UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

On the distribution and representation of schwa in Sliammon (Salish) : descriptive and theoretical perspectives Blake, Susan Jane 2000

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

Item Metadata


831-ubc_2001-610659.pdf [ 18.12MB ]
JSON: 831-1.0099657.json
JSON-LD: 831-1.0099657-ld.json
RDF/XML (Pretty): 831-1.0099657-rdf.xml
RDF/JSON: 831-1.0099657-rdf.json
Turtle: 831-1.0099657-turtle.txt
N-Triples: 831-1.0099657-rdf-ntriples.txt
Original Record: 831-1.0099657-source.json
Full Text

Full Text

On the Distribution and Representation of Schwa in Sliammon (Salish): Descriptive and Theoretical Perspectives  by Susan Jane Blake  B.A. Linguistics, University of British Columbia, 1987 M.A. Linguistics, University of British Columbia, 1992  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in  THE FACULTY OF GRADUATE STUDIES  (Department of Linguistics)  We accept this thesis in confonning to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA December 20, 2000 © Susan Jane Blake, 2000  In presenting degree  this  at the  thesis  in  partial fulfilment  of  University of  British Columbia,  I agree  freely available for reference copying  of  department publication  this or of  and study.  thesis for scholarly by  this  his  or  her  Department The University of British Columbia Vancouver, Canada  DE-6 (2/88)  requirements that the  I further agree  purposes  representatives.  may be It  thesis for financial gain shall not  permission.  Date  the  is  that  an  advanced  Library shall make it  permission for extensive  granted  by the  understood be  for  that  allowed without  head  of  my  copying  or  my written  ABSTRACT The goals of this dissertation are two-fold. My first major goal is descriptive, to contribute to the documentation of the Central Coast Salish language spoken by the Shammon, Klahoose and Homalco peoples. The primary source of data is my field notes collected from 1988 through 2000 in consultation with Elders resident in Sliammon, B.C. My second major goal is theoretical, to deepen our understanding of the distribution and representation of schwa in Shammon. Schwa, often characterized as a brief "neutral" vowel with special properties cross-linguistically, is central to an understanding of Sliammon phonology and morphology. In Chapter 2, it is hypothesized that schwa is featureless, and that it acquires its surface realization via colouration from adjacent consonants and vowels. The focus of Chapter 3 is on the prosodic structure of the language, and the independent hypothesis that schwa is also weightless. These generalizations are presented within the Nuclear Moraic Model of Shaw (1993, 1996). Chapter 4 focuses on schwa-zero alternations. Schwa is proposed to be non-lexical, and schwa epenthesis satisfies the constraint Proper Headedness which requires that a Foot contains a syllable which is headed by a vocalic Nucleus. Schwa epenthesis also satisfies the ban on steminitial consonant clusters. This chapter also provides evidence that Full Vowel Reduction involves the loss of phonological weight (i.e. a mora). The output of Full Vowel Reduction is distinct from schwa in its featural representation, but identical to schwa in its prosodic representation. Although schwa epenthesis is driven by the constraints on Proper Headedness, there is also evidence that Shammon has a number of strategies which conspire to avoid schwa in stressed open syllables. Chapter 5 brings together different cases of this, and shows that they receive a unified explanation with reference to the constraint *9]o, and its interaction with other constraints. Chapter 6 addresses two cases in which schwa epenthesis is systematically avoided: (i) the possessive -hV suffix and (ii) the plural /L'-/ prefix.  Chapter 7 provides a synopsis of the  descriptive and theoretical claims of the dissertation. Appendices IV-VII contribute to the basic documentation of Sliammon including: Consonant Contrasts, Root List, Lexical Suffixes, and Affixes/Clitics.  ii  Table of Contents Abstract  ii  Table of Contents  iii  Acknowledgements  xii  Abbreviations and Symbols  xiv  Guide to Pronunciation Guide  xix  Chapter 1: Introduction 1.0 General Introduction: Sliammon Language and Culture  1  1.1 Goals of the Dissertation  4  1.2 Previous Scholarship on Sliammon language  7  1.3 Theoretical Framework  8  1.3.1 Representations adopted in this Dissertation  8  1.3.2 Theoretical Models of Phonology  8  1.4 Overview of the Dissertation  8  Chapter 2: Introduction to Sliammon Sound System 2.0 Introduction.  14  2.1 Sliammon Sound System  14  2.2 Consonant System  15  2.2.1 Inventory  15  2.2.2 Consonant Allophones I  16 Stops and Affricates  16 Affrication of Stops  16 Aspiration  17 Palatal off-glide on Alveopalatals and plain velars  18 Secondary Labialization  20 Fricatives  21 iii Allophones of 0  21 Allophones of s  23 Allophones of x  24  w h ~ x in the environment of IvJ  25  w Sonorant Obstruents /j, j, g, g/  25 Pre-nasalization  25 Retracted IGI  26 Resonants  27 Nasals ~ Voiced Stops  27 Glottalized Resonants  27 Retraction of Coronals  28  2.2.3 Consonant Allophones II Labialization C° versus C  30 30  w Neutralization of contrast: C /u w Glottalization and Phonology of Laryngeals  31 32 Allophones of ?  32 Proposal  34 Glide Vocalization  35 Glottalization  38 Deglottalization  39 Glottal Restructuring  42 Floating feature: constricted glottis [cgl]  43 Epenthetic [h]  45  2.2.4 Obstruent/Glide/Vowel Alternations: /j, j, g, g/  46  2.2.5 Laterals/L, LV  49  2.2.6 Geminates  51  2.2.7 Consonant Deletion  51 Identical Consonants  51 Coronal Deletion  52 iv t-deletion  52 n-deletion  52 f-deletion  53  2.3 Theoretical Assumptions  54  2.3.1 Features  54  2.3.2 Feature Geometry  54 Representation of Sliammon Consonants  55 Labials  55 Coronals and Laterals  56 Alveopalatals  57 Velars  58 Post-velars: Uvulars and Laryngeals  61  2.4 Vowel System.  61  2.4.1 Surface Vowel Inventory  65  2.4.2 Vowel Quantity  65 Durational Contrast  65 Theoretical Assumptions  66 Representation of Durational Contrast: Nuclear Moraic Model 67 2.4.3 Full Vowels  68 Surface Realization of the Full Vowels  68 Retraction  68 Place Assimilation  74 Phonetic effect of Anterior Consonants  75 Interaction of Retraction and Place Assimilation  76 Vowel / Consonant Interaction Proposed Analysis  78 80  2.4.4 Schwa Colouration.  85 Schwa and Retraction.  87 Schwa and Labialization.  89  v Schwa and Place Assimilation.  90 Interaction of Retraction and Place Assimilation.  91 Trans laryngeal Harmony.  94 Summary of Allophones of Schwa.  95 Proposed Analysis.  96  2.4.5 Reduced Full Vowels.  97 Full Vowel Reduction (laxing) * Reduction to Schwa 2.5 Summary  99 100  Chapter 3: Prosodic Structure of Sliammon 3.0 Introduction  104  3.1 Moraic Structure.  106  3.1.1 Coda consonants are moraic  107 Compensatory Lengthening.  107 Stress Assignment and Vowel Reduction  110 Stative-it  111 Speakers'Judgements  112  3.1.2 Moraic status of Vowels  114 Long Vowels are bimoraic  114 Full Vowels are moraic  115 Schwa is non-moraic  117  3.1.3 Implications: CC [CaC] Roots and Minimality  117 The Problem  117 Discussion and Proposed Analysis  121  3.1.4 Summary  121  3.2 Syllable Structure  123  3.2.1 Simple Syllables  124 Onset  125 The Data  125 Proposed Analysis  126 vi Gemination: Onsets and Faithfulness to Moraic Structure 128 Nucleus  129 Syllables have Nuclei No Coda  131 131 The Data  132 Proposed Analysis: NoCoda  133  3.2.2 Complex Syllable-Internal Constituents *Complex Onset  134 134 The Data  134 Proposed Analysis.  137 Comparative Evidence for lack of Complex Onsets  138 Sliammon / Sechelt Data  138 Sliammon / hanqaminarh (Musqueam) Data  138 Proposed Analysis  140 Root Initial Consonant Clusters  141 Sliammon / hanqaminam (Musqueam) Data 141 Proposed Analysis  142 Loan Words: Evidence for lack of Complex Onsets  144 Apparent Exceptions  145 Initial sC sequences  145 Clitic Initial Constructions  148 Discussion and Proposed Analysis  150 Word-Internal Complex Onsets  150 The Problem  150 Proposed Analysis  151 Establishing Maximal Syllable Size: *Complex Coda  155 The Data  155 Proposed Analysis  158 Diminutive Reduplication and *ppp]a  159 The Data  159 vii Asymmetry in Word-initial and Word-final Consonant Clusters Minor Syllables in Sliammon. 3.2.3 Summary  162 164 167  3.3 Introduction to Sliammon Metrical Structure  168  3.3.1 Basic Observation: Primary Stress is Leftmost  168  3.3.2 Foot Form: Trochaic  170  3.3.3 Stressed Schwa  172  3.3.4 Location of Secondary Stresses  173  3.3.5 Lexical Suffixes and Stress Assignment  174  3.4 Summary  178  Chapter 4: Distribution of Schwa in Sliammon 4.0 Introduction  179  4.1 Excrescent versus Epenthetic Schwa  179  4.1.1 The Problem  180  4.1.2 Evidence from Syllabification  181  4.1.3 Discussion and Proposed Analysis  184  4.1.4 Further Implications: Echo Vowels  185  4.2 Distribution of Epenthetic Schwa  188  4.2.1 Some schwas are epenthetic  192  4.2.2 Schwa/Zero Alternations  196  4.2.3 Proposed Analysis  199  4.3 Full Vowel Reduction  201  4.3.1 The Problem  201  4.3.2 Full Vowel Reduction and Foot Structure  202 Bisyllabic Trimoraic Feet  203 Bisyllabic Bimoraic Feet  206 Representation of Full Vowel .Reduction  209 Contexts in which Full Vowel Reduction is blocked  210  4.3.3 Implications of Proposed Analysis  211 viii Imperfective Reduplication and Full Vowel Reduction  211 Vowel Height Assimilation is Independent of V Reduction  213 Summary of Full Vowel Reduction  214  4.4 Strengthening of Schwa: Schwa/Full Vowel Alternations  215  4.4.1 The Data  216  4.4.2 Discussion and Proposed Analysis  220  4.4.3 Implications of Strengthening  222  4.4.4 Theoretical Implications: No Long Schwa  223  4.5 Stable Schwa  226  4.6 Summary  228  4.7 Summary of OT constraints  229  4.7.1. Constraints  229  4.7.2. Effects of Constraint Ranking  230  Chapter 5: Constraints on the Distribution of Schwa in Sliammon 5.0 Introduction  231  5.1 Surface realization of Glottalized Obstruents  233  5.1.1 The Problem  233  5.1.2 Discussion and Proposed Analysis: *6]  a  5.2 Surface realization of Glottalized Resonants  240  5.2.1 Glottal Restructuring 5.2.2 Proposed Analysis: * 9 ]  238  240 245  0  5.3 Geminate consonants  246  5.3.1 Geminate Resonants  246  5.3.2 Geminate Obstruents  247  5.4 h-epenthesis  248  5.4.1 Root-LS: h epenthesis  249  5.4.2 Lack of [h] epenthesis after C-final Roots  250  5.4.3 Apparent Exceptions: CsC Roots  251  5.4.4 Proposed Analysis: *s]  252 ix  5.5 Implications: Imperfective Reduplication  254  5.5.1 Imperfective Reduplication: Strong Roots  254  5.5.2 Imperfective Reduplication: Weak Roots  255  5.5.3 Proposed Analysis: * 9 ]  257  a  5.6 Summary: *6]  258  5.7 Schwa in a stressed open syllable  259  a  5.7.1 The Problem  259  5.7.2 Proposed Analysis  259  5.8 Formal Characterization of *9]  260  0  Chapter 6: Implications and Conclusions  6.0 Introduction  262  6.1 Epenthesis of the Full Vowel [i] and the Possessive Affix -hV  262  6.1.1 Translaryngeal Harmony and the Possessive Affix  262 Translaryngeal Harmony  263 Discussion and Proposed Analysis  266  6.1.2 Position of the Possessive Affix  266 V-final Stems  266 C-fmal Stems  267 Stems ending in a C-cluster: [i] epenthesis  267 Proposed Analysis of the Position of the Possessive Affix  269  6.2 Implications: Non-reduplicative prefixes  274  6.2.1 Lack of Non-reduplicative C-prefixes  274 Loss of lexical nominalizing prefix s  275 Discussion and Proposed Analysis  276  6.2.2 I V - I Plural  278 Historical and Comparative Evidence  279 Synchronic Evidence for I V - I in Sliammon  281 Plural I V - I and Diminutive Reduplication  282 Plural I V - I and Diminutive Plural Reduplication  284  x Plural /L'-/and Imperfective Reduplication  285 Plural I V - I and Characteristic Reduplication  287 Discussion and Analysis  288  Chapter 7: Conclusion  7.0 Introduction  291  7.1 Summary: Representation of Schwa versus Full Vowels  291  7.2 Phonological Features  292  7.3 Prosodic Structure of Sliammon  292  7.4 Distribution of Schwa  293  7.5 Summary: Strategies to avoid stressed schwa in an open syllable  294  7.6 Status of unstressed schwa in open syllables  296  7.7 Other Implications  297  7.8 Topics for future research  298  References  300  Appendix I: Salish Language Classification and Map  313  Appendix II: Homalco, Klahoose, and Sliammon Language Consultants  315  Appendix III: Transcription Systems  318  Appendix IV: Consonant Contrasts  319  Appendix V: Sliammon Root List  373  Appendix VI: Sliammon Lexical Suffixes  392  Appendix VII: Predicate Complex and Affixes, Clitics and Particles  437  xi  Acknowledgements First I would like to extend my sincere gratitude to the following Homalco, Klahoose, and Sliammon Elders for sharing their knowledge of the language and culture with me: Mrs. Mary George, Mrs. Agnes McGee, Mrs. Phyllis Dominic, Mrs. Eva Hanson, Mrs. Marion Harry, the late Mr. Joe Mitchell, Mrs. Elsie Paul, Mrs. Sue Pielle, Mrs. Helen Hanson, Mr. Pete Harry, Mrs. Annie Dominick, and Mr. Dave Dominick.  I am grateful to the Sliammon Chief and  Council, the Sliammon Treaty Society, and community members for the opportunity to study their language. I would also like to express my deepest appreciation to Ms. Betty Wilson who, in collaboration with the Sliammon language working group, provided feedback on an earlier draft of this thesis. Special thanks to my research supervisor Dr. Patricia A. Shaw, also affectionately known as "Dr. Schwa", who has made this entire process a particularly rewarding intellectual experience. To committee members Dr. Douglas PuUeyblank and Dr. Paul D. Kroeber I also extend my sincere appreciation. Dr. Douglas PuUeyblank will be remembered for asking particularly challenging questions regarding the "theory", and is thanked for sparing me from a lengthy expose on the definition of "stem". Dr. Paul Kroeber, in addition to sharing his wealth of knowledge on Salish languages and Sliammon in particular, also passed on some valuable practical advise. He equates writing a dissertation with making pancakes. This idea is expressed in the following Russian proverb which was handed down to him from his father: pirvey blin komom "the first pancake is always lumpy" . I would also like to acknowledge questions and feedback from the other members of the examining committee present at my oral defense: Dr. Margery Fee, and Dr. Bryan Gick, (University Examiners), Dr. Anthony Dawson (Chair), and Dr. Sharon Hargus (External Examiner). I also appreciate questions and comments from Dr. Emmon Bach, and Dr. M . Dale Kinkade who first introduced me to the study of Salish languages and linguistics. I am grateful to my friend and colleague Dr. Honore Watanabe of Kagawa University for his careful commentary on the most challenging aspects of Sliammon morpho-phonology, and for sending cranes to mark each of the small steps along the way. I would also like to thank the other faculty and visiting scholars at University of British Columbia who have contributed to my studies: Mrs. Ingrid Brenzinger, Dr. G. Carden, Dr. Henry Davis, Dr. Rose-Marie Dechaine, Dr. Ewa Czaykowska-Higgins, Dr. Hamida Demirdache, Dr. xii  Laura Downing, Dr. Mark Hewitt, Dr. M . Dale Kinkade, Dr. Juta Kitching, Dr. Greg LaMontagne, Dr. Michael Rochemont. Thanks also to the Faculty and students of Girona Summer School of Linguistics (1996), le ragazze italiane, and to my many fellow graduate students at UBC, especially Darin Howe, Taylor Roberts, and Kimary Shahin. To Carmen de Silva and Edna Dharmaratne my most sincere appreciation for everything you have done. To my Musqueam friends and colleagues, and students and staff of the Musqueam/UBC First Nations Languages Program - you have all brought something very special to my life : hay ce:p qa si:?em ! I owe a special thanks to my family Marion, Charles and Margaret Blake for their continued moral support and encouragement.  It was my parents who first took me to Savary  Island as a child and introduced me to the traditional territory of the Sliammon people.  With my  sister Margaret, I treasure these early childhood memories of the sand dunes, the water, and the enchanted forest. To my closest friends, each and every one of you has played a special role: Harry and Patricia Bekke (and the kids: Lynx and Mousette); Luca Burroni, Barry and Maureen Clare and family; famiglia Curini-Galletti: Fabio, Lapo, Lucia and Cosimo; Aldona Dyk, Elena Fiume and Ron Yaworsky, daughters Alessia and Katrina; Jane B. Fuller, Caroline Mitchell, Kathleen and Lauren Stone, Monica Sanchez, Stephanie Trenchiansky,  Lance Woolard, and the Sons of  Joseph Zawenou. Thanks also to Sonya Usmiani, Dr. Tracey Simmons MacKinlay, the people of Lund and Savary Island, and especially to Willa Lane. I am also deeply grateful to Shawn David Clare who has managed "linguistics central" for the past 8 months with a sense of humour and levity which has had a profound effect on our collective sense of well-being. For you, pair-o-dice forever ! My doctoral research has been generously supported by the following funding organizations: University of British Columbia, Graduate Fellowship 1994 -1995; Social Sciences and Humanities Research Council of Canada, Doctoral Fellowship  #752-96-1924; M and E.  Jacobs Research Fund (1996) for research on 'Sliammon Metrical Structure', and the University of British Columbia Teaching and Learning Enhancement Fund Grant ("Cedar-to-CDRom") awarded to Dr. Patricia A. Shaw.  xiii  Abbreviations and Symbols Abbreviations a syllable boundary 1  first person  2  second person  3  third person  Act  activity suffix  Aff  Affix  Aux  auxiliary  C  any consonant  Caus  causative  cgl  constricted glottis  CH  Characteristic Reduplication  CJargon  Chinook Jargon  cl  compound ligature  CL  Compensatory Lengthening  COR  coronal  CTr  control transitive  CVCPL  CVC Plural Reduplication  CVPL  CV Plural Reduplication  C9CPL  CoC Plural Reduplication  dem  demonstrative  det  determiner  Dim  diminutive  DIM  Diminutive Reduplication  Dimgl  diminutive glottalization  DOR  dorsal xiv  Emph  emphatic  e.o.  each other  Erg  ergative (subject of a transitive predicate)  ex  extension  fern  feminine  Fr.  French  Ft  foot  Fut  future  gen  general  H  high pitch  I.  Indian  Imp  Imperfective  IMP  Imperfective Reduplication  Imper  Imperative  INC  Inchoative Reduplication  Ind  ^directive  Indef  mdefinite  Indep  independent pronominal  Instr  instrumental  Intr  intransitive  K  any obstruent  L  low pitch  LAB  labial  link  linking vowel  LS  Lexical Suffix  lit  literally  M  mid pitch  masc  masculine XV  neg  negative  Nom  nominalizer  Nuc  nucleus  NTr  noncontroi transitive  0  obstruent  Obj  object  Obi  oblique  OT  Optimality Theory  Pass  passive  past  past tense marker  PHAR  pharyngeal  PL  Plural  PMC  main clause Passive  Po  possessive  PSC  subordinate clause Passive  PrWd  prosodic word  ptc  particle  Q  interrogative, question marker  quot  quotative  R  any resonant  recip  reciprocal  RED  reduplicant  Rflx  reflexive  Rt  Root  Sb  subordinate  sg  singular  s.o.  someone  s.t.  something xvi  stv  stative  Su  subject  Tr  transitivizer  V  any full vowel (here Ii, u, al)  Sound Symbols  [*]  voiceless ejective lateral affricate  [S]  voiceless alveopalatal affricate  [t ]  voiceless interdental affricate  [y]  voiced palatal glide (resonant)  [«]  voiceless alveopalatal affricate  [f ]  voiceless ejective interdental affricate  [*]  voiceless lateral fricative  DI  voiced alveopalatal affricate  [i]  "darkened" lateral resonant (effect of post-velars)  [V]  voiceless labio-velar glide (resonant)  e  6  o c  palatalized consonant  9  schwa  V  primary stress  V  secondary stress  :  vowel length  h  aspirated consonant  xvii  C. Miscellaneous Abbreviatory Devices ~  alternates with. Used with variant output forms  !  fatal constraint violation  *  word boundary  ()  encloses optional element  (...)  marks foot boundary  *  'is not', i.e., in the constraint *Ci  *  constraint violation  *  ungrammatical morpheme boundary  //  underlying representation  =  introduces a Lexical Suffix  »  is more highly ranked than  [...]  infix in input representation  [... ]  phonetic form / output candidate marks the winning candidate output form  p  mora  V  Root  a  syllable  xviii  Pronunciation Guide Consonants /p/ voiceless bilabial stop as in English [pmc] pinch, [piknik] picnic, [pampkin] pumpkin. as in Sliammon [pa?a] one, [?dpAn] ten, [x dpx op] hummingbird. w  w  /p7 ejective bilabial stop as in Sliammon [pex ] flood, [poho] raven, [t^apaw] busy, [xap] crad7e basket. w  /t / interdental affricate 6  as in Sliammon [Tettan t 9m] Tm going to eat, [?3t na? nux ii] my canoe. e  w  e  /f/ ejective interdental affricate as in Sliammon [t^amq r] cloud, [f^orcis] seven, [0at'3m] spring salmon, w<  e  [manatee?] to beat a drum (in the traditional way), [ V A ^ ] bay. 6  I\l voiceless alveolar stop as in English [tent] tent, [tawn] town, [tipi] tepee. as in Sliammon [ti?ta ~ ti:ta] that one, [setqen] upper lip, [tihmot ~ ti:mot] really big.  1(1 ejective alveolar stop as in Sliammon [t'Xlek ~ t'/Chk] a hole, [qat'an ~ qafen] rat, [ta?gAf!] herring.  PfU voiceless lateral affricate (IPA ft); also informally referred to as "Charles Alias" as in Sliammon [Xapx ] broke, [XdihXsm] wer, [XmXan] real shy, [xaX] break (rope). w  xix  It! ejective lateral affricate (IPA t'i); also informally referred to as "Running Man" as in Sliammon [Xaiam] salt, [ztafsam] strong, [3Um£Am] square, [x eXay] mountain goat, [xat] to want, to desire, [n\At\ calm (water). w  Icl voiceless alveo-palatal affricate (IPA tj) as in English [cest] chest, [caepal ~ caepl] chapel. as in Sliammon [ciciye?] grandmother, [picu ~ pecu] basket, [hoc] I'm going.  Ibl ejective alveo-palatal affricate (IPA t'J) as in Sliammon [cs?no] dog, [CAX] ripe, cooked, [hi] rain, [cictiyax] sandpiper, [co?co?] wren, [&\q] robin, [sa?ci] tanned leather, [ x e & c Y ] autumn, fall, [pa?a6] net, [6ic] straight.  Ikl voiceless velar stop as in English [kelp] kelp, [kicon] kitchen, [kasc] catch. as in Shammon [kepu ~ k^po] coat, [kiks] cookie, [ki:ke?] bug, [lakle ~ Iflde] key, [p30k] bullhead (fish).  Ikl ejective velar stop as in Sliammon [iekce?] small basket for sewing; junk box, [wawafcila] limpets, [ki:iut ~ ii:ket] crow, [kew&egim] coyote.  Ik l voiceless labio-velar stop w  as in Sliammon [k a:nacim] sir down, [k assm] ruffed grouse, [£ UA1C U] salt water, w  w  W  W  [k uma?] ratfish, [senk u ~ senk o] ocean, [m90k ] blackcap berry, [qayk ^ bald w  w  w  w  1  eagle. Similar to English [kwest] quest, and [kwilt] quilt , except that Sliammon k functions as a w  single sound rather than a sequence of two sounds, as in English [kw]. xx  /k / ejective labio-velar stop w  as in Sliammon [i a?] sliced salmon, [k dta ~ k ut'a] barbecuing stick, [i a?wa] belly, w  w  w  w  stomach (external), [xayxele us] nightmare, [tA& tu]c tas] s/he is pulling it, w  w  w  [ju:lc vt] smash it up, [sesk im ~ sesk tm] shaking, [k as] hot. w  w  w  w  Iql voiceless uvular stop as in Sliammon [qax] many, [qayx] Mink (stage name), [moqsin ~ mXqsen] nose, [taqa] saiai berry, [§3q0ot] sigh, [iAxqzn] raspy throat, [pAq] white, [tKq] out.  Iql ejective uvular stop as in Sliammon [qat'an ~ qat'an] rat, [qa?3t| sea otter, [gaqet ~ garqet] it's opened, [s/qt] peel off s.t. (i.e. cherry bark); to split s.t, [tAq]  rot, [sXq] fifty cent piece, half  /q / voiceless labio-uvular stop w  as in Sliammon [q Al' ~ q af] he/she/they came, [fdq mot] saps running, w  w  w  [q^q^les] copper, [t'e?neq ] salmonberries, [maseq ] sea urchin, [?toq ] hard. w  w  w  Iq l ejective labio-uvular stop w  as in Sliammon [q et] beach, [q ex ] ashes, [q Xlq abs] raccoons, [ta?q a] devil fish, w  octopus,  [toq ] w  w  w  w  w  w  clear skies, [poq ] grey, brown, [X o?q ] sawbill duck, [t' dt' oq ] w  w  w  e  e  w  feather. /?/ glottal stop / laryngeal stop as in English [?dpan] open, [?o ?o] oh oh!, [?aepsl ~ ?aepl] apple. as in Sliammon [?asx ] seal [Veitan] eat, food, [?e:?] yes, [x a?] no, w  w  [x^ax^ne?] tidepool sculpin, bullhead, minoe, [xa?a ~ xa'a] butter clam, [ne:?et] be in t way, [sa?a] rwo, [sd?ot] choose it, [Terje] tree, relative.  xxi  /& voiceless interdental fricative as in English [0m] thin, [pze0] path. as in Sliammon  [?ay?aju03m]  language of our people; speak well, [0d6en] mouth,  [6e?0a] that one (fern.), [ma6k ] blackcap berry, [wal0] bullfrog, [qegA0] deer. w  Isl voiceless alveolar fricative as in English [silk] silk, [skin] skin, [smelts] smelts. as in Shammon [saftx ~ sattw] woman, [qasqps] fired of sitting, [Tines] tooth, w  [?a?bs] sea cucumber.  /if voiceless lateral fricative as in Sliammon [ta?amin] Sliammon people, [IAX] bad, [Xaiam] salt, [tal'ftom] wolf, [q a?wii ~ q a?wet] pitch, chewing gum, [pusiai] Adam's apple. w  w  Isl voiceless alveo-palatal fricative as in English [saeg] shag (cormorant), [sel] shell, [sip] ship. as in Shammon [se?] climb, go up,  [sqAihs]  his/her partner, [sim] dry, [saqGot] sigh,  [jisin] foot, [t'isaq ] snot, [tiiunis ~ tiimis] man, [miismus] cow, [k isk is] Steller's w  w  w  jayfx*] voiceless palatalized velar fricative [x a? lowlax m] spiked heel, high heels w  3  y  Rare in occurrence and not well documented. This sound occurs as a variant of Isl.  /x / voiceless labio-velar fricative w  as in Shammon [x a?] no, [x uXa?ju] trout, [niix if ~ ndx el] canoe, [tix 6a't] tongue, w  w  w  [lax™] falling snow, [?a?ax ] it's snowing. w  xxii  w  w  [w] voiceless labio-velar glide — variant o f /x / w  as in Sliammon  parjEcx^ot  ~ ?a?jecw?ot] how are you (sg)?, [?asx ~ ?asw] seal, w  [Tsnx ~ jenw] fish, [t^dmaTjuw ~ t^dmaTjux™] barnacle. w  Ixl voiceless uvular fricative as in Sliammon [xa?a] butter clam, [Xaws] new, [xawgas] grizzly bear, [xawsin ~ xawsin] bone, [sojcam] racing canoe, [y/Cxt] rib, [xexyeq] crab, [qaX] many, [^a^atsx] grasshopper.  /x / voiceless labio-uvular fricative w  as in Sliammon [X elfcay] mountain goat, [X eyt] stretch it, [x^ax^nE?] tidepool w  w  sculpin, bullhead, minoe, [t' dx o] ling cod, e  w  [pEx ] w  flood,fax™]to melt.  IhJ voiceless laryngeal / glottal fricative as in English  [hEmlak]  hemlock,  [hElp]  help,  [helo]  hello.  as in Sliammon [ha? mo] pigeon, [h£'?gm] strawberry, [hanaq os] wolf eel, aggressive, 3  w  [hahasamc] Tm sneezing, [hdmhom] blue grouse, [hEhEgAi]  [hEhawcis]  paddling,  for the first time, [tih ~ ti:] big, large, [?ah] sore, pain,  [qAht] to  lift up.  Iml bilabial nasal as in English [memo] memo, [mslt] melt, [msesk] mask, [maegnit] magnet. as in Sliammon [man ~ mAn] father,  [mA'gAJ  cougar, [cux men] a week, [nam?Am] to w  write, [tarn] what?  Iml laryngealized bilabial nasal as in Sliammon hamu [haPmo] pigeon, sama [sa?ma] mussel, [stmsim] it's already dried, tarn [tarn?] belt, [idtmoih] littleneck clam, x aX ayim [x ax ayim?] housefly. w  xxiii  w  w  w  ltd alveolar nasal as in English [not] note, [natikal] nautical. as in Sliammon [nigin] lunch, [niginaye] lunch basket, [nat ~ nAt] night, [jinis] foori, [x a:stAn] suet, [tan ~ ta:n] mother. w  Ihl laryngealized / glottalized alveolar nasal as in Shammon Xina [X£?na] oolichan oil, iaim [tzlno] dog, [tan ~ tan?] that one.  IV lateral liquid as in English [lek] lake, [lift] HA. as in Shammon [lastpol ~ laspol] soccer ball, [lik le ~ lik le] key, 9  3  [laplas] plank, long  board, [CE 1AS] three, [?elqay] barbecued deer meat, [?atnopel] car, automobile. 9  It/ laryngealized/ glottalized lateral as in Shammon ?afas [?a?bs] sea cucumber, [q 3l] to come, [paf ~ pal?] heron, crane. w  lyl alveo-palatal glide as in English [yes] yes, [yel] yell, [yoga] yoga, as in Shammon [yXxay ~ yexay] berry basket, [yetat] to call s.o.,  [ye?gay] i n n e r cedar  bark, [yeqet] disgusted with it, need it, [pi:paye ~ pe:paye] one person, [yiyqet] easy, inexpensive, [0a?qay ~  0a?qAy]  sockeye salmon.  lyl laryngealized/ glottalized alveo-palatal glide as in Shammon qaya [qaTye] water, sayja [saryje ~ sayje] leaf, [cixuy] children, [>tA7vXAy]  elder, [si:say ~ saysay] be scared.  xxiv  Ijl (voiced) alveo-palatal affricate (non-continuant resonant) as in English [JElifts] jellyfish. as in Sliammon p% [JIAVJ run, [jiceys] spear for cod and cod eggs, (jinjines] teeth, [SJESOI]  yesterday.  /w/ labio-velar gUde as in English [wasp] wasp, [was] wash as in Sliammon [wal0] bullfrog, [WA'XAS] frog, [paw?us] one dollar, [qaw6] potatoes, [yi?gaw] partically dried, dried up, [f/xaw] gone bad.  /w/ laryngeaUzed/glottalized labio-velar glide as in SUammon k awa [k a?wa] belly, stomach (external), [xewxa?gAt] chipmunk, w  w  [wiwlos] young man,  [me:mAw] cat,  [taw  ~ tu?]  ice.  IgJ (voiced) velar stop (non-continuant resonant) as in English [gem] game, [gost] ghost. as in SUammon [gaqet] ir's opened, [gfje] earth, land, [tigy9x ] nine, w  [pa?agii] one boat, [Gdga, Gdgi, hdga] go,  Vowels [i] high non-back tense oral vowel as in English [ski] ski, [pitsa] pizza, [siisi] sushi, [piasno] piano. as in Sliammon [tix 0ai] tongue, [xe&£] fall, [k isk is] Steller's jay, [sk ici] bothersome, w  w  w  pesty.  [i] high non-back lax oral vowel as in English [fts] fish, [swim] swim, [kmsip] kinship. as in SUammon [jinis] tooth, \jit\ run, [hi] rain, [616] straight. xxv  w  [e] mid non-back tense oral vowel as in English [gem] game, [snek] snake, [snel] snail. as in Shammon [qegA©] deer, [q eyq ay] sand, [senk u ~ senk o] ocean. w  w  w  w  [e] mid non-back lax oral vowel as in English [net] net, [kelp] kelp, (Jet] jet, [sel] shell. as in Shammon [?eit9n] eat, food, [kekce?] small basket for sewing, junk box, [qeqte?] youngest in family, [hewqen] swan, [k en] how many, [xexyeq] crab. w  [u] high back rounded tense oral vowel as in English [flut] flute, [glu] glue, [flu] flu, [pluto] Pluto, [siimask] sumac. as in Sliammon [k usem] green, blue, [musmus] cow, [ ruk ] to fly, w  <  w  [k uk pa?] grandfather, [wuk^ scoop net. w  w  [v] high back rounded lax oral vowel as in English [hvk] hook, [put] put as in Shammon [k v0ays] island, [k unvuk t] red blanket, [tvk t] pull it. w  w  w  w  [o] mid back rounded tense oral vowel as in English [foks] folks, [most] most, [pok] poke, [smok] smoke, [totem] totem. as in Shammon [PdFoq*] feafner, [toq^ hard, [4oq ] clear skies, [ibq toq ] oysters, w  w  w  [tdq mot] saps running, [hdmhom] blue grouse, [t' dx o] ling cod, w  e  w  [idtmom] littleneck clam, [k 6?ox ] smoked salmon. w  w  [o ~ a] mid back rounded lax oral vowel similar to the [o] in English [horn] horn , as in Shammon [x o?q ] sawbill duck, [Toq ioq ] warm water, [sjesot] yesterday, w  w  w  [?amam5?] chiton, [k dnot'] porpoise. w  xxvi  w  [a?] low non-back tense oral vowel as in English [laemp] lamp, [gasf] gaff, [klaem] clam. as in SUammon [kaepo] coat, [ce^aet'an] mouse, [kasmputs] rubber boots, v  [k ina?cx ] you carry it. w  w  [a] low central oral vowel as in SUammon [iik A ac] gunny sack. w  i  [a] low back oral vowel as in English [swan] swan, [swamp] swamp, [pat] pot, [kad] cod. as in Sliammon [sd?a] fwo, [?asx ] seal, [k dx a?] box, [t'dqt'aq] slow. w  w  w  [s] mid central lax oral vowel as in English [ago] ago, [ftfkas] focus . as in SUammon [tatemcx ] what are you doing? w  [A] low-mid back oral vowel as in English [ b A t ] but. as in Sliammon [WAXAS] frog, [sdpnAc] tail, [CAX] ripe, cooked, [qAit] many.  Diphthongs [iy~i:] as in Sliammon [tiymot ~ ti:mot] really big< /tih-mut/, [ey] as in SUammon (jiceys] spear for cod and cod eggs, [X eyt] stretch it. w  [uy] as in SUammon [need example]. xxvii  [uy]  as in Sliammon [cuy] child, [ci:cuy] children,  [ay] as in Sliammon [qayk ] eagle, w  [pAgay]  [ay ~  A  y  [qayx]  Mink (stage name), [x e?tay] mountain goat, w  halibut, [y/xay ~ yexay] berry basket, [pa?ay£] appendix.  ]  as in Sliammon /say=ana/ [s Aye?na] neck,  [ y e ? g A y ]  inner cedar bark.  [ A y ]  as in Sliammon taXxay [tKtxAy] elder,  [iw]  as in Sliammon [wiwlos] young man.  [ew]  as in Sliammon [xXXews] pluck a bird, feather a bird, [pewi] lard, rendered-down fat  [ew]  as in Sliammon [ x e w x a ? g A i ] chipmunk  [ew]  as in Sliammon [t^ewq] red elderberry  [uw]  as in Sliammon [yuw ~ yaw] it's been raining, dried up  [ow]  as in Sliammon [x al'owlaxyin ~ x a? lowlaxyin] spiked heels, high heels.  [aw]  as in Sliammon [paw?us] one dollar, [xawgas] grizzly bear, [qawG] potato,  w  [t^apaw]  w  3  busy  [aw]  as in Sliammon [xaws] new.  [aw]  as in Sliammon [?aw?awak ] 2ors of tobacco w  xxviii  [AW]  as in Sliammon mimaw [me:mAw] cat  [aw]  as in Sliammon [taw ~ tu?] ice.  [oy]  as in Sliammon [^aq o:ys] summer. w  Surface Long Vowels [i:]  as in Sliammon [krkeekik] blackbkd, [ki:kik ~ Ri:kek] crow, [ti:tolk um] small roots, w  \jr.t  cep]  you (pi) are running, [ti:ta ~ ti?ta] that one,  [QiiQa ~ ei?0a  ~ 0E?0a]  that one  (fern.), [ti:6os ~ tvtvs] Spring, [ki:ke?] bug, [ci:cuy] children, [ttcim]combhair, [si:say ~ saysay] scared. [e:]  as in Sliammon [he:ynAC ~ hiynAc] bottom of a basket, [pe:paye] one person, [ne:?et] be in the way,  [E:]  as  in  [me:mAw] cat.  Sliammon [q£:s£t] to tie it, [xE:xnAq ~  xE?ExnEq]  Owl's Grove (sacredplace),  |j£:q am] sweat, [q E:tam] front of the house (faces the beach q et), [?£:?] yes w  [u:]  w  w  as in Sliammon [tu:k umm ~ tu:k um£n] bailer, [ju:k ut] smash it up, w  w  w  [Tu:k ~ ju?k ] Indian rice, [su:k am] outer cedar bark, [Ju:pays] step over. w  w  w  [o:]  as in Sliammon [x d:q £t] s.o. snoring (in that state), [md:la ~  [a:]  as in Sliammon [ p i p E m c ] I'm working (rightnow), [A^fawvm] any berry,  w  w  md:lA]  mill.  [?a:y£?] house. [a:]  as in Sliammon [q d:q uq] woodricJc, [k d:na?cim] sit down, [x a:stAn] suet, /?ah-sx c7 w  w  w  w  [?d:sx c] Tm hurt (cf. ?ah be hurt, sore). w  xxix  w  El M a r Necesito del mar porque me ensena: no se si aprendo miisica o conciencia: no sS si es ola sola o serprofimdo o solo ronca voz o deslumbrante suposicidn de peces y navios. El hecho es que hasta cuando estoy dormido de algun modo magnetico circulo en la universidad del oleaje.  Pablo Neruda  Chapter 1: Introduction 1.0 General Introduction Sliammon is a Central Coast Salish language spoken just north of Powell River on the Malaspina Peninsula at Shammon, British Columbia, Canada (cf. Appendix I on Salish Languages, as well as Czaykowska-Higgins and Kinkade (1998)). The term "Shammon" is used here in this thesis as a cover term to refer to the language of the Homalco, Klahoose, and Shammon people currently living in Shammon, B.C. The language is currently spoken by approximately 40 of the 800 residents of Sliammon. According to a number of elders that I have consulted, the Homalco, Klahoose and Shammon consider themselves "one people with one language". Traditionally, the Shammon, Klahoose and Homalco people utilized a vast area of land and waterways for traditional harvesting of annual salmon runs, berry-picking, and hunting. Their traditional territory extends along both sides of the northern Strait of Georgia from Malaspina Strait, in the south, to Desolation Sound and Homfray Channel, to the head of both Toba and Bute Inlets in the north, and the islands between the Mainland and Vancouver Island. These islands include Texada, Harwood, Savary, Hernando, Mitlenatch, Marina, Cortes, West Redonda, East Redonda, Sonora, Stewart and Dent Islands, amongst others (cf. for example Barnett 1955; Kennedy and Bouchard 1983; Shammon Treaty Society). Today, many of the Homalco, Klahoose and Shammon people live at Shammon, B.C., located just north of the city of Powell River. There is a community at Squirrel Cove on Cortes Island, and many of the Homalco people moved to Homalco Reserve located on Vancouver Island (near Campbell River), when the last families left the Church House village site.  1  The name Shammon is an anglicization of /ta?amin/ [ia?amtn], a term which refers to the Shammon people. The word /t'isus-m/ [tesosam] is the place name which refers to Sliammon Bay, Shammon Creek, and to the Shammon village which is located near the mouth of the creek. One elder explained that this place name is clearly related to the term /DIM-t'isus/ [tetsos] which refers to 'a small saltwater fish' which occupies the large tidepool infrontof Sliammon. The language has also been referred to as Mainland Comox or Comox in previous ethnographic and linguistic research which was intended to indicate the dialect spoken by the Homalco, Klahoose, and Shammon peoples living on the Mainland in contrast to the Island Comox dialect, originally spoken on Vancouver Island. "Mainland Comox" is a designation which is dispreferred by current speakers of the language since they associate "Comox" with the name of the town of Comox, B.C. which is located in what is now Kwakwakawa'kw or Kwakwalaspeaking territory. Some people have referred to the language as [?ay?aju69m]. However, as one Sliammon elder explained [?ay?aju09m] means to 'speak well' (from the root /?J/ [?i? ~ ?i:] good ); so, it could be used to describe someone who speaks English well, or French well, or any other language, and is not restricted in its use to mean 'the language of the Homalco/Klahoose/Shammon people'. In short, different people have expressed varying opinions regarding the appropriate use and meaning of this word. There are also separate terms which combine a place name or the name of people living there with the lexical suffix =qin for 'mouth, language' to indicate the local variety of the language spoken by a person from that particular location: [ia?aminq£n] /fa?amin=qin/ 'Sliammon language', [td?q q£n] /tu?q=qin/ 'Klahoose/Squirrel Cove language', and [?d?p qen] /?u?p=qin/ h  'Homalco/Church House language'.  h  However, there also is some discussion regarding the  grammaticality of words with the addition of =qin meaning 'language of x', and therefore without consensus on the part of the speakers, it would not be appropriate to use this type of construction as a designation for the language. As researchers and visitors, we are looking to the Homalco, Klahoose, and Shammon people for guidance regarding an appropriate name for their language. In the meantime, in this present work I will continue to use the name "Shammon" as a cover term 2  which includes speakers of Homalco, Klahoose, and Sliammon living at SUammon, B.C. The community has plans to discuss an appropriate name for the language by consulting with Elders from all three bands: SUammon Band, Klahoose Band (Squirrel Cove), and Homalco Band in Campbell River. There seem to be very few differences between speakers which can be clearly attributed to "dialect" differences. There are a few lexical items which seem to have restricted distribution or specific pronunciation of individual lexical items, such as /janx / |Jenx ] 'salmon, fish' versus w  w  /JanX / jjen3C ] 'salmon, fish'; note that a SUammon consultant provided this word with a final w  w  velar fricative /x / whereas a Klahoose speaker systematically used the corresponding uvular w  fricative /x /. Nonetheless, these differences seem to be very limited. w  There may also be a  differences in the rate of speech which are readily perceived by native speakers of the language; these latter are not documented in the present work. There are concerted efforts on the part of the Sliammon, Homalco and Klahoose people to revitalize the use of their language. The SUammon Language Program in the Powell River School system (accredited by the Ministry of Education) teaches children from kindergarten to grade 12, the language spoken by the Homalco, Klahoose and SUammon people.  The emphasis of the  program is on spoken language and oral fluency in its cultural context. The primary source of the data in this dissertation is my fieldnotes collected from 1988 through 2000 in consultation with SUammon elders, especially: Mrs. Mary George, Mrs. Agnes McGee, Mrs. Phyllis Dominic, and Mrs. Eva Hanson. Additional data were collected and/or verified with the following elders and/or speakers: Mrs. Annie Dominick, Mr. Dave Dominick, Mrs. Helen Hanson, Mrs. Marion Harry, Mr. Pete Harry, the late Mr. Joe Mitchell, Mrs. Elsie Paul, Mrs. Sue Pielle, and Ms. Betty Wilson. The collaborative Sliammon/UBC orthography workshops held in SUammon during the summers of 1996-1998 also provided a forum for rechecking additional data. To these many elders who patiently contributed their expertise, I am deeply grateful. Although the majority of my fieldwork was carried out at Sliammon, I also benefited tremendously from working with two SUammon elders who were residing/visiting Vancouver, and therefore provided the opportunity of meeting on a weekly basis. 3  In addition to tape recordings, there are two video recordings which were filmed in collaboration with the SUammon community under the auspices of the U B C Teaching and Learning Enhancement Fund (TLEF), awarded to Dr. Patricia A . Shaw. This video film footage is to be incorporated into multimedia curriculum materials to enhance the community-based teaching of the SUammon language in traditional cultural contexts. The collaboration and team work involved in these sessions has been invaluable. 1.1 Goals of the Dissertation The goals of this dissertation are two-fold. M y first major descriptive goal is to contribute to the basic documentation of the language spoken by the Sliammon, Klahoose and Homalco peoples. It is my hope that the language data contained within this dissertation will be useful to the SUammon community and will help further their efforts in language education. M y second major goal is to present an analysis of the distribution and representation of schwa in Sliammon. This is of both descriptive and theoretical interest. 1  Schwa is the brief  "neutral" vowel [a] which shows special phonological properties in many languages. It is central to an understanding of the SUammon phonology and morphology. It is argued here that there are three different "kinds" of schwa in SUammon, as evidenced by their phonological behaviour: (i) excrescent schwas, (ii) epenthetic schwas, and (iii) reduced full vowels, which have the same prosodic structure as schwa. One of the major goals with respect to the distributional restrictions is to demonstrate that the surface constraint (or constraints) which bans schwa in stressed open syllables (informally abbreviated as, * a ] ) plays a central role in the organization of the grammar of the language. The a  phonological constraint *6]o when combined with other constraints within the grammar of SUammon has far-reaching implications for the phonological and morphological structure of the language. The symbol schwa [si] comes from the Roman alphabet; it is a lower-case e which has been rotated 180°. The name schwa or shva (Hebrew sh'wa) comes from traditional Hebrew grammar, according to Pullam and Ladusaw (1986) who cite Prokosch (1939:94). 1  4  I am assuming the general model of Optimality Theory in which constraint interaction, conflict and minimal violation determine the optimal output candidates, following Prince and Smolensky (1993), McCarthy and Prince (1993 et seq.), and many subsequent works that have been stimulated by the evolution of this research paradigm. This thesis presents continued research which I began in the context of my M.A. Thesis (Blake 1992) and addresses several outstanding issues which were raised there. In Blake (1992: 43-45), I observe that schwa in Shammon tends to occur in the following two contexts: (a) in an unstressed closed syllable (i.e. minimally a CaC syllable) or (b) in an open syllable which bears primary stress, as illustrated by the data in (1-2) (for an overview of the transcription system adopted here, see Guide to Pronunciation (pp.xv-xxvi); Chapter 2; and Appendices HI and IV). (1) Schwa in closed syllables: CaC Gloss  Schwa epenthesis  Output  a. PL-pq  paqpsq  p A q  b. t'in-?m  t'in?am  ten?am  to barbecue (sahnon)  Schwa epenthesis  Output  Gloss  XaXapx  Xa"Xapx  become broken  t' ak a  t'^a  edible root  nagi  nigi  you (sg) (Indep.)  Input  2  h  p A q  h  all white  (2) Schwa in stressed open syllables: Ca Input a. IMP-Xpx b. t ^ a c. ngi  3  w  w  e  w  v  Although there are no constraints on inputs within OT, I have provided Input forms (underlying representations) which conform to the following principles: (i) each morpheme has a single underlying representation or "Input form" for phonologically-conditioned allomorphs, and (ii) the Input only contains unpredictable information. It has been shown by many phonologists working within OT that when GEN creates other Inputs, the surface phonological and morphological constraints of the grammar will nonetheless converge to select the same optimal form. I therefore do not show multiple Inputs for the set of output candidates in this context. Here, the important point is that the surface distribution of schwa is determined by the constraint ranking regardless of whether or not it is present in the Input form. cf. Matthewson (1994: 38) for similar discussion regarding schwa in Lillooet (Salish). 2  This edible root grows in clumps and is prepared in a traditional rock-pit fire. It is steamed, peeled and eaten. M G calls them 'Indian bananas' due to their characteristic yellow colour, clustering bunches, and the fact that they are easily peeled (once cooked). This root, along with sea urchin, is considered a delicacy. The plant itself is a fine-stemmed fern. These roots may well be the rhizomes of the spiny woodfern. 3  5  Although schwa does appear to occur in stressed open syllables as shown by the data in (2) above, there are also a number of strategies which are employed in order to avoid stressed schwa in an open syllable, as shown by the data in (3). (3) Avoidance of stressed schwa in an open syllable Output  Gloss  63t' 9m  eaVt^am  jig for cod  O' restructuring  b. 0ym  Gayam  6a?yim  to sink  R' restructuring  c. t' xu  t^axu  ling cod  Gemination  Input a.  et'm e  8  e  d. p6=iq an  p90[h]iq an  pX0heq An  black hair  [h] epenthesis  d\ cx=iq an  caX[h]iq an  cexheq An  black hair  [h] epenthesis  e. pq=iq an  paq[?]iq an  pAq?sq An  blonde hair  [?] epenthesis  f. px m+[i]  pax im  pux em  steam  V-strengthening  w  w  w  w  w  w  w  w  w  w  w  w  These various different strategies include Glottal Restructuring (3.a-b), Gemination (3.c); [h] or [?] epenthesis (3.d-e), and Vowel strengthening (3.f). What all of these strategies in (3) have in common is that they prevent schwa from occurring in a stressed open syllable, thus avoiding violation of the constraint * a ] . 0  The contrast between (2-3) raises the following question: under what set of conditions does schwa occur in stressed open syllables? The proposal made in this thesis is that the constraint which aligns the head of the prosodic word (PrWd) to the left-edge of the lexical stem outranks the constraint against stressed schwa in an open syllable. This means that it is more important for primary stress to be properly aligned with the left-edge of the stem domain than it is to obey the constraint which bans stressed schwa in open syllables, h i this context, schwa will occur in a stressed open syllable, if no other strategy militates against it. In this thesis, I claim that the surface distribution of schwa in SUammon is predictable from surface prosodic constraints.  For example, schwa surfaces between a word-initial consonant  cluster in order to satisfy the high-ranking constraint against Complex Onsets in the language. 6  Schwa is also epenthesized in order to satisfy Proper Headedness, the requirement that the syllable which is the head of a Foot contains a nucleus, as shown by the contrast between stressed initial CaC syllables versus final Minor syllables (CC) which have no vocalic nucleus, as will be argued in Chapter 3. Shaw (1993,1995,1996c) characterizes schwa as non-lexical, featureless and non-moraic, drawing on evidence from a wide-range of languages including the Salish languages Stat'imcets (Lillooet) andNuxalk (Bella Coola). Kinkade (1998: 208) argues that epenthetic schwa in Upper Chehalis is both non-moraic and unspecified for phonological features, providing comparative evidence supporting the proposed representation of schwa in Sliammon. This dissertation also aims to contribute to the cross-linguistic studies of schwa in other Salish languages: Bagemihl (1991) on Bella Coola, Matthewson (1994), Roberts (1993), Roberts & Shaw (1994), Shaw (1993, 1994, 1996) on St'at'imcets (Lillooet),  Bianco (1996) on  Cowichan, Willet and Czaykowska-Higgins (1995) on Nxa'amxcin (Moses-Columbian), and Kinkade (1998) for Upper Chehalis, so that a cross-linguistic perspective on the behaviour of this vowel emerges.  1.2 Previous scholarship on Sliammon language Early scholarship on the Sliammon (Comox) language and culture includes: Haeberlin (1918), Sapir (1915), Boas and Haeberlin (1927), Swadesh (1950), Swadesh (1952), and Barnett (1955), amongst others. There are also a growing number of papers which appear in the pre-prints for the International Conference on Salish (and Neighboring) Languages; these include:  J. Davis  (1971a, 1970b, 1971, 1973), Bouchard (1971), Hamp (1971), Harris (1975), Hagege (1976), Kroeber (1988), Watanabe (1994b, 1996) and Blake (1997a, 1999), most of which are based on fieldwork on the language. Hagege (1981) produced a descriptive grammar of the language written in French (cf. Kroeber (1989) for a critical review). J. Davis (1970), Blake (1992), and Watanabe (1994a) are Masters Theses on the language, focusing mainly on basic aspects of the phonology and 7  Reduplication. Watanabe (2000) is a detailed description of the Morphology of the language which documents a number of aspects of the language which have not been discussed elsewhere. Research on the syntax of the language includes papers by J. Davis (1973, 1978a, 1978b), and Blake (1996, 1997a) on basic clause structure and Passives in Sliammon. Kroeber (1991, 1999) includes many syntactic examples of MComox (Homalco, Klahoose, Sliammon) in the broader context of his research on comparative Salish syntax, and are based on his own fieldwork on the language in the early 1980s. Harris' (1981) dissertation is on the Island Comox dialect; the speaker he worked with the late Mrs. Clifton, passed away several years ago. No other individuals are known to speak this dialect. There is also a growing body of fieldnotes on the language including: H.G. Barnett UBC Special Collections; W. Suttles (n.d.) SUammon fieldnotes; J. Davis (1969-1970); Kennedy and Bouchard (1971-1981); P. Kroeber (early 1980s); Blake (1988-2000); Watanabe (1990-2000) collaborative SUammon-UBC TLEF Project under the direction of Dr. Patricia A. Shaw (19961998), and S. Urbanczyk (1998) working on Klahoose, Cortes Island, B.C. Kennedy and Bouchard's (1983) publication entitled Sliammon Life, Sliammon Lands reports on detailed ethnographic research from (1971-1981). 1.3 Theoretical Framework 1.3.1 Representations adopted in this Dissertation The issue of representation is one which is particularly challenging. I will briefly discuss some of the most common issue regarding documentation, transcription, and levels of representation. Previous work on the language such as Sapir (1915), J.Davis (1970, 1971), Blake (1992), Watanabe (1994a) include the surface phonetic forms for most if not all examples.  One of the  potential drawbacks is that a phonetic representation may include a lot of information which is noncontrastive. Nonetheless, a distinct advantage of including aU of the surface phonetic forms is so that the reader can readily compare this level of representation with the more abstract phonemic representation, and future scholars will be able to make new hypotheses based on this surface phonetic data. It therefore seems preferable to me to include as much phonetic detail as possible. 8  At the same time, for practical reasons many forms are not transcribed with comparable phonetic detail. Any phonetic transcription is an abstraction to some degree.  The process of  transcribing what one hears is, of course, the first level of analysis. In the chapters to follow, the representation adopted for any set of data is somewhat dependent upon the focus of a particular section. The reader is therefore referred to Chapter 2, where I discuss many of the most frequently occurring allophonic processes which have been the focus of my research on the language. Because of the inevitable variation attested in the speech of any individual or of different individuals within a speech community, the reader is urged to cross-reference all of the available sources on the language in order to ascertain the range of variation in both underlying representation and surface forms. The representations adopted in this thesis are as follows: phonemic (Input) form and surface (Output) representation. In addition, I often provide for the reader an "intermediate" form which abstracts away from certain aspects of the consonant/vowel interaction in order to make the discussion at hand easier to follow. For example, reconsider the data presented earlier in (1), the first sets of which are reproduced for ease of reference here: (1) Schwa in closed syllables: CaC Input  Schwa epenthesis  Output  a. PL-pq  psqpaq  p A q  b. t'in-?m  trn?am  ten?3m  h  p A q  Gloss h  all white to barbecue (salmon)  The data in column 3 the "Output" represents the surface phonetic level, showing in this case the effects of C-V interactions. Even narrower phonetic transcriptions will sometimes appear in square brackets [ ], when they are used. The data in column 2 abstract away from some of the phonetic detail in Column 3, e.g. vowel height assimilation to adjacent consonants (i ~ e); schwa colouration ( a ~ A) ; aspiration of stops (q ~ q ) etc. The data in Column 1 represent an abstraction: the kinds h  of phonological information which are predictable are not present, e.g. I\l is regularly realized as [e] following a glottalized obstruent; a stop Iql is regularly aspirated ....etc. Column 4 provides a short  9  English translation. The reader is referred to Chapter 2 and the Appendices for further exemplification. In each section, I have tried to indicate clearly what phonological properties I am attempting to account for, as well as clearly indicate what properties are to be discussed elsewhere, or are simply not handled in the context of this thesis. Inevitably and regrettably, this thesis is also finite, and discusses only afractionof the very fascinating linguistic processes integral to this language. Nonetheless, an area in which I have tried to pay particular attention to detail is in my transcription of the phonetic vowel quality. I have also focused specifically on the realization of glottalized resonants and obstruents, aspiration of syllable-final consonants, the allophonic realization of schwa, and native speaker judgements regarding morafication and syllabification. Areas which remain particularly problematic both for description and for analysis include stress/pitch interaction and the placement of secondary stress. In cases where I could not decide whether or not there was a secondary stress, I did not mark stress. Therefore, because a form is not marked with secondary stress does not imply that secondary stress does not occur in this word/phrase rather it simply may not be indicated. Primary stress is marked throughout. See §3.3 for further discussion. It should also be noted that there is a considerable range in variation both within a single individual's speech and across speakers within a community, some documentation of which can be found in Davis' (1970) work. The range of variation is also documented in this thesis by listing the variants. For example, the word for coat is pronounced in a variety of ways: [kapo ~ kyaspo]. The tilda [ ~ ] separates one variant from another. The phonemic or Input forms adopted in this thesis and those labelled in Watanabe (2000) as //morpho-phonemic// are similar. Watanabe (2000) does not take into account the effects of syllabification and foot formation in driving vowel reduction, schwa colouration, or vowel height assimilation. His surface representations have phonemic vowels and he generally abstracts away from consonant/vowel interaction. These differences may appear to be significant on the surface but are the result of different focus and investigation of different linguistic properties of the language. I have made a number of choices with respect to representations which attempt to 10  mimrnize these differences. For example, to be consistent with Watanabe (2000a) I use /j, g/ in the present work rather than the archi-phonemes IY, W/ of Blake (1992). I have also adopted Ii, u, a/ rather than Ie, o, al for ease of comparison. 1.3.2 Theoretical Models of Phonology The analysis presented in this dissertation is predicated on two major theoretical claims made within the general theory of phonological systems. First, I adopt Shaw's (1996c) claim which argues that "an adequate theory of syllable structure must recognize both Nuclear headedness and moraic weight as independent structural properties." Shaw's claims are formulated in Optimahty Theoretic terms. In particular, she has proposed the addition of the constraint SYLL MORA which ensures that all syllables have phonological weight. The analysis presented in this thesis draws significantly on Shaw's claims. Second, I adopt the general model of Optimality Theory (OT), following Prince and Smolensky (1993), McCarthy and Prince (1993 et seq.). OT is a model of output constraints in which constraints are ranked with respect to each other. All constraints are in principle violable, and there is constraint interaction and conflict. An optimal output candidate will often violate a lower-ranking constraint in order to satisfy a higher-ranking, conflicting constraint. The reader is referred to the growing body of literature on OT for additional background and exemplification within this particular theoretical framework, and particularly to McCarthy and Prince (1994, 1995) on Prosodic Morphology within OT. 1.4 Overview of the Dissertation The next section presents an outline of the remainder of the dissertation. Chapter 2 presents a summary of the consonantal and vocalic sound system of the language, focusing on the phonological weight contrast between the full vowels Ii, u, al and schwa. Chapter 2 outlines the tenets of the Nuclear Moraic model of Shaw (1993, 1995, 1996a,b,c). In accordance with this theoretical model, I hypothesize here that schwa in Shammon is both nonmoraic and featureless (cf. Blake 1992, Kinkade 1992, Shaw 1993 et seq.). In addition, this  11  chapter establishes that the realization of Full Vowel Reduction (laxing) in unstressed syllables in Sliammon is distinct from the surface realization of schwa. Chapter 3 provides an analysis of the prosodic structure of the language with emphasis on the distribution of schwa. §3.1 motivates the moraic structure of coda consonants in the language, §3.2 provides a description of some of the basic issues regarding syllabification, emphasizing the interaction between syllable structure constraints and the morphology. §3.3 provides a preliminary introduction to metrical structure in the language, highhghting the fact that primary stress in SUammon, unUke many other SaUsh languages in the family, is fixed on the stem-initial vowel regardless of its inherent weight. Chapter 4 begins with a basic typology of the different "kinds" of schwa which are found in SUammon as evidenced by their phonological behaviour: (i) excrescent schwa, (ii) epenthetic schwa and (iii) non-alternating schwas, which are subsumed under (ii). Schwa is proposed to be non-lexical (i.e. not present in the Input), and epenthesis is driven by the constraint Proper Headedness which requires that the Foot contains a syllable which is headed by a Nucleus (specifically, a vowel). §4.3 provides evidence that Full Vowel Reduction (laxing) involves the loss of phonological weight, expressed in terms of moras. The output of Full Vowel reduction is distinct from the realization of schwa in terms of its featural representation, but is claimed here to have the same prosodic representation as schwa. This section provides independent data on Full Vowel Reduction which confirms the hypothesis made in Blake (1999). Although schwa epenthesis is driven by the constraints on Proper Headedness, there is also evidence that the language has a number of strategies which conspire to avoid schwa in a stressed open syllable (*9]a). Chapter 5 brings together different cases of surface allomorphy, and shows that they receive a unified explanation with reference to the constraint *6]o. Chapter 6 addresses further impUcations for the analysis developed in the context of this dissertation. §6.1 documents and analyzes the variant forms of the possessive -hV suffix in SUammon, along with its variant sites of affixation (sometimes a suffix, sometimes an infix). §6.2 makes important claims regarding the status and form of prefixes, interacting with pervasive constraints on the morpho-syntactic structure of the language. 12  The effect of these general  constraints on two prefixes in particular is discussed. One is the s-nominalizer, broadly attested across other languages in the Salish family, but conspicuously absent in Shammon. The second is a plural prefix/infix, here hypothesized to be /LV, cognate with the plural l-infix of Musqueam and Saanich. The existence of this prefix has not been previously recognized by others working with Sliammon, undoubtedly due to its highly variant realization, viz. [-i? ~ -u? ~ -a?]. What is shown here is that these variants follow directly from the convergence of hypotheses related to the realization of sonorant /L7, combined with constraints on prefixation. A crucial observation related to both these underlying prefixes is that both, being strictly "consonantal", would violate a pervasive constraint against initial complex clusters if they were simply prefixed, i.e. *s-C..., *L'C... Significantly, 9-epenthesis is not an available strategy to rescue either of these cases. The thesis concludes, therefore, with a discussion of how the operative higher order constraints here interact with the constraints governing the realization of schwa which have been motivated in previous chapters.  13  Chapter 2: Introduction to Sliammon Sound System How wonderful is the human voice! It is indeed the organ of the soul! .... The soul of man is audible, not visible. A sound alone betrays the flowing of the eternal fountain, invisible to man! Henry David Longfellow  2.0 Introduction The goal of this chapter is to present a detailed overview of the sound system of Shammon, presenting first the consonant contrasts and then focusing on the vowel inventory.  2.1 Sliammon Sound System Sliammon, like other Salish languages, has a large consonantal inventory (43 consonants). The contrasts indicated here are based on minimal and near-minimal pairs presented in Appendix IV, and are largely in agreement with work of J.Davis (1970), Blake (1992) and Watanabe (1994a, 2000). The proposed phonemic inventory of the language is presented in (1):  (1) Sliammon Phonemic Inventory (cf. J.Davis 1970, Kroeber 1989, Blake 1992, Watanabe 1994) Place  Labial  Manner Stops plain ejectives  P P  Fricatives  Dental Interdental t  e  Alveolar  Lateral  Alveo-Palatal Palatals  t  X  c  t'  f  t  t  e  s  t  s  e  Non-continuant Resonants Resonants plain glottalized  m m  n n  1 f  Uvular  Velar  k k  k k  w  w  x  w  J  g  J  k  w  L  y  w  u  i Vowels  [9]  a 14  q q  q q  x  x  Laryngeal  w  w  w  ? h  Each symbol in (1) is an abbreviation for a set of phonological features which will be discussed in more detail in §  2.2 Consonant System 2.2.1 The Inventory There are 19 Stops (13 stops, 6 affricates) plain and glottalized pairs at seven different places of articulation: Labial, Dental, Alveolar, Alveo-palatal, Velar, Uvular and Laryngeal: /p, p, t , t' , t, t', X, % c, &, k, k, k , k , q, q, q , q , II. Affricates are analyzed here as a species of e  e  w  w  w  w  stop since they clearly pattern with the class of [-continuant]s, and are distinguished by their release features, following Jakobson, Fant and Halle (1952), Czaykowska-Higgins (1988), Shaw (1991), and LaCharite (1993). Thefricativesin Sliammon are 16, s, i , s, x , x, X , hi and are exemplified in Appendix IV. w  w  There is no labial or labio-dentalfricativein Sliammon, nor is there a velarfricativeIxl (except as an infrequently occurring variant of /§/). The symbols /j, j , g, g/ are used to represent the set of features which display Obstruent/Glide/Vowel alternations. In Blake (1992), I used the archi-phonemes IY,Y\ W, W7, distinct from ly, y, w, w/, to represent the segments which alternate (/j/ [J ~ c ~ y ~ i ~ e]; /g/ [g ~ k ~ x ~ w ~ u ~ o]) along with their glottalized counterparts; the use of capitalized glide w  symbols was to emphasize the Resonant behaviour of these sounds. In the present work, the symbols I), j, g, g/ are adopted for ease of comparison with other research on SUammon such as Davis (1970), Kroeber (1989, 1991/1999), Watanabe (1994, 2000).  However, the arguments  regarding resonant status of I), j, g, g/ and their proposed featural representation is basically the same as those in Blake (1992). In addition to /j, j, g, g/, the class of resonants includes /m, m, n, n, I, l', y, y, L, L', w, w/. The archi-phonemes IL, VI are used to indicate a set of features which show a range of morphophonemic alternations ILI [w ~ y ~ i] and IVI [w ~ y ~ a? ~ ?a ~ i] to be discussed in §2.2.5. This traditional way of presenting the consonant inventory is supplemented with extensive discussion of what these symbols represent - since each symbol is an abbreviation for a set of 15  distinctive features.  The consonant and vowel inventory will be discussed in terms of  combinatorial specification §2.3, following Archangel! and PuUeyblank (1994). Contrast the large consonantal inventory with the relatively small phonemic vowel inventory in (1): Ii, u, al plus [9]. Although there are a large number of surface vowels in the system, these surface variants arise from consonant-vowel (C-V or V-C) interaction (retraction, labialization, place assimilation) and vowel reduction.  The surface realization of schwa is  discussed in detail in §2.4.  2.2.2 Consonant AUophones I The following consonantal allophones are not the focus of this study and are listed here so as not to give a false impression about the Shammon language. It is beyond the scope of the current dissertation to discuss and document the full range of complexity which characterizes the consonantal sound system of the language. The following consonantal allophones are mentioned here and each one of them is identified as a topic which warrants for future research (i.e. systematic ehcitation and acoustic studies).  The goal of this section is to define limitations on range of  consonantal phenomena to be discussed in the remainder of this thesis, while acknowledging the richness and range of complexity within the consonantal sound system. In general, these data include a level of phonetic detail which is not necessarily transcribed in the remainder of this thesis. Stops and Affricates Affiication of Stops The uvular stops /q , q7 are often accompanied by a fricative release, indicated here by the raised [x] after the stop. These affricated stops are written phonetically as [qx] and [qx], and appear to be restricted to syllable onset position, as shown by the data in (2-3).  16  (2)/q/-[qx] Output  Gloss  qa?um  qxa?om ~ qa?wum  eye  b. CsCPL-qawm  q9wqa?w9m  qx9wqxa?wi>m ~ d  eyes  c. CaCPL-maqin  maqmaqsn  niAq'taaqxan  lots of hair  d. pyqn  piqan  pi9qx9n  shoulder  e. CaCPL-pyqn  pdypiqan  pipisqxan  shoulders  Output  Gloss  Input a.  qawm  (3)/q/-[qx] Input a.  qayk  qayk  w  b. CaCPL-qayk  w  qx/Cyk ~ qXyk w  w  qayqayk  w  w  q eyq ayk^ x  x  bald eagle eagles  In contrast /q , q / are rarely affricated, a fact which is also noted by J.Davis (1970: 40). w  w Aspiration Stops (stops and affricates) are aspirated word-finally, as shown in (4) (cf. Davis 1970: 62) (4) word-finally Input  Output  Gloss  a. ?j a cx ?ut  ?9?}acx ?ut  ?a?jecx ?ot  b. ?j a cap?ut  ?9?jacap?ut  ?a?jEcep?ot  c. ?j can ?ut  ?iycan?ut  ?iy cen?©^  d. ?wk st ?ut ?]  ?9wk st?ut?9y  ?u-k st ?ot W  e. ?j-mut  ?i?mut  Yv.mvfi ~ Yiymvfi  w  w  w  w  w  h  h  w  h  How are you (sg) ? How are you (pi)? I'mfme We're all fine very good  Word-internal syllable-final stops are also aspirated, as shown in (5) (cf. Blake 1992, 1995).  17  (5) syllable finally Output  Input  Gloss  a.  t^ip-it-mut  t^ipitmut  t' e . pet . mut*.  very pointed  b.  CaCPL-pq  paqpaq  p A q  ail white  c.  ?atnupil  Vatnupil  ?at . no . pel.  car, auto  c'.  ?atmupil  ?atmupll  ?at . mo . pel  car, auto  d.  ngapty  nigapti  nX  women  e.  x up-x up w  w  3  h  11  .  p A q  1  h  .  h  x up . x up . w  e  w  h  . gAp  1  1  .  ti.  X op . x o^. w  h  w  hummingbkd  The examples in (6) show that aspiration may actually occur at a mora boundary . 1  (6) Input  Output  Gloss  a. ct-t  C9tt  cit¥  cut it  a', ct-t cn sm  catt can sam  cftVcin sam  I will cut it up  It should also be noted that it is sometimes quite difficult to distinguish the full release of a stop in the environment before another stop versus the presence of aspiration in this context. Palatal Off-glide on Palatals and Plain Velars The alveopalatal and plain velar (DOR) consonants are often followed by an audible palatal off-glide, as shown by the data in (7-11).  Bagemihl (1991: 635:fn 50) suggests that aspiration in Bella Coola may occur at a mora boundary: C-» f+sp  1  giy_jM.  18  (7) c - * [&] (note: [1] represents a dark/retracted 1 and not a voiceless lateral fricative i) Input  Output  Gloss  a. calas  calas  C^^IAS  three  b. calas-s  calass  cyeiAs:  Wednesday (three)  Output  Gloss  nis^ani  swim  Output  Gloss  (8) s - » [gy] Input a. ns-m  nasam  (9) k -»[ky] Input w  b. C3CPL-k3t=iq =uja w  pinky, small finger  ka?t'iq u?ja  kye?teq o?je  k^flca?t'iq ula  ky^tkye?t'eq oje  small fingers  Output  Gloss  hsnkyela  pot for cooking  w  w  a. k9t'=iq =uja  h  w  w  h  [P]  (10) k - * Input  a. hankala  hdnkala  b. kikak  kikak  c. cyk-?m=min  6ayk[a]?[a]min  cl;ky£?amin  trypan  d. tTl=iq  tMiq  tAlkyeq  nostril  fXitAlkyeq^  nostrils  tifcy  slim  Output  Gloss  w  e. C3CPL-tll-iq  w  crow  w  w  t9lt'9lkiq  w  tik  f. tit  (ii) g - t e P i Input a. tig9x /tigux w  w  b. DIM-qag0+[i]=ui  tigax / tigux  tigy-ux  qaqagiOut  qicpg'\ Qb\  w  w  nine  w  s  19  small potato  This fact is documented by J.Davis (1970: 38, 67) and may be explained by the articulation of these consonants which are characterized as having a raised dorsum (DOR [hi]).  The p] is  therefore a release feature associated with a high consonant before the transition to a back non-high vowel. Secondary Lahialization Consonants are pronounced with visible Up rounding before the rounded vowel /u/, which is indicated here by the raised [*"], as shown by the data in (12). This is a phonetic effect in these cases since there is no contrast between t' and i ™ or between t and t , for example. e  6  w  (12) Input a. t' ut' uq e  e  w  a'. CaPL-f ut' uq e  6  Output  Gloss  t'ewo't-ewoqw  feather  t' a^ ot oqw  feathers dog salmon  e  w  w  ,ew  b. AX =ay  && ay  t 6x Ay  V. CaCPL-fox^ay  fo'x^bjTay  %»6x 'K 'ox Ay  dog salmon (pi)  c. CaCPL-&iq =sn  *9q *aq S9n  X dq ^Aq sin  thighs  d. &?=uja  ^u?u?Ja  £ d?o?jE  fast picker, fast hands  d\ CaPL-ft?=uja  Ma?u?Ja  Aa^ ?o?jE  w  w  w  w  w  w  w  v  v  w  w  w  w  w  h  lots of fast pickers  3  The next examples show that labialization also results in a neutralization of contrast: /q , q / merge w  in favour of [q™] before the labio-velar vowel IvJ (=13.a). The related words in (13.a'-a") show that the Root /pq/ [p/q] white is q-final.  (13) Input  Output  Gloss  a. pq=uk t  p9quk t  pAq ok t  a', pq  paq  pXq  a", pq-pq  psq-paq  p/q pAq  w  w  w  w  20  white blanket, outfit white  h  h  h  h  all white  One minor and erratic phonetic effect (which will be mentioned here) includes the rounding effect that the consonant c has on a following vowel, as shown by the morphologically related examples in (14.a-c) versus (14.d).  (14) Lexical Suffix =cis/ =c9s hand :  Gloss  Output  Input  paddling (one paddle)  a. IMP-hiw=cis-ma  hi-hiw=cis-ma  heh9wcis ma  b. fti?=Cis  f u?=cis  t^dVcis  seven  c. ta?a=cis  ta?a=cis  taVacis  eight  09ya=cis  ©iecus ~ ©iyecis  five  3  y  h  but d. 0iya=cis  2 Fricatives Allophones of /& The articulation of the interdental fricative /©/ varies to some degree between [0 ~ s]. The symbol [s] is used to indicate an interdental s sound; a sound which is intermediate between 0 and s. The variant [g] appears to occur before or after a back vowel u, a, 9 and may simply be a slightly retracted articulation of /©/ so that the transition from an anterior consonant to a back vowel (or vice versa) is made with less articulatory effort.  Kroeber (p.c.) suggests that (14.d) may be /0iya=cs=us/ [Oiecsus] meaning further research.  21  five dollars. This point requires  (15) s in the environment of /a/  Gloss  Output  Input a. k Gays  k a'0ays  k i5GAys ~ k iSsAys  island  b. 6a?n  0a?an  0a?an ~ sa?an  cohoe salmon (late run)  c. CaCPL-Ga?n  0a?0a?an  0a?0a?an ~ sa?sa?an  lots of cohoe  d. Gat^m  Gat^sm  ©at^am ~ sat^am  Spring salmon  w  w  w  w  e.  C3CPL-6at' m  Oat' 0at' am  Git^Gat^am ~ sit^sat^am  Spring salmon (pi)  f.  DM-eat^mfi]^!  GaeafimM  GaGat^emu'l ~ sasat^enrCri  small Spring salmon  qigaG  qe-gAG ~  deer  6  g. qigae  e  e  qe-gAS^  (16) in the environment of IvJ Input  Output  Gloss  a. Gumin  Oilman  Gdm9n -^dman  eyebrow  b. CaCPL-Gumin  GamGuman  sgmsoman  eyebrows  c. GuGin  GuGin ~ 9n  GdGin  lips  d. C9CPL-GuGin  GaGGuGin  sissoGm  lots oflips  In the data in (17), /G/ is recorded systematically as [0]. (17) Input Gi?Ga  a. GyOa  Output  Gloss  GI?GA  that one (fern)  b. Giy=umix =tn  Giyumix tn  6eyomix tn  c. q up=uGin  q upuGin  q 6poGen ~ q d:p6Gen  beard  d. MP-maGiw+[?]  mamaGiw  marnaGew  to limp (limping)  e. Giq=nac  Giqnac  GeqnAc  dig roots  w  w  w  w  floor  w  w  w  h  22  The situation is also complicated by the fact that the Island Comox language is reported to have had [s] where the Mainland dialects (Homalco, Klahoose and Sliammon) use [6] (cf, Harris (1981) on Island Comox). The interdental fricative may actually have a grooved articulation which makes it sound much closer to [s] - this phenomena is surely the source of some misrecordings where s is written instead of 0. Variants of/s/  [s-x^  Historically, Proto-Salish (PS) *x fronted to s in Sliammon (cf. Thompson (1979a), Kuipers (1981, 1982), Galloway (1988), Kinkade (class notes), and Kroeber (1999:6-10) for a summary of the phonological sound correspondences). One elder spontanteously produced [x^J as a variant of III, as shown by the following examples. The forms in (18) are cited from a single speaker. (18) Input  Gloss  Output x a?lowlax m  spiked heels, high heels  a'. xaluwla=sn  x^ilowlastn  spiked heels  b. ?imax-ija / ?imax-ija  ?emax i?je ~ ? £ m a x i j a  ant (lit: fast walker)  b'. ?imas  ?emA§  walk  a.  x 9l'uwla=xn w  w  y  y  y  The examples in (18.a-b) are apparently old words judging from the comments of the elder, who learned these words from her father. These isolated examples may reflect the final stages of the historical sound shift since the Lexical Suffix (LS) =s9n foot, lower leg is most often pronounced as [=san ~ =§in] by other consultants, but is pronounced as [=x 9n] in these limited set of y  examples. Sapir (1915:30) records [x^] for  ant, and J. Davis (1970) also documents a small  number of instances of [x^l. Watanabe (p.c.) also records [x ix it'a-tsm] tip-toeing, walking on y  one's toes.  23  y Allophones of x  w  Davis (1970: 37) states that /x / is pronounced as [x™] intervocalically, and is often reduced to a w  voiceless [w] elsewhere. I have transcribed [x ] both intervocahcally, and in word-initial prew  vocahc position (i.e. syllable Onset position).  The reduction to [w] was also noted and  documented in the context of the Sliammon/UBC Orthography Workshops (1996-1998), and is written here as a voiceless labio-velar glide: [wj. (19) Input  Output  Gloss  a. sattx  sattx  saitw  woman  b. janx  Janx  jeuw  fish, salmon (generic)  c. Xpx  Xspx  Xdpw  broke  w  w  w  w  w  w  d. ?asx  ?asx  ?dsw  seal  e. tix eai  tix eai  tix eai~tiw6ai  tongue  w  w  w  w  w  Notice for example that the second person subject clitic cx  w  you (sg) is most often  pronounced [cw]; however, since it is a frequently occurring morpheme, and [w] is an allophone of /x /, it is written as cx elsewhere in the dissertation and will generally be written that way in order w  w  to mimmize differences in transcription, and facilitate comparative Salish research. (20) Second Person Subject Clitic: cx you (sg) w  Input a. IMP-tam cx  b. IMP-JX* a cx  c. papi-m cx  tatam cx  w  w  w  j9-yX" a cx papim cx  w  d. IMP-cag-t cx  w  w  w  ca-cag-[a](t) cx  w  Output  Gloss  tatamcw  What are you doing?  ji:Xaxw  Are you (sg) running?  papemcw  you're working  c;s6egACW  you 're helping him/her  The fact that /x / is often realized as a voiceless labio-velar glide [w] provides support for the w  existence of minor syllables in Sliammon (cf. § For example, /sa1tx / [saitw] woman is w  syllabified as sai . tw and ca-cag-a(t) cx [cecegAcw] you're helping him/her is syllabified as w  hi . ce . gA . cw . 24 h ~ x in the environment of IMI w  The fricative Dal alternates with a surface [x ] in the environment of Ivd, as shown by the variants in w  (21). (21) Input  Output  Gloss  a. puhu  puhu  pono~pdx o  raven  b. nuh-m  nuham  ndhom~ndx om  feast, invite for feast  w  w Non-continuant Resonants /J, j, g, g/ Pre-nasalization Prenasahzation of /g/ [ g] n  occurs phonetically in word-initial position  utterance/phrase initially in sentential contexts).  (usually  Prenasahzation is related to the timing of  articulatory gestures; the onset of voicing occurs before the velum has completely sealed off the nasal cavity. The air which escapes through the nasal cavity produces the pre-nasalized obstruent: [ g]. This is a phonetic effect which is variable, and has not been recorded for many lexical items. n  It will not be transcribed elsewhere in this thesis.  (22) [9g] Input  Output  Gloss  a. gi  gat  rjgai _ g^<j  shiny  b. ga-gt'  gawt'  5ga?o f~ gA'?awt'  oar  b'. CaCPL-gagt'  gawgawt'  gugo t'  oars  b". DIM-gf"  gigat'  gigAf*  small oar  c. gija/ gja ? check  gija  Ogije  soil, ground  w  w  Jgi:gije  lots of soil  ga?Pap  Oga^ap  driving, steering  gaPgaVt^ap  5ga?ga?t' ap  c'. CaCPL-gija  r  d. gVt^ap d'. C 3 C P L - g ? f a p 9  n  h  e  25  h  driving around pi.  Note that the corresponding pre-nasalized alveopalatal [ j] has not been documented, so that there ny  is an asymmetry between the realization of /j/ and /g/, as observed in Blake (1992). Further, this pre-nasalization is not observed before the voiceless velar stops Ik, k, k , k /. w  w Retracted /g/ Davis (1970: 44-45) records the following words which contain a retracted variant of /g/, written here as [G] (note: the proposed Input forms are mine). He analyzes these examples as cases of free variation in which /g/ is assunilated in the environment of either a preceding or a following (non-adjacent) uvular consonant [q]. (23) Data cited from J.Davis (1970:44-45) Input  Output  Gloss  a. gaqaO  [G/qa6]  married woman  b. ga-gq[aq]a6  [GawqaqaG]  married women  c. qax=igan  [qaxeGtn]  tell lies  [qa'aG+naiqawtx™]  bawdy house  d. qa?agin=aq=awtx  w  3  J.Davis also cites [ce? Gay] old time wooden spoon but is unable to explain the retraction in this 3  case. It is proposed here that this comes from /cag=?ay/ (help=tree) "wooden-helper" and that /g/ [G] is retracted in the environment of the following glottal (cf. §2.4 which shows that ? patterns with the post-velars in the retraction of the full vowels). It is proposed here that /g/ is retracted in the environment of any post-velar (PHAR) consonant, this class including /q, q, q , q , x, x , h, ?/ w  w  w  (cf. §2.3 on phonological features). Additional examples were recorded in the context of the present research: (24) a.  mga  msga  [mXGA]  The current most well-accepted translation of gaqa6 is  cougar  husband; gawqasam means playing house; living together. 26 Resonants Nasals ~ voiced stops Davis (1970: 34) notes that the nasals /m, n/ are sometimes strengthened to [b, d] in wordfinal position. These are clearly allophones of IvaJ and /n/. This was also noted for the following lexical items from the most elderly consultants, but does not appear to be present in the pronunciation of younger speakers. Sapir (1915) makes the same observation for Island Comox.  (25) [m] ~ [b] Input  Output  a. jaq -m w  Jaq -9m w  Gloss  jeq Am~ jeq Ab  sweat, perspire  Output  Gloss  w  w  (26) [n] ~ [d] Input a. ctux an w  catux an w  citux An ~ citux Ad w  w  wild blackberry Glottalized resonants There is some variation in the surface realization of glottalized resonants in word-final position, as shown by the data in (27) and discussed in greater detail in Chapter 5. (27) Input  Output  Gloss  a. cuj  cuy  cuy ~ cuy?  child, baby  b. tarn ga tan  tarn ga tan  tarn gA tXn ~ tAn?  what's that?  c.  AaXxay  XalbUy ~ XaXxAy?  old  d. IMP-q e-m+[?]  q 9-q e-am  q dq e9m  telling a story  e. tam  tarn  tAm ~ tAm?  belt, to tie  DIM-Aaxay+[?] w  w  w  w  27  w Retracted Coronals The next section documents retraction of the coronal consonants Isi and HI. Is/ is retracted to [s] in a number of contexts: (28) Input  Output  Gloss  a. mu?us  mu?us  md?os  head  a'. C9CPL-mu?us  m9?-mu?us  mamd?os  heads  The resonant IV is often pronounced with a back articulation [1]. [Note the symbol i is used for a voiceless lateral fricative; 1 for a velarized liquid]. (29) Input  Output  a. CaCPL-bplas  laptaplas  tepteplas  Gloss planks (<CJargon)  The data in (30) shows that retraction (PHAR) is spread throughout the word. (30) Input a. q alas  q alas  w  w  a'. C9CpL-q alas w  Gloss  q alAS  raccoon  q Xlq alAS  raccoons  w  q 9l-q alas w  Output  w  w  w  There are a small number of "retracted" Roots have been identified in SUammon; however, this area of the phonology requires further research . Consider the following retracted Root in 4  SUammon sal- (the retraction is indicated here by underlying the Root) which refers to a circular motion, as "you would turn a glass or the continual rotation of a Ferris wheel". In the surface form,  Retraction has received quite a lot  of  attention in the  discussion of  Interior Salish languages  (cf. van Eijk (1985/1997), Remnant (1990), Bessell (1992), Shahin (1995) on Lillooet (N.Interior Salish); Doak (1989), amongst others, and Czaykowska-Higgins/Kinkade (1998) for general discussion regarding retraction.  28  both the vowel [a] and the coronal consonants [s, 1] are noticably retracted: [§A!]. The proposal made here is that this morpheme has a PHAR feature associated with it, as indicated by the presence of PHAR in the proposed Input. (31) Sliammon Retracted Root salInput  Output  Gloss  a. sl+PHAR-t cx.w  sal Tf> cx'w  SATCX'  you turn (it) around  b. IMP-sl+PHAR-m  sa-sal-am  sdsalam  (its) turning  1  The following Cowichan (Hul'qumi'num') cognates may help to explain the source of Retraction in this Shammon Root. The Cowichan orthographic forms appear in angled brackets < > and are cited from Hukari and Peter (1995: 295). The Output forms have been transliterated based on the Guide to Pronunciation which appears in H&P (1995:340-341).  (32) Cowichan Cognates Orthographic Form  Output  Gloss  a. <sul'-ut>  sal'-at  sal'at  spin wool  b. <sul-sui'-tun>  sal'-saf-tan  sal'sal'tan  spinner, spindle whorl  ,  As shown by the data in (32), the Cowichan Root has a final glottalized t - the glottalization may well be the source of Retraction in the Shammon form.  This tentative proposal seems to be  supported by the observation that glottal stop and glottalized consonants in Shammon have a lowering effect on adjacent vowels in the language, as will be discussed in §2.4.  29  2.2.3 Consonant Allophones II The following issues require particular discussion since these consonantal processes occur pervasively in the language and the reader will need to keep track of these insertion, and deletion phenomena. Issue of Labialization C ° v s C  w  Davis (1970:27-28) notes that the labialized consonants are not perceived as rounded in the environment of a tense rounded vowel. He cites Sapir (1915) on Island Comox, and Newman (1969) on Bella Coola, both of whom make the same observation. Davis represents the rounded series as /C°/ at the phonemic level which is used to indicate visible labialization without an audible off-glide.  He differentiates [C°] from [C™] on the surface, and inserts the off-glide [ ] by w  phonological rule (viz. I°l - * [*]/  non-round vowel). The data in (33) illustrate this convention.  (33) Input  Output  Gloss  k i§k is  Stellar's jay  a. k°isk°i§  k isk is  a'. CVCpL-k°isk°is  k is-k isk is  k isk isk is  Stellar's jays  b. q°n =iq la  q 3?niq ia  q a? neq iA  knee  :  w  w  w  w  w  0<  w  w  w  w  w  w  w  w  a  w  q an-q 9?niq ta  q 9nq a? neq iA  knees  c. k°urht  k°urht  k°umt  kelp  c'.  k°um-k°umt  k°umk°umt  b . C9CpL-q n=iq ta 1  0  0,  w  w  w  w  w  a  w<  but  CVCpL-k°umt  h  h  lots ofkelp  The distinction between [C°] and [C ] is not transcribed in the remainder of this thesis. A raised w  [ ] is used throughout to indicate Up rounding, and does not differentiate between Up rounding w  versus Up rounding with an audible off-glide. The audible off-gUde is not perceived before a round 30  vowel; however, related plural CaC- reduplicative forms confirm that the consonants are labialized, as shown by the data in (34).  (34) a. C uC w  a'. C aCw Neutralization of contrast: C/C* -> C7u It should also be noted that there is a surface neutralization of plain vs. rounded contrast in velars and uvulars in the environment of a tense rounded vowel, as shown by the gaps in the data in Appendix IV. Evidence for /..Cu/ can be seen from the following plural and diminutive plural forms in (35-36). The surface form of the CaC- reduplicative prefix in (35.a) shows that the Root for ling cod is /t^xu/ with a plain Ixl when the triggering context for labialization is absent. The Ixl is realized as [x™] before the rounded vowel lul. Similar disambiguating behaviour is seen in the other forms here.  (35) Gloss  Output  Input  lots of small cod  ft-t^ax-feaxu?  t' lt' AXt' 0X O?  a', t^xu  t' axu  t^d^o  b. C3CpL-puxu=u1:  pax-paxui  pAXpAX 6t  small ravens  b'. puxu  puxu  pdx o  raven (cf. also pdho)  a.  DIM-CsCPL-f^xu+t?]  e  6  e  e  0  ling cod w  w  w  In contrast, the CaC reduplicative prefixes in (36.a'-b') show that these Roots contain a labialized consonant since labialization surfaces in contexts other than before a round vowel, as in x asw  x usam for example. w  31  (36) Output  Gloss  x us-9m  x dsom  I. ice cream, soapberry  a'. C3CpL-x us-m  x as-x us-am  x ASx osom  lots of soapberries  b. puk  puk  puk  book  Input a. x us-m w  w  w  w  w  w  b'. DIM-CaCpL-puk  v  w  w  pi-p9k -puk w  M  w  w  w  pe-puk puk w  w  lots of little books Glottalization and Phonology of Laryngeals Allophones of glottal stop /?/ One of the most difficult questions regarding the consonantal inventory of Shammon is what is the status of ? and h. Are these consonants a stop and fricative respectively, or are they both resonants, or is one an obstruent and the other a resonant? Davis (1970: 35) classifies /?, hi as sonorants, whereas Blake (1992) classifies 111 as a glottal stop, and Ihl as a fricative. In the case of/?/ there is conflicting evidence. In word-final position, a glottal abruptly stops the air flow and cuts off the preceding vowel, as in the pronunciation of the words in (37).  (37) Word-finally Input  Output  Gloss  a. DIM-ciya+[?]  ci-cya?  cic^yfe? ~ cic'yfe?  grandmother (Dim.)  a', ciya  ciya  ciya ~ ciye  grandmother  b. DIM-k upa+[?]  k u-k pa?  k uk pa?  grandfather (Dim.)  b'. k upa  k upa  k upa  grandfather  c.  pi-pcu?  pipcu?  small basket  c'. p£u  pacu  picu ~ picu  cedar root basket  d.  ni-nan  nenAn ~ nenAn?  nickname (Dim.)  nan  nan  name  w  w  DIM-pcu+[?]  DIM-nan+[?]  d'. nan  w  w  w  w  w  w  32  Word-final glottals (especially in stressed mono-syllables) are aspirated, providing evidence for their status as stops, as shown by the data in (38). Recall that stops and affricates are aspirated in syllable-final position as discussed in § Since these word-final glottals patterns with the stops in the language with respect to aspiration, it is proposed here that a full glottal stop [?] is characterized by the features [-cont, PHAR, LAR[cgl]].  Aspiration therefore targets all [-cont]  segments in the language. (38) Gloss  Output  Input a. ?wk st ?ut ?y  Tuk^t ?ot ?i?  b. ?i?  ?e?~?e?  w  h  we're all fine  h  yes  h  c. tala a k 90 na?  talaha k a6 na'?~talaha k 30 na?  Have you got money?  d. k^-t gy  k fst gi?~k ist gi?  count it!  w  w  w  h3  w  w  h3  h  h  This contrasts sharply with perception of some intervocalic glottals derived historically from resonants which typically lack complete closure characteristic of glottal stop.  Preliminary  spectographic data shows vowel formant structure, and creakiness characteristic of a glottal glide here hypothesized to be comprised of the features [SON, PHAR, LARfcgl]]. The symbol ['] is used here to represent this creaky voice articulation, following J.Davis (1970).  (39) Input  Output  Gloss  a. paL'  pa?a  pd'a~pd?a  one  b. saL'  sa?a  sd'a~sd?a  two  This observation was documented by Davis (1970), and therefore provides independent confirmation of these facts (cf. J.Davis (1970: 24-27) regarding glottal constriction).  Historically,  these glottals in Sliammon come from resonants, as shown by the comparative SUammon, Sechelt  33  (Central Coast Salish), and Thompson (Interior Salish) evidence in (40). The Sechelt data is cited from Beaumont (1985). The Thompson forms are cited from Thompson and Thompson (1992: 187); the Thompson vowel Id is realized as [e ~ as] cf. T&T (1994: 13-16) for allophones of Id.  (40) Comparative data: Sliammon / Sechelt / Thompson Input  Output  Gloss  a. paL'  pa'a  pd'a ~pd?a  one (SI)  a', pala  pala  pala  one (Se)  a", peye?  peye?  peye?  one (Th)  b. saL'  sa'a  sd'a~sd?a  two (SI)  b'. seye  seye  seye  two (Th)  Not only are these glottals ['] related to resonants from a comparative perspective but within the synchronic grammar of Shammon they show alternation with the resonants y and w, as shown by the morphologically related forms in (41).  (41)  Sliammon[7?~w~y]  Input  Output  Gloss  a. paL'  pa'a  pa'a ~ pa?a  one  a'. saL'  sa'a  sa'a ~ sa?a  two  b. paL'=us  pawus  paw?us ~ paw?us  one round object  b'. saL'^s  sawus  saTwus  two dollars  c. paL'-paL'  pay-pa'a  pe:pa'a ~ pe:pa?a  one person  c'. ssL'-saL'  say-sa'a  se:sa'a~ s^:sa?a  two people Proposal The proposal made here regarding the characterization of [?/'] follows from the theoretical perspective adopted regarding the nature of phonological features. If features are linguistic primes, 34  then what prevents combination of [PHAR, LAR[cgl]] with either of the features [-cont] or [son]? The feature [-cont] characterizes the stops and affricates in the language whereas the feature [son] characterizes the class of resonants in the language. If the features [PHAR, LAR[cgl]] are combined with [-cont], this produces a glottal stop whereas collocation of the features [PHAR, LAR[cgl]] with [son] produces a glottal resonant or glide (cf. Combinatorial Specification of Archangeli and Pulleyblank (1994)). If Sliammon has both glottal stops [?] and glottal glides ['] in its inventory, then what evidence is provided to the language learner to determine what kind of glottal is present in any particular form? There is at least some evidence based on morpho-phonemic alternation, as shown above, as well as by phonetic cues. In addition, a full glottal stop occurs word-initially (phrase-initially) and wordfinally whereas the glide ['] tends to occur between sonorants, as in (42). (42) [?] in onset position  [-cont PHAR LAR[cgl]]  [?] in word-final position  [-cont PHAR LAR[cgl]]  ['] in intervocalic position  [son PHAR LARfcgl]]  The hypothesis that there are glottal glides in the language receives some support from the facts regarding Glide Vocalization. Glide Vocalization The data in (43-44) show glide/vowel alternations (y~i/e and w~u/o) which are analized here as glide vocalization.  (43) Glide Vocalization: y ~ i/e Input  Output  a. i:ay=nac=tn  fay=nac=t9n  taynactan  a'. DIM-iay=nac=tn+[?]  ia-iy=nac=t[i]n  tafinactin  35  Gloss skirt small skirt  leaf  b. sayja  sayja  sa?yjA  b\ DIM-sayja+[?]  sa-syja?  sasi-je?  small leaf  c. qayx  qayx  qayx  Mink, stage name  c'. DIM-qayx=u4+[?]  qa-qyx^i  qaqex^ui  small (young) Mink  Output  Gloss potato  811  (44) Glide Vocalization: w ~ u/o Input a. qag©  qaw0  qawO  a'. DIM-qag0+[i]=ut(f)  qa-qag[i]6ui(t)  qaqAgyi e64(t )  small potato  a". CaCpL-qagO^uf  qaw-qawOui  qwdqAw66f  small potatoes  9  h  As shown by the data above, the palatal glide y alternates with the [-back] vowel IM [i ~ e], whereas the labio-velar glide w alternates with the vowel Jul [u ~ o].  Evidence from the form of  the Active Intransitive suffix /-?m/ when it is followed by the Instrumental suffix /=min/ suggests that vocalization of? does occur, surfacing as the corresponding PHAR vowel [a]. It is proposed here that the PHAR glide [son PHAR LARfcgl]] alternates with the corresponding PHAR vowel in the language, as demonstrated by the following ? ~ a alternations.  The Active Intransitive /-?m/ surfaces as [?am] in the following examples. (45) Input  Output  Gloss barbecue (fish)  a. t'in-?m  t'in-?9m  tEn?3m  b. sup-?m  sup-?3m  sdp?am ~  c.  ?ae-?m  ?a0-?9m  ?aG?Am  to give (s.t. atpotlatch)  d.  pA-?m  p3A-?9m  pXX?9m  pick (berries)  5  sdp9m  chop (wood)  5  The word S9p?am means hit something, and gat ak up is also used to mean chop wood. 6  w  36  When the Active Intransitive /-?m/ is followed by the Instrumental suffix =min, the identical instances of m-m are merged into a single [m], following Watanabe (2000). Notice that the glottal is both preceded and followed by a consonant, as shown by the forms in (46) Column 1.  (46) Input  ?-vocalization: [a]  Output  Gloss  a.  *ic-?m=min  iib-amm.  ii-^Emen  comb  b.  tg-?m=min  tu?-amin  tii?AmEn  freezer  taw  taw ~ hi?  ice  cst-q-amin  cftqamEn  knife  C9t-t-as  6lt tAS  he cut it  b\ tg c.  &-q-?m=min  c'. ct-t-as  h  d. DIM-msiq -?m=min  mi-msiq -arnin  memseq Amen  pins  d\ m s i q  m9siq  mXs£q  purple sea urchin  w  w  w  w  w  w  e. tat'-?m=min  iat'-amin  ia?tAm£n  herring rake  e\ tagat'  ia?gat'  ta?g9t  herring  e". DIM-tagat'+[?]  ia-igat'  fa'r g9t' 9  h  h  small herring  In the surface output form, the glottal [?] does not surface - a reflex of the PHAR vowel surfaces instead [-EntEn ~ AmEn ~ amsn], as shown by the data in (46.a-e) Columns 2 and 3. Conditions governing the realization of £ ~ A ~ a based on influence of adjacent consonants are detailed in §2.4. The proposal made here is that the PHAR glide ? vocalizes to [a], resulting in the observed surface output -amin. The implications of this proposal are far reaching but suggest that there is full symmetry in the relationship between Glide and Vowels in the language, as in (47).  37  (47) Glide Vocalization DOR[hi, -bk]  y~i  L A B DOR  w ~u  PHAR  ? ~a Glottalization Sliammon also has independent processes of glottalization which accompany Diminutive and Imperfective reduplication, as illustrated by the data in (48-49). It will be shown in §5.2 that glottalized resonants are not permitted in syllable-initial (non-moraic) position in keeping with the generalizations in Blake (1992, 1995). It is proposed here that glottalization which accompanies Diminutive reduplication targets theright-mostmoraic resonant within the stem domain, as shown by the data in (48). The floating glottal feature is represented here as [cgl] (constricted glottis) in Column 1.  (48) Diminutive Input  Output  Gloss  a. q at'-m  q atem  q atem  river  a'. DIM-q at-m+[i]+[cgl]  q a-q t'[i]m  q^q^fEih  smallriver,creek  b. tay^nac^n  tay=nac=t9n  iaynactan  skirt  tatinactm  small skirt  w  w  w  w  w  b\ DIM-i:ay=nac=tn+[i]+[cgl]'ia-1y=nac=t[i]n ::  ::  w  c. yaxay  yaxay  yXxAy  clam basket  c*. DIM-yaxay+[cgl]  ya-yxay  yey XAy  small clam basket  3  The diminutive data in (37.a-c) above show that vowel-final stems also take a final glottal stop. Glottalization which accompanies Imperfective reduplication targets theright-mostmoraic resonant (syllable-final) which is within the domain of Stem formation (cf. Appendix VII on the definition of morphological Stem, and Blake (in prep.) on Imperfective reduplication).  38  (49) Imperfective Output  Gloss  ?iftan  ?EitAn  ear  a'. IMP-?iitan+[cgl] c  ?i-?ittan c  ?8?E ftAnc  b. tg^qin  tag=qin  tiiwqen ~ t w q e n  to answer back  b\ IMP-tg=qin+[cgl]  to-tg[a]=qin  tat gaqen  answering back  c. hayi-m  haytam  hayiam  to flirt  c'. IMP-hayt-m+fcgl]  ha-hayiarh  hahayiAih  flirting  d. q as-Vm  q asam  q asAm  flower  d'. IMP-q as-Vm+[cgl]  q a-q asam  q aq asAm  blooming, flowering  e.  ?ah-am c  ?ahAmc  I got hurt  ?a-?ah-arh c  ?a?ahAihc  Input a.  ?iitan  w  w  ?ah-m c  e'. IMP-?ah-m+[cgl] c  <  3  3  w  w  Tm eating  h  w  w  w  w  Tm getting hurt  h Deglottalization The next section discusses deglottalization of a Root-final glottalized resonant in the environment before the Causative suffix /-stg/. Within Roots or Lexical Suffixes (LS), inherent glottalization associated with resonants is retained before another consonant, as shown by the data in (50-51).  (50) R o o t / L S Input  Output  Gloss older brother, best friend  a. nuwi  niiwi:  ndwt ~ ndTwai  b. FaHn  ^ay=tan  t^ayiTtan ~ t^aykan umbrella  c.  q aiaysan  q ai . ley . sin  q^i^'ay^sn  w  w  shoes  The examples in (51.a-f) show the retention of glottalization on =awtx , the Lexical Suffix for w  house, dwelling, building.  39  (51) Gloss  Output  Input a. tiws-am=awtx  tiwsEmawtx  place of learning  a". IMP-tiws-am+[cgl]  titiwSem  learning  b. cah-am=awtx  cehamawtx  church  b\ cah-am  cshAm  pray  w  w  w  w  c. Janx =awtx  jenx awtx  cannery  c'. janx  jenx ~jenw  fish  w  w  w  w  w  w  d. k u?ux =awtx  k u?ux awtx  smoke house  d'. k u?ux  k u?vx  smoked fish  w  w  w  w  w  w  w  w  w  w  e. ?axiG=awtx  ?axe6Awtx  bedroom, hotelroom  e\  ?axie  ?axeO  lay down  f.  pak it=awtx  w  w  w  pak itawtx  floating house  pa:k it  raft  w  w  f. pak it  w  w  w  When a Root which ends in a glottalized resonant is followed by a consonant-initial suffix, the glottalization associated with the resonant is lost (cf. ?j [?ay] good; -sx form of causative w  suffix /-stg/), as in the examples in (52). (52) Input  Output  Gloss  a. ?j-stg a cx ta hi  ?aysx a cx ta hi  ?iysx ACX t cti  Do you like the rain?  a'. ?j-stg can ta hi  ?ays(x ) can ta hi  ?iys cin t^ctf  I like the rain  w  w  w  w  w  w  3  b. ?j-stg a cx ta ?ax  ?aysx a cx ta ?ax  w  ?iysx ACX t ?ax Do you like the snow?  V. ?j-stg can ta ?ax  ?ays(x )can ta ?ax  w  ?iys cin t ?ax  w  w  w  w  w  w  w  w  3  40  3  w  w  I like the snow  At this point it is not clear whether or not deglottalization is triggered by all consonant-initial suffixes or whether the causative suffix -stg has a special deglottalizing effect . 6  When a glottalized resonant is followed by a subject clitic rather than a suffix (cx 'you sg.', st 'we') or particle (ga 'polite request, imperative'), deglottalization does not occur, w  as shown in (53). The subject clitic/particle in each case is underlined. (53)  a.  Gloss  Output  Input c a m sin c x  cam sin cx  w  a'. I M P - c a m sin c x  w  cem sin cx  w  cacam sin cx  cecsm Sin cx  w  Where are you going?  w  Where are you going?  w  What's that matter?  b. c m c x g a  cam cx ga  am cx ga  c.  cam ga  ctm gA~ cirh g§  w  c m ga  w  w  a  3  Why?  The proposal made here is that because the external argument (i.e. the subject pronoun) is introduced in the syntax (morpho-syntax), it is outside the domain of deglottalization. It is not within the same phonological domain as the Root/Stem which precedes it, and therefore the final glottalized resonant and following consonant are not adjacent to one another. Deglottalization does not occur between independent words either, as shown by the data in (54). The word q af come w  surfaces with a final [t ] before qa?men.  (54)  a. xat a k  w  a'. q f w  6  Watanabe  Gloss  Output  Input 0 q^t q?-m-an  xala  k a0 q A'f w  w  a  Do you want to come with me? come  q aT?~q AT? w  qa? men  w  (p.c.) proposes that this is an idiosynchratic property of the causative suffix. This is left for atopic  further research.  41  for  At this point, there is positive evidence for deglottalization before the causative morpheme. This area of the phonology is most complex (cf. also discussion in Watanabe (2000)), and requires further research (cf. Blake in prep). Glottal Restmcturing Glottalized obstruents /OV are realized as [?0'] in the environment after a stressed schwa in order to prevent stressed schwa from occurring in a stressed open syllable (Blake (1995, 1999), Urbanczyk (1999), and §5.1). In the data below, the reader will also note the regular realization of schwa as lowered to [a] before a glottal. (55) Input  Output  Gloss  a. xt^-t  X9?t' -t  xa?.  weigh it  a'. xt' -t  X9?t' -[9]t  xa?. Pat*  weigh it  c. xt' -t-'ut cn  xa?t' -tut can  xaW . tot. cin  I weighed it  e  6  e  e  6  d. xt^-t-'ut a cx  w  X3?1:' -tut a cx e  w  xaVf . to . tse . cx 3  w  Did you weigh it?  Chapter 5.1-5.2 provides detailed discussion of the surface realization of both glottalized obstruents and glottalized resonants.  The data in (56) shows that glottal insertion [?] also marks a very limited number of diminutive forms. (56) Output  Input sattx  a. sattx  woman  sattw ~ sattx  w  w  Gloss w  Uttiegirl  a'. sa[?]ttx  sa?ttx  sa? ttx ~sa?ttx  b. wiwlus  wiwlus  wiwlos  young man  b'. wiwlus+[cgl]  wiwlus  we?w los  young man at puberty  w  w  a  w  9  42  w Floating feature: constricted glottis [cglj Consider the realization of the Past Tense marker /-[cgl]ui/ in (57-60). What we observe from a comparison of the morphologically related forms in (57) is that the past tense morpheme /-[cgl]ui/ (written as //-?uL// by Watanabe (2000: 306)) systematically causes glottalization of a preceding resonant, as in (57.a'-d') Column 2, but fails to affect a precedingfricative(58.a'-d') or stop (59). Since glottalized resonants are not permitted in syllable-initial position, they are restructured as ?R between vowels (cf. Blake 1992, 1995, 1999 and §5.2). If this restruchrring cannot occur, then association of the floating [cgl] feature is blocked, given the high-ranking constraint which blocks glottalized resonants from Onset position (*R70nsef). Examples of stem-final resonants which are not glottalized include: /xpj-[cgl]ui can/ x6pjui can [xap^vt cen] I turned back (cf. /xpj/Xopi turnback).  (57) preceding Resonant Output  Gloss  k anewsAm  rest (-igs body)  a*. k n=igs-m-[cgl]ui a cx k aniwsamuiacx  ^anewsaTmotaecx*  Did you (sg) rest?  b. k tus-m ga  k t5tos9m gA  turn around (request)  Input k aniwsam  a. k n=igs-m  w  w  w  w  w  w  w  k 3 t u s a m ga  w  w  b\ k tus-m-[cgl]ut a c x w  w  k 9tusarhuia5x w  w  w  3  k 'utosa? mo ta3cx  Did you turn around?  w  ^  <  w,,  c. ns-m  nasam  nfsyam  swim  c\ IMP-ns-m-[cgl]ut can  nansamui can  ninsajT^motctn  I was swimming  d. sp=iq an  sa?piq an  sa?peq An  get hit on the head  S9?piq ariu1: c  sa?peq a? notc  w  w  d\ sp=iq an-[cgl]u i c w  ,  w  w  w  43  3  h  I got hit on the head  (58) preceding Fricative Input  Output  Gloss  a. IMP-gay-t-as  gagayatas  gXgaye-tAS  he's asking them  a'. gay-t-as-[cgl]ut  gayatasui  gayetasof  he asked him  b. k ay-as ga  k ayas ga  k ayis gA  hide it (request)  b\ k ay-as-as- [cgl]ui  k ayasasui  k aytsesot  he hid it  c. ic-t ga  tact ga  itct  cut it (request)  c\ ic-t-as-[cgl]ui  fd^tasui  iic^tasot  he (already) cut it  d. tk -t ga  tak t ga  t u l t gA  pull it (request)  d'. IMP-tk -t-as-[cgl]ut:  t9tk atasul  t3tk 9tasoi  he was pulling it  w  w  w  w  w  w  w  w  ga  3  w  w  w  9  3  w  The [cgl] feature associated with the Past Tense morpheme does not typically glottalize a preceding stop, as shown by the data in (59).  (59) preceding Stop Input  Output  Gloss  a. IMP-gay-t-as  gagayatas  gAgayfe-tAS  he's asking them  a'. gay-t-[cgl]ui c  gayatuf c  gayetoic^  I asked him  b. j t k - t  J9tk at  Jitk  shake it  w  w  b\ Jtk -t-[cgl]ui a cx w  1  w  jatk atu4acx w  w  w  At  h  Jitk atoiaxx w  w  wipe it  c. t ^ - t c\ t^-t-tcgljut a cx  Did you shake it?  w  f ok tutacx 8  w  w  f t5k toiaxx e  w  w  Did you wipe it?  The fact that glottalization is not always present in the Output suggests that it lacks segmental status, i.e. a root node. It behaves phonologically like afloatingglottal feature since it requires an ehgible host in order to be realized. Thefloatingfeature is represented here as [cgl], but is represented elsewhere in parentheses (?) to indicate that it does not always have a surface manifestation (cf. Watanabe 2000, for example).  44  When the past morpheme occurs after a vowel-final  lect suffix, it is often realized as -h-uf, as in  (60.a). (60.a) Input  Output  Gloss  a. K n-ei-[cgl]ut c  k 3(n)f3ihui c  k t50ehoic  b. pap-?[i]m-0i-[cgl]u'l c  pap?im0ihuf c  pap?Em6ehoic  pap?imtanapihuic  papTemtanapehoic  w  b. pap-?[i]m-t-anapi-[cgl]ut  w  w  c  I looked at you  h  I fixed it for you (sg)  h  11  I fixed it for you (pi)  I hypothesize that the intervocalic [h] is epenthetic - since it is the default consonant in intervocalic position (cf. § on h-epenthesis and § on the featural representation o f /hi). Since the floating ? only targets a resonant (consonants specified as [son]), it is subject to deletion in this context as w e l l . 7 Epenthetic Consonant [h] The consonant [h] is epenthesized between a vowel-final Root and before a vowel-initial Lexical Suffix. The epenthetic laryngeal [h] appears in square brackets in each o f the following examples. See §5.4 for further discussion o f the resolution o f vowel hiatus ( V - V sequences) in Sliammon.  (61) h-epenthesis Output  Input a.  lamatu=uk t w  lamatu[h]uk t  lamatu[h]uk t  lamatu  lamato  w  a', lamatu  Gloss w  sheep's wool, sweater sheep  Watanabe (p.c.) records two cases in which a vowel-final Root/LS takes -?ut rather than -h-ut: ?ui:q u-?ui and not w  7  *7uiq u-h-ut, and k as=uja-?ui not *k asuja-h-ul w  w  w  45  b. tala=awus=tn  tala[h]awust9n  tala[h]awust9n  eye glasses  b\ tala  tala  tala ~ taL\  money  c. kapi=aya  kapi[h]aya  k api[h]aye  coffee pot  c\ kapi  kapi  k api ~ kdpi  coffee  d. hrhu^ay  h9?mu[h]ay  hd?mo[h]Ay  cascara bark  d\ hriiu  h9?mu  hd?mo  pigeon  y  y  2.2.4 Obstnient/Glide/Vowel Alternations This section summarizes the Vowel/Glide/Obstruent alternations (/j7 Q ~ y ~ i/e ~ c], and /g/ [g ~ w ~ u/o ~ k ~ x ]) in Sliammon. Although these alternations have been discussed by w  previous scholars including Sapir (1915), Davis (1970), Hagege (1981), Kroeber (1989), Blake (1992, 1995), and Watanabe (1994, 2000), there is no agreement as to whether the corresponding underlying "segments" are obstruents or resonants. In order to emphasize the resonant behaviour of these sounds, in Blake (1992) I used the archi-phonemes /Y, Y ' , W, W7, distinct from /y> y> w, w/, as an abbreviation for the feature matrices for what in the present work are represented as /j, j, g, g/.  I propose here that the lack of agreement /j/Y/ or /g/W/ is symptomatic of  phonological theories which treat "segments" as primitives (cf. Archangeli and PuUeyblank 1994). I claim that the sets of features which show these surface alternations in Sliammon are a set of conflicting features.  Consider first the variant phonetic realizations of each of these sets of  alternations given in (62-63). Descriptively, [y] occurs either before another consonant, or at the end of a word. [T] occurs in pre-vocalic position, and the phoneme /j/ vocahzes to i ~ e (depending on the C-context) when it occupies the nucleus of a syllable, [c] is the surface realization when it occurs in a wordinternal coda followed by [t].  In summary, this segment  appears on the surface as  r j ~ £ ~ y ~ i ~ ] neutralizing in the appropriate contexts with /y, 2, i/. e  5  46  (62) iy Input  Output  Gloss  a. huT-it  ra  [ho'-Jit]  ready  b. huj  [y]  [hoy]  stop, finish  c. haj-haj-i-i  [y,j]  [hayhajii]  everybody's flirting  d. haji-am+fcgl]  [y]  [hayiAm]  flirt  e. ?J-?j=umis  [ij]  [?i?ajumis]  very beautiful  PI  [tic .ta?. je?.jts]  cheeks  f.  tj-taj-aj=us  The data in (63) shows that [w] occurs before another consonant, and that it alternates with [g] in prevocalic position . The phoneme /g/ vocalizes when it occupies the nucleus of a syllable, and it surfaces as [x ] in word-final position, [k] is the surface realization of /g/ when it occurs in w  a word-internal coda position followed by a voiceless non-continuant.  This appears to be in  keeping with the generalization the obstruent clusters agree in voicing (lack of feature [sonorant]). The segment /g/ surfaces as [g ~ k ~ x ~ w ~ u ~ o], neutralizing in the appropriate contexts with w  underlying /k, x , w, u/. w  (63) Input  Output  Gloss  a. hig=us  [g]  hegus  chief, rich in old way  b. DIM-C3CpL-hig=us  [w]  hehawhegvs  small chiefs  c. mga  [g]  mXg-9  cougar  d. mag-mga  [w]  mdwmXg-9  cougars  e. C C -iagf  [u,g]  •fuiiagit'  herring (pi)  f.  [x ]  k dnax  see him/her  3  PL  k n-ng w  w  w  47  w  Two points are crucial to understanding these segments. First, following Blake (1992) the Vowel/Glide/Obstruent alternations are governed by the prosodic organization of the syllable (cf. Blake (1992, 1995) and §3.2 on syllable structure in Shammon). (64) [j, g]  in syllable onset (non-moraic) position  [y, w] in syllable coda (moraic) position [i, u]  in syllable nucleus (Nucp) position  Second, the defining features for each of the surface variants of /J, g/ are given in (65). Their glottalized counterparts are identical, with the addition of LAR[cgl]. Their variant surface realizations result from the systematic non-realization of one (or, in the case of [x™], two) of their underlying features in a particular context. Namely: the feature [-cont] is not realized in moraic position (with the exception of c), and the feature [son] is lost in word-final position in the case of [x ]. The angled brackets are used here to indicate a feature which is not realized on the surface: w  <ra>. (65)  Distribution of f-elements for /J/  Distribution of f-elements for /g/  [J] - [son, -cont, DOR hi, -bk]  [g] - [son, -cont, DOR hi]; <LAB[rd]>  [i] - [son, DOR hi, -bk]; <[-cont]>  [u] - [son, DOR hi, LAB rd]; <[-cont]>  [y] - [son, DOR hi, -bk]; <[-cont]>  [w] - [son, DOR hi, LAB rd]; <[-cont]>  [c] - [-cont, DOR hi, -bk]; <[son]>  [x ] - [DOR hi, LAB rd]; <[son, -cont]> w  In accordance with Optimality Theory (henceforth OT), the loss of features is driven by the interaction of conflicting constraints which ensure the creation of optimal prosodic constituents (Prince & Smolensky 1993, McCarthy & Prince 1994, Kirchner 1995, among others). In the case of the Shammon data in question, a high ranking feature co-occurrence constraint drives the underpaying of lexically distinctive features. Align L ([-cont], o) determines which features are  48  underparsed, thus creating both optimal onset, and optimal coda constituents, as proposed in Blake (1995). Evidence that /j, g/ are specified as [son] is provided by the fact that they undergo glottalization, along with the other resonants /m, n, 1, y, w/ in the system, as argued in Blake (1992, 1995). Alternations between [g ~ w ~ x ] from /g/ provide evidence that the feature [son] is w  subject to deletion, and therefore provide evidence that [son] has autosegmental properties and is crucially not an "integral" part of the root node (contra McCarthy (1988), for example).  This  provides evidence for the position of the feature [sonorant] as a dependent of the Root node within the Feature Geometry which will be presented in §2.3.2.  2.2.5 Laterals/L, LV Following Blake (1992), the symbols PL, LV are used to represent sets of features which are realized as [y ~ w ~ f] and [y ~ w ~ f ~ ?] respectively. The distribution of each of these surface variants is dependent upon its position within the word, and on the quality of the adjacent vowel. The proposed status of /L, LV is motivated by the following morpho-phonemic alternations in (66-67), and by the fact that the independent phonemes /!/, IV , lyl, /w/ do not undergo these alternations.  (66) ILI Input  Output  Gloss  a. nx iL  nax tt  niix it  dugout canoe  b. nx iL-s  nax iy-s  nvx iys  his canoe  b'. nx iL-it  nax iy-it  ni5x iyit  their canoe  b". nx iL-ma  nax iy-ma  nux iyma  travel by canoe  c. t^amqn.  t' amq i  t' amq 4  cloud  w  w  w  e  c'. t' am-t' amq L-[i]m w  h  w  w  w  e  w  w  w  e  w  w  w  e  w<  w  t' am-t' amq [V]yim t^amt^amq^yim e  e  W  49  it's foggy  (67)  IVI  Input  Output  Gloss  a. paLVpaL'a  pa?a  pa?a  one  a'. saL'/saL'a  sa?a  sa?a  two  b. paL'=agif  pa?agrt  pa?agii  one canoe  b'. saL'^agrf  sa?agii  sa?agii  two canoes  c. paL'=us  pawus  paw?us ~ paw?vs  one round object  c'. saL'^us  sawus  sawus  two dollars  c". DIM-saL'=us  sa-swus  sasu?t)s  two sm. round objects  d. paL'-paL'  pay-pa?a  pe:pa?a  one person  d\ saL'-saL'  say-sa?a  se:sa?a  two people  e. paL'=lawi  pay=lawi?  paylawe?  one bottle (=lawi bottle)  e'. saL'=lawi  say=lawi?  saylawe?  two bottles  8  check  As observed from the data in (67), IVI becomes \ \ \ in word-final position. Word-internally it becomes [w] in the context of the round vowel /u/, [?] when it occurs in a total PHAR context (i.e. between a's), and [y] elsewhere. Notice that in (67.e-e'), IVI becomes [y] before the coronal lateral IV even though it is preceded by the vowel [a]. This reinforces that fact that the glottal [?] from IVI occurs in a total PHAR context. See also Appendix IV for further examples of the contrasts between IL, VI and /y, y , w, w, t/.  determining the underlying representation for the Roots one and two is somewhat problematic. These Roots often behave as though they are consonant-final. For example, they do not induce h-epenthesis with the addition of a vowel-initial LS; however, if the Roots are of the shape paL' and saL', then one might expect pat and sat when L ' occurs in word-final position. This issue is not resolved here; it remains a topic for further research.  50  2.2.6 Geminate Consonants The resonants [n, m, 1, y, w] are lengthened intervocahcally, as shown by the data in (68). (68) Resonants a. IMP-ix-m  faixAm  ia-ix-am  a'. IMP-"fx-m a tQ nx iL ia-fx-ama t0n9x it w  w  leaking  i3tx9m:at 0ni5x :it Isyr. boat leaking? 3  w  Obstruents are lengthened in intervocahc position after an initial stressed schwa, as in (69). (69) Obstruents Input  Output  Gloss  a. nx iL  nox ii  nvx :ii  dugout canoe  a'. nx iL-ma  n9x iy-ma  ntfx :i-ma?  travel by canoe  b. t' xu  t'9xu  t' ox :o  ling cod (fish)  b'. DIM-t^xu+t?]  t^i-t^xu?  t' it' x o?o  small ling cod  w  w  e  w  w  e  w  w  e  w  e  e  w  Gemination of intervocahc consonants after stressed schwa will be discussed in greater detail in §5.3. 2.2.7 Consonant Deletion The following deletion processes also affect the surface realization of consonants. Identical Consonants Davis (1970: 42) documents the fact that identical consonants CiCi generally reduce to a single instance of that consonant (Q). The data in (45-46) also show reduction of m-m to a single instance of m, as m lh-a-lm==m'ml catqamin [cttqamen] knife. A principled exception to this generalization is presented in (70) and involves Root-final -t followed by the -t of the Control Transitivizing suffix. (70) a. C9t-t-as  cartas  [cit tAS ~ cit . tas] 9  51  h  he cut it  Notice that if consonant deletion were to occur in this context yielding [£atAs], it would have the effect of leaving schwa in a stressed open syllable. As will be argued in Chapter 5, stressed schwa in an open syllable is systematically avoided, if possible. Coronal Deletion SUammon also has a number of consonant deletion processes which involve coronal consonants: t, n, t deleting before other coronal consonants. Consonants which undergo deletion are parenthesized (C) in column 2. t-deletion The data in (71) shows that the t-transitivizer undergoes deletion in the environment before a following c. (71) t-deletion t-deletion  Output  a. gq -t ga  g q t ga[?]  gdq t ga?~goq t  drag it (request)  b. gq -t-as  g9q tas  gdq tas ~ goq . tss  s/he drags it  c. gq -t can  gaq (t) can  gdq cm ~ gdq . cen  Tm dragging it  goq . cex  you drag it  Input  w  w  9  w  w  d. gq -t cax w  w  w  w  w  w  w  Gloss  w  gaq (t) cax w  w  w  w  w  w n-deletion Nasal n-deletion is illustrated by comparing related forms of the LS =iq an fop of head, w  high point. The LS is n-final, as shown in (72). (72) Input  Output  Gloss  a. tih=iq an  tihiq an  tiheq An  big head  b. ^ip=iq an  t' ipiq an  t' epe q An  pointed head  w  w  w  e  w  e  w  3  w  c. tf -[i]m=iq an  ta?^imiq an  ta?^emiq An  red head (red hair)  d. sp=iq an-?[i]m=min  s3?piq an?unin  sa?peq An?emin  fish club  9  w  w  w  w  w  w  52  When the coronal resonant n precedes either a t or 0, it fails to surface, as shown by the morphologically related forms in (73) (see Watanabe (2000) for some lexical exceptions). Example (73.b) shows that deletion of In/ before /0/ results in compensatory lengthening of the preceding full vowel. As discussed in §, compensatory lengthening occurs in stressed syllables and has not been documented in unstressed syllables.  (73) Input  n-deletion  Output  Gloss  a. xim=iq an-t-m  xnniq a(n)t3m  xemeq^tam  get clawed in the head  b. sp=iq an-0-as  ss?piq a(n)0as  sa?p£q a:0AS  he hit me on the head  c. x ulk -ay=iq an=tn  x ulk ayiq a(n)tan  x uUc ay£q at9n  hair ribbon  w  w  w  w  w  w  w  w  w  w  w  w  w  w i-deletion The example in.(74.a) shows that the past tense marker /-'ui/ is i-final (cf. § on other realizations of the past tense marker). The data in (74.b) shows that i deletes before -s. Both the past marker /-'ut/ and the third person possessive marker -s are within the affixal domain. Rootfinal i is retained before an s-initial suffix, as in 2tai-sx (Watanabe p.c). w  (74) i-deletion Input  i-deletion  a. gay-t-[cgl]ui c  Output  Gloss  gayftdic  I asked him  b. mna-?ui-s  ma?na-?u(i)-s  ma? na?6s  his/her child (former)  b'. mna  ma?na  ma?na  one's offspring, child  3  53  2.3 Theoretical Assumptions 2.3.1 Features The next section provides a general introduction to the phonolgical features which identify natural classes of sounds which pattern together in the language.  These features are listed in (75) and  discussed in detail in the following sections.  (75) Features Continuant [-cont] Sonorant [son] (all resonants) Consonant [-cons] (vowels and ghdes) Laryngeal: constricted glottis [LAR[cgl]] Labial (LAB) (primary place: labials) Round [rd] (labialized consonants: labio-velar and labio-uvulars) Coronal (COR) (interdentals, coronals, laterals) Dorsal (DOR) (alveopalatals, palatals, plain velars, uvulars) High [hi] (alveopalatals, palatals, velars) Back [-back] (alveopalatals, palatals) Low [lo] (low vowel /a/) Pharyngeal (PHAR) (post-velars: uvulars and laryngeals) Lateral (lat): X, X, f, 1, f, L, L' Nasal (nas) Distributed (dist)  2.3.2 Feature Geometry The phonological features in (75) are represented in the articulator-based Feature Geometry presented in (76), following earlier proposals that features are hierarchically ordered (cf. Sagey (1986), McCarthy (1988), Halle (1992), Archangeli and PuUeyblank (1994), amongst others). I have included only the part of the geometry which is relevant for characterizing the Shammon data 54  presented in this thesis. This is essentially the same model argued for in Blake (1992:8-9). The features as well as articulator nodes are privative - they are either present or absent from the representation.  (76)  Feature Geometry  RNo (Root Node) [-cons  [rd]  [dist] [-bk][hi][lo] Representation of Shammon Consonants The  following  sections  (§ l-§  provide  the  feature  geometric  representations which I am assuming for each of the consonants in the language (cf. also Blake (1992)). The representation of each of the consonants is central to an understanding of the consonant-vowel (C-V) interaction discussed in §2.4. The adjacent consonants also determine the surface realization of schwa which will be discussed in §2.4.4. Labials Labial consonants are specified as LAB and not as LAB[rd] since they do not seem to exert a rounding effect on a preceding schwa unlike velar and uvular consonants which have secondary  55  labialization.  The presence of the feature [rd] entails the existence of the Dorsal node  ([rd] D D O R ) . Geometry  P  P  m  m  RN  o  o  o  o  [son]  [son]  [son]  [nas]  [nas]  [nas]  [-cont]  [-cont]  [-cont] LN  [cgl]  0  [cgl]  [cgl]  o  o  0  o  LAB  LAB  LAB  LAB  PN LAB  o Coronals and Laterals The class of coronals includes dentals and alveolars. The dental consonants t and t' and the e  e  interdental fricative 0 are distinguished from the other coronals in the system by the nature of their consonant release; they are overtly marked as [distributed], as in (78).  (78) Coronals 0  t  (  s  n  n  o  o  o  o  o  0  [son]  [son]  [son]  [nas]  [nas]  [nas]  Geometry  t  RN  0  [-cont]  e  [-cont]  LN  [cgi]  f  e  o  [-cont]  [-cont]  [-cont]  o  O  O  [cgl]  [cgl]  [cgl]  o  o  0  0  o  o  o  o  COR  COR  COR  COR  COR  COR  COR  COR  COR  [dist]  [dist]  [dist]  [dist]  PN  56  (79) Laterals  Geometry  X  t  RN  0  o  o  fsonl [-cont] Rati LN  [-cont]  [-cont]  [lat]  Tlatl  Rat]  1  f  L  L'  o  0  o  0  [son]  [son]  [son]  [son]  [lat]  [latl  Hat]  [lat]  o  o  o  rcgii  rcgii  [cgl]  rcgii PN  o  0  o  0  o  0  o  COR  COR  COR  COR  COR  COR  COR  COR  DOR  DOR  DOR  [hi]  [hi]  rhii Alveopalatals The alveopalatals in (80) are marked as DOR[hi, -bk] following Blake (1992). Motivation for their DOR[hi, -bk] specification rather than a featural representation such as COR DORfhi] comes from the fact that alveopalatals front /a/ to [e] (see discussion of V-features, §2.4), whereas the class of coronal consonants do not. Alveopalatals are specified as [hi] since they affect the height of the non-low vowel  or /u/.  (80)  Geometry  5  c  s  T  I  y  y  RN  o  o  o  0  0  o  o  [son]  [son]  [son]  [son]  [-cont]  [-cont]  [son] [-cont]  [-cont]  [-cont]  >  Tlatl LN [cgll PN  0  o  o  [cgll  [cgll  [cgll  0  o  0  o  o  0  0  DOR  DOR  DOR  DOR  DOR  DOR  DOR  DOR  [Ml  [Ml  rMi  [Ml  [M]  rMi  [Ml  rMi  [-bk]  [-bk]  f-bkl  [-bk]  [-bk]  [-bk]  [-bk]  [-bk]  57 Velars Both the plain velars and the labio-velars determine the height of adjacent vowels, as will be discussed in §2.4.3-§2.4.4. These consonants are therefore marked as DORfhi]. (81)  Geometry  k  k  RN  o  o  k  w  0  k  w  0  x  o  [son] f-cont]  [-cont]  LN  rcgii PN  [-cont]  [-cont]  »  [-cont]  5  g  w  W  o  o  o  o  [son]  [son]  rson]  [son]  [-cont]  T-cont]  w  0  0  0  o  [cgll  [cgl]  [cgll  [cgll  o  o  0  o  o  0  o  LAB  LAB  LAB  LAB  LAB  LAB  LAB  LAB  Trdl  [rd]  Nl  [rdl  N]  [rd]  [rdl  Nl  DOR  DOR  DOR  DOR  DOR  [hi]  [hi]  [hi]  [hi]  o  o  DOR  DOR  DOR  DOR  DOR  [hi]  [hi]  H  [hi]  [hil Post-Velars: Uvulars and Laryngeals The uvulars and laryngeals are represented as in (82). Uvulars are characterized here as complex DOR PHAR whereas laryngeals are specified as PHAR, following Cole (1987), McCarthy (1991), Shaw (1991), amongst others.  58  (82) Uvulars and Laryngeals  Geometry  q  q  J  X  RN  o  o  0  [-cont]  [-cont]  q  q  w  X  w  w  o  ?  h  o  0  0  0  [-cont]  [-cont]  [-cont]  0  o  o  [cgl]  [cgl]  [cgl]  [son] [-cont] LN  [cgl]  0  0  o  LAB  LAB  LAB  LAB  [rd]  [rd]  [rd]  [rd]  o  PN  o  DOR  DOR  DOR  DOR  DOR  DOR  PHAR  PHAR  PHAR  PHAR  PHAR  PHAR  DOR PHAR  o  o  o  PHAR  PHAR  There is some difficulty in determining the appropriate representation of [h] in Sliammon. Schwa colouration before the consonants ? and h is not perfectly symmetrical.  Schwa systematically  lowers to [d] before the glottal [?], as shown by the data in (83.a-e) and discussed by Kroeber (1989). The diminutive examples in (83.c'-e') are provided in order to provide morphological evidence for the weak roots (CC) posited in (83.c-e). (83) Input ms?t  a. m?-t  Output  Gloss  md?t  get s.t.  h  b. sp=iq an  sa?piq an  sd?peq An  get hit on the head  c. X>fum  3b?ium  XdWom  wolf  c'. DIM-A:?4um=u<r  Xi-X9?'fumu i  XiXa?t6mi)4  young wolf, wolf cub  d. mq t'  m9?q t'  md?q f  onion, wild onion  w  w  w  w  <  w  e  w  e  9  d*. DIM-mq t' +[i]  mi-m3?q [i]t'  m8?ma?q et'  small onion  e. k nay  k 9?nay  k d? nAy  cover, lid  k i-k nay  k ik nAy ~ k tk nAy small lid  w  e  w  w  w  e'. DIM-k nay+[?] w  w  w  e  w  w  w  59  e  3  w  w  w  Notice, however, that schwa varies between [A ~ d] before h, as shown in (84).  This type of  variation is not recorded before [?].  (84) schwa before h Input  Output  Gloss  a. qh-t  qaht  q/ht ~ qdht  lift s.t.  a'. qh-t-'ui can  qahtui can  qXhtot cen  I lifted it up  a". qh-t can sm  qah(t) can sam  qdh cen sam  Tm gonna lift it up  Oah-t  eATit~6dht  prop s.t. up  b'. hu ga 6h-t  hu ga 6aht  hd gA eXht  go prop it up  b". 0h-?m=min  09h-?amin  0Ah?Amen  b.  eh-t  h  centre pole for tent  The data in (85) provides evidence that lil is retracted and lowered to [e] in the environment of [h]; therefore, /h/ is marked as PHAR. The different effects on schwa before /?/ may be attributed to the other marked properties of/?/. (85) Input  Output  Gloss paddle  a. hiw=cis  hiwcis  hewcis  a'. IMP-hiw=cis-ma  hi-hiwcisma  hehewclsma  get there by paddling  The fact that [h] is the epenthetic (Onset) consonant in the language provides support for its default representation; it is the least specified consonant in the system.  60  2.4 Vowel System The goal of the remainder of this chapter is to present the phonetic vowel inventory and show how the surface vowels are derived from the four vowel system: i, u, a, 9 . Schwa is treated as an epenthetic non-moraic Nucleus. Its prosodic distribution and properties will be discussed in more detail in Chapters 3-6 but its phonetic realization is addressed here. §2.4.2 provides evidence for a three way weight contrast in the language: schwa which is proposed to be non-moraic, full vowels which are moraic, and long vowels which are bimoraic, and are derived via Compensatory Lengthening. The hypothesis put forward here is that this weight contrast in Sliammon is encoded phonologically in terms of moraic structure, following the generalizations originally made in Blake (1992), and recast within the Nuclear Moraic Model of Shaw (1993 et seq.). §2.4.3 discusses the full vowels /i, u, a/ and the consonant-vowel interaction which accounts for the variant surface realization of each of these vowels. §2.4.4 documents the effects of adjacent consonants (and vowels) on the surface realization of schwa. §2.4.5 introduces the issue of Full Vowel Reduction which occurs in unstressed syllables; this is discussed further in Chapter 4.  2.4.1 Vowel Inventory Sliammon has a large number of phonetic vowels, as shown by the inventory in (86), following Davis (1971), Blake (1992), Watanabe (1994). (86) i  i t  i  u v o  e £  o  9  (ae)  A  a  a  61  The surface inventory includes tense and lax variants, front-central-back and hi-mid-low realizations. As shown by (86), there are no front rounded vowels nor are there any low rounded vowels in the system (i.e. [o] is treated as non-low). The phonetic vowels in (86) are allophones of a vowel system based on three underlying contrasts /i, u, al plus an epenthetic default vowel "schwa". The vowel contrast in Sliammon is low/non-low distinction (cf. Blake 1992). The non-low vowels Ii, ul are most often realized as [e, o] respectively. The allophones of the low vowel lal range from [e (a?) ~ a ~ a] and depend on the place of articulation of the adjacent consonants. I have changed my former usage (Blake (1992)) of le, o, a/ as the basic phonemic symbols to adopt Ii, u, al, in order to minimize phonemic transcription differences between authors writing on SUammon, and to also make it easier for those wishing to do comparative research in SaUsh. The underlying representation for each of the Full 1  vowels Ii, u, al is presented in (87).  (87) i, u, a i  Vo | \ [-cons]  Vo | \ [-cons]  i  i  PN  PN /\ LAB DOR [rd]  1 |  DOR [-bk]  a  u  Vo | \ [-cons]  1  PN  | DOR  [lo]  The surface height of the non-low vowels Ii, xxi is determined by the height of adjacent consonants, via consonant-vowel (C-V) feature sharing. The phonemes Ii, u/ are [i, u] next to alveopalatals/palatals, whereas they are [e, o] in a "neutral" context, and are retracted to [e, o] in the environment of PHAR consonants. The vowel system of SUammon has received a fair bit of  ^nkade (1997: 212, fh.l) makes a similar point with respect to the phonemic vowel inventory of Upper Chehalis.  62  discussion in J.Davis (1970, 1971), Kroeber (1989), Blake (1992, 1999), Watanabe (1994). The featural specification of the allophones of the full vowels is provided in (88).  m_  ,  fx!  RN  ,  /a/  Al/  M  [l  M  M  [Al  o  0  o  0  o  o  e  ,  M  M  [°1  0  o  0  o  f-cons] [-cons] [-cons] [-cons] [-cons] [-cons] [-cons] [-cons] [-cons] [-cons] [-cons] o  o  o  LAB  LAB  LAB  LAB  Trdl  [rd]  [rd]  [rd]  DOR  DOR  DOR  PN  DOR  DOR  [nil  fhil  r-bki  o  0  [-bk]  DOR  o  DOR  o  DOR  o  0  DOR  DOR  0  DOR  [hi] [-bk]  [-bk]  [-bk]  rioi  Po]  [lo]  (PHAR)  PHAR  [lo] PHAR  PHAR  PHAR  PHAR  The basic featural identity of Ixl is DOR [-bk]. Its variant realizations are outlined in (89). (89) Ixl is realized as [i] in the environment of a [hi] consonant (alveo-palatals and velars).  Ixl is realized as  [E]  in the environment of a post-velar consonant (uvulars & laryngeals), i.e. PHAR.  Ixl is realized as [E] in the environment of glottalized consonants. Ixl is realized as [e] elsewhere. Ixl is laxed to [i] in unstressed position in the environment of a [hi] consonant. Ixl is laxed to [E] in other unstressed contexts.  63  The basic featural identity of IvJ is LAB [rd] DOR. Its variant realizations are outlined in (90). (90) IvJ is realized as [u] in the environment of a [hi] consonant (alveo-palatals and velars). IvJ is realized as [o] in the environment of a post-velar consonant (uvulars & laryngeals), i.e. PHAR.  IvJ is realized as [o] in the environment of glottalized consonants. IvJ is realized as [o] elsewhere. IvJ is laxed to [v] in unstressed position in the environment of a [hi] consonant. IvJ is laxed to [o] in other unstressed contexts.  The basic featural identity of lal is DOR [lo]. Its variant reahzations are outlined in (91). (91) lal is realized as [e] after a [-back] consonant (alveo-palatals/palatals). lal is realized as [a ~ ae] in the environment of a non-sonorant lateral (X, X\ 4). lal is realized as [a] in the environment of post-velars (uvulars and laryngeals), i.e. PHAR.  lal varies between [a ~ A] in the environment of anterior consonants (coronals and labials). lal is realized as [a] elsewhere. lal is laxed to [A] in unstressed post-tonic position.  The allophones of schwa are given in (92), and discussed in further detail in §2.4.4. (92) schwa is realized as [v] in the environment of a tautosyllabic [hi, rd] consonant (labio-velars) schwa is realized as [o] (rounded midback V) before a non-high [rd] C (labio-uvular) schwa is realized as [t] in the environment of a [hi, -bk] consonant (alveo-palatal/palatal) schwa is realized as [e] between a [-back] consonant and a PHAR consonant schwa is realized as [i ~ t] in the environment of a [hi] consonant (plain velars) schwa is realized as [A] is the environment of a plain uvular, i.e. DOR PHAR schwa is realized as [a] in the environment of laryngeals (?, h), i.e. PHAR and LAR. 64  Sliammon also has a limited number of surface long vowels (i.e. [i:, e:, E:, U:, O:, a:]) which are all derived through compensatory lengthening, here treated as the loss of a moraic coda consonant. Analyses of how surface long vowels are derived are presented in detail in J.Davis (1970: 52-56), and Blake (1992, Chapter 3), and due to space limitations will not be discussed further here. The primary focus of the remainder of this chapter is on the weight contrast between schwa and the full vowels /i, u, a/, and their respective surface phonetic realizations. 2.4.2 Vowel Quantity The next section explores the evidence for a distinction in vowel quantity or weight, as represented by the prosodic constituant "mora". Sliammon exhibits a weight contrast between the allophones of schwa, and the allophones of the full vowels Ii, u, a/. Phonological Weight Contrast Phonetically the full vowels Ix, u, al are half-long [v] in stressed open syllables, as shown by the data in (93). This point is documented independently by Watanabe (p.c). Notice that the stressed vowel in (93.a) is also recorded with an off-glide [ei]. Diphthongization is another diagnostic for the constrast between schwa on the one hand, and the full vowels Ii, u, al on the other hand. (93) Output  Input  Gloss  a. pilaq  pilaq  pe-L\q (~ peiL\q )  bracket fungus  b. ?imin  ?imin  ?E-mm  door  c. qiga©  qiga6  qe-gA0  deer  d. k upa  k upa  k u-pa ~ k i i - p A  grandfather, grampa  e. ?upan  ?upan  ?d-pAn  ten  f. xaX-ng-mi c  xaXnumic  xdXno-mic  I love you  g- tapas  tapas  ta-pAS  cave  h. t V i q  f^at'iq  t^a-t'eq  a drop of water  ?amamu?  ?d-mamo?  chiton  w  i.  ?amamu?  w  h  w  65  h  w  Schwa, on the other hand, is noticeably shorter in duration, and surfaces consistently as a brief lax vowel, as in (94). (94) Input a. b.  tt et  tat  e^t  straight  co?omex tan  flooring, carpet  f^mtn  t^amtn  breast  saqt  s/qt  peel it off  m9?t  md?t  get it  C3?umix t9n  e. t'm=tn  w  g- m?-t  rain  eat it  d. c7=umix =tn  sq-t  tii ect  w  w  f.  Gloss  mak -t  c. mk -t  e  Output  w  2  w  Kroeber (1989: 108) also observes that schwa and the surface variants of schwa are generally "lax and a bit shorter than the allophones of non-schwa vowels, at least in stressed open syllables." Theoretical Assumptions The main purpose of the next section is to put forth a proposal which captures this observed durational contrast. The question which this observation raises is how is this length contrast encoded phonologically? Blake (1992) proposes that schwa and the following moraic coda consonant share a mora resulting in a mono-moraic syllable whereas a M l vowel and following moraic coda consonant are both moraic resulting in a bi-moraic syllable. This entails a syntagmatic rule of weight-by-position which is sensitive to whether or not the coda consonant is preceded by a full vowel or by schwa.  2  The labialization of schwa to [D] in this context comes from the full vowel /u/ in the following syllable. This is  an instance of translaryngeal harmony (see §  The root cs?- be on top of surfaces as [ce?-] in other  phonological contexts, as shown by forms like /£?=na5=tn/ [cE?nA6tm]  66  small blanket to sit on. Representation of Weight Contrast: Nuclear Moraic Model In this dissertation, this leading idea is recast within the Nuclear Moraic Model of Shaw (1993, 1994, 1996c). Within this model, schwa is Nuclear, and non-moraic.  The prosodic  representation for schwa is given in (95).  (95)  Nuc  The hypothesis put forward within the Nuclear Moraic Model is that schwa is weightless. Since it lacks phonological weight (i.e. a mora), the fact that it is perceptually shorter in duration is encoded in its phonological representation. In addition to the prosodic representation of this vowel (i.e. a bare nucleus), it is also claimed here that schwa lacks inherent phonological features and as such is subject to colouration by adjacent consonants (and vocalic nuclei), as argued in Blake (1992:3542). The range of phonetic colouration of schwa is sketched in §2.4.1 above, and explored in greater detail in §2.4.4.  Full vowels, on the other hand, are represented as both Nuclear and moraic, as in (96). The full vowels /i, u, a/ each dominate specific vocalic place features as specified in (88), and schematically represented here as [fj.  (96) Nuc  I [f]  67  The fact that schwa is shorter than the full vowels Ii, u, al is encoded here by a difference in moraic structure. The proposal that schwa is non-moraic in Sliammon is supported by the phonological behaviour of CaC weak roots versus CAC Strong Roots (cf. Blake 1992:40-42; Blake 1999). This proposed difference in prosodic structure (Nuclear non-moraic versus Nuclear moraic) will play a central role in determining the distribution of schwa. Prosodic structure of Sliammon is discussed in more detail in Chapter 3 where additional evidence is provided in favour of the weight contrast introduced here. This proposal regarding the distinction between schwa on the one hand, and the full vowels on the other hand draws significantly on the research of Shaw (1993 et seq.) on other neighbouring Salish languages (St'at'imcets (Lillooet), Nuxalk (Bella Coola), and hsnqsminsm (Musqueam Salish). This theoretical model provides insight into the behaviour of schwa in Sliammon, and provides an alternative analysis to Blake (1992).  2.4.3 Full Vowels Surface Realization of the Full Vowels The next section discusses the allophonic realization of the full vowels Ii, u, al. Retraction The Full Vowels are systematically lowered before and after uvulars, and laryngeals ?/h indicating that these post-velar sounds function together as a natural class. As seen in §, it is proposed that this class is captured by hypothesizing that these segments share PHAR specification. This PHAR specification triggers retraction of Ii, u, al, as in (97).  (97) Retraction lil  -»  [e] / PHAR  IvJ  -»  [o] / PHAR  lal  ~*  [a] / PHAR  68  This lowering takes place in both stressed and unstressed syllables, and is illustrated by the data in (98-106). Since M l vowels are generally laxed in unstressed closed syllables, the discussion here will focus on full vowels in stressed syllables. The [-bk] vowel I'll is retracted in the environment before and after post-velars (uvulars and laryngeals), as shown by the data in (98-101). It is retracted in stressed open and stressed closed syllables alike which shows that the presence of [E] cannot be attributed to a constraint on closed syllable shortening, for example.  (98) Retraction of I'll with uvulars Input  Output  Gloss  a. qiqti? (DIM?)  qiqti?  qEq.t£?  youngest in family  b. L'-DIM-xnq+[i]  x[i?]i-xn[i]q  XE . ?EX . n£q  Owl's Grove  c. mixai  mixat  rat.xAf  black bear  Output  Gloss  (99) Retraction of I'll with laryngeals Input a. ?ilq=ay  ?ilqay  ?E1 . qay  barbecued deer meat  b. ?ittan  ?ittan  ?£i. tan ~ ?E'f . tAn  to eat, food  b'. IMP-?ittan+[?] c  ?i-?ittan c  c. ?inhus  ?inhus  ?En. hos ~ ?£n . hos  new moon, month  d. ei?ea  ei?6a  0£?. 0a  that one (fern.)  e. higin  hi?gin  hi?. gm  strawberry  f. DIM-hk i?[i]q w  g- DIM-higus  w  hihk i?iq w  hi-hagus  w  Tm eating  h£h . k i . ? E q w  w  great-gr. grandmother  hE . hagus ~ hEhagvs small chief  The data in (100) shows that I'll is also systematically retracted in the environment of glottalized consonants.  69  (100) Retraction of I'll with glottalized consonants Input  Output  Gloss  a. t'in  t'in  ten  barbecued fish  a'. IMP-t'in-?m  ti-t'in-?3m  te. ten. ?9m  barbecuing fish  b. k in  k in  k en  how many  w  w  c. W i t  Wik*  w  d.  w  fit^iq  e. t'iniq  w  Wiq  worm t^e.t^eq  t'i?niq  w  te?. neq  w  mud salmonberries  Note that Retraction is obligatory, as indicated by the contrast between the grammatical examples in (101) Column 2 versus the ungrammatical examples in (101) Column 3. Focus on the quality of the stressed vowel in each example. Failure to retract the vowel I'll in the environment of a postvelar or glottalized consonant is clearly ungrammatical, judging from comments made by speakers of the language.  (101) Input a. qiqti?  Retraction  No Retraction  Gloss  qeq. te?  *qiq.te?  youngest in family  a'.  *qeq. te?  b. mixai  me. XA4  b'.  black bear  *me. XA!  c. ?ritan  ?e't. t A n  *?ii . t A n  to eat, food  *?ef. t A n  c'. d. ?inhus  ?en . hos ~ ?en . hos  d'.  *?rn . hos  new moon, month  *?en. hos  e. t'in  t'en  e'. f.  *mi. XA4  *t'in  barbecued salmon  *ten k in w  k en w  *k in w  70  how many  *k en  f.  w  g. t' it' ik e  e  w  f ^ . i M  *t^i. t' e£  fe?.neq  *fT? . neq  e  worm  w  g'h. tmiq  w  w  . neq  h'.  salmonberries  w  w  The vowel IvJ is retracted and realized as [o] before post-velars or glottalized consonants, as shown by the data in (102-104). The retraction of IvJ parallels the observed behaviour of lil discussed in (98-101) above.  (102) Retraction of/u/with post-velars  a.  Gloss  Output  Input juq^uqw/jaqw  juq-juq-  joq .joq w  lukewarm water  w  small soapberry  b. DIM-x usm+[i]+[?]  x u-x sirh  c. puhu  puhu  pd. ho .  raven  d. ?amamu?  ?amamu?  ?d. ma . mo? .  chiton  e. cu?cu?  cu?cu?  co?. co?  wren  w  f. IMP-?u<iq u+[?] w  w  w  ?u-?utq u? w  ?d . ?ot. q o?  digging clams  Output  Gloss  w  (103) Retraction of /u/ with glottalized consonants Input a. R ufiut'  k unut'  £ o?.nof  porpoise  b. humhum  humhum  horn. horn  blue grouse  c. fufmum  fuimum  i d i . mom  Uttleneck clam  w  w  w  Failure to undergo retraction is judged as ungrammatical, as shown by the data in (104).  71  (104) Retraction  No Retraction  Gloss  a. DIM-x usm+[i]+[?]  x ox . sim .  *x ux . sim .  small soapberry  b. puhu  po. ho.  *pu. hu.  raven  c.  ?d. ma . mo? .  *?d. ma. mu? .  chiton  Input w  ?amamu?  w  w  w  *?d . ma. mo? .  c\ CO? . cb?.  d. cu?cu?  *cd?. CO? hurhhum  hom . horn.  *hum . hom.  blue grouse  *hdm . hom.  e'. f.  wren (also: 6eSo?6o?)  *co?. CO?  d\ e.  w  iutmum  toi.  *tdi. mum .  mom .  littleneck clam  *tot. mom .  f.  The low vowel lal is retracted and realized as [a] in the environment of post-velars, as shown by the data in (105-106).  (105)  Retraction of lal in the environment of uvulars  Input  Gloss  Output qa-qyx ui  qd . qe . x vi  smaUMink  b. IMP-q asm+[?]  q a-q asam  q d . q a . sam  blooming, flowering  b*. DIM-q asm+[?]  q a-q s[9]m w  q dq . sam  little flower  c.  DIM-q atm[-i-]+[?]  q a-q t'[i]rh  q aq . t'erh  small river, creek  d.  IMP-x aj-t-awl  x a-x aj[a]tawi  x dx . je . tAwi  fighting each other  a.  DIM-qayx^ui w  w  w  w  w  w  w  w  w  w  w  w  w  w  w  w  w  72  w  w  w  w  (106) Retraction of/a/with laryngeals Input  Output  Gloss  a. IMP-?a0-m+[?]  ?a-?ae-9m  ?d . ?a . 0 A m  giving (at a potlatch)  b. ?ah  ?ah  ?dh  sore, pain  V. IMP-?ah-m+[?] c  ?a-?ah-9mc  ?d. ?a . hAmc  I'm hurting  c. IMP-?ax +[?]  ?a-?ax  ?d. ?ax .  it's snowing  d. DIM-k upa+[?]  k u-k pa?  k iik . pa?.  grandfather  w  h  w  w  w  w  w  w  w  The allophone [a] (DOR [lo]) and the retracted allophone [a] (DOR [lo] PHAR) are both written as a script 'a' in the remainder to this thesis. Retraction results from consonant-vowel feature sharing, as illustrated in (107). A capital Q is used to refer to the class of PHAR consonants in the language.  (107) Retraction b.  PN  /  /  PHAR  Q  I  PN  PN  I \ I \  DOR  u I I  t  PN  PN  \  LAB DOR  \/  /  PHAR  [e]  DOR PHAR  [lo]  [rd]  [-bk]  PN  I \ / I \/  [a]  [0]  As shown by the representations in (107), each retracted vowel [e, o, a] is characterized by its surface PHAR specification.  73 Place Assimilation The non-low vowels Ii, \xl are realized as [i, u] in the environment of DORfhi] consonants (alveo palatals and velars), as shown by the data in (108). (108) Input  Output  Gloss small child  a. DIM-cuj=ui  cu-cj=u4  cue. jut  b. IMP-jux -t+[?]  }U-JuX -9t  jii. jo . X A t  vomiting  c.  gu-guh-um  gu . gu . hom  barking  w  IMP-guh-Vm+[?]  w  w  (109) u  c  I PN  PN  I  l\  LAB DOR [rd] \ \  I DOR / | / [-bk]  [hi] The low vowel lal is realized as [e] after alveo-palatals, as in (110). (110)  [E]  Input  Output  Gloss  a. ?aya?-hV  ?aya-ha-?  ?a'. y E . . hci?  lie's got a house  b. DIM-sayja+[?]  sa-syja?  sa . si-. js?  small leaf  DIM-canu+[?]  ta-bnuf  Hb . no?  Utile dog  d. DIM-cag='ay+[?]  6a-^gay  bib.  small wooden spoon  c.  gAy  DIM-janx [-i-]=u°i+[?]  ja-jn[i]x =ut  jjfj.  f.  DIM-yaxay+[?]  ya-yxay  VEV . X A y  small clam basket  g- IMP-cah-m+[?]  ca-cah-9m  ^E . ^E . ham  praying  g'- IMP-cah-'Vg-m  6a-cah-ag-am  ^E . be . ha. gam  they're all praying  h. IMP-yax+[?]  ya-yaX  ye-VAX  sobering up  w  w  74  ne . x uf  small fish  e.  w  This is analyzed here as partial assimilation to the preceding alveo-palatal. This is illustrated by the representation in (111).  (Ill) c  I  a  I  PN  PN  I  I  DOR DOR / \ / \ [hi] [-bk] [lo]  [ce]  The fact that the feature [hi] fails to spread is encoded in the grounded constraints *HI/LO and It is also proposed here that spreading of [-bk] entails the loss of the feature [lo] (if [-bk]  *LO/HI.  then not [lo]). As noted earlier, the low vowel /a/ is never rounded. This follows from the general lack of low round vowels in system, and is analyzed formally as a high-ranking grounded constraint *LO/RD,  following Archangeli and PuUeyblank (1994).  Archangeli and PuUeyblank (1994) develop a model in which features (f-elements) freely combine in order to derive the inventory of consonants and vowels in a particular language. Combinatorial specification is constrained by phonetically-motivated grounded conditions which ban antagonistic articulatory gestures (i.e. a vowel cannot be both high and low at the same time *HI/LO)  and permit combinations which are compatible from an articulatory perspective. It will be  argued here that consonant-vowel interaction in Shammon is also constrained by grounded constraints. Effect of Anterior Consonants on /a/ The low vowel /a/ [a] occurs in free variation with [A] in the environment of labial and coronal (anterior) consonants.  The effect is gradient and variable. Examples of free variation  [a ~ A'] are given in (112). 75  (112) Input  Output  Gloss  a. tan  tan  tan ~ tXn  mother  b. man  man  man ~ mXn  father  c. nat  nat  nat ~ nXt  night  d. pipa  pipa  pipa ~  paper  e. ppa  papa  papa ~ p a p A  pipA  pepper  In the context of consonants specified for [hi], /a/ is realized as [A], as shown by the morphologically related forms in (113).  The vowel and surface form of the following [hi]  consonant Igj are underlined in Column 3. (113) Input  Output  Gloss  a. maga  maga  mXgA  cougar  b. DIM-maga=ui  ma-maga-wat  mXmAgdwai:  little cougar  c. CaCPL-maga  maw-maga  mdwmXga  cougars  d. mi-mag+[?]  mi-maw  me:mAW  cat  3 Interaction of Retraction and Place Assimilation In general, Retraction takes precedence over spreading of DOR[hi], as shown in (102.a; 108.b) for example. Retraction and the spreading of DOR[-bk] are sympathetic, as shown by the data in (1 lO.b). The data in (114-115) show some conflicting tendencies. The data in (114) shows that the vowel IvJ is realized as DOR[hi] when it occurs between DOR[hi] consonants (k _j and w  x _j) even though the vowel IvJ is immediately followed by the laryngeal constriction associated w  with /]/. This provides evidence that the glottalization associated with /]/ does not entail the 3  This Root also shows some irregular behaviour. It is written elsewhere in this dissertation as /mga/ maga  'cougar'; however, the diminutive form with the retention of the Root vowel is expected with strong roots of the form CACA, and not with weak C C A Roots.  76  presence of the PHAR node (cf. the representation of /j/ in (80)). The examples involve the surface realization of the LS=uja hand, lower arm.  (114) =uja [=u?Te] Output  Input  Gloss  a. Fifc =uja  t' ik u?ja  t^ek^Vje  b. Xpx =uja  Xipx u?ja  XXpx u?je  w  e  w  w  left-handed  w  w  break one's arm, hand  When /u/ is preceded by a labial or coronal consonant, the non-low back rounded vowel Ixxl is realized as [o] (its most "neutral" realization), as shown by the morphologically related forms in (115).  (115) =uja [=o?je] Input  Output  Gloss  a. ct'=uja  C9t'u?ja  &to?Je  cut one's hand  b. s4p=uja  salpurja  siip6?je  sUp out of one's hand  c. cm=uja  ca?mu?ja  6e? mo?j6  h  3  h  cold hands  The vowel lal also surfaces as [A] in the environment of a uvular, as shown in (116). (116) Output  Gloss  q alas  q dl:AS  raccoon  a'. DIM-q alas=ut  q a-q alasu"i  q A'q 3la-s6t  Utile raccoon  a". CsCPL-q alas  q al-q alas  q A'lq alAS  raccoons  b. xap  xap  xap ~  papoose basket  c. yaXay  yaxay  yXxAy ~ yexAy  berry-picking basket  d. qast  qast  qast ~ qXst  special person in yr. Ufe  Input a. q alas  w  w  w  w  w  w  w  w  w  w  w  w  w  11  XAp  The example which is of particular interest is the diminutive (DIM) in (116.a'). Notice that the low vowel a is realized as [A] is a stressed open syllable which is preceded and followed by a labiouvular consonant. Uvulars are proposed to be DOR PHAR. Notice that the resulting vowel [A] is not rounded which follows from the fact that there are no low round vowels in the language. If the root were q alas, we would expect a diminutive form in Ci- as well as labialization of the stressed w  vowel in a total labio-uvular context (cf. Blake (1992), Watanabe (1994, 2000), and Blake (in prep) on Reduplication in Shammon). The morphological evidence clearly points to the underlying vowel here being /a/; however, the surface realization [A] instead of [a] entails total Place Assimilation to the uvular. The low vowel /a/ is realized as DOR PHAR. The effects of uvulars (and post-velars in general) on vowels in Shammon merits further study. Full Vowel / Consonant Interaction The next section presents a summary of the effects of both preceding and following consonants on the full vowels /i, u, a/. The tables in (117-119) summarize the realization of the full vowels Ii, u, al in an initial stressed closed syllable.  (117) lil DOR[-bk] C2  LAB  COR  DOR[hi]  PHAR  e  e  8  CI LAB  e~  COR  e  e~i  i  8  DOR[hi]  i  1  i  8  PHAR  e  e  8  8  i  The specification DOR[hi] is used here to include alveopalatals, palatals and velars irrespective of their specification for the feature [back]. PHAR identifies the class of post-velars.  78  (118) /u7 LAB[rd] DOR LAB  COR  DORfhi]  PHAR  LAB  d  d~ ii  ii  o~d  COR  d  d  ii  0-0  u~ d  ii  d~d  d  d  5-6  C2 CI  DOR [hi] PHAR  *  *  o~o  It should be noted that the following consonant appears to exert a stronger influence on the resultant vowel quality than the preceding consonant does; although, in the case of the preceding PHAR and following DORfhi], Iwl is realized as [d] and not as *[u] showing that Retraction is less costly than DOR[hi] assimilation. This is a classic case of constraint conflict, and provides evidence for the ranking: *DOR[hi] »  *PHAR.  The summary of the surface realization of /a/ in (119) includes a distinction between alveopalatals/palatals (DOR[hi, -bk]) and the velars (DOR[hi]) in order to underscore the effects of a preceding alveopalatal on the realization of/a/, as discussed in (110).  (119) lal DOR [lo] LAB  COR  DOR[hi, -bk]  DOR[hi]  PHAR  LAB  d  d~X  d  d  d  COR  d  d  a~ a?  a~a  d  DOR[hi, -bk]  £  £  e~£  £  e  a~ a6  d  d  d  d  d  d  a~d  a~A  a  C2 CI  DOR[hi] PHAR  79 Proposed Analysis The surface vowels can be shown to arise from the interaction of constraints which drive Retraction and Place Assimilation. The intuition behind this analysis is that consonants and vowels which share Place specifications are more highly valued than ones which do not. Consonantvowel feature sharing occurs subject to a set of constraints which are phonetically grounded in the sense of Archangeli and PuUeyblank (1994). Sympathetic (compatible) articulatory gestures are enhanced and licensed whereas antagonistic articulatory gestures are banned by constraints on possible co-articulation. The grounded constraints which are clearly operative in Shammon are given in (120).  (120) Grounded Constraints a.  If a vowel is [-bk], then it is not [rd]  b.  If a vowel is [rd], then it is not [-bk]  *RD/-BK  c.  If a vowel is [lo], then it is not [hi]  *LO/HI  d.  If a vowel is [lo], then it is not [rd]  *LO/RD  *-BK/RD  A high-ranking constraint ensures that Retraction occurs (PHAR place is shared with an adjacent vowel) at the expense of spreading [hi]. This generalization suggests that the grounded condition *PHAR/HI is also operative in Sliammon. (121) If PHAR, then not [hi]  *PHAR/HI  This accounts for the height of vowels which occur after alveopalatals and before uvulars. The study of Glide Vocalization §2.4.6 also motivates the constraint in (122). (122) If [hi], then not PHAR  *HI/PHAR  80  The effect of adjacent consonants on Full Vowels is summarized in (123).  (123) Retraction  DOR [hi]  *DOR[hi] » * P H A R  Grounded Condition  I'll DOR[-bk]  [el  [il  [el  *-BK/RD  Al/ LAB[rd] DOR  M  [ul  [Ol  *RD/-BK  lal DOR[lo]  [al  [el  [el  *LO/RD  *LO/HI  *LO/HI  Full Vowel  The output of Full Vowel assimilation to adjacent consonants is subject to the grounded conditions in (120-121). These high-ranking grounded constraints reflect generalizations about Sliammon. There are no front rounded vowels in the language, nor are there any low rounded vowels. These grounded constraints are therefore proposed to be undominated constraints. The output of Retraction never yields a [hi] retracted vowel (cf. Bessell 1992a). Since full vowels are retracted whenever they are adjacent to a post-velar consonant, I assume that the presence of PHAR (retraction) prohibits the spread of DOR[hi]. This is reflected in the grounded constraint *PHAR/HI in (121)  Sharing of DOR features [hi] and [hi, -bk] occurs from adjacent alveopalatals and velars, the output is given in (123) Column 3. Notice that the low vowel lal is never [hi] expressing the fact that vowels can not be both low and high at the same time (120.c). (123) Column 4 shows the effect of consonants on either side of the full vowel. As observed from the data, Retraction (spread of the PHAR node) takes precedence over the spread of the Dorsal feature [hi]. Vowel assimilation to neighbouring consonants in the language seems to be subject to two different sets of constraints - first of all, Faithfulness to the lexical vowel features which establish the contrast between /i, u, al, and second of all, to the phonetic grounded conditions. A constraint which drives Vowel Assimilation to the neighbouring consonants may be in conflict with the grounded conditions. The grounded constraints must outrank VAssimilation since the opposite  81  ranking would create total assimilation without consideration for the phonetic viability of such a feature-sharing relation. The tableau in (124) shows the effects of this ranking.  FAITH I-Offl «sra.  V  Grounded Conditions  V Assimilation  C  -bk rd b. V C *! (*-BK/RD)  |\ 1 1 \| -bk rd c. V C  *!  |\ 1 1 \|  <-bk> rd  As shown by the optimal candidate in (124.a), C-V feature sharing is violated in order to satisfy the high-ranking Grounded constraint *-BK/RD. Violation of this constraint is ruled out as in (124.b). It is also necessary to consider what rules out candidate (124.c). Since lexical features of each full vowel are present in the output of Vowel assimilation, this means that Faithfulness and the Grounded Constraints both outrank Vowel Assimilation. It is also proposed here that the vowels never lose their featural content in order to assimilate and satisfy the grounding conditions. Faithfulness of the lexical features (FAITH [f]) associated with both consonants and vowels is high-ranking.  (125)  Faith[fJ, Grounding »  V Assimilation  What is the cost associated with Vowel Assimilation? There is no insertion of a feature not present in the Input; rather assimilation involves the insertion of a path (association line) between a vowel and adjacent consonantal place features (i.e. a DEP-PATH violation) (cf. Pulleyblank 1996:289-299).  82  (126)  Faith[f], Grounding »  V Assimilation »  DEP-PATH  Consider the partial tableau which shows how this works for the full vowels Ii, u, a/ when followed by a rounded uvular consonant (Q ). In order to consider a full range of candidates, the w  optimal candidate as well as other non-optimal ones, I adopt the following conventions here: [e, o, a] represent the retracted counterparts of the full vowels Ii, u, a/ as above. I will use the addition of a small raised  w  ([e , o , a j) to indicate a candidate which has undergone both w  w  w  retraction and labial assimilation (the leftward spread of [rd] from the consonant onto the preceding full vowel (or in the case of o sharing of [rd]). The candidates containing [E , a ] will shown to be w  w  non-optimal since they violate the grounded constraints *-BK/RD and *LO/RD respectively.  (127) Input: IC i Q / w  FAITH [f]  GROUNDING  V ASSIMILATION  DEP-PATH  V: DOR[-bk] Q :LAB[rd] DOR PHAR w  ®a. C £ Q  w  +  b. C e Q  w  c. C o Q  w  w  *! C-BK/RD)  DOR ""RD +  +  DOR  *  *! MAX-V[-BK]  The symbol ii used in (128.c) stands for a front rounded vowel (i.e. the consonant has spread all of its place features and the vowel u has retained its lexical [rd] specification).  83  (128) Input: / C u s7  GROUNDING  FAITH [f]  V ASSIMILATION  DEP-PATH  *(-bkl  *  u: LABfrd] DOR s: DOR[hi, -bk] •sr  a. C u s b. C i s c. C u s d.  4-  *! <LAB [rd]> *! (*RD/-Bk)  C o s  *i* DOR[hi -bk]  The tableau in (129) shows that /a/ shares a PHAR specification with a following labiouvular consonant, but that its [rd] specification is not spread onto the vowel due to the high-ranking grounded constraint *LO/RD , which bans low rounded vowels. So although the optimal candidate in (129.a) fails to undergo complete C-V feature sharing as shown by the violation of the constraint V-Assimilation, it does so in order to satisfy the higher-ranked constraints on Faithfulness and Grounding. Candidate (129.b) is ruled out since the [rd] specification has been spread onto an adjacent low vowel; this constitutes a violation of the grounded constraint *LO/RD. Candidate (129.c) is ruled out due to a violation of Faithfulness; the feature [lo] which defines the low vowel /a/ is not present in the Output. Candidate (129.d) is ruled out since consonant-vowel feature sharing (i.e. vowel assimilation) does not take place at all; failure to spread the PHAR node rules this out.  (129) Input: /C a Q / w  «• a. C a Q w  d. C a Q  GROUNDING  w  V ASSIMILATION *[rd]  w  b. C a Q c. C o Q  FAITH [fl  DEP-PATH  ' *  *! (*LO/RD)  w  *  *! <[lo]> *! PHAR, *[rdj  w  84  2.4.4 Schwa Schwa in Sliammon is subject to colouration from adjacent consonants and vowels. The allophones of schwa are brief in duration. Their quality results from feature sharing with adjacent consonantal and/or vowel place features, with some degree of variation. Schwa becomes [t ~t ~ i ] in the environment of alveopalatals, palatals and plain velars. Schwa is realized as [v] in the environment of labio-velars whereas it is shghtly lower and rounded [o] in the environment of labialized uvulars. Schwa is lowered to [A] in the context of plain uvulars, whereas it is lowered to a brief [a] in the environment of ?. It should be noted that [?] appears to have a systematic lowering effect on a preceding stressed schwa, a property not always exhibited by [h]. Schwa before [h] varies [A ~ a], as discussed in § above. Since schwa is realized as [a ~ i ~ * ~ v ~ o ~ A ~ a ], it is important to distinguish schwa and the allophones of schwa from the allophones of the full vowels Ii, u, al. The reader is referred to J. Davis (1970 et seq.), Kroeber (1989), Blake (1992), and Watanabe (1994) for similar discussion regarding the surface allophones of schwa. As argued in §, schwa in Shammon is analyzed here as a bare Nucleus, devoid of inherent phonological features, and lacking in phonological weight, following Shaw's (1996a, 1996b) analysis of other neighbouring Salishan languages (Lillooet, and Bella Coola). Kinkade (1998: 208) argues that epenthetic schwa in Upper Chehalis is both non-moraic and unspecified for phonological features, providing additional comparative evidence supporting the proposed representation of schwa in Sliammon. Matthewson (1994: 4) in her discussion of Lillooet schwa states that "consonants on both sides of hi colour its realization, in a non-discrete fashion, suggesting phonetic interpolation effects rather than phonological processes."  Shammon seems to exhibit some patterns of schwa  colouration which are systematic; these are (i) Retraction, (ii) Labialization, and (iii) the effects of preceding Alveopalatals. These three effects are therefore derived by constraint interaction. These effects are distinct from the effects of some preceding consonants on schwa - here I have recorded greater amount of variation, the forms are gradient, and do not seem to have the same status within  85  the grammar - their variability may well be attributed to phonetic interpolation rather than phonological constraints. The following section provides examples of the allophones of schwa and the range of variation in the output of this vowel. Schwa is realized as [a] in a 'neutral' context, as shown by the data in (130). Consider the realization of schwa in the initial stressed closed syllable. These examples are derived via CaCPlural reduplication; the schwa in question occurs within the reduplicative prefix. As shown by the placement of stress, the reduplicant occurs within the domain of the Prosodic Word.  (130) CaC- Plural Reduplication Gloss  Output  Input a. CoCPL-pma b. CaCPL-masiq  w  p9m  m9s-masiq  m9s . iriA . seq  w  purple sea urchins  w  eyebrows  09m . 0o . men  09m-0umin  c. CaCPL-0umin  wooden floats  . pa? . ma  pam-p9?ma  d. CaCPL-t' amq i  f 9m-t' amq l  t' 9m . f ^ A m . q 4  lots of clouds  e. CaCPL-sup=nac=min  s9p-supnacmin  s^p . sop . HAC . min  stumps  S9m-s9?ma  sam . sa? . ma  mussels  t9m-tumis  t9m  . to . mis'  young men  tan  lots ofmothers  e  f.  w  CaCPL-sma  g- CaCPL-tumiS  9  e  w,  e  w  h.  CaCPL-tan  t9n-tan  t9n.  i.  CaCPL-tala=aya-ap  t9l-tala[h]aya[h]ap  tsl .ta. la. ha. ye . h A p  your (pi) purses  j-  CaCPL-t'in  tan-t'in  t9n . ten  barbecued fish  k. CaCPL-t'iniq 1.  CaCPL-Xapatii root?  fen . t'e? . neq  Xgp-Aapatit  t6p . t<5 . pa . tit  cedar bark baskets knees  w  m. CaCPL-q n=iq ra  q an-q 9?niq ta  q 9'n .q a? .nuq w  w  n. CaCPL-?asx  ?9s-?asx  ?9s  . ?as . x  w  Wt  w  salmonberries  t9n-t'i?niq  w  w  w  w  w  w  86  w  w  seals Schwa and Retraction Schwa is realized as [A] in the environment before a plain (non-labialized) uvular consonant, as shown by the data in (131.a-f).  (131)  [A]  Input  Output  Gloss it's all white  a. CaCPL-pq  pgq-pgq  pXq .  b. CaCPL-mqsin  maq-maqsin  mXq . mAq . sin  noses  c. CaCPL-mixai  max-mixai  mXx . m e  black bears  d. CaCPL-sxm  S9X-S9X9m  SAX . SAJCAm  e. CaCPL-t^xu  t' 9X-t' 9XU  t' AX . t' 9 . X O  lots of ling cod  f.  Xaq-Xaqsin  XXq . XAq . sin  lots ofmoccasins  6  CaCPL-Xq=sin  E  6  pAq  6  .  XA$  racing canoes w  This is analyzed as C-V feature sharing; in particular, sharing of the consonantal place node (PN) as shown by the autosegmental representation in (132).  (132) 9  o  'PN  A DOR  PHAR  Schwa is also retracted and realized as [A] before a tautosyllabic glottalized consonant, as shown by the data in (133).  87  (133) [A] Input  Output  Gloss  a. Xam-Xam  Xam-Xam  XXm^Am  square  b. tn-t  iant  iXnt  to weave s.t.  c. mX  maX  mXX  calm (on water)  d.  ?ap-?aptan  ?Xp . ?ap . tan  green sea urchins  tam-tam  tarn. t A m  lots of belts  CaCPL-?aptn  e. CaCPL-tm  When schwa follows a uvular or laryngeal consonant it varies between [a ~ A], showing that the preceding segment does not exert as strong an influence on schwa as the one which follows it.  (134) [a~X] Gloss  Output  Input a. CaCPL-q alas  q al-q alas  q Xl. q a . las  raccoons  b. CaCPL-q ns  q an-q anas  q Xn . q A n . nts  humpback whales  c. CaCPL-q asm  q as-q asam  q as . q A . sam  lots ofgrouse  d. CaCPL-?atnupil  ?at-?atnupil  ?Xt. ?at. no . pel  cars  e. CaCPL-qap=awus  qap-qapawus  qXp . qa . pa .  bats  w  w  w  w  w  w  w  w  w  w  w  w  w  w  w  wvs  Schwa becomes [a] before a tautosyllabic glottal, i.e. one not followed by another V. This [a] is brief in duration, and is not as long as [a] from lal (cf. Kroeber 1989). Notice that this occurs both before 111 as in (135.a-c), and before the [?] associated with a glottalized resonant, as in (135.d-f). (135) [a] Output  Input  Gloss  pa?-pa?-pu?px  pX? . pa? . pu? . px lotsofkindhng  b. CaCPL-qVft  q a?-q a?i  q d? . q A?i  lots ofraspberries  C  X a?-x u?p  x d? . x 6?p  lots of awls  sa?ma  sd?mA  a. CaCPL-RED-pu?px  CaCPL-X u?p  d. sma  w  v  v  w  w  w  w  w  w  w  88  w  w  mussel  e. mria  m9?na  md?na ~md?nA  f.  qa?ya  qd?ye  qya  child, one's offspring water  The autosegmental representation in (136) illustrates lowering of schwa to [a] before ?. (136) 9  ?  o  I\ PN  LARfcgl]  I PHAR  Although the uvulars and laryngeals form a natural class of PHAR consonants, uvulars are distinct from laryngeals by virtue of their DOR specification (akin to DOR [-hi] within a model which admits binary features). As shown by the data in (135), glottal stop which is specified as both PHAR and LAR [cgl] has a significant lowering affect on schwa, an effect which is not shared with the plain uvulars. Schwa and Labialization Schwa is realized as [v] when it occurs either before or after a labio-velar consonant, as in (137).  (137) [v] in environment of labio-velar Output  Input psk -puk  Gloss  puk . puk  lots of books  a.  CaCPL-puk  b.  c CPL-t' k a  t' 3k -t' ak a  t' x5k . t' 9 . k a  c.  C3CPL-0k =nac=tn  09k -09k nactn  0uk . 0uk . n a c . t n lots of chairs  d.  C9CPL-tix 6at  t9x -tix 0ai  ti5x  e. CaCPL-k nay  k 9n-k 9?nay  k vn . k  w  f.  fc 9f-k att  l vi  w  e  w  w  9  w  w  w  C9CPL-k aft w  w  e  w  e  w  w  w  w  w  w  w  w  w  w  w  e  w  e  w  w  w  w  w  89  w  edible rootstalks  w  .tix . w  0A1  a ? . nAy  . k Aft  tongues lots of Uds lots ofplates  (138)  Schwa is both retracted and rounded before a labio-uvular consonant, as shown in (139). (139) [o] Output  Input  Gloss  a. C»CPL-Xiq =ana  foq^iq^ana  toq . ttq . ?a . na earlobes  b. C9CPL-Xq =inas  X9q -X9q inas  toq . to . q e . nAS hearts  c.  igq -i9q  4 o q . i:oq  w  w  w  CaCPL-i9q  w  w  d. C3CPL-£x =ay  w  w  w  w  w  w  w  5feX -3feX ay  w  w  w  arrows  w  tox . to . x w  w  A y  lots of dog salmon  (140) 9 o  PN  /I \ LAB DOR PHAR [rd] Schwa and Place Assimilation Schwa is realized as [i] between coronals, and as [t] in the environment of DORfhi] consonants (alveopalatals and plain velars). The relevant syllable is underlined in the Output column of (141).  90  (141) [i] Gloss  Output  Input a. C9CPL-6uein  eig.  eae-euein  b. cuj C9CPL-saitx  S9i-sattx  w  Go.  cuy . s£t- sal. tx  w  hps (pi)  em w  young women lots of slugs  fol-foVlaqan  %t 1. Aa? . la. qAn  d. CsCPL-jX=umix  j9A-j9?Airmix  . J£? . AO . mix  e. CaCPL-macin  macWnacm  mi6 . ma . 6tn  lice  f.  kap-kapu  k tp . k as . po  lots of coats  c. CaCPL-Xlaqan w  CsCPL-kapu  w  y  y  w  car, vehicle  (142) 9  o  / \ [-bk][hi] Interaction of Retraction and Place Assimilation J.Davis (1970) and Urbanczyk (1999) state that schwa becomes [e] when it occurs after an alveopalatal and before a laryngeal. As documented here in §, when schwa occurs after an alveopalatal and before a plain uvular stop, it is realized variably as [A ~ e]. Blake (1992, 1995), and Watanabe (1994, 2000) make a slightly different claim in which schwa neutralizes with the low vowel lal before a laryngeal, and then, like lal assimilates to the preceding alveo-palatal yielding [e]. The morphological status of the Root in question is taken here to be the deciding factor as to whether the vowel is underlyingly an lal or is a schwa. If the Root behaves as a Weak Root for morphological purposes, the surface [e] is analyzed as schwa whereas if the Root were to behave as a Strong Root, then surface [E] is analyzed as lal. Recall that surface [E] also comes from I'll in the environment of a following post-velar consonant, but that the allophones of schwa are perceptually shorter in duration than the allophones of the full vowels. 91  Examples in which schwa is realized as [e] are documented in (143). Notice that in each case both the alveopalatal and the following laryngeal ? belong to the same syllable. A comparison of the examples in (148.b-b'), repeated here as (143.d-e), shows that this is a necessary condition for assimilation.  (143) Output  Gloss  §9?mit  se?. met  dried (stative)  b. hu cap CaCPL-ju?  hu cap ]9?-ju?  hd . cep . je? . Ju?  you (pi) go home  c.  S9?t ga  se?t g A  go upstairs  Input a.  sm-it  s?t ga  d. c?=umix =tn  cb?umix t9n  co . ?o . m i x . t 9 n  rug on floor  e. c?=nac= tn  C9?nact9n  til  small blanket to sit on  w  w=  =  w  . nAc  . tin .  The surface realization [e] involves the rightward spread of DOR[-bk] and the leftward spread of PHAR given the present analysis, as in (144). (144) s  9  PN  PN / \  |  DOR  DOR  ? PN / PHAR  |  \ / [hi] [-bk] [SE?]  The examples in (145.a-c) show that for some speakers Retraction takes precedence over the effect of a preceding alveopalatal, since schwa is realized as [A] and not as [E]. The examples in (145.d-f) show that schwa is realized as [A] before a plain uvular, as in (131) above.  92  (145) [A] in the environment before a plain post-velar Input  Output  Gloss  a. cq  hq  cAq  robin  b. cx  c9x  CAX  ripe, cooked, done  c- jq  pq  jA'q  smooth  d. pq  paq  pA'q  white  e. Xq  £gq  XXq  rot  f. qx  qax  qXx  many  (146)  PN DOR / \ [hi][-bk]  Height assimilation of schwa to the following alveopalatal (spread of DOR[hi]) occurs consistently, but the effect of the preceding labialized consonant on schwa varies across speakers, giving the surface variation recorded in (147.a-b).  Again the tautosyllabic consonant which  follows schwa exerts a stronger effect on the surface realization of this epenthetic nucleus than the preceding consonant does.  (147) different speakers Input  Different Speakers  Gloss count s.t.  a. k ^ - t  k 9St  k t)st ~ k ist  b. k s=u0in-m  k 9su0in3m  k i5so0en9m ~ k fso6en9m  to tell a joke  c.  k 9n-k 9?nay  k tfnk a?nAy ~ k ink a?nAy  lots ofhds  w  w  w  CaCPL-k nay w  w  w  w  w  w  w  w  w  93  w  w Transiaryngeal Harmony Not only is the surface realization of schwa determined by adjacent consonants, it can also be influenced by the quality of adjacent vowels. The example in (148.a) shows that schwa may be coloured by a preceding vowel, and that assimilation occurs across an intervening laryngeal. The example in (148.b) shows that transiaryngeal harmony can also occur in the other direction; schwa becomes harmonic with a following vowel, and assimilation occurs across an intervening laryngeal.  (148) Input  Output  Gloss  a. tg-?m-t-'ui c  tu?amtui c  tu?omt6ic ~ tu?omt61:c  I froze it for her  b.  C9?umix t9n  bo . ?o . mix . tan  rug on floor  b'. cV?=nac=tn  ca?nact9n  btl. nAC . tin .  small blanket to sit on  b". £?-  <b?-  6?=umix =tn w  w  w  be on top of  94  2 A A.6 Summary of Allophones of Schwa The following table presents a summary of the allophones of schwa: Q indicates the place of articulation consonant which precedes schwa, and C2 is used to indicate the place of articulation of the consonant which follows schwa. The dotted line (-—) indicates that there are no clear examples in the present data base. The gap in (150) involves the plain velars in C2 position, a position in which plain velars are severely limited in distribution (cf. Appendix IV).  This is  directly due to the fact that Proto-Salishan *k/k were fronted to clh respectively. Examples of [k] and [k] within the synchronic grammar are loan words from either English or Chinook Jargon, or the neighbouring Wakashan language Kwakwala.  (150) C2 LAB  COR  Alveopalatal  Velar  Labio-Velar  Uvular  Labio-Uvular  Glottal ?  Ci  LAB  9  9~A  9  COR  9~ A  9 ~I  I  9 ~ 9'  A  V ~D  a  t ) ~ ii  A  i5 ~ o  d  A~ E  0  E  w  l  I~1  Alveopalatal  I  1~e  Velar  I  1  I  9 ~l5  A  d  Labio-Velar  V  V  I~V  •u ~ d  A  d  Uvular  A  A  A  A  6 ~ 9'  a~ A  Labio-Uvular  A  A  d ~d  X~d  0  d  Glottal  A  9~ A  A  A  A  a  *  9y  ~ ii  v w  [e-] ~ A A~ 9  w  9~A  w  As we have observed, schwa retracts to [A] in the environment of plain uvulars (PHAR DOR) whereas it retracts and lowers to a brief  [a]  in the environment before glottal stop (PHAR).  Retraction always takes place when schwa is adjacent to a post-velar consonant.  Schwa is  labialized when it precedes a labialized (i.e. [rd]) consonant. As seen from (147) above a preceding Labialized consonant does not tend to affect a following schwa to the same degree that a following labialized consonant does. Labialization spreads leftward but tends not to spread rightward (note 95  the variability mentioned in §  Labialization takes precedence over assimilation to a  preceding alveopalatal (DOR[-bk]). Proposed Analysis The analysis proposed here follows from same constraints proposed for the Full Vowels H, u, al and the fact that schwa lacks lexical feature specification. When schwa is followed by a tautosyllabic laryngeal consonant ?, it is systematically lowered to [d]. Retraction to the low vowel [d] interacts with Labialization in the following way. Retraction from a following glottal seems to take precedence over rounding from a preceding labialized consonant.  This not only shows that Retraction to [a] outranks Labialization but  underscores the fact that consonants which follow schwa seem to have a stronger effect than consonants which precede schwa. Notice that retraction to [a] and Labialization are in conflict since there are no low rounded vowels in the language, as discussed in §  Again this  follows from the grounded constraint which states that if a vowel is D O R [lo], then it is not round. This is abbreviated as * L O / R D . A preceding alveopalatal ( D O R [hi, -bk]) can affect the height and place of articulation of schwa depending on the place features of the following consonant, as shown by the data in (143-145) above. Alveopalatals generally share their [hi] specification with a following schwa except when schwa is followed by a post-velar ( P H A R ) consonant. Since retraction of schwa in the environment of a following post-velar consonant ( P H A R ) seems to be categorical (that is, it always takes place), the failure of [hi] spread from a preceding alveopalatal consonants is analyzed as a high-ranking grounded constraint which states that if the vowel (in this case schwa) is P H A R , then it is not [hi] ( * P H A R / H I ) .  A preceding alveopalatal also shares its [-bk] specification with a  following schwa as long as schwa is not followed by a tautosyllabic labialized consonant. In this case, rounding takes precedence over the spread of [-bk]. This is analyzed here as a grounded constraint which states that if a vowel is [rd], then it is not [-bk] ( * R D / - B K ) . This constraint also captures the fact that Sliammon lacks front rounded vowels in its inventory.  96  It is not fortuitous that schwa colouration and the realization of the underlying full vowels show many parallels; in particular, Retraction and the effects of a preceding Alveopalatal, One question we might address here is why does schwa colouration also involve Labialization whereas the realization of the full vowels does not? This is explained in a straightforward manner given the lexical representation of the Full Vowels and the grounded constraints proposed above. Labialization can not occur to either I'll or lal due to the grounded constraints *-BK/RD and *LO/RD. The effects of labialization on lul are not perceived since lul is lexically specified as LAB[rd] DOR. This observation is important since it provides us with a diagnostic for differentiating schwa (and the allophones of schwa) from a reduced full vowel. Full Vowel Reduction is discussed in the next section.  2.4.5 Reduced Full Vowels The prosodic properties of Full Vowel Reduction in Shammon will be discussed in detail in Chapter 4 where it will be argued that reduction is sensitive to stress. A full vowel is reduced in a closed unstressed (post-tonic) syllable in order to improve the resulting Foot structure (cf. §3.3 on Metrical Structure in Shammon and §4.3 on Full Vowel Reduction).  Representative data are  presented here in (151) in which full vowels alternate with their lax counterparts.  (151) Full Vowel Reduction Output  Input  Gloss  a. yax-t-anapi-as-'ui  yax-[a]t-anapi-s-ul  ye .xa .ta .na .ge.. sot He was thmking ofyou  a', yax-t-anapi c  yax-[a]t-anapi c  ye . xa . ta . na . rate.  b. hihw c xa^-ng-mi  hihiwc  b'. xaA-ng-mi c  xaAtiumic  xaAnumi  I remembered you (pi)  he:wc xat. no . ine. I love you very much xat. no . mi£.  I love you (sg)  1  Notice that if this proposal is correct, then we have evidence that the first person subject clitic c J in these contexts is syllabified with the preceding object suffix and is therefore part of the Prosodic Word domain. This will become important in determining the position of clitics within prosodic and morphological domains (cf. Selkirk 1995 on the position of clitics, and Watanabe 2000 on clitics in Sliammon).  97  I have tobacco  c. ?awuk -hV c  ?awu-hu-k C ..  ?A . wo . hok c  c\ ?awuk  ?awuk  ?a . wuk ~ ?A'wx)k  tobacco  d. t'ug-0-as  t'ug-[u]-6-as  to . gu . ©AS  she recognizes me  d'. t'ug-t c  t'ug-|>]<t>c  to . gt?c  I recognize her  e. DM-mixat^ui  mi-m<i>xaiut  mem . l a • tot  black bear cub  e*. mixai  mixai  me.XAJ  black bear  ?iita-ha-n c  ?et. ta . hAnc;  Tve got food  f. ?iitan  ?iitan  ?ei. tAn  eat, food  g- sup=nac-hV a  supna-ha-c a  sop . a§ . ha . ca.  Has he got a tail?  g'- sup=nac  supnac  sop. riAC  tail  w  w  f.  ?i=ttan-hV c  w  w  w  w  w  It is proposed that Full Vowel Reduction entails the loss of a mora associated with the Full Vowel, but that the full vowel retains its phonological features, following Blake (1999). Traditionally, full vowel reduction is often treated as reduction to schwa; that is, the loss of a mora and the phonological features which it dominates.  Within OptimaUty Theory, the  constraints which govern the prosodic structure can be ranked independently from the constraints on featural Faithfulness. It therefore seems entirely feasible to "adjust" the prosodic representation (in this case underparse a mora <u>) in order to create an optimal Foot without affecting the featural content which it dominates. In this way, constraint violation is minimal. It seems important therefore to consider whether or not Full Vowel Reduction in Sliammon is the same as reduction to schwa. It is claimed here that the output of Full Vowel Reduction has the prosodic structure of schwa and the featural content of a full vowel. In order to see this, consider the representation of the three "types" of output vowels (schwa, full vowel and reduced full vowel) presented in (152).  98  (152) Full Vowel  Schwa C  1 1 1 1 1  [f]  Nuc C  '  1  '  1  o  / \  i  Reduced Full Vowel  Nuc I  Nuc  1  1  1  <u>  |  1  [f]  [f]  [f]  Notice that the difference between the surface representation of schwa and the output of full vowel reduction is not whether or not the vowel has features (both do) rather the source of those features. The surface output of schwa is completely determined by the features associated with adjacent consonants (and vowels) whereas the output of Full Vowel Reduction retains its lexical featural content (DOR[-bk] for I'll [i], LAB [rd] DOR for lul [v ~ o], and DOR[lo] for lal [A]). The height of the reduced full vowel is still determined by the height of the adjacent consonants; however, the lexical content of the reduced vowel prevents total place assimilation, in keeping with the grounded constraints and faithfulness, as discussed above. Evidence that Full Vowel Reduction (laxing)  Reduction to Schwa  The next section presents two cases which supports this claim. Basically, the quality of a reduced full vowel is distinct from the surface realization of schwa in a number of contexts. The first case involves what happens to unstressed lil when it is followed by a labialized consonant (C ). If full vowel reduction were the loss of a mora and die phonological features w  associated with this vowel, then we would expect the surface reduced vowel (i.e. the Output) to be identical to schwa in the same context. The data in (153.a'-c') shows that schwa is rounded before a labialized consonant whereas (153.a"-c") Column 3 shows that schwa does not surface as a mid central unrounded vowel [a] in this context.  99  (153) Schwa Colouration Input  Gloss  Output  a. q x-t ga ta qigaO  q axt  a'. IMP-q x-t can  q 9-q xt can  w  w  w  w  q Axt~q dxt  butcher the deer!  q oq xt cm  I'm butchering it  w  w  w  w  w  3  *q aq xt cm w  a". nax ii  b. nx rt b'. CaCPL-nx if w  3  mix if ~nux it  dugout canoe  ni5x nx)x E'l  canoes  w  w  w  w  nax -nax if w  w  w  w  w  *nax n-ux et w  b".  w  c. DIM-Xx =ay+[?]  Xi-Xx ay  titx Ay  small chum salmon  c'.  3bx ay  &5x Ay  chum, dog salmon  w  w  w  w  w  tx =ay w  *Xax Ay  c".  w  Contrast this with the following data in (154) which shows unstressed I'll [e / i] in the same phonological context (i.e. before a labialized consonant). Notice that reduced I'll surfaces as [e / 1 ] , and crucially does not surface as [o I v], as shown by the starred (*) forms in (154.a'-e'). (154) Full Vowel Reduction/i/[e/i] V-reduction <u>  Output  a. masiq  (ma . siq <  massq  a'.  *ma. s a q  Input  w  w  u  u>u  )  Gloss  *masoq  w  w  (t\  b'.  •t^i. paq  c. Xiqiw  ( A I . qiw< )  XsqEW  c'.  *Xi. q a w  *^Eqaw  d. tiqiw  (ti . qiw< >u)  tEqEW  d\  *ti.  *tE'qaw  w  . piq <u> )  e  w  u  e  u>u  u  w  *t' Epoq  w  u  u  qaw  w  dark  horse  e. ?i?agik  (?i . ?a . gi£ <u> ) ?e?agtk  e'.  *?i. ?a . g a k  w  u  clothes  w  w  u  pointed nose  t' e-pEq  b. t' ip=iq e  purple sea urchin  w  u  w  *?e?aguk  w  100  The grounded constraints posited to explain the full vowel allophones §2.4.3 and the output of schwa colouration §2.4.4, also provide an explanation for why reduced Ixl [e / 1 ] fails to undergo rounding.  Since Ixl is lexically specified DOR[-bk], place assimilation and consonant-vowel  interaction is subject to the grounded constraint *-BK/RD which states that if a vowel is [-back], then it is not round. Reduction, which is construed as the underpaying of a mora (Max[p] violation), does not affect the correspondence relations of the features it dominates. Since schwa lacks inherent place features, it is free to undergo Labiahzation, thus explaining the observed contrast between the Output of Full Vowel Reduction and Schwa Colouration.  The second case which shows that the Output of Full Vowel Reduction is distinct from schwa colouration involves lul in an unstressed syllable between adjacent coronals, and a comparison with schwa in the same phonological context.  Consider the realization of schwa  between coronals. As shown by the data in (155) schwa is realized as [a ~ f]. (155) Input  Output  a. C9CPL-sattx b. C a C P L - t ' i n i q  w  w  sa4-sa1tx  w  t'9n-t'i?niq  w  Gloss  sfrsa rtx t  c  women  w  t9nt'e?neq  w  saknonberries  If full vowel reduction of lul were reduction to schwa (i.e. a bare nucleus), then we would expect it to surface as [a ~ i] between coronals. The data in (156) shows that this is not the case. Reduced lul is realized as [v ~ o] in keeping with the proposed analysis; crucially it retains its inherent DOR [rd] specification.  101  (156) Reduced IvJ [o~v] Input  Output  Gloss yesterday  a. jas-'ui  jasui  Jesot:  a*.  *Jas3i  *jESt! , *JES3+  b. Xpx -t-'ut c  Xapx atuic  X9px atvic  b\  *X9px ats4c  *X9px atiic  b".  *X9px at9ic  *X9px atalc  c. k unut'  k u?nut'  k d?not'  c'.  *& u?n9t'  *k d?n9t'  d. xat^-eut  xaf^Out  xdf eot  d'.  •xat^Gat  *xat^e3t, *xdt' eit  w  w  w  w  w  w  w  I broke it  w  w  w  porpoise  w  w  9  fit (clothes)  h  e  The explanation is parallel to the one presented above. When moras are parsed into Feet they are subject to constraints on well formedness, in particular Foot Binarity at the moraic level (FTBINp). The constraint conflict therefore is between the pressure to parse moras (Max[p.]) into well-formed Feet, and to construct Feet which obey Foot Binarity. Since the constraint conflict is of a prosodic nature, additional violations of Faithfulness (i.e. the loss of phonological features) will always entail non-minimal violation of the constraint hierarchy. In the case of IvJ, the loss of the features DOR[rd] would involve exactly these kind of non-mmimal violations.  The resulting candidate  would have more constraint violations, and therefore be less optimal than a candidate which incurs minimal violations (just enough in order to satisfy the constraint ranking). Consider the following tableau which illustrates this point.  102  (157) Jasut [Tesoi]  yesterday  Input: J a . su 4 u  u  u  FAITH [f]  FTBINp  MAX[p]  DOR  [rd] « r a. (ja>i . sot ) u  1 DOR  [rd] *! b. (Ja . su 4 ) u  u  u  DOR  [rd] c.  (ja  u  *!**  . s3t ) u  This is illustrated briefly here to show that the output of Full Vowel Reduction is distinct from Schwa Colouration. Detailed discussion and analysis of the prosodic properties of Full Vowel Reduction will be presented in Chapter 4.  103  Chapter 3: Prosodic Structure of Sliammon Raven did not come on Thursday, He sent nothing. Not a word. Not a sign. Nothing on Thursday. Nothing on Friday. Nothing on Saturday. Nothing on Sunday. Then be sent eagles. Phyllis Webb  3.0 Introduction This chapter is a pivotal chapter. It provides additional evidence for a phonological weight contrast between the full vowels and schwa, thus confirming the hypothesis made in Chapter 2, and it motivates the prosodic structures which will be assumed in Chapters 4 and 5. Arguments regarding the moraic structure of the language are presented in §3.1.  §3.2 introduces the basic  syllable-structure constraints, and §3.3 is a brief introduction to Sliammon metrical structure. Kenstowicz (1993), and Blevins (1995) both summarize the central role of the syllable within linguistic theory. Speech sounds are not simply ordered with respect to one another in accordance with the constraints on possible sequencing; rather speech sounds are proposed to be organized into higher prosodic units of Mora (p), Syllable (a), Foot (Ft), and Prosodic Word (PrWd), following Selkirk (1980a, 1980b), McCarthy and Prince (1986 et seq.). The modified version of the Prosodic Hierarchy which is adopted here is presented in (1), and re-introduces the Nucleus (N) as a linguistic prime, following work by Shaw (1992, 1993, 1995, 1996). (1) Prosodic Hierarchy PrWd  I  Ft a  I  N  104  Shaw (1995,1996) provides the syllable typology in (2): (2) Nuclear Moraic Model of Syllable Structure (Shaw 1993, 1995, 1996) Super-light  Light  Light  Heavy  Heavy  Super-heavy  non-moraic  mono-moraic  mono-moraic  bi-moraic  bi-moraic  tri-moraic  a.  b.  c.  d.  e.  f.  /  I  o  o  o l\  l\  /N / |  /N / |  / |  u  I p  j  /  |  /  C [3]  o  l\ l\  I  / I  C V  C C  o  l\\  l\\  /N / A  /N\ / | \  /N\ / /\ \  /lip.  / \/ C  o  l\  V:  / p - M  / I  I  C V C  /  iiiiii  / \/ | C V: C  This model allows for super-hght syllables which are characterized as Nuclear but non-moraic (=2.a). The model also groups mono-moraic nuclear CV syllables (2.b) together with the monomoraic non-nuclear CC syllables in (2.c). Although they differ in the presence/absence of a vocalic Nucleus, the claim made by this model is that they behave in a similar fashion with respect to their phonological weight.  The non-nuclear syllables such as the ones in (2.c) provide a  representation for obstruent-only syllables; these are also referred to as 'Minor Syllables', following Sloan (1988), Shaw (1996.a, 1996.b). As will be shown in §3.2, Shammon has minor syllables at the right-edge of the word. The reader is also referred to Shaw (1996.a) on Minor Syllables in Lillooet and Bella Coola, and to Bates and Carlson (1997) on Minor Syllables in Spokane (Salish). This model also predicts mono-moraic syllables of following form, following Shaw (1996.b): (3) Light: mono-moraic a. o / N\  / I \ / I VI I I  C [3] C  105  The structure in (3) represents a mono-moraic closed syllable which contains a non-moraic Nuclear schwa and a moraic coda consonant. Schwa most often occurs in this configuration, as discussed in detail in Chapter 5. If schwa is weightless, and syllables are maximally bi-moraic, then we may also expect to find evidence for bi-moraic syllables of the form in (4). (cf. Shaw (1996b) who questions whether or not schwa can license a complex coda.) (4) Heavy: bi-moraic a. a l\  \\  I N \ \  /  |  uu  C  [9]  CC  /III  As will be shown in §3.2, both CgC and CgCC syllables are attested in SUammon. What is of particular interest is the fact that SUammon exhibits a constraint against trimoraic CACC syllables whereas C9CC syllables are attested. This contrast provides additional evidence for the hypothesis that schwa is Nuclear and non-moraic whereas the full vowels Ii, u, al are Nuclear and moraic.  3.1 Moraic Structure Within this model, the mora (u) is the basic unit of phonological weight in keeping with a growing body of Uterature (Hyman (1985), Hayes (1995), Zee (1988), Bagemihl (1991), PuUeyblank (1994), amongst others).  Pulleyblank (1994) presents arguments that moras are  present in the Input rather than assigned by weight-by-position (cf. Hayes 1995). It will be argued in this section that SUammon displays a contrastive weight distinction. In particular, schwa is weightless whereas the full vowels Ii, u, al are mono-moraic.  For example,  Weak Roots of the shape C9C are proposed here to be mono-moraic whereas Strong Roots of the shape CAC are bimoraic. (Recall that "A" in CAC stands for a full vowel). The independent existence of Compensatory Lengthening (CL) in the language motivates preservation of underlying  106  moraic structure and provides evidence that coda consonants are moraic (cf. Hayes 1989 on CL; Blake 1992 on CL in Shammon). The goal of §3.1.1 is to establish the fact that coda consonants in Shammon are moraic. Once this point has been established, §3.1.2 shows that CaC syllables behave differently than CAC syllables. Since there is evidence that a single post-vocalic coda consonant is moraic, then this difference in behaviour is therefore attributed to a difference in phonological weight of the vowel; schwa is weightless whereas the full vowels Ii, u, al are mono-moraic. This corroborates the claim made in Chapter 2 where it was noted that schwa is shorter in duration than the full vowels.  The hypothesis that all post-vocalic coda consonants are moraic is central to the  discussion of the constraints on the distribution of schwa which will be developed in detail in Chapter 5.  3.1.1 Coda consonants are Moraic The purpose of this section is to show that post-vocalic coda consonants are moraic in Shammon.  Evidence is presented from Compensatory Lengthening facts §, Stress  assignment and Full Vowel Reduction §, and the stress properties of the Stative suffix -it §  The conclusion that coda consonants are moraic in the language finds additional  confirmation from judgements regarding prosodic constituency provided by speakers of Sliammon in § Compensatory Lengthening As shown by the data in (5) and following Blake (1992), the loss of a syllable-final glottal [?] gives rise to Compensatory Lengthening of the preceding full vowel nucleus, following Blake (1992). The relevant syllable is underlined in (5) Column 3. Although the data in (5.c-d) show that two variants appear to be in free variation, the existence of vowel lengthening provides evidence for the moraic status of the coda consonant.  107  (5) Compensatory Lengthening: loss of? Gloss  Output  Input  ne's gone driving  a. ga?-ga?t'ap  ga?ga?t'ap  g  a'. ga?t'ap  ga?t'ap  ga;f Ap  b. X aF=iq =uja  x af iq u?Ja  X A ' . t'e .q 6? . JE  joint (human body)  b'. x^tMq^uja  x af q u?5a  x A't' . q 6;j£  wrist  c. tyta  ti?ta  tf?tA~ti:tA  that one (gen.)  d. OyOa  0i?Oa  Qi2eA~ej;0A  that one (fern)  e  e  e  w  e  w  w  w  e  w  9  w  aJ!.ga?.t' Ap e  9  W  w  drive, steer  h  e  e  h  w  w  Loss of syllable-final [h] also gives rise to Compensatory Lengthening, as in (6).  The  morphologically related forms show that the Root is h-final.  (6) Compensatory Lengthening: loss of h Input  Output  Gloss  a. tih  tih  Mi-tii  big  a'. tih=us  tihus  tijaos  big head  tfheq  big nose  a". tih=iq  tihiq  a'". IMP-tih-INC  ti-tih-ih  titiheh  it is getting big  b. IMP-puh-INC  pu-puh-uh  piipuhuh  it is getting windy  b\ puh-?m  puh?9m  puh?Am -- pu:?am  to blow (wind)  b". puh-?m[i]  puh?im  puh?sm ~- pu:h£m  it's windy (state)  c. ?ah  ?ah  M  be hurt, sore  c'. ?ah-stg  ?ahsx  ?a:sx  hurts  c". ?ah-stg c  ?ahsx C  2i:sx c  I'm hurt  w  w  w  w  w  w  w  108  A third context illustrating Compensatory Lengthening entails the loss of a final n in the environment before t, or 6, as shown by the data in (7.a-b). The form in (7.c) shows that the Lexical Suffix (LS) =iq an head is n-final. w  (7) Compensatory Lengthening: loss of n n-deletion  Output  Gloss  a. sp=iq an-6-as  S9?piq a<n>6as  sa?peq a:6AS  he hit me on the head  b.  ximiq a<n>tam  Input w  w  xim=iq an-t-m w  w  napiq an  c. np=iq an  w  w  w  XEmeq a:t9m w  n9peq An w  get clawed in the head brain  To summarize, the loss of a syllable-final {?, h, n} causes Compensatory Lengthening, and therefore provides evidence that post-vocalic syllable-final consonants are moraic. Since h is a fricative, and n patterns with the class of Resonants, the hypothesis made here is that all coda consonants are moraic in Shammon, following Blake (1992).  This is illustrated by the  Input/Output representations in (8).  (8) Input  Output  p p  II C VC  p p  -  w C V:  It should also be noted that Compensatory Lengthening seems to be restricted to stressed syllables. One may also wonder why Compensatory Lengthening does not take place more frequently than it does. Since many cases of consonant deletion (Coronal deletion: n-deletion, t-deletion, t-deletion cf. § occur in order to reduce the phonological weight of the syllable in question, conservation of the moraic structure via Compensatory Lengthening is non-optimal in these contexts.  Furthermore, as will be argued in Chapter 4 there is a high-ranking constraint 109  PEAK PROM FT which ensures that the phonological weight of the head of the Foot is greater than or equal to the phonological weight of the non-head.  Conservation of moraic structure in an  unstressed syllable would therefore cause a violation of this constraint. Stress Assignment and Vowel Reduction One of the central claims made in Chapter 4 is that Full Vowel Reduction occurs in unstressed closed syllables in order to reduce the phonological weight of the non-head. The proposed analysis of the stress facts and of Full Vowel Reduction entail that coda consonants are moraic, as shown by the data in (9.a-d). The brackets in the Output indicate the foot structure: a period is used to indicate syllable boundaries, and u indicates the moraic status of each segment. The data in (9.a'-d') provides morphologically related forms which provide evidence for the representations assumed here. (9) Output  Input a. saL'^awus  sa?a[?]awus  (sa ?a ) (?a^wus ) u  a'. saL'  sa?a  b. fop=awus  u  Gloss two eyes  u  two  fopawus  sa?a ( A e . pa . wvs )  area below the eye  b\ fop  tip  AEp  under  c. Aax=ay  Aaxay  (£a . x A y )  elder (most respectful)  c*. DIM-Aax=ay+[?]  AaAxay  ^aAXAy ~ AaAxay  old person  c". IMP-^ax-INC  Aa-Aax-ax  <ta . £ a . XAX^)  getting old  d. k uyuk  k uyuk  (k ii yi)k )  Fish hook, troll  k ul oyuk vm  trolling  w  w  w  w  w  u  w  w  w  u  w  w  u  u  u  k u-k uyuk -am w  u  u  w  d\ IMP-k uyuk -m  u  w  M  w  If coda consonants were non-moraic, then there would need to be a different explanation for what drives Full Vowel Reduction.  110 Stative-it The special stress behaviour of the stative morpheme provides evidence that this suffix is bimoraic: -i t . Consider the data in (10) which show that the stative always bears secondary u  u  stressed when it follows a Strong Root of the form CAC. This fact is noted and discussed in detail in Watanabe (2000). What is of interest here is that the final CAC syllable is stressed, and resists reduction, as shown by the contrast between the grammatical examples in (lO.a-d) and the ungrammatical examples in (lO.a'-d').  (10) Stative Input  ^-conservation  a. x u q w  u  a  w u  -i t u  u  xu  . q iutu w  uu  -  (*<*0 < w  w  hu .Ji t uu  u  u  b\  °- »°™g  s  s  (  )  state  w  (hd:)(jlf)  already done  *(h°J ) lt  ta  u P u  -i t u  u  ta^.piutu  c\  (ta:)(petHta:Xpfet) tight h  h  *(tapet)  d. t% j - i t u  d'.  Gloss  *(x dq et)  b. h u ^ - i a V  c  Output  u  M  t%? . JVU u  ( « 0 Cjlt Ht' a?)0it ) shade h  e  h  *(t'a?jit) e  The fact that the stative bears secondary stress and resists full vowel reduction suggests that it is lexically footed (cf. Shaw et.al. 1999). Lengthening of the Root vowel preserves the bimoraic status of the initial syllable, since the Root-final consonant is parsed as an Onset to the stative suffix -it, as shown in (lO.a-c). The second variant in (lO.d) shows that the phonological weight of the second consonant of the Root is preserved when a vowel-initial suffix is added. The glottal portion [?] of j continues to occupy the coda of thefirstsyllable, and the j functions as the Onset to the second syllable. This satisfies the Onset constraint while also being Faithful to the moraic structure of the Root, as shown by the Foot structure in (11).  Ill  Output: [t^aTjit*] shade 1  (11) Input: t  El  £T  o  cr I  /  / N\  N\ | \ p H  /  / a  ?  p  J  u t  This kind of conservation of moraic structure provides evidence that post-vocalic consonants are moraic, and that the moraic content of Roots is present in the Input. Restructuring of glottalized resonants is discussed further in §5.2. See Appendix V for a list of Strong and Weak Roots in Sliammon. Moraic Structure: Speaker Judgements Additional confirmation that coda consonants are moraic comes from judgements regarding sub-syllabic constituency provided by speakers of the language. One elder consistently provides moraic units when asked to divide words into "syllables". She often taps out the number of rhythmic beats for each word. It is clear from a comparison of other forms syllabified by other speakers, and from syllable-sensitive processes in the language such as the vowel/glide/obstruent alternations discussed in §2.2.4, that the prosodic constituents provided by this speaker are smaller than a syllable. The fact that these are moraic-sized units is inferred by a comparison of a large number of forms which were morafied by this speaker. Relevant data was collected over a two year period and carefully compared with the judgements given by other speakers. A comparison of related data also enables us to rule out (a) morpheme-by-morpheme breakdown, (b) counting vowels, or (c) counting consonants as a possible interpretation of this speaker's judgements. A sample of the clearest data is presented in (12-13) below.  112  (12) Input  Moraic Structure  Output  Gloss  a. tk -t  tak  pull it (cedar root)  b. px-t  p9x  c. ?sp  ?as  w  d. plk -t  pal  w  w u  u  t  u  tul t  t  u  pAxf*  tear s.t.  ?dsp  finished  w  h  1  u  p  u  k 9t  u  palk 3t  roll it  Moraic Structure  Output  Gloss  SH  XEt'  /ron, metal  w  w  u  h  (13) Input a. Xif  9  X 1  b. x ip-t  t'6 1  |Ll  8  U  x^pit  sweep it  x ip?amin  duster, brush  11  w  x i  pst  w  b'. x ip-?amin  u  u w  w  x i p  ?a mm  w  u  u  u  u  These judgements regarding moraic structure provide further evidence for the non-moraic status of schwa. Compare (12.a) with (13.a) for example. The word tsk t pull it is parsed prosodically as w  tek  t not * t 9 k  w u  u  u  w u  . t . If the initial CaC were bimoraic u  t9uk  w u  , then it would be morafied in  a similar fashion to the first two moras in (13.a): xi t' . However, this is not the case. This kind e  u  u  of contrast provides additional evidence that schwa is non-moraic in these examples. The data in (14) provides further evidence for the moraic status of post-vocalic consonants. (14) Moraic Structure  Output  Gloss  a. cap©  ca  cepO  aunt, uncle  b. k aqt  ka^ q t  c. IMP-t^k-it  t' i t'  e  d. pl'ascan  p3?  bs  Input  w  p^ e  u  u  k aqt  w  u  e  u  u  w  u  kv  u  u  h  they're all screaming  u  ca n u  holler, scream  pa? lAscim 3  u  113  cone of tree (pine, fir) Onset consonants are non-moraic Onsets are non-moraic as seen from the data in (15), in keeping with the cross-linguistic generalization that Onsets do not contribute to the phonological weight of the syllable (cf. Hayes 1995).  (15) Gloss  Output  Input a. k aj-0ut-'ut a c x w  k a>06toi8ecx  w  w  Did you suffer?  w  6 moras; 3 Feet  a'. ( k a . J^Xe&u •touXteu• c x ) w  w  u  u  b. IMP-gasx-?m[i]+[?] a c x b'. (ga . g A S ) ( x l u  u  gagASX8?emAcx  w  u  Are you making lot ofnoise? 6 moras; 3 Feet  . ? e ) ( m A . cx ^) w  u  w  u  If following Hyman (1985), all consonants and vowels are moraic in the underlying representation, then satisfaction of the undominated constraint which requires that all syllables have Onsets in the language will be ranked higher than MAX[u], the constraint which keeps track of correspondence violations. In particular, a mora which is present in the Input but is absent in the Output incurs a MAX[p] violation, following McCarthy and Prince (1995 on Correspondence).  3.1.2 Moraic status of Vowels The data in §3.1.1 establishes that coda consonants are moraic. Now consider the moraic status of the vowels. The following section presents evidence from Stress Assignment in the language which shows that schwa behaves differently than the full vowels Ii, u, al.  This difference in  behaviour is captured by their difference in phonological weight: schwa is non-moraic whereas the full vowels are moraic, thus providing  independent confirmation of the hypothesis made in  Chapter 2, and additional support for the Nuclear Moraic Model of Shaw (1993 et seq.).  114 Long vowels are bimoraic Recall that surface long vowels are derived in Sliammon via Compensatory Lengthening. Long vowels are represented as bimoraic, in keeping with standard assumptions of Moraic Theory.  (16) Compensatory Lengthening Input  Output 1  Output 2  Gloss  a. tyta  ti?ta  ti:ta ~ t i . t A  that one (gen.)  b. 6y6a  eftea-ee'Tea  ei:6a  that one (fern.)  c. ma?-t-as  ma?tAs  ma:tAs  he got it  (17) NUC  A p p \/ V:  The data in (18.a'-c') shows that Strong Roots of the form C A C retain their bimoraic status with the addition of a bisyllabic Lexical Suffix. (18) Output  Gloss  tin  tih~ti:  big  a'. tih=u0in  tihuGin  ti:h60En  big mouth  b. q «uPu-  q up-  q op-  body hair  b\ q up=i6xan  q upi0xan  q d:pe0XAn  hair under arms  b". q up=u0in  q upu6in  q d:p60en  beard, facial hair  c. qwiuV  q it  q Et  beach  c*. q it=axan  q itaxan  q e:taxAn  front of house  Input a.  ti hu u  w  w  w  w  w  w  w  w  w  w  w  w  w  w  115  As shown by the autosegmental representation in (19), the second consonant of the Root functions as the Onset to the vowel-initial Lexical Suffix, and the full vowel of the Root is lengthened in order to maintain the bimoraic status of the Root.  (19) /tih=u0in/ [ti:h6een] big mouth  FT a  I  t  FT a a  / |  l\  / N / p. \i  /N I u  \l  h  / N\ / | u  I /  I I  o  e n  I  i:  I | \  0  Both feet satisfy Foot Binarity at the level of the mora (FTBDSTu) and therefore also satisfy Mhumality. This results in a surface candidate which has adjacent stresses: (cf)(cr a). Full vowels are moraic The stress facts in (20) also provide evidence that coda consonants are moraic in Sliammon since the bisyllabic word [t' if k it ] they're all screaming bears secondary stress. e  e  y  h  (20) Input  Output  Gloss  tr 't'k>lt e e  a. IMP-t^k-it  h  they're all screaming  a'. (PipPp) Cfcy )  2 feet, 4 moras  u  If moras are grouped together in a binary fashion in order to form trochees, then CAC . CAC is quadra-moraic and consists of two feet. If coda consonants were non-moraic, then we would predict a single bimoraic foot: (t' i t' . ki t), and not *(f i t' ) (kl t ). 6  e  u  e  u  e  u  116  u  u  u Schwa is non-moraic Compare the behaviour of the full Vowels in (20) with the behaviour of schwa in (21). As shown by the data in (21), two adjacent CaC syllables form a single foot. If schwa were moraic then the output in (21) should be parallel to the stress facts in (20) above. This is not the case, as shown by the ungrammatical examples in (21.a"-b"). (21) Input  Output  Gloss  a. crn-crh ta qya  cimcirh t  a'.  (&am . 4am^)  a".  *(^n n) (^um )  *4 moras  b. pq-pq  pAq pAq  all white  V.  (paq . paq^  b".  *(p9u lu) (P^uqu)  3  qa? ye 9  the water is cold  u  m  u  h  h  M  c  *4 moras  The goal of this section has been to present additional evidence which shows that schwa is distinct from the full vowels /i,u ,a / in terms of its phonological weight. This provides additional support that schwa is non-moraic (weightless) whereas the full vowels Ii, u, al are mono-moraic. This hypothesis has implications which are explored in the subsequent section.  3.1.3 Implications: CC Roots and Minimality The Problem Words in Shammon can be long involving complex affixation and reduplication; however, there are also restrictions on the minimal size offree-standinglexical items. McCarthy and Prince (1993: 44) provide the following explanation of the derived notion Minimal Word:  117  (22) The prosodic hierarchy and Foot Binarity, taken together, derive the notion "Minimal Word" (Prince 1980, Broselow 1982, McCarthy and Prince 1986, 1990a, 1991a, 1991b). According to the Prosodic Hierarchy, an instance of the category Prosodic Word (PrWd) must contain at least one Foot (Ft). By Foot Binarity, every Foot must be bimoraic or disyllabic. By transitivity, then, a Prosodic Word must contain at least two moras or syllables.  By observing some of the smallest stressedfree-standingwords in Sliammon, we can see that Minimality is generally respected. A word which consists of a Foot is either bimoraic (up) as shown by the data in (23) or is disyllabic (oo) as shown by the data in (24).  (23) Bimoraic Words Input a. pix  w  puq  Output  (piux )  pex  (t'i n )  ten  w  u  b. t'in c.  Bimoraic Foot  u  u  w  Gloss flood barbecued salmon  (pu q )  pdq  d. cuj  (cu y )  cuy  child  e.  xat  (xa 5t )  xaX  want  f.  tan  (tiunu)  tan  mother  Input  Disyllabic Foot  Output  ppa  (pa. PA)  papA  b. pcu  (ps. cu)  picu  c.  (q3. ji)  qXji ~ qaji  w  w  u  u  u  u  u  u  w  brown, grey  (24) Disyllabic Words  a.  qji  118  Gloss pepper  cedar root basket again  Furthermore, at the present point in time only two mono-moraic Roots of the shape CV have been recorded. They are 6u and hu which are both variants of the verbal auxiliary go. (cf. Appendix V, and Watanabe 2000 for a similar observation regarding the general lack of CV Roots in Sliammon). This observation seems to indicate that content words which are monomoraic are in some sense "too small", and are therefore generally excluded by constraints on Minimal Word in the language. The question then is how are 0u ~ hu licensed? It should also be noted that the verbal auxiliary 6u/hu occurs in predicate-initial position, and is often followed by a second-position enclitic (subject, imperative, quotative etc.,) which seem to be footed with the preceding auxiliary, as shown by the examples in (25). By subsuming the enclitic within the domain of the Prosodic Word, the resulting Foot satisfies Foot Binarity.  (25) Gloss  Output  Input a. hu ga t^ux^unis-m  huga t^ux uriisam  hdgA f dx onesam  go brush your teeth!  b. hu ga m?-t  huga ma?t  hdgA ma?!*  go get it!  c. hu ga s?  huga sd?  hdgA  go upstairs!  d. hu c IMP-nsm+[?]  hue nansam  hoc* ninsam  w  9  w  1  se?  1  Tm going swimming  The problem which we need to address here however, concerns the status of free-standing CaC Roots/Stems, such as those cited in (26). (26) Epenthesis  Output  Gloss  a. Ap  fop  XXp  deep  b. &  caf  &i  rain  c. m^  maX"  mXX*  calm (on water)  d. pq  paq  pXq  white  f. XX  xaX  xXX  break (e.g. a rope)  g. mq  maq  mXq  full (from eating)  Input  h  h  119  many, lots  h. qx  qax  qXx  i. Xq  Xaq  XXq  outside  Xaq  XXq  rotten (fruit, berries)  h  Notice that when these Roots occur in a sentential context, they are stressed, as shown by the data in (27). The predicate in question either occurs in word-initial position, as in (27.a-d) or in the position of the main verb with the presence of a verbal auxiliary or other predicative element (27.e-g). The example in (27.h) shows the CaC Root in an overt DP preceded by the article k . w  (27) Foot Structure  Cnitput  Gloss  (maX)  s f W  mXX  calm today  b. pq ta ?aya?  (paq ) ta?aye?  pXq  the house is white  c. xX ta x ifm  (xaX)  XXX  the rope broke  d. qx pu?px  (qax;) pu?px  qXx  lots ofkindling  e.  hihw  he:w (Xap*)  XXp  it's really deep  f.  k n a cx mq  k Xnascx (maq)  mXq  Are you full?  g-  hu ga tq  hdgA  (*aq )  XXq  go outside!  h.  hihw  bet  it's really raining  Input a. mt s P u i  w  w  t X e?lam 3  w  w  w  w  h  h  1  Xp  w  w  say-mut k bt w  w  h  he:w saymvt k (6ai) h  w  The (CaC) Root in question is footed as indicated by the presence of primary stress and the brackets to indicate the Foot boundary. It should be noted that the article ta and the syntactic nominalizer s are proclitics, and are therefore phonologically dependent on the Prosodic Word which follows in (27.a-c); these proclitics do not affect the footing of the (CaC) predicate which precedes them. Given the model adopted in this thesis, schwa is characterized as a bare Nucleus, and does not have any phonological weight associated with it (i.e. it is non-moraic), as argued in Chapter 2. The question, then, is do these stressed lexical words of the form Ca'C in (26-27) satisfy or violate 120  Minimality? Recall that Minimality is derived from the Prosodic Hierarchy and its interaction with FOOT BINARITY,  as in (22). It is clear from the examples in (26-27) that CaC words do not satisfy  Foot Binarity at the level of the syllable since they are clearly not disyllabic; however, do these words satisfy Foot Binarity at the moraic level? In other words, are they represented as in (28.a) or (28.b)?  (28) Input a.  cH  b.  6i  6[a]f  u  u  cta ]4  u  u  Output  Gloss  (cf<f)  rain  (cf^a)  rain  u  u Discussion and Proposed Analysis There seem to be two plausible lines of argumentation. First, although schwa often behaves phonologicaliy as though it is non-moraic, schwa in this context could be constrained to be both Nuclear and moraic in order to satisfy Foot Binarity (=28.b). The cost of such an analysis would be the insertion of additional structure; in particular, a violation of both DEP[NUC] and DEP[p]. This implies the following partial constraint ranking:  (29)  FTBINp  »  DEP[NUC], DEP[u]  In this case, Foot Binarity at the level of the mora must be satisfied at the expense of a DEP[NUC] and a DEP[u] violation. The other possible analysis is that the constraint FTBINp is ranked in such a way that it is violated in this context, and that these CaC forms represent degenerate feet in Shammon. Given Optimality Theory, FTBINp would be violated in order to satisfy some higher-ranked constraint in the grammar, yielding the output in (28.a). A plausible candidate is the interface constraint which ensures that every lexical content word receives stress. The following discussion pursues this hypothesis. 121  Prince and Smolensky (1993: 43) discuss Lx«PR(MCat) which requires that a member of a morphological category (MCat: Root, Stem, Word) corresponds to a prosodic category (PCat: Foot, PrWd).  They also propose that these relations between morphological categories and  prosodic categories can be achieved via Alignment (cf. also McCarthy and Prince 1993.b on Generalized Alignment). Alignment ensures that the left or right edge of a morphological category matches with the corresponding edge of the relevant prosodic category, making special reference to an edge (cf. also Chen 1987, Selkirk, Nespor and Vogel, McCarthy and Prince 1993). The alignment constraint in the case of Sliammon is given in (30).  (30)  Align L (MStem ; Foot) Align the left edge of every morphological stem with the left edge of a Foot (Ft).  If the interface constraint Align L (MStem; Ft) is ranked above Foot Binarity (FTBINu, o), then it will be more important to ensure that the left-edge of every stem is aligned with a Foot than it is to satisfy FTBINp..  Therefore, CaC Roots, which themselves are well-formed stems, are stressed  eventhough they fail to satisfy Foot Binarity, as shown by the partial ranking in (31).  (31) hi hi  [Hi] rain  Input: hi  u  ALIGN L ( M S T E M ; FT)  FTBINu  DEP[NUC]  «• a. (h%) b.  hip  *t  The optimal candidate in (31.a) is footed and satisfies the Align constraint at the expense of creating a mono-moraic Foot. Candidate (31.b) is ruled out since the lexical content word hi, which is itself a stem, fails to be aligned with the left-edge of the Foot. One of the questions which this proposal raises is why there is no augmentation in order to satisfy Foot Binarity?  If the constraints on Root Faithfulness are dominant, in particular 122  DEP[u]R t then strengthening will be ruled out, as shown by the tableau in (32). Weak Roots are 00  3  faithful to moraic structure.  (32)  bi &i  Input: hi  [cni] rain  ROOT FAITH: DEP[u]  u  ALIGN L (MSTEM ; FT)  FTBINp  DEPfNUC]  a. (c9i ) u  b. (69 4 ) U  c.  C94  U  *! *!  u  This may be considered an unorthodox proposal given the claims of Prince and Smolensky (1993: 109) who suggest that "Lx=PR and FTBIN are universally undominated." However, in the true spirit of OT, all constraints are violable. The proposal which is made here is that in a limited set of cases Foot Binarity is violated in order to satisfy Root Faithfulness. A Root in a non-derived domain can be sub-minimal because of Faithfulness.  This appropriately  characterizes  the  degenerate mono-moraic (CdC ) feet in Sliammon. This proposal seems to find support in the fact u  that these words are shorter in duration than free-standing bimoraic words, such as those in (23), and as documented in Chapter 2.  3.1.4 Summary This section has presented arguments for the moraic structure which is assumed in this thesis. Of particular importance is the claim that all coda consonants are moraic in Shammon. In addition, evidence from stress assignment provides support for the claim that schwa is non-moraic in contrast to the full vowels which are moraic, since CaC functions as light whereas CAC behaves as heavy. This claim has interesting implications regarding the licensing of free-standing C3C stems.  123  3.2 Syllable Structure This section outlines the basic syllable structure constraints in Shammon, building on the basic descriptive generalizations of Blake (1992). Two new observations are worth noting. First of all Sliammon makes limited use of so-called "minor syllables"; these are obstruent-only syllables of the shape CC. These non-nuclear syllables occur at the right-edge of mono-morphemic stems, such as sat . tx woman or t^am . q f w  w  cloud . These syllables are of particular interest  with respect to our discussion of the distribution of schwa in Shammon since these "extra" consonants do not trigger schwa epenthesis in order to satisfy the constraint that all syllables have Nuclei (cf. § The second observation relates to the general ban on Complex Onsets in the language. Although Shammon lacks word-initial Complex Onsets, there are a limited number of wordinternal st- Onsets which have not been discussed elsewhere.  3.2.1 Simple Syllables This section discusses "simple" CVC syllable structure in Sliammon, and establishes the constraints on the occurrence of syllable-internal "constituents" : Onset, Nucleus and Coda. Within the Nuclear Moraic model, the "Onset" is identified as the non-moraic consonant which precedes either a Nucleus, or a moraic consonant (e.g. in the case of a CCp Minor syllable). The term "coda" refers to the moraic consonant which occurs either after the vocahc nucleus and within the same syllable, or after the non-moraic onset in the case of CCp syllables. I will continue to use the convenient labels "onset" and "coda" in the discussion which follows. It will be argued here that all syllables in the language have a single Onset consonant, and that coda consonants are permitted. Section argues that although all content words surface with a vocahc nucleus, not all syllables do.  This section makes an important contribution to our understanding of  Shammon phonology in that it documents the existence of so-called Minor Syllables in the language, and explains the observed asymmetry between the numbers of consonants allowed at the beginning of words versus the number of consonants permitted word-finally (cf. Blake 1992 on 124  extra-metrical consonants at the right-hand edge of the word domain). Complex constituents, such as Complex Onsets and Complex Codas, are discussed in §3.2.2. Onset The Data Words in Sliammon generally begin with a single consonant, as shown by the monomorphemic lexical items presented in (33). (33) Input  Output  Gloss  a.  mixai  mexAi  me. x A i .  black bear  b.  puhu  pdho  p d . ho .  raven  c.  k uma  k uma  k u  .m a .  raffish  d.  q uw9t  q dwut ~  q d  . wut.  beaver  e.  q alas  q alAS  q a.  IAS .  raccoon  f.  waxas  w a x AS  wa.XAS.  w  w  w  w  w  q dwit  w  w  w  w  w  (messenger)  green frog  There are no vowel-initial words in the language, nor is there any evidence for vowel-initial Roots either. Roots are always consonant initial, as shown by the data throughout this dissertation, and by the Root List in Appendix 5. When a vowel-final Root is followed by a vowel-initial Lexical Suffix, an epenthetic [h] intervenes, as shown by the data in (34). (34) Root=LS Input a. waxat'i=aya e  wa~Xati[h]aya 9  a'. waxat' i e  b. kapi^aya b'. kapi  kapi[h]aya  Output  Gloss  waxat e[h]ay£  pipe case  waxat^e  pipe (for smoking tobacco)  k api[h]aye  coffee pot  k api ~ kapi  coffee  ,9  y  y  125  c. qnayu=aya  qdnayujjijaya  q/nayojjhjlye  sewing needle case  c'. qn=ayu  qsnayu  q/nayo  sewing needle  ^StO^k^  sheep's wool, I. sweater  lamato  sheep (< Fr. via CJargon)  d. lamatu=uk t w  lamatu[h3uk t w  d'. lamatu  Other cases of vowel hiatus within the affixal domain involve deletion, as shown in (35), and discussed in detail in Blake (2000). (35) Input a. ca-cag-a6i-as  Output  Gloss  6ecEga0is  s/he is helping you (sg)  b. ya-yaf-a6i-as  yayaiaSis  yeyeiaQts  s/he is calling you (sg)  c. ca-cag-at-anapi-as  cacagata?napis  cEcegat3?n3pis  s/he is helping you (pi)  d. ya-yai-at-anapi-as  yayafatsnspjs  yeyeiatanapis  s/he is calling you (pi) Proposed Analysis The Onset constraint is defined in (36) following Prince and Smolensky (1993:25) and McCarthy and Prince (1993). Since there is no evidence that this constraint is ever violated, it is proposed here that it is undominated in Shammon.  (36)  ONSET  Every syllable has an Onset  The constraint ONSET must dominate DEP[h] which militates against the insertion of epenthetic [h]. Since [h] is the least specified consonant in the system, and is arguably characterized as having a PHAR  specification, the constraint DEP[h] or DEP[PHAR] will keep track of the cost associated with  h-epenthesis.  126  (37)  DEP[h] An [h] which is present in the Output and not present in the Input will incur a DEP[h] violation.  DEP[h] belongs to the family of Faithfulness constraints within a correspondence model of Optimality Theory, following Prince and Smolensky (1993) and McCarthy and Prince (1994). An [h] which is present in the Output but is not present in the Input violates DEPfh] since there is a lack of correspondence between the Input and the Output. If ONSET and ROOT FAITH (i.e. MAX[f]) outrank DEP[h], then an epenthetic [h] occurs in order to provide an Onset, and also to avoid deletion of either the vowel belonging to the Root or the vowel belonging to the following Lexical Suffix (LS). As argued in Blake (2000), Lexical Suffixes in SUammon behave like Roots with respect to vowel hiatus, and are therefore analyzed as bound Roots. As bound Roots, Lexical Suffixes will be subject to Root Faithfulness constraints rather than to Affix Faithfulness, following Blake (1995, 1996, 2000). Consider the tableau in (38) which shows h-epenthesis between a Root and following LS =aya place, container.  (38) tala=aya [talahaye] purse  Input: taUv=aya  ROOT FAITH (MAX)  ONSET  DEPlTll  a. ta. l a . [h]a . ya *!  b. ta. la<a> . ya c. ta. l a . a . ya  *!  What is interesting about Sliammon is that vowels in hiatus are treated differently depending on their morphological affiUation, but the high-ranking Onset constraint is always satisfied.  127 Gemination: Onsets and Faithfulness to Moraic Structure The data in (39) show that an intervocalic resonant is systematically parsed by speakers as the coda to the preceding syllable and as the Onset to the following syllable. This occurs with CC and CCC Roots in (39), and with some examples of Roots containing a full vowel in (40).  (39) Syllabification  Output  Input  Gloss  a. q f a cx m  q uTaecx vm  q 9l'.  la?c . x am  Are you sg. coming?  b. q f a cap sm  q t>laecips9m  q 9l'.  lae . clp . S 9 m  Are you pi. coming?  c. wn-?m-min  wonamin  W 9 n . na . mm  a drill  d. kl0+[i]  kThe  161. ue  crooked  e. tTk+[i]  t'Xhk  Ul. h i  a hole  Input  Output  Syllabification  Gloss  a. ?ima6  ?emA0  ?em. mA0  grandchild  b. ?ayis  ?ayis ~ ?ayis  ?ay. yis  man's sister, cousin  c. ?aya?-s  ?aye?s  ?ay. ye?s  his house  w  w  w  w  w  w  w  w  w  (40)  d. C3C -Janx  w  PL  e. k ul=awtx w  w  j9n . jen . n9x  jinjenvx  w  k ulawtx w  w  w  k uT . law . tx w  w  lots of fish school  If all coda consonants are moraic in the language as evidenced by Compensatory Lengthening, and the stress facts in §3.1.1 above, then there is no inherent length contrast in postvocalic consonants in the language. If post-vocalic consonants were ambi-syllabic (non-moraic) then these would be the only non-moraic coda consonants in the language. Since this would be non-structure preserving in the sense of Kiparsky, it is proposed here that these consonants are moraic, and that gemination satisfies the constraint that all syllables have Onsets. In addition,  128  gemination also satisfies the constraint M A X [ f i ] which ensures a Faithful parse of underlying moraic contrasts. This proposal is confirmed in a number of different ways. Native speakers certainly syllabify these strings differently, and Harris (1981) writing on Island Comox treats these consonants as geminates. Both P. Kroeber and H. Watanabe (p.c.) both note that these consonants may be longer in duration, judging from their own transcriptions of SUammon. These array of facts leads me to hypothesize here that these consonants are moraic. Notice that gemination of an intervocaUc consonant has the effect of satisfying the highranking Onset constraint in the language while also maintaining the moraic structure of the Root (cf. Chapter 5). Nucleus This next section addresses the status of the Nucleus in SUammon. There is a growing body of literature which recognizes the Nucleus as the core of the syllable. For example, Levin (1985) argues that the syllable is projected from a single primitive category Nucleus which is the head of the syllable. Shaw (1992) adduces templatic evidence in favour of a Nucleus based on her analysis of reduplication in Nootka and Nitinaht (Wakashan), and Ojibwe (Algonquian). Early reference to the role of the Nucleus in syllable structure include Trubetzkoy (1939), and Fudge (1976), amongst others (cf. also references in Anderson (1985), and Kenstowicz (1994)). Bagemihl (1991), Shaw (1993, 1996) make reference to the Nucleus in their discussion of Nu^alk (Bella Coola) and St'at'imcets (Lillooet). Within OptimaUty Theory, the central role of the Nucleus is characterized by the constraint SYLL NUC, following Prince and Smolensky (1993: 87). (41.a) SYLL NUC  Syllables have vocaUc nuclei  Shaw (1996.c) captues this same generalization with reference to Proper Headedness: (41 .b) PROPHEAD o  A syllable is headed by a NUC [=SYLL NUC]  129  This is important since it will be argued in §  that although many/most syllables in  Sliammon satisfy SYLL NUC/PROPHEADa, there are a limited number of obstruent-only syllables in the language — syllables which clearly violate this constraint. The presence of a surface full vowel satisfies SYLL NUC / PROPHEADo in the data in (42). (42) Full Vowels Input  Output  Gloss  a. xit'  xit'  xii  iron, metal  b. k in  k in  k en  how many?  c. iuk  fuk  *iik  to fly  6  e  w  6  w  w  d. wuk  w  w  w  wuk  w  w  wuk  e. k as  k as  k as  f. pal'  pal'  pal'  w  w  w  scoop net hot (temperature)  w  heron  The presence of schwa in the Output in (43-44) also satisfies this constraint. Since constraints in OT are constraints on outputs, whether or not schwa is present in the Input is irrelevant to the satisfaction of SYLL NUC - the constraint evaluates whether or not each syllable contains a Nucleus in the Output. (43) Schwa Output  Gloss  q#  again, still  pscu  picu  cedar root basket  nax ii  nx5x tt  dugout canoe  d. tsk fi  tak fi  tvk ie  rabbit  e. tsq a  ts?q a  ta?q a  octopus, devil fish  f. qaya  qa?ya  qa?ye  water  g. mana  m9?na  ma?na  one's child, offspring  Input a. qaji b. pacu c.  n3x if w  w  w  w  w  w  w  w  w  130  (44) Schwa epenthesis Root  Schwa Epenthesis  Output  Gloss  qji  q[3]ji  qgji  again, still  b. pcu  p[a]cu  picu  cedar root basket  c. n x r r  n[9]x i<r  no5x i't  dugout canoe  d. tk 4i  t[3]k 4i  tuk 4e  rabbit  a.  w  w  w  w  w  w  e.  tq a  t[a]q a  ta?q a  octopus, devil fish  f.  qya  q[9]ya  qa?ye  water  m[9]na  ma?na  one's child, offspring  w  w  w  g- mna Coda The Data Sliammon has syllables which are closed by a moraic coda consonant, as shown by the data in (45). The syllabification in Column 2 is provided by speakers o f Sliammon.  (45) Closed syllables Input a.  DM-k upa+[?] w  b. xap c.  t' ix -t' ix  d.  mus-mus  e  w  e  w  e. t'aq-t'aq  It w i l l be argued  Syllabification  Output  Gloss  . k u k . pa? .  k uk pa?  grandfather  . xap.  xap  cradle basket  w  w  . f i x . t' ix 3  w  e  v  w  t' ex t' ex e  w  w  e  w  fish hawk  . mus . mus  musmus  cow  . t'aq . t'aq .  taqt'aq  slow  in §  that Sliammon generally  [k p, x t' , sm, qt] are not licit Onsets. w  w  e  131  lacks Complex  Onsets, so that Proposed Analysis Since coda consonants occur frequently in the language, this entails that the constraint NO CODA must be low-ranking in Sliammon. In addition, the family of Faithfulness constraints which ensure that a lexically specified post-vocalic consonant has a surface correspondence must be higher-ranking. The relevant constraints are given in (46). (46)  / NO CODA  Syllables are open (M&P 1993: 10)  MAX[f]  Each feature [f-element] in the Input is in a correspondence relation with a feature in the Output.  DEP[f]  Each feature in the Output is in a correspondence relation with a feature in the Input.  The constraint MAX[C] is a constraint which evaluates the cost associated with deletion of a consonant whereas DEP[C] is the constraint which evaluates the insertion of a consonant which is not in the Input. The constraint which ensures that syllables are open (NO CODA) is violated in order to satisfy Faithfulness (MAX and DEP constraints), as shown by the partial ranking in (47) and the tableau in (48).  (47)  FAITHFULNESS »  NO CODA  (48) xap [xap] cradle basket Input: xap  FAITHFULNESS  NO CODA  i®- a. xap  b. xa<p>  *! MAX[C]  c. xap[a]  *! DEP [a]  132  As shown by the tableau in (48), there is a faithful parse of the consonants and vowels in the Input and the correspondence relations between segments/features in the Input/Output are maintained. Since Faithfulness is ranked higher than NO CODA, deletion of the coda consonant is ruled out, as shown by candidate (48.b). Faithfulness also rules out vowel epenthesis which occurs in order to try and avoid a violation of NO CODA, as in (48.c).  3.2.2 Complex Syllable-Internal Constituents More traditional linguistic theories use surface phonotactics to determine the possible types of syllables in a language. It will be argued in §, that while Sliammon typically avoids Complex Onsets, a limited number of word-internal [st-] Onsets do occur in order to satisfy higher-ranking constraints on the Alignment of prosodic and morphological constituents. Violation of *Complex Onset is therefore optimal in a specific context - and occurs just when required to do so by a higher-ranking constraint. The ban on Complex Onsets cannot be viewed as a surface true generalization across the entire set of syllabified words.  Again it is constraint  ranking, constraint conflict and minimal violation which determine the most harmonic output. In §, it will be argued that Sliammon avoids heteromorphemic CA-CC syllables since they violate the constraint against trimoraic syllables in the language (*ppp] ). At the same 0  time, trimoraic syllables do occur in mono-morphemic Roots, such as CACC. High-ranking Root Faithfulness and the constraint on Root Contiguity drive violation of *ppp]cr.  It is therefore not  possible to make a statement regarding the output of syllabification which follows simply from the surface phonotactics and which does not take the morphological constituency into consideration. As pointed out by McCarthy and Prince (1993), a model of phonology in which constraints or generalizations about the language musr be surface true is untenable. Within OT, the optimal candidate may actually violate a number of constraints in order to satisfy some higher-ranked constraint within the grammar. The syllable typology which immerges is therefore complex and derived from the interaction of prosodic and morphological constraints.  133 ""Complex Onset Not only do syllables in Sliammon have an obligatory Onset but there is also a general lack of word-initial consonant clusters in Sliammon. The general ban on Complex Onsets is discussed c  in the next section. The Data Schwa occurs between the first two consonants of a C R ' V Root, as shown by the data in (49) Column 2. Given a theoretical framework which lacks constraints on Inputs, consider what would happen i f the speaker posits an Input representation in which schwa is not present in the Input. A s shown by the output forms in (49), schwa occurs between the first two consonants, and avoids a violation of the constraint *COMPLEX ONSET.  The ungrammatical forms in (49.a'-e')  show that Sliammon lacks word-initial C R ' Onsets. (49) C R ' V Roots  Input a.  sma  a',  sma  s a ? . ma  mna  Gloss  sa?ma  mussel  *sma  a", s m a b.  Output  * s m . ?a m a ? . na  marna  b'. mna  *mna  b". mna  *mn . ?a  c.  ?l'as / ?al'as  ?a? . las  ?a?lAS  c'. ?l'as  *?las  c". ?l'as  * ? 1 . ?as  d. q y a  q a ? . ya  qa?ye  d'. q y a  *qya  d'. q y a  * q i . ?a .  134  one's child, offspring  sea cucumber  water  e. k wa  k 9?. wa  w  w  k a?wa  e\ k wa  *k wa  e". k wa  *k u.?a.  w  stomach, belly  w  w  w  w  Note also that the resonants m, n , 1, y, w are not syllabic, as shown by the ungrammatical forms in (49.a"-e"). The glides y, w do not undergo vocalization in this position either in order to satisfy the constraint which requires a vocalic nucleus, as shown by the contrast between (49.a-e) and (49.a"-e"). 1  A survey of word-internal obstruent resonant clusters and the syllabification of these clusters shows that *CR Onsets are banned in word-internal syllables as well. (50) Syllabification  Output  Gloss  a. ?atnupil  ?at. nu . pil  ?atnopel  car, automobile  a'.  *?a . tnu . pil  b. saplin  sap. lin  saplen  bread  b*.  *sa. plin  c. Oicmus  6ic. mus  0icmos  back ofhouse  c'.  *0i. emus  d. lkli  lak. Ii  likle  key  d\  *la. kli  e. watla  wat. la  e'.  *wa. tla  Input  f. f.  DrM-x il'[-i-]m+[?] w  sweetheart  x ex lem  string, thread  h  x ix . lirii w  watla ~ wat la  w  w  w  *x i. x lim w  w  'It is not that glides never undergo glide/vowel alternations; they do. The factor which may be affecting the surface form in these cases is the presence of a glottalized resonant R'. Notice that CVR'V is systematically realized as Cv?Rv and not as *CvR?v (§5.2). Glottalization occurs adjacent to the stressed vowel and occupies syllable coda position, not the Onset of the weak member of a stress foot.  135  Not only does Sliammon lack CR Onsets but it also lacks OO (obstruent) Onsets as well, as shown by the data in (51). It should be noted that it is quite difficult to find mono-morphemic word-internal obstruent clusters since canonical Roots are predominately CVC/CaC or extended CVCV, CVCVC in shape. I have included diminutive forms since the word-internal consonant cluster occurs within the Root and not across a morpheme boundary, so as to avoid complications with syllabification potentially being affected by the presence of the edge of a morphological category.  (51) word-internal CC clusters Syllabification  Output  Gloss  a. x atq -m  x at. q 9 m  x atq om  thunder  a'.  *x a. tq 9m  tuk 4e  rabbit  k uk pa? w  grandfather  q aq s9m  little flower  x bx sim  small soapberry  Input w  w  w  w  w  w  w  b. tk i:i  t 9 k . 4i  b\  *t9.k 4i  w  w  w  w  w  c. DM-k upa+[?]  k u k . pa?  c\  * k u . k pa?  w  w  w  w  w  d. DHM-q asm+[?]  q aq . ssrh  d\  *q a . q s 9 m  e. DIM-xus-[i]m+[?]  x u x . sim  e'.  * x u . x sirh  w  w  w  w  w  w  w  w  w  w  w  w  w  w  The lack of CR onsets predicts the lack of obstruent obstruent (OO) onsets given the sonority sequencing generalization of Clements (1990). If a language allows an Onset cluster with a level sonority profile it typically allows an Onset cluster which has a rising sonority towards the nucleus (OR), given markedness. Notice that the word-internal cluster is heterosyllabic, as shown by the Sliammon data in (51 .a-e).  136 Proposed Analysis The relevant syllable structure constraints are provided in (52)  (52)  NO CODA  Syllables are open  (M&P 1993:10)  •COMPLEX ONSET  Syllables do not have complex onsets  The Sliammon facts in (49-51) show that it is better to incur an extra violation of the constraint NO CODA than it is to violate *COMPLEX ONSET.  This partial ranking is given in (53) and illustrated  by the tableau in (54).  (53)  *COMPLEX ONSET »  (54)  DIM-k upa+[?]  [k uk pa?] grandfather  w  w  k u-k pa? w  NO CODA  w  w  NO CODA  COMPLEX ONSET  «• a. k iik . pa? w  w  *!  b. k u . k pa? w  w  As observed in §, Faithfulness (here MAX [C]) is ranked higher than NO CODA in order to prevent deletion of a Root consonant, as shown by the tableau in (55).  (55) DM-k upa+[?] k uk pa? grandfather w  w  k u-k pa? w  w  w  ROOT FAITH: MAX [C]  **  a. k \ik .pa? w  w  b. k u<k >.pa?  *!  c. k ii.k <p>a?  *!  w  w  w  w  NO CODA  *  137 Comparative Evidence for the lack of Complex Onsets The next section provides comparative evidence which shows the lack of complex onsets in the language. As shown above, only a single consonant appears in word-initial position. In fact, one of the striking properties of Sliammon is the absence of the nominalizing prefix s- which is found in all of the other Salish languages (cf. Davis 1970:15 and §5.6). Shammon / Sechelt Data Compare the Sliammon (SI) and Sechelt (Se) forms in (56) which show the absence of this widespread prefix in Sliammon, cited from Blake (1992:49). Sechelt data are cited from Beaumont (1985), abbreviated RCB. The forms he cites are provided in the Sechelt practical orthography and appear in angled brackets <>. I have reconstructed the Input/Output forms based on the guide to pronunciation (Beaumont 1985:5-13). (56) Initial Complex Onsets: A comparison of Sliammon and Sechelt cognates Input  Output  Gloss  Language  a. nx iL  m5x ii  dugout canoe  5/  a'. s-nx ii:  sn9x ii  canoe (RCB)  Se  b. q ayx  q /yx ~ q ey x  firewood  SI  b'. s-q yx  sq Xyix  c. x s  X AS  w  w  w  w  w  c'.  w  <snexwilh>  w  w  w  w  9  <skw'eyex>  animal fat, lard SI  W  S-X 9S W  SX AS W  d. tumis  himis  d'. s-tumis  s tonus  firewood (RCB) Se  <sxwes>  <stumish>  138  grease (RCB)  Se  man  SI  man (RCB)  Se Sliammon / hdnqaminam (Musqueam) Data The historical loss of the widespread nominalizing prefix s- is shown again by a comparison of Sliammon and hanqsminam (Musqueam) cognates in (57.a-i'). Musqueam (Msq) is a related Coast Salish language spoken in the lower Fraser River delta, and is separated from Sliammon (SI) by Squamish and Sechelt territories . 2  (57) Initial Complex Onsets: A comparison of Sliammon and Musqueam cognates  Output  Gloss  Language  a. nax iL  nuVM ~ nvx :ii  dugout canoe  SI  a'. S-n9X 9"f  Sn9X 9t  canoe  Msq  b. qa?X"  qa?X"  sea otter  SI  b\ s-qe:X  sqae:X"  sea otter  Msq  c. k ak =aju  k ak a?Ju  squirrel  SI  c\ s-k aya?  sk aya?  squirrel  Msq  d. £gq  &Aq  robin, tiny bird  SI  d'. s-k qeq  sk qeq  robin  Msq  e. "fagat'  i"a?g9t'  herring  SI  e\  stew9t'  herring  Msq  w  w  w  w  w  w  2  s-*tew9t'  w  w  w  w  w  w  3  The hanqaminam (Musqueam) data are cited with permission from the collaborative Musqueam/UBC FNLG  materials: © 1999 Musqueam Indian Band and UBC FNLG. I gratefully acknowledge the contribution of all of the elders who have made this research possible; especially Adeline Point, the late Edna Grant, the late Arnold Guerin, the late Dominic Point: hay ce:p qa si:?em. These cognates exhibit other well-attested sound correspondences: SI t corresponds to Msq k .  3  w  SI g corresponds  to Msq w; SI s corresponds to Msq x. In addition, Proto-Salish *u became a/e in hanqaminam, as noted by Elmendorf and Suttles (1960), and u in Sliammon.  See also Kuipers (1981, 1982), Kinkade (1998) , Kroeber  (1991/1999) for comparative Salish sound correspondences, and Suttles (forthcoming) on hanqaminam.  139  f.  Saqay  Osqay  sockeye  SI  f.  s-Osqay  sOsqay ~s09qi?  sockeye  Msq  t^ek™  worm  SI  g'. s-t' ak  st' 9k  worm  Msq  h. W i q  t' et' eq  mud  SI  h'. s-t^iqal  st^iqal  mud  Msq  tu-mis ~tonus  man  SI  staniLx  warrior  Msq  ge  i.  w  e  6  tumis  i'. s-tamsx  w  e Proposed Analysis How do we account for the loss of the nominalizing s-prefix in Sliammon? What would happen if an s-prefix were posited? As observed in (49-51) above, Sliammon generally lacks Complex Onsets. If the constraint which bans complex onsets (*COMPLEX ONSET) is ranked above the constraint which requires a faithful parse of the prefix s-, then it would be better to delete the s-pfefix than to violate the constraint which bans complex onsets, as shown by the partial ranking in (58) and the tableau in (59).  (58)  *COMPLEX ONSET »  (59)  n3x it [ni5x ii ~m5x :i r] dugout canoe w  w  AFFIX FAITH (MAX)  w  s-n[Tj]x ii  c  AFFIX FAITH (MAX)  •COMPLEX ONSET  w  "sr a. <s-> m5x if w  b. s-nijx i4 w  *!  This is discussed further in Chapter 5 in which the s-nominalizing prefix is compared with the plural IV-I prefix.  140 Root-Initial Consonant Clusters If we assume that the distribution of schwa is predicted from the constraint ranking, contrast what happens to an initial C-prefix in Sliammon with what happens to CCVC Roots in the language. Sliammon / hanqaminam (Musqueam) Data The Sliammon and Musqueam cognates in (60) show that Musqueam retains Root-initial complex clusters while Sliammon has an epenthetic [a] between Ci and C 2 of the Root.  (60) Sliammon / Musqueam Comparative Evidence Output a. q a4e?=san w  •Complex Onset  Gloss  Language  *q 4e? . san  shoe(s)  SI  shoe  Msq  bow of boat  SI  bow of canoe  Msq  w  a'. q 4ey=xan w  b. qaion  *qron t  b'. sqian C.  A3X -t w  *AX a't w  beat him in a game Msq  c'. Ax -at w  d. t'aq am w  *t'q am w  d\ t'q 9m w  e. patt  *p"iat  e'. piet f. R as-t w  *k sat w  f. k x-et w  g. k a<r-t w  g'. k 4-et w  beat s.o. in a contestSI  *k <rat w  thimbleberry  SI  thimbleberry  Msq  thick (layer)  SI  thick  Msq  count it  SI  count them  Msq  spill it  SI  spill it  Msq  What we observe is that instead of deleting the first consonant of the Root, schwa always surfaces between C\ and C 2 of the Root in Sliammon. CCVC Roots undergo epenthesis in Sliammon in order to avoid a violation of *COMPLEX ONSET, as illustrated above. This is in contrast to the 141  treatment of the hypothetical cases discussed above which would involve deletion of a nominalizing prefix s-. Proposed Analysis If a speaker posits an Input such as /qtun/, the relative ranking of * COMPLEX ONSET and DEP[NUC] (the constraint which keeps track of the cost associated with schwa epenthesis) drives schwa epenthesis and selects candidate (61.a) over candidate (61.b).  (61)  qtun qaion bow of the boat  Input: qtun  COMPLEX ONSET  DEP[NUC]  " 5 * a. qafun  *!  b. qtun  The tableau in (62) shows that MAX[C] ROOT is also ranked higher than DEPfNUC] since deletion of either Ci or C 2 of the Root in (62.b-c) is clearly less optimal than candidate (62.a) which involves schwa epenthesis.  (62)  qtun qaton bow of the boat  Input: qtun  ROOT FAITH: MAX-C ROOT  DEP[NUC]  "3= a. qatun b. <q>tun  *!  c. q<t>un  *!  Notice that is the speaker posits /qatun/ as the Input, then all three constraints are satisfied in the optimal candidate: qatun. 142  By transitivity then, we have established the following partial constraint rankings:  (63)  *COMPLEX ONSET  »  AFFIX FAITH (MAX)  •COMPLEX ONSET  »  DEP[NUC]  •COMPLEX ONSET  »  NO CODA  MAX-C ROOT  »  NO CODA  MAX-C ROOT  »  DEP[NUC]  The data in (64) establishes the relative ranking of DEP[NUC] and NO CODA Consider the status of the following CVC Roots which can appear as unaffixed stems. (64) CVC Roots Input  Output  Gloss  a. tin  tin  ten  barbecued fish  • piq  piq  peq  wide  £uq  w  Xuq  w  Xoq  d. puk  w  puk  w  piik  c.  hard  w  book  w  e. tan  tan  tan ~ tXn  mother  f. man  man  man ~ mAn  father  Notice that Root Faithfulness and DEPfNUC] must dominate NO CODA since it is better to have a faithful parse of the coda consonant than to allow epenthesis, as shown in (65). (65)  tin [ten] barbecued fish (salmon)  t'in  DEP[NUC]  ROOT FAITH  a. t'in  *!  b. t'i.n[9] c. t'i<n>  *!  143  NO CODA  (66) Partial Constraint Ranking by Transitivity •COMPLEX ONSET  »  AFFIX FAITH (MAX)  •COMPLEX ONSET  »  DEP[NUC]  »  NO CODA  ROOT FAITH: MAX-C ROOT  »  DEP[NUC]  »  NO CODA  The next section provides additional evidence for the lack of Complex Onsets in Sliammon. Loan Words: Evidence for lack of Complex Onsets The phonology of words borrowed into Sliammon from English also provides evidence that Complex Onsets are generally banned in the language. Compare the English words in (67.a-b) which contain complex onsets with the corresponding Sliammon loan words in (68.a-b).  (67) English Source a. plXmz  plums  b. brdwk  broke  (68) Sliammon loan words Input  Output  Gloss  a. pirns  palmas  pslmss (~ pabmas)  plum, plums  b. plok -it c  palok i(t) c  palok ic  I'm broke (no money)  w  w  w  In the Sliammon loans, the initial consonant cluster is avoided by the presence of [a], as shown by the output forms in (68.a-b). The goal of the preceding section has been to establish the lack of Complex Onsets in the language. The next section addresses some apparent counter examples to this claim.  144 Apparent Exceptions to ""Complex Onset Initial sC sequences The following examples which were uttered in isolation begin with word-initial consonant clusters and therefore appear to be counter examples to the generalization that Sliammon lacks Complex Onsets. All of these examples in (69-70) involve an initial s. (69) Input  Output  Gloss  a. s q ajim  sq Xjim  he's so poor, in poor health  a", s IMP-q ajim+?  sq aq ajim  worse than poor, sickly  b. s k ici  sk i-ci  bothersome, a nuisance  c. s naq  snXq ~ s nXq  dear, loved one  a. s ca?at  sce?at  now  b. s canui  sce?noi  when?  c. s cams  seems  why?  d. s Jasu4  sjesoi: ~ s jesoi  yesterday  e. s 1 uk -'u4  s( 6k vi  at the end of a day  w  w  w  w  w  w  w  3  (70) time expressions  ?e  w  3  e  w  Although each example involves a sequence of two consonants (sC) from a linear perspective, the claim which is made here is that these two consonants do not form a constituent (i.e. Complex Onset) within the domain of the Prosodic Word (PrWd). Evidence that the initial s is segmentable is provided by the contrast between the data provided in (69-70) versus the data in (71) which lacks the initial s. Consider the contrast between the related words and phrases provided by one elder who consistently omits the initial s when she produces these words in isolation. Compare the data in (69-70) with the examples in (71) which lack the initial s- proclitic.  145  (71) Sliammon data Input  Gloss  Output  a. q ajim  q 9jim ~ q 9j9m  poor, poor in health  b. k ici  k i?ci  bothersome  c. cam ga  cim gA  w  w  w  w  w  ~  9  why?  cim ga 9  what's the matter?  c\ cam c x ga  cim c x  d. jasut  JESO"!  yesterday(cf. sjesoi)  e. t' uk  t' dk  today (cf. sPok")  w  6  w  e  ga  w9  w  breaking daylight  e\ MP-t^u^-INC  Since this s proclitic is a sentence-level constituent and is not part of the word, it is omitted in (71.a-e) when the word is pronounced in isolation. When the word occurs in a sentential context, the initial s reappears, as in (72.a'-a") and (72.d). The form in (72.c) shows that the non-reduced form of this phrase involves ?9s and that the s is syllabified as the coda of the first syllable; compare this with (71.c) above.  (72) Sliammon data Input  Gloss  Output  a. q ajim  poor  q 9jim ~ q 9j9m  w  w  w  hehew sq 9jim  a", hihw c ?9 k s q ajim  he':hewc (?9) k s  b. k ici  k i?ci  c. ?9s 6a?at  ?9s6e?At  d. hihw k s ca?amrf  he:w k s ce'?anot  when was it?  e. cam ga  cim  ~ cim ga  why?  ga  what's the matter?  w  w  w  w  w  ti?t9  tumis  He's a really poor man  a', hi hw s q ajim tayta tumis  q 9jim w  w  bothersome  w  w  e'. cam cx ga w  ?9S . 4e? . ?At  h  w  9  I'm poor (really tired)  gA  elm cx  w 9  9  146  now  f. jasuf  jesoi  yesterday  f. s jasul  s.jes.so'f  yesterday  g. t' uk  PoP  g'. M P - t ^ u ^ - T N C  t' dt' ok uk  breaking daylight  h. mamaia  mamaia  white person  h'. sq s mamata-s  sXq  s/he's a half breed  e  w  e  e  today w  w  smamA'iAS  In (72.f-f) this speaker produces both jesot ~ s.jeso'f for yesterday but clearly considers the initial s in the second variant outside of the domain of the first syllable, as indicated by the judgements regarding syllabification. A single consonant such as s which is not syllabified as part of the initial Onset, is proposed to be licensed moraically, as shown in (73) below (cf. Bagemihl 1991 on moraic licensing in Bella Coola). (73) a.  s^ fja s • soi:u) u  u  It should be noted that the elder who systematically omits the initial s in (71) above did produce two words with the initial sC. (74) Input a.  s q ay w  b. s k^-'u'l  Output  Gloss  s q ay  sq ay ~ sq Xy  sk i?ju4  sk i?Jo4 ~ s k i ? M  w  w  w  w  w  w  telephone this morning  Unfortunately, these two examples have not been systematically tested with other consultants nor have they been tested in different syntactic environments in order to detennine the nature of the initial s. Recall that there are a number of sources for initial s: nominalizing prefix s- which has generally been lost, the proclitic s found in subordinate clauses, and a reduced form of ?9s/?as.  147  The explanation for the existence of s C initial words involves a complex set of factors, such as those in (75). (75) • there are a number of s's which precede the stem in linear string • Markedness: coronal s often found as an exception cross-linguistically • Existence of word-internal sC onsets to satisfy metrical constraints • Contact with other Salish-speaking peoples where s-nominalizer is preserved • Increasing use of English which has many sC-initial words  What is clear is that Sliammon lacks word-initial C C onsets which do occur in many other Salish languages.  Ranking of * COMPLEX ONSET relatively high, but as with all constraints it can be  violated i f constrained to do so by some higher-ranking conflicting constraint. A case in point w i l l be discussed in § Clitic Initial Constructions The following clitic-initial constructions in the language also appear to be exceptions to the proposal that Sliammon lacks Complex Onsets. These constructions are particularly interesting since the first position in the phrase is typically occupied by the predicate or verbal auxiliary. A s shown by the data in (76-78), the subject clitics are the first element in the linear string. In fact, the subject clitic is followed by another second position clitic ( k u , k i ? , k a?) which indicates that the w  w  w  entire clitic group occupies this first position. These constructions are of interest from a phonological perspective since we need to determine how these phrase-initial consonants are licensed? The output forms in (76.a-g) were recorded with a very brief excrescent schwa following the clitic which seems to suggest that these clitics  are not  syllabified with  the  following  syllable,  148  and  therefore  do  not  violate  *COMPLEX ONSET. The data in (76) involves the first person subject clitic c and the clitic k u w  whereas (77) involves combination of c with k i?. w  (76)  First Person (sg) Subject Clitic: c I (usually an enclitic)  Output  Input  Gloss  a. c k u k an=nac-it  c k u k a? nascit  I'm sitting down  b. c k u ?i"ftan-it  c k u ?ettanit  I've eaten  c. c k u Xum  c k u torn  I've had enough; I'm almost there  d. c k u qaqam  c  e. c k u Xsct-am  c k u XCctam  f. c k u carhcam-am  c k u cf mce? m9m  g. c k u q 9l' gc=iq an-m  c  h. c k u cag-ux -an  c k u &egux m  I've already helped him  i. c k u q 9l' t'ap=us  c k u q Xf t'aspos  I'm getting (going) blind  j.  c k u q Xf tuk a? na  I'm getting (going) deaf  k. c k u Xax-INC  c k u Xaxax  Tm getting old  1. c k u 6ag-ux -an  c k u ce-gux in  w  w  w  w  w  w  w  w  w  c k u q 9l' t'uk =ana w  w  w  w  w  w  w  3  3  I'm hungry  qaqAm  ku  3  a  w  3  w  w  w  3  w  w  w  3  w  w  3  w  I'm sleepy I'm cold; I've gotten cold  e  w  k u q Xl' gA?&eq An9m I've gone bald w  w  w  w  w  w  w  w  w  w  a  w  w  w  I've helped him already  (77) First Person (sg) Subject Clitic: c J  Output  Input  Gloss  a. c k y cag-ux -an  c k i? ce-gux tn  I've helped him (just now)  b. c k y 6ag-t-'uf  c k i? cegatoi:  I've already helped him  b'. c k y cag-t-an-'ui  c k i? cigatanoi  I've already helped him  c. c k y k an=iws-it  c k i? k anewsit  I already rested  d. c k y ?rttan-it-'ui  c k i? ?ettan?lto1:  I've already eaten  c k i? hoy  Tm finished  w  w  w  w  w  w  w  e.  c k y huj w  w  w  w  w  w  w  h  w  w  149  (78) First Person (sg & pi) Forms: c J ; st we Input  Output  Gloss  a. c k a? ?9J-INC (stv.?)  c k a? ?a? Ji  I'm all better now  b. c k a? ta(?)gam-it  c k a? ta?gami^  I announced it  c. cx k a? ?9J-INC  cx k a? ?a? Ji  you're all better  d. st k a? ?3j-INC  st k a? ?a? Ji  we're all better  e. st k a? tag-am-it-'uf  stk a? tagamitoi  We announced it  w  w  w  w  w  w  3  w  w  w  w  w  3  3  w Discussion and Proposed Analysis It is proposed here that the sentence-initial clitic c is licensed moraically, as evidenced by the excrescent schwa which accompanies the release of this consonant. In particular, c is not the first member of a complex onset, and therefore does not constitute a violation of the constraint •COMPLEX ONSET.  As will be argued in §, Sliammon has obstruent-only syllables of the  form CCp. The other subject clitics cx and st are proposed to form minor syllables. Since these w  clitics are unstressed, and schwa epenthesis is proposed to occur within the domain of the stem in order to satisfy PROPER HEADEDNESS at the level of the Foot, these CC syllables are licensed without containing a vocalic nucleus. Word-Internal Complex Onsets The Problem In the previous sections, it is argued that Sliammon generally lacks word-initial and wordinternal Complex Onsets. The next section presents a systematic set of cases in which *COMPLEX ONSET  is violated, as shown by the data in (79).  A word-internal st- cluster is parsed as a  Complex Onset rather than spanning two different syllables, as shown by a comparison of the grammatical and ungrammatical examples in (79).  150  (79) Word-internal sC- Onsets and the Causative Suffix /-stg/ Output  Gloss  a. pt-st-agt  JLA" . stawi  have a race (with e.o.)  a'.  *]its . tawi  b.  k 9n-stu-mi c w  k i5n . sto . nuc  I'll show it to you  w  *k tins . to . mic  b'.  w  c. k 9n-st-anaq  Pvn . sta . nAq .  c'.  *k t5ns . ta . nAq .  w  person who shows off  w  d. k i-k 9n-st-ana-mut  k i . k 9n ..sta:. na? . mot  d'.  * k i . k 9ns . ta: . na?-. mot  e. IMP-tiwsam-st-anaq  ti. tiw . sem . stAn . nAq  e\  *ti. tiw . sems . tAn . nAq  w  f.  w  w  w  w  MP-tiwsam-stu-mi c  he's really showing off  w  ti. tiw . sem . sto . mic  teaching  I'm teaching you  *ti. tiw . sems . to . mic  f.  Compare the data in (71) with the data in (79) (both data sets from a single speaker), which show that there is an asymmetry between word-initial syllables versus word-internal syllables. The problem then is how do we account for the contexts in which Complex Onsets are banned, and the contexts in which Complex Onsets are permitted? All of the examples in (79) involve the Causative Marker /stg-/. The surface st- Onsets occur in both stressed and unstressed syllables so that an explanation for their distribution cannot be attributed to whether or not they occur in a strong/weak metrical position. Proposed Analysis These examples receive a principled explanation within an Optimality theoretic grammar given constraint ranking, conflict and minimal violation.  The constraint which bans complex  onsets (*COMPLEX ONSET) is violated in order to satisfy some higher-ranked constraint(s) within  151  the grammar. Consider the following analysis which is proposed here in order to account for the data in (79). Some morphemes prefer to be aligned with the edge of a relevant prosodic category rather than being parsed into different prosodic constituents (cf.  McCarthy and Prince (1993b) on  Generalized Alignment). I propose the following language-specific  instantiation of Alignment  which ensures that the left-edge of the Causative morpheme is aligned with the left-edge of a syllable, as in (80).  (80)  ALIGN L [CAUS, o]  Align the left edge of the Causative morpheme with the left edge of a syllable.  Consider the following representations which illustrate how satisfaction of this constraint works. The representation in (81.a) violates ALIGN L since the causative morpheme is parsed into two different syllables, whereas the representation in (8 Lb) satisfies the constraint since the causative morpheme is aligned with the left-edge of a syllable. The example in (81 .a'-b') illustrates how this applies to the form: jaX-st-awi have a race with each other, repeated here from (79.a). (81) Violates ALIGN L  Satisfies ALIGN L  a. * stem-s] [t...  b. stem] [-st...  a] [a a'. *jiXs . tawl  o] [a b'. jtX" . stawl  The optimal output satisfies ALIGN L at the cost of violating *COMPLEX ONSET, as shown by the tableau in (82).  152  (82)  k 9n-st-anaq w  [k i3n . sta . r i A q ] person who shows off w  k 9n-st-anaq w  •COMPLEX ONSET  ALIGN L [CAUS, a]  «s" a. k 9 n . sta . riAq w  *!  b. k 9 n s . ta . n A q w  In addition, if all coda consonants are moraic as argued in §3.1, then syllabifying VCCCV as VCC . CV will create a structure a preceding trimoraic syllable whereas syllabifying the same string as VC . CCV. does not. Notice that the output candidate which violates *COMPLEX ONSET also creates more optimal Foot structure. Consider the following metrical structures in (83) which illustrate this point (cf. §3.3 on Metrical Structure in the language).  (83)  «SP  a. jX-st-agi  [TiXstawi ~ JiXstAwf]  have a race; race e.o.  a',  (J9X . S t A W - I )  •COMPLEX ONSET  a".  *(j9luS  U  u  u  *(H*lW)Ft  • tAW t )  u  u  (  *Cj^uSu)  tAW  u  •CLASH  n^)  b. k n-stu-mi c  [k iinstomic]  b'.  (k 9n  b".  • ( k 9 n s . t o . nuc )  *(P-upu)Ft  b'".  • ( k 9 n s ) ( t o . mic )  •CLASH  [k^nstAnaq]  s.b. who shows off  w  I'll show it to you  w  w  u  •COMPLEX ONSET  . sto . micu) u  w  u  u  u  u  w  u  c. k n-st-anaq w  u  u  u  -  (k 9n  ".  • ( k 9 n s . ta . n 9 q )  *(upup)Ft  c .  •(k 9n s )(ta .  *CLASH  c  C  w  u  . sta . n9q ) u  *COMPLEX ONSET  u  w  u  u  u  w  u  u  u  153  u  n9q ) u  d . IMP-tiws-am-stu-mi c  [titiwsemstomLc]  d'.  (ti . ti w )(s£ mn)(st6 . mic )  d".  *(ti^)(ti^w^)(sem^s )(t6^)(mLc )  *(pup)a  d'".  *(ti . ti^w^)(s9m^s )(to^  *v-reduction  u  u  u  I'm teaching you  u  u  ^  ^  u  Ji  *COMPLEX ONSET  M  . mic^)  As shown by the syllabification and Foot structure presented in (83.a"), if the s were syllabified with the preceding syllable, this would also have the effect of increasing the moraic count of the entire word. In fact (83.a") is ill-formed because the Foot contains four moras (a non-minimal violation of FTBIN). The constraint ranking must also rule out candidate (83.a"') in which there are two adjacent Feet. By comparing the output candidate in (83.a') with (83.a'"), it seems that creating a bi-syllabic tri-moraic Foot is more highly valued than a sequence of two bi-moraic mono-syllabic feet - in other words, the amount of phonological material which occurs within the Foot domain is maximized, and this results in the violation of the lower-ranked *COMPLEX ONSET constraint. In addition, adjacent stressed syllables (*CLASH) are avoided . 4  This section presents evidence that *COMPLEX ONSET is violated in order to ensure that the causative morpheme is aligned with the edge of a syllable.  By observing the ungrammatical  examples in (83.a"'-d"'), failure to properly Align this suffix would also create structures in which the causative morpheme straddles not only two syllables but also two Feet. CRISP ALIGNMENT is therefore satisfied at the expense of a *COMPLEX ONSET violation.  4  The constraint *CLASH is not an undominated constraint in Sliammon. It can be violated just in case the word is  comprised of a Root and following bisyllabic LS, such as q up=i6xan [q 6:pe6xAn] hair under arms; tih=u0in w  w  [ti:h60en] big mouth ; sil=awtx [se:lawtx ~ se:lawtw] tent. In these cases, it is more important to stress the w  w  adjacent lexical heads than it is to satisfy the constraint which disprefers adjacent feet (*CLASH).  154 Establishing Maximal syllable size: ""Complex Coda It appears to be relatively easy to establish the simple syllable types in the language; however, it becomes more difficult to establish the upper limits on the size and shape of syllables in Sliammon. The Data Given the Nuclear Moraic Model of Shaw (1993 et seq.), i f schwa is nuclear and nonmoraic in contrast to full vowels which are Nuclear and moraic, then we predict an asymmetry between number of coda consonants licensed by schwa versus the number of coda consonants licensed by a full vowel. Further, i f feet are optimally bimoraic in the language (i.e. they satisfy FTBINp), then we expect to find (CsCC) and (CAC) Feet. This is in fact the case, as shown by the  data in (84-85) which involve mono-morphemic words in the language.  (84) CaCC  Input  Output  Gloss to burst; gun shot  a. t'k s  t9k s  t'i5k s  b. X6t  Xa&t  Xa6t  c. Xpx  Xspx  w  w  w  d. m6k e.  pit  f.  xfJc  w  w  m96k  w  pait  xat'k  w  g. i n t h. qmk i.  w  cpx  break  w  ma0k  w  xat'k  w  qamk  6apx  Xapx  blackcap berry  paft  "font w  sleep  thick  design, carved  w  fant w  qamk &apx  155  to w  weave capsize, tip over dirty  (85) CAC Syllabification  Output  Gloss  a. t'in  t'in  t'en  barbecued salmon  b. piq  piq  peq  wide  c. cuj  cuy  ciiy  child  Input  d. t' uk  t' uk  t' dk  day, light, bright  e. man  man  man ~ mXn  father  f.  tan  tan ~ tXn  mother  q ay  q ay  talk, speak  e  w  e  tan  g- q ay w  w  w  e  w  w  The data in (84-85) above satisfy the constraints SYLL NUC and SYLL MORA, as well as the constraint *[upu] which assigns a cost associated with super-heavy syllables. a  (86) SYLL NUC  Syllables have vocalic nuclei  (P&S 1993)  SYLL MORA  Syllables have phonological weight  (Shaw 1995, 1996)  *[ppp]  Syllables are not trimoraic  a  CaCC syllables do incur a *COMPLEX CODA violation, but do so in order to satisfy the highranking Root Faithfulness constraints. Now consider the following data which contain trimoraic CACC syllables. Notice that these examples involve mono-morphemic Roots in the language, as opposed to trimoraic syllables which may arise as a result of morphological concatenation. As will be shown in §, trimoraic syllables are avoided when they arise across a morpheme boundary whereas they are licit in the mono-morphemic forms in (87).  156  (87) CACC Output  Gloss  a. k ifa  k elt  upstream area  b. piwi  pewi  rendered fat, lard  c. kiks  kiks  cookie  Input  w  w  h  d. x uk t  x iik t  e. k umt  k umt  f. cap9  Sep©  w  w  w  w  w  w  kelp  h  aunt, uncle  g. ?asx  ?asx  h. ?aq t  ?aq t  w  nothing, none  h  seal  w  w  w  downstream area  h  The examples in (87) entail that ROOT FAITH outranks the constraint which bans trimoraic syllables.  (88) ROOT FAITH »  *upu]  0  (89) k umt [k umf] kelp w  w  Input: k umt w  ROOT FAITH  ^^^^^^^^^^^^^^^  «• a. k uhit w  b. k urii<t>  *!  c. k u<ni>t  *!  w  w  *[upuja  The optimal candidate in (89.a) violates the constraint against trimoraic syllables (*[upp] ) in order a  to satisfy the higher-ranking Root Faithfulness constraints which ensures that there is a correspondence relation between the Input and the Output.  157  Notice also that although schwa epenthesis takes place for purposes of stress assignment (cf. §4), schwa epenthesis does not occur in order to break up the final consonant cluster, as in (90). (90) Input  Gloss  Output  a. k ifa  k elt  *k iX[9]t  upstream area  b. piwt  pewi  *piw[3]'T  rendered fat, lard  c. kiks  kiks  *kik[9]s  cookie  w  w  w  d. x uk t  x uk t  *x uk [9]t  nothing, none  e. k umt  k umt  *k u?m[9]t  kelp  cep0  *cep[9]0  aunt, uncle  g- ?asx  ?asx  *?as[9]x  seal  h. ?aq t  ?aq t  *?aq [9]t  downstream area  w  w  w  f.  cap6 w  w  w  w  w  w  w  w  w  w  w  w Proposed Analysis Schwa epenthesis into a Root not only incurs a DEPfNUC] violation but also violates OCONTIGUITY of the Root which ensures that the contiguity relations in the Output are in correspondence with the contiguity relations in the Input (cf. McCarthy and Prince (1995:371) and Lamontagne (1996)on Contiguity). This effectively assigns a cost associated with insertion into the Root which interrupts the contiguity of the string. If O-CONTIG ROOT dominates DEP[NUC], then schwa epenthesis will be prevented within the Root unless constrainted to do so by some higher-ranking constraint (cf. § where ROOT CONTIGUITY is violated in order to satisfy PROPER HEADEDNESS). Consider the tableau in (91) which shows the effects of this partial ranking.  158  (91) cap0 [cep6] parent's sibling; aunt, uncle  5ap0  DEP[NUC]  0-CONTIGROOT  " 3 a. cap6 s  b. cap[a]e  *!  Contiguity must also dominate the constraint which bans trimoraic syllables, as shown by a comparison of the candidates in (92). (92) cap6 [cep6] parent's sibling; aunt, uncle Input: cap6  ROOT FAITH  «• a. ca p e u  u  O-CONTIGRT  *!  u  c. ca^p^e^  *!  d. ca <p >e  *!  u  u  DEP[NUC]  u  b. ca p[9]6 u  *[lW]a  u  Notice that this is different from what happens across a Root/Affix boundary, as shown by the data in the following section. Diminutive Reduplication and *nup]cr The next section shows that morphologically-triggered vowel deletion associated with Diminutive reduplication is blocked in cases where this would otherwise create super-heavy CACC syllables. Diminutive Reduplication In CV- diminutive reduplication, the Root vowel is deleted as shown by the data in (93), and discussed in Davis (1970), Kroeber (1989), Blake (1992), Watanabe (1994, 2000).  159  (93) Diminutive Reduplication Output  Gloss  x il'9m  x e? lAm  rope  a'. DIM-x il'[-i-]m+[?]  x i-x^im  ^ex^iem  string, thread  b. x us-m  x us9m  x dsom ~ x 6sAm  soapberry  b\ DM-x us[-i-]m+[?]  x u-x sim  x dx sim  small soapberry  c. t um=aju  t'uma?ju  t^dmaVju  barnacle  c\ DM-t um=aju=u'i+[?]  t^u-t^majuVui  ^dt^majuVol  small barnacle  d. q asm  q as9m  q as9m  flower  d'. DM-q asm+[?]  q a-q s9ih  q aq s9ih  little flower  e. canu  canu  ce?no  dog  e'. DM-canu+[?]  ca-cnu?  cecno?  little dog  f.  yaxay  yXxAy  clam basket  ya-yxay  yey XAy  small basket  Input a. x il'm w  w  w  w  w  e  ,6  w  w  yaxay  f. DM-yaxay+[?]  w  w  w  w  e  w  w  w  9  w  w  w  w  w  w  w  w  9  Vowel deletion associated with Diminutive Reduplication is blocked if it creates a sequence of three consonants following the full vowel (i.e. a trimoraic syllable), as shown by the contrast between the data in (94.a'-d') and the ungrammatical examples in (94.a"-d"). (94) Input  Output  Gloss  a.  wal0  wal6  wale  bullfrog  a'.  DIM-wal[-i-]e+[?]  wa-wa? . lie  wawa?ke  baby bullfrog  5  a".  *wa-w? . lie  *ppp]a  a'".  *wa-w . fie  *R70nset  B.Wilson (p.c.) has also documented w a w l a 0 for the diminutive of bullfrog.  160  b. cays  cayas  ceyis ~ ceyis  arm, hand  b\ DHM-cay[-i-]s+[?]  ca-ca?. yis  ceceyis  small hand  b".  *ca-c?. yis  b'"  *ca-c. yis  *R70nset  c. k urht  k urht  k urht  kelp  c'. DIM-k umt=u'i+[?]  k u-k um. tut  k uk umtiji  small kelp  c".  *k u-k m. tut  w  w  w  w  w  w  w  d. qayk  w  *ppp]a bald eagle  qa'yk  wh  w  d'. DlM-qayk =u<r+[?]  qa-qay. k -ut  d".  *qa-qy . k ui  w  w  w  qayk  w  h  w  qaqayk M w  young eagle *uu|i]a  w  Candidates (94.a"'-b"') are ruled out by the high-ranking constraint which bans glottalized resonants in Onset position. Candidates (94.c"-d") also involve sonority reversals within the coda: (94.c") *k uk m . tui and (94.d") *qagy . k ui. w  w  w  The following mono-syllabic Roots also retain the Root vowel, as shown by the data in (95). These CVC stems take the Ci- diminutive prefix, a fact which is also noted by Watanabe (2000). The ungrammatical forms in (95.a"-c") show that deletion of the Root vowel would create hetero-morphemic trimoraic syllables of the from CA-CC. (95) Input  puk  a. puk  w  w  a'. DIM-puk  w  Output  Gloss  piik  book  w  pi-puk  w  pepuk  w  small book  a".  *pi-pk  b. Xap / Xp  Xap  Xap  deep  b\ DHM-Xap  Xi-Xap  XiXap  little bit deep  b".  *Xi-Xp  *p\ip]o  w  *ppp]a 161  c. k as  k as  w  w  k i-k as  c".  *k i-k s  w  hot  k e'k AS  little bit hot  w  c\ DM-k as w  k as~k AS w  w  w  w  w  *JW]o  w  As observed in this section there is a general constraint against super-heavy syllables in Sliammon as shown by the Diminutive examples in (94-95). Mono-morphemic Roots of the shape CVCC violate the constraint *upu] in order to satisfy the constraints on Root Faithfulness. As a  argued in (91-92) above, mono-morphemic Roots do not take schwa epenthesis either, due to the relatively high-ranking constraint on Root O-Contiguity. Having discussed both simple syllable structure in Sliammon, and the constraints on maximal syllable size, consider the following residual issue. Asymmetry between beginning of words and ends of words One of the properties which characterizes Sliammon words is that the beginnings of words are restricted to a single consonant before the stressed vowel, as in (96). (96) Input  Output  Gloss  a. k isk is  k isk is  k isk is  Steller'sjay  b. k ut'a  k ut'a  k ut'a  barbecue stick  c. k as  k as  k as  w  w  w  w  w  w  w  e  w  w  w  d. mq t' w  w  hot  w  m[3]q t' w  e  mAcf? ~ ma?q t' w  e  onion, wild onion  Contrast this with what happens at the ends of words. As can be observed from the data in (97), words in Sliammon often end in long string of consonants.  162  (97) Input  a.  Pamq i  Famqn,  w  a'.  Output  Gloss  Pamq 1:  c7oud  w  *t'amq 9t e  b. saitx*  saitx  w  sa1tx ~ saitw  w  woman  w  V.  *satex  w  c. qtx  qs?tx  w  q/?tx ~ qa?tx  w  w  c'.  to burn (fire)  w  *qa?tsx  w  d. IMP-qtx  w  qaqtx  w  d'.  qaqtx  burning  w  *qaqtsx  w  e. pu?px  w  e'.  pu?px  w  pii?px  kindling  w  *pu?psx  w  How do we explain this observed asymmetry? Why is there only ever a single consonant at the beginning of words in Sliammon but the language tolerates many consonants word-finally? In particular, examples like those in (97.a-e) clearly exceed the constraints on maximal syllables established in §3.2.2 above. In Blake (1992), these extra consonants are treated as extrametrical; however, it was necessary to admit more than a single consonant at the right-hand edge of the word.  Since  extrametricality is typically limited to a single consonant or prOsodic unit at the edge of a domain, treating these consonants as "extrametrical" seems somewhat questionable.  163 Minor Syllables in Sliammon Syllabification of these examples in (98) Column 2 shows that the final CC is systematically treated as a separate constituent. The syllable boundary is marked with a period, and reflects the judgements of Sliammon speakers.  (98) Input  Syllabification  Output  Gloss  t^am . q "l  t*amq i  cloud  b. sa<ftx  sat. tx  sattx ~ sattw  woman  c. qtx  qa? . tx  a.  f amq L e  w  w  w  w  w  qaq . tx  e. pu?px  pu? . px  w  w  w  w  d. IMP-qtx  w  w  w  qA?tx ~qa?tx w  qaqtx  to burn (fire) burning  w  pii?px  w  w  kindling  Shaw (1993, 1995, 1996) discusses the role of minor syllables in Berber, Mon-Khmer, and Salish languages, with examples from Lillooet (St'at'imcets) and Bella Coola (Nuxalk). A minor syllable is comprised of an Onset consonant followed by a moraic coda consonant. Within Shaw's Nuclear Moraic Model which I adopt here, a minor syllable is mono-moraic, but non-nuclear. (99) a l\ I VL I  |  c c Blake (1995, 1999) claims that Sliammon has minor syllables, as shown by the data in (98). Shaw (1996) also refers to minor syllables as "headless" syllables since they do not contain a nucleus. In her discussion of headless syllables and their interaction with stress, Shaw (1996: 4)  164  states that a headless syllable has the following properties: (a) it cannot meet the requirements of either SYLL NUC or Proper Headedness at the level of the syllable, (b) it is constrained to metrically weak positions, and (c) it is mono-moraic (i.e. metrically light). Notice that there is a kind of complementarity between the locus of stressed schwa and the occurrence of minor syllables in Sliammon. Sliammon has minor syllables - syllables which violate the constraint SYLL NUC. This means that schwa epenthesis is not driven by the constraint SYLL NUC otherwise we would expect all syllables in Sliammon to have vocalic nuclei.  It is  claimed here that epenthetic schwa in Sliammon is inserted in order to satisfy Proper Headedness at the level of the Foot.  Shaw (1996.c) proposes that Proper Headedness is a family of three  independent and rankable constraints, as defined in (100).  (100) Proper Headedness of Shaw (1996c: 10) (cf. Ito and Mester 1992; Ola 1995)  a. PROPHEAD pw  A Prosodic Word is headed by a Foot  b. PROPHEAD FT  A Foot is headed by a Syllable  c. PROPHEAD a  A Syllable is headed by a NUC  [=SYLLNUC]  As can be seen from the data in (98), minor syllables in Sliammon do not occur in a stressed syllable (i.e. as the head of a metrical foot). Furthermore, schwa epenthesis occurs between Ci and Cj of the Root, as shown in (101).  (101) til hi  [hi] rain  Input: ti  PROPHEAD  DEP[NUC]  «• a. t[6]i b. bi  *!  165  (102) tt [tit] rain  Input: tt «•  PROPHEAD  O-CONTIG ROOT  a. v5[9]4  b. tt  *!  Evidence from § above shows that O-CONTIG ROOT »  DEP[NUC], therefore by transitivity  the partial ranking in (103) is established.  (103)  PROPHEAD »  O-CONTIG ROOT »  DEP [NUC]  Contrast this with an example like sattx [saitx™] w  (104)  . Input: saitx  «sf a. sat. tx  . PROPHEAD  w  woman.  O-CONTIG ROOT  nrrrNun  w  b. sat.t[9]x  *!  w  Notice that the optimal candidate in (104.a) violates SYLL NUC / PROPHEADo since the final syllable lacks a vocalic Nucleus but that it does so in order to satisfy O-CONTIG ROOT.  In addition to minor syllables which occur within the domain of the Root, the subject clitics also surface as CC minor syllables, as shown by the data in (105-106). Judgements regarding syllabification are given by the speakers.  166  (105) word-final minor syllables Gloss  Output  Input a. qp-t-'uf a cx b. xpj-'u't a cx  w  w  c. qmi -'ui a cx w  qgp . tu . ia . cx  w  .  q/pto i£ecx <  w  Did you touch it?  xap . ju . "fa . cx  w  .  XApjvriaecx  w  Did you turn back?  qam . k u . "la . cx  v  w  w  qXmk u4a3cx  .  w  Did you tip over?  v  (106) Input  Foot/Output  *9 epenthesis  Gloss  *(k 9 . ?e) (sit. S9t)  we're standing up  *(hiw . qe) (tot. S9t)  we answered it  a. k^is-it st  (k 9 . ?e) ( s i t . st )  b. tg=qin-t-'u'f st  (tiiw . qe) ( t o t . st )  c. rMP-0iq=nac-'uf st  (6e) (0eq. na) (cut. st ) *(0e) (0eq.na) (cui.Sst)  w  d.  IMP-qat' -ag=mix  e.  qms-t-'ui st  e  h  w  h  w  h  we're digging roots  s t ( q a . q a ) ( t ' a w ) ( m l x . s t ) *(qa.qa)(t' aw)(mlx .s9t) we're gathering pi. 6  e  w  h  ( q g m . s[a]) ( t o t . st? ) 1  e  w  * ( q 9 m . s[a]) ( t o t . S9t)  w e stored it away  Compare the output form in (106) Column 2 with the ungrammatical forms in (106) Column 3.  As can be observed from the foot structure, the minor syllable (stjj,) occurs in  unstressed position. If schwa epenthesis in Sliammon is driven by the need to satisfy Proper Headedness, then it is unnecessary to epenthesize schwa into a metrically weak (non-head) position. Basically, schwa is epenthesized in order to be stressed.  3.2.3 Summary This section provides the basic syllable structure to provide background for discussion regarding syllabification entailed in the remainder of the thesis. 6  This predicate means 'to gather people together from different places', and is related to the word kathaumixw '  a  gathering together of different peoples' which is the name given to the International Choral Festival held in Powell River once every two years.  167  3.3 Introduction to Sliammon Metrical Structure This section presents some of the basic properties regarding metrical structure in Sliammon. Although a complete analysis of Sliammon stress assignment is clearly beyond the scope of this thesis, the a number of the phonological properties of the language which are discussed in some detail (Schwa Epenthesis and Full Vowel Reduction) require some familiarity with the basic stress facts of the language. The most important generalization regarding stress in Sliammon for our present purposes relates to the position of primary stress.  3.3.1 Basic Observation: Primary Stress is Leftmost Primary stress in Sliammon occurs on the first syllable of the stem, as shown by the data in (107). This is true of the vast majority of the data collected in this study, and accords well with the descriptive generalizations of Davis (1970), Kroeber (1989), Blake (1992) and Watanabe (1994, 2000). Primary stress is marked by an acute accent over the vowel in the stressed syllable (v).  Primary stress is left-most, as shown by the data in (107) (107) Input  Output  Gloss oolichan oil  a. Xina  Xi?na  Ae?na ~ Xena  b. pilaq  pilaq  peiL\q ~ peUq*  bracket fungus, mushroom  c. k ut'a  R ut'a  k dta ~ k uta  barbecuing stick  w  h  w  d. R u?ux  k u?ux  e. watla  watla  w  w  w  1  w  w  w  R d?ox ~ k ii?t>x w  w  w  wat la ~ h  wat lA h  palaf  pal'Af  g. pcu  p[a]cu  picu ~ picu  3  9  h. mnat' i  m[9]nat' i  manatee  i.  k[a]nika  kinike ~  e  knika  e  smoked salmon, fish sweetheart skunk  f. pal'at'  9  w  ~ pecu  cedar root basket drum  kinekA  168  coloured person  Primary stress also occurs on reduplicative prefixes, as shown by the data in (108.a-f). This provides evidence that the reduplicative prefixes (Plural, Diminutive, and Imperfective) in Sliammon are within the domain of the stem (cf. Davis (1970), Kroeber (1988, 1989), Blake (1992, in prep.) and Watanabe (1994, 2000) on Reduplication, and Sapir (1915) on Reduplication in Island Comox). See Appendix VII on the identification of the morphological stem domain.  (108) Output  Gloss  max-mixaf  m/xmexAi  black bears  mixai  mexAi  black bear  Input a.  CaCpL-mixai  a'. mixat b. DM-janx [-i-]=u i+[?]  Ja-jn[i]x uf  jejnex ui  small fish  b\ janx  janx  jenx  fish, salmon (generic)  w  w  c. DIM-C9CpL-puk c\ p u k  <  w  w  w  pi-pak -puk w  puk  w  w  w  pe-puk puk w  w  piik  w  w  lots of little books book  w  d. DIM-CaCpL-higus  hi-hsw-higus  hehgwhegTJs  small chiefs  d\ higus  higus  hegTJS ~ hegus  chief  e.  IMP-t^uk^-INC  t' u-t' uk -uk  t' dt' ok uk  breaking daylight  e\  t' uR  t' uk  t' dk  day, light  f.  MP-?ajus m-t-agi  ?a-?ajus9mtawl  ?a?ajus9mtawi  exchanging gifts  f.  ?ajusm-t-agi  ?ajus9mtawi  ?aji)S9mtawt  exchange gifts (e.o.)  e  w  e  e  e  w  w  w  6  e  e  w  w  w  Stressed syllables in Sliammon tend to be markedly higher in pitch than their unstressed counterparts, a fact which is also discussed by Watanabe (2000). The observed pitch patterns are indicated in (109), where [H] indicates a syllable which bears a relatively high pitch, and [L] indicates a syllable which is lower in pitch.  169  (109) Input  Output  tpitch]  Gloss  a. ?ima6  ?ima0  ?emA6  [HL]  grandchild  b. ?usa  ?usa  ?6sa  [HL]  blueberry  c.  qa?qa  qa?qa  qa?qa  [HL]  mat, mattress  d.  t'gm  t'[9]g[9]m  t9g9m  [HL]  sun, moon  e.  qsnaf  q[9]snay  qgs nAy  [HL]  dress, shirt  q[9]?ga  qa?gA  [HL]  cane, walking stick  f. qga  3  Although there does seem to be a strong correlation between high pitch and metrical prominance, J.Davis (1970) observes that pitch and stress can also occur independently o f one another. The reader is also referred to Watanabe (1998, 2000) for interesting findings regarding the interaction between stress and pitch. In his discussion, he defines many o f the complex issues and problems for further research.  3.3.2 Foot Form Trochaic Sliammon has left-headed (trochaic) feet as shown by the data in (HO.a-g).  The  ungrammatical forms in (1 lO.a'-g') confirm that feet are not right-headed (iambic).  (110) Left-dominant, Trochaic Feet Input a.  ciya  ciya  a'. w  k ufa w  b\  c  \  Gloss  ciye-ciya  grandmother, granny  *ciye ~ *ciya  b. k ut'a  c.  Output  k 6t'a~k uta w  w  barbecuing stick  *k ota ~ *k uta w  wikal'i  wikal'i  w  wike?le ~ wika?le *wike?le ~ *wika?le  170  hermit crab  d. mnat !  m[9]nat' i  19  drum  manatee  e  *m9nat'e  d'.  e  e. MP-cag-t-as  ca-cag[a]tas  ie&egatAs  e'.  s/he is helping her  *6eiegatAs  f. ya't-t-anapi-m-'u't  ya i[a]tanapimu'i  ye'tatanapemu't  <  f.  you (pi) got called  *yeiatanapem6'i  g. x up=inas=tn w  x upinastn  x dpenastn  w  w  brooch  *x openAst9n w  g'-  In sentential contexts, mono-syllabic content (lexical) words are also stressed, as shown by the examples in (111). The grammatical markers (particles and clitics), on the other hand, are typically unstressed and phonologically dependent (cf. Appendix VII on affixes and clitics).  ( I l l ) Sentential Contexts Output  Input  Gloss  na?  Xe?na[h]A k 90 na? Do you have oolichan oil?  na?  na?  b. ?wk st Tut ?j  ?y  ?iik st ?ot ?i?  b\ ?J  ?y  c. pq t9 ?aya?  pgq  a. Xina a k  w  0 na?  a', na? w  w  w  h  We're all fine good  pXq t9 ?aye?  the house is white white  h  c\ pq  p9q  p/q  d. hu ga s?  S9?  hd  d\  S9?  se?  S?  belong to  gA  se?  go upstairs! go upwards  Since mono-syllabic lexical items are stressed in sentencial contexts, they are also marked with primary stress in isolation, as in (112).  171  (112) Input a. t'in b. rix <  w  c. pix  ten  barbecued salmon  iix  lose a loved one  w  flood  w  w  e. <hik  wuk fuk  w  £uq  Gloss  pex  w  d. wuk  f.  Output  scoop net  w  to fly  w  Z6q  hard  g. tuq  fdq  clear skies  h. k as  k as  hot  i.  k a6  k a£  dogfish  j.  ?ax  ?ax  falling snow  k. ci  tit  rain  1. pq  pXq  white  m. p6k  p96k  bullhead (fish)  n. t^ms  t' Ams  soaked  o. fk s  t9k S  to burst  w  w  w  w  w  w  w  w  w  w  e  w  w  p. m6k  w  m9'6k  w  blackcap berry  3.3.3 Stressed Schwa Many Salish languages avoid stressing schwa if there is a full vowel in the syllable to its right (cf. Thompson and Thompson (1992), Czaykowska-Higgins and Kinkade (1997:15-16) and the references cited therein, Bianco (1995, 1996), Shaw et. al. (1999), amongst others). Stress in Sliammon is always leftmost even if it means stressing schwa rather than an adjacent full vowel, as shown by the data in (113) (cf. Blake 1992, 1995, and Urbanczyk 1999).  172  (113) Schwa epenthesis  Output  Gloss  a. qya  q[s]?ya  qa?ye  water  »' b • qga  q[9]?ga  qarga~qargA  cane, walking stick  c. mna  m[9]?na  ma?na  child, offspring  a. tk fi  t[9]k li  tuk fe  rabbit  Input  w  Wl  w  3.3.4 Location of Secondary Stresses Now consider longer words which have more than a single degree of stress. The primary stress is located at the left-hand edge of the word, as shown by the data in (114). Secondary stress is marked by the grave accent [v]. The syllable which bears primary stress tends to be higher in pitch, as indicated in Column 4 (H=high, M=mid, L=low pitch).  (114) Primary and Secondary Stresses Input  Output  a. qms-t-'u'r c  Gloss  . sa . xM . c  qXmsat6ic  h  [HLM]  / stored it away  . f?u . k uk .  fWofc ufc  [HLM]  breaking daylight  q9m  b. IMP-t^u^-INC  [pitch]  h  w  w  w  w  c. DM-Xatx+[i]  Xa-Xat[i]x  AaXatEX ~ AaA9tex  [HLM] grasshopper  d. DIM-&?him=ut  Xi-XgVtumu't  MaVfomu'T  [HLML] wolfcub  e. fn-t-'ut a cx  igntulacx™  •IXntoifficw  [HLML]  w  Did you weave it?  Some forms are recorded with adjacent stressed syllables, as shown by the data in (115). (115) a. laplas  laplas  laplas  plank, long board  b. kamputs  kamputs  kasmputs  rubber boots  173  The stative forms in (116.a-e) also have adjacent stresses, and are accompanied by [HH] pitch pattern, as documented by Watanabe (2000).  (116) Stative /-itu / U  Input  Output  [pitch]  Gloss  [HH]  burping, belch  [HH]  I'm burping  a. q ac-it  q acit  q a:£it  a'. q ac-it c  q aci<t>c  q a:£ic  a". IMP-q a& c  q a-q a£ c  q aq acc  b. q iq -?m+[i]  q^q^im  q^q^em  [HH]  s.b. is nailing  c. Xux -it  Xux it  Ao:x et  [HH]  crying  w  w  w  w  w  w  w  w  w  w  w  w  h  w  w  w  w  h  c'. MP-foax -it-'ui st  Xu-Xux itui st  tot.. x e.toi. st  d. tap-it  tap-it  ta:pet  d'. rMP-tap-INC  ta-tap-ap  tatapAp h  e. gan-[i]m  ganim  ga:nem  e'. IMP-gan-[i]m+[?]  ga-gan-irh  gagAnem  w  w  I'm burping  h  we were all crying  w  h  [HH]  tight getting tighter  [HH]  orphan having no parents  3.3.5 LSs and Stress Assignment One area which has not been discussed in any detail elsewhere, but warrents mention here is the role which Lexical Suffixes play in stress assignment in Sliammon. Bi-syllabic Lexical Suffixes tend to be stressed in Sliammon, as shown by the data in (117).  From a  historical/comparative perspective, Lexical Suffixes in Salish languages may originate from independent Roots with the loss of the initial consonant, following Egesdal (1981), Martina (1987), and Kinkade (1998). Blake (2000) argues that LSs in Sliammon behave like bound Roots with  174  respect to the resolution of vowel hiatus in the language. It is proposed here that the root-like status of Lexical Suffixes in Sliammon may also explain the observed stress properties . 1  (117) LSs and stress assignment Output  Input  Gloss  a. t^iyc^agic  Fiyca?gi2  twisted spine (t^iyi twisted, =agic spine)  b. xawsin=agic  xawsina? gLc  spine (xawsin bone)  c. Xpx =agic  XXpx a?gic  break one's back (Xpx break)  d. IMP-q ay=axi0  q aq ayAxe0  talking in one's sleep (=axi© bed)  e. ?usa=aja  ?6saha?je  blueberry leaves (?usa blueberry)  f. qnayu=aya  qXnayohaye  sewing needle case (qsnayu sewing needle)  g- pq=aya  parqaye  stove pipe (paq smoke, =aya container)  h. ngin=aya  niginaye  lunch basket (nagin lunch)  h  3  w  w  w  w  w  w  The LSs =ui young of a species usually bears secondary stress, as shown by the data in (118). (118) Output  Gloss  qayk  qXyk  bald eagle  a'. DM-qayk =u l+[?]  qa-qay^ut  qaqAyk ui  small eagle  b. k urht  k umt  k umt  kelp  b'. DM-k umt-u l+[?]  k u-k umtuf  k uk umti)t  small kelp  c. tagat'  ialgai  ia?g3t' ~ ta?gAt'  herring  c'. DjM-'lagat'=u'l-+[?]  ia-igaivft  iatgatbi  small herring  d. janx  janx  jenx  fish (generic)  Input a. qayk  w  w  w  <  w  <  w  w  d'. DIM-janx [-i-]=u'H[?] w  w  w  w  w  w  w  w  w  w  wh  ja-jnix ui w  jej nex M 3  w  small fish  'Kinkade (1973), Czaykowska-Higgins (1996, 1998), Shaw et al. (1999), Tamburri Watt (1999, 2000) discuss LSs and stress assignment in a number of other Salish languages.  175  e. x ax ni?  x ax ni?  x ax ne?  bullhead  e'. x ax ni?=u'r  x ax ni?ui  x ax ne?6f  small bullhead  f.  tym  Xysm  Xi?9m  cockle  f.  DIM-Xym[-i-]-ui  Xa-Ayimuf  takifembi  small cockle  g-  t^umaju  t^umaVju  t^dma?)!)  barnacle  g'- DIM-t^umaju^ul-  t^u-t^majiirvrr  t^dt^majuVoi  small barnacle  h. sma  S9?ma  sa?ma  blue mussel  h*. DM-sma=ui+[?]  si-sma?uf  sisma?6f  small blue mussel  pux u  pdx o  raven  pi-px^Wui  pepx o?u'i  w  w  w  i.  w  w  w  w  pux u w  w  w  i'. DIM-puxu=u i+[?] <  w  w  w9  W3  w  w  small raven  ,  Other LSs, such as =inas, do not bear secondary stress if it creates a stress clash, as shown by the comparison between the data in (119-120).  (119) =inas chest Gloss  Output  Input  heart beat  a. ntx-[i]m=inas  n9txim=inAS  nXt. xe . me . n A S  b. IMP-jk -m=inas w  j9-jk -[a]m=inAS  jey . k a . me .  b'. IMP-jl  w  j3-jl  Jeyk  rubbing  Output  Gloss  w  w  w  w  .  nAS.  heart burn  (120) =inas chest Input  a. q up=inas  q up=inAS  q d . pe . n A S .  hair on chest  b. xat =inas  xat' =inAS  xa . t°e .  breast bone, sternum  c. Xk =inas  X9k =inAs  tv . k i . nAS  w  e  w  w  e  w  w  w  nAS  heart  Although primary stress is most often leftmost as discussed above, it is a point which requires further investigation; exceptions to the generalization that primary stress is leftmost 176  involves words which contain Lexical Suffixes. I have recorded some examples in which primary stress is right-most rather than left-most, as shown by the data in (121). Notice that these examples have a [MLHL] pitch pattern; primary stress is correlated with the syllable which bears the highest pitch.  (121) Input  Output  [pitch]  Gloss  a. q uw=ana  q uw=a?ana  q 6wa?ana  [MLHL]  ear  b. DIM-q up=ana  q u-q p=a?ana  q 6q pa?ana  [MLHL]  hat sticking out ears  c. x aw=iq =uja  x a?w=iq =u?ja  x a?weq 6?Je  [MLHL]  finger  d. Xax=iq =uja  Xax=iq =u?ja  Xaxeq d?je  [MLHL]  thumb  e. kit'=iq =uja  kit'=iq =u?ja  kyefeq d?jE  [MLHL]  pinky, little finger  w  w  w  w  w  w  w  w  w  w  w  w  w  w  w  w  w  w  w  w  These examples all involve Lexical Suffixes, as indicated by the equals sign (=) which precedes the LS (cf. also Appendix VI). The examples in (121.c-e) contain two adjacent LSs. The surface stress patterns resemble the same stress patterns associated with compounds in the language, as shown by the single example in (122). Compounds are rare in Sliammon, as discussed by Hagege (1976), and Watanabe (2000).  (122) a. pq-at-cayis  paqaicayis  pAqiAiceyis  palm (white ofhand)  a', pq  paq  pXq  white  a", cayis  cayis  ceyis  hand  For our present purposes, I will continue to assume that primary stress is left-most but acknowledge that a complete analysis of Sliammon metrical structure is still wanting, and there are many questions which remain for future research. In particular, although the locus of primary  177  stress is fairly well established, the generalizations regarding secondary stress placement are not well understood.  3.4 Summary This chapter provides discussion and independent motivation for the prosodic structures which are assumed in the remainder of the thesis. As argued in §3.2, the distribution of schwa is often detennined by the requirement that each Foot contains a vocalic nucleus, and is not driven solely by constraints on syllabification, since Sliammon has vowelless syllables in word-final position.  178  Chapter 4 Distribution of Schwa in Sliammon Raven has blue eyes, like the waters of [Harwood Island] on a good day. He also carries a black magic umbrella. This makes me want to sing. Caw caw. Or cry. Phyllis Webb  4.0 Introduction In this chapter, I argue that there are three different schwas in Sliammon, as evidenced by their phonological behaviour: (i) Excrescent (or transitional) schwa, which is written as a small raised schwa [ ] to differentiate it from all other schwas, which are written as [a], (ii) Epenthetic 9  schwa (E-schwa), which displays a ~ 0 alternations, and (iii) Non-alternating schwa (S-schwa) which does not show any surface alternation, and is subsumed under (ii). The terms E-schwa, and S-schwa are descriptive, pre-theoretical terms used here following van Oostendorp (1999) in order to underscore "the various roles which schwa can play in the phonology of a language". Van Oostendorp (1999) also includes the term R-schwa to refer to schwas which shows full vowel/schwa alternations. It will be argued in §3.3 that Full Vowel Reduction in Sliammon is distinct from reduction to schwa. Although the resultant reduced full vowel shares the same prosodic representation as schwa (i.e the reduced full vowel is proposed to be Nuclear and nonmoraic), it retains its distinctive feature specification. The traditional designation "Reduced Vowel" is maintained in order to emphasize the fact that it does not neutralize with schwa.  4.1 Excrescent versus Epenthetic Schwa In terms of accounting for the distribution of schwa in Sliammon, it is necessary to distinguish 'epenthetic' schwa from 'excrescent' schwa, following work by other Salishan scholars, such as Bagemihl (1991), Bessell (1992), Matthewson (1994), and Kinkade (1997). In the section  179  which follows, I will outline the phonological distribution and surface alternations which do provide evidence for the distinction between epenthetic schwa on the one hand, and excrescent schwa on the other hand.  4.1.1 The Problem In Sliammon, some schwas surface consistently in each token of a given word. This holds of different instances of the same word from the same speaker as well as across speakers, as in (1-2).  (1) Same Speaker Input  Schwa [a]  Output: Multiple tokens  Gloss  a. i x  i[9jX  iXx  bad  b. i i  i[9]i  iii  rain  c. mt  m[s]5t  mXX  calm (on water)  d. qya  q[9]?ya  qa?ye  water  e. q asm  q as[s]m  q as9m  flower  p[s]quk [i]t  pXq^k^it  Input  Schwa [s]  Output: Multiple speakers  Gloss  a. pq  p[9]q  pAq  smoke  b. pq  p[a]q  pXq  white  c. sxm  s[9]x[9]m  sXxAm  racing canoe  d. q l'  q [9]f  q Xl'  they came  w  w  f. pq=uk t+[i] w  w  w  11  dressed in white  (2) Different Speakers  w  w  e. m0k f. ngin  w  m[9]0k n[9]gin  w  w  m90k  blackcap berry  w  nigin ~ nigm  180  packed lunch  Contrast this with the data in (3-4). In these examples the schwa [ ] is very brief and more variable 3  in its surface realization. For example, a speaker may pronounce a very brief schwa in one token of a given word whereas in other instances this brief transitional vowel is not recorded, as in (3). (3) Same Speaker Input  Output 1  Output 2  Gloss  a. x il'-m  x e?bm  x e? bm  rope  b. yagay  ye?gAy  ye? gAy  inner cedar bark  c. paL' q umay  pa?a q d?mAy  pa?a q d? mAy  one year (one snow)  d. Games'a?  OXmse?  OXmse?  twenty  w  w  w  w  3  3  w  w  3  3  The output forms in (4) Columns 3-4 show that the presence of this brief schwa [ ] also 3  varies across speakers. (4) Different Speakers Input  Speaker 1  Speaker 2  Gloss  a. tm=us=tn  t3?mustsn  ta?most9n  ta? most9n  headband  b. qsnaj  qasnay  qXsnAy  qXs nAy  shirt, dress  c. Vatnupil  ?atnupil  ?atnopel  Vat^opel  auto, car  d. lkli  lakli  likle  lik^le  key  e. Jaja  Jarja  jE? Je  relative, friend  f. L'-xi-xniq  x[i?]i-xniq  3  3  3  xe?exneq  xe?ex neq 3  Owl's Grove  These variable transitional schwas are very brief in duration, audibly shorter than the instances of schwa which surface consistently. The transitional schwas are inserted either after a glottal, as in (3) or between consonants when the second consonant is a resonant, as in (4).  4.1.2 Evidence from Syllabification There are some schwas which are syllabified as separate syllable peaks, and other schwas which are not. This contrast is shown by the data in the syllabification columns of (5-6). Recall from § and §2.4.4, that schwa lacks inherent features, and as such is subject to colouration 181  from adjacent consonants which accounts for the range of surface outputs in (5) Column 3 and (6) Column 2. In the data in (5), the stressed initial schwa [a] functions as a syllable peak.  The  syllabification of each form is given in Column 4. (5) Schwa functions as a syllable peak Input  Output  Syllabification  Gloss  a. np=sn  napsan  napsm  nap. san  b. tU[-i-]  talik.  t'akk  tal. lik  a hole  c. £t-t-as  cartas  6at tas ~ 6at tAS  cat. tas  he cut it  tatk atas  tatk atAS  tat. k a . tas .  he's pulling it  d. HMP-fk -t-as w  3  w  3  w  tripped  w  Contrast this with the syllable judgements in the next set of data. The data in (6.a-b) shows presence of a brief schwa [ ] i n the environment before a word-initial resonant. In (6.c-e) a brief 3  schwa [ ] occurs either before or after a resonant when it is adjacent to a voiceless obstruent. In 3  (6.f) it occurs between two identical stops. A comparison of the syllabification of forms i n (6) Columns 3-4 shows that these instances of schwa are not considered separate syllable peaks by Sliammon consultants, in contrast to the data in (5) above (cf. §3.2 for constraints on syllabification), [h] is the least-marked epenthetic consonant in the language, and is inserted here to show that the ungrammatical examples in Column 4 are not ruled out independently by the Onset constraint. (6) Brief schwa * Syllable peak Syllabification  •SyllPeak  Gloss  [yapt ~ yapt]  yapt  *[h] . yapt  break it  b. y&-[i]  [yi& ~ zyit]  yi£  *[h] . yit  mil  c. mX-mut  [marmot]  maX. mut  *ma .  very calm  d. MP-xai-mut  [xaxai mot]  xa . xaf . mut  *xa . xa . f . mut  he's really tall  e. qayx  [qay x]  qay. x  *qa. y x  Mink  f. ct-t-as  [cartas]  cat.tas  *ca.t .tas  he cut it  Input  Output  a. yp-t  3  3  3  3  9  . mut 9  9  3  182  It seems worthwhile to note that these brief schwas [ ] in (6.c-e) occur between an 3  obstruent and following resonant or vice versa, and provide a transition between consonants with different major class features. Recall from §2.3.2 that obstruents are unmarked for the feature [sonorant] whereas resonants are lexically specified as [sonoranfj. Contrast this with what happens in (7).  Here a brief schwa [ ] occurs between a front 3  (i.e.[-bk]) vowel [i ~ e ~ e] and a following post-velar consonant. (Note: the symbol [I] represents a backed/retracted variant of IV, and not a voiceless lateralfricative.)In this case, the brief schwa [ ] provides a transition between a vowel and a following consonant produced at opposite places of 3  articulation; for example, between the [-bk] vowel and a uvular stop, as in (7.a-b). When the tongue moves from the anterior portion (front) of the oral cavity towards the uvula, it moves through a neutral position. This brief neutral transition is perceived as a transitional schwa: These are cases of diphthongization and will not be discussed further here.  (7) Input  Output  Gloss  a. RED-q ay+[?]  q i-q ay  q e q ay  talking  b. piqin  piqin  pi qen  shoulder blades  c. cq  caq  ce q  fence  d. calas  calas  CE 1AS  three  w  w  w  w  3  w  3  3  3  So what we have observed in this section is that [a] and [ ] are syllabified differently; [a] 3  functions as a syllable peak whereas [ ] is ignored for purposes of syllabification. The next section 3  characterizes this observed difference in terms of epenthetic versus excrescent schwa.  183  4.1.3 Discussion and Proposed Analysis In his paper on syllable structure in Bella Coola, Bagemihl (1991: 600) discusses the characteristics of excrescent vowels, following work by Levin (1987) . These generalizations are 1  presented in (8) below and are citedfromBagemihl (1991): (8) Excrescent vowels a. The quality of an excrescent vowel is variable, frequently tends towards schwa, and is generally determined by phonetic coarticulation effects. The surface quality of an excrescent vowel does not necessarily correspond to any of the underlying vowel qualities of the language. b. The insertion of an excrescent schwa is triggered by the need for a transition between adjacent articulations, and is not inserted in order to syllabify stray consonants. Phonological rules do not refer to an excrescent schwa.  The central claim to be made in this section is that these brief transitional [ ] schwas in Sliammon 3  show the same cluster of properties attributed to excrescent vowels in (8.a-b) above. Given the present analysis of the phonology and morphology of the language, there is no evidence that these excrescent vowels are referred to by phonological constraints within the grammar. I conclude that these brief transitional schwas are phonologically inactive, and therefore excrescent. I propose a sub-classification of two types of excrescent vowels, the second constituting what are commonly referred to as "echo vowels".  2  'Matthewson (1994), Bianco (1996), and Kinkade (1997) present similar argumentation, citing Bagemihl's (1991) research on syllable structure in Bella Coola. I use the term "echo vowel" to refer to the brief vowel often heard after laryngeal closure (cf. Bessell 1992). This is distinct from the "linking" vowel in Sliammon associated with Control Transitive Allomorphy (cf. J.Davis 1970 et seq., Kroeber 1989, Blake 1999, Watanabe 2000).  184  4.1.4 Further Differentiation: Echo Vowels There is often an echo vowel, represented here as a raised excrescent vowel [ , ], after a e  a  syllable final glottal, as shown by the data in (9) Column 3. It is typically a copy of the preceding vowel.  (9) Echo Vowels Input  Output  Syllabification  •Syllable peak Gloss shallow  a. ?il'-it  ?i?lit  ?e?%t  ?e?. let  *?e . ? . let  b. qawum  qaVwum  qa? wum  qa? . wum  *qa . ? . wum eye  c. tlc+[i]  te?l[i]c  ta? lic  ta?. lie  *ta . ? . lie  d. qwut  q9?wut'  qaVawut  qa? . wuf  *qa . ? . wut' uvula  e. q l'  q 9l?  q 3l?  q 9l  *q al. ?  ?e  w  e  w  a  a  56  w  9  e  a  round  a  9  a  w  w  6  9  come  Echo vowels, like excrescent schwas, are not considered separate syllable peaks by Sliammon consultants, as shown by the contrast between the data in (9) Columns 4-5.  Echo vowels may  represent a broken vowel in which the glottal constriction (laryngealization), written as [?], represents the most prominent glottal pulse during the articulation of a creaky vowel. The glottal constriction may be articulated with the supra-laryngeal tract in the vowel configuration, a configuration which may continue slightly after laryngealization has ceased (or the glottal stop has been released), producing an echo vowel: v? (cf. Bessell (1992:6) for similar argumentation for v  Makapmxcin (Thompson Salish), and Chomsky and Halle (1968: 315-316) for a discussion of glottal constriction). Since excrescent vowels and echo vowels are not referred to by phonological constraints nor do they figure into the prosodic structure of the language (i.e. they do not function as syllable nuclei), they are not discussed further in subsequent chapters. It is worth emphasizing that anyone working on the language for the first time will need to make the distinction between excrescent schwa versus other instances of schwa.  185  (10) []  excrescent schwa  [a]  schwa  3  The next section presents the distribution and analysis of epenthetic schwa (E-schwa) which shows schwa ~ zero alternations.  4.2 Distribution of E-schwa 4.2.1 Some schwas are epenthetic Diminutive CV- reduplication is accompanied by deletion of the Root vowel, as discussed by Kroeber (1989), Blake (1992), Watanabe (1994), and illustrated by the data in (11). The reduplicant is a CV- prefix (a monomoraic open syllable) and in these examples is immediately followed by the Root. The Root vowel does not surface in the diminutive, as shown by (1 l.a'-d') Columns 3 and 4. The use of the angled brackets <> in Column 3 shows which vowel is deleted. (11) Deletion of the Root vowel in Diminutive Red+Base  Input a.  x il'm  a'.  DM-x il'm+[i]+[?]  b.  x us-m  V.  DM-x us-m+[i]+[?]  c.  yaxay  c'.  DM-yaxay+[?]  V-deletion  Output  Gloss  x il'am  x e?lAm  rope  w  w  w  DM-(x ilim) w  w  w  w  DM-(yaxay)  string, thread  x i-x <i>hm w  w  x usam  x dsTjm  soapberry  x u-x <u>sim  x dx sim  small soapberry  yaxay  yAxAy  clam basket  ya-y<a>xay  yeyxAy  small basket  w  DM-(x usiih)  w  w  w  w  w  w  The following diminutive and non-diminutive pairs in (12.1) and (12.2) show that some schwas are clearly epenthetic, an observation also made by Watanabe (2000). When the Root is followed by a consonant-initial suffix (e.g. =nac, or =sn) or is comprised of more than two consonants, then this gives rise to a string of word-internal consonants. An epenthetic schwa [a]  186  appears after the second consonant of the Root, as shown by the data in (12.1.a'-b') and (12.2.a'-c'), in order to provide a more optimal syllabification.  The ungrammatical forms in (12.1a"-b") and  (12.2a"-c") show that these forms do not satisfy the prosodic constraints on syllable and Foot as well as the outputs in (12.1a'-b') and (12.2a'-c') do. (cf. Chapter 3 and Chapter 7 (§7.6)).  (12.1) Diminutive Reduplication Input  Syncope  a. q up=sn w  Gloss  q dpsin  hair on legs  q dq p[9]sin  bit ofhair on legs  w  w  a'. D M - q u p = s [ i ] n + [ ? ]  Output  q u-q <u>psin w  w  w  w  *q dq p . sin w  a.  w  *q dq [9lpsin w  b. Vatnupil b'. DIM-?atnupil  ?a-?<