parallelism constraint on discourse- morpheme boundary functions (87, 128)1, 2, 3 first, second, third person p1 pluralA subject of transitive poss possessiveabs absolutive pred predicateacc accusative presupp presupposed knowledgeadhort adhortative prog progressiveanti antithesis quot quotativeappi applicative RC relative clausecomp complementizer recip reciprocalconj 1. conjunction; red redirective2. conjunctive inflection S 1. subject of intransitive;cons consequential 2. sentence constituentdef definite sg singulardeic deictic su subject (indicative)det determiner TO topical object marker (-tali)emp emphatic pronoun tr transitiveemph emphasiserg ergativeevid evidential ConsultantsF Fountain dialect AA Alice Adolphfoc focus BF Beverley Frankfut remote future, possibility DU Dorothy Ursaki (Nla’kapmxcmn)GF grammatical function GN Gertrude Nedhyp hypothetical LT Laura Thevargeincip incipient RW Rose Whitleymd indirectiveindic indicativeinterrog interrogativeintr intransitiveM Mount Currie dialectMS Northern Interior Salishnom nominalizernow demarcation of timeNP noun phraseO direct objectobj objectobl obliqueVAcknowledgmentsThe Lillooet Tribal Council gave permission to study St’át’imcets. Thanks are due to thefollowing consultants for their help—and for enduring my ghastly pronunciations andsentences—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 andgenerous to the linguists who invade his home every two weeks. Kukwstum ‘ckál ‘ap, Insnek ‘wntk ‘w7a.Thanks to my thesis supervisor, Hamida Demirdache, who has done the work of threepeople this year, and who has given freely of her time and ideas to me and to the otherstudents who desperately needed them. I’m lucky also to have had the following professorsvolunteer 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 generallyshared his expertise with Salish throughout the year. Lisa Matthewson also read most of thethesis and saved me from making several errors. Dorothy Ursaki provided someNla’kapmxcin data for chapter 2. My other professors at UBC included Laura Downing,Doug Pulleyblank, and Michael Rochemont. Myles Leitch and Doug Pulleyblank haveassisted me at various times in persuading an IBM-compatible computer to do linguisticspapers 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 CoastLanguages (SSHRCC grant 410-924629) supported fieldwork on St’át’imcets, and I’vebenefited phonologically and syntactically through collaboration with the project researchers:Susan Blake, Henry Davis, Lisa Matthewson, and Pat Shaw. Henry and Lisa first encouragedme to work on St’át’imcets, and the results presented here, such as they are, were madepossible 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 donemore than anybody to help me with the fieldwork—intellectually, emotionally, financially, andotherwise. This thesis would not have been possible without Henry, and without a heart-to-heart tsétstecwts with him several months ago.Some of my other friends and fellow students at UBC include Eleanor Blain, HelmiBraches, 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 VanessaValerga. I’m thankful for the courage that several of them had in daring to feed a vegetarianand 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 atYork University, and I’ll always be grateful for their help and enthusiasm, which continues tothis day.My largest debt is to Erwin, Andrea, and Brent (my family), who—though twothousand miles away for the Last two years—have remained my best friends. I love them.viChapter 1Introduction1.1. Goals and outline of thesisThe goal of this thesis is twofold: first, to describe some of the symmetric and asymmetricbehaviours of transitive and intransitive subjects in St’át’imcets, a Northern Interior Salishlanguage; second, to consider how the Principles and Parameters framework (Chomsky 1981;1982; 1986; 1992; etc.) can explain the asymmetries. Although many Salish languages areknown to display ergativity in their third-person subject inflection, the extent to which theselanguages are syntactically ergative is not well documented—perhaps because their accusativityhas been more salient. The question has not been investigated for St’át’imcets, and this thesisshows that there is at least one aspect of St’át’imcets syntax—relativization—that appears to beergative. 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 areshown in this thesis to derive from more general constraints on discourse. Salish languages areoften presented as though they were radically different from other languages, but with respectto the complex and subtle data examined in this thesis, St’át’imcets resembles other knownlinguistic systems.The results of this study have broader implications. Understanding the St’át’imcetspivots for relativization and coordination will shed light on discourse binding (Matthewson1993; in prep.) and St’át’imcets grammatical relations generally. Once the pivots for theselanguage-particular processes have been ascertained, they will help to explain, for example, towhat 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 beto antipassivize the predicate, deriving an S from an underlying A in order to satisfy theergative 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 determined1what syntactic requirements they are feeding; hypotheses concerning their functions may betested 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 beuseful 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 beeninvestigated 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 explainthe nature of parametric variation in syntax.The thesis is organized as follows. Chapter 1 is the introduction. Chapter 2 examinesrelativization 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 beneither ergative nor accusative; rather, it is most strongly constrained by a rule of one-nominalinterpretation and a requirement that arguments of transitive predicates share the samediscourse function across conjoined clauses. Chapters 2 and 3 are largely descriptive, and theconclusion 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 collectingthe 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 thegeneralizations and some of the data that illustrate them. The thesis is therefore organized insuch a way that it should be accessible to readers whose interests lie mainly in Salish languagedata, and also to readers interested primarily in syntactic theory. Readers of the latter categorywho skip directly to chapter 4 should bear in mind that the data presented there havenecessarily been idealized, and that the preceding chapters may be consulted if informationabout the variation across speakers and across elicitation sessions is desired. Finally, there arefour appendices that offer grammatical paradigms, a key to the orthography, biographies of thelanguage consultants, and a table of the elicitation sessions.The next subsection of this introductory chapter briefly introduces the problems thatlanguages displaying ergativity have posed for some theories of syntax, and the following2subsection outlines the morphosyntax of St’át’imcets. The chapter concludes with anexplanation of the data and methodology that were used in preparing the thesis.1.2. Grammatical relations and ergativityThe grammatical relations of transitive subject, intransitive subject, and direct object, are theprincipal concern of this thesis. Following Dixon (1979), they are henceforth abbreviated asA (transitive subject), S (intransitive subject), and 0 (direct object). The abbreviation A fortransitive subject is best regarded as mnemonic for ‘agent’, since transitive subjects aretypically agentive. The GFs frequently pattern together in systems that are called ‘accusative’and ‘ergative’; the relevant groupings of these relations are schematized below:(1) Accusative ErgativeA nominativeS S0Processes that do not systematically group A, 5, and 0 are called ‘neither accusative norergative’.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 methodof case-assignment. For example, English can be seen to have accusative morphology in itspronoun system (i/me, he/him, etc.) as well as in syntactic properties that group the A and Sroles into a single category ‘subject’. An example of the latter is the so-called that-traceeffect: embedded subjects (i.e., A and S) may not be questioned if the complementizer that ispresent (2a,b). This restriction does not hold of 0, however, as (2c) indicates. Note that allof the examples in (2) are grammatical if that is absent.(2) a. who1 do you think (*that) tj likes Fred? (A)b. who1 do you think (*that) tj is dying? (S)c. who1 do you think (that) Fred likes tj? (0)ergativeabsolutive0 accusativeA3The correlation between morphological case-marking (nominative/accusative) andgrammatical relations (subject/object) in languages like English has been taken to support theidea that nominative case is assigned from a unique case-assigner (tensed Infi) to a uniquestructural position (specifier of IP). Ergative languages pose a problem for this theory of case-assignment. Anderson (1976), for example, claims that subjecthood is best defined not bymorphological categories like case inflection and verb agreement, but by syntactic propertieslike control, raising, reflexive, and coordination. He shows that many languages that aremorphologically ergative are in fact syntactically accusative, since they treat {A,S} inopposition 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 inRelational Grammar, or structurally, as it is in the Principles and Parameters framework. Thefact that languages can have ergative morphology but accusative syntax challenges theories thatassume a structural correlation between case and grammatical functions. For some recentapproaches 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 thatcertain syntactic constructions like imperatives and jussives universally refer to a ‘deep subject’that comprises {A,S}, Dixon shows that other processes like coordination and subordinationmay have either accusative or ergative properties, as determined on a language-particularbasis. Languages like English that are thoroughly accusative happen to have their pivotsdefined on the same set as the ‘deep subject’ {A,S}, but other languages select an ergative{S,0} pivot for some aspects of their syntax. St’át’imcets may be a language of the lattercategory. 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 thatassumes a notion of subject that is defined on {A,S}, e.g., Keenan and Comrie (1977: 80 ff.).41.3. Outline of St’át’imcetsSt’át’imcets is spoken in southwest mainland British Columbia, in an area 160-300 kilometersnorth-northeast of Vancouver. A grammatical sketch of the language is given in § 1.3.1, andsome differences between the Mount Currie and Fountain dialects are noted in § MorphosyntaxThis section introduces some morphosyntactic characteristics of St’át’imcets that will berelevant for considering the kind of data to be presented in the thesis. Data are presented inthe orthography that is used in van Eijk (1981, 1983) and Peters et al. (1992), with theexception of additional hyphens (-), which indicate morpheme boundaries. A key to theorthography is in appendix B. Right-aligned below each form cited in the thesis are thespeakers’ initials and the token number(s) in the database, or other information identifying thesource (see § 1.4). Appendix A provides pronominal paradigms and some other grammaticalinformation, 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 objectarguments are marked by obligatory pronominal affixes on the predicate (Davis 1993b: §2).Overt NPs are optional, as the following example shows:(3) tsiin-tsi-lhkantell-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 rigidlypredicate-initial; specifically, VOS order (verb-object-subject) is preferred in elicitation bymany 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) tsuw’-n-ás ti sqáycw-a ti k’ét’h-akick-tr-3sg. conj det man-det det rock-det‘the rock kicked the man’* ‘the man kicked the rock’(AA 2223; LT 2324)Some speakers are not as strict as others in their post-predicate word order, although they stillprefer VOS over VSO order:(5) áts’x-en-as ta sqáycw-a ta smülhats-asee-tr-3erg det man-det det woman-det‘the woman saw the man’ (preferred)‘the man saw the woman’ (also possible)(RW 48; Matthewson 1993a: 2-3)For another example of this variation in basic word order between speakers, see (111) inchapter 3 (p. 54).VSO order may be obtained if the predicate has restrictions concerning the animacy ofits arguments. Consider the following example:(6) (tsicw) áts’x-en-as ti syáqts7-a i tsftcw-a1go see-tr-3erg det woman-det p1. det house-det‘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 thissentence 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 followingexample illustrates:(7) *ats’x.en4tas i ucwalmIcw-a ti syáqts7-asee-tr-3p1. su p1. det people-det det woman-det‘the people saw the woman’(AA 2230)1 The initial auxiliary tsicw ‘go’ was given only by LT, not by AA.6Nominals may not precede the predicate unless introduced by the focus particle nilh, asshown in (89).2 The asterisk (*) outside of the parentheses surrounding nil/i indicates that theentire sentence is ungrammatical if nil/i is absent:(8) *(nilh) ti syáqts7-a ts’um’-qs-án’-as ti sqáycw-afoc det woman-det lick-nose-tr-3erg det man-det‘it was the woman that kissed the man’(AA 2217)(9) *(nilh) ti sqáycw-a qwatsátsfoc det man-det leave‘the man left’(AA, LT 2214)Emphatic pronouns, however, may behave predicatively and appear sentence-initially withoutthe 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) hastaken it as evidence against a noun/verb distinction; see van Eijk and Hess (1986) for adifferent perspective.Nominals usually appear with what may be regarded as a discontinuous determiner, inthat the first element of an NP is preceded by an article that encodes such categories assingular and plural, and followed by an enclitic -a. Examples of both singular ta and plural iappear above in (6). The enclitic -a is dropped if the progressive auxiliary wa7 intervenesbetween the article and the nominal; for examples, compare sentences (52-53) to (51) inchapter 2 (p. 26), and see also the footnote in chapter 3 giving AA’s version of sentence (86)(p. 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 andFountain dialects respectively, but they seem simply to be in free variation for the consultantsof 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 possibleexplanation for this innovation is that these speakers have been in extensive contact withspeakers of SecwepemctsIn (Davis 1993b: §3.1 n. 5), a language that allows nominals toprecede the predicate (Gardiner 1993; Gardiner et al. 1993).7This is exemplified below for two speakers, who volunteered these sentences as translations forthe English glosses indicated:(10) áts’x-en-lhkan i tsuw’-n-ás ta maw-a ti smülhats-asee-tr-lsg.su when.past kick-tr-3sg.conj det cat-det det woman-det‘I saw when the woman kicked the cat’(RW 1629)(11) áts’x-en-lhkan ti smülhats-a tsuw’-n-ás ta maw-asee-tr-lsg. su det woman-det kick-tr-sg. conj det cat-det‘I saw the woman kick the cat’(GN 1630)The vowel of the singular determiner can also be rounded if it precedes a word having aninitial labial consonant; for an example, see (66) in chapter 2 (p. 32). Also see § 1.3.2 belowfor more information about the differences between Mount Currie and Fountain speech.The indicative paradigm (called ‘plain’ by Kroeber 1991), from which the main clauseperson markers are drawn, is said to be morphologically ‘split-ergative’. The 3sg subject of atransitive predicate in a main clause is marked on the predicate by the ergative suffix -as, asshown in (12). Direct objects and subjects of intransitive predicates, however, do not induceovert agreement on the predicate, as indicated in (12) and (13) by absolutive -Ø (examples (12-13, 15-17) are adapted from van Eijk 1985: 172, 174):(12) tsün-O-astell (trans)-3sg. abs-3sg. erg‘she told him’(LT 2425)(13) tsut-Osay (intr)-3sg. abs‘she said’(LT 2426)The null symbol -Ø is omitted from cited examples, as in (4-9) above, unless clarity requiresit. In addition to displaying ergativity in its morphology, St’át’imcets appears to displaysyntactic ergativity in relativization, as chapter 2 will show.8In main clauses, non-third-person subject markers usually cliticize to the first prepredicate 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 otherwisehost the subject marker -lhkan if no auxiliary were present:(14) hüy’ lhkan ka7lh tsukwincip 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 andintransitive subjects are inflected alike, in opposition to direct objects:(15) tsun-ts-ká1’ aptell-lsg. acc-2p1. nominative‘you guys told me’(LT 2427)(16) tsüt-kal’apsay-2p1. nominative‘you guys said’(LT 2428)(17) tsüt-kansay-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 fromsubjects of intransitive; the 1 sg direct object in (15) is indicated by the affix -ts, while the 1 sgintransitive subject in (17) is marked by -kan.Given that NPs do not show overt case, when a single overt nominal appears in asentence with a transitive predicate that bears third-person subject and object affixes, it mightbe 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 asthe subject, as the glosses for the following sentences indicate:9(18) wa7 k’al’em-mfn-as ta {smülhats-a/syáqts7-a}prog wait-appl-3erg det woman-det‘he is waiting for the woman’*‘the woman is waiting for him’(AA, GN 1312; LT 2329)(19) (wa7) qvl-mmn-as ti syáqts7-aprog bad-appl-3erg det woman-det‘he doesn’t like the woman’*‘the woman doesn’t like him’(AA 2233; cf. ON 1313; LT 2331)In order for a sentence to be interpreted as having an overt NP subject and a null 3sgpronominal object, the sentence is passivized, as shown in (20-2 1). Another example is thepassive in (22), which was volunteered as a form having the overt NP as subject, and itcorresponds to the non-passive in (18) above.(20) áts’x-en-as ta sqáycw-asee-tr-3erg det man-det‘he saw the man’(AA, BF, ON, LT, RW 29)(21) áts’x-en-em 1 ta sqáycw-asee-tr-pass obi det man-det‘he was seen by the man’(BF, ON, RW 28)(22) nilh t’u7 s-e-s k’al’-em-mfn-em ti syáqts7-afoc 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 hascome 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 interpretedas the absolutive.10The rule of one-nominal interpretation will be seen in chapter 3 (pp. 47 ff.) to be a fairlystrong constraint in St’át’imcets. Its effects can also be observed in the two other MSlanguages, 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 Kroeber1991) are similar in form to the plain clitics, except that they lack the -(lh)k- indicativemarker. Conjunctive inflection is used mainly for adverbial clauses and interrogativecomplements, as well as for transitive complements of negation (cw7aoz ‘not’), and in othernominalized environments. Adverbial clauses are introduced by the complementizers lh‘hypothetical’ or i ‘when.past’. Examples follow:(24) saw-en-tsál-itas [ lh swan’ulh-ás ni7 qmut ]ask-tr-lsg. obj-3pl. subj hyp whose-3sg. conj deic hat‘they asked me whose hat that was’(AA, LT 2299; van Eijk 1985: 272)(25) láni7 [ i t’fq-as ]then when.past come-3sg. conj‘it is then that he came’(LT 2431; van Eijk 1985: 272)Relative clauses usually receive conjunctive inflection, although some speakers will acceptplain (main clause) inflection in them; see sentence (49) in chapter 2 (p. 25).Finally, factual inflection (called ‘nominalized’ by Kroeber 1991) appears in non-initialconjuncts 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 possessiveaffixes, and transitive subjects are marked by conjunctive affixes. Nominalization also appearsin complement clauses that are introduced by the complementizer kw, which is glosseddet(’erminer) because of its formal similarity to the indefinite determiners. This general use ofthe indefinite determiner as a complementizer in St’át’imcets resembles the Coast Salishpattern, and differs from Nla’kapmxcIn and Secwepemctsfn, which choose definite and11indefinite 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 lhkan zewát-en [kw s-t’iq-s ]prog lsg.su know-tr det 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). SeeKroeber (1991: 165) for a similar chart (but without introductory particles) for the Salishfamily.(27) Inflection of clause types in St’át’imcetsClause type Inflection [ Introductory particlemain plain noneadverbial, interrogative, conjunctive lii ‘ hypothetical’negation 1 ‘when.past’cw7aoz ‘not’complement; nominalized kw;non-initial conjunct nilh ‘then; so, then’The facts outlined above represent the core of St’át’imcets grammar. Naturally, thereis 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 determinerssometimes appear to trigger metathesis in the roots to which they attach. There is also analternation between -lhk/-k as the indicative marker for non-third-person subjects, which forbrevity is usually not glossed as a separate morpheme, but is treated as part of the rest of thesubject affix. This morpheme appears as -lhk usually after vowels and resonants, and as -kelsewhere; 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 beconsulted for more information if desired.121.3.2. Regional variationTwo dialects of St’át’imcets are recognized by van Eijk (1985): the Fountain dialect, which isassociated 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 (MountCurrie, near Pemberton). The two groups are separated by the Lillooet mountain range, butare connected by about 100 kilometers of waterway and, today, by road. The dialects aremutually intelligible, the primary differences being a few lexical items, most of which arerecognized by the consultants of this thesis. Such variation is exemplified by the wordssmülhats (F) and syáqtsa7 (M) ‘woman’, as in (18) above. There is also a phonologicaldifference 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 concerningthe 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 (andother sentences starting at p. 43).Apparently there is a syntactic difference between the Fountain and Mount Curriedialects. Although certain pronominal markers are second-position clitics, as noted on page 9with respect to (14), Mount Currie speakers allow the person marker to appear sentence-initially:(28) kan xát’-min’ ku kál’watlsg. 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. The structure underlying(28) would therefore be:13(29) wa7 lhkan xát’-min’ ku kál’watprog lsg.su want-appi det medicine‘I want some medicine’(LT 2433)This analysis may not be correct, however, as one example of a fronted subject marker cooccurring with progressive wa7 has been volunteered:(30) kan t’u7 wa7 s-tálh-leclsg. su part prog nom-upright-body‘I am standing’(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 beinginnovations.Another area of variation that deserves further study is the determiner system. Thesingular ti/ta alternation has been claimed by van Eijk (1985: 223)—and challenged above onpage 8—to characterize the Mount Currie and Fountain dialects respectively. Similarly,proper nouns are almost invariably preceded by the nominalizer s- or the determinernominalizer 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 determinerkw alone, often unrounding the latter to k or eliding it completely—although the unroundedsequence 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 itlooks indeed more like an isogloss, since the full kw-s sequence before proper nouns andelsewhere is generally ungrammatical for Mount Currie speakers:(31) á7ma k(*ws) Mary ihel s-Janecute det-nom Mary than nom-Jane‘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-assee-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 syntacticphenomenon, rather than one of phonological reduction, but the matter needs furtherinvestigation before conclusions can be drawn.3The regional differences outlined above are the only ones that have been obvious, andwhere these differences exist in the data cited, they are mentioned in footnotes.1.4. Data and methodologyData for this thesis were collected from native St’át’imcets speakers during the periodNovember 1993 to May 1994, under the auspices of the Project on Lexical Interfaces withPhonology and Syntax in North West Coast Languages, as part of an ongoing syntacticdatabase of St’át’imcets that was begun by Henry Davis and Lisa Matthewson in October1992. Most of the original data for this thesis are from Gertrude Ned and Rose Whitley, andthus—despite minor variation between these speakers—can be taken to represent the Fountaindialect. Most (if not all) of the original sentences cited here have been checked and recheckedduring sessions subsequent to their initial collection, and the crucial data have additionallybeen confirmed in independent sessions with Alice Adolph and Laura Thevarge, speakers ofthe Mount Currie dialect. (See appendix C for biographies and genealogies of the languageconsultants.) The original elicitation of a sentence and its subsequent confirmations wereusually recorded in a single database record, and so several speakers are often associated witha single token number. Sentences confirmed with the Mount Currie consultants mayoccasionally differ with respect to lexical items like smálhats/sydqtsa7 ‘woman’, but thesedifferences are usually not noted. To get a clearer picture of which consultants originallysupplied a particular sentence, consult appendix D for a table of the elicitation sessions, whichFountain speaker RW characterizes sentences like these as typical of Mount Curriespeech (Henry Davis, p.c.).15indicates which consultants were present at individual sessions. Note also that differentsentences occasionally have the same token number because variants of a basic sentence havesometimes been recorded in the same database record.Only data as are sufficient to illustrate generalizations are given in the thesis, but otherdata that confirm these generalizations are in the aforenamed database. Some idealization ofthe data has been necessary in chapter 3 (where subtle grammaticality judgments are notalways consistent across speakers and across elicitation sessions), but it is always statedexplicitly. If it seems like undue attention is given to explicating the varying grammaticalityjudgments 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 grammaticalityjudgments of the consultants, and where differences among speakers and sessions exist, theseare noted in the text and footnotes.Textual data are not cited because texts cannot furnish two kinds of information that arenecessary to the thesis: first, whether a sentence has more than a single interpretation; andsecond, whether a sentence would be ungrammatical if it had a slightly different form than theattested 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 iscrucial for being able to describe a language’s syntactic restrictions, and thus for trying tounderstand what constitutes a speaker’s knowledge of language.16Chapter 2Relativization2.1. IntroductionThis chapter examines the structure of non-oblique-centered relative clauses (RC5) inSt’át’imcets, filling in some of the gaps of previous descriptions.4 Transitive and intransitivesubjects will be seen not to pattern alike with respect to relativization, thus presenting aproblem for a theory of relativization that assumes a unified notion of ‘subject’, e.g., Keenanand Comrie (1977: 80 ff.). Additionally, data are presented that clearly show that absolutivecentered relative clauses contain a gap corresponding to the relativized constituent—not covert3sg 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 arelative pronoun as Indo-European languages do, relativization nevertheless involves extractionof the relativized constituent. A fuller analysis of the structure of RCs remains to be done notonly for St’át’imcets, but for Salish languages generally. Kroeber (1991) gives an excellentoverview of the variety of forms of RCs in Salish, although he does not have as much data onSt’át’imcets as on the other Northern Interior languages, Nla’kapmxcmn (Thompson) andSecwepemctsIn (Shuswap). The data presented in this chapter resolve all of the questionsraised by Kroeber (1991) and Matthewson (1993b) pertaining to pronominal inflection insideRCs.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 individuallanguages differs to an extent such that a purely syntactic definition of RC is difficult tomaintain universally. The following definition, from Comrie (1981: 136), will suffice for thischapter:Parts of this chapter are from Roberts (1994); thanks to Paul Kroeber for his detailedcomments 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: 224-226), 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 arenon-restrictives [= appositives]. . . . A relative clause then consists necessarily of ahead and a restricting clause. The head in itself has a certain potential range ofreferents, but the restricting clause restricts this set by giving a proposition that must betrue 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 thehead, while who Fred hired is the restricting clause. Note that the restricting clause has a gapwhere 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 thegrammatical relation of object within the restricting clause (a property of the internal syntax ofthe 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 RCrefers 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 internalsyntax 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 examplecorresponding to the English sentence above is Sally met who (ever) Fred hired. St’át’imcetshas headless, post-head, and pre-head RCs, although the status of the latter construction as atrue RC remains unclear. Nothing is said about them here, though it would be worthwhile toexamine these structures in light of the findings that are presented in this chapter concerningthe other two RC types.Kroeber (1991: 176) notes that appositives [= non-restrictive RCs] are not widelyattested in Salish, and Comrie (1981: 132) states more generally that the distinction betweenrestrictive and non-restrictive RCs is found only sporadically across the world’s languages.Likewise, attempts to elicit appositives in St’át’imcets have not met with succeSs.5 Onestraightforward test for the existence of appositives is to determine whether an RC may modifya proper noun. Unlike a restrictive RC, whose restricting clause serves to narrow thereference specified by the head, an appositive merely supplies additional information about aThanks to Henry Davis for checking these data.18fully-identified head. Proper nouns are fully referential, hence they may normally head onlyappositives—as in the English sentence John, who was only hired this morning, was fired thisafternoon. Here, there are not several persons in the universe of discourse6 named John, withthe RC serving to narrow the reference of the main clause to a single individual. Rather, thereis a single person named John in the universe of discourse, and the RC merely suppliesadditional 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 exampleshows:(33) *pzán4hkan [ s-John [ta ats’x-en-ácw-a i-nátcw-as 1]meet-tr-lsg. subj nom-John det see-tr-2sg. conj-det when.past-day-3sg. conj‘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 inthe following example:(34) [ áts’x-en-lhkacw s-John i-nátcw-as I müta7 [pz-án-lhkan lhkünsa]see-tr-2sg. su nom-J. when.past-day-3sg. conj conj meet-tr-lsg. subj now‘I met John, who you saw yesterday’(BF 828)Because a structure like (33) cannot be used for appositives, it can be safely assumed that all ofthe 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 HierarchySubject > Direct Object > Indirect Object > Oblique > Genitive > Object ofComparison6 Calabrese (1990: 12) defines the universe of discourse for an utterance U as ‘the setof referents, properties and presuppositions which the speaker believes the hearer presupposesin the time instant tin which U is uttered.’19If a language allows relativization of more than one of these roles, the Accessibility Hierarchypredicts that they will fall contiguously on the scale. For example, if a language can relativizeobliques (by which Keenan and Comrie mean arguments, not adverbials), then all of the higherGFs 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 strategiesof 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 employdifferent 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 thescale 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, sincethey are harder to relativize (Comrie 1981: 156).The Accessibility Hierarchy is a good place to begin the examination of St’át’imcetsRCs, since Salish languages can challenge and test it in several respects. Relativization inSt’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 andergative languages, and who argue that the notion of ‘subject’ is valid even in an ergativesystem, where the single argument of an intransitive predicate patterns with the object of atransitive predicate—rather than with the agent, as it does in an accusative language. Recallthe chart in (1) in chapter 1 (p. 3).Salish lacks a distinct ‘indirect object’ role, and so this member of the AccessibilityHierarchy may be disregarded (Kroeber 1991: 232). Kroeber (1991: 233) argues moreoverthat genitive (possessor) must be higher on the hierarchy than Keenan and Comrie suggest, andalso that the ‘oblique’ category needs to be subdivided into oblique object, instrument, andlocative. His revised scale for Salish therefore has the following form:(36) Accessibility Hierarchy for SalishSubject > Object > Possessor > Oblique Object > Instrument > Locative20Kroeber (1991: 232) does not discuss objects of comparison, due to lack of data. Little will besaid of them here, either, beyond showing that—regardless of whether they are a distinctgrammatical role in Salish (a separate, as yet unaddressed question)—they are difficult if notimpossible to relativize. See (73-75) below (p. 34).The only parameter of form that differs among RC types in St’át’imcets is the presenceor 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 oftransitive, subject of intransitive, direct objects, and possessors.2.2. Pronominal inflection of relative clauses2.2.1. Subject of transitiveIn 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 [ ta pzántsih*(as)a 1]see-trans-lsg. su det man-det det meet-trans-2sg. obj-3sg. conj-det‘I saw the man who met you’(AA, GN, LT 1328)If the object agreement were omitted from (37), it would be interpreted as 3sg, since thisperson is not overtly marked. In such structures where the object is 3sg and the subject isextracted, it is preferred for some speakers (notably RW) to use either a passive constructionor the topical object marker -tall (which is in complementary distribution with the 3sg -assubject suffix, and which only occurs in extraction contexts—see (91-93) on page 46 in chapter3). The latter strategy is exemplified below in (38). Sentences (39-40) show that it applieswith equal force to wh-extraction, and it is preferred especially when coreference between thewh-word and a possessive pronominal—indicated in the following glosses by matchingsubscript indices—is intended (Matthewson 1993a: 3, 19):21(38) áts’x-en-as [ ti sqáycw-a [ ti s-Bill ]]see-tr-3erg det man-det det hit-tr- (TO/3sg. conj]-det nom-Bill‘he saw the man that hit Bill’(GN 141; GN, RW 218; LT 2581)(39) swat ku ats’x-en-táli i skicez-f-hawho det see-tr-TO pl.det mother-3p1.poss-det‘who1 saw their1 mother?’(BF, GN, RW 41)(40) *swat 1w áts’x-en-as i skicez-I-hawho det see-tr-3sg. conj p1. det nwther-3p1.poss-det‘who1 saw their1 mother?’(BF, GN, RW 41)Interestingly, neither -tall nor passive may mark a relativized subject when the object is non-third-person, as in the following examples corresponding to (37) above:(41) *áts’xenlhkan [ ta sqáycw-a [ ta pz-an-tsi-táli-ha 1]see-trans-lsg.su det man-det det meet-trans-2sg.obj-TO-det‘I saw the man who met you’(AA, GN, LT 1329)(42) *âts’x.en4hkan [ ta sqáycw-a [ta pz-.n-tsi-m-a ]]see-trans-lsg. su det man-det det meet-trans-2sg. obj-pass-det‘I saw the man who met you’(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 aGF-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), therewill not otherwise be any indication of which argument has been focused.7 This seemsplausible, except that it does not explain why subject morphology is obligatorily retained in anRC like (37). If the subject morphology were gapped, for example (*ta pzántsiha), thesentence should still be unambiguously interpretable as ‘the one who met you’, since therewould be an overt 3sg nominal to serve as the subject (the head of the RC, ta sqáycwa ‘theCraig (1977: ch. 7) gives the same analysis of a similar effect in Jacaltec.22man’). 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 thegrammar 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 intransitiveIn all Salish languages except those of the Southern Interior, person morphology correspondingto a relativized subject of intransitive is absent (Kroeber 1991: 235), but because 3sgabsolutive is regularly marked by -0, the only way to determine whether the morphology isindeed absent in an intransitive-centered RC is to cleft a non-third-person nominal. Assumingthat the residue of a cleft is an RC (Kroeber 1991: 184-187), the morphology corresponding tonon-third-person targets in such constructions is indeed absent. Compare (43) and (44) in thisregard, where the bracketed constituent in (44)—the residue of the cleft—is argued to be aheadless RC:(43) tsIcw-kango-lsg. subj‘I went’(LT 2438; Kroeber 1991: 262, citing van Eijk 1985: 279)(44) tsukw t’ u7 s7ents [ ti tsIcw-aonly part lsg. emph det go-det‘Only I went’(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 subjectmorphology 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 snüwa ti ãts’x-en-ts-acw-afocus 2sg. emph det see-tr-lsg. obj-2sg. conj-det‘it was you who saw m&(ON 116, 1320, 1330, 1331; AA, ON, RW 1403;AA 2226, 2227; LT 2333)Rather, there is obligatory 3sg agreement on the predicate:(46) nilh snüwa ti áts’xents*(as)aSfocus 2sg. emph det see-tr-lsg. obj-3sg. conj-det‘it was you who saw me’(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. *nilh snülap [ i ats’x-en-ts-ál’ap-a ]focus 2pl. emph p1. det see-tr-lsg. obj-2pl. conj-det‘it was you guys who saw me’(AA, GN, LT, RW 1182)b. ?nilh snülap [ i ats’x-en-tsal-itás-a ]focus 2pl. emph p1. det see-tr-lsg. obj-3pl. conj-det‘it was you guys who saw me’(AA, ON, LT, RW 1183)c. nilh snülap [ i ats’x-en-ts-ás-a ]9focus 2pl. emph p1. det see-tr-lsg. obj-3sg. conj-det‘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) beingpreferred to plural agreement (b), despite the fact that the focused pronoun is plural. Another8 The emphatic pronoun may appear sentence-initially without the focus marker nilh.In this respect, emphatic pronouns differ from other nominals, which cannot appear prepredicatively without a preceding nilh (as noted in § 1.3. 1).See the previous footnote.24reason that singular agreement is preferred here is that plural morphology is in complementarydistribution with overt nominals and plural determiners (van Eijk 1985: 277-278).Under a 3sg-agreement analysis, then, sentence (44) would be more fully representedas follows:(48) tsukw t’u7 s7ents [ ti tsicw-ø-a jonly part lsg. emph det go-3sg. intr. su-det‘Only I went’A possible argument against the 3sg-agreement analysis, though, is that the residue of cleftsgenerally receive conjunctive, not indicative, inflection, and so absence of a 3sg pronominal inan RC would have to indicate a real gap, since conjunctive inflection is overt for 3sg (examinethe 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, asthe following example shows:(49) nilh ta sqáycw-a ta ats’x-en-(lhk)-án-a cillelfocus det man-det det see-tr-indic-lsg.su-det run‘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 theirS-centered RCs do contain a gap, it would be desirable to be able to show that the otherspeakers have a gap in these structures as well. Another argument for the existence of a gap inS-centered RCs is needed.The correctness of the gap analysis is suggested by the fact that the intransitive-centeredRC (44) is indeed ungrammatical with conjunctive inflection, regardless of whether there islsg or 3sg agreement:(50) a. *tsu t’u7 s7ents [ ti tsfcw-an-a ]only part lsg. emph det go-lsg. conj-det‘I am the only one who went’(AA, ON, LT, RW 1200)25b. t’u7 s7ents [ ti tsfcw-as-a ]only part lsg. emph det go-3sg. conj-det‘I am the only one who went’(AA, GN, LT, RW 1201)The fact that even the default 3sg agreement illustrated in (46) and (47) for subject oftransitives is not possible for intransitive subjects shows that relativization does not refer to asingle ‘subject’ role that subsumes both transitive and intransitive subjects.Further evidence against the 3sg-agreement analysis in (48) is the behaviour of pluralintransitive subjects when relativized, since plural morphology is overt in both the plain andconjunctive paradigms. Because plural morphology is in complementary distribution withovert nominals and plural determiners, as noted above, this can only be checked with headlessRCs. The following examples take this form, and they confirm that the morphologycorresponding to the relativized intransitive subject is obligatorily absent:(51) áma-s-kan [ i qwatsats(*as)a ]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) áma-s-kan [i wa7(*as) güy’t ]°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) áma-s-kan [i wa7(*as) q’flhil ]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 thepredicate is unaccusative, as in (51), or unergative, as in (52-53). 11 The fact that all of thesesentences are ungrammatical with 3sg agreement constitutes strong evidence that the10 Pursuant to the discussion in § 1.3.1, the enclitic half of the discontinuous determineris absent here and in the following sentence because progressive wa7 is present.gay ‘t ‘sleep’ patterns with unergatives according to Davis (1993a).26grammatical variants of these sentences—and S-centered RCs generally—contain a gapcorresponding to the relativized constituent.2.2.3. Dfrect objectRelativization of direct objects mirrors relativization of subject of intransitive, in that themorphology corresponding to the target is absent:(54) (nilh) snüwa ti ats’xen(*tsfn)ánafocus 2sg. emph det see-tr-2sg. obj-lsg. conj-det‘it was you that I saw’(ON 119, 120; AA, ON, LT, RW 1405, 1406)Both Kroeber (1991: 259-263) and Matthewson (1993b: 16-17) state that object morphology inobject-centered RCs may be freely retained or omitted, but this is not certain. ON apparentlyaccepted overt object morphology at one point (token 120), but ON and RW together rejectedthis variant of (54) during its most recent elicitation; AA also independently rejects it. Ifobject morphology in these structures is indeed obligatorily absent, a possible analysis,suggested by Matthewson (1993b: 17), is that the missing object morphology does notrepresent 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 ats’x-en-O-án-afocus 2sg. emph det see-tr-3sg. obj-lsg. conj-det‘it was you that I saw’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-centeredRCs.A test for a gap is suggested by van Eijk’s (1985: 278-279) statement that conjoinedproper noun complements generally require a plural affix on the predicate. This is exemplifiedbelow:27(56) wa7 k’wzüsem{wit/*ø} wi s-John müta7 s-Bill12prog work-intr- (3pl. su/3sg. intr) p1. det nom-John conj nom-Bill‘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 thesentences in (46-47) above. Because third-person object agreement is overt when the object isplural and the subject is non-third person (the 3p1 object affix is variously -tani and -wit—seethe object suffix paradigm in appendix A, p. 106), it is possible to distinguish a gap fromobject 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 s-John mita7 s-Billsee-tr-{3pl. obj/3p1. obj/3sg. obj]-lsg. su p1. det nom-John conj nom-Bill‘I saw John and Bill’(LT 2335, 2336, 2337)When the object is clefted, however, as in (58), the pattern of agreement is exactly opposite:(58) nilh wi s-John müta7 s-Bill i lhkünsafoc pl.det nom-J conj nom-B pl.det see-tr-{-Ø/3p1.obj]-lsg.conj-det now‘it was John and Bill that I saw’(LT 2338, 2339)The fact that the 3p1 agreement morphology that was obligatory in (57) must be absent from itscorresponding cleft in (58) establishes that there is a gap in the RC residue corresponding tothe clefted object. Note that this test requires a cleft construction and an RC residue; anordinary 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 areconjoined as a single object, since they require plural agreement. In (59) below, a 2sg12 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 has3sg and 3pl agreement in free variation. Additional comments on the differences between LTand AA with respect to person/number agreement follow in the text.28emphatic pronoun is conjoined with a proper noun. Unlike the clefting of a single 2sgemphatic pronoun as the subject of a transitive, as in (45-46), there cannot be 3sg agreement.Rather, agreement is obligatorily second-person:’3(59) áts snüwa müta7 s-Marysee-tr- {2sg. obj/2pl. obj/3sg. obj]-lsg. conj 2sg. emp conj nom-Mary‘I saw you and Mary’(LT 2340, 2341)When the conjoined object in (59) is clefted, however, the RC residue contains an obligatorygap corresponding to the object:(60) (nilh) snüwa müta7 s-Mary {ti/i}foc 2sg. emp conj nom-Mary det/pi. det see-tr- (G/2pl. obj/2sg. obj] -lsg. conj-det‘it was you and Mary that I saw’(AA 2260, 2261, 2264; LT 2342, 2343)This comprises additional proof that object-centered RCs have a gap corresponding to thetarget, not covert 3sg agreement, contra Kroeber’s (1991: 235) claim that object pronominalsare ‘probably never obligatorily deleted in object-centered RCs.’The tests above do not work for all speakers. AA, for example, requires 3sgagreement 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/absolutiveasymmetry, requiring person/number agreement for conjoined absolutive arguments, but freelyalternating between person/number agreement and default 3sg agreement for conjoinedergative arguments. Although AA and LT are both Mount Currie speakers, the facts aboveconfirm that speakers of the same region may nevertheless vary with respect to subtlephenomena like agreement with compounded arguments. It is important to note that this13 LT has alternately preferred 2sg and 2pl agreement. In checking similar data withthe Fountain speakers, GN stated that the number varies depending upon the number of theaddressee, i.e., whether the referent of the proper noun is present as an addressee; if so, 2plagreement is preferred.14 The singular and plural determiners ti and i have been in free variation for both AAand 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 theprevious footnote.29represents true idiolectal variation since each speaker is systematic in her own agreementpattern, 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 (45-47).Fountain speakers GN and RW are yet different with respect to agreement withconjoined arguments. All conjoined proper noun complements—whether as ergatives orabsolutives—may freely have either 3sg or 3pl on the predicate. Conjoined argumentscontaining an emphatic pronoun are more complicated: there appears to be obligatory personagreement with the emphatic pronoun only with the transitive subjects and direct objects; theseconjoined 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 beenelicited once, and need to be confirmed individually with the consultants.The gap found in object-centered RCs for several speakers therefore parallels the gapfound in intransitive-subject-centered RCs. Stated more generally, absolutive-centered RCs inSt’át’imcets contain a gap corresponding to the target.2.2.4. PossessorWord order of possessors is very restricted. The sentence in (61) shows that while a possessorand a head may be preposed together in a focus construction, (62) reveals that neither elementmay be brought into this position alone, since it creates a discontinuous constituent:(61) nilh [ ta sqáxa7-s-a s-Mary ] ta ciilel-afoc det dog-3sg.poss-det nom-Mary det run. away-det‘it was Mary’s dog that ran away’(AA, LT 2304; Gardiner et al. 1993: 144)(62) a. *njlh ta sqáxa7-s-a ta ctilel-a s-Maryfoc det dog-3sg.poss-det det run. away-det nom-Mary(AA, LT 2305; Gardiner et al. 1993: 144)30b. *nilh s-Mary ta cülel-a ta sqáxa7-s-afoc nom-Mary det run. away-det det dog-3sg.poss-det(AA, LT 2306; Gardiner et al. 1993: 144)Similarly, a wh-possessor may not be extracted. The possessor and head may be preposedonly as a constituent:(63) a. [ swat skicza7 ] qwatsátswho mother leave‘whose mother left?’(AA, GN, LT, RW 425; Gardiner et al. 1993: 145)b. *swat ku qwatsáts ku skIcza7-swho det leave det 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 NISlanguages, Nla’kapmxcIn (Thompson) and SecwepemctsIn (Shuswap), which do permitpossessor extraction in certain contexts (Gardiner et al. 1993: 139-145).However, there is evidence for possessor extraction of some kind in St’át’imcets, sinceit is possible to relativize a possessor with unaccusative and unergative predicates; examplesare given in (65-67) below. For the sake of comparison, (64) is a non-relativized structurecorresponding to the relativized construction in (65). Note that the morphology correspondingto 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 thehierarchy than direct objects—which were shown in the preceding section to be gapped underrelativization.(64) plan zuqw ta skiiza7-s-a ta smülhats-aalready die det child-3sg.poss-det det woman-det‘the woman’s child has died’(GN 1335; LT 2440)(65) wa7 flal ta smüthats-a ta züqw-a ta sküza7*(s)aprog cry det woman-det det die-det det child-3sg.poss-det‘the woman whose child died is crying’(ON 1336; LT 2441)31(66) qwits ti sqáycw-a tu wa7 alkst ti sem7ám*(s)al5rich det man-det det prog work det wfe-3sg.poss-det‘the man whose wife is working is rich’(GN 1340; LT 2442)(67) qwits ta smülhats-a ta xzüm-a ta tsftcw-s-arich det woman-det det big-det det house-3sg.poss-det‘the woman whose house is big is rich’(GN, RW 1354; LT 2443)These constructions are difficult for speakers to process, though GN, RW, and LT accept themas grammatical. Interestingly, AA does not find these sentences grammatical, but morereadily 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 oran 0, presumably because possessors are already fairly inaccessible to relativization, and it isyet 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, ifcorrect, since it has not hitherto been obvious. Kroeber (1991: 187), for example, regardsrelative clauses and the residues of cleft constructions as nondistinct in Salish, since theyappear identical in form. The data examined above suggest the summary in (68):(68) Possessor-extraction contexts in St’ at’ imcetsI Grammatical I Ungrammaticalrelativization focus; wh-questions;ordinary clausesFinally, note that possessor relativization does not show stage- vs. individual-level-predicate 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 predicatesapparently do not (Gardiner et al. 1993: 140-141). A similar asymmetry originally appeared15 The determiner tu is underlyingly the usual ti/ta, the vowel merely havingassimilated to the initial labial of the following wa7.32to exist for possessor relativization in St ‘at ‘imcets, as in (69), which relativizes the possessorof an argument of a stage-level predicate, wáz’arn ‘to bark’:(69) wa7 we7áw’ ta smiMhats-a ta *(wa7) wáz’-am ta sqáxa7-s-a’6prog shout det woman-det det prog bark-intr det dog-3sg.poss-det‘the woman whose dog barked is shouting’(ON, RW 1356)Possessor relativization is ungrammatical without the progressive auxiliary wa7, but this illformedness is merely aspectual in nature (Henry Davis, p.c.). The sentence is improved whenwa7 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 orany other auxiliary. More importantly, however, although the Nla’kapmxcmn consultant forGardiner et al. (1993), DU, made the distinctions reported by them on 2 June 1993, during themost recent elicitation of these data she did not. The relevant sentences are the following:’7(70) swat k zum k jtx”-swho unr big unr house-3sg.poss‘whose house is big?’(71) swat k wác’ama k sqaqa?-swho unr bark unr dog-3sg.poss‘whose dog barked?’(72) swat k q’iyx k sqáqa?-swho unr leave unr dog-3sg.poss‘whose dog left?’(DU, Nla’kapmxcfn, 19 April 1994)16 The variant of this sentence without wa7 must also have the enclitic portion of thediscontinuous determiner (ta wáz’-am-a)—as discussed in § 1.3.1—although this variant isirrelevant, since the sentence without wa7 is in any case ungrammatical (as noted in the text).Consult Thompson and Thompson (1992) for explanation of the grammaticalabbreviations in these examples.33Gardiner et al. (1993: 140-141) report that (70) (their example (7))—having possessorextraction 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 representedin (70-72) remain stable, then Nla’kapmxcfn and St’át’imcets may be more alike with respectto possessor extraction than has hitherto been suspected. Unfortunately, possessor extraction ismarginal enough that it will likely be difficult to ascertain consistent judgments for thisconstruction in either language.2.2.5. Object of comparisonRecall from page 21 that Kroeber (1991: 232) did not have any data on relativization ofobjects of comparison. These are ranked lowest on Keenan and Comrie’s (1977) AccessibilityHierarchy (35), and indeed it is difficult to relativize them in St’át’imcets. This is notsurprising, given the difficulty in relativizing possessors, which are argued by them to behigher on the scale. (73) is a main clause comparative, and (74-75) attempt to relativize theobject of comparison:(73) p’a7cw s-zác-al’qwem’-s ta sqáycw-a lhel ta smülhats-amore nom-long-appear-3sg.poss det man-det than det woman-det‘the man is taller than the woman’(AA, LT, RW 1051)(74) ?wa7 tayt ta p’á7cw-a s-zác-al’qwem’ sqaycw lhel ta smülhats-aprog hungry det more-det nom-long-appear man than det woman-det‘the woman whom the man is taller than is hungry’(AA, LT, RW 1052)(75) ?wa7 tayt ta sqáycw-a wa7 p’á7cw-a zác-al’qwem’-a lhel ta smülhats-aprog 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).34The consultants therefore can be seen to differ in the ease with which they relativize the lower-ranked 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, thesedata may be irrelevant to the predictions of the Accessibility Hierarchy if objects ofcomparison are not a distinct grammatical role in Salish—a separate question that has not beeninvestigated.2.3. SummaryConsider the general patterning of morphology that corresponds to non-oblique targets ofrelativization, as summarized from the preceding discussion in (76):(76) Morphology with non-oblique targets of relativizationErgative Absolutivesubject of transitive I possessor18 subject of intransitive I direct objectmorphology corresponding to target is obligatory gap corresponding to targetobligatorily retainedA clear ergative/absolutive pattern has emerged. Absolutive-centered RCs obligatorily omitperson morphology. Ergative-centered RCs, however, are complicated in two respects: first,subject morphology corresponding to the target is retained, and second, there exists specialmorphology (the topical object marker -tall) that can occur only in this type of RC. Not onlyare ergative-centered RCs therefore marked more explicitly than absolutive-centered RCs (byretaining the 3sg subject suffix -as), but they can also employ a distinct strategy ofrelativization (affixation of -tall). These are exactly the properties claimed by Keenan andComrie (1977) to show that such a relativized constituent is lower on the AccessibilityHierarchy than higher roles, which are relativized more easily. The Accessibility Hierarchy18 Possessor appears under the ergative column because it behaves the same as A withrespect to retention of morphology when relativized. It is not clear whether this fact has thesame explanation in both ergative- and possessor-centered RCs, but see Allen (1964) for asurvey of languages in which there is a formal correspondence between ergatives andpossessors.35for 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 > LocativeKeenan and Comrie (1977: 80 ff.) are reluctant to assign distinct positions on the AccessibilityHierarchy 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 applyonly 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 auseful description of the facts in St’át’imcets relativization, but a structural explanation for theranking in (77) is offered in §4.4.An ergative pivot is apparently at work in St’át’imcets relativization, at leastmorphologically, and this presents a problem for a theory of relativization that assumes anotion of subject that is defined on {A,S}. The status of the morphology in ergative-centeredRCs is not clear, though; if -as were not a subject suffix, for example, then the ergativecentered RCs would more plainly have a gap, exactly as do the absolutive-centered RCs. Theanalysis of RCs in §4.4 will assume that ergative-centered RCs do indeed involve extraction ofthe relativized constituent, and such an analysis is surely desirable, not only on generaltheoretical grounds, but because there is indirect evidence for ergative extraction inHalkomelem, a Coast Salish language that omits the person morphology corresponding torelativized ergatives (Gerdts 1988: 82-83)—exactly the environment in which St’át’imcetsseems to retain such morphology. Nevertheless, in having a restriction on relativization ofergatives (whether it is analyzed as morphological or syntactic—and the division is not alwaysclear), St’át’imcets is only one of many such languages, including Tzutujil (Dayley 1985: 210-2 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), CoastTsimshian (Mulder 1988: §3.10; 1989: 426-428), Inuktitut (Creider 1978: 99 ff.; Johns 1992:3672), Yup’ik Eskimo (Payne 1982: 85-87), and Greenlandic Eskimo (Woodbury 1975, cited byKeenan and Comrie 1977: 83; cf. Smith 1984).However the facts presented in this chapter are analyzed, it is important to bear in mindthat relative clauses in Stát’imcets resemble RCs in other languages, in that they exhibit anobligatory gap. This gap mirrors the extraction of relative pronouns and empty operators inIndo-European languages like English.37Chapter 3Conjunction3.1. IntroductionThis chapter examines the properties of pronominal coreference across sentential conjuncts inSt’át’imcets. As explained in chapter 1, Dixon (1979) shows that languages may vary in theirchoice of syntactic pivot for processes of conjunction (‘coordination’ in his terminology) andsubordination. If a syntactic process treats A and S alike, in opposition to 0, it is said to havean accusative pivot. On the other hand, if S and 0 (but not A) are uniquely identified by asyntactic operation, the process is said to have an ergative pivot. Recall the chart in (1) onpage 3. It is also possible for a language to be neither ergative nor accusative with respect tothese 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 subordinationtypes in St’ät’imcets, relativization. This chapter examines conjunction (the other majorprocess 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 inthis construction. It will be seen that conjunction, unlike relativization, is neither ergative noraccusative. Although one speaker sometimes appears to have an accusative pivot, it isprobably best regarded not as a syntactic pivot, but instead as a constraint on the discourseroles of coreferent NPs (Matthewson 1993a). For all speakers, coreference is most forcefullyconstrained by the one-nominal interpretation law (Gerdts 1988), which compels a uniqueinterpretation in conjuncts that contain a single overt nominal.St’át’imcets has many ways of marking conjunction; for a sampling of subordinatingand coordinating conjunctions, see van Eijk (1985: 172-173, 211-213, 217-219, 252, 270-273). In order to restrict the domain of investigation and thus facilitate the comparison ofsentences, 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 in38chapter 1 (p. 7) and seen throughout chapter 2 in cleft constructions. The conjunction nil/i isclearly subordinating, rather than coordinating, as it induces subordinate inflection(specifically, nominalization) in the non-initial conjunct. Inflection of clause types inSt’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 havingboth a subject and an object are rare in discourse (van Eijk 1985: 262)—since St’át’imcets is aradical head-marking language and thus allows arguments to be referenced solely bymorphology on the predicate—entirely third-person sentences without at least one overtnominal argument are highly disfavored discourse-initially. For this reason, in eliciting mostof 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 firstconjunct. The context that was used in each case is always cited below. Even transitivesentences having a single overt nominal—for example, (18-19) in chapter 1 (p. 10)—aredisfavored discourse initially, and are regarded as felicitous only when there is a discoursecontext that can supply a subject for such sentences. For elaboration of this point, see thediscussion concerning (124-126) in chapter 4 (pp. 63 ff.).3.2. Conjuncts with non-third-personIt is easy to show that there is no single pivot for conjunction that is either ergative oraccusative by constructing null-pronominal sentences whose second, transitive conjunct hasone pronominal affix (marking either A or 0) that is uniquely non-third-person. Consider, forexample, sentences (79-80), which have (78) as their context:(78) qwatsáts i smelhmülhats-aleave p1. det women-det‘the women1 left’39(79) p’ an ‘t-wit nilh s-7ats’ x-en-tsál-itas19return-3p1. intr conj nom-see-tr-lsg. obj-3p1. su‘they1 returned and they saw me’(AA, GN, RW 1382)(80) ‘an’ t-wit nilh s-7ats’ x-en-tánih-anreturn-3pl. intr conj nom-see-tr-3p1. obj-lsg. conj‘they1 returned and I saw them’(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 subjectin (79), and the object in (80). In each sentence, the 3pi argument of the second conjunct maycorefer with the 3pi subject of the first conjunct, as indicated in the glosses by subscriptindices. Note, importantly, that (79) and (80) differ minimally in the person morphology ofthe second conjunct; the conjoined predicate in neither of these sentences needs grammatical-function-changing morphology, e.g., passive -m or the topical object marker -tall. IfSt’át’imcets conjunction had an ergative or an accusative pivot, it might be expected that aprocess like passive or antipassive would operate in either (79) or (80) in order that such a GFchanging operation could feed the syntactic pivot (Dixon 1979: 120-130).There are similarly no asymmetries in sentences having two intransitive conjuncts; apronominal 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 eachconjunct of the following examples; it may not be deleted under identity with the clitic in thefirst conjunct:19 LT (2445, 2446) prefers that this sentence and the next one begin with the pastcomplementizer I, which also triggers conjunctive -as on the predicate: i p’án’t-wit-as ‘whenthey returned’. Coreference obtains as indicated in the text by the subscript indices.40(81) qwatsats*(kalh) nilh s-p ‘an’ t*(ka1h)2Oleave-ipi. intr conj nom-return-lpl.poss‘we left and then we returned’(AA, GN, RW 1389)(82) mátq-kalh nilh skwis*(kalh)walk-ipi. intr conj nom-fall-lpl.poss‘we walked and then we fell’(AA, GN, RW 1391)(83) ka-kwfs-kalh-a nilh s-ka-güy’ t*(ka1h)a2looc-fall-lpl. intr-ooc conj nom-ooc-sleep-lpl.poss-ooc‘we fell and then we slept’(AA, GN 1509)(82) has an unergative predicate in its first conjunct, as does the second conjunct of (83); theremaining conjuncts in (8 1-83) have unaccusative predicates. Because coreference betweenconjuncts is free in all of these combinations of different kinds of predicates—and becausenone of them show signs of GF-changing operations—there is again no evidence for a syntacticpivot.Conjunction in St’át’imcets therefore appears to be neither ergative nor accusative,unlike relativization, which is ergative (chapter 2). In this respect, St’át’imcets probablybehaves 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, interms of the ergative/accusative continuum. That is, processes such as coordination[= conjunction] may not operate in terms of well-defmed constraints like thoseapplicable to Walmatjari and Dyirbal. . . . Certainly, some languages have aconsiderable set of well-defmed syntactic constraints, which facilitate a clear judgmentof their position on the ergative/accusative syntactic scale; but others have more fluidconditions that provide slimmer evidence for judgment. For instance, coordination20 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)leave-ipi. intr conj part conj nom-return-lpl.poss‘we left and then we returned’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)afall-ipi. intr conj part nom-ooc-sleep-lpl.poss-ooc‘we fell and then we slept’41may largely follow semantic, stylistic, or discourse-organization preferences, ratherthan 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 objectCoreference across conjuncts is more restricted in sentences having third-person subjects andobjects with various combinations of proper nouns. Because consultants have the mostconsistent judgments for coreference in sentences having two transitive conjuncts, these will beexamined first (3.3. 1). Section 3.3.2. then examines sentences in which one of the conjunctsis intransitive; the data here are less firm, and so it is harder to base generalizations on them.3.3.1. Conjoined transitives3.3.1.1. Parallelism constraint on discourse functionsSentences having two transitive conjuncts display striking evidence for a parallelism constrainton 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) Pz-án-twal’ wi s-Bill müta7 s-John. Wa7 wi7 cmán’-twal’-wit.meet-tr-recip p1 nom-Bill conj norn-John prog ernph enerny-recip-3p1.su‘Bill1 and Johns met each other. They’re enemies.’22 A reciprocal context with same-sex participants (‘Bill and John met each other’) isused in order to minimize the possibility that one participant will be more prominent in thediscourse and unduly affect the coreference possibilities in the sentences of interest. Initialattempts to elicit conjoined transitives used asymmetric discourse contexts, with male andfemale participants and predicates like ts ‘urn ‘qsán’ ‘kiss’, and coreference judgmentsconsequently turned out to be asymmetric in unusual ways that suggested that the consultantwas basing them on facts that she thought would be more likely to obtain between men andwomen (as she later confirmed).42(85) áts’ x-en-as nilh s-qvlqvl-ts-mf&-as23see-tr-3erg conj nom-bad-mouth-appl-3sg. conja. ‘he1 saw him, and then {he1 swore at him/*he swore at him1}’b. ‘he saw him1, and then {he swore at him1/*he swore at him}’(AA, GN, RW 1595)Because the first conjunct of (85) has no overt NPs, it may be ambiguously interpreted as towho is seeing whom. However, given a specific interpretation of the first conjunct, there is atendency, for all speakers, to interpret the subject and object of the second conjunct asreferring to the same subject and object of the first conjunct.A similar phenomenon can be observed in complex clauses. In examining pronominalcoreference into relative clauses in St’át’imcets, Matthewson (1993a) independently foundthat—although coreference into relative clauses is not restricted by Condition C of the bindingtheory (see Chomsky 1981; 1986; etc.)—it is restricted by a constraint on the discoursefunctions of the coreferent items. Consider the following paradigm:(86) ts’um’-qs-án’-as kw-s Mary ta sqáycw-a ta xwis-ás-alick-nose-tr-3erg det-nom Mary det man-det det love-3sg. conj-deta. ‘Mary1 kissed the man that she loves’ (topic. . . topic)b. *‘the man who loves her1 kissed Mary1’ (non-topic. . . non-topic)c. *‘Mary1 kissed the man who loves her1’ (topic. . . non-topic)d. *‘the man that she1 loves kissed Mary1’ (topic. . . non-topic)(GN, RW 386; Matthewson 1993a: 19; Matthewson et al. 1993: 229)Kinkade (1989; 1990) shows that topics (‘old’, presupposed information in the discourse) inSalish 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-canonicalpositions (object and subject respectively) if the predicate is passivized or affixed by the topicalobject marker, though in St’át’imcets the latter strategy may be used only in extractioncontexts, 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 bythe topical object marker, the subjects in the interpretations (a-d) correspond straightforwardly23 Qvlqvl-ts-mmn’ ‘to swear at (tr)’ here and throughout represents the Fountainpronunciation; Mount Currie speakers retract and glottalize the lexical suffix -ts ‘mouth;language’, as noted in § topics, and objects to non-topics. As indicated, the preferred interpretation is the one inwhich ‘Mary’ and the coreferent null pronominal are the topics of their clauses (a). The otherinterpretations (b-d) differ in this respect, and are dispreferred. Reading (b), with ‘Mary’ andthe coreferent null pronominal as non-topics, is also allowed for some speakers, and was thefirst translation offered for (86) by AA, presumably in accord with her fairly strict VOSorder.24Data like those in (85) and (86) suggest that St’át’imcets is subject to the followingrestriction (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 (eithertopic of the discourse or non-topic). In addition, there is a preference for bothcoreferential elements to fulfill the topic of the discourse function.Constraint (87) can be observed in the two other MS languages, Nla’kapmxcfn andSecwepemctsfn (Matthewson et al. 1993: 228-230), and a similar parallelism restrictionobtains 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 thissection, are ungrammatical only for RW. For the other consultants (AA and GN)25, theinterpretations marked by the asterisk (*) should be taken only as less obvious and mildlydispreferred. When translating the sentences in this section into English, for example, theseconsultants spontaneously provide glosses in accord with (87), but will concede that the nonparallel reading is possible if prompted for it.24 AA prefers the only the determiner k in front of the proper noun, and also aprogressive auxiliary inside the relative clause, which requires the enclitic -a portion of thediscontinuous determiner to drop.(i) ts’um’-qs-án’-as k Mary ta sqáycw-a ta wa7 xwis-aslick-nose-tr-3erg det Maiy det man-det det prog love-3sg. conj25 BF shares the same judgments as AA and GN. Thanks to Henry Davis for checkingthese data with her.44The parallelism constraint is expected to hold across conjuncts if one conjunct has twoovert NPs. This is indeed the case, as (88-89) show:(88) áts’x-en-as nilh s-qvlqvl-ts-mIn’-as kw-s John kw-s Bill26see-tr-3erg conj nom-bad-mouth-appl-3sg. conj det-nom John det-nom Billa. ‘he1 saw him, and then {Bill swore at John/*John swore at Bi111}’b. *‘h saw him1, and then {Bill swore at John/John swore at Bill1}’(AA, GN, RW 1597)(89) áts’x-en-as kw-s John kw-s Bill nilh s-qvlqvl-ts-mfn’-assee-tr-3erg det-nom John det-nom Bill conj nom-bad-mouth-appl-3sg. conja. ‘Bill saw Johns, and then {he1 swore at him/*he swore at him1}’b. *‘John saw Bill1, and then {he swore at him1/he swore at him}’(AA, GN, RW 1596)Recall from chapter 1 that the preferred word order in elicitation is VOS, especially insentences 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-orderpreference immediately accounts for the ungrammaticality of interpretation (b) for (88), sinceit would require VSO order in the second conjunct. Interpretation (a), however, respects theVOS 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 thefirst conjunct (‘Bill saw John’), in which case the interpretation of the second conjunct must26 Here and throughout this chapter, the cited forms are the ones that originate fromGN and RW (Fountain speakers). Although AA and sometimes LT are cited with each formas one of the consultants, the sentences that they examined differed minimally in that propernouns were preceded by either the determiner k(w) alone or the nominalizer -s alone, but neverby both together, as are the proper nouns in this sentence. It is hard to be sure whatcharacterizes this variation between speakers; see § 1.3.2 for further examples and discussion.45have a null pronominal subject coindexed with the subject of the first conjunct.27 Aninterpretation 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 apassivized predicate in the second conjunct, as exemplified below:(90) áts’x-en-as kw-s John kw-s Bill nilh s-qvlqvl-ts-mfn’-emsee-tr-3erg det-nom John det-nom Bill conj nom-bad-mouth-appi-passBill1 saw Johns, and then he1 was sworn at by him’(AA, GN, RW 1596)Another example of passivization switching discourse topics is in (107) below. Note that theungrammatical interpretation in (89a) cannot similarly be saved by affixation of the topicalobject marker -tall; such a sentence is ungrammatical under any reading:(91) *áts’xenas kw-s John kw-s Bill nilh s-qvlqvl-ts-mfn’-talisee-tr-3erg det-nom John 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, theSt’át’imcets cognate appears only in extraction contexts (Matthewson 1993a: 4-5). This wasnoted with respect to examples (38-39) in chapter 2 (p. 22). Consider also the passivesentence (22) in chapter 1 (p. 10), repeated below as (92):(92) nilh t’u7 s-e-s k’al’-em-mIn-em ti syáqts7-afoc 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 ti syáqts7-afoc part nom-prog-3sg.poss wait-intr-appl-TO det woman-det‘he is being waited for by the woman’(LT 2430)27 LT, who often has VSO order in elicitation, allows interpretation (a) like the otherspeakers, but additionally permits the reading of (b) which has parallelism, ‘John saw Bill1,and then he swore at him1’.463.3.1.2. Interaction with the one-nominal interpretation lawThe remaining class of sentences with two transitive conjuncts are those that contain a singleNP in one or both of the conjuncts. Recall from chapter 1 10) that St’át’imcets is subjectto 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 interpretedas the absolutive.That is to say, in sentences having a single NP and only third-person morphology on thepredicate, the NP is interpreted as absolutive, i.e., as the only argument (subject) of anintransitive 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’-assee-tr-3erg det-nom John conj nom-bad-mouth-appl-3sg. conja. ‘he saw Johnj, and then {he swore at him/*he swore at him}’b. *‘John saw him1, and then {he swore at him/he1 swore at 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 onlybe grammatical when ‘John’ is the object of the first conjunct, the null pronominal object ofthe second conjunct in such a case must likewise corefer with the object ‘John’ of the firstconjunct, in accord with the parallelism constraint (87).The obvious question that arises with a constraint like parallelism is whether it interactswith the one-nominal interpretation rule, since it is possible for the parallelism constraint andthe one-nominal interpretation law to impose conflicting demands in a single sentence. Therelevant context is a sentence having a single overt NP in each conjunct, i.e., structures of thefollowing form, where pro marks a null pronominal argument, and NP1 is not coreferentwith NP (the order of elements within brackets is irrelevant):47(96) [s. . . pred.. . . NP1 . . pro. . . j conj [s. . . preci. . . NP . . . pro. . .]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, butparallelism will be violated. On the other hand, if {NP1,NP} bear different GFs or discoursefunctions, parallelism will be satisfied (assuming also the appropriate indexing of pro in eachconjunct), 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) áts’x-en-as kw-s John nilh s-qvlqvl-ts-mfn’-as kw-s Billsee-tr-3erg det-nom John conj nom-bad-mouth-appl-3sg. conj det-nom Billa. ‘he1 saw John, and then (i) hej swore at Bi111/(ii) *Bil1 swore at him’b. *‘John saw him, and then (i) he swore at Bi111/(ii) Bill1 swore at him’(AA, GN, RW 1593)That (97) has a grammatical interpretation at all is striking, since it illustrates that theparallelism 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-nominalinterpretation law.The relatively low ranking of the parallelism constraint (87) expresses twogeneralizations about coreference across transitive conjuncts in St’át’imcets, the most obviousof which is that, for all but one consultant, parallelism is merely a preference; speakers permitinterpretations that violate it, even in sentences about which the one-nominal interpretation lawhas nothing to say. More importantly, however, it captures the difference between thegrammars of different consultants. Given only sentence (85), for example, it is unusual thatviolations of parallelism should be ungrammatical for a single consultant, but possible for theothers. However, sentences like (97)—in which parallelism and the one-nominal law imposeconflicting demands—show that PC is violable even for the more restrictive consultant, whowould otherwise appear to have a grammar different from other speakers of the same48language. (97) therefore reveals that all of these speakers share the dominant ranking of theone-nominal interpretation law.One way of encoding the relationship between these constraints is to employ theformalism of Optimality Theory (Grimshaw 1993; Prince and Smolensky 1993), whichhypothesizes that grammatical constraints are in principle violable, and that minimal violationis allowed when it leads to the satisfaction of higher-ranked constraints. Although theformalism of this theory is useful for explicating the coreference data of St’át’imcets, it is usedmerely for expository convenience, and it has nothing of interest to say about the nature ofthese constraints, whose interpretations are tied to syntactic structures (examined in §4.3).Because the one-nominal constraint has precedence over the parallelism constraint inevaluating whether a given interpretation is grammatical, the former is ranked above the latter,as formalized in (98), where >> is interpreted as ‘dominates’:(98) ONO >> PCEach possible interpretation of (97) is a member of the ‘candidate set’, and is evaluated withrespect 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 theirranking. Each occurrence of an asterisk (*) before a constraint indicates a single violation ofthat constraint; lack of such a mark indicates that the constraint is satisfied or irrelevant. Theoptimal candidate—which is the grammatical interpretation—is identified by the dollar sign ($).(97) is therefore evaluated as follows:(99) áts ‘x-en-as kw-s John nilh s-qvlqvl-ts-mmn ‘-as kw-s Bill (=97)a. ‘he saw John, and then he swore at Bill1’ (=97a.i) ONO *PC $b. ‘he1 saw John, and then Bill swore at him’ (=97a.ii) *ONO PCc. ‘John saw him1, and then hej swore at Bill1’ (=97b.i) *ONO PCd. ‘John saw him1, and then Bill1 swore at him’ (=97b.ii) **ONO *pNotice that the total number of constraint violations is not relevant to the evaluation ofcandidates. The candidates (a-c) each have a single constraint violation, yet only one of them49(a) is grammatical. The fact that candidates (b-c) satisfy PC is not sufficient, since it is betterto satisfy the higher-ranked ONO, even if it requires that the lower-ranked PC be violated, asit is in (a). The contrast between (a) and (b-c) illustrates a clash between competingconstraints, 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 tableauin (101) shows, one reading (c) should be preferred over the other (a):(100) áts’x-en-as nilh s-qvlqvl-ts-mmn’-as kw-s Billsee-tr-3erg conj nom-bad-mouth-appl-3sg. conf det-nom Billa. ‘he1 saw him, and then (i) he swore at Bill/(ii) *Bill. swore at him’b. ‘he saw him1, and then (i) hej swore at Bill1/( i) *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. ‘he1 saw him, and then he swore at Bill1’ FI (=lOOa.i) ONO *pb. ‘he1 saw him, and then Bill1 swore at him’ (= lOOa.ii) *ONO PCc. ‘he saw him1, and then he swore at Bill1’ 0 (=lOOb.i) ONO PC $d. ‘he saw him1, and then Bill swore at him’ (= lOOb.ii) *ONO *PCThe checkmark (0) indicates the interpretations that ON and RW allow. Candidates (b, d) arecorrectly 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 thissentence with AA and LT, interpretation (c) was uniquely grammatical, and interpretation (a)was not allowed, in accord with the candidate evaluation in (101). Grammaticality judgmentsfor these sentences are subtle, and it is likely that the data concerning this point will continueto vary across speakers and across elicitation sessions.The constraint-ranking motivated above accords with the fact that the one-nominalinterpretation law is a salient property of St’át’imcets grammar, applying generally throughoutthe language, in simple sentences as well as in compound and complex ones. Havingestablished that this is the case, it remains to be explained. Of course, it will likely remain50unexplained why the parallelism constraint has varying importance across speakers, although itaccords with Dixon’s (1979: 129) statement, quoted above on page 41, that ‘. . . processessuch as coordination [= conjunction] may not operate in terms of well-defined constraints.[Sjome languages. .. have more fluid conditions... [F]or instance, coOrdination may largelyfollow semantic, stylistic, or discourse-organization preferences, rather than conforming to anystrict syntactic matrix.’ St’át’imcets conjunction takes on yet more of this character in the nextsection, which examines intransitive conjuncts.A table summarizing the various combinations of null and overt NPs and theircoreference possibilities across transitive conj uncts is given in (152) in chapter 4 (p. 82).3.3.2. IntransitivesThis section is divided into two parts: § examines data that have an intransitivepredicate in the first conjunct, and § presents data that have an intransitive predicate inthe second conjunct. Intransitive in first conjunctIf subject of intransitive (S) is in the same structural position as subject of transitive (A), bothS 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 Johnprog wait-intr det-nom John‘Johns is waiting’(103) p’an’t kw-s Bill nilh s-7áts’x-en-asreturn det-nom Bill conj nom-see-tr-3sg. conj‘Bill1 returned and (i) he1 saw him/(ii) *he saw him1’(AA, GN, RW 1586)51The second conjunct of (103) does not have any overt nominals, and so the one-nominalinterpretation law (94) is not implicated here. For this sentence, all speakers28 interpret thesecond conjunct in accord with parallelism, suggesting that A and S are indeed in the sameposition. However, consider (104) below, which is complicated by the addition of a single NPin the second conjunct.(104) p’an’t kw-s Bill nilh s-7áts’x-en-as kw-s Johnreturn det-nom Bill conj nom-see-tr-3sg. conj det-nom John‘Bill returned and (i) he1 saw John/(ii) ?John saw him1’(AA, GN, RW 1584)Reading (i) was spontaneously volunteered by all speakers as a translation of (104), althoughreading (ii) was unexpectedly accepted by GN and RW after it was suggested to them (inviolation of the one-nominal interpretation law). The same judgments are given by BF (HenryDavis, p.c.). For one speaker, AA, reading (ii) is ungrammatical, as expected, given the highranking of the one-nominal interpretation law demonstrated in § is not clear why several speakers would accept interpretation (ii) for (104), especiallysince a sentence having the same structure, but with different names, was on two occasionsinterpreted by GN and RW as expected. The context for (106) is (105):(105) wa7 k’ál’-em kw-s Maryprog wait-intr det-nom Mary‘Mary is waiting’(106) p’an’t kw-s Bill nilh t’u7 s-áts’x-en-as kw-s Maryreturn det-nom Bill conj part nom-see-tr-3sg. conj det-nom Mary‘Bill1 returned and (i) he saw Mary/(ii) *Mary saw him’(AA, GN, RW 1392)In (106), the one-nominal interpretation rule pressures ‘Mary’ in the second conjunct to be thedirect object, thus excluding the reading in (ii). This reading can be saved, however, if thepredicate is passivized:28 Except BF and LT, who accept both interpretations (i) and (ii). Thanks to HenryDavis for checking this point with BF.52(107) p’an’t kw-s Bill nilh t’u7 s-áts’x-en-em s-Maryreturn det-nom Bill conj part nom-see-tr-pass nom-Mary‘Bill returned and he1 was seen by Mary’(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 permitan interpretation that is excluded by the one-nominal rule suggests that ONO indeed holdsstrongly throughout the corpus. It would be worthwhile to recheck the problematicinterpretation in (lO4ii), though. Intransitive in second conjunctThe 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 overtnominal in one conjunct or the other. As expected, their possible interpretations are stronglydetermined by the one-nominal interpretation law:(108) Pz-án-twal’ wi s-Bill m.ita7 s-John. Wa7 wi7 cmán’-twal’-wit.meet-tr-recip p1 nom-Bill conj nom-John prog emph enemy-recip-3p1.su‘Bill1 and John met each other. They’re enemies.’(109) qvlqvl-ts-mIn’-as kw-s John nilh s-qwatsáts kw-s Billbad-mouth-appl-3erg det-nom John conj nom-leave det-nom Billa. ‘he swore at John, and then (i) Bi111/(ii) *he. left’b. *‘John swore at him1, and then {Bill/he} left’(AA, GN, RW 1589)(110) qvlqvl-ts-mfn’-as kw-s John nilh s-qwatsátsbad-mouth-appl-3erg det-nom John conj nom-leavea. ‘he1 swore at John, and then left’b. *‘John swore at him, and then left’(AA, GN, LT, RW 1590)The (b) glosses in each of (109-110) are excluded because they violate the rule of one-nominalinterpretation, which requires ‘John’ to be the direct object, as it is in the (a) readings. Thesecond conjunct of (109) must have ‘Bill’ as subject of the intransitive predicate, since it is an53overt 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 parallelismconstraint, under the assumption that S and A are in the same structural position. It isimportant 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 overtNPs. 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 s-Bill nilh s-qwatsátsbad-mouth-appl-3erg nom-John nom-Bill conj nom-leavea. ‘Bill1 swore at John, and then he,9 left’b. ‘John swore at Bill1, and then heji*1 left’(AA, GN, RW 1592)All consultants accept both coreference possibilities indicated in (a), in violation of theparallelism 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 overttransitive subject ‘Bill’, was spontaneously offered by RW as a translation for (111).Moreover, during the most recent elicitation, the parallel interpretation was in fact preferredby both GN and RW to the non-parallel one. On these grounds, then, the formerinterpretation can perhaps be taken as preferred to the reading in which the null pronominal Sis 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 nilh s-qwatsátsbad-mouth-appl-3erg conj nom-leave‘he1 swore at him, and then he1j left’‘he swore at him, and 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 difficult54to explain, given a discourse context like (108) having enough overt NPs that could plausiblybe coreferent with the null pronominals in (112).The conclusion to be drawn concerning coreference between conjuncts having anintransitive predicate is that the one-nominal interpretation law holds strongly, as it does acrosstransitive conjuncts, but that parallelism is at best a tendency.3.4, SummaryCoreference across conjuncts is neither ergative nor accusative in St’át’imcets. It is moststrongly constrained by the one-nominal interpretation law (94), and the data that are presentedin §3.3. 1.2-3.3.2 in support of this point further establish ONO as a salient property of thegrammar 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 isintransitive, though, PC does not hold strongly (if at all), and the grammaticality of therelevant sentences is more difficult to establish firmly and consistently. This conclusion isconsistent with that of Matthewson (1993a), who found that intransitive subjects patterned withtransitive subjects only across discourse, not intrasententially. The data in this chapter involveonly intrasentential coreference, and so the failure of parallelism to hold for intransitivesubjects is not unexpected. An important question is therefore how to capture the asymmetricbehaviour of transitive and intransitive conjuncts with regard to coreference possibilities. Justas in chapter 2, the facts presented here again present a problem for a theory that treats A andS together as members of the category ‘subject’, e.g., Anderson (1976) and Keenan andComrie (1977: 80 ff.). However, while chapter 2 showed that S and 0 pattern alike withrespect 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 tohave either an ergative or an accusative pivot.It should be stressed that in having a parallelism constraint that applies only totransitive conjuncts, St’át’imcets is not unusual. In examining a similar range of facts inEnglish, Oehrle (1981) comes to nearly identical conclusions. Oehrle’s propositional55congruence (his ‘rule R’), though, requires additionally that the predicates in each clause beidentical and that the rule apply with a specific intonation.29 Oehrle notes that having anintransitive predicate in one conjunct makes the reference of its subject difficult to establish inisolation (because the different predicate disrupts the congruence across conjuncts), althoughits 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 Englishand St’át’imcets, in rather complex constructions requiring subtle intuitions about theirambiguity and grammaticality, strongly suggests that the parallelism constraint—whatever itsultimate explanation—is a plausible candidate for Universal Grammar.29 The intonational properties of sentences in this chapter have not been studied, and soit is possible that some of the seemingly problematic data are due to misunderstood prosodicproperties of the utterances.56Chapter 4Structural Asymmetries4.1. IntroductionThis chapter shows that the various asymmetric behaviours of subjects in St’át’imcets, asdescribed in chapters 2 and 3, may be given a structural explanation within the broaderPrinciples and Parameters framework of Chomsky (1981; 1982; 1986; 1992; etc.). Section4.2 addresses the one-nominal interpretation law, §4.3 examines the appearance of parallelismon 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 thatare topics are structurally higher than NPs that are focused.30The 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 rigiddefinitions of topic and focus here would not serve well the purpose of this chapter, whichcrucially assumes only that, however topic and focus are defined, the NP representing theformer 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 generaldefinitions.Reinhart (1981: 57-58) identifies two majors approaches to topichood. One approachdefines the topic as the noun phrase whose referent a particular sentence is about, and the otherdefines the topic as the old information in the discourse. Reinhart argues for the formerapproach (‘pragmatic aboutnes&), and suggests the following test for topichood (Erteschik-Shir1993: 1; 22, §4):30 The idea to apply this structural asymmetry to an analysis of St’át’imcets is duegenerally, and in each specific case examined below, to Henry Davis and HamidaDemirdache. (The particular structures given below were suggested by Henry Davis.) Thischapter could not have been written without their help.57(113) Topic testSpeaker A: Tell me about XSpeaker B: . . . X. . . (X = topic)In speaker B’s answer to A’s question, X is the topic, since B is telling A about X. The topicis 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 ofsubject to topic can of course be altered by special morphology, phonology, syntax, or bywhatever else a language might use to alter the topic-focus relation, and examples fromSt’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 thesisas follows:31(114) Topic Constraint* TOP1 [SUBJECT [vp. . .The topic constraint does not exclude overt (S-structure) topicalization (e.g., Fred1 I like t,)nor left-dislocation (e.g., Fred1 I like him); rather, the topic constraint merely maps syntacticpositions to discourse functions at a level of derivation beyond S-structure, here assumed to belogical form (LF). Because this chapter will adopt a version of the VP-internal subjecthypothesis, it is necessary to revise the topic constraint as in (115), in order that the subject ofa clause may be base-generated in Spec/VP and still be mapped to topic:(115) Topic Constraint (revised)* TOP1 [SUBJECT [v’. . . NP1...]]31 The topic constraint captures a valid cross-linguistic generalization, though it hasapparently never been derived. Its explanation may be related to the fact that subjects arealways generated higher than objects, though the question is beyond the scope of this thesis.58The 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 thetopic.The topic constraint forbids the object from being the topic, but as a negative constraintit does not specifically compel the subject to be the topic. Consider, however, that the topic isgenerally regarded as being in an A’-position, hence operator-like; see, for example, thesentences in (140, 146) below (pp. 74, 78). The ban on vacuous quantification (Chomsky1982: 11-13) would exclude a sentence like the following, since the topicalized NP does notbind anything in the sentence:(116) *Fred, I like herThe ban on vacuous quantification conspires with the topic constraint (115) to ensure that thesubject and topic are coindexed. Specifically, if the topic and object are coindexed, the topicconstraint (115) is violated. Although the topic constraint will be satisfied if the topic does notbind any NP in the sentence, such quantification would be vacuous. Therefore, a well-formedstructure is derived only when the topic and subject are coindexed. The mapping of subject totopic that emerges from the interaction of the topic constraint (115) and the ban on vacuousquantification can be satisfied either by LF movement of the subject to topic position, or bycoindexation of a discourse topic with a null pronominal subject. Both options will beexemplified in this chapter.The next question to be addressed concerns the syntactic position of the topic, somerecent proposals for which are illustrated below:a. Internal Topic b. External Topic c. Topic Phrase(117)CP E Top-PTopic1 IP Topic1 CP e Top’ti... . . . (pro1)... Top CP/IP59Aissen (1992: 47) suggests that Mayan languages have two topic positions, one for an internaltopic, which is in Spec/CP (1 17a), and another for an external topic, which is prefixed to anentire clause CP under a node labeled E (11Th). The internal topic position is a landing sitefor movement of a topic from within the sentence, while the external topic position contains abase-generated topic that may be coindexed with a pronominal within the sentence. The nodeE 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 phrasestructure 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 maximalprojection Top-P, and the NP that is the topic—because it is a maximal projection—appears inspecifier position. The topic NP may be overt, as in the case of S-structure topicalization, orit may be non-overt, as it is in (c). In this diagram, a null topic e occupies Spec/Top-P, and insuch a case its reference would be determined by discourse. 32It 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 wordorder, 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 aremapped at LF from overt NPs) and discourse topics (which are coreferential with NPs outsideof the sentence), neither the internal nor the external topic position in (1 17a, b) seemsappropriate. Instead, the topic phrase structure of (1 17c) is adopted here because it is broadenough to subsume both sentential topics and discourse topics. Other structures forrepresenting topics could be accommodated by the analysis here, though the only crucial (anduncontroversial) requirement of the representation of topic is that it be structurally higher thanthe non-topic/focus of a sentence—which is examined next.3332 Another structure that is not considered here, suggested by Authier (1992) andWatanabe (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 fromthis analysis. Although there might be more than one topic in the discourse, the notion oftopic that is considered here is that of topic of a sentence, i.e., the single NP which thesentence is about or is predicated of (in the sense of Erteschik-Shir 1993 and Reinhart 1981).60The 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 whetherit may answer a wh-question. In the following examples, the capitalized constituent is thefocused 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 correlationthat has been observed in intonationally unmarked sentences is that a focused NP typicallyappears 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 NPshas a tripartite structure that includes the quantifier, restrictive clause, and nuclear scope. Thequantifier has scope over variables in the restrictive clause, while unbound variables in thenuclear scope get bound by ‘existential closure’, which is the presence of a covert existentialquantifier. In theories that attempt to derive such logical representations from syntacticstructures, the division between the restrictive clause and the nuclear scope is drawn near theVP (Erteschik-Shir 1993; Diesing 1992: 9-10, 49-53; Partee to appear: 5). Further, as arguedby Erteschik-Shir (1993) and Partee (to appear), and as suggested by Diesing (1992), thenuclear scope contains the focus. The topic and non-topic/focus constituents therefore appearin the following structural relation, (1 19a) representing Erteschik-Shir and Partee’ s proposal,and (1 19b) representing Diesing’s:(119) a. [cp topic [vp focus]]b. [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 the61nuclear 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 focusThis section gives a structural interpretation of the one-nominal interpretation law (Gerdts1988: 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 interpretedas the absolutive.When the predicate is intransitive, a single 3rd person nominal must be interpreted asabsolutive, since the predicate takes a single argument. What (120) leaves unexplained is whythe same fact obtains when the predicate is transitive, since the nominal might just as easily beinterpreted as the subject. In the following sentence, for example—repeated from (20) ofchapter 1 (p. 10)—the single overt nominal is uniquely interpreted as the object.(121) áts’x-en-as ta sqaycw-asee-tr-3erg det man-det‘he saw the man’(AA, BF, GN, LT, RW 29)(121) may not be interpreted as ‘the man saw him’. This restriction may be structurallyencoded given the reasonable assumption that NPs will not appear overtly in a sentence if theirreferents are presupposed in the discourse. As discussed in §1.3.1, St’át’imcets is a radicalhead-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. The followingexample is repeated from (3) of chapter 1 (p. 5):62(122) tsün-tsi-lhkantell-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) arealways 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, asthe following example shows (repeated from (12) of chapter 1 (p. 8)):(123) tsün-ø-astell-3sg. abs-3sg. erg‘she told him’(LT 2425)In sentences like these that do not have overt lexical arguments, the arguments of the predicateare generally assumed to be represented by null pronominal constituents (the empty categorypro). Each pro is in an argument position, and is licensed by a pronominal affix on thepredicate.34Note, however, that although sentences like (123) are grammatical, they aredispreferred discourse-initially, since the third-person referents are not identified, and becausepro cannot be used deictically. The inability ofpro to refer independently has been taken to bea characteristic of anaphors, as the contrast below suggests:(124) a. [Somebody in the audience gets up and leaves]Speaker: He is weirdWhere is the bastard going?Where is John going?I guess his patience ran outb. [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.63c. [Same situation]Speaker: *Hielfis weirdThe italicized NPs refer to the person who is leaving. Crucially, none of the NPs in (124a) issyntactically bound, since the antecedent comes not from the sentence or even from thediscourse, but merely from the speech situation. The same situation in (b, C), however, showsthat an anaphor (himself) may not be used deictically—in contrast to the non-anaphoricelements in (a). The pro of St’át’imcets therefore behaves as an anaphor in this respect, sinceit likewise may not appear unbound in a sentence, and minimally requires a discourseantecedent. 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, forexample, is subject to the same restriction. Consider the following contrast from Italian:(125) a. Quando Carlo1 1’ ha vista, pro1 è arrossitowhen Carlo her have. 3p see.prtc pro be. 3p blush.prtc‘when Carlo saw her, he blushed’b. Quando pro1 1’ ha vista, Carlo1 è arrossitowhen pro her have.3p see.prtc Carlo be.3p blush.prtc‘when he1 saw her, Carlo1 blushed’(126) a. *Quando 1’ ha vista Carlo, proj è arrossitowhen her have. 3p see.prtc Carlo pro be. .3p blush.prtc‘when Carlo saw her, he1 blushed’b. *Quando pro1 1’ ha vista, è arrossito Carlo1when pro her have. 3p see.prtc be. 3p blush.prtc Carlo‘when he1 saw her, Carlo blushed’(Calabrese 1990: 10)The sentences in (125) contain a preverbal subject, Carlo, in either the adverbial clause (a) orthe main clause (b); in each case, the pro subject that is in the other clause may take itsreference 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, post64verbal subject Carlo is not possible, and because these sentences do not contain a topic, thepro subject lacks an antecedent.A St’át’imcets sentence like (123), then—which has no overt nominals—may be freelyused non-initially, since two participants may be tracked throughout a discourse by mappingGFs to discourse roles according to one of the schemata in (119) and by using other methodsto indicate the switching of these roles.35 Once a discourse topic has been established, it neednot be expressed subsequently by an overt NP. Bearing this in mind as we return to a sentencelike (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 orperson about which the discourse is concerned, the presupposed information—would normallybe referenced solely by a third-person pronominal affix on the predicate. Because this NP is anew referent—and because the topic of a sentence represents a presupposed referent, as thetopic test in (113) shows—the most natural interpretation of this sentence is therefore the onein which ta sqáycwa ‘the man’ is the non-topic or focus. Given the mapping in (119), then, itmust be inside the VP as the object:(127) [cp [vp áts’x-en-as ta sqáycw-a]]see-tr-3erg det man-detUnder this approach, the one-nominal interpretation law as stated in (120) is not a rule ofSalish grammar, but simply a description of facts that are derived from (i) an independentlyneeded mapping that gives the result of (119), and (ii) the principles that license null NPs asarguments.36 The interaction of these general sub-theories is what compels the single NP in(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 languagesmakes entirely correct predictions for St’át’imcets, which exhibits properties of both radicalhead-marking languages like Mohawk and lexical argument languages like English.654.3. Parallelism4.3.1. TransitivesHaving derived the effects of the one-nominal interpretation law, most of the effects ofparallelism 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 theparallelism 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 (eithertopic of the discourse or non-topic). In addition, there is a preference for bothcoreferential elements to fulfill the topic of the discourse function.The effect of parallelism is illustrated below in (130), which has the context (129). Theseexamples 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‘Bill1 and John met each other. They’re enemies.’(130) áts ‘x-en-as nilh s-qvlqvl-ts-mIn ‘-assee-tr-3erg conj nom-bad-mouth-appl-3sg. conja. ‘he saw him, and then {he1 swore at him/*he swore at him}’b. ‘he saw him, and then {hej swore at him/*he1 swore at him}’(AA, GN, RW 1595)Recall from the previous section that because null pronominals are bound variables (which donot have independent reference), they are disfavored discourse-initially; rather, they must bebound 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 nullpronominal subject in each conjunct, in accord with the parallelism constraint. The nullpronominal that is assumed to be in subject position is not itself a topic, but instead takes its66reference from the discourse topic because of the unmarked correlation noted above betweensubjects and topics.Before examining the structure of this sentence, it is worth considering whether theeffects 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 toviolate the ban on vacuous quantification, it follows that the pro subject in both conjuncts of acompound sentence like (130) will be bound by the same topic in Spec/Top-P (which is thehighest phrase in any sentence), in accord with the topic constraint. Parallelism is thereforenot the result of a principle like (128); instead, the effect of parallelism is derived by theinteraction of the topic constraint with the requirement that pro be bound, and the fact that theuniverse of discourse contains only two persons, Bill and John.37 Consider how the topicposition and the coindexation between subject and topic account for the parallelism exemplifiedin (130). This sentence has the structure in (131)—as suggested by Henry Davis (p.c.)—witheach conjunct assumed to be in an asymmetrical relation with respect to each other. Subjectsare shown in their S-structure position (Spec/IP) for the sake of clarity, though §4.4 willdemonstrate that transitive subjects are base-generated in Spec/VP and raised at S-structure inorder 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—especiallydouble object constructions—in order to determine whether they provide evidence for aparallelism constraint that is independent of the topic constraint.67Top-P(131)VP NP XA/ \ pro1 nilh VP NPáts’x-en-as Projsee-tr-3erg profs-qvlqvl-ts-mIn’ -as projnom-bad-mouth-appl-3sg. conjThe conjunction nilh is labeled simply as a head X, since its exact category is neither easy todetermine 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 theconjunction 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 analysisaccommodates 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). Ina structure like (131), the conjunct to receive non-subordinate inflection (1P) is the one that isnot a complement of the conjunction nil/i, while the non-initial conjunct that is selected by nilh(1P2) receives subordinate inflection.Returning to the facts of (130) that the structure in (131) is intended to explain, there isa null topic in this structure, represented by e, which is in Spec/Top-P. The referential indexof the null topic is chosen from one of the topics in the discourse (129). This coindexation68occurs at LF (Huang 1984: 550), and it resembles covert left-dislocation, corresponding toEnglish S-structure left-dislocation in a sentence like Sam1, he1 swore at Fred. Regardless ofwhether the null topic in (131) is coindexed with Bill or John from the discourse (129), notethat the subject in Spec/IP of each conjunct is coindexed with e in accord with the topicconstraint (115). Huang (1982: 359-360, §5.4.1, 444 n. 14; 1984: 542-543) identifies thesame 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 locallyA’-bound by the null topic. The same analysis is adopted here. Huang assumes moreover thatthe variable is the trace of the moved topic. However, the null topic is represented heresimply as a base-generated empty category that must be coindexed with a pro variable in ordernot to violate the ban on vacuous quantification (Chomsky 1982: 11-13). This coindexation iscompelled because pronominals are not deictic, and hence must be topic-bound. Even if thetopic were a null operator that had been moved from an A-position, it would still need to findan antecedent from the discourse.38In a sentence in which there is an overt subject and object, the topic is not taken fromthe discourse, since the sentence already has a subject that may move to the topic position atLF. Consider the following sentence, which is repeated from (88) of chapter 3 (p. 45).(132) áts’x-en-as nilh s-qvlqvl-ts-mfn’-as kw-s John kw-s Billsee-tr-3erg conj nom-bad-mouth-appl-3sg. conj det-nom John det-nom Billa. ‘he1 saw him, and then {Bill swore at John/*John swore at Bill1)’b. *‘h saw him, and then {Bill swore at John/John swore at Bill1}’(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 VSOorder. The subject Bill in the second conjunct will therefore control the reference of thesubject of the first conjunct.38 See Cinque (1990: §3) for motivation for distinguishing pronominal variables frompure variables (i.e., traces of moved operators).69Before examining the structure of this sentence, it should be noted that the site of overtnominals in radical head-marking languages is a subject of debate: it has been claimed thatovert nominals appear in adjoined positions and are either coindexed with pronominal affixes(Jelinek 1984) or with empty categories in argument positions (Baker 1991). Neither approachis 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 overtpossessor NPs. This fact can only be explained if overt nominals occupy argumentpositions.39 Davis’s and Matthewson’s structure is assumed here, and so the overt nominals of(132) appear in argument positions in the following structure:(133)Ipi xPVP NP X 1P2_________________pro nilh v NPáts’x-en-as prosee-tr-3erg Bills-qvlqvl-ts-mmn’-as kw-s Johnnom-bad-mouth-appl-3sg. conj det-nom JohnAt LF, the subject Bill from 1P2 moves to Spec/Top-P, and it is then coindexed with the prosubject in 1P in accord with the topic constraint (115). This coindexation is essentially LFtopicalization, resembling English S-structure topicalization in a sentence like Fred1, Sallybelieves t to be a fool. The LF representation is as follows:Gardiner (1993: §5) reaches the same conclusion for overt nominals inSecwepemctsfn.70Top-P(134)Bi11 IPIpi xPVP NP X 1P2_________________prof nilh VP NPáts’x-en-asprosee-tr-3ergs-qvlqvl-ts-mfn’-as kw-s Johnsnom-bad-mouth-appl-3sg. conj det-nom JohnThe objects are free to corefer with each other, as indicated. The same analysis extends to(89) of chapter 3 (p. 45), which differs from (132) (= 88) only in having the overt subject andobject in the first conjunct rather than in the second.A question concerning (134)—given that XP is in an adjoined position—is whether LFextraction from XP violates the Condition on Extraction Domains (Huang 1982: §6.4), whichforbids extraction from a domain that is not properly governed. The resolution of this questionis only a technical matter, since it will be seen below in (135) and (142) (pp. 72, 76) thatextraction is indeed permissible only from the non-initial conjunct. See also the discussionpertaining to (146) below (p. 78) for one manner of addressing this issue. Note moreover thatextraction of Bill in (134) cannot be regarded as an across-the-board violation, since at LF, thetopic Bill obligatorily binds an empty NP in each conjunct: pro in 1P, and trace in 1P2. Thebinding relation in each case is an A’ relation.Parallelism obtains in sentences that have only null NPs, and in sentences in whichthere are two overt NPs in a single conjunct. In all of the cases examined above, the subject ismapped to topic, and because there is only one binder (the topic in Spec/Top-P), it must bindtwo elements, in accordance with the topic constraint. The sentences that remain to beexplained are those that do not exhibit parallelism; they differ from the former sentences inthat one of the conjuncts contains a single overt NP. Recall from chapter 3 (pp. 48 ff.) that71sentences having a single overt nominal in each conjunct provided evidence for constraintranking, since the one-nominal interpretation law was respected at the expense of violating thelower-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 Billsee-tr-3erg det-nom John conj nom-bad-mouth-appl-3sg. conj det-nom Billa. ‘he1 saw John, and then (i) hej swore at Bi111/(ii) *Bill swore at him’b. *‘John 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 ofthe 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 ownconjunct:(136)Ipi xPVP NP X 1P2__________________pro nilh NPáts’x-en-as kw-s Johnsee-tr-3erg det-nom John pros-qvlqvl-ts-mmn’-as kw-s Billnom-bad-mouth-appl-3sg. conj det-nom BillBecause there are only two persons in the universe of discourse, it is not possible to leave theovert NPs in situ at LF by coindexing the topic with some preceding NP that might have had adistinct reference from John and Bill. Nor is there an overt subject that is available formovement to Spec/Top-P. Null pronominals are non-referring, as discussed above, and soneither of the pro subjects may raise to topic. Moreover, coindexation of a null topic with oneof the overt NPs, as below, is also not possible:72* Top-P(137)IPIpi xPVP NP X 1P2_________________pro nilh NPáts’x-en-as kw-s Johnsee-tr-3erg det-nom John pro• s-qvlqvl-ts-mfn’-as kw-s Bill1nom-bad-mouth-appl-3sg. conj det-nom BillWhatever index is borne by the null topic in (137), a Condition C violation will result in oneof 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 moveone of the overt object NPs to topic position, violating the topic constraint. In the followingstructure, the object Bill from 1P2 moves:* Top-P(138)Bill1 IPIP1 xPVP NP X 1P2pro1 nilh VP NPáts’x-en-as kw-s Johnsee-tr-3erg det-nvm John pro1s-qvlqvl-ts-mfn’ -as t1nom-bad-mouth-appl-3sg. conjThe subject pro of 1P is coindexed with the topic, Bill. Under the topic constraint (115), thesubject of the other conjunct is similarly indexed with the topic, exactly as in (131, 134)73above. However, this coindexation results in a strong crossover configuration in 1P2, since thetrace of Bill is a variable, and the closest binder is not its antecedent Bill in topic position, butrather the A-binder pro in SpecIIP2. Condition C therefore excludes (138), since the trace ofBill is not A-free. (135) must instead have the following structure, with pro in Spec/1P2disjoint in reference from the trace that it c-commands:Top-P(139)Bill IP------Ipi xPVP NP X 1P2pro1 nilh VP NPáts’x-en-as kw-s Johnsee-tr-3erg det-nom John___________________prods-qvlqvl-ts-mf& -as t1nom-bad-mouth-appl-3sg. conjThe 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 wasdescribed in chapter 3 by ranking the parallelism constraint below the one-nominalinterpretation law; recall tableau (99) on page 49. In the analysis here, though, the nonparallel reading is derived by a conspiracy between the one-nominal interpretation law, whichforces the single overt NP in each conjunct to remain within VP, and Condition C, whichrequires that the trace of the object that is moved to topic position be disjoint from the subjectof its clause.A strong crossover configuration is similarly reproducible in English by topicalization(Lasnik and Uriagereka 1988: 154):(140) a. who does heji* like tb. John, heji* likes t74Both the wh-trace in (a) and the trace of the topicalized NP in (b) are variables, hence must beA-free. English S-structure topicalization therefore mirrors St’át’imcets LF topicalization—aswas already noted with respect to (134), which had an overt subject and object in the secondconjunct.Apparently, the single overt NP John in the first conjunct of (135) may instead raise totopic 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)John. IPIpi xPVP NP X 1P2A/ pro1 nilh VP NPáts’x-en-as tsee-tr-3erg______________________projs-qvlqvl-ts-mIn’-as kw-s Billnom-bad-mouth-appl-3sg. conj det-nom BillIt 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 raisedto 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 asingle corresponding LF representation.4° In fact, LF-raising of an object is possible onlyfrom the non-initial conjunct 1P2, although this can be ascertained only by examining asentence having a single overt NP in the first conjunct, but no overt NP in the secondconjunct. 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., wellformed) derivation for any given sentence.75(142) áts’x-en-as kw-s John nilh s-qvlqvl-ts-mIn’-assee-tr-3erg det-nom John conj nom-bad-mouth-appl-3sg. conja. ‘he1 saw Johns, and then (i) he1 swore at him/(ii) *hj swore at him1’b. *‘John saw him1, and then {hej swore at him1/he swore at him}’(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—isimmediately excluded. The non-parallel interpretation (142a.ii) is excluded because it wouldrequire the second conjunct to violate the topic constraint (115). Since there are no other overtNPs in this sentence that can move to Spec/Top-P (John is excluded, as will be seenpresently), the topic is taken from the discourse, and so the empty topic is coreferent with Billfrom the context (129). This sole grammatical reading (142a.i) therefore has the followingrepresentation:Top-P(143)e1 IPIpi xPVP NP X 1P2___________________pro1 nilh VP NPáts’x-en-as kw-s Johnsee-tr-3erg det-nom John___________________pro1s-qvlqvl-ts-mfn’ -as prosnom-bad-mouth-appl-3sg. conjCrucially, the NP John in the first conjunct may not map to topic position. Such a structure,shown below, violates the topic constraint (115) in 1P, and corresponds to the ungrammaticalreading in (142a.ii):76* Top-P(144)John. IPVP NP X 1P2A/ \ pro1 nilh VP NPáts’x-en—as tsee-tr-3erg______________________prods-qvlqvl-ts-mmn’ -as pro1nom-bad-mouth-appl-3sg. conjThe generalization illustrated by the contrast between (135) and (142) is that an object NP maybe extracted to topic position—violating the topic constraint (115)—only if it is at the peripheryof the sentence. (Stated differently, the VP-internal overt nominal may escape the nuclearscope and raise to topic only if it originates in the non-initial conjunct.) The validity of thisrestriction is further supported by the data below in (148), and it is also motivated by thecross-linguistic observation that topics typically appear at clause-peripheries, rather than inembedded positions (e.g., As for Fred, I like him). The correlation seems natural, in thatsubjects 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 isadjoined to 1P. Consider the following sentence, in which parallelism obtains in an adjunctclause only for the subject:(145) ts’um’-qs-án’-as lh áts’x-en-as ta sqáycw-a s-Marylick-nose-tr-3erg hyp see-tr-3erg det man-det nom-Marya. ‘she1 will kiss him if Mary1 sees the man’ (topic. . . topic)b. *‘she1 will kiss him if the man sees Mary1’ (topic. . . non-topic)c. * ‘he will kiss her if Mary sees the man’ (non-topic. . . topic)d. *‘he will kiss her if the man sees Mary1’ (non-topic. . . non-topic)(GN, RW 505; Matthewson 1993a: 29; Matthewson et al. 1993: 230;cf. BF, RW 567)77Readings (b, c) are not parallel. The interesting contrast is between (a) and (d); both areparallel, but only the former has the peripheral NP Mary acting as both the subject and thetopic, 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 arenot especially strict in requiring VOS order, the contrast between (a) and (d) provides furtherevidence that the topic is always the peripheral element in the structure.With this description in hand, it remains to be explained how an appropriate LFrepresentation is derived for such sentences. Reconsider (135), which had a single overt NP ineach conjunct, but apparently two well-formed LF representations (139, 141). It is now clearfrom the discussion of (142) that (141)—with the object raising from the initial conjunct—isnot 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 regardthe 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 thisscrambling as A’-movement. First, reconstruction is always from an A’-position, and there isevidence 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 itclearly holds at S-structure), it would be puzzling why LF movement of an object NP to topicposition (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 beexplained by the generalization that a phrase that has been moved to an A’ -position does notconstitute as strong a barrier for extraction as does a phrase in an A-position. As evidence forthis claim, consider the following contrast (from Fiengo et al. 1988: 89, §3):(146) a. *[vowel harmony]1,I think that articles about tj have been publishedb. ?[vowel harmony], I think that [articles about tj], you should read tj carefullyThe contrast between (a) and (b) shows that the effects of the Condition on ExtractionDomains (CED) are weakened if extraction occurs from a topicalized constituent (which is in78an A’-position). In (a), extraction of vowel harmony directly from the embedded subjectarticles about vowel harmony violates the CED, since subject position is not properlygoverned. 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 furthertopicalized. The generalization is stated by Fiengo et al. (1988) as in (147):(147) x is a barrier only if it is not an A’-binderBecause articles about vowel harmony is an A’-binder in (146b) (it A’-binds tj), it is not abarrier, and so vowel harmony may be further extracted. See Fiengo et al. (1988) for anattempt 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 theextraction in the English sentences (146). The non-initial conjunct in St’át’imcets is probablybest regarded as an ungoverned domain, since—like an adjunct—it is in an adjoined position(specifically, the XP in which it is contained is adjoined to the initial conjunct 1P). Thesubject position is English is similarly ungoverned. In both languages, A’-movement(topicalization in English and scrambling in St’át’imcets) therefore facilitates furtherextraction, which would not otherwise have been allowed if it had occurred from an in situungoverned domain.41The structural approach to coreference across conjuncts has a particular advantage overthe 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 itto simultaneously be the topic of the sentence while showing the usual one-nominalinterpretation effect within its own conjunct.41 A test that could help to confirm whether a scrambled NP had indeed become thediscourse topic (suggested by Henry Davis, p.c.) would be to include a ‘post-context’ having apro subject after the sentence of interest. More research on the properties of scrambling inSt’át’imcets that might confirm or refute this analysis remains to be done.79(148) áts ‘x-en-as nilh s-qvlqvl-ts-mIn’ -as kw-s Billsee-tr-3erg conj nom-bad-mouth-appl-3sg. conj det-nom Billa. ‘he1 saw him, and then (i) he swore at Bill1/( i) *Bjll. swore at him’b. ‘he saw him, and then (i) hej swore at Bill1/( i) *Bill swore at him’(AA, GN, RW 1594)The interpretations in (a.ii) and (b.ii) are not allowed, because they would require that thesingle overt NP Bill be the subject of the second conjunct, violating the one-nominalinterpretation 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 asthe object. Because there is no disjoint, overt nominal elsewhere in the sentence that can bemapped to topic, the null topic is coindexed with John from the context (129). This topic isthen coindexed with the pro subject of each conjunct, satisfying the topic constraint (115).Top-P(149)ej IPIpi xPVP NP X 1P2/ pros nilh VP NPáts’x-en-as pro1see-tr-3erg______________________pross-qvlqvl-ts-mIn’ -as kw-s Bill1nom-bad-mouth-appl-3sg. conj det-nom BillIn a sentence like (135), which has two overt NPs—one in each conjunct—there is a singleinterpretation for the sentence. There is also only one possible reading in a sentence in whichthere 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 mayserve as the topic. The latter interpretation is (148a.i), in which Bill is the topic of thesentence at the same time that it is the object of the second conjunct. After Bill moves to80Spec/Top-P (perhaps after having scrambled to an A’-position, as suggested above), it iscoindexed with the subject of 1P, and the object of this conjunct is free to take the referenceof John from the context (129), as must the subject of 1P2. The well-formed structurecorresponding to (148a.i) is therefore as follows:Top-P(150)Bill1 IPIpi xPVP NP X 1P2/ prof nilh VP NPáts’x-en-as prossee-tr-3erg______________________projs-qvlqvl-ts-mfn’-as t1nom-bad-mouth-appl-3sg. conjCrucially, the subject of 1P2 may not corefer with Bill, since this would create a strongcrossover configuration in 1P2: the trace of Bill would be locally A-bound by the pro subject,resulting in a Condition C violation.* Top-P(151)Bill1 IPIpi xPVP NP X 1P2/ pro1 nilh VP NPáts’x-en-asprosee-tr-3erg pro1s-qvlqvl-ts-mIn’ -as t1nom-bad-mouth-appl-3sg. conj81Whatever principle allows the overt NP in the second conjunct of (135)—which has a singleovert 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 clauseand/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 asscrambling of the peripheral object to an A’-position (e.g., adjoined to XP) before mapping itto topic position (in the case of objects). Other approaches are conceivable, but it is clear thata 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). Aswas noted in § (p. 50), the tableau for this sentence in (101) uniquely identified asgrammatical the reading with the overt object as non-topic, incorrectly dispreferring thereading in which the object is topic. The structural account proposed in this chapter, on theother hand, predicts exactly the attested ambiguity.The interpretations that are allowed across transitive conjuncts having variouscombinations of overt nominals are summarized in the chart below, in which pro represents anull pronominal, and NP represents an overt R-expression.(152) Topic 1PConstraint___________ ____________________ ________(115’) subject object subject 1 object(130) pro1 prod pro1 prod(132) pro1 prod NP1(89) NP1 NP pro1 proj(135) * pro1 NP prod NP1(142) pro1 NP pro1 proj(148a.i) * pro1 prod prod NP,(148b.i) pro1 pro1 proj NPThe structures that violate the topic constraint are (135) and (148a.i). The common propertyof these structures is that they have a single overt NP in the non-initial conjunct (1P2). Notethat the topic constraint is only violable in this non-initial conjunct. Sentence (89), which hastwo overt NPs in the first conjunct, obeys the topic constraint in all conjuncts; note moreover82that LF extraction of the subject to Spec/Top-P must be allowed from the initial conjunct, andthat this is presumably allowed because the overt NP subject is the highest element in thatclause.For completeness, a table of the excluded interpretations for the same set of sentencesis given below:(153) Topic 1PConstraint__________ ____________________ ________(115) subject object subject object(130) * pro, pros prod pro1(132) * pro1 prod NP NP,(89) * NP, NP pros pro1(135) pro1 NP3 pro, NP(142) * pro1 NP prod pro1(148) pro1 pro1 pro1 NPParallel interpretations for sentences having a single overt NP in the second conjunct areungrammatical, despite satisfying the topic constraint in that conjunct.4.3.2. IntransitivesSection 3.3.2 showed that parallelism—the preference that subjects corefer—does not holdstrongly across conjuncts if one of them is intransitive.42 One case where it does hold formost speakers, though, was (103), repeated here as (155), and which has the context (154):(154) wa7 k’ál’-em kw-s Johnprog wait-intr det-nom John‘John is waiting’(155) p’an’t kw-s Bill nilh s-7áts’x-en-asreturn det-nom Bill conj nom-see-tr-3sg. conj‘Bill1 returned and (i) he saw him/(ii) *he. saw him’(AA, ON, RW 1586)42 The effect of the semantics of intransitive predicates in determining coreference hasnot been investigated, though it should be. In an examination of Italian, for example,Calabrese (1990: §6) shows that unaccusative verbs do not have consistent focuscharacteristics, these properties apparently being determined by whether or not the verb isstative.83Note that the subject of the first conjunct is preferably coreferent with the subject of the secondconjunct (the interpretation in (i)). If the overt nominal subject of the intransitive firstconjunct occupies Spec/1P1, coreference between the subjects in each conjunct followsstraightforwardly. The overt nominal Bill moves to the topic position, which is then coindexedwith the pro subject of 1P2, as schematized below:Top-P(156)Bill IPIpi xPVP NP X 1P2A/ \ t1 nilh VP NPp’an’treturn pro1s-áts x-en-as prodnom-see-tr-3sg. conjLike sentences having two transitive conjuncts (4.3. 1), sentences having at least oneintransitive conjunct allow LF movement of an NP from the initial conjunct to Spec/Top-P aslong 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 5-structure (Spec/IP) gives the correct result for (155), but not always for (111), repeated forconvenience as (157):(157) qvlqvl-ts-mfn ‘-as s-John s-Bill nilh s-qwatsátsbad-mouth-appl-3erg nom-John nom-Bill conj nom-leavea. ‘Bill1 swore at John, and then he1pj left’b. ‘John swore at Bill, and then hej,* left’(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, the84subject is mapped to topic, and this topic corefers with the subject of the intransitive in thesecond conjunct. In interpretation (b), the NP John would have to be scrambled and/ortopicalized in order that VSO order could be derived in the first conjunct, and thisrepresentation then apparently requires that the derived subject John corefer with the prosubject of the intransitive clause. Of particular interest, though, is the reading in (a) markedby ?, which is meant to indicate not that this reading—which is not parallel—is marginal, butthat its grammaticality has varied across consultants and across elicitation sessions, and hasbeen difficult to settle with certainty. This variability is not surprising, since St’át’imcetsexhibits ergativity, which entails that intransitive subjects pattern with direct objects withrespect to certain processes. The variability seen with sentences like (157) suggests that theintransitive subject functions sometimes as the topic, but at other times as the non-topic. Asimple way to explain the variability is to base-generate intransitive subjects within VP. Notethat intransitive subjects get the case of an object (absolutive), hence it need not raise at Sstructure in order to get ergative case in Spec/IP, as must a transitive subject. Intransitivesubjects may remain in situ at LF when functioning as non-topics, and so the topic constraintwill 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 topicconstraint and consequently must corefer with the subject of the initial conjunct.43Returning to (157a), the parallel interpretation has the following structure, with theovert transitive subject Bill from 1P moving to topic position at LF and then being coindexedwith the intransitive pro subject in 1P2.‘ It is uncontroversial for the object to receive case within VP, the problematic GF ofcourse being the intransitive subject. No specific claim will be made concerning whetherintransitive subjects get case in Spec/VP, as sisters of the verb, or perhaps through a relationwith some other functional projection outside of the VP; note that the last possibility seemsunlikely, since absolutive agreement (unlike ergative agreement) is not overtly marked on thepredicate. It would be desirable if transitive and intransitive subjects shared the same positionat D-structure, since some aspects of the syntax of St’át’imcets—such as the formation ofimperatives and jussive complements—are clearly accusative.85Top-P(158)Bill1 IPIpi xPVP NP X 1P2t nilh VP NPqvlqvl-ts-mI&-as s-Johnbad-mouth-appl-3erg nom-John pro1s-qwatsátsnom-leaveFor this particular reading, then, the intransitive subject is in Spec/1P2,and so the structure of(158) is accusative, since the transitive subject in the first conjunct is also in Spec/IP. In orderto derive the non-parallel reading in (157), however, the intransitive subject must remain inSpec/VP (but see the previous footnote), because in this lower position it need not becoreferential with the topic Bill, as the following structure shows:Top-P(159)Bill1 IPIpi xPVP NP X 1P2_____t. nith VPqvlqvl-ts-mfn’-as s-Johnbad-mouth-appl-3erg nom-Johns-qwatsáts prodnom-leaveThere is no VP-external subject in 1P2 that needs to be coindexed with the topic, and the topicconstraint (115) is also satisfied, since the pro inside VP in the second conjunct does notcorefer with the topic Bill; instead, it is free to corefer with the other non-topic, John. In86summary, the two readings of (157a) therefore share LF movement of the subject Bill from thefirst conjunct to topic position. Their different interpretations are derived from the position ofthe intransitive subject, which in a split-ergative language like St’át’imcets may appear inSpec/IP when behaving accusatively, and within VP when behaving ergatively. As notedabove, perhaps the semantics of the intransitive predicate can account for the variation, thoughthis remains to be investigated.In conclusion, the facts concerning coreference across conjuncts in St’át’imcets may beexplained by mapping grammatical functions to discourse functions at LF. Specifically, thegrammatical subject is canonically mapped to topic position (Spec/Top-P) in accord with thetopic constraint (115) (although the topic constraint is violable in the second conjunct ofsentences that have a single overt, peripheral NP). In so far as coreference across conjuncts isdetermined by this mapping of grammatical functions to discourse functions (Reinhart 1981and Erteschik-Shir 1993), coreference in St’át’imcets may be regarded as having an ergativepivot, since it has been seen that the topic constraint is a constraint on the mapping of ergativeNPs (transitive subjects) to topic position. Additionally, St’át’imcets displays a split betweentransitive subjects (ergative) and intransitive subjects (absolutive); specifically, absolutivesubjects display variability in the way they satisfy the topic constraint. Note that in anaccusative language like Italian (Calabrese 1990), there is a similar split between unaccusativesubjects on the one hand, and transitive and unergative subjects on the other. This isinteresting, since in St’ at’ imcets, all intransitive subjects are unaccusative, in the sense thatthey get the case of D-structure objects. It remains a topic for investigation to determine inwhat further respects intransitive subjects behave alike in both types of languages.The structural asymmetries proposed above may be fruitfully extended to explain theergative/absolutive extraction asymmetries in relative clauses, the subject of the next section.4.4. Extraction in relative clausesChapter 2 established that there is an ergative/absolutive asymmetry in relative clauseformation in St’át’imcets. Specifically, pronominal morphology corresponding to the87relativized constituent is obligatorily retained in ergative-centered RCs, but obligatorily gappedin absolutive-centered RCs. Consider the 0-centered RC in (58), repeated here as (160), andthe S-centered RC data of (44, 50), consolidated below as (161):(160) nilh wi s-John müta7 s-Bill i lhkiinsafoc pl.det nom-J conj nom-B pl.det see-tr-(-Ø/3p1.obj]-lsg.conj-det now‘it was John and Bill that I saw(LT 2338, 2339)(161) tsukw t’u7 s7ents [ ti tsIcw{ø/*_an/*_as}a]only part lsg. emph det go- {-ø/lsg. conj/3sg. conj)-det‘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 thesestructures suggests that RCs in St’át’imcets do indeed involve extraction of a constituent fromthe position that is relativized. Because this extraction is not seen in the form of an overtrelative 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’imcetssentence has an overt, plural determiner wi). The 0-centered relative clause forms the residueof the cleft construction; it is also a complete DP because of the discontinuous, pluraldeterminer i. . . -a. The head of the NP that is selected by this determiner (NP1) is empty(though coreferent with John and Bill) and omitted from the diagram for conciseness, asindicated by the dashed line between DP and CP. The cleft residue is therefore morespecifically a headless RC, and so (160) might be more literally translated as ‘it was John andBill, the ones that I saw’. Further discussion follows the diagram:88‘P(162)XP DPX DP. D CPnil,,Z prof IPwi s-John mita7 s-Billp1. det nom-J conj nom-Bpl.det...-detI VP prodlsg.conj ats’x-en t1see-trBecause the target of relativization is absolutive, the trace of the null operator (tj) receivesabsolutive case from the predicate inside VP. The operator itself is assumed to be a nullpronominal pro in Spec/CP. Because pro is non-referring and never used deictically, it needsan antecedent; following Browning (1987) and Cinque (1990: 160), the reference of a prooperator is identified via agreement between CP and the head of the relative clause (here, thenull 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 subjectmorpheme -an in I(nfl). As indicated by the arrows, the predicate raises to I (in order to takethe lsg subject suffix -an) and finally to D, where it appears with the discontinuousdeterminer. Intermediate steps in the raising of V as are necessary to satisfy the HeadMovement Constraint (HMC) (Chomsky 1986: 71) are also assumed, but omitted from thediagram for clarity.A similar derivation applies to the S-centered RC in (161). The constituent tsukw t’u7is a focus marker analogous to nilh (van Eijk 1985: 279, §38.5; 1987: 72), and so it appearsunder X, selecting the lsg emphatic pronoun s7ents. The intransitive subject pro is base89generated within VP, per the discussion in the previous section. After it moves to Spec/CP (itis an operator), absolutive case is assigned inside VP to the trace of the operator (tj):(163)XPDPX DP D CPA A/ \ / \ ti...-a prof IPtsukw t’ u7 s7entsdet ... -detonly part lsg. empVPAtsicw tgoThe predicate tsicw ‘go raises to I, then to C, deriving a complete S-structure. Intermediateraising 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_______ _______ti tsfcw -a pro1 IPtsukw t’u7 s7ents det go-detonly part lsg.empAti tiThe topic node has also been omitted from the above structures, since it is irrelevant.90The 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 morphologycorresponding to the relativized constituent must be retained, or else the topical object marker-tall must appear on the predicate.44 Recall (38) from chapter 2 (p. 22), repeated here:(165) áts’x-en-as [ ti sqáycw-a [ ti tup*unh..*{_tali/as}(h)a s-Bill ] jsee-tr-3erg det man-det det hit-tr- (TO/3sg. conj]-det nom-Blil‘he saw the man that hit Bill’(ON 141; ON, RW 218; LT 2581)All consultants45 require either -tall or -as in ergative-centered RCs, and so the appearance ofboth morphemes needs to be explained. The conclusion of chapter 2 (2.3) emphasized thatthis sort of restriction on relativization of ergatives is common in many languages outside ofSalish, and so it should be possible to give a principled account of this ergative/absolutiveasymmetry. An explanation is indeed possible by adopting a version of the VP-internal subjecthypothesis (Koopman and Sportiche 1991 and the references cited there). It has already beenassumed that intransitive subjects are generated within VP. Suppose that ergatives likewiseoriginate in VP, specifically in Spec/VP (for the sake of concreteness, though they could justas easily be in some other projection selected by I, depending on how one analyzes thetransitive affix), and that they move to Spec/IP in order to get ergative case from the ergativemarker -as in I—the absolutive case that is available within VP being needed by the object. Asentence like (5), repeated here as (166), therefore has a structure like (167) after case-assignment to the subject ta smiflhatsa ‘the woman’ that has raised to Spec/IP:(166) áts’x-en-as ta sqáycw-a ta smülhats-asee-tr-3erg det man-det det woman-det‘the woman saw the man’(RW 48; Matthewson 1993a: 2-3)‘ The passive morpheme -m is sometimes, but not always, interchangeable with thetopical object marker -tall. The former appears to occur in more restricted environments thanthe latter, but the reasons for the variation remain unclear.‘ Except ON (Henry Davis, p.c.).91‘P(167)I’ NPI VP ta smülhats-a1det woman-det-as V’ NP3e>/\ats’x-en ta sqáycw-asee-Ir det man-detV-to-I raising derives the S-structure, and the subject is mapped to topic at LF by moving toSpec/Top-P:Top-P(168)NP IPta smülhats-a I’ NPdet woman-detI VP tts’x-en -as V’ NPsee-tr-3e,5/,,,//”\t ta sqáycw-adet man-detReturning to the variant of the ergative-centered RC in (165) with the 3sg subjectmarker -as, the representation of the complex NP after operator movement would be as in(169). The null operator originates in the VP-internal subject position (marked below by ti),from which it moves to Spec/IP in order to receive case from -as in I (the predicate’sabsolutive case having been assigned to the object, Bill). The operator ultimately lands inSpec/CP, the A’-position from which it binds the variable in Spec/IP (the case-marked t2):92(169)Aii sqáycw-a D CPdet man-detti ... -a pros C’det ... -detC IPI’ NPI VP t21-as NP3sg.conjti itup-un’ s-Billhit-tr nom-BillThere is thus an explanation for why relativization of ergative requires that the personmorphology 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 soneither could function as a variable. The pro in Spec/CP in such a structure would violate theban 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—andbecause its trace at S-structure receives absolutive case from the predicate, a well-formedoperator-variable chain is created. In this analysis, then, both ergative- and absolutivecentered RCs have extraction of an operator. Note, incidentally, that Jelinek’s (1984) analysisof 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 actualarguments). In this account, however, which does not regard relativization as having asyntactically ergative pivot (since all RCs have extraction of an operator), the ergative93pronominal 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 thefollowing contrast, in which a wh-word is extracted from the subject position of an embeddedclause (Lasnik and Uriagereka 1988: 90:(170) a. who1 [is it likely [tj will win the race?]]b. *who1 [is it likely [tj to win the race?]]The embedded subject position is case-marked in (a)—because of the finite Infi in theembedded clause—but not in (b), which has an infinitival embedded clause. Both sentencesare otherwise identical, and so the ill-formedness of (b) must be due to the fact that the whtrace lacks case.The structure in (169) does not yet have a topic. The head of the RC, ti sqáycwa ‘theman’, may not lower to Spec/Top-P, downward movement being generally prohibited. Normay the null operator in Spec/CP raise to Spec/Top-P, since pro—being non-referential—isexcluded from topic position. (The null pronominal pro is as a bound variable, hence itself isalways topic-bound, a point that was illustrated above (p. 66) for both St’át’imcets (130) andItalian (125-126).) An additional reason that the relativized ergative is excluded from servingas the topic is that the head of the RC is the focus of a subordinate predication (Erteschik-Shir1993: 35-39, §6.2.2); since the RC head ti sqáycwa ‘the man’ is coreferential with the ergativeoperator—the person understood to be hitting Bill in (165) is the man—moving the operatorinto topic position would illegitimately entail that in the subordinate predication the focus andthe topic would have the same referent. Another option is that the object, Bill, becomes thetopic in (169); this possibility is forbidden by the topic constraint (115), however, and in anycase it will be reserved for the variant of (165) that contains the topical object marker -tallrather than the 3sg subject marker -as. (As will be seen below, sentences with -tall signalfocus of the subject rather than the object.) Instead, it will be assumed (following ErteschikShir 1993) that the entire clausal portion of the relative clause in (169) denotes the subordinate94topic. The only remaining step in this derivation is therefore V-to-I, I-to-C, and C-to-Draising, as indicated below by the arrows:DP(171)Ati sqáycw-a D CPdet man-detti...-a pro1 C’det ... -detC IPI’ NPi v t21-as• NP3sg. conjtup-un’ s-Billhit-tr nom-BillThe restriction on extraction/relativization of topics is not unique to St’át’imcets. Thesame restriction is particularly salient in Japanese, for example, since topics are overtlymarked by the morpheme -Wa, an example of which is given in (172). A homophonousmorpheme indicates contrastive focus, however—as shown in (173), in which the first NPbears the topic -Wa, and the second NP bears the contrastive-focus -Wa. These and subsequentJapanese sentences are from Uechi (1994).(172) Nagano-wa hito-ga ooiNagano-top people-nom many‘speaking of Nagano, there are many people there’95(173) Nagano-wa hito-wa ooiNagano-top people-contr many‘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, theconstituent that answers a wh-question is the focus of the sentence. The reason for this, ofcourse, is that a wh-question has the communicative goal of seeking new information, and sothe answer to a wh-question comprises non-topical, or new, information. In contrast, the topicof 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 intuitionis syntactically realized in Japanese. Both RC heads and wh-words may be marked by thefocus 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-focusinterpretation is possible for the -wa-marked NPs (174b, 175c):(174) a. Tamako-ga sukina momoTamako-nom like peach(i) ‘the peach Tamako likes’ (wide focus)(ii) ‘the peach TAMAKO likes’ (narrow focus)b. Tamako-wa sukina momoTamako-nom like peach‘the peach which TAMAKO, not others, likes’ (contrastive focus only)(175) a. Kyoo-wa dare-ga kimasu ka?today-top who-nom come interrog‘as for today, who is coming?’b. *Kyoowa dare-wa kimasu ka?today-top who-top come interrog‘as for today, speaking of who, is s/he coming?’c. Kyoo-wa dare-wa kuru kedo, dare-wa kimas-en ka?today-top who-contr come but who-contr come-not interrog‘as for today, who is coming, but who is not coming?’96A similar paradigm can be constructed in English. The as for x construction carries topicalinformation; examples are the following:(176) a. As for Sally, she writes poems and short storiesb. As for donuts, Fred eats them for breakfastA topic may appear in a wh-question if the topic and the wh-word have distinct referents, andas long as the topic appears higher than the wh-word:(177) a. As for Sally, what does she write?b. *What, as for Sally, does she write?c. As for donuts, who eats them?d. *Who, as for donuts, eats them?However, an as for x construction may not be used to ask a wh-question directly, since the whword may not simultaneously be the topic and the focus of a sentence. The followingsentences are parallel to the Japanese sentence in (175b):(178) a. *As for who, (she) writes poems and short stories?b. *As for who, does she write poems and short stories?c. *As for what, Fred eats (them) for breakfast?d. *As for what, does Fred eat (them) for breakfast?These data are reminiscent of the topic/focus structure subordinate predication in the RCdiagrammed in (169). Just as the focused head ti sqdycwa ‘the man’ could not lower to topicposition (since the topic and focus would illegitimately have the same referent), neither maythe wh-word in (178) be both the topic and the focus of the sentence.To summarize, the obligatory retention of pronominal morphology corresponding torelativized ergatives in St’át’imcets can be explained by Case theory and by a structure inwhich ergative NPs occupy a higher position at S-structure than do absolutive NPs. The lackof a topic in these relative clauses probably represents a universal phenomenon, as suggestedby the Japanese and English data cited above. See also Schachter (1973) for instances of otherlanguages in which RCs and focus/cleft constructions resemble each other.97The final task is to explain the difference between an ergative-centered RC with -as asopposed to one with the topical object marker -tall. The discourse facts that need to becaptured in the syntax are described thus by Matthewson (1993a: 4-5):[The] parallelism between the object of an ordinary transitive clause and the subject ofa -tall clause extends to their discourse function; these are the slots where entities newto the discourse are introduced. Conversely, that entity which is already ‘underdiscussion’ and is the ‘topic of the discourse’ tends to occur in the subject position ofan 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 isassociated with the focus, pursuant to Diesing’s (1992: 9-10) Mapping Hypothesis or somesimilar 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 differentderivation is needed for those structures. The most sensible way to facilitate the requiredderivation is to base-generate -tall within VP, and to suppose that—like 3sg -as—it has its owncase to assign. The D-structure for the variant of (165) that contains -tall would therefore beas follows:(179)Ati sqáycw-a D IPdet man-detti...-adet ... -detI VPNPtup-un’-táli s-Billhit-tr-TO nom-Bill98The morpheme -tall may occupy a separate projection within VP—perhaps Agr-Object. Forthe sake of clarity, however, it is shown here as affixed to the predicate within VP, since thisaccount requires only that -tall occupy a projection lower than I. Within VP, there aretherefore two cases available: one from the predicate (which is received by the object, Bill, asusual), and the other from -tall, which is free to be assigned to Spec/VP. Because Spec/VP isa case-marked position, the operator pro can move directly from Spec/VP to Spec/CP. Thepredicate raises and ultimately lands in D, and the S-structure in (180) is derived:(180)Ati sqäycw-a D CPdet man-detti tup-un’-táli-ha pro1 C’det hit-tr-TO-detC IPI’I VPV’ NPt3 s—Billtjnom-BillThe operator pro may not move to Spec/Top-P for the same reasons that it could not do so inthe relative clause with -as in (169). However, because the object of a -tall clause is the slotin which presupposed information appears, the object Bill must raise at LF, landing inSpec/Top-P. From there, it A’-binds its case-marked trace. The LF representation wouldtherefore be as follows:99DP(181)Ati sqáycw-a D cPdet man-detti tup-un’-tá1i-ha NP CPdet hit-tr-TO-detpro1 C’s Bulknom-BillC IPIII VPV1 NPtj tktiSuch a derivation violates the topic constraint (115), since an NP under V’ (i.e., the trace ofthe object tk) is coreferential with the topic. However, this violation alone does not precludethe analysis outlined here, since the topic constraint (115) is intended to represent only thecanonical mapping of subject to topic. Another violation of the topic constraint was seenabove in (150) (p. 81), in which LF topicalization of an object was licensed by its peripheralposition in the clause.It remains to be explained why -tall forces the object to raise to topic. When thesubject is focused, as it is in -tall clauses (which include wh-questions, clefts, and RCs), itmust also be extracted. Perhaps because the VP in a -tall clause already contains a focusedconstituent (i.e., the trace in subject position), the object must move to Spec/Top-P at LF inorder that there not be two focused constituents in the same domain.The topic constraint can be violated also in English, as an intonationally prominentsubject is focused. This is exemplified below:100(182) a. Fred hates Billb. FRED hates BillSentence (a) with normal intonation has the unmarked mapping of subject to topic, in whichFred is the topic. In (b), though, with stress on the subject, Fred, this NP is interpreted as thefocus, not the topic.Clearly, a language may have various strategies for altering the subject-to-topicmapping, and -tail represents another such strategy that is available in St’át’imcets. Becausethere are two cases available to be assigned within VP in a -tali clause, the relative clauseexamined here could conceivably have the same representation at S-structure as at LF. Such astructure, shown above in (180), would resemble the one with -as in (169), in that the topic ofthe subordinate predication in (180) would be the clausal portion of the RC (followingErteschik-Shir 1993). In a structure like (181), though, the function of the topical objectmarker -tall is precisely to allow the object of a transitive clause to function as the topic withinthe 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 inSpec/VP, within the domain that is mapped to the nuclear scope. Because the nuclear scope isassociated with the focus, it is possible to explain why some speakers—notably RW (andprobably also LT, based on the limited elicitation with her regarding these data)—prefer andusually require -tall in sentences in which the subject has been focused (whether relativeclauses, clefts, or wh-questions). The head of an RC is the focus of the subordinatepredication, according to Erteschik-Shir (1993), and because this head corefers with theextracted operator, it would naturally be preferable if the variable was within the domain offocus (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 thestructure in (169), then, the variable t2 corefers with the focused RC head ti sqáycwa ‘theman’, but the variable itself will not be in the domain of focus, since it has escaped VP. Thisfact may be the reason why ergative extraction with -as instead of -tall is ungrammatical for101some speakers. The more general question concerns why focusing the subject with -talirequires it to be extracted; it is likely related to the topic constraint, though this question is leftto future research.This analysis explains the retention of ergative morphology and topic-focus structureonly with respect to relative clauses, and more specifically with regard to a subset of data. Itremains to be explained why some consultants (GN is one) can extract an ergative NP withneither -tali nor -as on the predicate. In the meantime, the present analysis offers a frameworkwithin which such additional data may be addressed. Further fieldwork will determinewhether this system is adequate to explain the additional facts while remaining a restrictive andcoherent theory.4.5. SummaryThis chapter has shown that the asymmetric behaviours of subjects and topics in St’át’imcetsmay be explained within the Principles and Parameters model of syntax (Chomsky 1981; 1982;1986; 1992; etc.). Radical head-marking languages and languages displaying ergativity havebeen integrated into this approach to syntax with limited success, since the model has beenbuilt mainly upon well described, accusative Indo-European languages. However, recent workon 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 Sportiche1991 and the references cited there) is beginning to reveal how typologically diverse languagesconverge structurally.Section 4.2 demonstrated that the effects of the one-nominal interpretation law may bederived by identifying the nuclear scope of a sentence as the structure below VP. Section 4.3showed that parallelism in coreference across conjuncts is the result of an LF principle—thetopic 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 byemploying Case theory to ensure that ergative NPs are higher than absolutive NPs at Sstructure. The topical object marker -tali appears when the subject is focused, and it moreover102requires that the subject be extracted. Most of the analyses in this chapter assume thatgrammatical functions are mapped to discourse functions at LF, and that this mapping isconstraint by the topic constraint.All of the constraints and asymmetries outlined in this chapter have appeared in theliterature in various forms. Although very specific, all appear to underlie a broader linguisticsystem, since they manifest themselves universally in discourse. It would be surprising to finda language, for example, that regularly marked a topic by repeating it throughout the discourseas an overt NP, or that had a device for questioning given information in the discourse. All ofthese discourse-based notions are amenable to a structural analysis. That this analysis applieswith equal validity to languages as genetically and typologically diverse as St’át’imcets,Italian, Japanese, and English—and others sharing their characteristics—strongly suggests thatit is capturing not merely language-specific features of sentence structure, but properties ofUniversal Grammar.103Chapter 5ConclusionThe principal contribution of this thesis has been to describe some hitherto uninvestigatedsyntactic properties of St’át’imcets, with a view to determining specifically to what extent thelanguage is syntactically ergative. Chapter 1 outlines some morphosyntactic characteristics ofSt’át’imcets, showing how it shares with other Salish languages the property of morphologicalergativity in its pronominal inflection. Dixon (1979) identifies relativization and conjunctionas two major processes in which languages manifest either syntactic ergativity or syntacticaccusativity, and so chapter 2 explores relativization and chapter 3 investigates coreference inconjoined clauses. Both processes appear to be ergative in certain respects.Relativization in St’át’imcets has the appearance of syntactic ergativity, since ergativeNPs 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 usualpronominal 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 thanabsolutive RCs. In contrast, all speakers have an obligatory gap corresponding to therelativized 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-nominalinterpretation law (Gerdts 1988) and by the topic constraint (Erteschik-Shir 1993), which givesthe effect of a parallelism constraint on discourse functions (Matthewson 1993a). Whencoreference is examined in terms of a mapping of grammatical functions to discoursefunctions, it might be regarded as behaving ergatively, since the topic constraint appliesspecifically 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, but104chapter 3 supports the conclusions of Kinkade (1989; 1990) and Matthewson (1993a) that theprinciples governing pro-drop and coreference are heavily discourse-based.Chapter 4 accounts for the asymmetries outlined in the previous chapters within thePrinciples and Parameters syntactic framework. The analyses there are tentative, and willsurely need to be refined as additional aspects of Stát’imcets are documented. Data fromSt’át’imcets have only recently been brought to bear on this syntactic framework (principallyby Davis et al. 1993, Gardiner et al. 1993, Matthewson et al. 1993, and in other references bythese authors), and the importance of fieldwork on this language cannot be emphasizedstrongly enough. The relative clause and conjunction data in this thesis represent a smallcontribution in this regard.Each chapter emphasizes the similarity of the St’át’imcets facts to those found in betterdescribed languages outside of the Salish family. St’át’imcets poses several problems forsyntactic theories, but these problems have many precedents, and this fact must only reinforcethe view that underlying the surface dissimilarity of genetically and typologically diverselanguages is a unique, shared linguistic system.105Appendix AMorphologyThe words and morphemes listed here are given in orthography, a key for which is in the nextappendix. An underline — indicates that the following morpheme is an enclitic; if the underlinefollows the morpheme, then it is a proclitic. Page references are to van Eijk (1985), whichshould be consulted for detailed information on St’át’imcets.Personal affixationPossessive affixes (p. 170)sg p11 n- -kalh2 -tsu (after lh, s) -lap-su_(elsewhere)3 -ts (after lh, s) -i-s_(elsewhere)Object suffixes (p. 171)1 3pl -tumc-al -ts-al -tumul3sg, 2sg/pl -tumc -ts -tumulh2 lsg -tumi(n) -tsi(n) -tumulh3sg/pl -tumi -tsi -tam-al’ap3 1 sg -tan-ilsg, 2sgIpl -Ø -wit3sg/pl -øsgsubject transitivizersuffix_________ ________I llp1106Subject suffixes (p. 171)a. Indicative (plain)sg p1trans. intr. trans. mtr.1 -(lh)k-an see §22.3.4 (p. 176), and ‘Passive’, below -(lh)kalh2 -(lh)k-acw -(lh)k-al’ap3 -as I -ø -it-as -as-wit -twit-as -it-as -witI IIlsg/pl 2sg/pl 3sg/plobject suffixesb. Subjunctive (conjunctive)sg p1trans. I intr. trans. mtr.1 -an see §22.3.4 (p. 176), and ‘Passive’, below -at2 -acw -al’ap3 -as -it-as -as-wit -twit-as I -it-as -wit-as1111lsg/pl 2sg/pl 3sg/plobject suffixesPassive (p. 176)The theme argument must appear closer to the predicate than the agent. Passivized predicatesare 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 listedbelow with the object suffixes that they select:object ( = theme) [ trans sg p11 I -tumc-al-em -tumul-emII -ts-al-em2 I -tumi-m -tam-lhk-al’apII -tsi-rn3 I -turn -tan-em-witII -em107(In)transitivizers (pp. 130 ff., 150-152)transitivizers intransitivizersI II-s (ts after s or lh) -Vn, -Vn’ -cal-en-s -nun/-nun’ -em, em’-mm/-mm’-cit-min-cit/-min’-cit-øPersonal (emphatic) pronouns (p. 191)sg p11 s-7ents s-nfmulh2 s-nüwa s-nuláp3 s-nilh wi s-nilhWh-words (Davis et al. 1993)who swat kuwhat s-tam’ kuwhere n-ka7 lhwhen (i)-kanm-ás-(as) lhwhy kánemhow s-kás-ts-as lhinduces conjunctive inflectioninduces conjunctive inflectioninduces nominalizationinduces conjunctive inflectionDemonstrative pronouns (deictics) (p. 198)_____________________sg p1this, these ts7a izávisible that, those ti7 iz’that, those (farther) t7u izüthis, these kw7a kwelhainvisible that, those ni7 nelhthat, those (farther) ku7 kwelh108Determiners (p. 223)sg ki ta /(t)i na mi ku kwp1 i nelh kwelh wiEncitics (pp. 231 ff.)variable wordsknown__________present absentcollective I individualunknownpresent absentpropernounswith a I without aa reinforcement (with cwilh, kaj det 239an’ evidential evid 232hem’ antithesis anti 237ha interrogative interrog 237k’a possibility, surmise evid 234kelh remote future, possibility fut 233ka obligation, expectancy oblig 233ku7 guotative guot 234malh adhortative adhort 231ga7 presupposed knowledge presupp 236tu7 definite past def.past 231t’th demarcation of time now 232well, but, so part 232wen emphasis emph 238wi7 emphasis emph 238cwilh after all, it turned out to be cons 235Procitics (pp. 252-253)(7e)lh ‘before’; links independent (non-subordinate) sentencesku links attributes to their objects; can often be dropped109Multi-clausal sentences (pp. 270 ff.)nilh conj (subordinating conjunction)kw (ku] det; goes with s- nominalizers nominalizer; induces possessivet detcomplementizers; normally lh hyptrigger conjunctive inflection i when.pastPrimary prepositions (pp. 254-256)Deictics and articles starting with 7 change the 7 to k when following these prepositions:1 in, on, at, with, among (oblique)e_ 1. toward, along; 2. by (with agent in passive constructions);(this preposition is often dropped)ken around, viaihel 1. from, out of; 2. than (in comparisons)Focusnilh foc focus marker-tali TO topical object: follows transitivizer in A-centered RCs with3rd person object; subject affix is omittedtsükw t’u7 only predicate17wa7 even predicateAuxiliarieswa7 progressive progka ... a out-of-control ooc110Appendix BKey to OrthographyData in this thesis are presented in the orthography that is used in van Eijk (1981, 1983) andPeters et al. (1992), with the exception of additional hyphens (-), which indicate morphemeboundaries. This orthography is the one that is currently the most widely used in St’át’imcetslanguage courses. For a key to the earlier Bouchard and Powell alphabets, consult Peters etal. (1992: appendix B).Orthography Americanist Phonetic Orthography Americanist PhoneticRepresentation Representationa e,a qw q’’e ,I,i xi i xw0 0 g Vu U gwçwao a r yv A 7p p p1 p’t t tits ts’ c’,ts’k k k’ k’kw k’’ k’w k”5 q’ q’z q’wqW1 1 z’m m 1’ 1’n II m’ m’w w n’y y w’ w’h h y’ y’c X r’cw X” g’lh + g’w c,’”q q ii e,I111Appendix CConsultantsThis appendix lists the principal language consultants for the Project on Lexical Interfaces withPhonology and Syntax in North West Coast Languages. The short biographies and genealogieswere prepared from information offered by each individual, and they are given here with theirpermission. All of the consultants are presently participating in linguistics courses inSt’át’imcets and Nla’kapmxcmn at the University of British Columbia and Simon FraserUniversity.Alice Adolph Kawá7tu. Born in Léqem’ts (near Mount Currie) to Francis Felix Leoand Evelyn Sam on 13 March 1939, and raised in nearby Xit’lólacw;attended St. Mary’s Residential School in Mission, B.C., from ages sixto 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 inVancouver, but still a member of the Cáclep Band. AA was raised byher paternal grandparents, Felix Leo (Lha7q) and Susan Felix Leo(Xánaq’a7) until she was six, and she spoke only St’át’imcets until sheattended residential school. Her maternal grandparents were Sam Jim(Sitii) and Caroline (Maggie) Jim of Sqátin (Skookumchuck); Sam Jim’smother was Kawá7tu, and Maggie Jim’s parents were Peter Williams (anhereditary chief) and Lucy Williams (Mamahisi), all of Sqátin.Beverley Frank Cdzll’. Born in Sék’wel’was (Cayoose Creek) to Baptiste (Han) andCatherine (KatlIin) Frank, where she still lives; attended Lillooet PublicSchool for eight years before transferring to Kamloops Indian ResidentialSchool, from which she graduated in 1959; attended the University ofBritish Columbia, from which she received a Bachelor of Educationdegree in Elementary Education. Until leaving for residential school inKamloops, BF spoke St’át’imcets daily with her parents and communityelders. Her father was from Sék’wel’was, and her mother was fromNxwéysten (Bridge River); her maternal grandfather was fromNqwátqwa7 (Darcy), and her grandmother, Cázil’, was fromNxwéysten.Gertrude Ned Born in T’ft’q’et (Lillooet) to Bernice Adolph on 1 February 1930, andraised by her maternal grandparents, Jack James of Nxwéysten (BridgeRiver) and Susan James of T’ft’q’et; attended St. Mary’s ResidentialSchool in Mission, B.C.—where she and RW were classmates—from age112seven until the seventh grade; thereafter attended high school inKamloops; 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 atMountain View Lodge, where she specialized in long-term care aid, andfor several years she has taught St’át’imcets at both the high school leveland 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 1June 1931; attended St. Mary’s Residential School in Mission, B.C.,from ages eleven through sixteen; returned to Lil’wat7ül until she wastwenty-four, when she was married; now living in Vancouver. HarryDan was from Seabird Island and Placida Pascal was from Lil’wat7ül.LT was raised by her parents and maternal grandparents, Joseph Pascal(hereditary chief of Lil’wat7ül) and Louise Pierre. Her paternalgrandparents were Dan and Rosemarie Michel of Seabird Island andSqátin (Skookumchuck) respectively. LT spoke only St’át’imcets untilshe 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 inLytton, and Charles Walkem—who was the leader of the Cook’s FerryBand—was born in Spences Bridge. His uncle, George Walkem, was atone time the premier of British Columbia. DU’s paternal grandfather,Hugh Blake Walkem, was born in Montreal and was a surveyor for theCanadian Pacific Railway; Hugh Walkem’s father, Charles Walkem,was born in Ireland. See Smith (1989: 102). DU spoke onlyNla’kapmxcfn until she attended residential school.Rose A. Whitley K’wstátqwa7. Born in Cáclep (Fountain) to Sebastian (Nk’yáp) andAdeline Peters on 18 May 1930; attended St. Mary’s Residential Schoolin Mission, B.C., from 1938 to 1946; spent her summer holidays inCáclep and in the United States picking berries. RW’s paternal great-grandparents were Peter Qwá7na and Pauline (Mutátkwa), JosephTsil.hüsalts (Cáclep’s first chief) and K’wswapáw’s, and Eustache Peter(Qatsk) and Rosalee (CaIts’a7). Her maternal great-grandparents wereBilly Fountain (Yawá7tulh) and Seraphine (TIcnek), Joe Joseph(Lil’wat7ül) and Josephine (K’wstátqwa7). RW’s family ancestry ismainly ScwepmectsIn and Lil’wat7ülmec. Her parents also attended St.113Mary’s Residential School, and so RW grew up speaking bothSt’át’imcets and Sám7ats (English). She married Edward Napoleon ofthe Lillooet Band (T’ft’q’et), where she still lives. RW has numerousgrandchildren and one great-grandson, and she holds a Bachelor of Artsdegree in Native Studies and Social Work Practice from Evergreen StateCollege (1987).114Appendix DElicitation SessionsThis appendix is an index to the St’át’imcets syntax database for the Project on LexicalInterfaces with Phonology and Syntax in North West Coast Languages, from which the datafor this thesis are drawn.Session Tokens Consultants Date1 1-27 BF, ON, RW 8 October 19922 28-42 BF, ON, RW 29 October 19923 43-59 RW 12 November 19924 60-78 GN 26 November 19925 79-114 GN, RW 3 December 19926 115-149 GN 8 December 19927 150-167 BF,GN,RW 7January 19938 168-191 BF (168-184), ON (all but 184) 21 January 19939 192-230 GN, RW 4 February 199310 231-256 GN, RW 18 February 199311 257-334 GN, RW 4 March 199312 335-372 GN, RW 18 March 1993373 BF13 374-375 BF 25 March 199314 376-434 GN, RW 1 April 199315 435-5 18 GN, RW 21 April 199316 5 19-536 BF 22 April 199317 537-601 BF, RW 6 May 199318 601-669 ON, RW 20 May 199319 670-743 ON (675-743), RW 16 September 199320 745-796 ON, RW 30 September 199321 797-835 BF 13 October 199322 836-918 ON (836-875), RW 14 October 199323 9 19-954 BF 27 October 199324 955-976 ON, RW 28 October 199325 977-1035 BF 4 November 199326 1036-1094 RW 25 November 199327 1095-1224 ON, RW 9 December 199328 1225-1298 ON 6 January 199429 1299-1353 ON 20 January 199430 1354-1445 ON, RW 3 February 199431 1446-1502 RW 17 February 19941503-1528 ON32 1529-1583 AA 2 March 199433 1584-1643 ON, RW 3 March 19941644-1712 BF, ON, RW (variously)1713-1808 ON, RW34 1809-1834 AA 8 March 199411535 1835-1870 AA 16 March 199436 1871-2034 BF, GN, RW (variously) 17 March 199437 2035-2064 GN, RW 24 March 199438 2065-2083 AA 30 March 199439 2084-2191 GN, RW (variously) 31 March 199440 2192-2212 AA 6 April 199441 22 13-2222 AA 7 April 199442 2223-2279 AA 13 April 199443 2280-2306 AA 18 April 199444 2307-2322 AA 20 April 199445 2323-2415 LT 22 April 199446 2416-2461 LT 27 April 199447 2462-2557 GN, RW 28 April 199448 2558-2641 LT 3 May 199449 2642-2680 AA 4 May 199450 268 1-2695 LT 5 May 199451 2696-2737 LT 10 May 199452 2738-2746 AA 11 May 19942747-2759 BF, RW (variously)53 2760-2806 AA 12 May 1994116ReferencesAissen, 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. Iterated CPs and Embedded Topicalization. Linguistic Inguiiy23(2): 329-336.Baker, Mark C. 1991. On Some Subject/Object Non-Asymmetries in Mohawk. NaturalLanguage and Linguistic Theory 9: 537-576.Banfield, Ann. 1973. Narrative Style and the Grammar of Direct and Indirect Speech.Foundations ofLanguage 10: 1-39.Browning, Marguerite A. 1987. Null Operator Constructions. PhD thesis, MassachusettsInstitute of Technology.Calabrese, Andrea. 1990. Some Informal Remarks on Focus and Logical Structures inItalian. 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. Dordrecht: ForisPublications.—. 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. MIT Occasional Papers inLinguistics 1.Chung, Sandra. 1978. Case Marking and Grammatical Relations in Polynesian. Austin:University of Texas Press.Cinque, Guglielmo. 1990. Types ofA’ -Dependencies. Cambridge: MIT Press.117Comrie, Bernard. 1979. Degrees of Ergativity: Some Chukchee Evidence. Pp. 2 19-240 inErgativity: 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. 1977. The Structure of Jacaltec. Austin: University of TexasPress.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 atthe University of British Columbia, 19 November.—. 1993b. A Configurational Pronominal Argument Language. Paper presented at WECOL23. Seattle: University of Washington, 22-24 October.—, Dwight Gardiner, and Lisa Matthewson. 1993. A Comparative Look at WH-Questions inNorthern Interior Salish. Pp. 79-95 in Papers for the 28th International Conference onSalish 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 theClause: 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. 1985. A Unified Theory of Syntactic Categories. Dordrecht: ForisPublications.Erteschik-Shir, Nomi. 1993. The Dynamics of Focus Structure. Ms., Ben Gurion Universityof the Negev.Fiengo, Robert, C.-T. James Huang, Howard Lasnik, and Tanya Reinhart. 1988. The Syntaxof Wh-in-Situ. Pp. 8 1-98 in Proceedings of the Seventh West Coast Conference onFormal Linguistics. Hagit Borer, ed. Stanford: Stanford Linguistics Association.118Gardiner, Dwight. 1993. Structural Asymmetries and Preverbal Positions in Shuswap. PhDthesis, Simon Fraser University. To be published by the University of BritishColumbia Press.—, Lisa Matthewson, and Henry Davis. 1993. A Preliminary Report on Word Order inNorthern Interior Salish. Pp. 139-157 in Papers for the 28th International Conferenceon Salish and Neighboring Languages. Seattle: University of Washington.Gerdts, Donna B. 1988. Object and Absolutive in Halkomelem Salish. New York: GarlandPublishing.Grimshaw, Jane. 1993. Minimal Projection, Heads and Optimality. Technical Report #4 ofthe Rutgers Center for Cognitive Science, Rutgers University.Heim, Irene. 1988. The Semantics of Definite and indefinite Noun Phrases. New York:Garland Publishing.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 Languageand 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).— and Bernard Comrie. 1977. Noun Phrase Accessibility and Universal Grammar.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 IndianLanguages. Paper presented at the 24th International Conference on Salish andNeighboring Languages, Steilacoom, Washington, August 16-19.—. 1990. Sorting Out Third Persons in Salishan Discourse, international Journal ofAmerican Linguistics 56(3): 34 1-360.Koopman, Hilda, and Dominique Sportiche. 1991. The Position of Subjects. Lingua 85:211-258.119Kroeber, Paul. 1991. Comparative Syntax of Subordination in Salish. PhD thesis,University of Chicago.Lasnik, Howard, and Juan Uriagereka. 1988. A Course in GB Syntax: Lectures on Bindingand 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., Universityof 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 NeighboringLanguages. Seattle: University of Washington.Mulder, Jean Gail. 1988. Ergativity in Coast Tsimshian (Sm’algyax). PhD thesis, Universityof 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 andErgative Languages. PhD thesis, Massachusetts Institute of Technology.Oehrle, Richard T. 1981. Common Problems in the Theory of Anaphora and the Theory ofDiscourse. Pp. 509-530 in Possibilities and Limitations of Pragmatics: Proceedings ofthe Conference on Pragmatics, Urbino, July 8-14, 1979. Herman Parret, MarinaSbisà, and Jef Verschueren, eds. Amsterdam: John Benjamins B.V.Partee, Barbara H. To appear. Topic, Focus and Quantification. Proceedings of the 1991SALT Conference. Steve Moore and Adam Wyner, eds. Ithaca: Department ofModern Languages and Literatures, Cornell University.Payne, Thomas E. 1982. Role and Reference Related Subject Properties and Ergativity inYup’ik Eskimo and Tagalog. Studies in Language 6(1): 75-106.Peters, Desmond, Sr., et al. [1992]. Introducing St’át’imcets (Fraser River Dialect): APrimer. Kamloops: Secwepemc Cultural Education Society.Prince, Alan, and Paul Smolensky. 1993. Optimality Theory: Constraint Interaction inGenerative Grammar. Technical Report #2 of the Rutgers Center for CognitiveScience, Rutgers University.120Reinhart, 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 ofBritish Columbia.Ross, John Robert. 1967. Constraints on Variables in Syntax. PhD thesis, MassachusettsInstitute of Technology. Distributed by Indiana University Linguistics Club. Pagereferences here are to the edition published as Infinite Syntax! (Norwood, NJ: AblexPublishing, 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 Campbelland Audrey Ward. Murphy Orlando Shewchuk, ed. Merritt: Sonotek Publishing.Smith, Lawrence R. 1984. On the Nonergativity and Intransitivity of Relative Clauses inLabrador Inuttut. Pp. 289-3 15 in Cook and Gerdts, eds. (1984).Thompson, Laurence C. 1979. Salishan and the Northwest. Pp. 692-765 in The Languagesof Native America: Historical and Comparative Assessment. Lyle Campbell andMarianne Mithun, eds. Austin: University of Texas Press.— and M. Terry Thompson. 1992. The Thompson Language (University of MontanaOccasional Papers in Linguistics, no. 8). Missoula: University of Montana.Thráinsson, Höskuldur. 1991. Long-Distance Reflexives and the Typology of NPs. Pp. 49-75 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. Ms.,University of British Columbia.van Eijk, Jan P. 1981. CuystwI Malh UcwalmIcwts: Teach Yourself Lillooet: UcwalmIcwtsCurriculum for Advanced Learners. Mount Currie: Ts’zil Publishing House.—. 1983. A Lillooet-English Dictionary. Ms., Mount Currie Cultural/CurriculumDevelopment.—. 1985. The Lillooet Language: Phonology, Morphology, Syntax. PhD thesis, Universiteitvan 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): 319-331.121Watanabe, 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 GreenlandicEskimo. 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