UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Intonation and Focus in Nte?kepmxcin (Thompson River Salish) Koch, Karsten 2008

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

Item Metadata

Download

Media
24-ubc_2008_fall_koch_karsten.pdf [ 19.22MB ]
Metadata
JSON: 24-1.0066828.json
JSON-LD: 24-1.0066828-ld.json
RDF/XML (Pretty): 24-1.0066828-rdf.xml
RDF/JSON: 24-1.0066828-rdf.json
Turtle: 24-1.0066828-turtle.txt
N-Triples: 24-1.0066828-rdf-ntriples.txt
Original Record: 24-1.0066828-source.json
Full Text
24-1.0066828-fulltext.txt
Citation
24-1.0066828.ris

Full Text

Intonation and Focus in N+e?kepmxcin (Thompson River Salish)  by KARSTEN KOCH B.A.,The University of British Columbia, 1995  A THESIS SUBMITTED IN P A R T I A L F U L F I L L M E N T OF T H E REQUIREMENTS FOR T H E D E G R E E OF DOCTOR OF PHILOSOPHY  in  T H E F A C U L T Y OF G R A D U A T E STUDIES  (Linguistics)  T H E UNIVERSITY OF BRITISH C O L U M B I A (Vancouver) July 2008  © Karsten Koch, 2008  Abstract In this dissertation, I examine the marking of focus and givenness in Nte?kepmxcin (Thompson River Salish). The focus is, roughly, the answer to a wh-question, and is highlighted by the primary sentential accent in stress languages like English. This has been formalized as the Stress-Focus Correspondence Principle. Given material is old information, and is deaccented in languages like English. Nte?kepmxcin is a stress language, but marks focus structurally. However, I argue that the structure has a prosodie motivation: the clause is restructured such that the focus is leftmost in the intonational phrase. It follows that Salish focus structures lack the special semantics that motivates the use of English structural focus (clefts). As a theoretical contribution, I show that the Stress-Focus Correspondence Principle does not account for focus marking in all stress languages, nor does the "destress-given" generalization account for the marking of given information. This is because focus surfaces leftmost, while the nuclear stress position is rightmost. Instead of "stress-focus", I propose that alignment with prosodie phrase edges is the universally common thread in focus marking. This mechanism enables listeners to rapidly recover the location of the focus, by identifying coarse-grained phonological categories (p-phrases and i-phrases). In Thompson River Salish, the focus is associated with the leftmost p-phrase in the matrix intonational phrase. The analysis unifies the marking of focus across languages by claiming that focus is always marked prosodically, by alignment to a prosodie category. The study combines syntactic analysis of focus utterances with their phonetic realization and semantic characteristics. As such, this dissertation is a story about the interfaces. This research is based on a corpus of conversational data as well as single sentence elicitations, all of which are original data collected during fieldwork. The second contribution of this dissertation is thus methodological: I have developed various fieldwork techniques for collecting both spontaneous and scripted conversational discourses. The empirical contribution that results is a collection of conversational discourses, to add to the singlespeaker traditional texts already recorded for Nte?kepmxcin.  Table of Contents Abstract  ii  Table of Contents  iii  List of Tables  viii  List of Figures  ix  List of Symbols and Abbreviations  xi  Acknowledgements  xiii  Dedication  xv  Chapter I: Overview and Background  1  LI  Overview  1  1.2  Organization of the dissertation  6  1.3  Background: What is focus?  7  1.4  1.5  1.3.1 The syntactic notion of focus  8  1.3.2 The semantic notion of focus  10  1.3.3 The prosodie expression of focus  11  1.3.4 Focus projection  15  1.3.5 Focus summary and preview of results  18  What is givenness?  20  1.4.1 A syntactic representation  20  1.4.2 The semantics of givenness (Schwarzschild 1999)  21  1.4.3 The prosodie expression of givenness  22  1.4.4 Givenness summary and preview of results  23  Representing S T R E S S - F O C U S and D E S T R E S S - G I V E N phonologically 1.5.1 Phase theory and a two step parse into p-phrases  25  30  1.6  Bilingualism and intonation  31  1.7  Intonation in Salish  34  1.8  Methodological, empirical and theoretical contributions  36  Chapter II: N-îeîkepmxcin Basics  38  2.1  Phonemic inventory  39  2.2  Word order and other syntactic background  40  2.2.1 Basic word order: V S O  40  2.2.2 Different types of predicates: verbs, nouns, adjectives  43  2.2.3 Demonstratives and headless relative clauses as arguments  45 iii  2.3  2.2.4 The topic projection at the left edge  45  2.2.5 Determiners  47  2.2.6 Relative clauses and their morphology (Kroeber 1997)  52  2.2.7 Two focus structures  54  Summary  59  Chapter III: Focus Structure  60  3.1  "Stress-focus" theories and predictions for Thompson River Salish  60  3.2  Focus projection and predictions for wide focus in Thompson Salish  62  3.3  A corpus study: Focus type and syntactic realization  63  3.4  3.3.1 Subjects  63  3.3.2 Method  63  3.3.3 Results: Wide C P focus  67  3.3.4 Results: Wide V P focus  68  3.3.5 Results: Narrow verb focus  71  3.3.6 Results: Narrow subject focus  72  3.3.7 Results: Narrow object focus  75  3.3.8 Results: Number quantifier focus  76  3.3.9 Results: Summary  79  Discussion  79  3.4.1 Implications for the "stress-focus" generalization 3.4.2 Implications for focus projection  79 .'  81  3.5  Clefts versus nominal predicate constructions: The role of contrast  88  3.6  Metalinguistic focus: A residual "stress-focus" effect  95  3.7  Summary  97  Chapter IV: A Case Study of Nuclear Stress in MeVkepmxcin  99  4.1  Nuclear stress  99  4.2  Informal observations about nuclear stress in  4.3  Phonetic observations about nuclear stress  OTeVkeprnxcin  101 104  4.3.1 Background  104  4.3.2 Subjects  106  4.3.3 Method  106  4.3.4 Results and discussion  107  4.4  Phonological observations: the direction of heavy NP shift (HNPS)  111  4.5  Summary  115 iv  Chapter V : Phonetics of Intonation  117  5.1  117  5.2  5.3  5.4  Background: Acoustic markers of prominence 5.1.1 What I will be comparing  118  5.1.2 Acoustic correlates of prominence: overview  123  5.1.3 Acoustic correlates of focal accenting (nuclear pitch accent)  125  5.1.4 Acoustic correlates of deaccenting  131  5.1.5 Declination  136  5.1.6 Effects of data source  136  Case study: neutral versus narrow focus  137  5.2.1 Subjects  138  5.2.2 Method  138  5.2.3 Statistical analysis  139  5.2.4 Results: the leftmost lexical stress  141  5.2.5 Results: the right edge  150  5.2.6 Results: declination  159  5.2.7 Discussion  165  Case study 2: The effects of data source  174  5.3.1 Subjects  175  5.3.2 Method  175  5.3.3 Statistical analysis  175  5.3.4 Results  175  5.3.5 Discussion  183  Conclusion  184  Chapter VI: Phonetic Evidence for Prosodie Phrasing  186  6.1  Background: Prosodie phrasing in English and Salish  186  6.2  Prosodie phrasing in MeVkepmxcin  189  6.3  The problem with clitics  197  6.4  Summary  206  Chapter VII: Motivating Focus Structure  208  7.1  The syntactic structure of clefts: Biclausal  208  7.2  The semantics of clefts  212  7.2.1 No exhaustivity effects  213  7.2.2 Corroboration from narratives  216  7.2.3 No presupposition of existence  218 V  7.2.4 Corroboration from narratives  220  7.2.5 Determiners: The source of the difference?  221  7.2.6 What makes focus structures special?  224  7.2.7 Tree diagrams for focus structures  227  7.2.8 What makes focus structures special, part 2: No overt nominals  229  7.2.9 A semantics for the cleft predicate  236  7.2.10 Semantics of clefts: Summary  239  7.3  Prosodie motivation: A PF discourse constraint  240  7.4  Purely syntactic focus marking: the predicate and the focus  243  7.5  Summary  243  Chapter VIII: Rethinking the Stress-Focus Correspondence  245  8.1  A review of prosodie phrasing in Thompson River Salish  246  8.2  Rethinking "stress-focus" in Generalized Alignment terms  246  8.2.1 Focus alignment in English: A n example  248  8.2.2 Nie?kepmxcin: Clefts as alignment of focus with prosodie phrase edges... 250  8.3  8.4  8.2.3 Focus in two tone languages (Truckenbrodt 1999, Downing 2003)  256  8.2.4 Limits of the system: Eligible prosodie categories  259  Rethinking "stress-focus" in English  260  8.3.1 Some phonological background  260  8.3.2 A n edge-alignment account for English  266  8.3.3 Focus marking as edge-alignment: Summary  269  Prosodie versus syntactic motivation for clefts (Focus = Predicate)  270  8.4.1 Additional evidence: Null foci and focused clitics  271  8.4.2 C P focus expressed via clefts or nominal predicate constructions  276  8.4.3 Association with focus  276  8.4.4 Extraposition from complex nominal predicates  280  8.4.5 Predicate-argument flexibility (Davis 2007)  282  8.4.6 Narrow object focus is not expressed using verb-object word order  283  8.4.7 The case of auxiliaries  285  8.4.8 Focused DPs are not simply moved leftward  286  8.4.9 Summary  286  8.5  On the relationship between focal accenting and given deaccenting  288  8.6  Summary  292  Chapter IX: Conclusion  294 vi  References  301  Appendices  320  Appendix A l :  Scripted conversation - "Going to the grocery store"  321  Appendix A2: Scripted conversation - "Going to the clothing store"  331  Appendix B: Spontaneous dialogue - "Visitors and babysitting"  337  Appendix C I : Spontaneous conversation - Responding to questions about a multimedia display - newness vs. contrast  341  Appendix C2: Spontaneous conversation - Responding to questions about a multimedia display - frog sequence 1  350  Appendix C3: Spontaneous conversation - Responding to questions about a multimedia display - frog sequence 2  354  Appendix C4: Spontaneous conversation - Responding to questions about a multimedia display - bear sequence  360  List of Tables Table 1.1  Focus and givenness marking in English and Salish  29  Table 2.1  The Salish language family  38  Table 2.2  Phonemic inventory  39  Table 2.3  Determiners in NteVkepmxcin  52  Table 2.4  Relative clauses and their morphology  54  Table 3.1  Stress-focus systems and predictions for MeVkepmxcin  62  Table 3.2  Focus type and syntactic realization - a corpus study of 338 cases  79  Table 3.3  Focus construction and focus type  90  Table 3.4  N'te?kepmxcin versus "stress-focus" systems  98  Table 4.1  Declination measures by speaker in default wide focus utterances  111  Table 5.1  Comparisons to be made  123  Table 5.2  Acoustic cues for left edge narrow focus accent (nuclear pitch accent),  130  Table 5.3  Acoustic cues for deaccenting of given material  135  Table 5.4  F E leftmost lexical stress: summary of acoustic cues, and t-test results  149  Table 5.5  P M leftmost lexical stress effects: summary of acoustic cues, t-test results  150  Table 5.6  FE rightmost stress effects: summary of acoustic cues, and t-test results  158  Table 5.7  P M rightmost stress effects: summary of acoustic cues, and t-test results  159  Table 5.8  F E declination effects: summary of acoustic cues, and t-test results  Table 5.9  P M declination effects: summary of acoustic cues, and t-test results  164 165  Table 5.10  Main effects of data source: the leftmost lexical stress  179  Table 5.11  Significant post-hoc pairwise comparisons of data source (leftmost)  179  Table 5.12  Main effects of data source: the rightmost lexical stress  180  Table 5.13  Significant post-hoc pairwise comparisons of data source (rightmost)  180  Table 5.14  Interaction effect of data source by speaker: FO maximum (Hz)  181  Table 5.15  Significant post-hoc pairwise comparisons of data source on FO peak  181  Table 5.16  Main effects of data source: global utterance characteristics  182  Table 8.1  Syntactic versus prosodie focus marking: Additional evidence  288  Table 9.1  Predicted typology and gaps: Focus and i-phrase alignment  298  List of Figures Figure 1.1  Pitch tracing and waveform: "No, it was [Karstenlppc [that washed theml^."... 19  Figure 1.2  Contrast!ve focus accent and post-focal deaccenting in English  Figure 1.3  Pitch tracing and waveform: "When Mary is hungry, she picks some fruit... " .25  Figure 2.1  Pitch tracing and waveform: "(Our mother helped our brotherJFoc '  Figure 2.2  Pitch tracing and waveform: "[My husband was cleaning up the snowlpoc" •••42  Figure 2.3  Pitch tracing and waveform: "[1 heard it's going to rain tonightJFoe"  43  Figure 2.4  Pitch tracing and waveform: "[Eddie is heavy]poc."  44  Figure 2.5  Pitch tracing and spectrogram for example (10)  46  Figure 2.6  Relative clause in NteVkepmxcin: "the (thing that) I found"  51  Figure 3.1  Metalinguistic focus: "They [havelpoc lots of dogs."  97  Figure 3.2  Metalinguistic focus: "The ones that are fishing there are three wo[men|Foc."..97  Figure 4.1  Stress on s?â?a? 'crow' (spectrogram shows 0-5000 Hz)  101  Figure 4.2  Pitch tracing and waveform for the sentence My mother drinks coffee  105  Figure 4.3  FO declination for F E across multiple peaks  108  Figure 4.4  FO declination for P M across multiple peaks  109  Figure 4.5  Pitch tracing and waveform: "|Our mother helped our brotherjpoc"  110  Figure 4.6  Pitch tracing and waveform: "[My husband cleaned up the snow|poe"  110  Figure 5.1  Maximum left edge FO by speaker and focus type  142  Figure 5.2  Left edge FO range (semitones) by speaker and focus type  143  Figure 5.3  Left edge time of FO peak as a percentage of vowel duration  145  Figure 5.4  Left edge maximum amplitude (db)  146  Figure 5.5  Left edge average vowel amplitude (db)  147  Figure 5.6  Left edge vowel duration (ms) by speaker and focus type  148  Figure 5.7  Maximum right edge FO by speaker and focus type  152  Figure 5.8  Right edge FO range by speaker and focus type  153  Figure 5.9  Right edge time of FO peak as a percentage of vowel duration  154  Figure 5.10  Right edge peak amplitude (db) by speaker and focus type  155  Figure 5.11  Right edge stressed vowel duration (ms) by speaker and focus type  157  Figure 5.12  Pitch contour across FO peaks for FE, by focus type  160  Figure 5.13  Pitch contour across FO peaks for P M , by focus type  161  Figure 5.14  Amplitude contour across peaks for FE, by focus type  162  Figure 5.15  Amplitude contour across peaks for P M , by focus type  163  Figure 5.16  Peak amplitude declination values for F E by focus type  168  Figure 5.17  Peak amplitude declination values for P M by focus type  169  20 41  ix  Figure 5.18  Focal accent and post-focal deaccenting in English  170  Figure 5.19  Pitch tracing, waveform and amplitude: "No, it was [Rosslpoc  171  Figure 5.20  Pitch tracing, waveform and amplitude curve: "It's just |waterIppc ..."  172  Figure 5.21  Left edge vowels are marginally louder in scripted utterances  176  Figure 5.22  Right edge vowels are marginally louder in scripted utterances  177  Figure 5.23  Right edge vowels are marginally longer in scripted utterances  177  Figure 5.24  FO declination does not differ significantly by data source  178  Figure 6.1  Pitch tracing and waveform: "|Our mother helped our brotherlpoc"  190  Figure 6.2  Pitch tracing and waveform: " | M y husband cleaned up the snowipoc-"  191  Figure 6.3  Pitch tracing and waveform: "[We're gonna' go look for some grouselpoc-" •• 192  Figure 6.4  Pitch tracing and waveform: "[Bill's younger brother gave him a sweaterjp" 193  Figure 6.5  Pitch tracing and waveform: "Peter has two houses."  194  Figure 6.6  Pitch tracing and waveform: "[No.[ He only has [oneJFoc big pig-"  195  Figure 6.7  Pitch tracing and waveform: "[No,] it was IKarstenJpoc that washed them."... 196  Figure 7.1  A syntax for the extraposition of English clefts (Hedberg 2000: 913)  223  Figure 7.2  Tree for nominal predicate construction  228  Figure 7.3  Tree for cleft construction  229  Figure 7.4  Residue clauses as DP subjects (headless relative clauses): N P C  234  Figure 7.5  Residue clauses as DP subjects (headless relative clauses): cleft  234  Figure 7.6  Residue clauses as DPs with bare C P complement: N P C  235  Figure 7.7  Residue clauses as DPs with bare C P complement: cleft  235  Figure 8.1  Pitch tracing, waveform and phrasing for ( 19B):  254  List of Symbols and Abbreviations Abbreviations used in the glosses (based on Thompson and Thompson 1992, 1996, Kroeber 1997) are as follows: '-' = affix or clitic  MDL = middle  '=' = lexical suffix  NCM = non-control middle  APPL = applicative  N c r = non-control transitivizer  AUG = augmentative reduplicant  NEG = negation  AUT = autonomous,  NOM = nominalizer  CAUS = causative  o = object  CNSQ = consequential  OBL = oblique  COMP = complementizer  PERF = perfective  CONJ = conjunctive (i.e. subjunctive)  PERS = persistent (emphatic particle)  DEM = demonstrative  PL = plural  D, DET = determiner  poss, PS = possessive  DIM = diminutive  PROG, PRG = progressive  DRV = directive transitivizer  PRP = proportional  DVL = developmental  Q = y/n question marker  EMPH = emphatic  RED = reduplicant  EVID = evidential  REEL = reflexive  ek'^u 'reportive, hearsay' nke 'conjectural, I guess, apparently' nuk" 'perceptual, usually other than sight' FUT = future  REL = relational  IM = immediate  REM = reaffirmative  IMP = imperative  RPT = repetitive  INCH = inchoative  SG = singular  INSTR = instrumental  STAT  CLEFT = cleft predicate  SUBJ.EXTR  IRL = irrealis,  TRANS/TR  LOC = locative  TS = transitive subject  = stative prefix = subject extraction suffix = control transitivizer  For reasons of space and clarity, I often do not provide full morphological breakdowns for nouns, adjectives, adverbs, and so on.  Data are presented in the orthography developed in Thompson and Thompson (1992, 1996), and Kroeber (1997). The phonemic key to the orthography is as follows; symbols not listed have the standard IPA interpretation: c = [tJlor[c] ç=[ts] c=[ts'] e=[e, as, a, e, a] y=[i, ei, ai] o = [o, o] s = Lllor[s] s=[s] u = [u, o, o] y = ly, iJ. See Thompson and Thompson (1992) in particular for the phonetic realizations of phonemic vowels across contexts. N4^e?kepmxcin [z] is more lateral than English [z], though there may be considerable regional or speaker variation.  Acknowledgements This dissertation is the final stage in a long path of education and research that has seen many different people help me along the way. I wish to express my thanks to some of those people here; invariably I will miss others, whose aid and encouragement was no less appreciated. Without the commitment and hard work of Flora Ehrhardt and Patricia McKay, two fluent speakers of N+eVkepmxcin, this project could not even have started. We have spent many hours together over the past number of years, and I thank them for their infinite patience and willingness to teach me about many aspects of their language and culture. Cunnwénne me^f xé?e! I hope that our relationship is one which will continue for a long time - k"'uk"scémx"'. Many thanks also go to my first NieVkepmxcin teacher, Mandy Na'zinek Jimmie; and to Patricia Shaw for organizing the F N L G courses. Eric Rosen and Paivi Koskinen were among my first linguistics professors, and along with Geoffrey Hall, helped me apply successfully to graduate school. Without the help of Guy Garden and Joseph Stemberger I would not have secured funding from NSERC to enable my graduate education. In my early years of graduate school, Bryan Gick, Patrick Moore and Martina Wiltschko offered me the opportunity for valuable research, writing, presentation and publication experience. Rose-Marie Déchaîne introduced me to fieldwork methodology in Yoruba, and thus inadvertently shaped the content of my second qualifying paper. Lisa Matthewson, Patricia Shaw and Joseph Stemberger oversaw my initial qualifying papers drafts, and I am especially thankful to Hotze Rullmann and Henry Davis for shaping the final versions of these papers. Many thanks go to my dissertation committee members Hotze Rullmann and Eric Vatikiotis-Bateson, whose influence will be felt throughout this work, as well as in my future directions. My dissertation work was supervised by the tireless Henry Davis, who helped to shape the research program behind this work and provided invaluable feedback, pushing and challenging me to greater levels of insight and explanation. He has always taken a genuine interest in my research and the outcomes, and considered what the implications are not only for other Salish languages, but for language in general, invariably suggesting new avenues for exploration that I had not thought of. My additional examiners, Patrick Moore, Douglas PuUeyblank, and Daniel BUring, offered valuable and detailed commentary that shaped the final content of this work, and no doubt its future forms as well. Other graduate students in my cohort were also valuable discussants and coresearchers. Jason Brown always knew how to distill interesting data to reveal its most core importance; Jeff Muehlbauer, Clare Cook, Solveiga Armoskaite, Tyler Peterson, Sonja  Thoma, Ryan Waldie and James Thompson shared many of their own ideas and research, and saw relevance to my own interests; Peter Jacobs and Marion Caldecott provided valuable insight into Salish phonetics, phonology, as well as research methodology. And I enjoyed conversations with many others not of my "generation." The research behind this dissertation was made possible through funding from Jacobs Kinkade Research Grants through the Jacobs Research Fund, as well as funding from SSHRCC Grants to Henry Davis, Lisa Matthewson, and Hotze Rullmann. Many thanks go to my parents, Herbert and Brita, who learned to no longer ask when I expected to be finished; but, rather more proactively, bought me my "graduation gift" two years ago — the laptop computer on which I wrote this dissertation. Without their interest and unqualified support I would not have been able to complete this research. Many friends and family offered support throughout the years, whether through curiosity in my strange linguistics research work, or by simply being around to take my mind off of the academia, if only for a short while. Thank you! Finally, it takes two to tango, and without Monique I would not have been able to dance my way through this disseration project to its final completion. Merci, ma belle!  Dedication  for future learners and speakers of N*e?kepmxcin  Chapter I: Overview and Background 1.1  Overview Intonation, or the speech melody, phrasing, and pauses that overlay our utterances,  plays a crucial role in the marking of information structure in many - perhaps all languages. In this dissertation, I provide the first comprehensive study of intonation and focus in N'teVkepmxcin (Thompson River Salish). A study of focus is an interface study, and the present account will take us through many of the areas of the language faculty: syntax, semantics, phonology and phonetics. Although NleVkepmxcin is a stress language (Egesdal 1984, Thompson and Thompson 1992), acoustic phonetic evidence indicates that foci do not bear the primary sentential stress. Instead, utterances are restructured so that focused elements align with the left edge of the clause; nuclear stress (underlined), however, is at the right edge. (1)  Neutral focus example (all new information): ?éx  xe?  cax-t-0-és  t  PROG  D E M clean-TR-3o-3TS  n-sxaywi  DET ISG. PS-husband  e  swux'^t.  DET  snow  " M y husband was cleaning up the snow." ' (2)  Object focus example (nominal predicate construction): A : Sté? what  x^'ûy k FUT  coMP  s-ta?xâns-9p  tk  sVap.  N0M-eat-2PL.P0ss  OBL.IRL  evening  "What are you people going to eat this evening?" B: [pmsjFoc nee? beans  ISG.EMPH  x^uy  e  n-s-'ta?xâns.  FUT  DET  ISG.POSS-NOM-eat.  "I'm gonna' eat [beanslpoc-" (literally "[Beanslpoc's the (thing that) I'm gonna' eat.") As a result, speakers of N'teVkepmxcin violate the Stress-Focus Correspondence Principle which has been proposed as a model of focus marking in stress systems (Reinhart 1995; also Selkirk 1995, Vaissière 1995, Schwarzschild 1999, Szendroi 2003, Gussenhoven  ' See the List of Symbols for keys to the orthography, and to the abbreviations in the gloss; see Thompson and Thompson (1992, 1996) for further detail. For reasons of space and clarity, I often do not provide full morphological breakdowns for nouns, adjectives, adverbs, and so on.  2004 on the Effort Code, Féry and Samek-Lodovici 2006, Hartmann 2007). NieVkepmxcin speakers also violate the generalization that old, or given, information is deaccented after the focal stress. These two generalizations are described by the constraints below, from Féry and Samek-Lodovici (2006). (3)  STRESS-FOCUS:  a focused phrase has the highest prosodie prominence in its focus  domain. (4)  DESTRESS-GIVEN:  (Féry & Samek-Lodovici 2006: 135-6) A given phrase is prosodically non-prominent. (Féry and Samek-Lodovici 2006: 135-6)  The idea that stress languages mark focus through the use of pitch accents is strongly expressed by Hartmann (2007): Intonation languages use pitch accents as the principle means of focusing.... A pitch accent triggers expansion of the pitch range. After the nuclear accent, the pitch range is considerably compressed. (Hartmann 2007: 225-226) The data analyzed in this dissertation, however, are not accounted for by the generalizations in (3) and (4), as expressed by Hartmann. I therefore propose a reanalysis of the model of focus marking in languages like English. In the remainder of this section, I give some more details about this proposal. Rather than "stressing" focused information, we place focused items at the edges of speech phrases (Truckenbrodt 1999 on Chichewa, Gussenhoven 2004 on Basque). For example, in Thompson Salish, focused items are at the left edge of the sentence (in the first phonological phrase), while nuclear stress is rightmost. In English, focused items are at the right edge; since this happens to be where stress is assigned in English phrases, there is an apparent "stress-focus" correspondence. Thus, the universal interface between a syntactic focus and its phonological realization is edge-onQXittd, not stress-oriented. A second result is that an apparent structural focus strategy like the clefts employed in Salish focus (Kroeber 1997, 1999) is in fact prosodically motivated (Szendroi 2003 on Hungarian structural focus). To support this claim, I follow Kroeber (1997, 1999) in showing that clefts in NteVkepmxcin are not generated by movement; thus there is no syntactic motivation for structural focus. Moreover, Thompson Salish clefts lack the special semantic interpretations associated with English clefts (Davis et al. 2004 on St'ât'imcets and Straits  Salish; Perçus 1997 and Hedberg 2000 on English clefts); this fact rules out a semantic motivation for focus structures. Instead, N+eVkepmxcin clefts best satisfy the discourse prosodie requirement that focus be aligned with the left edge of the clause. The findings allow the reduction of what have previously been considered two parameters in the marking of focus to a single linguistic universal: focus is marked prosodically, through alignment to phrase edges. Regarding the first parameter, languages have been observed to mark focus prosodically (English, German), or structurally (eg. Brody 1995 on Hungarian, Kroeber 1997 on N1^e?kepmxcin). Prosodically, languages like English employ prosodie heads (pitch accents) to mark the focus, while languages like Chichewa or Korean employ phrasal boundaries (phrase edges). (5)  Focus marking cross-linguistically: Two parameters [to be revised] Focus Marking  Head-oriented  Edge-oriented  English  Chichewa  Hungarian  German  Korean  NteVkepmxcin  Following Szendroi (2003) on Hungarian, I propose that cases of structural focus are in fact prosodically motivated. This eliminates one parameter for focus marking. The second parameter concerns which prosodie units are relevant for the marking of focus: prosodie phrase edges, or prosodie heads. Languages like Chichewa (Truckenbrodt 1999), Korean (Jun 2003: 239-241, ft. 17; also Selkirk 1996, cited in Beck 1999), or Basque (Gussenhoven 2004 on contrastive focus) insert a phrase boundary before or after a focused X P , thus aligning focus with prosodie phrase edges. Languages like English, which obey the "stress-focus" correspondence (Reinhart 1995, and many others), align focus with prosodie heads. I reformulate S T R E S S - F O C U S within the framework of Generalized Alignment theory (McCarthy and Prince 1993; Truckenbrodt 1999) to capture this variation under a single system: focus is predicted to align with different prosodie categories, edges or heads. However, since prosodie units have only a single head, at either the left or right edge, I propose that this second parameter of focus marking can also be reduced to a single relevant  unit: focus seeks out prosodie phrase edges, not heads. Like the "stress-focus" accounts, this is a stronger hypothesis than Generalized Alignment to either heads or edges, so I pursue it to see to what extent it can account for a classic "stress-focus" system, English. In languages like English, prosodie heads happen to line up with the same edge as foci (rightmost). Since constraints regulating the default marking of stress (prosodie heads) have been independently proposed (Truckenbrodt 1999; Féry and Samek-Lodovici 2006), there is no need to double this information in constraints referring to focus marking. Thus, the "stress-focus" correspondence is epiphenomenal. (6)  Focus marking cross-linguistically [new proposal] Focus Marking Prosodie, Edge-oriented  Additional Parameters Focus Alignment p-phrase  i-phrase  DESTRESS-  NUCLEAR  GIVEN  STRESS  English German  Right —  V  Right  Hungarian^ Romance  Left —  Nfe?kepmxcin Chichewa Korean'*  Right  Right  X  Left Right  Right Right  X[?]^ —  Right[?I  y/  Parameters independent of focus marking account for cross-linguistic differences: (i) alignment of the focus and/or nuclear stress to left versus right phrasal edges (ii) "destress-given"  ^ The analysis of Chichewa proposed in Truckenbrodt (1999) and Downing (2003) suggests that given material is not differentiated by reduction of accent; a recent study by Downing et al. (2007) suggests that focused p-phrases containing high tones are also marked by higher pitch. This opens the possibility that D E S T R E S S - G I V E N is operative. ^ Generalizations for Hungarian are taken from Szendroi (2003: 44). Generalizations for Korean are taken from Jun (2003: 239-241).  In regards to point (i), English and German are both languages with righmost nuclear stress. Focus and nuclear stress both align to the right edge of the intonational-phrase (iphrase). In Chichewa, focus aligns with the edge of a p-phrase, but not an i-phrase (Kanerva 1990, Truckenbrodt 1999, Downing 2003 on Chichewa), while in Korean, focus also aligns with a p-phrase (and sometimes also an i-phrase - Jun 2003). In Hungarian, focus aligns with the left edge of i-phrases, the same location as nuclear stress (Szendroi 2003). And in NteVkepmxcin, focus is left-aligned in the i-phrase, while nuclear stress is right-aligned. Regarding point (ii), in languages like English, the constraint D E S T R E S S - G I V E N is operative, independent of the system of focus marking. As a result, given material (old information) is not parsed into phonological phrases at the interface of syntax and phonology (Selkirk and Kratzer 2007); this prevents given material from bearing phrasal accent. I will argue that we can use this insight to claim an apparent "stress-focus" alignment in languages like English. The stress-focus correspondence is thus just the result of focus being the only material eligible for a phrasal head, since given material is not parsed into p-phrases. An example is presented to illustrate. In (7), the entire sentence Sam ate a squash is the focus since it answers the wh-question of (7A); focus is indicated by subscript ' F O C ' The focus also receives the nuclear stress (shown by underlining), which falls in the default rightmost position (eg. Chomsky and Halle 1968, Cinque 1993, Selkirk 1995, on nuclear stress in English). There are two phonological phrases, and the focus and the main stress are both right aligned. Word-level stress is shown by acute accent ('). (7)  A : What happened? (  X  )  i-phrase  ( X ) (  X  )  p-phrase  B: [Sam  ate a squashIpoç.  In (8), however, the focus is the subject Sam, since this is the answer to the whquestion in (8A). The primary sentential stress is once again on the focus. In addition, ate your dinner is old, given information in (8B) (shown with a subscript ' G ' ) . Hence, ate your dinner is not parsed into a p-phrase at the interface of syntax and phonology. Sam is thus the only eligible head for phrasal stress at this point in the derivation. Selkirk and Kratzer (2007) suggest that given material is recursively parsed into a p-phrase by the phonological component, but since each p-phrase can bear only one head, Sam will bear the only p-phrase accent - this accent serves as head of both p-phrases. Both focus and phrasal stress are once again right-aligned, since there are no other potential p-phrase heads further to the right  within the i-phrase; and the larger p-phrase containing the focus is right-aligned in the i phrase. I'll be assuming this two-step model of the syntax-phonology interface in this dissertation (see section 1.5.1 for more discussion).^ (8)  A : Who ate my dinner? ( X  )  i-phrase: STEP 2  ((X )  )  p-ph recursively parsed by phonology: STEP 2  (X )  p-phrase at interface: STEP 1  B: [SamlFoc [ate your dinnerlc.  1.2  Organization of the dissertation The dissertation is broadly organized into three sections, plus an appendix. The first  portion (chapters 1-3) gives background information and introduces focus data from NfeVkepmxcin. The second portion (chapters 4-6) provides phonetic analysis to support some of the observations made in chapters 2 and 3. The final two chapters discuss the theoretical implications in more detail. In the remainder of this chapter, I summarize what is meant by focus and givenness. I also review the literature on intonation in Salish; as well as research on intonation and bilingualism, since the Nte?kepmxcin consultants whom I worked with for this dissertation are also fluent in English. Chapter 2 gives some background on word order and syntactic structure in N"le?kepmxcin; I show that matrix predicates are always leftmost in the NieVkepmxcin intonational-phrase. In chapter 3,1 present a corpus study of focus type (wide focus or narrow focus) and its syntactic realization in conversational data in Thompson River Salish; we shall see that focus is always associated with the predicate, which is at the left edge of the NieVkepmxcin clause. The middle portion of the dissertation turns to phonetic analysis. Chapter 4 is a study of phrasal stress in N"te?kepmxcin, and I offer detailed observational, acoustic phonetic, and phonological support for the claim that nuclear stress is rightmost in neutral utterances. I continue with the acoustic phonetic analysis in Chapter 5, where I present the results of an analysis of narrow focus cases in Nfe?kepmxcin; the results indicate that focus is not marked with additional prosodie prominence, nor is given information deaccented. Finally, Chapter 6  ^ In principle, this process can happen in a parallel evaluation system, but I'll be presenting this as a serial model because it makes the point more clearly.  provides evidence for prosodie phrasing in Thompson River Salish, and I claim that predicate and arguments are parsed into individual phonological phrases (Beck 1999 on Lushootseed Salish, Barthmaier 2004 on Okanagan Salish; Hayes and Lahiri 1991 on similar parsing in Bengali, Schafer and Jun 2002 on Korean, Nespor and Sandler 1990 on Israeli Sign Language). The last portion of the dissertation examines the motivation for the structural focus observed in NteVkepmxcin. Chapter 7 considers, and rejects, semantic motivations and syntactic movement accounts. In Chapter 8,1 consider two alternative hypotheses: (i) focus marking is purely syntactic, by association with the predicate (Davis 2007), and (ii) focus marking is prosodie, by alignment with the left edge of the intonational phrase. Because any syntactic expression of focus is necessarily expressed in the phonology, at a minimum by linearization, the two hypotheses are not mutually exclusive. That is, the syntactic derivation is recoverable from a linear acoustic output, as well as a linear visual output (eg. Yehia et al. 2002, Santi et al. 2003, Callan et al. 2001,2004a, 2004b). However, I give evidence which suggests that focus marking is optimized for rapid phonological identification of the focus (Selkirk 1995, Buring 2003, 2006, on "focus projection"; Kjelgaard and Speer 1999, Jun 2003, Fodor 1998 on prosodie parsing preceding syntactic parsing; Callan et al. 2004b on listeners internally simulating the speech act of speakers).  1.3  Background: What is focus? Intonation is used to signal a variety of discourse notions. Of primary concern in this  dissertation are the notions of focus and of givenness. Various terms have been applied to distinguish focused and given information from other portions of the discourse. Terminology includes the distinctions between given/new information (Halliday 1967b, Chafe 1976, Prince 1981), topic/comment (Hockett 1958, van Kuppevelt 1994) or theme/rheme structures (Firbas 1966, Harlig and Harlig 1988, Rialland and Robert 2001), focus/background or focus/presupposition (Chomsky 1971, Jackendoff 1972, von Stechow 1990, Krifka 1992), and focus/ground (Vallduvi 1990). In view of the overlapping use of terminology, it is worthwhile defining what I mean by focus and givenness in this dissertation. A recent volume by Féry et al. (2007) aims at securing more precise definitions of these aspects of information structure; I refer the reader to Krifka (2007) in particular for a more thorough discussion of focus, givenness, and other concepts in information structure.  1.3.1 The syntactic notion of focus In this study, I make the standard assumption that focus is a syntactic category. The constituent that is the focus is identified by f(ocus)-marks, or a [FOCUS] feature (Jackendoff 1972, Selkirk 1984, 1995, Brody 1995). Rooth (1992) introduces a focus operator in the syntactic representation. This syntactic focus feature mediates between the phonological expression of focus and its semantic interpretation. I will indicate the focused constituent with a subscript ' F O C ' For the sake of clarity, I will not mark subconstituents of the focus with additional f-marks, as Selkirk (1995) does; in addition, recent work suggests the possibility of doing away with many of these f-marks (Schwarzschild 1999, Féry and SamekLodovici 2006, Selkirk 2007). Different types of focus can be identified in different ways. New information focus, or presentational focus, is classically diagnosed as the answer to a wh-question (Halliday 1967b on "informational focus," Jackendoff 1972, Selkirk 1995, Buring 2003, and many others). Bart in (10) is a new information focus, since this D P answers the question who. In English and many other stress languages, the focus is marked prosodically by bearing the primary sentential stress (the nuclear stress). I indicate this by underlining, while acute accent (') indicates word-level stress. (9)  New information focus A wh-question expression focuses a constituent, and an appropriate answer to a wh-question must focus the same constituent.  (10)  (Selkirk 1995: 553)  A : Who cooked dinner? B: [Bartjpoc cooked dinner. Wh-questions do not need to be overt in the discourse; information that answers an  implicit wh-question is also focused (van Kuppevelt 1994, Buring 2003b, Krifka 2007). New information focus is often distinguished from contrastive focus (eg. Rochemont 1986, Gussenhoven 2004, Féry and Samek-Lodovici 2006, Selkirk 2007; É. Kiss 1998 on "identificational" focus in Hungarian/English). (11)  A : I heard Janice found some mushrooms. B: No, [KeHylpoc found some mushrooms.  Contrastive focus sequences are distinguished from new information focus in that they involve dual and symmetric frames in which only one element differs. The background is shared between both configurations. In (11), this frame is x found some mushrooms, and Janice and Kelly are the contrastive foci in a type of anaphoric relationship (Ladd 1980, Rochemont 1986, Rooth 1992:80, Féry and Samek-Lodovici 2006:135). In (12), the frame is X farmer and the adjectives American and Canadian are the contrastive foci. In this case, the adjectives are contrasted within a single sentence, and use of the frame x farmer results in phrasal accent on American or Canadian rather than farmer in the two DPs (Rooth 1992: 80, ex. 11). (12)  A n |Américan]poc farmer was talking to a [Canadianjpoc farmer. Like in the Nie?kepmxcin example in (2), structural focus can also be used for  contrastive focus in English. In English clefts, the contrastively focused cleft head {Bart in 13B) bears the nuclear stress of the clause. (13)  A : Did Sam cook dinner? B: No, it was [Bartjpoc who cooked dinner. A contrastive focus can be embedded within a new information focus. In these cases,  marking the contrastive focus is more important than marking the new information focus (BUring 2003, Féry and Samek-Lodovici 2006). In the example below, a new pair of black boots answers the wh-question and thus constitutes the new information focus. However, the nuclear stress falls on black; the adjective is the contrastive focus in the frame a new pair ofx boots.^ (14)  A : Natalie bought a new pair of brown boots. What did Andrea buy? B: Andrea bought [a new pair of [blacklppn bootsjpoc. * As Buring (2003) points out, it is not always clear whether stress on black here is a  result of a local contrastive focus frame, or because black constitutes the only non-given item within the new information focus. Thus, deaccenting of given material will similarly result in nuclear stress on black (eg. Schwarzschild 1999). It is not my aim to decide between these two approaches; I will continue to treat these as cases involving a local contrastive focus. The important point is that, in either case, we expect the nuclear stress on the contrasted adjective black.  1.3.2 The semantic notion of focus Both new information focus and contrastive focus as discussed in the previous section fall under what Krifka (2007) identifies as expression focus. Expression focus is pragmatic; it steers the conversation, or otherwise manages what is in the Common Ground of discourse (eg. Karttunen 1974, Stalnaker 1974). For example, the new information focus in (10) satisfies A ' s need to add the information about who cooked dinner to the Common Ground. However, it does not immediately affect truth conditions - that is, the meaning of the propositions in the Common Ground. (lOB) still means that Bart cooked dinner. This contrasts with what Krifka calls denotation focus, in which semantic operators are associated with focus. These change the truth conditions of the utterance. For example, only adds the truth conditional meaning that the focus denotation is the only alternative that makes the proposition under discussion true (Krifka 2007: 25). 1 primarily look at expression focus in this dissertation, though I will discuss association with focus in Chapter 8. There is not always a clear distinction between expression focus and denotation focus (Krifka allows that the latter may have developed from the pragmatic principles governing expression focus). In the data in this dissertation, we shall see that focus particles often surface with both new information and especially contrastive focus cases (section 3.5). Both types of focus identified by Krifka have in common "the presence of alternatives that are relevant for the interpretation of linguistic expressions" (2(X)7: 18). This is the insight of Rooth's Alternative Semantics account of focus (1985, 1992; Buring 1997 on extension to topics). I will assume that focus in Ni-e?kepmxcin is also associated with alternatives. In Rooth's theory, focus semantic values are generated alongside ordinary semantics values. The values for (lOB) are shown below; the focus semantic value is a set of alternative propositions in which the focused subject DP varies (eg. Michèle cooked dinner, Mabel cooked dinner, Bart cooked dinner). (15)  [BartIFOC cooked dinner. a. Ordinary semantic value:  [[Bartlpoc cook dinner]" = cook(Bart, dinner)  b. Focus semantic value:  [(BartIFOC cook dinner]' = {cook(x, dinner) | x £ Dg} where D^ is the domain of individuals  Jackendoff (1972) introduced the idea of a "presup," or what Rooth called a "presupposition skeleton." This is a lambda expression in which the focus of the sentence corresponds to a variable:  (16)  [Bârt]Foc cooked dinner. Presup:  Xx. x cook dinner  The Structured Meaning Approach to focus interpretation develops this idea, splitting the clause into a focus, and a background that corresponds to Jackendoff s presup (von Stechovi' 1990, Krifka 1992). The background is applied to the denotation of the focus to yield the ordinary meaning of the sentence; the background is a function taking the focus as its argument. Background and focus are represented as a pair, <Background, Focus>. (17)  <Xx. X cook dinner, Bart> Unlike Alternative Semantics, the Structured Meaning Approach allows direct access  to the focus denotation (Bart in this case) to yield the ordinary meaning. And, alternatives can still be introduced independent of the <B, F> structure in (17). We will see that narrow subject or object focus in Salish splits the clause neatly into focus and background, though the syntactic role of predicate and argument is reversed from the semantic relationship expressed in (17): the focus is made the predicate, while backgrounded information is a clausal argument of this predicate (Kroeber 1997). 1.3.3 The prosodie expression of focus Current theories on the marking of focus in stress languages generally have in common the correspondence between stress and focus. The observation is that focused material bears the dominant sentential or phrasal accent. I will call these "stress-focus" accounts, though "stress" is somewhat misleading. Assuming that "stress" is a property of prosodie words (Sluijter and van Heuven 1996a), and that pitch accents are assigned at the level of the p-phrase and i-phrase (Nespor and Vogel 1986), then "stress-focus" is better described as "accent-focus." For the sake of consistency with previous literature, I will continue to refer to "stress-focus" here. In any case, some examples of "stress-focus" accounts are shown below.  ( 18)  Proposals on the marking of focus a. Basic Focus Rule: A n accented word is F(ocus)-marked. (Selkirk 1995: 555) b. Stress-Focus Correspondence Principle: The focus of a clause is a(ny) constituent containing the main stress of the intonational phrase, as determined by the stress rule. (Reinhart 1995: 62) c. Focus: A Focus-marked phrase contains an accent. (Schwarzschild 1999: 173) d. F O C U S - P R O M I N E N C E : FOCUS needs to be maximally prominent. A prosodie category C that contains a focused constituent is the head of the smallest prosodie unit containing C. (Truckenbrodt 1995, Buring 2(K)3) e.  STRESS-FOCUS:  a focused phrase has the highest prosodie prominence in its focus  domain. (Féry & Samek-Lodovici 2006: 135-6) The correspondence between focus and stress has been assumed as a universal feature of stress languages (eg. Vaissiere 1995, Hartmann 2007): Intonation languages use pitch accents as the principal means of focusing. Most intonation languages use the H * L falling tone as a pitch accent to mark focus, where the * following the H tone signals that the tone on the accented syllable is high. (Hartmann 2007: 225) In the remainder of this subsection, I give a brief typology of possible focus strategies in stress languages. In English, the nuclear stress (underlined) surfaces on the focused constituent, without any change in surface word order. (19)  a. A : Who squashed a peach? B: [FrankIFOC squashed a peach.  [subject focus]  b. A : What did Frank do with the peach? B: Frank [squashedlppc the peach.  [verb focus]  c. A : What did Frank squash? B: Frank squashed ]a peach]FOC-  (object focus]  While English can mark narrow focus without changes in word order, Hungarian employs movement to bring the focused constituent into the default nuclear stress position.  Default nuclear stress is leftmost, on the verb bemuttatam^ as shown in the wide focus question and answer pair in (20) (Szendroi 2003). (20)  A : What happened? B: [Tegnap este  bemutattam  yesterday evening PRT.introduce.I  Petert Marinak.|poc Peter  "(Yesterday, I introduced Peter to Marylpoc"  Mary (Szendroi 2003: 71, ex. 55)  Narrow focus constituents {Marinak in 21B) move into a syntactic focus projection to the left of the verb - that is, to the nuclear stress position. While this has been conceived of as syntactically driven movement to satisfy a l+Focus] feature (Brody 1995), it also receives a natural explanation as phonological ly driven movement, to satisfy the "stress-focus" principles of (18) (Szendroi 2003; but see Horvath 2005 for a reply). (21)  A : Who did you introduce Peter to ? B: Tegnap  este  [Marinak^ Ippc  yesterday evening Mary  mutattam  be  Petert t^.  introduce.!  PRT  Peter t„  "Yesterday evening, I introduced Peter to lMary]poc" (Szendroi 2003: 65, ex. 45) Romance languages like Portuguese display a similar surface pattern (Cruz-Ferreira 1998, Costa 1998; Samek-Lodovici 2005 on Italian; Zubizaretta 1998 on Spanish), except in the other direction. Like Thompson Salish, as we shall see, nuclear stress is rightmost in default wide focus utterances (22). As expected under a stress-focus account, narrowly focused material appears at the right edge of the clause (23). (22)  A : What's going on? B: Eu prefiro que I  prefer that  ela  venha.  she  come  "II would prefer her to comejpoc-  (Cruz-Ferreira 1998)  ^ Left adjunctions like Tegnap este 'yesterday evening' can precede the verb/nuclear stress constituent.  (23)  A : Who would you prefer to come? B: Euprefiroque I  prefer that  venha [elajpoctg  come shCe  "I would prefer [herJFoc to come."  (Cruz-Ferreira 1998)  Finally, in German, non-focal constituents scramble away from the nuclear stress position (Krifka 1998b; also Neeleman and Reinhart 1998 on Dutch). This "selfless" movement allows focus constituents to surface in the nuclear stress site and carry sentential stress (Krifka 1998b). Sentential stress in German is immediately before the final verb; in (24) , the direct object focus den Roman 'the novel' is base-generated in this location. Thus, the focus carries the primary stress by default. (24)  A : What did Hans read to Maria? B: Hans  hat  der  Maria [den  Hans  has  the.DAT Maria the.ACC  Roman Ipoc  vorgelesen.  novel  read  "Hans read Maria [thenoyeiJFoc"  (Krifka 1998b: 88,ex. 33)  In (25), however, the indirect object der Maria is the focus. The direct object den Roman has scrambled leftward so that now der Maria surfaces before the verb and receives the nuclear stress. Once again, the focus carries primary stress, but this time as a result of scrambling of the non-focal constituent. (25)  A : Who did Hans read the novel to? B: Hans hat Hans has  [den  Romanic  [the.ACC novelji.  [der  Marialpoc  [the.DAT Maria]  "Hans read the novel to [MariaJFoc"  t^  vorgelesen.  \  read  (Krifka 1998b: 88, ex. 34)  In this section, I have illustrated three strategies employed in stress languages to satisfy the stress-focus correspondence. English leaves focus in-situ, and changes the location of the nuclear stress to the focus constituent. In Hungarian and Portuguese/Romance, focused constituents move to the edge of the clause where the nuclear stress is located (leftmost in Hungarian, rightmost in Portuguese). And in German, unfocused material scrambles out of the nuclear stress position so that the focus may surface there (see also Ishihara 2001: 172 for similar claims for Japanese).  (26)  "Stress-focus" strategies (i)  move the nuclear stress to the focus (English)  (ii)  move the focus to the nuclear stress position (Hungarian, Portuguese)  (iii)  scramble non-foci away from the nuclear stress site (German)  1.3.4 Focus projection Selkirk (1995) discusses the phenomenon of focus projection in English. A single rightmost nuclear stress can indicate a focus on a variety of ever larger constituents, each time "projecting" upwards through the syntax. The idea is that a listener is able to identify which constituent carries the nuclear stress, and use this information to help recover the focus of a speaker's utterance. Example (27d) illustrates the phenomenon of focus projection. Default nuclear stress falls rightmost in English (Chomsky and Halle 1968, Cinque 1993), on the deepest syntactic constituent. The nuclear stress on peach in (27d) can be used to answer any of the questions in (27a-c), which mark CP, VP, or narrow object focus. This is what Selkirk (1995) identified as focus projection: the primary stress indicates an f-mark on the object in (27d), by the rule in (28). This f-mark can optionally "project" upwards through the syntax from the object, to the verb, to the V P , and all the way to the CP, by the rules in (29). The focus is identified by the rule in (30). Thus, focus projects from the nuclear stress position. (27)  a. What happened?  [sentence-wide CP focus question]  b. What did Frank do?  [wide V P focus question]  c. What did Frank squash?  [narrow object focus question]  d. [Frank [[squashedf [a  peachf] ]f\ FQÇ. [optional focus projection]  (28)  Basic Focus Rule (English): "An accented word is f-marked."  (29)  Focus Projection (a) f-marking of the head of the phrase licenses the f-marking of the phrase (b) f-marking of the internal argument of a head licenses the f-marking of the head  (30)  Defining the Focus The Focus of a sentence is defined as an f-marked constituent not dominated by any other f-marked constituent.  (Selkirk 1995: 555, 561)  In this dissertation, I will only identify the focus ' F O C , ' and not individual f-marks in the syntax, though nothing hinges on this method of representation. Recent work by Buring (2003,2006) aims to do away with Selkirk's focus projection system. I'll briefly introduce some of the key points of his focus marking system here, as the issue will be relevant to the discussion in chapter 3 (section 3.4.2). Projection from heads to phrasal categories (29a) is what BUring calls "vertical" focus projection. In fact, Buring shows that Selkirk's rules for focus projection are too narrowly defined in (29), since they do not allow projection from accented adjuncts or specifiers. He restates Selkirk's rules of focus projection as the "restricted view:" (31)  Restricted vertical focus projection: Only heads and arguments can project focus.  (Buring 2003)  Buring gives empirical data for English and German showing that, in fact, any syntactic category (including transitive subjects, adverbs, and indirect objects) can project focus vertically to another dominating syntactic category. The breadth of the focus is indicated by the preceding wh-question, by the rule in (9). Yet, in each case below, the accent falls on a subconstituent not predicted by (31). As Buring notes, the more unusual accentuation patterns in the examples below are due to given material being deaccented; but this accent still projects focus vertically. (32)  Focus projection from accented transitive subject DP to C P Q: Why did Helen buy bananas? A : IcpBecause |ppJohn| bought bananas  (33)  (Buring 2003: ex. 17)  IFOC-  Focus projection from accented adverb to V P Q: What will she do if her call doesn't go through? A : She'll [vpcall him (Advpagain]Ipoc-  (34)  (Buring 2003: ex. 23)  Focus projection from accented indirect object to V P A : They accused Sinatra of having given money to the mob. What did Dean Martin do, to avoid having his image ruined, too? B: He Ivpgave [^pmoney] jppto fppthe Salvation Armyl|]poc. (BUring 2003: ex. 20)  Buring therefore proposes unrestricted focus projection: (35)  Unrestricted vertical focus projection: Any subconstituent can project focus.  (Biiring 2003)  It should still be noted that the subconstituent projecting the focus in (32-34) does so by virtue of carrying the most prominent phrasal accent. For the present study, we will want to know what categories count for focus projection in Thompson Salish, in terms of vertical focus projection. Neeleman and Reinhart (1998) describe this as the syntactic heads or projections which form the "focus set." The projection from arguments to heads, as described by Selkirk in rule (29b), has been termed "horizontal" focus projection by Buring. Buring argues that this asymmetry between heads and arguments is a matter of default prosody, and not special rules of focus projection (see also Truckenbrodt 1995, 1999). The asymmetry is regulated by a constraint against predicates bearing pitch accents (in chapter 4, we will see similar surface facts for MeTkepmxcin: nuclear stress falls rightmost, typically on arguments of the verb, and not on the lefmost predicate): (36)  *STRESS-PREDICATE:  Verbs/predicates/heads don't bear prominence.  (Buring 2003; see also Kahnemuyipour 2004, Selkirk and Kratzer 2007 on deriving this constraint) In wide focus cases (VP or CP), stress falls on the argument instead of the predicate, since by the rule below, the focused phrase must still bear prominence on a subconstituent: (37)  F O C U S - P R O M I N E N C E : FOCUS  needs to be maximally prominent.  A prosodie category C that contains a focused constituent is the head of the smallest prosodie unit containing C. (Truckenbrodt 1995, Buring 2003) Crucially, prominence is still defined as phrasal accent, carried on prosodie heads. Since prominence is assigned by default prosody, though, one might wonder why the focus rule needs to refer to prosodie heads at all. This matter will be taken up in chapter 8. For now, we can note that in Nte?kepmxcin we still expect horizontal focus projection from arguments (which we will see bear the nuclear stress when rightmost) to heads (the predicate).  1.3.5 Focus summary and preview of results In this section, I have reviewed some basic ideas about focus that feature prominently in the literature. I will assume that focus is a syntactic category marked by a FOCUS feature, and that semantically it is associated with alternatives. The data presented in this dissertation will, however, reveal trouble for the "stressfocus" accounts. Focus projection will necessarily also proceed in a different manner in Nte?kepmxcin. In stress languages, focus has been widely held to bear the most prominent phrasal accent (the nuclear stress). This phrasal accent can "project," or indicate to a listener that the focus is equal to one of an ever larger group of dominating syntactic constituents. This prosodie focus marking enables listeners to keep track of or recover focused information in the discourse; in Krifka's expression focus cases, this amounts to managing the content of the Common Ground of a discourse. In cases of Krifka's denotation focus, the prosodie marking will be required for the proper interpretation of the truth conditions of the utterance. I will show that the stress-focus correspondence fails to account for focus marking or focus projection in Me?kepmxcin. Narrow focus utterances employ the same prosodie patterns as neutral, all-new utterances. Focus does not receive additional prosodie prominence, nor is given material deaccented. As far as focus projection is concerned, the focus projects from the first phonological phrase in the utterance containing the matrix predicate. The focus may be equivalent to this p-phrase; it may be a subconstituent of this pphrase (even if it does not bear the phrasal stress, unlike English); or, in the cases of wide focus, the focus may itself contain the first p-phrase. A n example is shown on the next page.  The contrastive focus Karsten is the leftmost lexical element in the clause, it does not carry a large pitch excursion; compare Rois in the English example in figure 1.2. And, the given material ?ex ncewm is not deaccented: pitch is reset to a higher level on this phrase, and the stressed verb also carries an amplitude peak (again, compare given material that called me today in the English example below). Thus, nuclear stress falls on the rightmost lexical element here, while the focus is the leftmost lexical element. The focus e Karsten is the only lexical information in the first p-phrase, so a listener can recover the focus by identifying the first p-phrase.  •mm amplitud peak paijtial decliijiation reset,—.  lack of additional prominence  cé  xe? e  CLEFT  DEV D  0.5  Figure 1.1  [KârstenJFoc Karsten  [?éx PRG  1  Time (s)  n-céw-m]G LOC-wash-MDL  1.5  Pitch tracing and waveform: "[No, I it was I Karsten Ippr: [that washed themlo."  1.99406  In the English example, contrastive focus on Rois is marked with a large pitch excursion and amplitude peak. After the focus, FO drops more than 50 Hz (-4.8 semitones) from the peak on Rois, to the rightmost word of the utterance, today; amplitude peaks drop from 73 db to 63 db, a -10 db change. This pattern of H * L prominence on the focus and post-focal deaccenting is widely reported for stress languages; see Hartmann (2007) for discussion.  Q.  No, it was [Roislpcx: [that 0  called  0.5  me 1  todayJc. 1.61086  Time (s)  Figure 1.2  Contrastive focus accent and post-focal deaccenting in English  The discussion of these figures brings us to to the next section, where I review givenness.  1.4  What is givenness? Roughly speaking, given material is old information that is already in the Common  Ground of a conversation. 1.4.1 A syntactic representation Given material is also assumed to carry a syntactic feature, the given feature ' G ' (Féry and Samek-Lodovici 2006, Selkirk 2007, Selkirk and Kratzer 2007). Like the FOC feature, G mediates between the phonological expression of given material on the one hand and its  semantic interpretation on the other. I will assume that given material is syntactically marked, and I will indicate this with a subscript ' G ' where relevant. 1.4.2 The semantics of givenness (Schwarzschild 1999) Schwarzschild ( 1 9 9 9 ) proposes that given constituents are entailed by prior discourse. He gives the following definition for G I V E N material. I give both the informal and formal version. (38)  Definition of G I V E N (informal version) An utterance U counts as G I V E N iff it has a salient antecedent A and a. if U is type e, then A and U corefer; b. otherwise: modulo 3-type shifting, A entails the Existential F-Closure of U . (Schwarzschild 1 9 9 9 : 1 5 1 )  (39)  Definition of G I V E N (formal version) An utterance B counts as G I V E N iff it has a salient antecedent A and a. if the semantic type of B is e, V(w, g) S c 3h ( [[A]f = [[B]f-^] h. if the semantic type of B is conjoinable: V<w, g) Gc 3h |ExClo(ffAl]^)(w) ^ ExClo([[B|]«'')(w)l (Schwarzschild 1 9 9 9 : 1 5 2 ) In the example below (from Schwarzschild 1999), John is given in the question,  while the relation praise is not. That is, the question does not entail 3x3y[y praised xj. Therefore, the nuclear stress falls on the verb praised and not on him. (40)  A : What did John's mother do? B: She [[praised] himclpocThe next case has the nuclear accent on the given pronoun him. The pronoun is also  the focus, though, since it answers the preceding wh-question. Moreover, given material entailed by the preceding question includes She, and crucially 3x[She praised x), where x corresponds to the wh-word Who. Crucially, him is not in a given relationship with other material: the question does not entail She praised him, 3y[y praised him), or 3R[She R-ed him). Thus, focal accent falls on him (see Schwarzschild 1 9 9 9 : 1 5 7 - 1 6 1 for the full account).  (41)  A : Who did John's mother praise? B; She praised IhimJoFocSchwarzschild's treatment also allows non-constituents, such as her [...] convertible  below, to be identified as given (1999: 146-147, 161-162; Taglicht 1982, 1984). (42)  A : John drove Mary's red convertible. What did he drive before that? B; He drove [her [bluejpoc convertiblejpocThe sentence John drove Mary's red convertible entails, for example, the NP:  3X3P[P(her X convertible)] (Schwarzschild 1999: 161). The adjective blue is not entailed, so focal accent falls on blue. Note that the fact that given material can be a non-constituent is problematic for a syntactic G feature (section 1.4.1). This issue is beyond the scope of the present study, however, and I will continue to use a subscript G as a convenient way to identify given material where necessary. Schwarzschild's definition of givenness allows us to identify given material in the Common Ground of utterances, but does not identify degrees of givenness. Work by Baumann and Grice (2006) has suggested that the degree of accessibility can also affect the prosodie marking of given material. In German, given material in whole-part relationships with priorly mentioned material, or given material that is predictable from context but not previously mentioned, may carry some level of accentuation, but a different accent from new information. Given material in this dissertation is compatible with Schwarzschild's definition, though I will not be providing any detailed semantics to illustrate. For the sake of clarity, I will just use subscript ' G ' to indicate given material where necessary. 1.4.3 The prosodie expression of givenness Krifka (2007: 37-40) identifies three ways (in addition to the use of anaphors) to mark givenness: Deaccentuation, the reduction of the prosodie realization of expressions that are given in the immediate context; deletion, which can be seen as an extreme form of reduction; and the realization of an expression in a non-canonical position, typically before the canonical position. Given material in Thompson Salish can be deleted (Kinkade 1983, Gerdts and Hukari 2004 on 3"* person transitive subjects being null and topical, for example). N-teTkepmxcin  speakers also use word order to mark given material (though, in narrow focus cases, given material appears after its normal canonical position rather than before). However, deaccenting of given information is not routinely attested. This suggests that there is no link between deletion and reduction as suggested by Krifka in the above quote; rather, D E S T R E S S GIVEN  and D E L E T E - G I V E N are two different processes. In stress languages, much attention has been paid to the deaccenting of given  information, particularly after the nuclear stress position. This generalization has been strongly stated by Hartmann (2(X)7), who I quote again here: Another very general feature of focus intonation is the drop in pitch after an early nuclear accent. The postfocal contour is deaccented, due to the fact that there are no more accent targets following the focus. Thus, the pitch range, which is expanded on the focus constituent, is compressed post-focally. (Hartmann 2007: 225-226) These observations are illustrated by the English sentence in figure 1.2. The "deaccent-given" generalization is formally described in the constraint D E S T R E S S - G I V E N proposed by Féry and Samek-Lodovici (2006):* (43)  DESTRESS-GIVEN:  A given phrase is prosodically non-prominent.  (Féry and Samek-Lodovici 2006:135-6) 1.4.4 Givenness summary and preview of results I will identify given material with a subcript G, and assume Schwarzschild's account for the semantic identification of given information. In terms of the prosodie marking of given material, we shall see that focus in Thompson Salish is associated with an early position in the clause (leftmost), yet post-focal material does not appear to be deaccented, as Hartmann suggests for stress languages. This was already illustrated in figure l . I . Deaccenting of given information does not appear to be a universal property of stress languages. Gumperz (1982) identifies Indian English and Caribbean English as nondeaccenting languages (see also Ladd 1996 on Italian, Romanian, and an overview of  * It should be noted that this constraint fails to account for the observation expressed by Hartmann that given information is deaccented above all post-focally (BUring, p.c); that is, D E S T R E S S - G I V E N is not a directional constraint. However, this is a general problem for "destress-given" approaches as far as I can tell, and I do not address the issue further.  languages that lack deaccenting; Ortiz-Lira 1993, 1995 on Spanish). The following example from Gumperz (cited in Ladd 1996: 176) illustrates: (44)  If you don't give me that cigarette. I will have to buy a cigarette. While the second instance of cigarette is clearly given, it continues to carry the  nuclear stress. For a speaker of a standard variety of English, this accentuation pattern induces the interpretation that the second instance of cigarette refers to a different kind of entity than the first. Compare the typical intonation pattern found on such cases, where accent falls on buy rather than the final instance of cigarette: (45)  If you don't give me that cigarette, I will have to buy a cigarette. Similar examples can be constructed in Thompson Salish. Example (46), notably, was  elicited as a translation from the English counterpart, where the pronoun it marks the given status of the final object DP (and lacks phrasal accent). Despite the potential bias toward the speaker producing an English intonation pattern in the Salish version, however, sq'^iyt 'fruit' is not only produced sentence-finally, but is also accented in both occurrences in the Nte?kepmx version. Neither occurrence of sq^iyt (shown in the dashed boxes) has reduced pitch or amplitude curves in figure 1.3; the nuclear stress remains rightmost. (46)  When Mary is hungry, she picks some fruit and then she eats it.  e  teyt us e Mary ?É snésc ( ["'yéwm tl : sq^iyt s?ùpis  DET hungry 3 D Mary &  pick  go  2  Figure 1.3  D  Time (s)  fruit  3  & eat  [sq-iyt]G DIT fruit  4.92594  Pitch tracing and waveform: "When Mary is hungry, she picks some fruit and then she eats [the fruitlc."  1.5  Representing S T R E S S - F O C U S and D E S T R E S S - G I V E N phonologically The marking of focus in "stress-focus" accounts is through alignment with the  prosodie heads of phrases: phonological phrases, and intonational phrases. For the purposes of this dissertation, I will adopt the prosodie hierarchy proposed by Nespor and Vogel (1986).  (47)  The prosodie hierarchy (Nespor and Vogel 1986, Hayes 1989) Utterance (U) t Intonational Phrase (i-phrase) I  Phonological Phrase (p-phrase) I  Clitic Group (cl-gp) I  Prosodie Word (PWd) I  Foot (Ft, «I>) I  Syllable (a) Prosodie categories are in an exhaustive, hierarchical relationship. Each prosodie constituent (from the foot onward) has a single head, at the left or right edge. These ideas are captured in the Strict Layer Hypothesis (Selkirk 1984, 1995b, Samek-Lodovici 2005). For this dissertation, I adopt the following constraints: (48)  Conditions on prosodie structures: a.  HEADEDNESS:  Each prosodie constituent has one and only one head, at the left or right edge, (see McCarthy 2003:111 on "End Rule" constraints) b. L A Y E R E D N E S S : no prosodie constituent is dominated by a constituent lower in the prosodie hierarchy. Example: A syllable (a) does not dominate a foot (<î>). c. E X H A U S T I V I T Y : no prosodie constituent immediately dominates a constituent that is not immediately below it on the prosodie hierarchy. Example: A PWd does not immediately dominate a syllable (a). d. N O N R E C U R S I V I T Y : N O prosodie constituent dominates a constituent of equal rank in the prosodie hierarchy. Example: No foot ($) dominates another foot (<ï>).  Selkirk (1995b) notes that while H E A D E D N E S S and L A Y E R E D N E S S appear to be universally undominated, E X H A U S T I V I T Y and NONRECURSIVITY are sometimes violated. I will primarily be concerned with p-phrases and i-phrases in this dissertation, since this is the level where phrasal accent is assigned. Pitch accents are assigned at the p-phrase, and the nuclear pitch accent at the i-phrase. Prosodie heads build on each other at each level of the prosodie hierarchy; thus, p-phrase heads are built on Prosodie Word heads, PWd heads are built on the head foot of that word, and so on. In English, phrasal accents fall on the rightmost head in their phrase, as expressed in the constraints HP and HI ( H E A D P - P H R A S E , H E A D I-PHRASE)  below (from Féry and Samek-Lodovici 2006:134, adapted from McCarthy  and Prince 1993, Truckenbrodt 1999). Thus, HP and HI account for the observation in HEADEDNESS  in (48a) that phrasal heads fall at either the left or right edge of a prosodie  phrase, and not on potential intermediate heads. (49)  a. " H P : "  Align the right boundary of every p-phrase with its head. ALIGN(P-PH, R ; PHEAD, R )  b. "HI:"  Align the right boundary of every i-phrase with its head. A L I G N ( I - P H , R ; IHEAD, R )  In "stress-focus" languages like English, the focus carries both the p-phrase accent and the i-phrase accent. In the object focus example below, the focus peach carries the p-phrase accent and the i-phrase accent. The verb phrase squashed a peach is parsed into one p-phrase, while the subject Frank is parsed into another (eg. Kahnemuyipour 2004, Selkirk and Kratzer 2007). (50)  A : What did Frank squash?  [object focus |  (  X  )  i-phrase  [nuclear pitch accent]  ( X )(  X  )  p-phrase  [pitch accent]  B: Frank squashed  [apéaçhjpoc-  The representation of "destress-given" is less clear at the outset. There are two possibilities: given material may still carry a p-phrase accent, just so long as the nuclear pitch accent has shifted elsewhere; or given material carries no phrasal pitch accent at all, just word-level stress. These two possibilities are represented in (b) and (c) below. In (b), the nuclear pitch accent has shifted to the middle p-phrase headed by the verb, to satisfy "stressfocus." In this case, the given object peach still carries phrasal accent, but less than when it is  itself the object focus in (50B). In (c), the given object peach carries no phrasal accent at all; this achieved, roughly, by not parsing peach into a p-phrase at all (Selkirk and Kratzer 2007; a more detailed summary of their account is given in section 1.5.1). (51)  a. A : What did Frank do with the peach? [verb focus]  b.  (  X  ( X ) (  X  )  ( X  )  i-phrase  )  p-phrase  [hypothetical phrasing]  B: Frânko |squashed ]FOC the péacho X  ( (  X  ) )  i-phrase p-phrase  c. B: Frânko [squâshedlppc the péacho. I will argue that, at least for in situ focus languages like English where "destressgiven" is active, a representation along the lines of (c) is what we must adopt (I'll revise it somewhat in section 1.5.1). This is the position implicit in, for example, Hartmann (2007): "The postfocal contour is deaccented, due to the fact that there are no more accent targets following the focus." Lacking "targets" means that there are no more potential p-phrase heads after the focus.^ Because given material appears post-focally in Thompson River Salish, (c) is also the representation assumed if "destress-given" is active in NieVkepmxcin. There are two reasons to suppose (c) is on the right track. First, phonologically, a structure like (b) is ill-formed: this is because the i-phrase head falls on the middle of three p-phrases. Since prosodie heads fall at either the left or right edge (that is, phonology cannot count), structure (b) is not possible. In (c), where peach is not parsed into a p-phrase, this problem does not arise - the i-phrase head simply falls on the rightmost p-phrase head. Secondly, numerous scholars distinguish the prosodie properties of pitch accents at the phrasal level (p-phrases and i-phrases) from word level stress (Halliday 1967b, Vanderslice and Ladefoged 1972, Beckman and Edwards 1994, Sluijter and van Heuven 1996a, 1996b, Astruc and Prieto 2006). For example, Sluijter and van Heuven (1996a, ^ It should be noted that the "destress-given" constraint, as far as I can tell, has no account for why deaccenting of given information appears to be primarily restricted, above all, to post-focal deaccenting (BUring p.c.). This is a general problem that I will not be addressing in this dissertation, as it is not crucial for the present study.  1996b, 1997) give phonetic evidence that phrasal accent differs from pure word-level stress. The former is marked primarily by additional pitch movements (eg. phrasal pitch accents) in comparison to word-level stress. Deaccented given material, crucially, lacks the pitch movements characteristic of phrasal stress. This again suggests that a representation along the lines of (c) is correct. That is, "destress-given" is truly lack of phrasal pitch accenting on given material. The representation in (c) can be achieved purely by appealing to "destress-given," and does not require reference to a "stress-focus" constraint. That is because, once given material is deaccented, the stress will fall on the focus by default. We may wonder if we need representations for both given and focused material, then, or if just one will do. However, there is evidence that we need both categories. First, in many languages contrastive focus is distinguished from new information focus (eg. Gussenhoven 2004, Féry and Samek-Lodovici 2006, Selkirk 2005, 2007, Selkirk and Kratzer 2007). While "destress-given" can account for stress falling on new information focus cases (as in 50c), we thus still need a category of contrastive focus. There are also instances where a constituent can be both given and focused. The following example is from Schwarzschild (1999: 172). Here, the final DP the rising of the tides is given since it is also the subject of the first clause; yet it is still accented, by virtue of being focused (see also Féry and Samek-Lodovici 2006: 133, for discussion). (52)  The rising of the tides depends upon the moon being full and the boat being empty depends upon [the rising of the tideslppc g. Thus, we need both discourse categories, focus and givenness. Recent work by  Selkirk (2007; also Selkirk and Kratzer 2007) suggests a three-way split, with focus marking for contrastive foci, givenness marking, and new information focus the unmarked category. However, in this dissertation, I will continue to mark all focus with the subscript ' F O C ' I will claim that in Thompson River Salish, there is no "destress-given" constraint active, so information structure is necessarily marked by appealing to the category of focus; furthermore, contrastive focus and new information focus are treated with the same constraints.  Table 1.1  Focus and givenness marking in English and Salish Focus marking  English  Givenness  Contrastive  New information  (Destress-given)  X  —  X  Salish  X  —  I will show that focus in Thompson River Salish is marked by alignment to prosodie edges, while in English focus is marked by alignment to prosodie heads ("stress-focus"). The generalization that emerges is that languages which do not deaccent given material also do not have a "stress-focus" constraint; instead, alignment to prosodie edges is the key to focus marking.  1.5.1 Phase theory and a two step parse into p-phrases In this dissertation, I'm going to be assuming a two-step process in the interface of syntactic phrasesand phonological phrases. This will give us a slightly more refined representation of (51c). Numerous theorists have proposed that, at the interface of syntax and phonology, prosodie phrases are parsed into phonological phrases (p-phrases) (Truckenbrodt 1995, 1999, Legate 2003, Fox and Pesetsky 2005, Ishihara 2007, Selkirk and Kratzer 2007). Prosodie parsing has thus been proposed to be a two step process (Legate 2003, Selkirk and Kratzer 2007). In Step 1, syntactic phrases are parsed into p-phrases at the interface of syntax and phonology. In Step 2, the remaining prosodie units (PWds, i-phrases, etc.) are parsed by the phonological component alone. I will adopt this model because of its restricted view of the syntax-phonology interface (eg. syntactic phases are exported only to p-phrases, and not other phonological categories); and because discourse constraints like S T R E S S - F O C U S and DESTRESS-GIVEN  are assigned as interface constraints to Step 1.1 will be appealing to other  interface constraints in this dissertation. In the case of given material, Selkirk and Kratzer (2007) suggest that DESTRESS-GIVEN  is an interface constraint which prevents given material from being parsed  into a p-phrase at the interface of syntax and phonology. Instead, given material is parsed recursively into p-phrases in Step 2, by the phonological component. This means that it shares a phrasal head with the focused p-phrase of Step 1 ; since p-phrases can have only a single head, no further accents can fall on given material. Thus, the representation of (51c) in this two-step interface model looks as follows:  (53)  p-phrases at the interface and in the phonological component: two steps X  )  i-phrase: STEP 2  (  X  )  recursive p-phrase: STEP 2  (  X  ( (  )  p-phrase at interface: STEP 1  Franko [ squashedJpoc the péachoAgain, this approach doesn't capture the fact that D E S T R E S S - G I V E N applies above all to post-focal information, but that is a general problem that I will not be addressing in this dissertation.  1.6  Bilingualism and intonation The present study is based on conversational data from two female speakers of  MeTkepmxcin; in both cases, NieVkepmxein was the first language they learned. However, both are also fluent speakers of English. This raises the possibility of cross-linguistic "contamination" of the intonational properties of Nte?kepmxcin observed in the collected data. I review some literature on this subject in the present section. In short, there are no clear guidelines to be drawn. It may be that English intonation influences the intonation that speakers of Nte?kepmxcin employ, but it is equally possible that their dialect of English is influenced by features of Nte?kepmxcin intonation. However, the statistical results of the phonetic study in chapter 5 indicate that the intonational properties of Nte?kepmxein are significantly different from those of English. There has been very little research on intonation in bilingual speakers. The question is important since speakers may incorporate the intonation system of one language into the other. In a "substratum" theory of language contact, intonation is a key feature of the native language that persists in the non-native language, while "adstratum" proponents find that immigrants (for example) incorporate prosodie features of the target language into their native language (Vildomec 1971, discussed in Thomason and Kaufman 1988:9). In the present study, there are four possible results of language contact between N'teVkepmxcin and English, as far as intonation is concerned. First, bilingual speakers may have no noticeable cross-linguistic influence in their intonation systems. Secondly, features of English intonation may have been incorporated into N+e?kepmxcin. Thirdly, intonation features of NfeVkepmxcin may have been incorporated into the variety of English spoken in Lytton. Note that these latter two possibilities are not mutually exclusive, such that a fourth  possible result is that bilingual speakers may borrow features from both languages and come up with a single intonation system for both of the languages they command. Colantoni and Gurlekian (2004) analyzed declarative sentences for two speakers of Buenos Aires Spanish. They found that the prenuclear pitch accent has early alignment, unlike other varieties of Spanish, which have a late-aligned accent (though they note that the early alignment is used in other environments, namely to signal contrastive focus, in these other varieties of Spanish). In addition, Buenos Aires Spanish has pronounced down-step on postnuclear peaks in intonational phrases (DESTRESS-GIVEN), again unlike other varieties of Spanish (eg. Ortiz-Lira 1993, 1995). The authors conclude that these features of Buenos Aires Spanish have been transferred from Italian, including through bilingual populations during the 20* century. It should be noted, however, that it is very difficult to establish the intonational features both of early 20* century immigrant Italian or Buenos Aires Spanish from written descriptions alone, so these results are speculative at best (McMahon 2004). Moreover, Colantoni and Gurlekian's hypothesis is based on examination of descriptive statistics only, not significance tests; and as the authors note, they looked only at FO but not other potential indicators of phrasal accent like duration or amplitude. McDonough (2002) compared tonal contours of yes/no questions, focus constructions and declaratives in Navajo, a tone language. Again, the author inspected descriptive statistics only, and only for FO. The investigation examined the claim that Navajo lacks focus or other tonal intonation, and found no evidence for pitch perturbations at edges or elsewhere to differentiate focus or yes/no structures from normal declaratives. McDonough attributes the lack of intonational focus in Navajo to its being a pronominal argument language (Willie and Jelinek 2000), and to the fact that languages tend to mark accents on arguments over predicates (eg. Bollinger 1986, Schwarzschild 1999, Buring 2003,2006, Kahnemuyipour 2004, Selkirk and Kratzer 2007). Regarding bilingualism, McDonough's study shows that Navajo intonation is quite different from English, suggesting a lack of influence from American English intonation patterns. Queen (2001) looked at Turkish-German bilinguals' intonation in interrogatives, focus constructions and phrase-final rises. This study is also based on descriptive statistics. Subjects were four 10-12 year-old bilinguals, two monolingual German children and one adult male Turkish speaker. Queen found that the degree of intonation mixing varied both by speaker and by structure. Thus, bilinguals maintained language-specific interrogative marking. The author speculates that this is due to "the particular morphosyntactie patterns that govern interrogative formation in the two languages" (2001: 68), since the German intonation pattern was maintained even in code-switching cases where a German wh-word was used in an otherwise Turkish utterance; or when the Turkish question-particle was used  in a German matrix utterance. In focus constructions in the bilinguals' German, the speakers employed a non-German postposing, which was accompanied by Turkish intonation when it occurred (this finding was based on 9 sentences). Finally, for phrase-final rises, the bilinguals used both Turkish L % H % and German L * H H % contours in both German and Turkish. These contours appear to have taken on meanings of "discourse continuation" and "discourse cohesion" respectively, though results are not consistent across the four speakers. The results suggest that bilinguals may mark their bilingual status with intonation contours that differ from either monolingual language, but in specific constructions rather than across-the-board. These findings underline Thomason and Kaufman's (1988) remarks on the importance of the sociolinguistic situation when considering language change through contact. Similar findings were obtained by Cichoeki and Lepetit (1986) in their examination of declination in French-English bilinguals in Welland, Ontario. Cichoeki and Lepetit studied 14 bilingual grade 5 children in three groups: high French use (n=4), equal French-English use (n=6), and high English use (n=4). A l l children attended a francophone school. They had subjects read 30 semantically unrelated sentences varying in length from three to nine syllables. Using a multivariate statistical analysis, the authors determined that the three groups used different declination rates when the sentence-initial subject was a pronoun (including French ce), rather than a full lexical NP, in which case declination did not differ significantly for the three groups. When sentences started with a pronoun, speakers in the French-dominant group (F) showed least declination, those in the French-English group (FE) used greatest declination, and subjects from the English-dominant (E) group were in between. The authors hypothesized that the F E and E speakers allowed the pronoun to carry a high tone H , not possible for the F group. They concluded that the "important result for declination theory is that a syntactic constraint is operative in intonation" (1986: 245). They note that the F E group in fact leads the E group in the use of the "putatively more Englishlike variant" (245), suggesting that "the F E group may be insisting on the relative strength of their bilingualism within the community" (1986: 245). Unfortunately, the authors do not provide statistical results to indicate which of the groups, FE, E or F, are significantly different from each other, merely that they are not all identical; they do note that "the significant effect is due to the F E group" but give no statistical figures. What these results show, like Queen's (2001) study of Turkish-German bilinguals, is that cross-linguistic influences in intonation can be linked to a specific syntactic construction, and can result in a different intonational pattern from that used in either source language. The results also suggest that declination can be subject to language or dialect-specific phonetic implementation.  Penfield (1984) analyzed recordings of spontaneous English conversation among bilingual Mexican-Americans from El Paso. In a purely descriptive study, she found that the intonation contours in "Chicano English" differed from Standard English in a variety of ways (she did not explicitly attribute these contours to the influence of Spanish). For example, Chicano English speakers use sentence-final rising contours to signal emphasis or contrast on the final word in declaratives, which would get a falling contour in English. Penfield identifies five such intonational features of Chicano English, and, like Cichocki and Lepetit (1986), concludes that "prosody not only marks ethnic membership but, in the case of codeswitching Chicanos, it also marks the degree of identification with this membership" (1984: 57). Finally, Atterer and Ladd (2004) looked at German speakers' intonation of English, and found that they timed the anchoring of English L + H contours using their native German timings. Though these were second language learners of English and not bilinguals, the results suggest that there is a global L + H contour, which German speakers recognize in English, but implement using the phonetic timing specific to German. Overall, the results do not give any clear indication of what kind of influence, if any, we ought to expect from the effect of bilingualism on the intonation of Nie?kepmxcin. The most common theme in the literature reviewed above is that bilingual speakers can employ particular intonational features, in particular constructions, as a sociolinguistic marker. However, this does not help us identify potential influences in the present study. I thus confine myself to describing and analyzing the prosodie properties of focus constructions in MeVkepmxcin, with frequent comparison to English. The results suggest that prosody plays out quite differently in N'teVkepmxcin than in standard dialects of English.  1.7  Intonation in Salish Intonation has not been documented to very great extent in any of the Salish  languages. To be sure, some grammars make cursory observations (to which I'll return below for N"le?kepmxcin; but see Montler 1986 on Senchothen), but these typically don't allow us to draw theoretical cross-linguistic comparisons, nor are they accompanied by detailed examples or phonetic analysis. Some exceptions among more recent work are Bennett and Beck (1998), Beck (1999), and Beck and Bennett (1997) on discourse prosody in Lushootseed Salish; Jacobs (2007) on question intonation in Squamish; Bar-el and Watt (1998) and Watt et al. (2000) on correlates of stress (and phrasal accent) in Squamish; Benner (2006) on Senchothen prosody; Caldecott (2006) and Oberg (2007) on St'at'imcets; and Barthmaier (2004) on prosody in Okanagan discourse.  Some general observations that have emerged are that the Salish languages generally have similar acoustic correlates of stress as you would find in stress languages like English: higher pitch, greater amplitude and longer duration are associated with accented syllables (Bar-el and Watt 1998, Watt et al. 2000, on Squamish; Benner 2006 on Senchothen, Caldecott 2006 on St'ât'imcets, Thompson and Thompson 1992 on Nfe?kepmxcin). A notable exception is Upriver Halkomelem, which has lexical tone (Galloway 1991, 1993; whether this is tone, or a pitch accent type system like Swedish or Japanese, is not clear - see Brown and Thompson 2005, 2006, for more discussion). Because these observations on "stress" are based on words produced in phrases (at a minimum, the phrase containing the single word), they actually reflect the properties of phrasal accent rather than word-level stress (eg. Sluijter and van Heuven 1997, 1998, for discussion). At the phrasal level, prosodie phrases generally follow a declination of amplitude and FO from left to right. The beginnings of prosodie phrases can be marked by complete or partial reset of declination, and boundaries are also marked by pauses of various lengths (Beck 1999, Beck and Bennett 2007 on Lushootseed, Barthmaier 2004 on Okanagan, Benner 2006 on Senchothen). Edge tones (in the sense of Pierrehumbert 1980, Ladd 1996, etc.) can mark continuation or uncertainty (see the discussion of contrastive topics in chapter 2, section 2.2.4). For example, the descriptions of phrase level prosody in NteVkepmxcin provided by Thompson and Thompson (1992) describe general declination patterns and phrase final boundary tones (what they call "phrase-end melodies" - 1992: 24). They identify four general patterns: (54)  General prosodie patterns in some MeVkepmxcin clauses (Thompson and Thompson 1992: 24) (i)  non-terminal clauses: "ends below mid-range without dramatic rise or fall"  (ii)  general (sentence-final, including "factual questions"): "mid-high with last primary stress, abrupt drop to low"  (iii)  consulting: "mid-high with last primary stress, light rise: request for confirmation, check on validity of assumption"  (iv)  inconclusive: "mid or high-mid with last primary stress, somewhat lower pitch on any following syllables, but no fall to low: disinterest, incompleteness, non-final item in a series (common in prepared recording of citation forms)"  The descriptions don't tell us much about focus, pitch accents, or any relationship that exists between the two. That is the topic of the present study. Thompson and Thompson  remark that what is "needed is deeper study of discourse patterns and the intonational phenomena that accompany them" (1992: 183), a challenge that this dissertation takes up.  1.8  Methodological, empirical and theoretical contributions The current research makes significant empirical, methodological and theoretical  contributions to language research in the Salishan, and other Amerindian, languages. Because I am examining the discourse concepts of focus and givenness, the analysis is based on conversational language data. However, this data first had to be collected through original fieldwork, a process which included the application of new elicitation methodologies in the fieldwork setting. Empirically, therefore, this project contributes a significant amount of new, high quality recordings and transcriptions of natural and elicited dialogue in Nte?kepmxcin. This is significant because most fluent speakers of NteVkepmxcin are in their 60's or older. At the same time, very few texts of conversations between fluent speakers of the language have been collected (or at least, collected, published and analyzed). Numerous traditional narratives involving a single speaker recounting a story have been recorded by Thompson & Thompson, and Egesdal (for example, see Thompson & Thompson 1992:199-227; Thompson and Egesdal 1993; Egesdal 1984), yet naturalistic dialogue remains an empirical gap which this project has begun to fill. These conversational texts will be of broad interest to a variety of researchers, and some examples are given in the appendix. Moreover, because of their nature as "everyday conversations," these texts also provide an excellent potential resource for teaching materials. Absence of conversational texts is not limited to NteVkeprnxcin, but is a broader empirical gap in the documentation of many Amerindian languages, given the traditional methodology of single speaker text collection (eg. the work of Franz Boas, James Tait and other early field linguists in the Pacific Northwest). More recent research has often focused on single sentence elicitation, as part of a Generative Linguistics approach to elicit both positive and negative data, often of more complex sentences or structures not likely to be found in traditional texts. Such data also does not provide the sort of stretches of discourse which are necessary for the examination of discourse notions like focus and givenness (at least not a priori; we shall see in chapter 5 (section 5.3] that language consultants are able to closely approximate discourse conditions in single sentence utterances when provided with sufficient context (see also Matthewson 2004(). Thus, by contributing conversational data, this project addresses an empirical gap in Amerindian language research more generally (but see for example Muehlbauer 2(X)5, Cook and Muehlbauer 2005, Muehlbauer and Cook 2005 on prosody in Plains Crée discourse).  Methodologically, the current research required the development of new techniques for gathering conversational texts in a fieldwork setting. The corpus analysis of chapter 3 and the phonetic studies in chapters 4 and 5 are based on a corpus of new conversational recordings. The recordings fall into three basic categories. First is spontaneous conversation (Appendix B, C). Second is scripted conversation (Appendix A). Third is single sentence elicitation (How do you say X?). The first two categories required new fieldwork techniques. Because language consultants often feel uncomfortable or unable to produce conversation on demand, the current research developed methodologies to facilitate conversation. Spontaneous conversation was prompted in several ways: providing discourse topics (Appendix B), developing audio-visual material for speakers to respond to (Burton 2005, Koch 2007c, Caldecott and Koch 2007, Koch and Caldecott 2007 - Appendix C), or engaging in "everyday" activities (playing games, looking at photographs, etc.). While these techniques resulted in large volumes of naturalistic data, more targeted techniques were used to elicit specific discourse structures: in scripted conversation, consultants were provided with a dialogue (in English), and asked to role play, performing the dialogue in Nte?kepmxcin (Appendix A). Finally, single sentence elicitation was used to check very specific sentences, and to provide negative data as well. While the methodologies themselves will be familiar from psycholinguistic or laboratory phonetic work on more common languages like English (eg. Anderson et al. 1991 on the "map task"), their application in a fieldwork setting is more novel. These methodologies will be relevant for researchers investigating things like information structure, discourse coherence, and intonation, not only in other Salish languages, but in fieldwork research in general. Finally, the current research makes new theoretical contributions in the study of intonation and information structure in conversation in Nte?kepmxcin. This research will add to the typological knowledge of a small but growing body of literature on intonation in Salish (eg. Bennett & Beck 1998, Caldecott 2006, Benner 2006, Jacobs 2007), as well as the study of intonation cross-linguistically. The lack of a stress-focus correspondence contradicts what has previously been considered to be a universal in stress languages (Vaissière 1995, Hartmann 2007). Instead, it suggests that Lindstrom and Remijsen's observations are on the right track, made in their study of Kuot, another stress language that does not employ pitch accent to mark discourse information: "English and other well-studied European languages [may be| simply typologically unusual in the extent to which intonation expresses speaker attitude" (2005: 843-4).  Chapter II: NteVkepinxcin Basics In this chapter, I establish that NteVkepmxcin is a stress language with rightmost nuclear stress, and a default word order of (Aux)VSO. Nie?kepmxcin is a member of the Northern Interior branch of the Salish language family. It is spoken in the southwest of British Columbia, Canada, in an area bounded roughly by the Fraser and Thompson Rivers to the west and north. The traditional language area continues south over the Merritt plateau and into northern Washington State in the United States. Most fluent speakers are in their 60's or older. The language is most closely related to St'ât'imcets (Lillooet) and Secwepmc (Shuswap), which it borders to the west/ northwest and north, respectively. In the east, the Nte?kepmx territory shares borders with the Okanagan language area, and, to the south, Halkomelem. Table 2.1  The Salish language family (adapted from Kroeber 1999: 4)  A. Bella Coola  D. Tsamosan (Olympic) 1. Inland  B. Central (Coast) Salish Comox Pentlatch  Upper Chehalis Cowlitz 2. Maritime  Sechelt  Quinault  Squamish  Lower Chehalis  Halkomelem Nooksack Straits  E. Interior Salish 1. Northern  Northern Straits  Lillooet (St'ât'imcets)  Clallam  Me?kepmxcin (Thompson)  Twana Lushootseed  Shuswap 2. Southern Columbian  C. Tillamook  Okanagan Kalispel Coeur d'Alene  The present study is based on a corpus of conversational recordings collected during fieldwork with two female speakers of the iqamcm, or Lytton, dialect of NteVkepmxcin. Both are bilingual, also being fluent speakers of English, and are in their late 60's.  2.1  Phonemic inventory The phonemic inventory of Nte?kepmxcin, based on the Spuzzum dialect, is taken  from Thompson and Thompson (1992, 1996). The dialect of the speakers in this study seems to be very similar in terms of phonemes, with the notable exception being the replacement of (YI by Ij] (orthographic 'y'). N'te?kepmx Izl is considerably more lateral than English Izl, and Thompson and Thompson classify it as a resonant (1992: 8, for more description); the pharyngeal fricative series  V V ^ / also seems to pattern with résonants rather than  fricatives in terms of being glottalized (Thompson and Thompson 1992). Table 2.2  Phonemic inventory (adapted from Thompson and Thompson 1992) alveo-  CONSONANTS  Stops Ejectives  labial  alveolar  P  t  kk'^  qq*  p  t'  k k"^  q q*  s [f]  x x*  xx^  y Ijl  w  S"?^  y  w  T'l^  Lateral Eject. Nasal Glottalized  m  n  m  n ç |ts|  Ejective  uvular pharyngeal  c [t/l  s Isj  Lateral  t (w)  Lateral  z 1  (w)  Glott. Lateral  z f front  VOWELS  high  i  mid  e  low  central  back  I  u 9 a  glottal ?  c [ts']  Fricatives  Glottalized  velar  t.  Affricates  Approximant  palatal  a  o  2.2  Word order and other syntactic background In this section, I review the basic word order o f Thompson Salish (VSO). I discuss  predicate-argument flexibility, the determiner system, and the structure and morphology of relative clauses. I also introduce two structures which have been identified by Kroeber (1997) as focus constructions, a hypothesis that will be empirically tested in the following chapter. 2.2.1 Basic word order: VSO NieVkepmxcin is a strongly predicate-initial language. Usually, this initial predicate is a verb like kdntes 'help' in (1), or a light verb (auxiliary) like the 'progressive' ?ex in (2). '° The basic word order, at least in the Lytton dialect that is the subject of the present study, is verb-subject-object (Davis 2005 on V S O order in Lower St'at'imcets, a neighbouring Salish language). Examples in this section are wide focus (CP focus) utterances, where all information is new. These would be suitable responses to a wh-question like "What happened?" or "What's going on?" This provides us with a default word order, and default intonation pattern, with which to compare other types of focus. Focus is indicated with a subscript FOC; in the examples below, the whole clause comprises the focus. Nuclear stress is indicated by underlining the word bearing the nuclear pitch accent. The pitch tracings and waveforms are from Praat (Boersma and Weenink 2007). At the moment, I present only informal observations on the location of nuclear stress; after introducing the primary Nte?kepmxcin data in this and the next chapter, I provide detailed phonetic arguments for the rightmost nuclear stress position in chapter 4. (1)  V  [2"" pos. clitici  S skixze?-kt  O  [k9n-t-0-és  xe?  e  e  smd?-kt]Foc  help-TRANS-3o-3TS  DEM  DET mother-lPL.POSS D E T b r o t h e r - l F L . P S  "[Our mother helped our brotherJFoc" (*"Our brother helped our mother.")  See the List of Symbols for keys to the orthography, and to the abbreviations in the gloss, and Thompson and Thompson (1992, 1996) for further detail. For reasons of space and clarity, I often do not provide full morphological breakdowns for nouns, adjectives, adverbs, and so on.  300  N  I  B ix  150-  2.91247  Time (s) Figure 2.1 (2)  Pitch tracing and waveform: "[Our mother helped our brotherIpp^."  Aux [2"" pos. clitic]  V  S  [?éx  xe?  cax-t-0-és  f  PROG  DEM  clean-TR-3o-3TS  DET ISG.PS-husband DET  "IMy husband was cleaning up the snow]Foc-"  n-sxaywi  O e  SWU2Ç^-]FOC  snow  300  N  X  150-  h.  caxtés clean  DET  1  Figure 2.2  n-sxaywi my-husband  e swùx^'t. DET snow  2.72333  Time (s)  Pitch tracing and waveform: "I My husband was cleaning up the snowlpoc'  Second position clitics, including evidentials, clause-typing morphology, and the ubiquitous discourse level deictic xe?in(\) and (2), follow the first prosodie word. The deictic xe? could either be doubling one of the arguments, or referring to a discourse-level situational argument. I'll take the latter view (following Thompson and Thompson 1992: 135, 142); since transitive sentences with two ;ve?deictics are not possible, it suggests that its role is not to refer to subject or object arguments when these arguments are themselves overt. Nuclear, or primary sentential stress, typically appears rightmost (underlined), on the object in basic transitive sentences (1-2) (Chomsky and Halle 1968, Cinque 1993, on English). In intransitive sentences, the subject is generally the rightmost constituent, and so the subject receives nuclear stress (3). At this point, I just make an informal observation about the location of nuclear stress; the claim is investigated in detail, and given phonetic support, in the study presented in chapter 4. (3)  V [kstni-m rodfish-MDL  e DET  "(One man is fishingIp^-."  pé[p]ye?  te  ^UTS^JO^.JFOC  one(DiM|  OBL  man.  Where adjuncts, like the adverbial phrase "tonight," appear rightmost, they receive the nuclear stress. (4)  Aux [x*uy FiJT  [2"" position clitics]  V  Adv  ek^u xe?  téki  tk  sitist]mr.  EViD DEM  rain  OBL.IRL  night  "|I heard it's going to rain tonightlpoc''  Time (s)  Figure 2.3  Pitch tracing and waveform: "[I heard it's going to rain tonight]poc  2.2.2 Different types of predicates: verbs, nouns, adjectives Bare nouns (5) or adjectives (6) can also function as initial predicates. Rightmost subjects again carry the nuclear stress. (5)  N  12"" position clitic]  [sqâqxa  xe?  e  Hérmannlpor-  dog  DEM  DET  Hermann  "[Hermann is a doglpoc''  S  (6)  A  S  [xzuméyxkn  e  Eddie] FOP.  big  DET  Eddie  "[Eddie is heavy  IpQç."  30&  N  X B h.  150  1.41218  Time (s)  Figure 2.4  Pitch tracing and waveform: "(Eddie is heavylpoç."  Nominal predicates can be complex, comprising (for example) a modified noun, like xzum xe? tk spzu? 'big bird' below. I'll assume that the deictic xe?is not syntactically a part of the complex NP; but, as a second position clitic, xe?h positioned prosodically after the first word. (7)  [[Npxzum big  xe?  tk  spzu?]  e  sqsca ]FOC  DEM  OBL.IRL  bird  DET  chickenhawk  "(The chickenhawk is a big birdlpoc''  2.2.3 Demonstratives and headless relative clauses as arguments Overt subjects or objects can be argument DPs, as seen so far. Demonstratives can apparently also function as arguments," as can headless relative clauses. 1 will assume that headless relative clauses are headed by a null NP. (8)  [nqfxcetnJFnr xé?e. key  DEM  "That's a [keylpoc" (9)  ye-niin-0-0-e  xe?  [DP e  good-REL-3o-lsG.TS DEM f  n-snukV? ]].  DET  ISGPOSs-friend  0 [cp s-cuvi^=étx'^-s  DET NP  NOM-build=house.3sg.poss  "I liked the house that my friend built." (more literally: "I like the (one that) my friend house-built.") 2.2.4 The topic projection at the left edge So far, all of the examples presented have been predicate-initial. Contrastive topics constitute the only instance where a DP precedes the predicate (Gardiner 1998 on neighbouring Shuswap Salish). In these cases, however, the contrastive topic is set off in its own intonational phrase, typically ending in a high boundary tone and followed by a pause. In (10), the speaker is describing a game of "hide-and-go-seek." The contrastive topic, e pepye? te sqaczeîiyxs  'one of their fathers,' precedes the matrix verb punms 'find,' but  constitutes its own intonational phrase (i-phrase). Its last prosodie word sqacz.e?iyxs carries a high boundary tone to mark continuation, and is followed by a pause of approximately 600 milliseconds and an intake of breath. The second intonational phrase resets declination, starting at a high pitch, and generally declining throughout.  " It is probable i\\a\xe?e is still a situational deictic in (8), and that the clausal subject is just null (see Thompson and Thompson 1992: 135, 142, for discussion).  (10)  [e  pépye?  te  sqacze?-iyxs]i -phrase»  DET  one|DiM|  OBL  father-3PL.POSS,  [pûn-in-0-0-s  ek'^u xe? e  siséye  find-REL-TR-30-3TS EVID D E M DET tWO|AUG]  te OBL  smuimumleç]i.phraseWOman|AUG||DIMl  "One of the fathers, he found two little girls."  Time (s)  Figure 2.5  Pitch tracing and spectrogram for example (10)  Thus, even in the case of contrastive topics, the following generalization holds: the intonational phrase bearing the matrix predicate is always predicate-initial. The undominated constraint in (11) describes this state of affairs (see Krifka 1998 on the constraint V E R B RIGHT  (11)  in German): PREDICATE-LEFT:  Align the matrix predicate with the left edge of an intonational phrase. (11) is meant as a decription of the facts presented thus far, and not as an Optimality Theoretic constraint to derive syntactic ordering. I will assume that the syntactic derivation provides verb-initial structures, but employ the terminology P R E D I C A T E - L E F T as a convenient description of these facts.  It should be pointed out that the "fronting" of DPs as well as various adverbial phrases has previously been described as "unmarked fronting" (Kroeber 1999: 391; see also Thompson and Thompson 1992: 159-161); "unmarked" here means that it is morphologically unmarked, inducing no special morphology on the verb. However, I observe that such fronting is always prosodically marked, by setting the fronted material in a separate i-phrase marked with boundary tones. 2.2.5 Determiners In this section, I discuss four important morphemes in the determiner system of Nie?kepmxcin. The following discussion is based on Thompson and Thompson (1992), Kroeber (1997), and Jimmie (2002, 2003), as well as my own observations. The distribution of determiners will be used to argue against a focus movement account for NieVkepmxcin. I will gloss the first two determiners, (h)e and f(e)&s DET. (h)e introduces DPs that are present and visible. Often this determiner reduces to (a] or zero. Its counterpart is remote i(e), which introduces DPs that are removed in space or time (either not visible or referring to the past or future). Both of these determiners introduce subject and object in transitive clauses (12ab), as well as subjects of intransitive clauses (12cd). (12)  a. k"'éw-0-0-es float-TRANS-3o-lSG.TS  xe?  e  John  e  syép  DEM  DET  John  DET  tree  "John let the tree float down the river." (DPs present) b.  sik-t-0-és  xe?3  hit-TRANS-3o-3TS D E M  t  n-sinci?  DET  ISG.POSS-younger.brother  i  n-snùfc"'e?  DET  ISG.POSS-friend  "My younger brother hit my friend on purpose." (DPs absent) c. qft-t wake-iM  ?éyl  e  sk*ùk*mi?t.  now  DET  child  "The child finally woke up." (DP present) d.  cuw^éix"^  we?e  "t  ^u?sqâyx"'  build=house  DEM  DET  man  "The man builds houses." (DP absent)  Irrealis k marks complements that are unrealized, as in contexts of negation or hypothetical situations. I gloss this morpheme as IRL. (13)  6, xzum  nke  k  e?-citx^  oh, big  EVID  IRL  2sG.POSS-house  "Oh, I guess your house is big." Finally, oblique t(e) serves a variety of functions typical of a general preposition like English o/(Kroeber 1997 glosses this as a preposition). I gloss it as OBL. The oblique marker introduces patients in morphologically intransitive (14) and ditransitive clauses (15), as well as indefinite 3rd person agents in 'passive' constructions (16). t{e} also serves to introduce instruments (17). (14)  kéx-m  xe?  t  dry-MDL  D E M DET  n-skixze?  te  epls  IsG.POSS-mother  O B L apple  " M y mother dried some apples." (15)  x"'uy  xe?  )tqu?-xi-t-sm-s  f  n-skixze?  FUT  DEM  sew-APPL-TRANS-lSG.O-3TS  DET  ISG.POSS-mother  te OBL  n^tpice? ISG.POSS.shirt  " M y m o t h e r w i l l be s e w i n g me m y shirt."  (16)  qây-9-t-0-m  xe?  te  snùk^e?-s  shot-DRV-TRANS-30-IDF.TS  DEM  OBL  friend-3sG.P0SS  " H e got shot b y his f r i e n d . "  (17)  c9q-t-0-éne  xe?  e  n-k^9n=us-tn  hit-TRANS-3o-lSG.TS  DEM  DET  LOC-look.at=opening-iNSTR  te  ?e?=use?  OBL  RED.egg=berry  "I hit the w i n d o w w i t h an e g g . "  In the above instances, we can think of te as a combination of the oblique t- plus the determiner e (Kroeber 1997). That oblique t- can combine with the other determiners is most 48  evident with the irrealis determiner A: (18). When the oblique marker combines with the remote determiner f, as in the ditransitive in (19), the result for my consultants is typically loss of the t- portion through assimilation (Kroeber 1997 reports the same tendency). (18)  x^uy  xe?  n-t-sém-s  t  FUT  DEM  give-TRANS-lSG.0-3.TS  DET  tk  katni-in-tn  OBL.IRL  rod.fish-MDL-INSTR  n-smci? ISG.POSS-younger.brother  "My younger brother is gonna' give me a fishing rod." (19)  ?éx  xe?  cù-l-x-a-O-ne  ^  n-snuk^e?  PROG  DEM  sh0W-APPL-DRV-30-lSG.TS  DET  ISG.POSS-friend  0  i  OBL DET  n-citx* ISG.POSS-house  "I'm showing my friend my house." The combination of oblique t~ and irrealis k is particularly common in complex nominals. Thompson and Thompson (1992: 153) call this tk a single 'descriptive' marker (it is frequently used in descriptive complex predicates like (20)), but I gloss it as a combination of the two primitive morphemes OBL and IRL. It should be noted, however, that tk does not indicate that the noun dog stands in an oblique relationship with its modifier big (as Thompson and Thompson's use of "descriptive" suggests). Rather, I assume that the noun is the head of these complex structures, and not the initial adjective (following Davis, Lai and Matthewson 1997, on complex nominals in St'ât'imcets Salish). (20)  [NP xzum big  xe?  tk  DEM  OBL.IRL  [N sqâqxa]] dog  e  Bérnie  DET  Bemie  "Bernie is a big dog." In addition, t(e), as noted by Kroeber (1997), also introduces all relative clauses.'^ Kroeber (1997:385) distinguishes this use of t(e) from the oblique marker, identifying it as 'attributive', though he speculates that it may be the same morpheme as oblique t(e). I simply Exceptions include headless relative clauses, which are introduced by the matrix determiner k, (h)e or i, and locative relative clauses, which are introduced by the relevant preposition (see Kroeber 1997; (22b), (23)).  gloss all instances of t(e) as OBL. Relative clauses are formed by fronting the determiner from the clause-internal argument corresponding to the head of the relative clause (for detailed argumentation, see Kroeber 1997, Koch 2(X)4, 2006; Davis 2004 on St'at'imcets). This is achieved by fronting of a DP with null N P complement to the specifier of CP. The determiner combines with the oblique marker that introduces the relative clause. This is readily apparent where the determiner involved is present (h)e, since oblique t and present e combine to give te. (21)  né-x-t-sm-e  he  se?Iis t-ea  cu-t-éx^  U  give-APPL-TR-lSG.o-iMP  DET  knife  fix-TR-30-2SG.TS  td  OBL-DET  "Give me the knife that you fixed." However, when the remote determiner f i s involved, oblique t is lost (22a), presumably due to similarities in pronunciation of these two segments, as previously mentioned. For my consultants this is almost always the case, and Kroeber (1997) reports this as a strong tendency for his consultants also. In locative relatives, the preposition from inside the relative clause fronts to introduce the relative clause, and not the oblique marker (22b). (22)  a.  q"'in-t-0-éne  xe?  talk.tO-TRANS-30-lSG.TS D E M  i  sfe^ùk^mi?t  DET  child  0A-QX  wik-t-0-ne  U  OBL-DET-PROG  see-TRANS-30-lSG.TS  tj  "I talked to the c h i l d that I s a w . "  b.  ?éx  kn  x^f?-m  PROG  ISG  look.for-MDL  te  npùytn  n-Ci  x^ûy  wn  V^oyt  ti  OBL  bed  in-DET,  FUT  ISG.CONJ  sleep  tj  "I'm looking for a bed where I'm gonna' sleep." In headless relative clauses, only the matrix determiner surfaces; the oblique marker and relative-clause internal determiner are lost. Presumably there is an independent restriction against a double determiner structure without an intervening nominal. In the example below, 'dog' is an N P predicate, and takes a headless relative clause as subject. The structure of these is important because given material surfaces in clausal DP arguments.  (23)  iNpsqâqxa]  [DP  dog  e  0  DET  NP  [cp (*t-ed)  pûn-m-0-ne  (OBL-DET)  tj ]  find-REL-30-1 SG.TS  "I found a dog." (more literally: 'The (thing that) I found was a dog.") I will assume that this determiner movement'^ creates an operator-variable configuration that allows the relative clause to be interpreted as a type <e,t> predicate (lambda abstraction, or predicate abstraction), following Heim and Kratzer (1998). That is, the oblique-determiner combination is the same as a relative pronoun like who or what or that in English relative clauses. The following structure illustrates a relative clause in Thompson River Salish; the NP may be overt or, as below, null.  DP D  NP  I  e the  punmne Ubi^sp Ifounddt Figure 2.6 (24)  Relative clause in Nte?kepmxcin: "the (thing that) I found'  Predicate (lambda) abstraction: If a is a branching node whose daughters are a relative pronoun and p, then [[all = kxE D,. [[^]\\  (Heim and Kratzer 1998: 96)  More precisely, a DP with null NP complement; the DP raises to specifier of CP.  The determiners e and f, oblique t{e), and irrealis k can also act as complementizers. They may introduce adjunct clauses and complement clauses (see Thompson and Thompson 1992 and Kroeber 1997 for further examples).'^ (25)  taté?  k  n-s-tÇ^-iyx-nwéin  NEG  COMP  1 SG.POSS-NOM-run-AUT-NCM  "I can't run." (more literally: "It's not the case that I can run.") (26)  ?éx  xe?  PROG  DEM tell-TRANS-lSG.O-3TS DET  te  pilax-t-sm-s  s-x*ùy-s  \ nés  COMP NOM-FUT-3SG.POSS go  n-snuk^e? ISG.POSS-friend  u  t  ^tqamcin  t  John  to  DET  Lytton  DET  John  " M y friend told me that John was going to Lytton." (27)  e  wuxn us,  COMP  snow  x^uy kt  3CONJ FUT  iPL  faxi cold  "If it snows, we're gonna' get cold." The determiner system of Nieîkepmxcin is summarized in the table below. Table 2.3  Determiners in N'te?kepmxcin  he/9l0  specific; present, visible (DET)  tfe)  remote (in space or time) (DET)  t(e)  oblique (OBL)  k  unrealized/irrealis (IRL)  2.2.6 Relative clauses and their morphology (Kroeber 1997) Kroeber (1997) gives a detailed description of the morphology involved in relative clauses in N'te?kepmxcin (see Kroeber 1999 for a comparison across the Salish language family). I summarize the main points here; as we will see, focus structures employ this agreement morphology as well.  Kroeber (1997: 381) notes that fis "quite rare" as a complementizer, an observation with which I concur.  Relative clauses are typically head initial, with the restricting clause following the head (though the other order is sometimes also attested). The head may also be omitted entirely. When subjects of intransitive clauses are restricted, the verb in the restricting clause bears no special morphology (28). The same is true for transitive clauses with relativized objects (29). The indicative morphology of matrix clauses is used in these cases. (28)  pùn-m-0-ne  n-ci?  [DP 0 ^u?sqâyx"'[cpt-Cs  find-REL-3o-l SG.TS  at-there  DET  man  OBL-DET  "I found lopa man [cpfishing there]]."  (29)  [DP  k^is-0-0-ne drop-TR-3o-lTS  ^  patak  DET  potato  ?x  ^atm'-m ts||. t^  PROG fish-MDL  (Kroeber 1997: 390)  [CP 0  %  O B L DET  ?ûpi-n-0-x*  ta]]  eat-TR-3o-2TS tj  "I dropped Ippthe potato [cpthat you ate]]." When the head of the relative clause corresponds to the transitive subject of the restricting clause, the subordinated verb is marked with special subject extraction morphology -emus, but only if the object of the verb is also in the 3"^ person (Kroeber 1997: 391). Use of-emus distinguishes relativization of transitive subjects from transitive direct objects. (30)  swét who  [DP k  [cp ?ùpi-t-emus  IRL  eat-TR-SUBJ.EXTR  <!  he?ûse?]]  DET  egg  "Who ate the egg?" Nonlocative obliques that are relativized induce nominalization morphology on the verb in the relative clause. These obliques include instruments, patients of morphologically intransitive verbs (like fa?xans 'eat' in (31)), and patients of ditransitive verbs. (31)  ye-mine  xe?  g00d-REL.30.1 SG.TS  DEM  [^pC DET  "I liked (DPthe fish (cpthat I ate]|."  sqyéytn [cpt-tj  n-s-ia?xâns  fish  ISG.POSS-NOM-eat tj  OBL-DET  ti ]]  Finally, relativization of locatives induces conjunctive'^ morphology on the predicates of the relative clause. In addition, the preposition is fronted to introduce the relative clause, along with the determiner corresponding to the NP that has been relativized. (32)  (w)?éxkn  x^i?-m  PROG  look.for-MDL O B L  ISG  t  wn  î^oyt t, IJ  ISG.CONJ  sleep  [ope  npùytn  DET bed  [cpn-ei in-DET  x™ûy FUT  t,  "I'm looking for a bed where I'm gonna' sleep." The generalizations are summarized in the table below. We will see the same sort of morphology surfacing in focus structures, as discussed in detail by Kroeber (1997). While the focus is associated with the predicate, given material surfaces in headless relative clause arguments. Table 2.4  Relative clauses and their morphology (Kroeber 1997)  G R A M M A T I C A L RELATIONSHIP O F  G R A M M A T I C A L AGREEMFJ^JT O F  RESTRICTED N P TO R E L A T I V E C L A U S E VERB  R E L A T I V E C L A U S E VERB  Subject of intransitive verb  indicative  Object of transitive verb  indicative  Subject of transitive verb  -emus suffix on verb  (when object is 3"^ person) Nonlocative oblique  nominalized  Locative  subjunctive (conjunctive)  2.2.7 Two focus structures There are two common narrow focus structures in Thompson Salish (Kroeber 1997), and indeed across the Salish language family (Kroeber 1999 for comparison). In predicate constructions, a bare noun or adjective acts both as the matrix predicate and as the focus.'* The focus is usually followed by a clausal DP argument which contains the backgrounded, or  '^ Subjunctive morphology is glossed 'conjunctive' in the Interior Salish tradition to avoid confusion with 'subject' in the glosses. '* Kroeber (1997) has thus called these "bare clefts." However, I stick with the less misleading term "predicate constructions," following Davis et al. (2004).  given, portion of the utterance. Like a headless relative clause, the cleft clause is typically introduced by a determiner and carries the same subordinating morphology seen in the previous section. I'll refer to these clausal argument DPs as "residue clauses," as a convenient cover term for the clausal arguments found in both types of focus structures. Residue clauses serve as syntactic subjects; thus, clefts are truly bielausal, with the focus base generated and not moved from within the cleft clause (see Kroeber 1997, 1999, Koch 2007b, for detailed argumentation; also Davis et al. 2004, on St'ât'imcets Salish; ). The syntactic structure of clefts will be taken up in more detail in chapter 7. (33)  A:  Sté?  x^ùy k  s-^a?xâns-9p  tk  s7âp.  what  FUT  NOM-eat-2PL.POSS  OBL.IRL  evening  IRL  "What are you people going to eat this evening?" B:  |pins]Foc  nee?  x^uy  e  n-s-ia?xâns.  bean  ISG.EMPH  FUT  DET  Iso.POSS-NOM-eat.  "I'm gonna' eat (beansJFOC-" (more literally "The (thing that) I'm gonna' eat is Ibeansjpoc") In example (33), the bare noun pins is both the predicate and the object focus; the subordinated verb ia?xans 'eat' is introduced by a determiner e and prefixed with nominalizing morphology n-s-}'' Nuclear stress remains rightmost (this informal observation will be tested phonetically in chapter 4). The second type of focus structure is what Kroeber calls an introduced cleft. Introduced clefts consist of a cleft predicate ce or ?e which "introduces" the focused D P (the head of the cleft). DPs such as proper names are typical arguments and cannot be predicates in Me?kepmxcin (though they can in the neighbouring Salish languages St'ât'imcets and Secwepmctsin, when not introduced by a determiner - Davis p.c). Therefore they require a cleft predicate at the left edge of the clause when focused; recall that Nfe?kepmxcin is a " The position of the future marker x'^uy is also somewhat anomolous in (33), coming before the determiner that introduces the clause whose verb x'^iiy modifies; but Kroeber (1997, 1999:390) has noted that residue clauses with future markers are sometimes not introduced by a determiner at all, or sometimes only erratically, with the consultants he has worked with. I concur with this finding, adding that sometimes consultants will have the future marker preceding the determiner, as in example (33). Similarly, my consultants sometimes omit the determiner introducing residue clauses beginning with another auxiliary, 'progressive' (w)?ex{tg. 35A).  predicate-initial language, so DPs may not be clause-initial.'* In (34), the DP e Monique is the object focus, and follows the cleft predicate ce and the second position clitic xe?. In the residue clause, the given verb wiktne 'I saw' is preceded by the determiner e. (34)  A:  swét  xe?  k  wîk-t-0-x^.  who  DEM  IRL  see-TR-3o-2SG.TS  "Who did you see?" B:  cé  xe?  CLEFT D E M  [e  Mom'quejfoc  DET Monique  e  wik-t-0-ne.  DET  see-TR-3o-lSG.TS  "I saw IMoniqueJpoc." (literally "It was (Moniquelpoc that I saw.") Again, there is a divorce of the primary stress (rightmost) from the focus (the leftmost lexical element). This divergence of the focus from the nuclear stress position is unexpected under common accounts of focus marking. Like in English, the clausal argument in focus structures may be omitted entirely ( D E L E T E - G I V E N ) . However, we shall see that it is not deaccented (DESTRESS-GIVEN). Thus, we cannot see deletion as the most extreme form of deaccenting (eg. Krifka 2007). In (35), has been working at the café is given in the question, and is not produced in a residue clause in B's answer. (35)  Residue clauses may be omitted in focus structures A : Swét who  met  0  ?ex  cw-um  ne: ... na-,  n"ia?xanséyq'^.  CNSQ  IRL  PROG  work-MDL  in.DET... i n . D E T - , café  "Who's been working at the café?" B: cé CLEFT  ek*u  e  Sally.  EVID  DET  Sally  "Sally." [has been working at the cafe] It should be noted, as Kroeber remarks, that there is nothing unexpected about these focus structures per se, especially nominal predicate constructions. We have already seen that  Except, of course, as contrastive topics, as discussed earlier. In these cases, however, they are set into their own intonational phrase, and the matrix verb is still initial in its intonational phrase.  bare nouns can act as predicates (section 2.1.2), and that headless relative clauses can act as arguments (section 2.1.3), independent of one another: It should also not be understood to imply that [focus structures] as defined constitute a distinct construction type in Salish languages. Headless relative clauses are solidly attested in Salish languages in ordinary DP positions other than subject (hence, outside of cleft constructions), and ... Salish languages readily allow nominal expressions to act as predicates even when the subject of the sentence is something not obviously clausal; that [focus structures] like the [onesI cited above should exist is simply an automatic consequence of these facts of Salish constituent structure, requiring no special stipulations. (Kroeber 1999:261-262) The examples below illustrate these facts. In (36), the bare noun sqâqxa 'dog' acts as the predicate, taking a D P subject e Hermann. In (37), the headless relative clause e punmne 'what I found' acts as an object argument for the transitive matrix verb 'eat.' Example (38) shows a nominal predicate construction, which simply combines the bare noun predicate of (36) with the headless relative clause argument of (37). Finally, (39) shows a cleft with a DP focus (DPs cannot be predicates, and so must be introduced by the cleft predicate). Since Kroeber notes that nominal predicate constructions (NPCs) and clefts are used as focus constructions, I have marked the focus with a subscript FOC; Kroeber'S "cleft-focus" observation will be tested empirically in chapter 3, and observations about the nuclear stress (underlined) will be tested phonetically in chapter 4. (36)  sqâqxa  xe?  e  dog  D E M DET  Hermann, Hermann  "Hermann is a dog." (37)  ?ûpi-0-0-ne  xe?  e  eat-TR-30-lSG.TS  D E M DET  pùn-m-0-0-ne. find-REL-TR-30-lSG.TS  "I ate what I f o u n d . "  (38)  [sqâqxajpoc  xe?  e  pùn-m-0-0-ne.  dog  DEM  DET  find-REL-TR-30-lSG.TS  "I found a Idoglpoç." (more literally "The (thing that) I found was a [doglpoc-")  (39)  cé  xe?  CLEFT D E M  [e  sqâqxajpoc  e  pùn-m-0-0-ne.  DET  dog  DET  fi n d - R E L - T R - 3 0 - l SG.TS  "I found the [doglpoc." (literally "It was (the dogjpoc that I found.") As Kroeber notes, however, clefts are certainly unique constructions in other respects. Though cleft predicates take two arguments (the focus, and the residue clause), they are the only transitive predicate to lack transitivizing morphology. They also have a strict word order of cleft predicate - focus - residue clause, whereas typical transitive sentences allow some degree of flexibility in post-predicative word order. We can also note that both focus structures abide by the predicate-initial structure of Nte?kepmxcin. In nominal predicate constructions (38), the focus sqaqxa 'dog' is leftmost, and since bare nouns can be predicates, the matrix predicate, again sqaqxa, is also initial. No further morphology is required. In the case of focused DPs (39), there is a conflict for the leftmost position between the focus and the predicate; but DPs cannot be predicates, and so focused DPs must be introduced by a cleft predicate at the left edge of the clause. This sets N'Ie?kepmxcin apart from its neighbours St'at'imcets and Secwepemctsin Salish, where proper names are nominalized and can act as predicates when not preceded by a determiner (Davis, p.c; Thoma 2(X)7 on proper name predicates as covert clefts). Finally, focus structures neatly split the clause into a < B A C K G R O U N D , FOCUS> configuration as in the focus semantics proposed in the Structured Meaning Approach reviewed in section 1.2.2 (von Stechow 1990, Krifka 1992; also Jackendoff's 1972 'presup'). The surface order is of course different in Salish, corresponding to <FOCUS, B A C K G R O U N D > . In the Structured Meaning focus semantics, the background is the function taking the focus as its argument. The ordinary semantics operate just the other way around here: in the actual surface structure of Salish focus constructions, the focus is the predicate which takes the background information as a subject clause. I will assume that residue clauses, like relative clauses elsewhere, are formed by clause-internal movement of the D P which corresponds to the focus (see figure 2.6). This turns the residue clause C P into a predicate of type <e,t>, by lambda abstraction (24) (Heim and Kratzer 1998: 96). In the nominal predicate construction, the focus is the matrix predicate taking the residue clause as its argument. The structure creates a clear division between focus and background.  [sqâqxa]Foc  xe?  e  pûn-m-0-0-ne.  dog  DEM  DET  find-REL-TR-30-1 SG.TS  FOCUS  BACKGROUND  dog  Xx. I found x  "I found a (doglpoc-" (more literally "The (thing that) I found was a [doglpoc-") In clefts, the focused DP is part of the cleft predicate V P . The residue clause containing the background is the clausal subject. Focus and background are linked by the cleft predicate. (41)  9  r  ce  xe?  CLEFT D E M  [e  MoniqueJFoc  e  wik-t-0-ne.  DET M o n i q u e  DET  s e e - T R - 3 0 - l SG.TS  FOCUS  BACKGROUND  Monique  Xx. I saw X  "I saw IMoniquelpoç." (literally "It was [Moniquelpoc that I saw.")  2.3  Summary This chapter has provided background information about Nie?kepmxcin. I showed  that Thompson Salish is a stress language with verb-initial word order. Bare nouns and adjectives can also act as predicates, and predicates are always initial in their intonational phrase (PREDICATE-LEFT). Cleft structures and nominal predicate constructions are common, and employ the same verbal morphology found in relative clauses. In the next chapter, I test Kroeber's observation about clefts and nominal predicate constructions that I introduced in section 2.2.7: clefts and NPCs are used to mark the focus of the sentence ("cleft-focus"). We will see that this observation is difficult to reconcile with another common account of the marking of focus, the "stress-focus" generalization. This is because nuclear stress is rightmost (an observation which will be tested phonetically in chapter 4), but clefts and NPCs have the effect of restructuring the clause such that the focus is at the left edge.  Chapter III: Focus Structure In chapter 2,1 observed that nuclear stress is rightmost in Nte?kepmxcin. In contrast to the rightmost location of sentential accent, however, is the observation that narrow focus is marked by restructuring the focus toward the left edge of the clause (Kroeber 1997, 1999), by employing clefts or nominal predicate constructions (NPCs). This divergence of the focus from the nuclear stress position is unexpected under many of the common accounts of focus marking in stress languages. Moreover, these focus structures are very generally and commonly employed across the Salish language family, including for wh-questions (see Kroeber 1999 for a thorough treatment; Kuipers 1%7, Kuipers 1974, Gerdts 1988, P. Davis and Saunders 1997, Davis, Gardiner and Matthewson 1993, Galloway 1993, Gardiner 1993, van Eijk 1997, Suttles 2004, Davis 2008). It is likely that examination of nuclear stress in other languages in the family would find a similar divergence of focus and stress (though Benner 2006 suggests leftmost nuclear stress in Senchothen). The purpose of this chapter is thus to test Kroeber's "cleft-focus" observation through a corpus study of focus types and their surface syntactic form. How are various focus types syntactically realized: in the default verb-initial order, or as clefts or NPCs? I begin by examining the default case (wide CP focus) where all information is new, as well as wide V P focus. Then I compare the surface structure of various types of narrow focus: narrow verb focus, subject focus, object focus, and number quantifier focus. The results support Kroeber's generalization: while CP focus, V P focus and narrow verb focus are realized in default V S O order, narrow focus on subject, object, or number quantifier is indeed predominantly marked through the use of left edge clefts or NPCs. The generalization for all types of focus is that the left edge of the clause is important for focus marking, and for focus projection, whereas the rightmost nuclear stress position does not play a role.  3.1  "Stress-focus" theories and predictions for Thompson River Salish Current theories on the marking of focus in stress languages generally have in  common the correspondence between stress and focus. I examined these claims in section 1.3.3. The main proposals are repeated here: (1)  Proposals on the marking of focus a. Basic Focus Rule: A n accented word is F(ocus)-marked. (Selkirk 1995: 555) b. Stress-Focus Correspondence Principle: The focus of a clause is a(ny) constituent containing the main stress of the intonational phrase, as determined by the stress rule. (Reinhart 1995: 62)  c. Focus: A Focus-marked phrase contains an accent. (Schwarzschild 1999: 173) d. F O C U S - P R O M I N E N C E : FOCUS needs to be maximally prominent. A prosodie category C that contains a focused constituent is the head of the smallest prosodie unit containing C. (Truckenbrodt 1995, Buring 2003) e. S T R E S S - F O C U S : a focused phrase has the highest prosodie prominence in its focus domain. (Féry and Samek-Lodovici 2006: 135-6) The examination of English, Portuguese, Hungarian and German focus marking revealed three strategies to ensure that the location of focus and nuclear stress coincide: (2)  "Stress-focus" strategies (i)  move the nuclear stress to the focus (English)  (ii)  move the focus to the nuclear stress position (Hungarian, Romance)  (iii)  scramble non-foci away from the nuclear stress site (German)  In chapter 2, we saw that Nte?kepmxein has a default verb-initial order with rightmost nuclear stress. The claim about the location of nuclear stress will be tested empirically in chapter 4. In the meantime, we can predict three possible ways that narrow focus on object, verb or subject will be marked in NteVkepmxcin, assuming that the "stressfocus" generalization holds. In an English-type system, we don't expect a change in word order, just a change in the location of the primary stress.'^ If Nte?kepmxcin is a Romance-type system, we expect focused constituents to move to the right edge of the clause, where the nuclear stress position is. And in a German-type system, we expect non-focal constituents to scramble away from the right edge so that the focus can surface there. These predictions are summarized in table 3.1. For narrow object focus, we never expect changes in word order. Since objects are in the nuclear stress site in default V S O word order, they are already in the ideal "stress-focus" position. Of course, English focus structures like clefts or pseudo-clefts do involve a different word order; but the nuclear stress will still shift to the focused constituent. In addition, use of such structures is motivated, for example, by the semantics they provide (Perçus 1997, Hedberg 2000 on English clefts); we shall see in chapter 7 that Salish focus structures lack this semantic motivation. For the purposes of table 3.1,1 am just concerned with English in situ focus.  Table 3.1  Stress-focus systems and predictions for NteVkepmxcin Predictions for Nte?kepmxcin  System type  English  B A S I C WORD  OBJECT  SUBJECT  ORDER  FOCUS  FOCUS  vsO  vsO  vSo  Vso  vsO  vsO  v[tJoS,  [tvlsoV, \ ^  vsO  vsO  vo,S[t,]  V E R B FOCUS  (in situ) Romance (movement) German (scrambling) note: underlined C A P I T A L S show the location of focus and nuclear stress It should be pointed out that the predicted surface orders of the Romance and the German systems are the same, so we would have to develop some language-internal diagnostics to decide between the two hypotheses (Pulleyblank, p.c). In general, it is not clear whether Romance movement and German scrambling are syntactically- or phonologically-driven movement systems (for varying views, see Krifka 1998, Costa 1998, Zubizaretta 1998, Szendroi 2003, Davis 2007, etc.), but the issue is not critical for the present purposes.  3.2  Focus projection and predictions for wide focus in Thompson Salish Chapter 2 showed that the default word order is verb-initial, and that this order is used  for clause-wide focus (CP focus). For wide focus on the VP, we also want to know what sort of focus marking is to be expected. In chapter 1 (section 1.3.4), I reviewed the concept of focus projection. Selkirk (1995) for example noted the phenomenon of focus projection in English, where a single rightmost nuclear stress can indicate focus on a variety of ever larger constituents, each time "projecting" upwards through the syntax. Wide V P focus or wide CP focus thus employs the default prosodie marking, with nuclear stress on a rightmost object, for example. The example below is repeated from section 1.3.4:  (3)  a. What happened? b. What did Frank do?  [sentence-wide CP focus question] [wide V P focus question]  [narrow object focus question] c. What did Frank squash? d. [Frank [[squashedf [a peachfj IJfIf, FOC- [optional focus projection]  Thus, the same syntactic form that marks narrow object focus should also project to mark both types of wide focus. Because the nuclear stress falls on objects, another way of stating this observation is that focus projects from the nuclear pitch accent. (4)  Focus projection and predictions for Nte?kepmxcin a. Wide V P and CP focus employs the default word order (VSO), and b. Wide V P and CP focus employs the same word order as narrow object focus.  3.3  A corpus study: Focus type and syntactic realization In this section, I present results from a corpus study of collected conversational  recordings in Thompson Salish. Focus types were identified and coded for syntactic form: either default verb-initial order, or non verb-initial order (clefts or nominal predicate constructions). For each focus type, I present several typical examples, as well as examples that deviate from the norm. For an introduction to clefts and nominal predicate constructions, see section 2.2.7. 3.3.1 Subjects The language data was collected from two female speakers of Nte?kepmxcin in their late 60's (FE and PM). Both are speakers of the Lytton dialect, and fluently bilingual in English. 3.3.2 Method Different instances of focus were identified from a corpus of conversational recordings. Recordings were made at the residence of either the language consultants or of the researcher, using a Marantz P M D 670,671 or 660 digital audio recorder. Each consultant was recorded on a separate channel using a Countrymax Isomax E M W Lavalier lapel microphone. The microphone was attached onto the exterior of the consultants' clothing, approximately at the sternum.  Conversational recordings fell into three general categories (Caldecott and Koch 2007, Koch and Caldecott 2007). First were spontaneous conversations. These included conversations initiated by the consultants, as well as conversations in which one consultant was asking prepared questions of the second consultant, who was free to answer spontaneously, as she liked. Questions were either general questions about the consultant's day to day affairs, or were questions about a media display (photographs, drawings, or computer animations) which the consultant was looking at. Scripted conversations were ones where each consultant had a prepared part of a conversation. Consultants then engaged in role-playing to hold the conversation. Scripts were prepared in English, so the task involved translation as well as engaging in mock dialogue. This format allowed for more targeted data gathering than spontaneous conversation. Finally, single sentence elicitations were used to ask consultants how to say a particular sentence (or small set of sentences), given a particular situation (Matthewson 2004). This technique allowed for very precise targeting of particular linguistic data. Gussenhoven distinguishes contrastive, or "corrective" focus, from new information, or "presentational," focus (2004: 86 for discussion). While some researchers (i.e. É. Kiss 1998, Féry & Samek-Lodovici 2006, Selkirk 2007) treat new and contrastive focus as two different primitives, others do not formally distinguish between the production and interpretation of these two apparent types of focus (i.e. Selkirk 1995, Rooth 1992). Krifka (2007) also notes that both types of focus are what he calls expression focus. That is, neither changes the truth conditions of the utterance (see section 1.3.2 for discussion). It is also plausible that the apparent different prosodie properties of contrastive focus may in fact be related to higher emotional arousal (eg. Banziger and Scherer 2005) rather than a distinction in focus type. For the purposes of this study, I also did not differentiate between contrastive focus and presentational focus, as there were no obvious differences in their syntactic realization as verb-initial or non verb-initial. Where a given utterance contained both a contrastive focus and a new information focus, I counted it according to the type of the contrastive focus for the purposes of this study, since contrastive focus in English is marked preferentially over new information focus (Féry and Samek-Lodovici 2006). The generalizations made in this chapter seem to hold of both types of focus; I'll return to this issue in section 3.5, since other research in both N+e?kepmxcin (Koch 2007c) and St'ât'imcets (Thoma 2007) suggests that clefts have a contrastive semantics not found in nominal predicate constructions. Also, I'll show that constrastive contexts are often marked by the introduction of second position focus particles. The presence of focus particles (i.e. even or only in English) does introduce  different trutli conditions, what Krifka calls denotation focus. Thus, the line between expression focus and denotation focus is often blurred in contrastive contexts. Following Selkirk (1995), I assume that focus is a syntactic category which can be identified by a series of f(ocus)-marks in the syntactic derivation. In view of recent work which aims to eliminate intermediate f-marks (Schwarzschild 1999, Buring 2003, 2006, Féry and Samek-Lodovici 2006) I indicate only the focus constituent with subscript FOC and square brackets [...] in the target utterance. I do not mark intermediate f-marks, to avoid excessive subscripts, though nothing hinges on this theoretically. Nuclear stress is indicated by underlining the word containing the nuclear stress. Word-level stress is indicated by an acute accent (')• In the examples in this chapter, I often provide discourse preceding the target utterance (where space allows), since examples are taken from larger portions of dialogue. In some cases I provide a preceding wh-question only in English italics to reflect the previous discourse (where the wh-question was not overt, where hesitations etc. make the preceding relevant discourse excessively lengthy, or where the question was asked by a non-native speaker of the language - the elicitor, or a computer animation). Focus was identified in one of several ways. First, I adopted the common diagnostic that a new information focus is the answer to the wh-word in a question, like (5B) and (6B) (eg. Jackendoff 1972, Selkirk 1995, Buring 2003 etc.). The wh-question need not be overt, since new information can be introduced in a series of declaratives, as in (6C) (van Kuppevelt 1994, Buring 2003b, Krifka 2007). (5)  A : ?e and  swét  k  taVxans  tk  seplfl  who  IRL  eat  OBL.IRL bread  t  snwénwen.  DET morning  "Who ate some bread this morning?" B: ce CLEFT  ek^u  [e  Patricia]Foc  k  +a?xâns  EVID  DET  Patricia  IRL  eat  tk  seplil  +  snwénwen.  OBL.IRL  bread  DET  morning  "I Patricia J (6)  ate some bread this morning."  A : Where did she go? B: [x^es-x^esit AUG-walk  we  syapyép  te  tam-tmix^Jpoc-  to.DET  tree[AUG]  OBL  AUG-land  "She went |walking in the forestlpoc-"  C: fPûn-m-0-s find-REL-3o-3TS  e  qémes]Foc-  [Pùn-m-0-s  e  qémes]Foc-  DET  mushroom.  find-REL-3o-3TS  DET mushroom.  "She Ifound some mushroomsJpoc- She [found some mushrooms Ipoc-" note: C answers a covert VP-focus wh-question What did she do? Secondly, a contrastive focus sequence occurs, for example, when the speaker contradicts or updates a portion of the information in a previous utterance (7), possibly by responding to a yes/no question (8). (7)  A : Qe?nim-0-0-ne hear-TR-3o-l SG.TS  xe?  k  s-wik-t-0-îyxs  DEM  COMP  NOM-see-TR-3o-3PL.TS  ?et  tu?  xe?  e  spé?ec  ACCM  PERS  DEM  DET bear  u cf?  ?ef.  to there  ACCM  "I heard they saw a bear too." B: He?ây.  [SésyejFoc  xe?  tk  spé?ec  Yes.  Two[DiM]  DEM  OBL.IRL  bear  9  ?es-wik-t-0-iyxs.  DET  STAT-see-TR-30-3PL.TS  "Yes. They saw [twolpoc bears." (8)  A : w?éx  n  met  tu?  xe?9  M^?ex-s-t-0-és  PROG  Q  CNSQ  PERS  DEM  PR0G-CAUS-TR-30-3TS  k  mus  tk  cikn-s.  IRL  four  OBL.IRL  chicken-3sG.P0SS  "Does he still have his four chickens?" B: Té?e, NEG,  [pi?éye?]Foc  ^u?  xe?  one[DlM]  PERS  dem  e  s-w?x-ùm-s  te  eikn.  DET  NOM-PROG-MDL-3SG.POSS  OBL  chicken  "No, he only has [onejFoc chicken." Utterances were coded for one of six focus types: C P focus, V P focus, narrow verb focus, subject focus, object focus, or number quantifier focus. Focus type served as the independent variable. Each utterance was then coded for syntactic type: default verb-initial word order, or non verb-initial order. Non-verb-initial order could be realized either as clefts or nominal predicate constructions as introduced in section 2.2.7. Occasionally, left-  extraposed contrastive topics (see section 2.2.4) were employed instead of clefts to mark narrow subject focus cases. The corpus analysis yielded a total of 338 utterances. Results are reported as absolute numbers and percentages for each focus type. 3.3.3 Results: Wide CP focus Sentence-wide, or C P focus, answers the question What happened? so that (roughly) all information in the sentence is new. In chapter 2,1 showed that the default word order in wide C P focus cases is verb-initial. This hypothesis was tested here empirically, by examining the actual realization of wide focus cases across the collected corpus. As expected, wide C P focus cases were typically verb (9-10) or auxiliary initial (11-12). (9)  A : What's going on in the picture? B:  V [kstni-m  e  pé[p]ye?  rodfish-MDL  DET one[DiM]  te  ^u?sqâyx'^]Fnr.  OBL  man.  "|One man is fishinglpoc( 10)  [at the beginning of a discourse] [ki?-kéy  ek^u  e  téwn  i:  spiVxav^^JFor.  AUG-quiet  EVID  DET  town  DET  day.  "[Yesterday sure was a quiet day in townlpoc' (11)  A:  What happened?  B:  Aux [?éx  V xe? cax-t-0-és  PROG D E M clean-TR-3o-3TS  S  O  t  n-sxâywi  e  swux'^tJFor.  DET  IsG.PS-husband  DET  snow  "[My husband was cleaning up the snowlpoc-" (12)  [deciding what will happen next during a translation activity] [Nes  kéy-nm-s-t-s-nJFor.  go  follow-deliberate.progress-CAUS-TR-2SG.o-l SG.TS  "[I'll M o w youjpoc-" Occasionally, however, a predicate construction was used to answer wide focus questions. The predicate in these cases is typically a quantified NP (13) or locative  demonstrative (14). As the literal translations show, these responses are odd in English, but acceptable in Salish (Koch 2007c; see also Davis et al. 2004 on St'ât'imcets and Northern Straits Salish). I will return to this issue when discussing the semantics of clefts in chapter 7. (13)  A : What was going on yesterday? [x^it  ek^u  xe?  many  EVID  DEM  k  ?éx  n  IRL  be  in  i DET  tk  séytknmx  OBL.IRL  people  téwnjFnr. town  "[Everybody was in townlpoc-" (more literally ?? 'The (ones that) were in town were lots of people.") (14)  A : kén-m  met  xé?e.  what.happen-MDL CNSQ D E M  "What happened?" B: Urn, um, ne  [né?  ek^'u  xe?  k  S"^y-ép  us  there  EVID  DEM  IRL  burn-iNCH  3.CONJ  n?^pénk-s  in.DET under-3.P0SS  e: ...  e  D E T . . . DET  s-ia-  e...  e  sq'^iyt]Fnr.  NOM-eat-  DET... DET fruit  "[A fire started under the fruiMlpoc-" (more literally ?? 'The (place that) a fire started under the fruits was there.") Of 64 instances of CP focus that were identified in the corpus, only 6 (9.4%) deviated from the verb-initial word order by employing predicate constructions. 3.3.4 Results: Wide VP focus Wide V P focus, for example, answers the question What did X do? where agent X is known in the discourse. Thus, a verb (and its object, if it is transitive) are focused, but the subject is not. Nuclear stress remains rightmost. In (15B), the focused V P is a new information focus answering the wh-question. In (16B), the V P is contrastively focused with the V P in the preceding question. Both cases results in standard verb-initial utterances. (15)  A : sté? what  k  s-zéy-tn-s  e  IRL  NOM-do-iNSTR-3SG.POSS  DET Flora  "What did Flora do yesterday?"  Flora  +  spi?xâwt.  DET  day  B: [q^ic-m]Foc launder-MDL  ek*u  xe?e  e  Flora  i  spi?xâwt.  EVID  DEM  DET  Flora  DET  day.  "Flora [did laundry[ppc yesterday." (16)  A : ?^6y-t sleep-iM  k"'  n.  2sG  Q  [whispering]  "Are you asleep?" B: [ Q i i wake  kn  nuk^  ?éyi]Foc.  [QIÎIFOC  kn  nuk^.  IsG  EViD  now.  wake  ISG  EVID.  "I'm [awake nowjpoc- I'm [awakeIFQÇ." In the next two examples, consultants are describing photos about a woman, Michelle. The subject is known, but the V P is new information in each case. The first example has an unergative verb 'to laugh,' while the second has an intransitive middle 'win' with oblique object 'lots of money.' Both examples are in the standard auxiliary-initial word order. The topical subject is elided. Nuclear stress is rightmost. (17)  Mmm. [?ex  ?e-?ùy-ifi]For.  Mmm. PROG  AUG-laugh-MDL  "Mmm. She's [laughingjpoc." (18)  [X*ùy xe?  ^x^-um  tk  x^it  tk  snûye.]Foc (laughter!  FUT  win-MDL  OBL.IRL  much  OBL.IRL  money  DEM  "She's [gonna' win lots of moneylpoç!" [laughter] In the next example, B's response is split into two intonational phrases (indicated by the comma). The gap is phonetically marked by a !4 second pause, and declination is reset in the second phrase. In the first intonational phrase, nuclear stress falls rightmost, on the phrase-final demonstrative (interesting because this is a functional element). The second phrase, 'the man,' has only one stress, which is the nuclear stress by default. (19)  A : sté?  met  xe?  k  s-cwu-s  what  CNSQ  DEM  IRL  NOM-WOrk-3. POSS  e  tâkte  k  sm?é?m-s  c?éyt.  DET  doctor  IRL  wife-3.POSS  today  "What did the doctor;'s wife do today?"  B: l)tîx^-0-0-es yell-TR-3o-3TS  ek^u  jçePe,  e  ^u?sqâvx^]Fnr.  EVID  D E M , DET man.  "She [yelled at him^Foc-" (literally: "She yelled at himi, the man;.") The term V P focus also describes cases where adjectives serve as the main predicate and focus. The case below is interesting for stress-focus accounts, because the focused portion in B's reply is restricted to the adjectival root plus proportional suffix i'^mi?me? 'small.' However, stress falls on the lexical suffix for 'house,' =(m)éix^}^ Contrastive focus does not induce a stress-shift from the lexical suffix to the root 'small.' (20)  A : 6, oh,  xzum nka  k  e?-citx^.  big  IRL  2sG.POSS-house  EVID  "Oh, I guess it's a big house." (literally "Oh, I guess your house is big.") B: teté? NEG  k  s-xzum-s.  [k^miVmeVlFor-mefx^ ^u?  IRL  NOM-big-3sG.POSS.  small.PRP=house  xé?9.  just D E M  "It's not big. It's [smallVor." Of the V P focus cases in the corpus, 76 were in the default auxiliary- or verb-initial word order. Only one employed a cleft structure. In that case, the speaker was describing a series of photographs describing the weekend activities of a man named Chris. The head of the cleft is a null element here; it could be the given topic, 'Chris,' or perhaps a null situational deictic like 'here' or 'now.' In the latter case, the focus may extend all the way to the left edge of the clause (as indicated by the bracketing below). Note that the clefted focus is in this case a null element; this is surprising for "stress-focus" accounts, but not if stress and focus diverge. I'll introduce some more examples along these lines in section 8.4.1. (21)  [?é  5Cu? 0  e  s-cuk^-s  CLEFT  PERS 0  DET  NOM-finish-3SG.POSS DET  "He's  e  s-cw-um-sjpnr. NOM-WOrk-MDL-3SG.POSS  [finished work now[poc-"  (more literally: "It is [0[ that he is finished with his work.")  ^° The initial [m] in the lexical suffix appears to be a speech error here, and is corrected by the speaker in the following sentence. The usual form for this lexical suffix is =éix^.  3.3.5 Results: Narrow verb focus Narrow verb focus answers a question like What did Gary do with the beans? Under a "cleft-focus" account, we might expect a response like It was picking that Gary did with them. Such a construction with a clefted verb is bad in English, and Nte?kepmxcin as well.^' Instead, narrow verb focus employs the default verb-initial word order. Subject and object are often elided. In this case, the verb may end up sentence-final, in the default nuclear stress position, as in B's reply below. (22)  A : Ké-s-t-0-s  t^m  what.happen-CAUS-TR-3o-3TS  xé?e  PERF D E M  e  s-cwu-s,  ne  s-pâq^.  DET  N0M-W0rk-3SG.P0SS, in.DET NOM-watch  "What did he do with the work (he did) in the book?" B: w?éx PROG  nke  hxl  xe?  ne?  [k^én-0-0-eslFnr.  EVID  PERS  DEM  there look.at-TR-3o-3TS  "He was just [looking atjpoc it." The narrowly focused verb may also be sentence-initial, with nuclear stress falling on a rightmost deictic. (23)  Teté? k  s-pâq^-0-es.  NEG  NOM-watch-TR-3o-3TS. bum-TR-3o-3TS  IRL  [S'^ey-t-0-és]Foc  t^ra  xe?  né?e.  PERF  DEM  there  té?e.  "He didn't look at it. He [bumedlpor it." (24)  Teté? k  s-nfk-0-es.  [?ùpi-0-0-s]Foc  ?tu?  ciy  NEG  NOM-cut-TR-3o-3TS.  eat-TR-3o-3TS  PERS  similar D E M .  IRL  "He didn't cut it. He [atelpoc it like that." Nuclear stress may fall on a rightmost object that remains overt. There does not appear to be any deaccenting effect (I'll take this matter up in more detail in chapters 4 and  5).  ^' A n exception is in "only" clefts, which employ the "only" cleft predicate cuk"^, and can take clefted VPs as their single argument (see example 59 and footnote 23 in section  8.4.3).  (25)  A : What is the woman doing to the cow? B: ? é x PROG  xe?  k"^-- [k'^up-O-O-esJFoc e  DEM  k"'- push-TR-3o-3TS  smutec  DET woman  e  mosmos.  DET  cow  "She's [pushinglpoc the cow." In the present corpus study, all 19 cases of narrow verb focus retained the standard verb- or auxiliary-initial word order. 3.3.6 Results: Narrow subject focus Subject focus answers a question like Who fell into the river? or Who fixed my bicycle? The present corpus study confirms Kroeber's observation that narrow foci are restructured at the left edge, contrary to the "stress-focus" predictions of section 3.1. Nuclear stress, however, is retained at the right edge (typically old information in the residue clause). In the first example below, B's response employs an introduced cleft to focus the subject DP e Flora as the leftmost lexical element. The cleft clause contains old information, and is subordinated: it is introduced by complementizer e and the verb 'wear' carries -emus subordination morphology. Again, there is no deaccenting effect (I'll give acoustic phonetic support for this observation in chapters 4 and 5). (26)  A : ?e and  swét  xe?  k  s-Wm-s-t-O-emus  who  DEM  IRL  STAT-wear-CAUS-TR-3o-suBJ.EXTR  e  ?es-tip-tep-t  te  n^npâxn.  DET  STAT-black-AUG-IM  OBL  veSt  "Who is wearing the black vest?" B: ce CLEFT  xe?  [e  FloraJFoc e  s-tum-s-t-O-emus  DEM  DET  Flora  STAT-Wear-CAUS-TR-30-SUBJ.EX  DET  e  ?es-tip-tep-t  te  nfcnpâxn.  DET  STAT-black-AUG-IM  OBL  vest  "fFloralpoc is the one wearing the black vest."  (27)  A : swét who  xe?  k  cu-t-émus  e  s-i-a?xâns  DEM  IRL  make-TR-SUBJ.EXTR  DET  NOM-eat  "Who made the food?" B: cé CLEFT  e  [Kârsten]Foc  e  cu-t-émus  DET  Karsten  DET  make-TR-suBJ.EXTR D E M  e  s-ta?xâns-c.  DET  NOM-eat-3sG.POSS  xe?  "It was [KarstenJFoc that made his food." Since subjects are typically DPs, which cannot be predicates, introduced cleft predicates are usually employed to mark narrow subject focus. Among the exceptions in the current study are quantified subject DPs, which usually surface as predicate constructions: (28)  A : Who's shopping? B: [ x ^ i t much  te  séytknmxJFoc 0  ?éx  téw-m.  OBL  people  PROG  buy-MDL  DET  "[Lots of peopleJFoc are shopping." (more literally: "The (ones that) are shopping are lots of people.") (29)  A : ?e and  swét  mef  tux^e?  e?-s:cméyt  who  CNSQ  of.DET  2sG.P0SS-kids  k  ye-mm-t-0-9mus  xé?e.  IRL  g00d-REL-TR-30-SUBJ.EXTR D E M  "And w h o of y o u r kids l i k e s i t ? " B:  ( t é k m J F o c us all  xe?e  3.CONJ D E M  e DET  s-ye-min-t-0-iyxs. NOM-g00d-REL-TR-30-3PL.TS  "They [aiUpoc l«ke it." Assuming that the cleft predicate selects a focus argument of type e, this patterning of quantifier foci as predicate constructions is not surprising. Under the standard analysis (eg. Bach 1989), quantified noun phrases are not of type e. Therefore, they cannot be the focused argument of the cleft predicate ce or ?e. In section 3.3.8, we shall see that number quantifiers pattern the same way. The strong universal quantifier in (29) is unusual in this regard; it's not clear whether it's really acting as a predicate here; in St'ât'imcets Salish the universal quantifier can't act as the main predicate (Matthewson 1996).  Emphatic pronouns may also surface as nominal predicate constructions (though they may be the head of clefts with a null cleft predicate - Thoma 2007 on St'at'imcets Salish). (30)  A : ?e CLEFT  swét  k  ta?xâns  tk  seplil  t  who  IRL  eat  OBL.IRL bread  DET  snwénwen. morning  "Who ate some bread this morning?" B: [Ncéwe?]Foc ISG.EMPH  e  taVxans  te  seplil t  snwénwen.  DET  eat  OBL  bread DET  morning  "[IIFOC ate some bread this morning." (more literally: "[IJpoc was the (one that) ate some bread this morning.") Of 56 subject focus sentences in the data, only 4 (7.1%) retained basic auxiliary- or verb-initial order. Most (52, or 92.9%) had subjects appear at the left edge, confirming Kroeber's "cleft-focus" generalization. However, it should be noted that 7 of these cases (12.5%) had the focused subject in the contrastive topic position. This indicates that, unsurprisingly, other discourse factors such as topic-marking are also at work, and may sometimes conflict with straightforward focus-marking as clefts or nominal predicate constructions. One of the exceptions in which basic verb-initial word order was retained is shown in the example below. A and B are discussing a series of pictures illustrating Stef's day at the lake. B's reply focuses narrowly on the subject, Stef, yet this focus is not clefted in this case. This may also be because of additional discourse factors, namely that B is metalinguistically correcting A ' s use of 'one person' to the more specific Stef, and so does not alter the basic word order of A ' s utterance. (31)  A : W?éx PROG  xe?  ne?  D E M there  ?es-'lâq-ix  e  STAT-sit.PLURAL-AUT DET  ?e'l ?es-té'l-ix  e  pépye?.  and STAT-stand-AUT  DET  one[DIM|  "People are sittin' there id one person is standing." B: ?es-tét-ix STAT-stand-AUT  [e  SteflFoc-  DET  Stef  "I Stef IFOC is standing."  séytknmx people  3.3.7 Results: Narrow object focus Results for narrow object focus were similar to those for subject focus: focused objects were typically restructured to the left edge, in clefts or nominal predicate constructions. Recall that all of the stress-focus mechanisms examined in section 3.1 predicted that narrow object focus should employ the default word order (Aux)VSO, because the focused object appears in the nuclear stress position in this default word order. Yet the results contradict this hypothesis. We already saw an example of a nominal predicate object focus ((33) in chapter 2), and a clefted object focus in ((34) in chapter 2). While the focus is at the left, the nuclear stress position is on the right. Below are some further examples. (32) shows a cleft with a focused object DP e sq^axts 'her leg', and example (33) shows a nominal predicate kapi 'coffee' that also doubles as the focused object. (32)  tém  ek^ute?  NEG.SUBJ EVID that  k  s-maV-t-s  *  IRL  NOM-break-iM-3sG.POSS  DET  kéyx-s. hand-3sG.POSS.  "I heard she didn't break her arm." cé  ek*u  [e  sq^âxt-s]Foc  k  CLEFT  EVID  DET  leg-3SG.POSS IRL  maVt. break-iM  "It was her [leglpoc she broke." (33)  A : Sté? WHAT  tk  s-?ùq™e?  OBL.IRL  NOM-drink  k  e?-s-cw-ùm  t  IRL  2SG.POSS-NOM-make-MDL  DET  snwenwen. morning  "What did you make to drink this morning?" B: IKâpiJpoc xe? coffee ?e'l and  DEM e DET  e DET  n-s-cw-um. ISG.POSS-NOM-make-MDL  n-s-?ûq'^e?  t  snwenwen.  ISG.POSS-NOM-drink  DET  morning  "I made IcoffeelpQc to drink this morning." (more literally: 'The (thing that) I made was [coffee Ippc, and that I drank this morning.")  Of object focus cases in the corpus, 43 (79.6%) were produced as left edge clefts while 11 (20.4%) remained in situ in the basic verb-initial word order. Thus, the majority of instances confirm Kroeber's "cleft-focus" generalization. 3.3.8 Results: Number quantifîer focus Narrow focus on number quantifiers typically results in the quantified DP being focused at the left edge of the clause as part of a nominal predicate. There is a variation on this pattern, however. The basic pattern is shown in the next example (see also 7B). The complex nominal predicate mus (Xu?xe?)tk sifcu?'four shoes' in (34B) is the left edge predicate containing the focus mus 'four.' The second position clitics ^u?A^e?follow the first word of the complex predicate, mus. The residue clause is introduced by the determiner e and carries nominalizing morphology s-...-s on the verb tx^'up 'buy.' (34)  A: e  sxaywi-s...  DET  cu-t  e  husband-3sG.POSS... say-iM DET k  s-k^n-9m-s  tk  COMP NOM-grasp-MDL-3SG.POSS  sxaywi-s husband-3sG.POSS ciyksttk  OBL.IRL five  sitcuV-s.  OBL.IRL shoes-3sG.POSS  "Her husband said that she got five pairs of shoes." B: [Musjpoc tu? four  PERS  xe?  tk  si+cu?  DEM OBL.IRL shoes  e  s-tx^-up-s.  DET  NOM-buy-INCH-3SG.POSS  "She only bought |fourjpoc pairs of shoes." (more literally: 'The (thing that) she bought was just Ifourjpoc pairs of shoes.") The next example follows a similar pattern. Note that the focused root 'three' in the last line of (35B) does not carry even word-level stress, which falls on the lexical suffix =eyus; this suffix, meaning 'leg' or 'pants,' carries old semantic information, but is still stressed. This is unexpected under "stress-focus" and "destress-given" accounts. Nuclear stress remains rightmost; contrastive focus on the root 'three' does not attract the stress from the lexical suffix =eyus. (35)  A : Cu-t  xe?  met  e  Natalie  k  s-k^n-am-s  DEM  CNSQ  DET  Natalie  COMP  N0M-get-MDL-3SG.P0SS  ... umm...  ciyci  tk  OBL.IRL... umm...  new  OBL.iRLpants-3sG.POSS  Say-IM  tk  "Natalie said that she got new pants."  sqéyus-c.  B: Teté? k  s-[pi?éy-...  NEG IRL NOM-one-...  péye?]Foc-s  tu? k s-tx^-ùp-s ...  one-NOM  PERSIRLNOM-buy-INCH-3SG.PS...  "She didn't buy just [onelpoc pair (more literally: "It was not just [onelpoc (pair) that she bought...") [Ke?t]Foc=éyus  tk  sqéyus  Three=pants  OBL.IRL pants  e  s-tx'^-up-s.  DET  NOM-buy-iNCH-3sG.poss  "She bought Ithreelpoc pairs of pants." (more literally: 'The (thing that) she bought was [threelpoc pants.") In the second pattern, the nominal predicate may also be separated in the surface order, with the number quantifier remaining at the left edge but the N P right extraposed. In (8B), pi?éye?'one'  is at the left edge while te cikn 'chicken' is extraposed to the right, past  the residue in the clausal DP e sw?xûms 'that he has.' In the next example, B's reply restructures the clause such that thé predicate is the complex nominal 'six potatoes,' yet only the quantifer surfaces at the left edge. The N P portion, tk stqols 'potatoes,' is right extraposed past the cleft clause e nsx^'ox^st 'that I want' For the "destress-given" generalization of information marking, this type of extraposition is rather surprising, since a given element, tk stqols, is being "moved" into the rightmost nuclear stress position. This extraposition can be seen as an interpretive strategy to isolate the narrow focus at the left edge of the clause, though it seems to be optional in cases like these (see section 8.4.4 for more discussion). The subscripted't' represents the 'trace' of the extraposed NP. (36)  B: ké?  k  is.it.case IRL  e?-s-w?x-ùm  tk  stqols.  2SG.POSS-NOM-PROG-MDL  OBL.IRL  potato  "Do you have potatoes?" A : he?ây, yes,  k^'inex  k  how.many IRL  e?-s-x^6x^-t. 2sG.P0SS-N0M-want-IM  "Yes, how many do you want?" B: [^aqmekstJFoc tp e six  tp DET  n-s-x^6x"'-st  [tk  stqolsrQp.  ISG.POSS-NOM-want-REFL  [OBL.IRL  potato[p  "I want I six IFOC "It potatoes." (more literally: is [sixlpoc tp that I want [potatoesdp.") The rightmost N P is in fact extraposed, and not an in situ argument of the subordinated verb in the residue clause. This can be seen when the embedded verb is  transitive (37): the right extraposed NP is always introduced by the oblique marker {te or tk), and not by the determiners e or f which introduce direct arguments of transitive verbs. Thus, like English, structures which focus a modifier of the NP, but "strand" the NP as an argument inside the residue of the clausal DP, are not attested (38b). Unlike English, the NP portion of the predicate may be extraposed to the right (38c provides a rough surface approximation). (37)  A : w?éx  n  xe?  x^f?-m  tk  - uh -  ciyci  PROG  Q  DEM  look.for-MDL  OBL.IRL  - uh -  new  tk  - uh -  qemut-s.  OBL.IRL  - uh -  hat-3SG.POSS  "Was she looking for some new hats?" B: [Ke?i-és]Foc three  nke  xe?  th  EViD  D E M t^  k  s-pùn-m-0-0-s  IRL  NOM-find-REL-TR-3o-3TS  jtk  (/*e)  qemut-slt,.  [OBL.IRL  (/*DET)  hat-3SG.POSS|h  "She found [threelpoc hats." (38)  a. The (thing) that she was looking for was three hats. b. *The (thing) that she was looking for hats was three,  [also ungrammatical in N l ]  c. *Three t^ was [the (thing) that she was looking for] hats^. [grammatical in N i . ] These split nominal predicate structures are attested independently of narrow focus structures, as can be seen in the two pairs of attributive sentences below. The N P smuiec 'woman/lady' can either precede or follow the subject Mary, but in either case is part of the complex nominal predicateputi (xe?) tk smuiec 'pretty lady.' (39)  a. puti pretty  xe?e  tk  smuiec  e  Mary  DEM  OBL.IRL  woman  DET  Mary  e  Mary  [tk  smuiecl^  DET  Mary  (OBL.IRL  woman ]„  "Mary is a pretty lady." b.  puti  xe?  tw  pretty  D E M t^  "Mary is a pretty lady." In the data collected for this study, there were 68 cases of narrow focus on a number quantifier. Only 1 case (shown below) was produced using the default verb-initial order.  A : cké?9  (40)  met  Whieh  xé?  k  s-wik-t-0-s  CNSQ DEM IRL  NOM-see-TR-3o-3TS  e DET  st?âsze. squirrel  "Did they see any squirrels?" B: He?ây, yes,  wik-t-0-iyxs  ek*u  xe?  see-TR-3o-3PL.TS  EVID  DEM  [pi?éye?]Foc  te  st?âsze.  onefDiM]  OBL  squirrel  e DET  S-...  NOM-...  "Yes, they saw [onejpoc squirrel." 3.3.9 Results: Summary Wide C P focus, V P focus and verb focus typically retains the basic verb-initial order. Narrow focus on objects, subjects and quantifiers generally results in a cleft or nominal predicate construction with the focus restructured to the left edge of the utterance. Table 3.2 summarizes the findings. Implications are discussed in the next section. Table 3.2  Focus type and syntactic realization - a corpus study of 338 cases Focus Constituent  Word order  CP  VP  Verb  Object DP  Subject DP  QP  V - or Aux-  58  76  19  11  4  1  (90.6%)  (98.7%)  (100%)  (20.4%)  (7.1%)  (1.5%)  6  1  0  (9.4%)  (1.3%)  (0%)  (79.6%)  (92.9%)  (98.5%)  initial Non Verbinitial  3.4  Discussion  3.4,1 Implications for the "stress-focus" generalization The findings presented in section 3.3 support Kroeber's (1997, 1999) observation that left edge clefts and nominal predicate constructions are employed to mark narrow focus in Nte?kepmxcin. The dissociation of narrow focus from the nuclear stress position is unexpected under a "stress-focus" account, in which a variety of strategies may be employed to ensure that the focus and nuclear stress coincide.  Unlike English speakers, speakers of Thompson Salish do not retain the default word order, in which "stress-focus" is satisfied by simply altering the location of the nuclear stress to the narrow focus constituent. Neither is there movement of the focus to the rightmost nuclear stress position (as in Portuguese). Unlike German, there is also no scrambling of unfocused material away from the right edge. Narrowly focused subjects and number quantifiers surface at the left edge. Even objects, which are in the rightmost nuclear stress position in the default V S O word order, are generated in left edge clefts or NPCs when focused. This demonstrates the primary finding of this chapter, namely that a prosodie condition which aligns the focus with the nuclear stress cannot be what is driving the "cleft-focus" generalization in Nie?kepmxcin. Thus, the Stress Focus Correspondence Principle (Reinhart 1995) or other principles like it are at best language-specific, and not universal principles for stress languages. Narrow verb focus, too, is not marked by any sort of movement of the verb to the rightmost nuclear stress position. On the other hand, focused verbs are not clefted either. Instead, the default verb-initial word order is retained. The generalization that emerges is that the left edge of the clause is relevant for focus marking, while nuclear stress is not relevant. Verbs are already at the left edge, so default word order is used to mark V P , verb, or CP focus. Other narrow foci have to be clefted or turned into predicates to appear at or near the left edge. Thus, another generalization that emerges is that the focus is closely associated with a syntactic constituent, namely the predicate. Recall the generalization made in chapter 2, namely that matrix predicates are always leftmost in their intonational phrase. (41)  PREDICATE-LEFT:  Align the matrix predicate with the left edge of an intonational phrase. Now we can add a second generalization, which we will call F O C U S - L E F T . (42)  FOCUS-LEFT:  The focus is leftmost in an intonational phrase, [first attempt]  These constraints are similar to how Krifka (1998b) describes the German system of focus marking (see 24-25 in chapter 1). Recall that in German, nuclear stress surfaces immediately before the verb, and the focus also optimally surfaces in this preverbal surface position. Krifka describes this situation as a competition between the constraints V E R B -  RIGHT  and F O C U S - R I G H T . Since V E R B - R I G H T is undominated, the focus does the next best  thing, and surfaces preverbally, or as rightmost as possible. Focus is marked in a similar way in Thompson Salish. I have claimed that there is an undominated requirement that matrix predicates be initial in an intonational phrase. When verbs or VPs are focused, the focus therefore aligns with the left edge of the clause in the default V S O order. For wide C P focus too, which necessarily includes the entire clause, the focused C P is by default aligned with the left edge. When narrow focus falls on subjects, objects, or number quantifiers, focus is marked using a cleft or a nominal predicate construction. For clefts, used to mark focus on DPs like proper names, the cleft predicate is at the left edge of the clause. This is because DPs cannot be predicates in Thompson Salish, and indicates that P R E D I C A T E - L E F T is undominated. However, using a cleft ensures that the focus is still the leftmost lexical item in the clause, therefore satisfying F O C U S - L E F T "as best as possible" given the undominated status of PREDICATE-LEFT.  That is, the cleft predicate is a functional element, and cannot be focused.  In the case of nominal predicate constructions, both P R E D I C A T E - L E F T and F O C U S - L E F T are satisfied, since nominal predicates are both the focus and the matrix predicate of the sentence. 3.4.2 Implications for focus projection For focus projection, results also did not turn out as expected under a Selkirk-type system (1995 - see section 1.3.4 for discussion). The first prediction (4a) was that wide focus (VP or CP) would be marked using the default verb-initial word order. The results (table 3.2) confirm this first prediction, and are thus far compatible with a system where, like in English, nuclear stress on the righmost object in a V S O configuration can project to the V P or C P level as well. (43)  Hypothetical focus projection from rightmost object to V P and C P in NteVkepmxcin Icplvp wik-t-O-nCf see-TR-3o-l SG.TS  xe?  Ippe Monikjf  DEM  DET  Jfpoc-  Monique  "[cpl IvpSaWf iDpMoniQuelflflf.Foc-" However, the predicted link between wide focus and narrow object focus (4b) was not found, since narrow object focus was marked using left edge clefts or NPCs. If focus projects horizontally from the nuclear stress position, we would have to say that, in cases of V S O  word order, focus M U S T project beyond the object, and cannot be confined to narrow focus on the object. The same situation holds of intransitive predicates where the subject is rightmost. In the standard verb-subject order, nuclear stress falls on the post-verbal subject. However, narrow subject focus is expressed structurally. Once again, if the nuclear stress is relevant for focus projection, in standard V-S order we would have to stipulate that focus M U S T project beyond the subject (in fact, "project" downward to the VP). (44)  Impossible focus configurations in Nie?kepmxcin verb-initial order: Lack of focus projection beyond the nuclear stress a. Narrow object focus: Who did you see? #[cp [vp w îk-t-0-ne  xe?  see-TR-3o-lSG.TS  [^p e Monik]poc ] J.  D E M DET Monique  intended: "(cpl [ypSaw [ppMoniquelpoclJ." b. Narrow subject focus: Who ran? #lcp[vpt*i''^-iyx ]  [DP  run-AUT  e  ÉuTsaâyx^JFoc ]•  DET  man  intended: "[cp [DP The manjpoc (vp ran]]." This is an undesirable and ad-hoc stipulation: objects must be restructured to the left edge of the clause when narrowly focused, but in those eases, nuclear stress is not relevant at all in indicating the focus. The focus cannot project from the nuclear pitch accent in clefts or NPC structures, since the rightmost element (the cleft clause, or residue clause) is not part of the focus. We would have a situation where focus projected from the nuclear pitch accent in the default word order, but did not project from the nuclear pitch accent in focus constructions. (45)  Narrow focus cannot be "projected" from rightmost nuclear pitch accent in clefts [cp[vp  cé  xe? [ppe Moniklpoc  CLEFT  D E M DET  " I c p i t Ivpwas  1  Monique  [ppMoniquelpocI that I saw)."  e  wik-t-0-ne 1 j .  DET  see-TR-3o-l SG.TS  The more straightforward account of Nte?kepmxcin simply abandons the idea of focus projection from pitch accents. I conclude that horizontal focus projection (from accented complements to unaccented heads) is not operative in Nie?kepmxcin. This conclusion is in line with Biiring's (2003, 2006) finding that "horizontal focus projection" in English and German is in fact a matter of default prosody, and does not require special rules of focus projection. If, in NieVkepmxcin, nuclear pitch accents are not relevant for marking focus, we don't expect to see any "horizontal focus projection" effects from pitch accents either. The dissociation of the focus from the nuclear stress position suggests that nuclear stress assignment is a prosodie phenomenon, and not a syntactic one. Davis (2007) has stated this in the strongest possible terms: (46)  Postulate 1 The Nuclear Stress Rule is a purely prosodie phenomenon.  (Davis 2007)  This differs from previous conceptions of nuclear stress, in which it was related to the deepest syntactic constituent (eg. Cinque 1993). As stated in (46), surface position, prosodie phrasing, and rules for headedness (left or right) would be the only factors determining where the nuclear pitch accent falls; any relation to syntax is only indirect. This was the strong position claimed for Nie?kepmxein in chapter 2, where I observed that nuclear stress falls on the righmost constituent in the surface order, whatever its syntactic category. However, to the extent that prosodie constraints refer to syntactic categories, there continues to be a role for syntax in constraining where the nuclear pitch accent will fall. Evidence that certain syntactic categories (like prepositions in English - eg. German et al. 2006) are more resistant to carrying the nuclear stress, for example, suggests that (46) is too strong. It may turn out that in Salish, too, certain syntactic categories are disprefered as hosts for the nuclear pitch accent; for example, rightmost adverb phrases may resist carrying the nuclear stress in favour of preceding direct objects. However, this issue will have to await future research. Thus far we have seen that the location of the nuclear pitch accent is not important for focus projection. So what is relevant? Focus is marked by pitch accent in English and German, and focus projects vertically. Since any syntactic category can bear the nuclear pitch accent, focus can therefore project from any syntactic category (BUring 2003). So, accented heads, arguments, adjuncts or specifiers are all sources for focus projection (see section 1.3.4 for examples and discussion). The situation in Salish is rather different, as noted by Davis (2007). The generalization for all types of focus is that focus aligns to the left edge, which is also the  position of the matrix predicate. Arguments, adjuncts and specifiers do not appear at the left edge of the intonational phrase containing the matrix predicate. We can make the strongest possible statement: focus always and only projects from the matrix predicate. The focus projection line is thus always from verb (or predicate), to verb phrase (or predicate phrase), to IP, to CP. Thus, in Salish, the "focus set" (Neeleman and Reinhart 1998) is restricted to {Predicate, PredP, TP, CP} (Davis 2007), where "PredP" is any predicate phrase (NP, A P , VP).This is a syntactically oriented view of focus projection, argued for by Davis (2007): (47)  Postulate II Focus projection is a purely syntactic phenomenon.  (Davis 2007)  In languages like English or German, on the other hand, any syntactic category can project focus vertically onto a dominating syntactic category. (48)  Unrestricted vertical focus projection: Any subconstituent can project focus.  (Buring 2003)  But we have made the strong claim that in Thompson Salish, focus projects from the predicate (see also Davis 2007): (49)  Restricted vertical focus projection (Salish) (first attempt| The head of the predicate can project focus. {Pred, PredP, TP, CP} The examples below illustrate focus projection from the predicate in Thompson  Salish. In the case of verb, V P or CP focus in the default V S O word order, the leftmost position is occupied by the verb. The focus minimally includes the verb, but may include the V P or the entire CP. I've deliberately kept the syntactic representations fairly simple for the present purpose of illustration, by not labelling additional intermediate projections such as AgrS or AgrO (subject and object agreement nodes), and so on. (50)  Focus projection in the default V S O order a. Narrow verb focus (no projection beyond verb head) A : What is the cow doing to the cat? B: ?éx PROG  xe?  [v mi?x-e-t-0-és]Foc e  DEM  kick-DRV-TR-30-3TS  "The cow is [ykickinglpoç the cat."  DET  mosmos  e  pus.  COW  DET  cat  b. V P focus projected from verb head to V P A : What is the woman doing? B: [cp?ex[vp[v ?es-k'^én-s-t-0-s] PROG  V  [ope  STAT-l00k-CAUS-TR-30-3TS  pus] ]mr ].  DET cat  "She's [looking at the çat]Foc-" c. CP focus projected from verb head to C P A : What's happening in the picture? B: [cpw?éxne?  [vp[v ?es-téi-ix]  PROG there  [ppne tmix^ [ope srnuleç]]]Foc-  STAT-stand-AUT  in.DET land  DET woman  "[A woman is standing on the ground therejpoc" In nominal predicate constructions, the focus and matrix predicate coincide; focus may be confined to this constituent, or may project on to the clausal level from there (51). (51)  Focus projection in nominal predicate constructions a. Narrow focus on the N P (no projection beyond N P predicate) A : What did you make to drink this morning? B: [NpKapiJFoc coffee  xe?  e  DEM  DET  n-s-cvy-um. IsG.POSS-NOM-make-MDL  "I made [coffee IFQÇ." (literally: 'The (thing that) I made was [coffee]Foc." b. C P focus projected from complex NP predicate A : What was going on yesterday? B: [CP[NP x ^ i t  \j many  ek"'u  xe?  tk  séytknmx]  EVID  D E M OBL.IRL people  k  ?ex  n  t  IRL  be  in  DET town  tewnJFoc-  "(Everybody was in town]Foc-" (literally ?? "[The (ones that) were in town were lots of peopleIpoc-")  For clefts (52), the focus is the cleft head that serves as internal argument to the cleft predicate; here, focus necessarily includes this predicate complex since it projects from the cleft predicate head to the V P including the cleft head. From there, focus can again optionally project to the clausal level. (52)  Focus projection in introduced clefts a. Narrow focus on the cleft head (no projection beyond cleft predicate VP) A : / heard that it was Fred who painted it. B: Té?9. [cp[vp[vcé] no,  V  [opt  CLEFT  Ross] ]FOC e  DET Ross  pint-a-t-0-mus].  DET paint-DRV-TR-30-SUBJ.EXTR  "No, it was (RQSS]FOC that painted it." b. C P focus projected from cleft predicate A : What's going on in this picture? b. [cp[vp[vcé]  xe?  \j \l CLEFT  ne? [ope  pétusk^'u  DEM there  DET  n  i  ?éx  ut  k9tni-rTi]]]Foc.  in  DET  PROG  Ipl.coNJ  rodfish-MDL  lake  "[This is the lake we went fishingJFoc-" Since the verb is a subconstituent of the residue clause (eg. cwum 'make' in 51a), we predict that neither bare nor introduced clefts can be used to mark V P focus, since we cannot project focus from the cleft head to a subconstituent of the cleft clause. The results (table 3.2) confirm this prediction. While some 10% of wide focus cases were expressed using focus structures, only one marginal case of V P focus was expressed this way. The fact that not more than 10% of wide focus cases were expressed using focus structures may be because they are biclausal (section 7.1), so they are less economical to produce when standard V S O structures are available. In the case of complex NPs (modified by adjectives or number quantifiers, for example), the focus may comprise only one part of the clefted nominal predicate. In these cases, the focus may be restricted to a subconstituent of the predicate phrase that is not the predicate head. This suggests that (49) is too strong focus projection; the predicate phrase is important, but not the predicate head. Also, an additional principle needs to be invoked to cover cases of narrow focus that are only a subconstituent of the predicate phrase:  (53)  a. Restricted vertical focus projection (Salish) [revised I The predicate phrase can project focus. b.  {PredP, TP, CP}  Narrow focus (Salish) Focus may be restricted to the predicate phrase or any subconstituent of the predicate phrase.  This differs from "stress-focus" languages like English, where the focus may be restricted to the constituent bearing the nuclear stress, and bears no relationship to the predicate phrase. In the example below, a simple number quantifier ciykst 'five' is the narrow focus, though it is embedded within a complex nominal predicate 'five pants.' There does not appear to be any deaccenting after the focus (a fact I show in chapter 5). (54)  Narrow focus on a number quantifier within a complex predicate A : K^inex  xe?  how.many D E M  tk  sqéyus  k  s-tx^up-s.  OBL.IRL  pants  IRL  NOM-buy-iNCH-3sG.POSS  "How many pants did she buy?" B: [NP [Ciykstjpoc five  xe?  tk  sqéyus-c]  e  DEM  OBL.IRL pantS-3SG.POSS DET  s-k'^n-gm-s. NOM-get-MDL-3SG.POSS  "She got [fiyejpoc P^irs of pants." We have seen that, while stress languages like English or German allow for focus projection from any syntactic constituent, focus in Salish projects from the predicate phrase. In addition, narrow focus can be restricted to any subconstituent of the predicate. In turn, this suggests a syntactic view of focus projection (47). The results of the corpus study in this chapter are not explained by the common prosodie account of focus marking, as embodied in "stress-focus" proposals. We may ask if there is not a role for prosody, however. For one thing, there are two interfaces of the grammar where focus projection is relevant: (i) at the semantics-syntax interface, and (ii) at the phonology-syntax interface. The former may be syntactically driven (assuming that focus is marked as a syntactic category), and satisfies the interface between syntax and logical form. The primary role of the interface between syntax and phonological form (PF), however, is to communicate information from speaker to hearer. Focus projection in this case serves to allow the listener to reconstruct where the focus is in the sentence, and is necessarily mediated phonologically since communication is packaged in a speech signal.  Where the locus of the nuclear pitch accent mediates focus projection in English, I have suggested that alignment with the left edge of the intonational phrase may be a phonological principle mediating focus projection in Thompson Salish. I'll return to this question in chapters 7 and 8.  3.5  Clefts versus nominal predicate constructions: The role of contrast We have seen that the two focus structures, clefts and nominal predicates (NPCs), can  be employed to mark both informational focus or contrastive focus. In fact, due to the FOCUS-LEFT  constraint, narrow new information focus is necessarily expressed structurally,  and not by employing the standard verb-initial word order. We might wonder, however, what difference in interpretation exists between clefts and NPCs, if any. In addition, we might wonder how contrastive focus is expressed. In St'at'imcets Salish, Thoma (2007) found that clefts are used in contrastive contexts, while nominal predicate constructions are not. (55A') is not felicitous in the contrastive situation below, where the speaker has to identify only one of the available alternatives. (55)  Context: two pictures, one with a sleeping dog, another with a bear climbing a tree. Q: swat who  ku^î'^uytâl'men  [St'at'imcets Salish]  DET=sleepy  "Who is sleepy?" A: nit CLEFT  ti=sqâx?=a  ti=wà?  ?"'uyt  DET=dog=EXIST  DET=IMPF  sleep  "It's the dog who's sleeping." A': # sqaxa? dog  ti=wâ?  V^uyt  DET=iMPF  sleep  " A dog is sleeping."  (Thoma 2007)  Similar facts describe simple contrastive contexts in Nte?kepmxcin (Koch 2007c). To test for the difference between clefts and NPCs, consultants were presented with computer displays of either one object (new information condition) or two objects (contrastive condition) (Appendix C). In the experiment, animated computer characters asked questions about the objects on the screen. In the new information condition, the computer character  asked "What's this?" Consultants responded with the name of the item, using either a simple bare nominal predicate, or a cleft. (56)  A : sté? What  met  xé?e.  indeed  DEM  "What's this?" B: [ÉQéwe?]Foc onion  xé?e.  [nominal predicate construction]  DEM  "Those are [onions]poc." B': cé CLEFT  xe?  |e  q"léwe?]Fnr.  DEM  DET  Onion  [cleft]  "Those are [onions]Foc." In the contrastive focus condition, the animated character asked, for example, "Are these flowers?" while pointing at a book. Consultants responded by identifying the book, and continuing with a contrastively focused demonstrative to correctly point out the actual flowers; this last utterance was the target for analysis here, where two objects were being explicitly contrasted. Simple nominal predicate constructions were heavily dispreferred in this context; when they were used, NPCs always contained additional lexical items asserting the existence of a contrast between two items. It should be pointed out that questions were presented either as clefts (eg. 57A) or as nominal predicate constructions (eg. (6) in Appendix CI), with no obvious correlation to clefts or NPCs in the answer; that is, the syntax or semantics of the question did not force clefts in the answer. (57)  A : cé CLEFT  n  xe?  e  spaqm.  [pointing at book]  Q  DEM  DET  flowers  k  spaqms.  spaq* xé?e.  flowers.  book D E M .  "Is this flowers?" B: teté? NEG  xe?  D E M IRL  "That's not flowers. That's a book." T A R G E T : cé  in  CLEFT E M P H  [x?é]Foc  e  spaqm.  DEM  DET  flowers.  "IThatlpoc is the flowers."  Table 3.3 summarizes the results of the experiment. Only a single simple NPC was employed in the target eontrastive context. Table 3.3  Focus construction and focus type This is X .  (No.) T H A T ' S X.  [new focus]  [contrastive focus]  Cleft  33 (46%)  41 (73%)  Nominal predicate construction (NPC)  39 (54%)  1 (2%)  NPC plus other lexical items  0 (0%)  14 (25%)  Let us look at the additional morphology that indicates a contrastive interpretation. First, the results indicate that the cleft predicate itself, ce or ?e, is associated with contrast in a way that simple nominal predicate constructions are not. The cleft predicate may provide a weak implicature of exhaustivity; in chapter 7, we shall see that this implicature is easily cancelled, as shown for St'ât'imcets and Straits Salish by Davis et al. (2004; see also Shank 2003 for a semantic analysis of the [semantically] cognate cleft predicate n/f in Northern Straits Salish). In addition, the demonstratives xe?and x?e can not be predicates in Thompson Salish, so must appear in cleft structures when focused; however, this fact does not account for all of the clefts used in the contrastive condition, since often the N P s?ix'*'f 'others' was employed instead of a demonstrative (eg. the T A R G E T in 59B), yet was still clefted as a D P rather than acting as a bare nominal predicate.^^ Secondly, in more complex cases, nominal predicate constructions are extensively used for eontrastive focus; quantifiers in particular seem to be used in NPC structures rather than clefts (see examples in section 3.3.8). This fact follows if cleft predicates select for a focus of type e, and quantified NPs are not of type e (eg. Barwise and Cooper 1981, Bach 1989). However, results here indicate that additional morphology is employed to overtly mark the presence of a contrastive context. In (57B), the speaker responds using the emphatic marker m (in this case, in a cleft structure). Emphatic /n is a second position clitic. Though Thompson and Thompson (1996: 209) describe m as "rare" and "not completely  This points to the fact that the target clefts in this task are all equational, equating either a demonstrative with a DP, or equating two DPs. In chapter 7,1 will argue for a different semantics for cleft predicates that have a residue clause as an argument.  understood," I find that it is quite common exactly in this contrastive conversational context. Consultants state that a clause without m "is ok with just one item," but that a clause with ih "is for [when you are talking about] two things." The contrastive demonstrative x?é seen in (57B) also surfaced only in the contrastive focus condition, and never in the new information focus condition. When asked about using this demonstrative instead of the usual xé?e in the simple new information focus condition (with only one item on display), consultants commented: "what is the other thing you are talking about? You have to say what it is...." Thus, the demonstrative x?e seems to presuppose the existence of another item for comparison. In other cases, overt DPs that presuppose contrast, such as 'the others' or 'another one,' were employed in addition to the focus structures. In (58B), the DP e s?ix^i 'the others' serves as the subject of the nominal predicate AeFuse?'egg.' In addition, the prepositional phrase u CJ7 'over there' serves to overtly introduce contrast of the locations of the two objects, "here" and "over there." (58)  A : cé n CLEFT  met  xe?  e  ?e?ûse?.  Q  CNSQ  DEM  DET  u cf?  e  s-?ix*t.  to there  DET  NOM-other  [pointing at potatoes]  egg  "Are these eggs?" B: .... he?ùse? .... egg  "Those [others over there] are eggs." What Thompson and Thompson (1992: 139) call an aspectual marker, meioi (59B) (the 'consequential' CNSQ, in their terms) is used to indicate "change from present situation: anyway, anyhow; despite the evidence, contrary to expectations." Like other morphological items in contrastive focus contexts, this particle may presuppose contrast between situations in some way - the exact semantics of these elements remains a topic for future research. (59)  A : cé CLEFT  n  met  xe?  e  stqols.  Q  CNSQ  DEM  DET  potato  "Are these potatoes?"  [pointing at eggs]  B: Teté? xe? k NEG  stqols-c.  D E M IRL potatO-3.POSS.  He?ûse?  xé?e. 23  Egg  DEM.  "That's not potatoes. Those are eggs." T A R G E T : cé  rn  CLEFT E M P H  mei  e  s-?ixn  u cf?  e  stqols.  indeed DET NOM-other to there DET potato  "[That'slFoc some potatoes there." (lit. "It is those others over there that are potatoes.") The "perfective" particle ^am seems to introduce exhaustive interpretations (60, 61B, 62b; Thoma p.c.), even where only two salient alternatives are discussed (eg. babysitting or not in 62a, being here or returning home in (63)).  BUring (p.c.) points out that the first line of (59B), he?ûse?xé?e,  though not the  target in this task, may also be taken as contrastive (with the NP 'eggs' contrasting with 'potatoes'). Why is it not always clefted, then, but instead often expressed as a nominal predicate construction (see also the response 57B spâq'^xéîe)!  Because the contrastive stress  in the English answer here could fall on either the N P {Those are eggs) or both demonstrative and N P (Those are eggs), I did not make this utterance a target in this task. My feeling is it typically falls on the N P alone (Those are eggs), in which case we would expect either a nominal predicate construction or a cleft to be possible (to the extent that the English and Salish cases are parallel in their marking of contrast - an interesting area for future exploration here). Another answer may be due to the expression of definiteness/indefiniteness: though Salish determiners do not encode definiteness like in English (eg. Matthewson 1996), when something like definiteness must be expressed, a cleft can be used, while an NPC typically expresses indefiniteness (Davis p.c). In (59B), 'eggs' have not been introduced in the prior discourse (eg. the question in (59A)), so are expressed using an NPC. Rather than thinking of this in terms of (in)definiteness, more recent work by Matthewson (p.c.) accounts for these differences in terms of familiarity: determiners encode familiarity, while NPCs do not. Thus, the notion of contrast may be reduced to concepts like exhaustivity, familiarity, (in)definiteness and discourse newness/oldness, or at least interact with the marking of these notions (eg. Horvath 2007).  (60)  Context: Discussing what she did with all the papers that she cleaned up at her home. ?e  s-?ûs-t-0-ne  t^m  xe?  tékm  us.  and  NOM-discard-TR-3o-l SG.TS  PERF  DEM  all  3.CONJ  sqyéytn.  "And I threw them [all] out." (61)  A : ké? is.it.case  k  e?-s-w?x-ùm  tk  IRL  2SG.POSS-NOM-PROG-MDL  OBL.IRL salmon  "Do you have any salmon?" B: té? NEG  PERS  te?  k  n-s-w?x-ùm  tk  sqyéytn.  DEM  IRL  I S G . P O S S - N O M - P R O G - M I D D L E O B L . I R L Salmon  " I don't have any salmon." cik  t^va  use.up PERF  k  sqyéytn.  IRL  salmon  "The salmon is all gone." (62)  Context: Speaker A has said that she still babysits some of her grandchildren. Speaker B's reply is below. a. B : cuk"^ stop  ^am  xe?  9  n-s-?ém=it  ncé?.  PERF  DEM  DET  lSG.POSS-NOM-GUARD=agent ISG.EMPH  " I quit babysitting myself." b. B : tékm all  t.sm  xe?  sk^ul  e  scmémi?t  PERF  DEM  school  DET  children  " A l l the kids are at school now." (63)  Context: Discussing B's houseguests. A: M m . Mm.  ?éx  n  met  PROG  Q  indeed PERS  "Mm. Are they still here?" B: Nwén already  t^m  pent  x'^uy.  PERF  return go  "They already went back."  t\xl  n?éye. here.  Finally, the emphatic particle and second position clitic iu?, which can mean 'just' or 'only,' often surfaces in contrastive focus contexts (8B, 20B, 22B, 24, 34B, 61B, 63A). This particle seems to introduce the semantics of exhaustivity which is provided for free by the structure of English clefts.^'* Like emphatic m and x?e, ^uPalso seems to overtly specify a contrastive context, though the exact semantics of this particle needs further research. A further example is shown in the discourse below; A and B are discussing starfruit, and the fact that they haven't tried it. In B's reply, nuclear stress remains on the verb (striking in comparison to the English translation). (64)  Context: discussing starfruit A : Tete? NEG  témn  k-ex  s-msténe.  on.the.contrary  IRL-PROG  N0M-try.TR.30.1 SG.TS  "I haven't tried it." B: Té? NEG  t\xl  té?  k  s-msténe.  PERS  DEM  IRL  N0M-try.TR.30. ISG.TS  "I haven't tried it either." It is worth noting that the focus particles m, x?e, mef, ism and ^uPare all second position clitics, meaning that they are prosodically positioned after the first word in the clause. Thus, like the focus, these focus particles seek out the left edge of the clause. I'll return to this point in chapter 8 (section 8.4.3). Thus, it seems that simple bare nominal predicates are not suited to contrastive contexts, except when used in conjunction with other morphemes or lexical items that explicitly encode the semantics of contrast elsewhere in the clause. Cleft predicates, on the other hand, are preferred in contrastive contexts, though additional morphology or lexical material is typically employed as well. In more complex sentences, the split between the two forms (clefts and NPCs) in terms of their contrastive use breaks down. This is because, as noted previously, quantifiers are expressed in NPCs rather than clefts; and focused DPs (eg. proper names) are always expressed using clefts, since they cannot be predicates, no matter whether the focus is presentational or contrastive.  Interestingly, a phonological cognate of this particle seems to serve as the cleft predicate ^ in Musqueam and Chilliwack Halkomelem, as reported in Kroeber (1999: 370; citing Suttles 1984, and Galloway 1977).  In this section I reported the results from a series of tasks in which consultants responded to questions posed by computer characters. The goal was to determine whether the two focus structures differed in their use in new information versus contrastive focus contexts. While both were equally used in new information focus contexts, clefts were preferred in contrastive contexts. We also observed the use of other emphatic clitics in the contrastive context: emphatic m, demonstrative x?é, aspectual mei, perfective ^dih, and /iu?. The use of these additional particles presumably affects the truth conditions of the utterance, and so represents instances of Krifka's denotation focus. The structural form of the focus, as a cleft or NPC, does not change however.  3.6  Metalinguistic focus: A residual "stress-focus" effect There is one context where the focus does attract the greatest prosodie prominence of  the utterance, even in Nie?kepmxcin: metalinguistic focus. This has also been informally observed for St'ât'imcets Salish by Davis (p.c.) and Caldecott (p.c), who suggest that the "stress-focus" effect in metalinguistic cases may be a universal feature of human language. Metalinguistic focus is necessarily of the corrective or contrastive variety, since it involves a speaker correcting a portion of her own or someone else's utterance. These types of utterances have also been termed "metalinguistic negation" (Ducrot 1972,1973, Horn 1985, 1989). In English, this type of focus can attract the nuclear stress of the utterance to a word (65a), or just a part of a word (65b). (65)  Metalinguistic focus attracts nuclear stress: English a. Around here, we don't Uke coffee, we love it.  (Horn 1989: 382)  b. I didn't manage to trap two mon-geese — I managed to trap two mon-gooses. (Horn 1989: 371) The same is true in Nte?kepmxcin. In the first example (66a), the speaker is correcting the transitive ending she utters on the verb root q'^in- 'talk.' She begins with -em, but changes it to -es. The corrected suffix has the highest amplitude and FO peak in the utterance, having attracted the nuclear stress. In (66b), speaker B objects to A ' s phrasing of possession (see the literal translation); B uses instead the verb w?xum 'to have,' which is metalinguistically focused and attracts the nuclear stress. Finally, in example (c), speaker B corrects A ' s use of smuiec, insisting on the reduplicated form smufmuiec instead to mark the plural; once again, this metalinguistic focus attracts the nuclear stress, on the augmentative reduplicant here, which is phonetically marked by a dramatic 5 semitone pitch rise and the absolute amplitude peak of the clause.  (66)  Metalinguistic focus attracts nuclear stress: NteVkepmxcin a. ?e  swét  and who  met  tk  sxayxiVwi,  q^in-t-ém~  indeed  OBL.IRL husband[AUGl  q^in-t-0-és  e  sm?é?m-s.  talk.tO-TRANS-30-3TS  DET  wife-3.P0SS  talk.to-TRANS-em-  "Which of the husbands talk-- talked to their wives?" b. A : x*?it  e  many DET  sqaqxa-s. ... dog-3.POSS  "They have lots of dogs." (more literally: 'Their dogs are many.") B: x ^ i t e s-w?xùm-s te sqaqxa. many DET NOM-have-3.POSs O B L dog "They [havelpoc lots of dogs." c. A : [ke?ke?tés three[AUG]  ne?  tk  smûtec]Foc  there  OBL.IRL woman  0  ?ex  ketnim.  DET  PROG  rodfish.MDL  "The ones that are fishing there are [three women]poc " B: ke?ke?tés three[AUG]  xe?  ne?  tk  smutmutec  0  ?ex É;etnirn.  DEM  there  OBL.IRL woman[AUG] DET PROG rodfish.MDL  "The ones that are fishing there are three wo[men[poc." The waveforms for these examples illustrate the type of dramatic deaccenting of nonfocused information that is typical of languages like English, but not generally found in Nte?kepmxcin (as we shall see in chapters 4 and 5; see also the figures in chapter 2). Figure 3.1 shows (66b), and figure 3.2 the waveform for figure (66c). While the metalinguistic focus carries a large peak, the non-focused information is greatly reduced in prominence, a pattern we do not typically find in Thompson Salish focus contexts. In figure 3.2, it is in particular the post-focal material 7ex ketnim 'that are fishing' that is dramatically deaccented, again a pattern that is common in English (and many other languages), but not in Nte?kepmxcin.  0.4639-  -0.4178-  Time (s)  Figure 3.1  2.28366  Metalinguistic focus: "They [havelpoc lots of dogs."  0.2592-  -0.4701-  Time (s)  Figure 3.2  3.53255  Metalinguistic focus: "The ones that are fishing there are three wo|men]poc."  I will not discuss metalinguistic focus prosody any further in this dissertation.  3.7  Summary The present chapter has shown that clefts or nominal predicate constructions are used  to mark narrow focus, while default word order is used to mark CP, V P or verb focus. In addition, CP focus is sometimes marked using clefts. The results confirm Kroeber's (1997, 1999) "cleft-focus" generalization. In all cases of focus marking, the left edge of the clause is the relevant position, and not the rightmost nuclear stress position identified in chapter 2. The conclusion is that "stress-focus" cannot be the primary motivation for cleft or NPC focus structures. Predictions of "stress-focus" accounts were not realized in Nte?kepmxcin: narrow focus was not marked  by shifting the loeation of the nuelear stress, nor by moving the focus to the nuclear stress position, nor by scrambling unfocused information away from the nuclear stress site. Focus projection also operates somewhat differently in Nfe?kepmxcin than in English. We saw that in wide focus cases, focus projects from the predicate phrase rather than from the nuclear pitch accent, while narrow focus is restricted to any subconstituent of the predicate phrase. In a "stress-focus" language like English or German, on the other hand, wide focus projects from any subconstituent bearing the nuclear pitch accent. Narrow focus can in turn be restricted to any constituent bearing the nuclear pitch accent (NPA). Table 3.4  Nfe?kepmxcin  Nte?kepmxcin versus "stress-focus" systems Focus type  Wide focus projects from:  Narrow focus restricted to:  Structural  Predicate Phrase  Predicate or subconstituent  "cleft-focus" English etc.  of the Predicate Phrase  Intonational  Any subconstituent of the  Any constituent bearing the  "stress-focus"  wide focus bearing the NPA  NPA  I concluded by observing some differences between clefts and nominal predicate constructions. Clefts are preferred in contrastive contexts, though the use of additional morphology plays the most important role in indicating a contrastive interpretation. For now, it is time to turn to the phonetics of some of the focus cases observed in the present corpus study. I made the observation that nuclear stress is rightmost in Thompson River Salish, and in chapter 4 I will offer detailed support for this claim. I have also observed in this chapter that, in narrow focus clefts or NPCs, nuclear stress continues to surface rightmost (usually on given material in the residue clause). This observation will be phonetically tested in chapter 5, against the hypothesis that given material is deaccented ("destress-given"). It is also possible that there is a residual "stress-focus" effect in clefts or NPCs. Under this hypothesis, focused material in left edge clefts or NPCs receives greater prosodie prominence than the left edge in the default V S O word order employed in neutral (wide CP focus) sentences; this greater prosodie prominence would surface despite the left edge focus position not being associated with the nuclear stress position on the right. These hypotheses are tested phonetically in chapter 5.  Chapter IV: A Case Study of Nuclear Stress in Nie?kepmxcin In chapters 2 and 3,1 made the informal observation that nuclear stress is rightmost in NteVkepmxcin, as it is in English (Chomsky and Halle 1968, Cinque 1993, Selkirk 1995). In this chapter, I formally test this observation. I begin by presenting more informal evidence from Thompson and Thompson (1992) and Egesdal (1984), before I provide acoustic phonetic evidence from neutral wide CP focus utterances. Because these utterances lack narrow foci or given information, they constitute the "neutral" focus case and are expected to carry the default stress pattern (Gussenhoven 1984: 17-18, 65-68; Hayes and Lahiri 1991: 56, Selkirk 1995). Finally, I give phonological evidence that nuclear stress is rightmost (following Wasow 2002, Anttila 2007).  4.1  Nuclear stress Nuclear stress is the most prominent prosodie peak in an utterance. Phonetically,  prominence is defined as a combination of increased FO, amplitude and duration (eg. Fry 1958), which is typically measurable on the vowel bearing the nuclear stress. Phonologically, I will define the nuclear stress as the locus of the nuclear pitch accent in an intonational phrase, adopting the well-known model of phonological constituents in the prosodie hierarchy (Nespor and Vogel 1986, Hayes 1989). Using this model will allow more straightforward comparison of the phrasal structure of NteTkepmxcin to other languages, since it has been commonly employed by researchers across a wide variety of languages. Other models may differ in terminology (BUring 2003, 2006, Selkirk and Kratzer 2007); see Shattuck-Hufnagel and Turk (1996: 206, Figure 2), for comparison with other theories. For example, Selkirk (eg. 1995b) distinguishes major phrases and minor phrases, where Nespor and Vogel have only p-phrases. For the purposes of this dissertation, nothing hinges on the existence of minor phrases, so I adopt the simpler and more widely used model of Nespor and Vogel.  (1)  The prosodie hierarchy (Nespor and Vogel 1986, Hayes 1989) Utterance (U) I  Intonational Phrase (i-phrase) I  Phonological Phrase (p-phrase) I  Clitic Group (cl-gp) I  Prosodie Word (PWd) I  Foot (Ft) I  Syllable (a) In English, the nuclear stress is assigned rightmost, on the most deeply embedded syntactic constituent (Chomsky and Halle 1968, Cinque 1993). Pitch accents are assigned at the level of the p-phrase (the head of each p-phrase bears a pitch accent), and the nuclear pitch accent is carried by the head of the intonational phrase. The example below shows default parsing for the English wide-focus sentence John saw Monique. There are three prosodie words (PWd); PWd stress is shown by acute accent (').Verb and object are typically parsed into a single p-phrase (Chomsky 1971, Jackendoff 1972, Gussenhoven 1983, Selkirk 1995, Kahnemuyipour 20(M, Selkirk and Kratzer 2007), while the subject John forms a second p-phrase. This gives us two pitch accents at the pphrase level. Finally, there is a single right-headed i-phrase carrying the nuclear pitch accent, on Monique. (2)  (  X  )  i-phrase  (nuclear pitch accent)  (X )(  X  )  p-phrase  (pitch accent)  [John saw Monique [poç. Thus, in English, nuclear stress is rightmost. Another way to describe this state of affairs is to say that intonational phrases are right-headed. That is, when more than one pphrase is parsed into one intonational phrase, the rightmost p-phrase pitch accent will also carry the nuclear pitch accent at the i-phrase level.  In the remainder of this chapter, I show that nuclear stress in Thompson Salish is also rightmost.  4.2  Informal observations about nuclear stress in MeVkepmxcin At first glance, NteVkepmxcin is a stress system similar to English. In their grammar  of the language, Thompson and Thompson remark that word-level stress "seems to manifest itself as a complex of loudness, force, and pitch differences, rather similar in type to the phonetic reality of English stress" (1992: 21). This is shown in the Praat picture (Boersma and Weenink 2007), where the initial [a] of the word s?â?a? 'crow' is louder, longer, and higher in pitch than the unstressed second |a|. The pitch tracing is superimposed on the spectrogram.  0  0.581134 Time (s)  Figure 4.1  Stress on sPaPa? 'crow' (spectrogram shows 0-5000 Hz)  Similarly, Watt et al. (2000) examined stressed Inl and /a/ vowels in a carrier phrase in Squamish Salish, and found that the acoustic correlates of stress are increased length, higher pitch and greater amplitude (see also Bar-el and Watt 1998). Because they analyzed words from a carrier phrase, their findings probably better reflect the correlates of phrasal accent (possibly the nuclear stress) than strictly word-level stress (Sluijter and van Heuven 1997, 1998, for discussion). Watt et al.'s findings thus also suggest that Squamish speakers use pitch accent in a way similar to English.  Returning to Thompson, Egesdal ( 1 9 8 4 : 1 0 9 ) also identifies NteVkepmxcin as a "stress-timed language." In his study of the stylized use of speeeh in Nte?kepmxcin narratives, Egesdal found that classic correlates of stress (duration, pitch and amplitude) were among the suprasegmental features manipulated by storytellers as "rhetorical or performative devices" ( 1 9 8 4 : 6 ) . Indeed, "rythmic or assonant stylized speech" is employed to "convey salient information toward advancing the narrative, occurring at crucial or exciting points" (Egesdal 1984: 102). It is perhaps all the more surprising, then, that stress is not employed to mark focus, but this is precisely what I will claim (see also Davis 2 0 0 5 : ft. 18, on neighbouring St'at'imcets Salish). Sentence-level accent, like English, appears to be rightmost. I introduced this idea in chapter 2 , based on informal observation. Thompson and Thompson ( 1 9 9 2 : 1 4 8 ) also describe the final position in the Nte?kepmxcin clause as having "emphatic force," or being "mildly emphatic." In their desciption of clausal intonation patterns, they always identify the last stressed syllable as standing out ("mid-high with last primary stress"), followed by a drop to the end of the phrase ( 1 9 9 2 : 2 4 ; see ( 5 1 ) in section 1.7). I explicitly formalize these observations as a case of rightmost nuclear stress. Given the surface verb-object order, this is not surprising: we expect main stress to fall on the object, since cross-linguistically phrasal accent falls on arguments rather than heads (Schwarzschild 1999: 127 " H E A D A R G " constraint, Buring 2 0 0 3 " * S T R E S S - P R E D I C A T E " constraint, Kahnemuyipour 2 0 0 4 , Selkirk and Kratzer 2 0 0 7 ) . The following data is from Egesdal ( 1 9 8 4 : 14, ex. 2. le), and presents evidence from a traditional narrative that corroborates the idea of rightmost nuclear stress. The character Robin is foretelling "the coming of fish at different points in a legend." Egesdal makes the following observations about these lines of narrative: "Stressed I'll vowels are lengthened in each phrase except the first. These same vowels also show a sharp rise in pitch." This observation is indicated by underlining the final vowel in each phrase below; in each case, this is the proposed nuclear stress position. Egesdal gives four lines for each line of discourse: underlying form - gloss - translation - surface form. I have added underlining to indicate the nuclear stress in both NteVkepmxcin and English. Word-level stress and phrasing is as marked by Egesdal. (3)  yu?=cin-m  cat  toast=fish-MDL  spurt  "Toast fish—spurt." (as in fish dipping in hot oil) (yu?icinrn ci+]  yu?=cin-m toast=fish-MDL "Toast fish." [yuVicinm]  tfék  e  pisui  arrive DET  fish  "The fish are here." \mk  e pi:sut]  Z9m=cin-m toast=fish-MDL "Toast fish." [z9rh=ci:nni] sip-e-t-e bend-DRV-TR-iMP "Bend it!" [fishing rod over the fire to malce it strong] [si:pet8] k"én-t-e  t  grasp-TR-iMP DET  e?-ce?qintn 2sG.POSS-angling.rod  "Grab your angling rod!" [k'^én-t-E i 8?-cE?qé:ntnj  (adapted from Egesdal 1984: 14, ex. 2. le)  The first / i / case in line 1 (the only / i / lacking longer duration and higher pitch) is the only / i / which is not the rightmost stressed vowel in its phrase; hence the / i / in line 1 is predicted to lack these prominent acoustic characteristics if nuclear stress is rightmost. Interestingly, in later lines, we see that nuclear stress ends up rightmost despite being hosted by old information; in many languages, we have seen that old or given information is deaccented ("destress-given"). Line 3 has a rightmost subject DP e pisuf 'the fish' which carries the nuclear stress, despite being old information. In the previous two lines of discourse, Robin has announced that it is time to toast fish, using a lexical suffix for 'fish'; nevertheless, the DP 'fish' in line 3 gets nuclear stress, even though we might consider only the verb 'arrive' to be new information here.  Line 6 has a similar phenomenon. Here, we find a rightmost objeet DP 'angling rod' that is given in the previous line of discourse (though not overtly) and a narrowly focused verb 'grab.' 'Grab' is narrowly focused since is contrasts with 'bend' of the previous line. The narrow contrastive focus is expected to attract greater prosodie prominence (Féry and Samek-Lodovici 2006), while old information is expected to be deaccented ("destressgiven"). Once again, however, the rightmost given object DP gets nuclear stress, and not the contrastively focused verb. The apparent violations of "stress-focus" and "destress-given" are a common theme in this dissertation, and will be empirically tested in the next chapter. It is also worthwhile noting that the final lengthening observed by Egesdal occurs on the final stressed syllable in each of lines 2-6, even when it is not the absolutely final syllable. In each case of lengthening at least one or two unstressed syllables follow the lengthened vowel. This suggests that final lengthening is also a marker of nuclear stress. Egesdal's observations about the phonetic markers of prominence leads us to the next section, where I investigate the acoustic phonetic properties of neutral utterances to determine if there is support for the perceptual observation of rightmost nuclear stress.  4.3  Phonetic observations about nuclear stress  4.3.1 Background When looking for phonetic correlates of nuclear stress, we can consider pitch (fundamental frequency, or FO, in Hertz or semitones), loudness (amplitude, or energy, in db) and duration (ms). However, it is not enough to simply look for the point in the utterance with the greatest absolute values of these measures. The task is complicated by declination across the breath group: FO and amplitude (db) show a gradual downward trend from left to right, as air is expelled from the lungs and subglottal pressure decreases (Cohen and't Hart 1965, Pierrehumbert 1979, 't Hart et al. 1990, Strik and Boves 1995, Trouvain et al. 1998, and many others; Vatikiotis-Bateson and Fowler 1988 on articulatory declination). This general trend can be observed for N+e?kepmxcin in the pitch tracings and waveforms shown in the figures throughout chapter 2. Declination from left to right means that a rightmost peak, such as the nuclear stress accent in English, will still be perceived as more prominent even though its absolute pitch and amplitude is lower than in preceding stressed vowels (eg. Pierrehumbert 1979: 363). A n English example is shown in the pitch tracing and waveform below. The figure depicts the neutral utterance My mother drinks coffee by a female speaker of English. Though nuclear stress is on the rightmost object, coffee, its absolute amplitude and FO are lower than  on the stressed vowel in mother. While mother has a pitch peak of 187.5 Hz and amplitude peak of 76.6 db, coffee has a pitch peak of 162.7 Hz (a difference of 2.46 semitones) and amplitude peak of 73.3 db (3.3 db less).  300  N X B a.  ISO-  Time (s)  Figure 4.2  L I 8531  Pitch tracing and waveform for the sentence My mother drinks coffee.  Studies of English can give us some idea of the type of declination effects to expect in NteVkepmxcin, if, as predicted, nuclear stress is rightmost there too. Sorenson and Cooper (1979) found declination of 6.7 Hz/syllable in utterances of 8 syllables (as reported in Pierrehumbert 1979: 368). This length closely matches the average number of syllables in the present set of neutral CP focus Nte?kepmxcin utterances to be examined (9.2 syllables for FE, 8.6 syllables for PM). If nuclear stress is not rightmost in Nfe?kepmxcin, we would expect greater declination than this. Pierrehumbert (1979) tested for the perception of peak FO declination across utterances with 2 or 3 stressed syllables. Utterances consisted of nonsense syllables with artificially manipulated pitch and amplitude contours. The rightmost syllable was perceived as most prominent even when pitch declined from left to right peaks; the decline was typically between 6.8 and 11.1 Hertz (0.73 to 1.16 semitones). This is less declination than reported by Sorenson and Cooper, but perhaps a function of Pierrehumbert's study being  perceptual and using artificial stimuli. The present study, on the other hand, examines production of real speech in NteTkepmxcin. Regarding amplitude declination, Pierrehumbert (1979) notes that declines of 4 db are common within an intonation group. Trouvain et al. (1998) examined average amplitude declination for two speakers of English, in utterances containing between 1 and 6 stressed vowels. By examining their figures, we see an approximate amplitude declination between left and right peaks of between 1 to 6 db. 4.3.2 Subjects The language data was collected from two female speakers of N+e?kepmxcin in their late 60's (FE and PM), as also reported in chapter 3. Both are speakers of the Lytton dialect, and fluently bilingual in English. 4.3.3 Method Instances of neutral, wide CP focus were identified from the corpus of conversational recordings that served as a source for the data reported in Chapter 3. Recordings were made at the residence of either the language consultants or of the researcher, using a Marantz P M D 670, 671 or 660 digital audio recorder. Recordings were made at 44.1 kHz. Each consultant was recorded on a separate channel using a Countrymax Isomax E M W Lavalier lapel microphone. The microphone was attached onto the exterior of the consultants' clothing, approximately at the sternum. To account for declination effects, only utterances which were completed in a single breath group were considered. For the present case study, neutral cases (wide C P focus), where everything in the utterance was new information, were used to provide a default intonation marking. These CP focus cases give us a baseline intonation pattern where we expect to find the default nuclear stress pattern. This criteria yielded 20 utterances for the first speaker, FE, and 19 for the second speaker, P M . For the present acoustic phonetic analysis, I compared the declination between the leftmost stressed syllable and the righmost stressed syllable (see Sorenson and Cooper 1979, Pierrehumbert 1979, 't Hart et al 1990, Strik and Boves 1995, on this "topline" declination measure). Stressed vowels were segmented in Praat (Boersma and Weenink 2007). Utterance length was also identified. For both individual vowels and entire utterances, a variety of acoustic measurements were then made by using automated scripts in Praat. Pitch measurements of primary interest were the maximum and minimum FO, the standard deviation of FO, and the timing of the FO  maximum and minimum (expressed as a percentage of the vowel duration). Where the Praat algorithm mismeasured FO (eg. in the presence of glottalization, etc.), measurements were done by hand via visual inspection of the waveform, and automated measurements were disregarded. The average and maximum intensity (in decibels) was also recorded, as was vowel and utterance duration. The number of syllables per utterance was counted, to provide a measure of speech rate. 4.3.4 Results and discussion I report means (M) and standard deviations (SD) individually for each speaker, due to their differing pitch ranges, though both showed the same fundamental patterns. Basic characteristics indicated that both sets of utterances were comparable. F E had an average utterance length of 9.20 syllables (n=20, SD=3.58), or 2.39 seconds (n=20, SD=0.75). Her speech rate averaged 3.85 syllables/second (SD=0.77). For P M , utterances averaged 8.63 syllables in length (n=19, SD=3.51), or 2.15 seconds (n=19, SD=0.90). P M ' s speech rate was, on average, 4.12 syllables/second (SD=1.04). For FO declination, F E had a mean decline from left to right peaks of 2.51 semitones (n=20, SD=1.93 semitones), or from approximately 211 Hz to 183 Hz. P M showed a similar FO decline (2.34 semitones, n=19, SD=1.86 semitones), or from an average of 168 Hz to 147 Hz. Rate of decline between peaks was considerably less than that the 6.7 Hz/syllable reported in Sorenson and Cooper (1979, as presented in Pierrehumbert 1979): F E averaged 3.84 Hz/syllable, and P M 3.15 Hz/syllable. If nuclear stress in N+e?kepmxcin were not rightmost, as in English, we would expect larger topline FO declines. Amplitude declination was also not greater than that previously reported for English. Amplitude peaks, for FE, actually increased slightly from left to right stressed vowels, shown by a negative declination of -0.24 db (n=20, SD=4.18 db). Average amplitude declined 0.78 db (SD=4.25 db). For P M , amplitude peaks did have a positive decline (M=1.25 db, n=19, SD=5.86 db), as did average amplitude declination (M=2.38 db, SD=5.60 db), but well within the 1-6 db range reported by Trouvain et al. (1998) for English. The results suggest that rightmost nuclear stress is marked by relative suspension of amplitude declination. There was a final lengthening effect, as observed by Egesdal (3). Rightmost stressed vowels were on average 81.0 ms longer for F E (n=20, SD=76.6 ms) than left edge stressed vowels. For P M the difference was 58.7 ms (n=19, SD=54.6 ms). Lengthening occured when the stressed vowel was on both the final syllable or a non-final one (as also noted by Egesdal - see (3) and discussion), though final stressed syllables incurred greater lengthening (M=123.3 ms, n=10, SD=83.4 ms for FE; M=65.6 ms, n=13, SD=62.0 ms for PM).  Lengthening was less when the rightmost stress fell on a non-final syllable (M=38.5 ms, n=10, SD=38.0 ms for FE; M=43.7 ms, n=6, SD=33.3 for PM). Finally, there is one more suggestive result. When we consider stressed vowels in between the right and left edges, we can document declination across more than two peaks. In particular, we can calculate the expected FO value on the rightmost peak based on the declination the first two peaks. For both speakers, the final vowel carried an FO peak above the declination line expected from extrapolation of the first two peaks ("left" and "middle" in figures 4.3 and 4.4). If nuclear stress is rightmost, as predicted, this is exactly the pattern that we anticipate. Whiskers indicate error bars of one standard deviation in either direction.  Figure 4.3  FO declination for FE across multiple peaks: final peak has a mean FO above the anticipated declination line  actual declination  expected declination  Left  Middle  Right  Vowel position Figure 4.4  FO declination for P M across multiple peaks: final peak has a mean FO above the anticipated declination line  As a whole, these results support the informal observations that Thompson Salish is a language with rightmost nuclear stress. The nuclear stress is marked by comparable or lesser FO declination than English, another language with rightmost nuclear stress; by relative suspension of amplitude declination; and by final lengthening of the syllable bearing the nuclear pitch accent. These results are illustrated in the pitch tracings and waveforms in figures 4.5 and 4.6, where the nuclear stress is indicated inside the dashed line.^^ Both utterances show a general downward trend in FO and amplitude from left to right.  The steep pitch climbs indicated at the end of each DP in figure 4.5 are a local phonetic effect due to the glottal stop ending these words; that is, these rises are not perceived as part of the general downward declination trend. In addition, the pitch tracing may be exaggerated due to Praat misreading the pitch cue (eg. Gussenhoven 2004: 5-6 on algorithm errors in computing pitch).  3oa  150-  kantés help  xe? DEM  e DET 1  Figure 4.5  skixze?-kt mother-our  e smçi?-kt DET brother-our 2.91247  Time (s)  Pitch tracing and waveform: "|Our mother helped our brotherlpoc"  300  150-  •t n-sxaywi DET my-husband Time (s)  Figure 4 . 6  e swux'^t DET snow 2.72333  Pitch tracing and waveform: "(My husband cleaned up the snow Ipoç.  The findings are summarized in the table of means below; standard deviations are reported in brackets. Table 4.1  Declination measures by speaker in default wide focus utterances FE  M H A S U R E  G  D  4.4  E  E  N  C  E  L  R  A  L  M  E  A  S  U  R  E  PM  S  Number of utterances examined (n)  20  19  Utterance length (syllables)  9.20 (3.58)  8.63 (3.51)  Utterance duration in (ms)  2.39 (0.75)  2.15 (0.90)  Speech rate (syllables/second)  3.85 (0.77)  4.12(1.04)  FO declination (semitones)  2.51 (1.93)  2.34(1.86)  FO declination rate (Hz/syllable between peaks)  3.84  3.15  Peak amplitude declination (db)  -0.24(4.18)  1.25 (5.86)  Average amplitude declination (db)  0.78 (4.25)  2.38 (5.60)  Lengthening of rightmost stressed syllable total (ms)  81.0 (76.6)  58.7 (54.6)  Lengthening (sentence-final syllable only)  123.3 (83.4)  65.7 (62.0)  Lengthening (non-sentence-final syllable only)  38.6 (38.0)  43.7 (33.3)  I  N  A  T  I  O  N  M  E  A  S  U  R  E  S  F  R  O  M  L  E  F  T  T  O  R  I  G  H  T  P  E  A  K  Phonological observations: the direction of heavy NP shift (HNPS) A third piece of evidence for the location of nuclear stress in Nte?kepmxcin is  phonological: heavy NP shift is rightward in languages where nuclear stress is also rightmost (Wasow 2002, Anttila 2007). Heavy N P shift is often grouped with extraposition from heavy NPs, another process which shifts heavy prosodie material to the rightmost nuclear stress position in English (eg. Rochemont and Culicover 1990). In this section, I'll review recent arguments to this effect by Anttila (2007) for English, another language with rightmost nuclear stress. In addition, I'll review evidence (also mentioned in Anttila) that Japanese shows evidence for HNPS to the left, the same direction as nuclear stress (McCawley 1977, Cinque 1993, Yamashita and Chang 2001).  Then I'll show that Thompson Salish has heavy NP shift to the right, in line with Davis's (2005) findings for neighbouring Lillooet Salish. Anttila (2007) explores the role of prosody in driving the well-known dative alternation. In English ditransitive verb constructions, full argument NPs may occur in either order; pronouns, however, are not generally produced in final position. This is accounted for because pronouns, as clitics, do not generally carry phrasal stress and thus are poor candidates for occupying the rightmost nuclear stress position (shown by underlining). (4)  a. Pat gave ffoodj |to ChrisJ ~ Pat gave [Chris] [foodJ. b. Pat gave [it] [to Chris] ~ *Pat gave [Chris| [h].  (Anttila 2007: ex. 4)  As the size of two full NPs increases, the same principle holds. Wasow (2002) and Anttila (2007) show that, when there are two full NPs, only the relative weight of the two NPs determines which surfaces at the right. In the ditransitive construction in (5), the theme occupies the rightmost nuclear stress position when heavier than the goal (a). When the goal is heavier than the theme (b), the goal is rightmost. Finally, if both goal and theme are equally heavy, either order is possible (e). The data in (5) comes from a corpus study by Antilla. (5)  a. Goal < Theme: I gave (my sister) (twenty dollars) b. Goal > Theme: I gave (the money) (to my little sister) c. Goal = Theme: I gave (my sister) (the money) ~ I gave (the money) (to my sister) (Anttila 2007: ex. 19) The coincidence of the larger prosodie unit and the nuclear stress position maximizes  the number of word-level stresses that are dominated by the nuclear stress; Wasow (2002) calls this the Principle of End Weight, and Anttila (2007) the Stress-to-Stress Principle. There are several important predictions made by these principles (Anttila 2007). First, if the nuclear stress is "lured away," for example to a sentence-final adverb, the weight effect should be mitigated or eliminated. Anttila (2007: ex. 20) reports several examples from a corpus study that seem to confirm this prediction: where an adverb follows the arguments of a ditransitive verb and bears the nuclear stress, pronouns may follow full NPs. (6)  ... showing [people] [him] through our life,  (from Bresnan and Nikitina 2003: 19-20)  For the present study, the prediction is as follows: if left edge predicates in focus structures (which narrowly focus subject, objects, and so on) attract the nuclear stress, there should not be any rightward HNPS effects in narrow focus constructions in Thompson Salish. However, we have already seen that extraposition from complex nominal predicate constructions is rightward (section 3.3.8; see also section 8.4.4). Secondly, Anttila (2007) predicts that function words, which do not bear stress, should not count for HNPS. This reduces nuclear stress to a prosodie phenomenon, rather than a syntactic one, where, for example, individual nodes (including function word nodes) would count for heaviness. In chapter 2,1 claimed that nuclear stress in Thompson Salish was also a prosodie phenomenon, since the rightmost lexical constituent bears the nuclear pitch accent irrespective of its syntactic status (subject, object, adjunct). Finally, and importantly for the present issue, Anttila (2007) notes that his study predicts that nuclear stress and HNPS coincide. That is, languages where nuclear stress is leftmost should exhibit heavy NP shift to the left. This is counter to processing theories of HNPS, which argue that heavy NP shift is a result of more complex and newer information being processed later in the sentence (eg. Arnold et al. 2000 for detailed discussion). Anttila proposes that Japanese may be one such language with leftmost nuclear stress (following Cinque 1993, and McCawley 1977), and leftward HNPS. Indeed, Yamashita and Chang (2001) report that Japanese has a "long before short" preference, such that heavy NPs are uttered before shorter NPs (but see Ishihara 2001, Shiobara 2004, for an alternate view of the site of nuclear stress). Hungarian is another language with leftmost nuclear stress, on the verb (Vogel and Kenesei 1987, Szendroi 2003: 44, Neeleman and Szendroi 2004). As expected under Anttila's account, heavy NPs in Hungarian are preferably before shorter NPs when in the pre verbal nuclear stress position (Hawkins 1994: 131, Yamashita and Chang 2001: B54). Heavy N P shift is often grouped with extraposition from heavy NPs, another process which shifts heavy prosodie material to the rightmost nuclear stress position in English (eg. Rochemont and Culicover 1990). (7)  I wore l^pa ring th | last night [pp of shiny silver],,. Nte?kepmxcin exhibits both heavy NP shift and extraposition from NPs to the right,  but not to the left (Davis 2005 on St'at'imcets; ft. 16 on extraposition from NP). Davis (2005) noted that Upper St'at'imcets, which has a canonical VOS word order, also allows a V S O interpretation just in case the object is prosodically heavier than the subject.  (8)  a. x'^a'-an-as seek-TR-3TS  [na-zâxalV^m-a  sâma? sqayx"^] [k"  s-Gertie]  |ABS.DET-tall-EXis  white man]  NOM-Gertie|  [ D E T  (i) "Gertie was looking for the tall white man." not (ii) *'The tall white man was looking for Gertie." b. x^it-an-as  [k*  seek-TR-3TS  [  D  E  T  s-Gertie]  [na-zaxafq'^am-a  sâma? sqayx"^]  NOM-Gertie]  [ABS.DET-tall-EXis  white  man]  (i) "The tall white man was looking for Gertie." or (ii) "Gertie was looking for the tall white man."  (Davis 2005: ex. 55)  In section 2.1.1,1 showed that V S O is the canonical order in N+eîkepmxcin. The examples below all show sentences with VOS interpretation; the rightward-shifted subject NPs are prosodically heavier than the objects. (9)  o  V céw-0-0-es  xe?9  wash-TR-3o-3TS D  E  M  [DP  i  [  D  nk^anustn] E  T  window]  S 1+  n-sk^ùze?-s  [ D E T  t  ISG.POSS-offspring  D  smuiec] E  T  woman]  "My daughter was the one that washed the windows." (literally: "My female offspring washed the windows.") (10)  V  O  qVw-0-0-es  xe?9  pick-TR-3o-3TS  D  E  M  [ope D  E  S  sq^yt] [ o p i T  fruit  D  E  T  skixze?-s  i  mother-3sG.POSS  D E T Mary  Mary]  "Mary's mother picked the fruit." Extraposition from complex nominal predicates is also common. In the canonical order, subjects follow these complex N P predicates. However, extraposition from the complex NPs targets the rightmost nuclear stress position. Non-heavy predicates (eg. those consisting of a single adjective or noun, as in (c-d)) may not be extraposed to the right. Extraposition is also never leftward (e). In each example, the second position clitic xe? follows the first prosodie word.  (11)  INP t*mi?me?éyxkn  a.  small  xe? D  E  M  tk O  B  L  sqaqxa] .  I  R  dog  L  e D  Hermann.  E  Hermann  T  "Hermann is a small dog." b. [fjp k:'^mi?me?éyxkn small  xe?  td ]  D E M t^  e D  Hermann E  T  [tk  Hermann  O  B  sqaqxajn. L  .  I  R  L  dog  "Hermann is a small dog." c. *  td  e  Hermann  tj  D E T Hermann  xe?  [AP  D E M  fc^mi?me?évxkn]H small  intended: "Hermann is small." d. *  td  e  Hermann  xe?  td  D E T  Hermann  D E M  [NP  sqâqxa]H dog  intended: "Hermann is a dog." e. *[tk O  B  sqâqxa]d L  .  I  R  L  dog  xe? [^p ^'^mi?me?éyxkn D  E  M  small  td ] t^  e D  Hermann. E  T  Hermann  intended: "Hermann is a small dog." Thus, weight-driven movement in Nte?kepmxcin targets the right periphery, as we would expect if the nuclear stress is rightmost (Anttila 2(X)7).  4.5  Summary This chapter has provided a study of nuclear stress in Nte?kepmxcin. Informal  observations by myself, Thompson and Thompson (1992) and Egesdal (1984) suggested nuclear stress was rightmost. Acoustic phonetic analysis of neutral, all new information sentences confirmed these observations; this was achieved by comparing the declination patterns of Thompson Salish with English, which also has rightmost nuclear stress. Finally, phonological evidence (Anttila 2007) is also consistent with nuclear stress in N"le?kepmxcin being rightmost: heavy NP shift and extraposition from heavy NPs is to the right - to the nuclear stress position - and never to the left. In the next chapter, I compare the acoustic phonetics of narrow focus utterances with the neutral wide focus cases examined in this chapter. If information structure in N'le?kepmxcin is marked prosodically according to "stress-focus" and "destress-given"  accounts, we would expect the prosodie characteristics of narrow focus utterances to differ from those of the neutral wide focus cases examined in this chapter. We shall see, however, that this is not the case: information structure is not marked through local prosodie prominence.  Chapter V: Phonetics of Intonation In chapter 4,1 gave phonological and acoustic phonetic evidence that nuclear stress (the primary sentential accent of the utterance) is rightmost in N'te?lcepmxcin. On the other hand, in chapter 3, we saw that focused constituents consistently appear at the left edge of the Nte?kepmxcin clause, an unexpected result for "stress-focus" theories. Nevertheless, we may expect that narrow focus constituents are still marked by additional prosodie prominence on the focus, and lower prosodie prominence on given material ("destress-given"). I test this hypothesis in the present chapter. I compare the acoustic phonetics of intonation in neutral (wide focus) contexts, with narrow focus contexts in Me?kepmxcin. Section 5.1 examines previous research on acoustic markers of prominence in human language. There are two questions here: (i) what sorts of acoustic parameters are relevant for marking discourse status (focus and givenness), and (ii) what degree of difference marks prominence on one item rather than another? After establishing predictions for the acoustic marking of focus, section 5.2 presents a case study comparing intonation contours in neutral focus utterances (the default intonation) with narrow subject or object focus sentences in NteVkepmxcin. In the present focus-oriented framework, "neutral" focus corresponds to wide CP focus, where all information is new (Gussenhoven 1984: 17-18, 65-68; Hayes and Lahiri 1991: 56). I then examine the acoustics of narrow object and subject focus, to determine if the narrowly focused constituent is marked by additional prominence. In section 5.3,1 detail a second case study which compares the effect of data source (spontaneous conversation, scripted conversation, or single sentence elicitation) on intonation. Section 5.4 concludes.  5.1  Background: Acoustic markers of prominence Before examining the acoustic phonetics of intonation in NteVkepmxcin, it will be  useful to determine what sorts of physical characteristics of the acoustic signal are important for marking prominence. Classic markers of prominence in stress languages are pitch (fundamental frequency), duration, amplitude and vowel quality (Fry 1958, and many others). In this study, I will examine aspects of pitch, duration and amplitude, the markers of prominence identified by Lieberman (1967: 144). A second question is what sorts of differences indicate relative prominence. For example, if we mark informational prominence with pitch accent, how much additional pitch do we require on focused items so that they are perceived as focused? This perceptual  difference has been called a "just noticeable difference" or JND (eg. 't Hart 1981). We may also look at production differences in this regard. In reviewing the literature on these two questions, I hope to gather a useful set of parameters that can be used as a guideline when investigating acoustic prominence. While no one to my knowledge has developed a successful formula to determine which combination of pitch, amplitude and duration will be perceived as more prominent than another (but see Lieberman 1960 on word stress cues), a summary of previous research findings will provide a useful reference point for individual acoustic parameters at least. 5.1.1 What I will be comparing In chapter 2, we saw that default word order in Nie?kepmxcin is verb-initial. The corpus study in chapter 3 showed that this is the order typically used in wide CP focus utterances, where all information is new. Wide CP focus cases will therefore serve as a baseline, default intonation pattern with which to compare narrow focus utterances. If narrow focus is marked prosodically, we expect to see some perturbation in the intonation contour of narrow focus utterances when compared to the neutral wide focus cases. The corpus study in the previous chapter showed that narrow subject and object focus is marked through a left edge cleft or nominal predicate construction. Unfocused (or old, given) information is in the residue clause at the right edge. If a "stress-focus" generalization holds in Nte?kepmxcin, then we would expect the left edge of focus constructions to carry greater acoustic prominence than the same left edge stress in neutral contexts. If a "destressgiven" generalization holds, we also expect the right edge of clefted focus sentences to be deaccented relative to the right edge in the neutral broad focus context. The constraints below show one recent instantiation of these discourse prosodie principles. (1)  Stress-Focus: A focused phrase has the highest prosodie prominence in its focus domain. (Féry and Samek-Lodovici 2006: 135-6)  (2)  D E S T R E S S - G I V E N :  A  given phrase is prosodically non-prominent. (Féry and Samek-Lodovici 2006: 135-6)  Finally, the differences in the left and right edge should also be manifested in greater declination between the prosodie peaks at the left and right edge in narrow focus cleft sentences. On the focus, we are testing for the relative strength of phrasal pitch accents, while  on given material we are looking for the absence of phrasal pitch accents (see the discussion in section 1.5). As such, changes in FO are the primary cue (as opposed to word level stress, where amplitude and duration are thought to be of greater importance - Sluijter and van Heuven 1996a, 1996b, Shue et al. 2007). Numerous scholars distinguish the prosodie properties of pitch accents at the phrasal level (p-phrases and i-phrases) from word level stress and unstressed items (Halliday 1967b, Vanderslice and Ladefoged 1972, Beckman and Edwards 1994, Sluijter and van Heuven 1996a, 1996b, Astruc and Prieto 2006). I follow these accounts in defining stress as a property of the head of a prosodie word, while accent is a property of phrasal prominence. Accents are realized on word-level stresses, but add additional prosodie characteristics. (In fact, this view suggests that "stress-focus" accounts would in fact be more accurately called "accent-focus.") In the present study, we will not be concerned with unstressed items, but will distinguish two levels of accentuation (simple p-phrase accents versus nuclear pitch accents), and the possibility of stressed but unaccented items. I follow the prosodie hierarchy proposal of Nespor and Vogel (1986) and Hayes (1989), who propose the following prosodie constituents (see Shattuck-Hufnagel and Turk 1996: 206, Figure 2, for comparison with other theories). (3)  The prosodie hierarchy (Nespor and Vogel 1986, Hayes 1989) Utterance (U) I  Intonational Phrase (i-phrase) I  Phonological Phrase (p-phrase) I  Clitic Group (cl-gp) I  Prosodie Word (PWd) I  Foot (Ft) I  Syllable (a) In the default C P focus case in (4), there are three prosodie words (PWd), whose heads are marked by acute accent (')• Verb and object are typically parsed into a single p-  phrase (Chomsky 1971, Jackendoff 1972, Gussenhoven 1983, Selkirk 1995, Kahnemuyipour 2004, Selkirk and Kratzer 2007), while the subject John forms a second p-phrase. This gives us two pitch accents at the p-phrase level. Finally, there is a single right-headed i-phrase carrying the nuclear pitch accent, on Monique. Example (4) thus shows three levels of stress: a stressed but unaccented verb saw, the pitch accented subject John, and the direct object Monique carrying the nuclear pitch accent. The horizontal line separates phrasal accents from word-level stress. (4)  (  X  )  i-phrase  (nuclear pitch accent)  ( X )(  X  )  p-phrase  (pitch accent)  [John saw Moniquejp. When the object is focused in a cleft (5), it attracts the main sentential stress (the nuclear pitch accent). The old, or "given," information in the residue clause that John saw is deaccented. This is achieved by not parsing that John saw into a p-phrase at the interface of syntax and phonology (Selkirk and Kratzer 2007). This means that, in Step 1, only the focus Monique is parsed into a p-phrase. The given material that John saw is parsed recursively into a p-phrase by the phonological component in Step 2, and then into an i-phrase. Deaccenting after the focal accent is a strategy employed in diverse languages (Schwarzschild 1999; Benkirane 1998 on Moroccan Arabic, Botinis 1998 on Greek, Gr0nnum 1998 on Danish, Kratochvil 1998 on Beijing Chinese, 't Hart 1998 on Dutch, and many others). This means that John and saw have word level stress, but no phrasal pitch accents (see section 1.5: ex. 48-49, for more discussion about adopting this particular representation). Givenness is shown by a subscript ' G . ' (5)  ( (  X (  X  )  (  X  )  )  i-phrase: STEP 2  )  recursive p-ph: STEP 2 p-phrase at interface: STEP 1  It was [Moniquejpoc [that John sawjg. In the default intonation case (wide CP focus - (4)), the rightmost stress is most prominent. In the focused cleft case (5), the leftmost (lexical) stress is most prominent. So, we would expect, all else being equal, that clefting focused items will result in greater prosodie prominence at the left edge in comparison with the default stress marking. Acoustically, we want to know what cues distinguish the nuclear pitch accent from simple pphrase pitch accents.  Secondly, the given material in (5) does not receive phrasal stress due to deaccenting. Thus, we would expect given material at the right edge of a cleft to be marked with lesser prominence than the right edge in the default stress case (4). Acoustically, we would thus like to know what cues discriminate pitch accents (default intonation) from just word-level stress ('deaccented' material) (Sluijter and van Heuven 1996a, 1996b, 1997). Thirdly, there should be greater declination between the left and right edge stresses in the narrow focus cases as compared to the default intonation in (4), since in (5) we are dropping from a nuclear pitch accent at the left to unaccented material at the right (Eady and Cooper 1986). Measuring declination as the difference in peak heights (the left and right phrasal accents in (4), and the left phrasal accent and rightmost stress in (5)) will reflect differences in the accentual level on these stressed syllables rather than "true" declination, but accentual differences are precisely what we are interested in. This has been called a "topline" declination measure ('t Hart et al 1990; also Sorenson and Cooper 1979, Pierrehumbert 1979, Strik and Boves 1995). It is worth noting that all three measurements can be recuced to the result of a single discourse principle (Buring, p.c); that is, G I V E N  D E S T R E S S -  accounts for lack of accent on given material, greater accent on the focus (by default),  and a resulting greater declination between the two. However, because a constituent can be both focused and given, I continue to employ notions of both focus and givenness (see section 1.5 for more discussion). Now let us consider the parallel Thompson Salish examples. In chapter 2,1 showed that the neutral word order in Nie?kepmxcin is verb initial. Chapter 4 showed that nuclear stress is rightmost. In the acoustic phonetic analysis that follows, I will consider only wide C P focus utterances containing a single i-phrase (for example, those which were split into two breath groups were not analyzed). A n example is shown below. (6)  ( (X  )  (  X  )  i-phrase  (nuclear pitch accent)  X  )  p-phrase  (pitch accent)  |Wik-t-0-ne  xe?  e  MoniquelFoc-  see-TR-3o-3s  D E M  D E T  Monique  "(I saw MoniqueIPQÇ." In (6), there are two phrasal stresses, one on the verb, and one on the direct object Monique (I'll provide more evidence for Thompson Salish phrasing in chapter 6). This means that there are two pitch accents at the p-phrase level. This differs from languages like English, where verb and object are typically parsed into a single p-phrase (4) (but see Beck 1999 on Lushootseed Salish, Barthmaier 2004 on Okanagan Salish, for similar claims; Hayes  121  and Lahiri 1991 on individual p-phrases for verbs and arguments in Bengali, Schafer and Jun 2002 on Korean, Nespor and Sandler 1999 on Israeli Sign Language, Ishihara 2007 on Japanese). In (6), the nuclear accent is rightmost at the i-phrase level. A t the prosodie word level, there are also just two stresses in this example (shown by acute accent '). When focusing an object or subject DP, a cleft structure is employed (Kroeber 1997, 1999). The surface form closely parallels clefts in English (Perçus 1997, Hedberg 2000); I'll take up this matter in more detail in chapter 7. (7)  Cleft predicate  Cleft head  Cleft clause/residue clause  It  is  [«]FOC  that has the property 11  It  is  [Monique IPOC  that I saw.  [e Moniquejpoc  [e  9 y  xe?  ce C  L  E  F  T  D  E  M  D  E  T  Monique  D  wik-t-0-ne]G. E  T  see-TR-3o-l  S  G  .  T  S  "[I sawJo [Monique]Foc." (literally "It is [Moniquelppc [that I sawlg.") If we find the same "stress-focus" and "destress-given" constraints in NteVkepmxcin as we do in English, we should see the following sorts of prominence marking in a Thompson Salish cleft: (8)  Expected prominence relations in clefted foci [hypothetical] (  X  ( cé C  L  E  F  T  (  X  )  (  X  )  [e MoniqueJFor [e  D  D  M  E  T  Monique  i-phrase: STEP 2  )  recursive p-phrase: STEP 2 p-phrase at interface: STEP 1  xe? E  )  D  wik-t-0-neJG. E  T  see-TR-3o-lSG.TS  "I saw [MoniquejpQç." (literally "It is [MoniqueIpo^ [that I sawl^.") If information structure is marked acoustically in Nfe?kepmxcin like in English, then the acoustic cues should be parallel to the English case. In terms of focus marking, Monique is now at the left edge of the clause (the first lexical element, and the first p-phrase accent). The left edge accent of the clause should thus be acoustically distinguished by carrying a nuclear pitch accent instead of just a p-phrase pitch accent in the default case in (6). In terms of givenness marking, the right edge of the clause should now lack a pitch accent, while the 122  right edge carries the nuclear pitch accent in the default case. Again, if this sort of deaccenting occurs in N+e?kepmxcin, it should be marked by measurable acoustic correlates. Finally, the declination between stressed items within each utterance in the focused cleft cases should be greater than declination in the default case. Table 5.1  D  (  E  W  F  I  Comparisons to be made  A  D  U  E  L  T  C  P  U  F  T  O  T  C  E  U  R  S  A  N  C  E  N  A  R  R  O  W  F  O  )  1.  Left edge stress  2.  Right edge stress (nuclear stress)  3.  Declination within each utterance  C  U  (  S  C  L  E  U  T  F  T  T  O  E  R  R  A  N  N  P  C  E  C  )  Narrow foci at left edge <^  Given material at right edge Declination within each utterance  5.1.2 Acoustic correlates of prominence: overview Now the question arises as to what sorts of acoustic measures are relevant for the comparisons proposed in Table 5.1, and what constitutes a perceptual difference for each. In many respects, what we know about which combinations of acoustic cues are important for nuclear stress is still best summed up in the abstract of Fry (1958): Differences of stress are perceived by the listener as variations in a complex pattern bounded by four psychological dimensions: length, loudness, pitch and quality. The physical correlates of these perceptual factors are the duration, intensity, fundamental frequency and formant structure of the speech sound waves. Fry (1958: 126) Fry (1958) reported on an experiment using artificial stimuli varying on the dimensions of duration, intensity and FO. Fry suggested that pitch outweighs duration and intensity as a cue for stress (a common theme in much of the literature to be examined). Changes in fundamental frequency. Fry found, tend to be judged as "all or nothing:" the fact that a change took place is more important than the magnitude of that change. Fry also found that stress is judged to be incrementally greater as duration and intensity increase, with duration playing a larger role than intensity, but FO the dominant factor. As Sluijter and van Heuven (1996a, 1996b, 1997) point out though, many experiments like Fry's have failed to differentiate correlates of stress from correlates of pitch accents (1996b: 2471 for discussion). This is because all utterances, even single word utterances, are parsed into prosodie phrases, which carry pitch accents, and many studies  have confounded these two variables. Sluijter and van Heuven therefore compared the acoustic markers of pitch accented syllables with stressed but unaccented syllables. They showed that FO movement characterizes pitch accents but not lexical level stress in stressaccent languages like Dutch and English (1996a, 1996b; also Choi et al. 2005, Shue et al. 2007; de Moraes 1998:188 on Brazilian Portuguese; Suomi et al. 2003 on Finnish). FO has therefore been perhaps the most consistently linked acoustic cue for pitch accents. Many studies of intonation in general have concentrated primarily on the role of FO (eg. 't Hart, Collier and Cohen 1990, Gussenhoven 2004, and many others). In a stress language like Thompson Salish, we thus expect pitch to play an important role in cueing phrasal accent. Overview of aspects of FO change serving as accentual cues. Gussenhoven (2004) provides an excellent overview of FO cues that signal greater emphasis: (9)  FO cues for emphasis and example language (Gussenhoven 2004) a. higher FO peaks (English) b. greater FO excursions (English) c. later FO peaks (English, German) d. earlier FO peaks (Serbo-Croat) e. different pitch accents in narrow focus versus wide focus (European Portuguese) f.  higher FO register (Dutch)  Gussenhoven reports on various experiments examining what aspects of FO are interpreted as signalling greater emphasis on one constituent over another. Basic cues are a higher peak FO, and a greater pitch excursion (a greater range between the maximum and minimum FO). The experiments reviewed, in this and the following sections, that look at height of the FO peak find a difference of 2 to 7 semitones as a signal of emphasis (results vary by language and level of phonological accentual difference). In Nêhiyawêwin (Plains Crée), it is falling pitch which is taken as the marker of prosodie prominence (Edwards 1954, Muehlbauer 2005), and can be characterized as a low FO accent L * (Muehlbauer 2005). Gussenhoven also remarks that a later FO peak can "substitute" for greater pitch excursion as a prominence cue (2004: 91-92). Later peaks are presumably interpreted as allowing for more time to implement a greater pitch excursion; indeed, accented syllables in American English are produced with both greater pitch excursions and later peaks (Sluijter and van Heuven 1996a, Choi et al. 2005, Shue et al. 2007). In Hamburg German (Peters 2002), narrow focus is also indicated by later pitch peaks.  In some cases, earlier peak accents are also characteristic of focal accents. Gussenhoven (2004: 93) reports that Serbocroat (Smiljanic and Hualde 2000) has pitch peaks that are 100 ms or so earlier on focus accents. Contra Shue et al. (2007), Eady and Cooper (1986) found that American English speakers used earlier pitch peaks in sentence-initial narrow focus accents. Gussenhoven surmises that earlier pitch peaks are more difficult to implement; presumably this can reflect the speaker's use of what Gussenhoven calls the Effort Code, the idea that important information is highlighted through additional articulatory effort (2004: 85-93). Differences in timing of pitch peaks may also indicate that different pitch accent tunes are employed in narrow versus wide focus contexts. In European Portuguese, for example, broad focus is marked with a H+L* pitch accent, while contrastive narrow focus is marked with *H-i-L (Gussenhoven 2004: 61, 86-87; Frota 1998).^^ Dutch listeners also perceived higher registers as signalling more emphasis (though British interpreted higher registers as less emphatic - Chen et al. 2002). I will continue by reporting on what specific acoustic correlates have been found to indicate focus accenting across languages. 5.1.3 Acoustic correlates of focal accenting (nuclear pitch accent) There are few studies, to my knowledge, that directly answer the question regarding our measurement 1 (what is the acoustic difference between a p-phrase accent at the left edge, and an i-phrase accent in the same position?). There are a few examples, sometimes estimated from tables provided by previous researchers (for example, a number of the studies in Hirst and Di Cristo 1998 provide useful tables or figures in this regard). American English: Eady and Cooper (1986). This study directly addresses our measurement 1. Eady and Cooper had six subjects read both questions and statements with either "neutral focus" (wide focus), or narrow focus on the initial or final element. They found that narrow focus on the initial element resulted in 31.2% increase in duration (1986: 407, Table II). Interestingly (and unlike most other studies reviewed here), FO peaks on the initial word did not increase significantly from the neutral condition when the initial word was narrowly focused; instead, there was a significant drop in pitch after the focal accent (deaccenting) of about 46 Hz (-5.6 semitones), between the focus and the next stressed word. This drop was 3.6 semitones greater than in the neutral case. Subsequent stressed words were  It may be possible to characterize this difference as an early peak alignment in the narrow focus cases (PuUeyblank, p.c).  deaccented approximately 10-20 Hz (-1.4 to -2.7 semitones) when compared to the neutral focus sentences (1986: 407-408). The authors conclude: ... sentence-initial focus does not result in an increased FO value on the focused item. Instead, the presence of emphatic stress at this position is realized by an increase in word duration ... and by a very sharp post-focus FO drop. (1986: 408) Eady and Cooper also determined that initial narrow focus had an earlier pitch peak (at 0.35 of the vowel duration, standard deviation 0.13) than the same word in a neutral or given context (0.61 and 0.63 of vowel duration respectively, standard deviations 0.14 and 0.11 ). Earlier pitch peaks are more difficult to implement, and thus may reflect additional effort on the speaker's part to mark focused information (Gussenhoven 2004: 85,93). Curiously, another similar study (Shue et al. 2007) found that American English speakers employed a later pitch peak in this position. This suggests that timing of pitch accents is not a consistent cue, or may vary with dialect. American English: Cooper et al. (1985). In a related study, the authors presented two experiments to test the acoustic correlates of contrastive stress. In experiment 1, the authors had six subjects read a prepared sentence in answer to a questions. The location of the focus was manipulated in the question. Like Eady and Cooper (1986), Cooper et al. found that focus on the initial word did not result in significantly higher FO peaks, but did manifest itself in a durational increase of 38% - 41 %, and a sharp drop of 4.14 semitones to the next stressed word (1985: 2146, table III). A follow-up experiment with longer test sentences produced similar results, with a durational increase of 36.9% on initial focused words. Again, they found no evidence for higher FO peaks on initial words, but FO excursion appeared important, with a "dramatic" post-focus drop of 4.8 semitones to the next stressed word (1.5 semitones greater than in the neutral case) (1985: 2151, table VI). American English: Eady et al. (1986). In a follow-up set of studies, the authors again investigated the influence of duration and FO in differentiating neutral focus sentences from narrow focus sentences (exactly what we will be examining). In their experiment 1, seven subjects were asked to answer questions with prepared sentences; the questions manipulated the focus type. For initial words, there was no difference in duration or peak FO between words in the neutral focus condition and narrow focus condition. A second experiment, using shorter speech stimuli (comparable to the data in the present experiment), did find significant effects. This time FO peaks were 1.5 semitones higher on narrowly focused initial words in comparison to the neutral focus condition. A n additional 2.3 semitone drop to the next stressed word in comparison to the neutral condition, and a 27% earlier pitch peak, also indicated greater pitch excursion (i.e. pitch range) on narrowly  focused items. In addition, initial focused words were 34.4% longer than in the neutral case (1986: 244, table 3). The authors concluded that the acoustic effects of focus were more apparent in shorter utterances, comparable in length to those in the present study. American English: Shue et al. (2007). Shue et al. examined accented versus unaccented syllables for ten speakers of American English, for the sentences "DaGAda gave Bobby doodads" and "Dagada gave Bobby DOOdads." The authors looked at both statements and echo questions. These examples again very closely parallel the cases to be examined in the present study. The focal accent on Dagada (English subject focus) places the focus at the left edge of the clause, like we would expect in narrow focus Salish clefts, while the material at the right edge is given. The focal accent on doodads, in comparison, has the nuclear accent at the right edge, just like default wide CP focus cases in Thompson Salish. Shue et al. found that stressed syllables at the left edge carrying a H * nuclear pitch accent were, for males, "about 15 Hz higher" (about 2 semitones) than without the nuclear accent, while the H * peak was realized only at the start of the following syllable (about a 100ms delay) (2007: 2626-7). Finnish: Suomi et al. (2003). A similar experiment manipulated the location of focal accent in Finnish sentences read by speakers in a laboratory study. The authors distinguished strong accent (eg. contrastive focus), moderate accent, and deaccented words, though the measurements occurred quite late in the utterance as opposed to the initial position of interest in the present study. Strong accent corresponds to our focus condition, while moderate accent corresponds to our neutral condition. The study examined words with the syllable structures C V C V , C V V C V , and C V C V V . Strong accents (eg. focused words) had a 2.0 semitone greater pitch peak than accents in the neutral condition, and a 3.4 semitone greater FO fall (2003: 130). Vowels in the strong accent condition were from 24 - 37% longer for V , , or 11 - 34% longer for V j (2003: 122), than vowels under moderate accent or no accent. Dutch: 't Hart (1981). In a perception experiment, 't Hart tested listeners' ability to distinguish the difference in prominence-lending pitch movements in utterances containing two stressed syllables, 't Hart manipulated the amount of pitch rise and the end pitch of the speech stimuli. He found that listeners were broadly grouped into three levels of perceptual ability: quite a few failed to discriminate pitch-rise differences of less than 4 semitones ("nondiscriminators"); another group falsely based the pitch comparison on the final pitch point ("final pitch discriminators"); and the remaining subjects could discriminate differences of 1.5 to 2 semitone in peak FO heights. The results were essentially the same when repeated with piano tunes rather than speech stimuli, 't Hart concluded that, in order to be interpreted as a more prominent accent, the second pitch movement in running speech should be at least 3 semitones greater in range than the first.  Swedish: Gârding (1998). Gârding reports on a study of Southern Swedish, comparing pitch contours in a default utterance ("focus-free" in her terms) with rightmost sentential accent, and the same utterance with narrow focus on the initial pitch accent. This comparison closely parallels the Thompson Salish cases to be examined here. Placing narrow focus on the leftmost pitch accent resulted in an additional 20 Hz or so (2.3 semitones) in the upward pitch movement (1998: 125, Figure 4). Danish: Gr0nnum (1998). Gr0nnum reports on a study comparing neutral intonation in sentences with three stress peaks, with contrastive focus on one of the three stress peaks. For the left edge, the additional focal prominence resulted in 2 semitones greater FO peak, and an FO fall that was 4 semitones greater, than phrasal stress alone on the left edge (1998: 142-3, Figure 6). Greek: Botinis (1998). Botinis reports on a production experiment for Greek, and finds that focal stress added an additional 12.1% duration over and above simple phrasal stress (1998: 302). Focal accent also seems to result in much larger pitch movements, with a greater FO peak of 100 Hz or more, though exact details are not recoverable from the graphic representations provided (1998: 303, Figure 7). Hungarian: Fonagy (1998). Fonagy reports that contrastive focus ("focalisation with implication" in his terms) in Hungarian results in a sharp 7-8 semitone pitch rise on the first syllable of the focal constituent, followed by a large 13-14 semitone fall. Unfortunately comparison with the default case is not provided, so we are unable to determine how much greater these rises and falls are for a p-phrase accent alone (1998: 340-341, Figure 3). Beijing Chinese: Kratchovil (1998). The author reports the results of a study of the sentence méiyou mqing 'there was no love,' with non-emphatic phrase-final stress (underlined), with méiyou àiqing where emphatic stress falls on the first word. The test consists of a single pair of sentences taken from a corpus recording of a single speaker, and should be treated with caution regarding generalizability. Shifting the primary accent to the first word resulted in overall higher tone (presumably because the first word carries higher tone than then second), but crucially an additional 120 Hz (7.4 semitones) on the focused méiyou FO peak, and an additional 95 Hz difference (4.3 semitones) in the fall between the focused méiyou FO peak and the deaccented aiqing FO peak (1998: 428-429, Tables 2-3). Nêhiyawêwin (Plains Crée): Cook and Muehlbauer (2005). In a general descriptive study examining acoustic cues to various constituents in Plains Crée, the authors find that pitch (FO) plays an important and consistent role. Cook and Muehlbauer report that right-edge focus has "exaggerated pitch movement" that is clearly visible in pitch tracings, as well as final lengthening and glottalization.  Summary: correlates of nuclear pitch accent. Given the above findings, the best estimates for acoustic markers of nuclear pitch accent on left edge focus are as follows. 2 semitones greater FO peak seems to be a conservative estimate for the increase in maximum FO on focused items. In absence of a higher FO peak, we might look for a greater FO range, again conservatively estimated at 3.6 semitones (Eady and Cooper 1986). Timing of the FO peak may be offset in either direction by a rather large value (100 ms), but we'll adopt the more conservative estimate from Eady and Cooper (1986) and Eady et al. (1986), where FO peak was reached 27% earlier in initial words carrying the nuclear stress. Alternately, we can look for delay of pitch peak into the following syllable, which should be indicated by overall late timing of the FO peak in the vowels of interest. There are fewer estimates of durational increases, but we may expect duration of vowels carrying the nuclear stress to increase anywhere from 12% to 34% in comparison to vowels carrying just a p-phrase stress. The table on the next page summarizes the results.  Table 5.2  Acoustic cues for left edge narrow focus accent (nuclear pitch accent), as compared to wide focus accent (p-phrase pitch accent)  D  (  E  W  F  I  A  D  U  E  L  T  CP  U  T  T  E  R  A  F  O  C  U  S  )  N  C  E  N  O  R  N  O  A  M  R  I  R  N  O  A  W  L  Left edge stress  American English  F  P  R  O  E  C  D  U  I  S  C  A  U  T  T  E  T  E  R  A  N  C  E  C  O  N  S  T  R  U  (  C  L  E  F  T  C  T  I  O  N  )  Narrow foci at left edge  •  No additional FO peak, +3.6 semitones greater fall (Eady and Cooper 1986)  •  +1.5 semitones peak FO, +2.3 semitones greater fall (Eady et al. 1986)  •  +2 semitones peak FO (Shue et al. 2007)  Swedish  •  +2.3 semitones peak FO (Carding 1998)  Dutch (perception)  *  ^3 semitones greater pitch excursion on prominent peak ('t Hart 1981)  Danish  •  +2 semitones peak FO, +4 semitones greater FOfall (Gr0nnum 1998)  Hungarian  •  7-8 semitones FO rise, 13-14 semitone FO fall (Fonagy 1998)  Finnish  •  +2.0 semitones peak FO, +3.4 semitones greater FO fall (Suomi et al. 2003)  American English  •  earlier pitch peak (0.35 vs. 0.61/0.63) (Eady and Cooper 1986)  •  earlier pitch peak (0.37 vs. 0.64) (Eady et al. 1986)  •  later pitch peak (100 ms) (Shue et al. 2007)  Hamburg German  •  later pitch peak (Peters 2002)  Serbocroat  •  earlier pitch peak (100 ms) (Smiljanic and Hualde 2000)  American English  •  +31.2% duration (Eady and Cooper 1986)  •  +34.4% duration (Eady et al. 1986)  Greek  •  +12.1% duration (Bonitis 1998)  Finnish  •  +11-37% duration (Suomi et al. 2003)  5.1.4 Acoustic correlates of deaccenting The experiments reported in this section all tested laboratory speech. Experimenters had speakers produce a target word in a carrier phrase. The focal accent of the utterance was varied to be either on the target word, or on another word in the carrier phrase. Results from these tests thus illustrate what acoustic measures characterize pitch accented syllables from stressed but unaccented syllables. The target word was stressed on either the first or second syllable. To what extent these properties are generalizable to spontaneous discourse is a matter for debate, but at least some research suggests that the similarities of "lab speech" and spontaneous speech are more pronounced than the differences (Klatt 1976: 1209, Lickley et al. 2005). I'll return to this issue in section 5.1.6 and section 5.3. Dutch, American English: Sluijter and van Heuven (1996a, 1996b). In an experiment, Sluijter and van Heuven (1996b) had 12 Dutch speakers produce a two-syllable minimal pair differing only in stress placement (/'ka:nDn/ 'cannon' versus /ka:'non/ 'canon') in a carrier phrase, Wil je [target] zeggen ' W i l l you say [target].' The words were produced either with a pitch accent ([+Focus| condition), or with a contrastive focus on the verb say in the carrier phrase instead ([-Focus] condition). Words were produced both in lexical form, and as réitérant 'nana' syllables. A similar experiment with 6 speakers of American English (1996a) looked at four two-syllable verb/noun pairs that are distinguished by stress placement (eg. 'permit'), again in the carrier phrase Will you say [targetJ. The key findings relevant to the present study can be summarized as follows. Sluijter and van Heuven found that accented syllables [+Focus[ have a significantly greater intensity than unaccented syllables. In Dutch this difference was found to be 2.9 db, regardless of which of the syllables in the two-syllable Dutch words examined carried the stress (1996b: 2475, Table II). In American English, Sluijter and van Heuven found the effect of stress on intensity to be small (1-3 db), but found a "considerably larger effect of [focus] accent" - an estimated 5 db, based on their graphic presentation (1996a: 3, Figure I). The second finding of relevance concerns the value of duration as a cue for deaccenting. Duration was found to be a very reliable cue of stress, but remained relatively unaffected in the additional presence of pitch accents (1996b: 2475). Accented syllables were, however, still slightly longer (significantly so) than unaccented ones. Sluijter and van Heuven measured the duration of syllables while the present study measures just vowel duration, but the percentage change from accented to unaccented position is still relevant. In the two-syllable Dutch words examined in Sluijter and van Heuven (1996b: 2475, Table II), accented syllables were 11.9% longer than unaccented syllables when the first syllable carried the stress, and 6.1% longer when the second syllable carried the stress. In American  English (1996a: Table I), accented syllables were 19.6% longer when the first syllable was stressed, and 17.3% longer when the second syllable was stressed. Because Dutch has a verbfinal order and hence a phrase-final nuclear accent when the verb is contrastively focused, the Dutch data measures deaccenting on prenuclear stresses; the American English data is therefore more relevant, since deaccenting was measured in the postfocal position, just like the right edge of Salish clefts. It is worthwhile noting that the experimental methodology employed by Sluijter and van Heuven depends on  D E S T R E S S - G I V E N  being operative in the target language; if given  information is not deaccented, then the correlates of pure word-level stress can not be straightforwardly measured. Catalan: Astruc and Prieto (2006). A similar experiment with 6 female Catalan speakers found that accented syllables had about 32 Hz (3.5 semitones) greater peak pitch, 11 % greater duration, and 4 db greater intensity than unaccented but stressed syllables (calculations adapted from their Figure 1). Astruc and Prieto noted that increased duration tended to be tied to increased pitch excursions on accented syllables, presumably to provide more time to realize the greater pitch movement. Finnish: Suomi et al. (2003). Another experiment in the spirit of Sluijter and van Heuven, this time on Finnish, found no significant differences in length when comparing vowels under moderate accent (our neutral condition) and word-level stress only (eg. deaccented vowels) (2003: 120, 122). However, pitch did play an important role, with moderate accent (eg. neutral condition) words having a 4.0 semitone movement to the pitch peak, and 4.1 semitone fall thereafter; in contrast, word-level stress alone showed no tonal movement (2003: 130). Scottish EngUsh: Turk and White (1999). This series of studies also looked at the effects of contrastive accent on duration, but in Scottish English. For one-syllable words, the authors found a 23% increase in length for accented versus unaccented but stressed words, while two-syllable words showed a 16% increase in duration when accented (1999: 184, Figure 3). A later experiment also found 23% increases in length, this time on the initial accented syllable in three-syllable words (1999: 197, Figure 7). Interestingly, this accentual lengthening spread rightwards and sometimes leftwards to neighbouring syllables within the Prosodie Word boundary. American EngUsh: Shue et al. (2007). As discussed in the previous section, this study tested 10 speakers of American English, who were asked to produce utterances varying the location of the focal stress. For stressed syllables at the right edge, females had approximately 40 Hz greater FO on accented than on unaccented syllables (about 3.5 semitones), but no peak delay (2007: 2627, Figure 2).  American English: Okobi (2006). This dissertation examined the acoustic correlates of stress and accent in American English (5 adult male speakers) on both novel and real words. Focus accents were either on the target word ("My grey [Dldilppc / [STAtuelppc drove here"); or on the preceding adjective ("My | B L U E ] F O C didi / statue drove here") or possessor ("[YOURJFOC  grey didi / statue drove here"). Like the Sluijter and van Heuven experiments,  this allowed comparison of pitch accented syllables with unaccented syllables. Novel words were didi, dodo, and dada, with stress either on the first or second syllable. Because the present study also examines the correlates of accent in Thompson Salish across vowel types, Okobi's results (computed from the measurements provided in his appendix B) should prove informative here. Okobi found that higher FO, greater peak intensity and longer duration correlated with pitch accented (2006:43-46), as compared to deaccented syllables.^' The differences were greater when stress was on the second than the first syllable (an effect attributed by Turk and White 1999 to greater intrinsic prominence at the right edge of a prosodie constituent, the PWd in this case), and the differences were also generally greater for the real words than for the nonsense words. When the first syllable was stressed, pitch acccented syllables were 4.7 semitones higher in their FO peak, 4.2 dB louder, and 18.4% longer than deaccented syllables, for nonsense words. Real words were 7.3 semitones higher in pitch, 4.95 dB louder and 15.6% longer when pitch accented than when deaccented. When word-level stress fell on the second syllable, pitch accents were 6.7 semitones higher, 6.15 dB louder, and 27.7% longer than deaccented syllables for nonsense words. For real words, a pitch accent resulted in 7.9 semitones greater FO, 6.15 dB more amplitude and 27.5% longer duration. Averaged across conditions, pitch accented syllables were 6.65 semitones higher, 5.36 dB louder, and 22.3% longer than unaccented, but stressed, syllables. American English: Eady and Cooper (1986). Eady and Cooper had six subjects read both questions and statements with either "neutral focus" (wide focus) or narrow focus on the initial or final element. There was a significant drop in pitch after the focal accent (deaccenting) of about 46 Hz (-5.6 semitones) between the focus and the next stressed word. Subsequent stressed words were deaccented approximately 10-20 Hz (-1.4 to -2.7 semitones) when compared to the neutral focus sentences (1986: 407-408).  Okobi also found greater amplitude as measured in the first harmonic, H I * , but I will not be considering spectral measures in this study.  American English: Cooper et al. (1985). In a related study, the authors reported on an experiment in which six subjects read answers to questions manipulating the site of contrastive focus. Cooper et al. found that final post-nuclear unaccented words had a 2.5 semitone lower FO peak (1985: 2146, table III). In addition, the durational difference between an unaccented final word and a final word carrying the nuclear stress was an average of 16.7%. A follow-up experiment with longer sentences and comparison to a neutral (wide focus) condition found somewhat different results. Unaccented final words were not shorter than in the neutral focus condition (1985: 2145, table IV). However, they were 2.3 semitones lower in FO peak (1985: 2151, table 6). American English: Eady et al. (1986). In follow-up study, the authors tested for the difference between neutral (wide CP) focus and narrow focus. In their experiment 1, deaccenting on the rightmost word produced no difference in FO peak or duration from the neutral sentence condition in this study. A second experiment, using shorter speech stimuli, deaccented righmost words had 1.4 semitones lesser FO peaks and 8.3% lesser duration than in the neutral case (1986: 244, table 3), but no difference in FO peak timing (0.23 and 0.22 of word duration respectively). Western Arabic (Morocco): Benkirane (1998). Benkirane reports that the rightmost nuclear stress is characterized by approximately a 6 semitone pitch rise, and an octave fall (12 semitones) to the end of the utterance. Unaccented syllables have no notable pitch movement (1998: 351-352, 356). Warlpiri: Butcher and Harrington (2003). In a study with two speakers of Warlpiri, the authors investigated the acoustic effects of words in a [focused] and [unfocused] condition. Focused words carried a prominent pitch accent, while unfocused words lacked this pitch movement. The differences in FO between the two conditions ranged from about 3.5 to 5.5 semitones (results estimated from their figure 1). Summary: correlates of deaccenting. Best estimates for acoustic correlates for deaccenting based on the above findings are as follows. Maximum FO peaks are expected to be approximately 3.5 semitones (or more) lower on deaccented material (eg. Shue et al. 2007). Maximum amplitude is expected to be 3 db (or more) lower, while deaccented vowels are from 6% to 28% shorter in duration.  Table 5.3  Acoustic cues for deaccenting of given material in nuclear stress position  D  (  E  W  F  I  A  D  U  E  L  T  CP  U  T  T  E  R  F  O  C  U  S  A  N  C  E  N  A  R  R  O  W  F  O  C  U  S  U  T  T  E  R  A  N  C  E  ( C L E F T )  )  Right edge stress  Given material at right edge  Catalan  •  - 4 db (Astruc and Prieto 2006)  Dutch  •  -2.9 db (Sluijter and van Heuven 1996b)  American English  •  -5 db (Sluijter and van Heuven 1996a)  •  -5.36 dB (Okobi 2006)  •  -3.5 semitones FO peak (Shue et al. 2007)  •  -6.65 semitones FO peak (Okobi 2006)  •  -1.4 to -2.7 semitones FO peak (Eady and  American English  Cooper 1986) •  -1.4 semitones FO peak (Eady et al. 1986)  •  -2.3 to -2.5 semitones FO peak (Cooper etal. 1985)  Catalan  •  -3.5 semitones FO (Astruc and Prieto 2006)  Western Arabic (Morocco)  •  - 6 semitones FO (Benkirane 1998)  Finnish  •  -^.0 semitones FO (Suomi et al. 2003)  Warlpiri  •  -3.5 to -5.5 semitones FO peak (Butcher and Harrington 2003)  Dutch  •  -6.1 % to -11.9% duration (Sluijter and van Heuven 19%b)  American English  •  -17.3% to -19.6% duration (Sluijter and van Heuven 1996a)  •  0% to -16.7% duration (Cooper et al. 1985)  Finnish  •  -8.3% duration (Eady et al, 1986)  •  no significant change in duration (Suomi et al. 2003)  5.1.5 Declination Our final measurement concerns the amount of topline declination between left and right edge stresses within each utterance. As air is expelled from the lungs during the course of an utterance, subglottal pressure drops, and is associated with an intrinsic accompanying decline in FO and intensity from left to right (Cohen and't Hart 1965, 't Hart et al. 1990, Strik and Boves 1995, Trouvain et al. 1998, etc.). Because we are interested in the effect of declination on peak prominences, I will measure declination as the difference in acoustic prominence between stressed syllables within an utterance. This means that we won't have a "true" measure of declination, but a "topline" measure indicating differences in accentual level (eg. 'tHartetal. 1990). The wide focus case will give us a default declination level for comparison. In narrow focus clefts, under a stress-focus account, the nuclear stress will shift leftward to the focus. This should result in increased prominence on the left edge, while deaccenting of given material at the right edge of clefts ("destress-given") will result in decreased prominence there. As a result, declination between stressed syllables at the left and right edge in narrow focus clefts should be greater than in the default case. I have not found any studies that directly address the question of which phonetic differences in declination contours are relevant for perceiving differences in prominence. We can test to see if the various measures of interest (FO peak, FO range, and amplitude declination) differ between the two focus types, and then consider if any differences are likely to be perceptually salient. Section 4.3.1 presented some background on declination measures for English, and neutral declination values for Nfe?kepmxcin are reported in section 4.3.4 and table 4.1. 5.1.6 Effects of data source A secondary question to be answered in this study is the effect of the type of data on the acoustic realization of intonation. Data came from either spontaneous utterances, scripted role-playing conversations, or single sentence elicitations. Knowing that single sentence elicitations, for example, exhibit similar acoustic characteristics as utterances taken from spontaneous conversation would suggest that speakers produce natural sounding utterances even when not in full conversational mode - that is, given sufficient context, a speaker is able to "imagine" herself in the midst of a conversation and produce an appropriate utterance (see also Matthewson 2004). Since single sentence elicitation or scripted conversation can more  rapidly and efficiently produce relevant linguistic data, we would like to know if these types of data also exhibit similar acoustic characteristics. This issue is addressed in a second study, reported in section 5.3. We might expect that the acoustic characteristics of spontaneous conversation will differ from those of both scripted conversation and single sentence elicitations. However, some previous research suggests that the acoustic differences between naturally produced speech and "lab speech" are less pronounced than the similarities (Klatt 1976: 1209, Lickley et al. 2005). Lickley et al. (2005) had speakers read test sentences in an examination of postnuclear FO minima in Dutch falling-rising questions. Of particular interest in this study, the authors investigated concerns that read speech in "laboratory phonology" studies is not a valid method for characterizing the intonation of spontaneous speech. On the other hand, even if there are "higher level" cognitive differences in the planning of read versus spontaneous speech (Levelt 1989), there may not be differences in "lower level" processes where "the planned utterance is translated into phonological/phonetic code .... That is, once a speaker has chosen a contour, it is a reasonable assumption that the contour's phonetic properties are largely or wholly predictable from phonetic and phonological factors alone" (2005:172). To test these hypotheses, the authors had 4 speakers produce a task-oriented dialogue using the Map Task (Anderson et al. 1991), which resulted in 21 questions directly comparable to those tested in the reading task. Both examination of the descriptive statistics and some statistical analysis failed to find any difference between read speech and spontaneous speech produced in the task-oriented experiment. The authors conclude that read speech can be "used as a source of evidence in experimental work that addresses phonological and phonetic questions" (2005:179), with its obvious practical advantages of using tightly controlled speech materials. Similar findings in the present study would prove useful for field linguists investigating intonation, or other phonetic and phonological phenomena.  5.2  Case study: neutral versus narrow focus In this section, I report the results of a detailed acoustic phonetic study of different  focus types in Thompson Salish; neutral, wide CP focus on the one hand, and narrow object or subject focus on the other.  5.2.1 Subjects The language data was collected from two female speakers of Nie?kepmxcin in their late 60's (FE and PM). Both are speakers of the Lytton dialect, and fluently bilingual in English. 5.2.2 Method Different instances of focus were identified from the corpus of conversational recordings that served as a source for the data reported in Chapter 3. Recordings were made at the residence of either the language consultants or of the researcher, using a Marantz P M D 670, 671 or 660 digital audio recorder. Each consultant was recorded on a separate channel using a Countrymax Isomax E M W Lavalier lapel microphone. The microphone was attached onto the exterior of the consultants' clothing, approximately at the sternum. To account for declination effects, only utterances which were completed in a single breath group were considered. For the present case study, neutral cases (wide CP focus), where everything in the utterance was new information, provided a default intonation marking (Gussenhoven 1984: 17-18, 65-68; Hayes and Lahiri 1991: 56, Selkirk 1995). Narrow subject and object focus, which is either clefted or produced in a nominal predicate construction, was compared to the default case to determine if and how the acoustic signal differed. Thus, one factor in the analysis was focus type: neutral focus, or narrow focus. For each utterance, stressed vowels were identified in Praat (Boersma and Weenink 2007). Of primary interest were the vowels of the left and right edge. The first lexical vowel at the left edge is the verb in the default case, and typically the focused noun in the narrow focus utterances. These were the vowels that were compared when testing the hypothesis that narrowly focused items carry greater acoustic prominence. Absolutely leftmost stressed vowels were also measured (sometimes these were functional items, like the cleft predicate or auxiliaries). These were the vowels that were measured for the purposes of calculating utterance declination from left to right stresses, since they had higher absolute FO and amplitude peaks than following material. At the right edge, the rightmost stressed vowel in the default case is the nuclear stress, while in the focus cases it is old/given information in the cleft clause. In addition, other stressed vowels throughout the utterance were identified, in order to provide a better overall picture of the declination contour throughout the utterance, as well as a better account of variability. Thus, the second factor in this analysis was vowel position: leftmost lexical stress (the verb, or the focus), rightmost stress (the nuclear stress position), or other.  Utterance length was also identified in Praat. For both individual vowels and entire utterances, a variety of acoustic measurements were then made by using automated scripts in Praat. Pitch measurements of primary interest were the maximum and minimum FO, the standard deviation of FO, and the timing of the FO maximum and minimum (expressed as a percentage of the vowel duration). Where the Praat algorithm mismeasured FO (eg. in the presence of glottalization, etc.), measurements were done by hand via visual inspection of the waveform, and automated measurements were disregarded. The average and maximum intensity (in decibels) was also recorded, as was vowel and utterance duration. The number of syllables per utterance was counted, to provide a measure of speech rate. 5.2.3 Statistical analysis Results were analyzed for means (M), standard deviations (SD), and statistical significance. Numbers of observations (n) and degrees of freedom (df) are also reported where relevant. The choice of analysis method is not immediately obvious. Since the data reflect repeated observations from each subject, a repeated measures A N O V A may seem like a promising candidate; but with only two subjects, this technique lacks statistical power (Ladd and Schepman 2003, Lickley et al. 2(X)5). Because many phonetic experiments are based on a similar model, in which many observations are collected from a small pool of subjects, there is precedence for adapting statistical analyses to meet the needs of the experimental design. Ladd and Schepman (2003: 86-87) analyze the data from each of their two subjects by treating each speakers' data in a separate between-subjects A N O V A . Another possibility, employed in an experiment with seven subjects by Lickley et al. (2005:167), is to treat speaker as a between-items factor. The primary objection to these modifications is that they violate an underlying assumption of the A N O V A framework, namely that each data point is independent: "Independence simply means that the observations within or between the [groups] are not paired, correlated, matched, or interdependent in any way" (Hopkins, Hopkins and Glass 1996: 207). In the present design, we can be reasonably certain that the observations comparing the left edge of wide focus sentences with the left edge of narrow focus sentences are fairly independent, since they come from different utterances (though still from the same vocal tract). Similarly, the measurements comparing the rightmost nuclear stress position across utterances are relatively independent. However, observations taken within a single utterance (the left edge, the right edge, and other stressed vowels in between) will be less independent since they originate in the  same breath group. In this case, the mathematical A N O V A model "is likely to underestimate the true variability of the results, leading to tests that are biased towards rejection of null hypotheses" (Keppel and Wickens 2004: 142-3, citing Scariano and Davenport 1987). Since the present hypothesis is that speakers of Thompson Salish do not mark focus or givenness with pitch accent, we expect not to find a significant result in an A N O V A framework; as such, violations of independence actually work against the present hypothesis since they increase the likelihood of its rejection. The A N O V A therefore seems a suitable method for testing for main effects and interactions. Nevertheless, reporting a series of non-significant results is not necessarily desirable. Rather than using the A N O V A framework for analysis, another option is the t-test. The t-test easily allows the null hypothesis to be set to one which anticipates a difference between means (Keppel and Wickens 2004: 72). Since the null "stress-focus" hypothesis predicts that focused items will be marked with additional acoustic prominence, we can adapt the t-test to examine the anticipated difference for each acoustic parameter. Failure to mark focus acoustically will then be indicated by a significant statistical result. Since we are interested in specific comparisons (left edges in wide focus versus narrow focus utterances, and right edges, but not intermediate points), we can ignore the other stressed syllables which are the primary cause of concern for the violation of the assumption of independence. The use of a null hypothesis where two means are expected to be unequal, and an experimental hypothesis where the means are expected to be equal, is somewhat unusual. However, it should be made clear that there is nothing in the statistical model requiring a null hypothesis of ^i-^  = 0. Null hypotheses can perfectly well "specify outcomes rather than  absence of an effect" (Keppel and Wickens 2004: 72), and an honest examination of the focus marking literature presents just such a case. The overwhelming result of research into focus marking is that focus is marked by additional prosodie prominence; this is the null hypothesis, and its rejection would be an interesting result. Thus, where possible, I specify a null hypothesis based on "stress-focus" generalizations in the literature. This methodology risks missing the possibility that it is some combination of FO, amplitude and duration which marks prominence, rather than values of individual variables (Vatikiotis-Bateson, p.c). A multivariate or regression model could get at this possibility, but is beyond the scope of the present study. However, even in English, no one to my knowledge has developed a model which predicts which combination of acoustic parameters mark prominence, and virtually all of the studies reviewed in the previous sections also examine acoustic variables individually (and often only one variable: FO). Thus, the present statistical design is broadly comparable to other research in the field. Furthermore, by reporting means and standard deviations for multiple aspects of all three indicators (FO, amplitude and  duration), I also go some distance beyond many other studies in terms of providing a complete picture of speech prosody in NteVkepmxcin. For the present case study, planned comparisons of the means between the left edge focus position and the right edge nuclear stress position were carried out using independent sample t-tests for each variable (using pooled variances). Due to the number of variables analyzed and number of comparisons performed (29 t-tests are reported on in total, for this case study), a conservative p-value of 0.001 was chosen for significance, to avoid an inflated family-wise error rate. With p=0.001, the family-wise error rate is limited to 0.029, close to the standard value of 0.05. To indicate trends in the data, however, I distinguish three levels of significance in the tables used to illustrate results: p<0.05 and p<0.01 are marginally significant (indicated with * and ** respectively), while p<0.001 is the chosen significance level (indicated by ***). Since the phrasal accents that are under investigation are pitch accents, change in FO parameters ought to be the primary phonetic indicator of both increased accent level on focused material, and deaccenting of old information. 1 will begin by considering the local effects of pitch at the left edge focus position, and then at the right edge nuclear stress position. I will conclude by examining declination effects across utterances as a whole. I report means (M), standard deviations (SD), number of observations (n), and degrees of freedom (df) where appropriate. 5.2.4 Results: the leftmost lexical stress The left edge: general comparisons. In the two categories to be compared (neutral wide C P focus, and narrowly focused DPs), general measurements found that the utterances in the two groups were similar. Mean utterance duration was 2.33 seconds in the default cases (n=41, SD=8.29 sec), and 2.48 seconds in the narrow focus sentences (n=65, SD=7.65), a non-significant difference (t=0.898, df=104, p>0.3). The stressed vowels to be compared (in the first lexical item from the left edge) were an average of 2.41 syllables from the left edge in the default case (n=41, SD=2.76), and 2.48 syllables from the left in the narrow focus case (n=65, SD=1.92). Again, this difference was non-significant (t=0.137, df=104, p>0.8). These findings suggest that any differences found between the two utterance types is not likely to be due to any declination effects, but rather will reflect other factors (such as information structure). The left edge: pitch (FO). In section 5.1, we saw that increased phonetic prominence could be marked by higher FO peaks, later (or earlier) FO peaks, and greater FO excursions.  At the left edge, narrowly focused subject or object DPs were not marked with higher FO peaks than left-edge verbs in the default wide focus utterances. For FE, left-edge verbs in the default case had a mean maximum FO of 202.5 Hz (n=22, SD=16.9 Hz). Focused NPs in narrow focus clefts in fact had a slightly lower FO peak, on average (M=193.8 Hz, n=34, SD=11.1 Hz). The same pattern was followed by the second speaker, P M . PM's left edge verbs in the default case had a mean maximum FO of 164.8 Hz (n=19, SD=15.0 Hz). Focused NPs had, on average, a lower FO peak (157.9 Hz, n=31, SD=19.1 Hz). A note on boxplots. The boxplots (or "box and whisker" plots) throughout this chapter display the data separated by speaker and focus type. Figure 5.1 shows the results for peak FO values on left edge lexical stresses. The dark line in the box represents the median value, and the box show the interquartile range (the middle 50% of the data points). The whiskers in either direction represent values 1.5 times the interquartile range. Outliers are indicated as open circles or stars beyond this range (see Howell 1992: 48-51 for a more detailed description of boxplots).  Figure 5.1  Maximum left edge FO by speaker and focus type  Under the null hypothesis, narrowly focused items were expected to carry an FO peak that was at least 2 semitones greater than the left edge verbs in the default focus cases.  (10)  Null hypothesis (HQ): Left edge focused items have 2 semitones greater FO peaks |Li, - >  2 semitones  This hypothesis was not supported. Independent sample t-tests for both F E (t=-9.00, df=54, p<0.001) and P M (t=-5.27, df=48, p<0.001) were significant, allowing us to reject the null hypothesis that Thompson speakers mark narrowly focused items in left edge clefts with greater FO peaks. The size of the FO excursion (the FO range between maximum and minimum), in semitones, was also compared for the two focus cases. Again, both speakers followed a similar pattern. In the default wide focus utterances, FE had an average FO excursion of 2.18 semitones (n=22, SD=1.03 semitones) on left edge verbs, while narrowly focused DPs at the left edge had a similar FO excursion (M=2.53 semitones, n=34, SD=1.30 semitones). For P M , the left edge in the default case had an FO range of 1.55 semitones (n=19, SD=1.10 semitones), while narrowly focused DPs at the left showed an FO range of 1.64 semitones (n=31, SD=0.94 semitones).  7 5  5  8  4  0  8  8 0 N =  22  34 FE  19  31  I I narrow focus  PM  Speaker Figure 5.2  Left edge FO range (semitones) by speaker and focus type  Based on previous research, the null hypothesis was that narrowly focused items at the left edge would show a greater pitch excursion of at least 3.6 semitones, when compared with left edge verbs in the default wide focus case. (11)  Null hypothesis (HQ): Left edge focused items have 3.6 semitones greater FO range |Xi - |j,2 > 3.6 semitones This null hypothesis was not supported. Neither speaker marked narrowly focused  DPs with greater pitch excursions. Independent sample t-tests for both F E (t=-9.89, df=54, p<0.001) and P M (t=-12.02, df=48, p<0.001) were significant, allowing us to reject the null hypothesis that Thompson speakers mark narrowly focused items in left edge clefts with FO excursions that are at least 3.6 semitones greater than pitch excursions in the default sentences. Next, left edge lexical stresses were examined for the timing of FO peaks. Previous research suggests that, as a percentage of total vowel duration, narrow left edge focus may be marked by earlier pitch peaks of as little as 27% (Eady and Cooper 1986), though other studies have reported differences of 100 ms or more (which would amount to an 80% earlier FO peak, given that average vowel duration was 124.7 ms). (12)  Null hypothesis (Ho): Left edge focused items have a 27% earlier or later FO peak lii-\X2> 0.27 In the present study, lexical items at the left edge were similarly marked in terms of  timing of FO peaks in both focus conditions. For FE, FO peaks occurred at 21% of vowel duration in left edge verbs in the default case (n=22, SD=30%). In narrowly focused DPs, F E marked FO peaks slightly earlier (M=17%, n=34, SD=27%). In both cases, however, the majority of FO peaks occurred very early or at the onset of the vowel: in neutral focus cases, over half the FO peaks occurred in the first 10% of the vowel, while in narrow focus cases two-thirds of FO peaks occurred in the first 10% of the vowel. For P M , FO peaks in the default case occurred on average at 45% of the vowel duration (n=19, SD=33%). P M ' s focused NPs had somewhat earlier pitch peaks (M=29% of vowel duration, n=31, SD=29%). In the default wide focus cases, P M ' s FO peaks were quite evenly distributed throughout the vowel duration, but in the narrow focus cases, 45% of PM's FO peaks occurred in the first 10% of the vowel duration. For both speakers, average differences in peak timing were only about half of the expected difference of 27%, and much less than 100 ms (given that the average duration of 144  the vowels was 124.7 ms). Since most pitch peaks occurred early in the vowel, there was no evidence for delay of FO peaks until onset of following syllables (as found by Shue et al. 2007 for American English).  Oneutral focus •narrow focus  Speaker Figure 5.3  Left edge time of FO peak as a percentage of vowel duration  Because the data were not normally distributed (most FO peaks occurred near the start of the vowel) and did not have equal numbers across conditions, I did not perform t-tests for this variable. The left edge: amplitude (db). Since amplitude is a correlate of stress, we would expect narrowly focused DPs carrying a nuclear pitch accent to carry greater amplitude. Review of previous studies, which tend to focus on FO, did not reveal any specific predicted differences for this variable for a left edge nuclear stress. (13)  Null hypothesis (HQ): Left edge focused items have a greater amplitude peak ji, -  ^2  > 0 db  For FE, narrowly focused DPs did have a greater amplitude (M=76.0 db, n=34, SD=4.97 db) than left edge verbs in the default case (M=74.2 db, n=22, SD=4.64 db), a  difference of 1.8 db. However, it is not clear whether this difference is perceptually salient. For example. Draper et al. (1989: 20-21) report that humans can t