Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Laryngeal processes in Chipewyan and other Athapaskan languages Gessner, Suzanne C. 1999

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

Item Metadata

Download

Media
831-ubc_1999-0503.pdf [ 16.29MB ]
Metadata
JSON: 831-1.0089167.json
JSON-LD: 831-1.0089167-ld.json
RDF/XML (Pretty): 831-1.0089167-rdf.xml
RDF/JSON: 831-1.0089167-rdf.json
Turtle: 831-1.0089167-turtle.txt
N-Triples: 831-1.0089167-rdf-ntriples.txt
Original Record: 831-1.0089167-source.json
Full Text
831-1.0089167-fulltext.txt
Citation
831-1.0089167.ris

Full Text

L A R Y N G E A L P R O C E S S E S IN C H I P E W Y A N A N D OTHER ATHAPASKAN LANGUAGES by S U Z A N N E C. G E S S N E R B . A . , The University of Regina, 1993 A T H E S I S S U B M I T T E D IN P A R T I A L F U L F I L M E N T O F THE REQUIREMENTS FOR THE D E G R E E OF . M A S T E R OF ARTS in T H E F A C U L T Y OF G R A D U A T E STUDIES (Department of Linguistics) We accept this thesis as conforming to the required standard  T H E UNIVERSITY O F BRITISH C O L U M B I A August 1999 ©Suzanne C . Gessner, 1999  In  presenting  degree freely  at  this  the  thesis  in  partial  fulfilment  University  of  British  Columbia, I agree  available for  copying  of  department  this or  publication  of  reference  thesis by  for  his  this  thesis  and study. scholarly  or for  her  Department  of  l—i hi) ui<,hcs  The University of British C o l u m b i a Vancouver, Canada  •ate  DE-6  (2/88)  Atqirtr  I further  purposes  the  requirements that the  agree that  may  representatives.  financial  permission.  of  It  gain shall not  be is  Library  permission  granted  by  understood be  for  an  advanced  shall make for  the that  allowed without  it  extensive  head  of  my  copying  or  my  written  Abstract This thesis investigates laryngeal processes in Chipewyan and other Athapaskan languages. Athapaskan languages provide an interesting testing ground since they exhibit a three-way laryngeal distinction in stops (voiceless unaspirated, voiceless aspirated and glottalised), as well as a two-way distinction (voiced vs. voiceless) in fricatives. Data from a previously undocumented dialect of Chipewyan is presented to bring new evidence to bear on the cross-linguistic picture within Athapaskan. This dialect shows significant diachronic changes. Acoustic analysis reveals that several of the stops traditionally classified as voiceless unaspirated are phonetically voiced. Furthermore, the results show a front-back asymmetry in voicing. Other findings include merger of the alveolar and palatal stop series, and merger of interdental stops with interdental fricatives. The acoustic findings are used to develop a featural specification of Chipewyan consonants adapted from Rice (1994). The phonological behaviour of these stops has interesting implications for the phonetics-phonology interface. Several morphophonemic processes are examined from a cross-linguistic and comparative historical perspective to test the tenets of feature specification, privative features, constraint definition and interaction. Firstly, Pro to-Athapaskan had a two-way laryngeal contrast stem-finally (maintained, e.g., in Hupa), which has been neutralized in many daughter languages (e.g., Koyukon). Languages such as Chipewyan have undergone a process of stem-final spirantisation. These related processes of laryngeal neutralisation and spirantisation will be examined in an Optimality Theory context using constraints developed by Steriade (1997). Secondly, Athapaskan languages exhibit a phonological process of continuant voicing whereby voiceless noun stem-initial continuants become voiced with the addition of the possessive prefix. This process, displayed by previously documented dialects of Chipewyan, is analysed in an Optimality Theory framework. The research dialect of Chipewyan does not exhibit the process due to a restructuring of the morphosyntactic system of possession marking. Finally, tone and tonal processes, found in most Athapaskan languages, are the synchronic residue of Proto-Athapaskan laryngeal behaviour. Two examples of tone assimilation are discussed: Navajo, where inherent high tone spreads rightwards in verbs, and Chipewyan, where inherent high tone spreads leftwards in nouns.  ii  Table of Contents Abstract  ii  List of Figures  v  Acknowledgements  vii  Chapter 1: Introduction  1  Chapter 2: Historical, Comparative and Theoretical Perspectives  7  2.1 Introduction  7  2.2 Proto-Athapaskan literature review  7  2.2.1  Leer (1979)  7  2.2.2 Krauss (1978)  9  2.3 Northern Athapaskan consonant inventories: Chipewyan vs. Slave  10  2.4 Consonant distribution and consonant variability  12  2.5 Laryngeal features in Athapaskan languages (Rice 1994)  13  2.6 Vowel inventory  16  2.6.1  Previous analyses  17  2.6.2 Research dialect vowel inventory  18  2.6.3 High front vowel [i]  19  2.6.4 M i d front vowel [e]  19  2.6.5 M i d front vowel [e]  21  2.6.6 High back vowel [u]  22  2.6.7 M i d back vowel [o]  22  2.6.8  22  L o w back vowel [a]  2.6.9 Diphthong [ay]  22  2.6.10 Tone  22  2.6.11 Long vowels  23  2.7 Morphological overview  23  2.8 Conclusion  25  Chapter 3: Chipewyan Consonant Inventory: Phonological and Acoustic Evidence  26  3.1 Introduction  26  3.2 Previous acoustic results (Hogan 1976)  27  3.3 Obstruents: stops and affricates  28  3.3.1  Bilabial stop  29  3.3.2 Dental stops  29  3.3.3  31  Interdental stops  3.3.4 Alveolar stops  33  3.3.5  Lateral stops  34  3.3.6 Palatal stops  35  3.3.7 Velar stops  36  3.3.8  Labiovelar stops  36  3.3.9 Glottal stop  37  3.4 Fricatives 3.4.1  38  Interdental fricatives  39  3.4.2 Alveolar 3.4.3  fricatives  Lateral fricatives  39 39  3.4.4 (Alveo-)Palatal  fricatives  iii  40  3.4.5 Velar fricatives  41  3.4.6 Laryngeal fricative  42  3.5 Sonorants  42  3.6 Conclusion  44  Chapter 4: Laryngeal Neutralisation and Spirantisation  47  4.1 Introduction  47  4.2 The Licensing by Cue hypothesis (Steriade 1997)  47  4.3 Data: glottalisation patterns in Athapaskan  49  4.3.1  Stem-initial distribution  49  4.3.2 Stem-final distribution: retention of plain and glottalised stops  50  4.3.3 Stem-final neutralisation  50  4.4 Analysis 4.4.1  51 Previous analysis  51  4.4.2 Optimality Theory analysis 4.5 Spirantisation 4.5.1  52 54  Background  54  4.5.2 Analysis  55  4.6 Conclusion  56  Chapter 5: Continuant Voicing  57  5.1 Introduction  57  5.2 Background information  57  5.2.1  Possessive prefixes  57  5.2.2 Target consonants  57  5.2.3 Theoretical assumptions  58  5.2.4 Previous analyses  59  5.3 Continuant voicing 5.3.1  60  Data  60  5.3.2 Optimality Theory analysis  61  5.4 Research dialect: no continuant voicing  65  5.5 Conclusion  68  Chapter 6: Tone Assimilation  70  6.1 Introduction  70  6.2 Navajo background information  71  6.3 Navajo tone assimilation  72  6.4 Optimality Theory constraints  74  6.4.1  Faithfulness constraints  74  6.4.2 Prosodic Constraints  74  6.4.3 Alignment Constraints  77  6.5 Tone assimilation tableaux  78  6.6 Tonal effects of the Chipewyan possessive suffix  84  6.7 Tone assimilation analysis  87  6.8 Conclusion  88  Chapter 7: Conclusions and Issues for Further Research  90  References  95  Appendix  98  iv  List of Figures Figure 1: M a p of Northern Athapaskan languages  2  Figure 2: Initial [b]; [bis] 'knife'  98  Figure 3: Intervocalic [b]; [tabi] 'net'  98  Figure 4: Initial [d]; dene 'people'  98  Figure 5: Initial [t]; te&e 'bed'.  99  Figure 6: Initial [4  te0 'coal'  99  Figure 7: Intervocalic [d]; ?edu 'hot'  99  Figure 8: Intervocalic [t]; yati 'language'  100  Figure 9: Intervocalic [f]; det'anco 'eagle'  100  Figure 10: Initial [t6]; t©e 'stone/rock'  :  Figure 11: Initial [t6]; t0i?ah 'pillow'  100 100  Figure 12: Initial [t'6]; t'0ay 'bowl'  101  Figure 13: Initial [t'0]; t'Guwe 'milk'  101  Figure 14: Intervocalic [d&]; ?ed6ay 'smoked/dried fish'  102  Figure 15: Intervocalic [t6]; ?et6en 'caribou'.....  102  Figure 16: Intervocalic [t'6]; sit'Gene 'my bone'  102  Figure 17: Initial [dz]; [dzen] 'muskrat/rat'  103  Figure 18: Initial [dz]; [dz§] 'gum'  103  Figure 19: Initial [c]; [c£] 'excrement'  103  Figure 20: Initial [c]; [ca] 'hat'  104  Figure 21: Intervocalic [dz]; [?edzah] 'cold'  104  Figure 22: Intervocalic [c]; [sici] 'my nose'  104  Figure 23: Intervocalic [c]; [sicane] 'my wife'  105  Figure 24: Initial [A,]; [Ayne] 'mouse'  105  Figure 25: Initial [X]; [Xis] 'grease'  105  Figure 26: Initial [t]; [fax] 'rope'  106  Figure 27: Intervocalic [k]; [naXor] 'smile/laugh'  106  Figure 28: Intervocalic [t]; [neXah] 'your buttocks'  106  Figure 29: Initial [j]; fta] 'here'  107  Figure 30: Initial [J]; Dole] 'ball'  107  Figure 31: Initial [c]; [ca] 'rain'  107  Figure 32: Initial [6]; [celtiie] 'tobacco/cigarettes'  108  Figure 33: Intervocalic []]; [?ejere] 'cow'  108  Figure 34: Intervocalic [c]; [dac£] 'raven/crow'  108  Figure 35: Initial [g]; [gokus] 'pig'  109  Figure 36: Initial [g]; [gah] 'rabbit'  109  Figure 37: Initial [k]; [kcj^] 'house'  109  Figure 38: Initial [k]; [ko0] 'clouds'  110  Figure 39: Intervocalic [g]; [yagole] 'butterfly'  110  Figure 40: Intervocalic [k]; [hokamo] 'it tastes good'  110  Figure 41: Intervocalic [k]; [sikd6] 'my neck'  Ill  Figure 42: Labiovelar [k ]; [sik e] 'children'  111  w  w  Figure 43: Labiovelar [k ]; [6ek aze] 'girl, young woman'  111  Figure 44: Initial [?]; [?asi] 'something'  112  w  w  v  Figure 45: Post-consonantal [?]; [har^s?j] 'I want'  112  Figure 46: Initial [0]; [0e] 'ice'  112  Figure 47: Initial [8]; [Se] 'skin'  113  Figure 48: Initial [i]; [fa] 'many'  113  Figure 49: Initial [1]; [lah] 'work'....  113  Figure 50: Initial [y]; [yake] 'window'  114  Figure 51: Intervocalic [y]; [tGiyah] 'hair'  114  vi  Acknowledgements I am extremely grateful to my primary Dene consultant, Max Deranger. This thesis would not have been possible without him. I am also indebted to my second consultant, Rose Desjarlais. I would like to thank my supervisor, Pat Shaw, for her detailed and insightful comments, and committee members Doug Pulleyblank and Henry Davis for all their help. M y thanks also go to the following people: Guy Carden for help with the acoustic analysis, Leora Bar-el for the use of her computer, Gunnar Hansson for help with the final formatting, and my fellow graduate students Tanya Bob, Marion Caldecott, Ikuyo Kaneko, Eun-Sook K i m , Sun-Young Oh, Yumiko Nakamura, Matt Ritchie, Eric Rosen and Uri Strauss for their comments and suggestions. Finally, I would like to thank my parents, Leonard and Ella Gessner for all their support.  vii  Chapter 1 Introduction  1  Athapaskan inventories include three types of stops, traditionally classified as voiceless unaspirated, voiceless aspirated and glottalised, as well as voiced and voiceless fricatives and voiced sonorants. The three-way laryngeal distinction exhibited in Athapaskan languages provides an interesting testing ground for laryngeal processes. The goals of this thesis include the following: (i) presentation of data from a previously undocumented dialect of Chipewyan to bring new evidence to bear on the cross-linguistic picture of Athapaskan languages; (ii) investigation of the acoustic properties of Chipewyan obstruents, focusing mainly on the stop series; (iii) identification of specific issues related to the phonetics/phonology interface; and (iv) examination of several morphophonemic processes from a cross-linguistic and comparative historical perspective to test the tenets of feature specification, privative features, constraint definition and interaction. The Athapaskan language family includes three subgroups: Northern, Pacific Coast and Apachean. The languages in each family are listed in (1). (1)  Athapaskan language family (Krauss and Golla 1981)  Northern Ahtna Tanaina Ingalik Holikachuk Koyukon Kolchan (Upper Kuskokwim) Lower Tanana Tanacross Upper Tanana Han Kutchin Tutchone Slavey-Hare Bearlake Hare Mountain Slavey  Dogrib Chipewyan Beaver Sekani Tahltan-Kaska-Tagish T sets aut Carrier and Babine Chilcotin Sarcee  Pacific Coast Upper Umpqua Tututni-Chasta Costa Galice-Applegate Chetco-Tolowa Hupa Mattole Sinkyone-Wailaki Cahto  ADachean Chiricahua Jicarilla Mescalero Navajo Western Apache Kiowa-Apache Lipan  Chipewyan is a Northern Athapaskan language spoken across a wide geographical area 2  extending through northern Manitoba, northern Saskatchewan, northern Alberta, and in part of the North West Territories south of Great Slave Lake. The Northern Athapaskan languages  The font used throughout this thesis is First Nations New font, developed by Bowen Hui and Cody Shepherd under the auspices of a TLEF grant awarded to Patricia A. Shaw, UBC 1997 & 1998. - Speakers of Chipewyan prefer to call their language [dsne] or [dene suslens] 'real or true language' but there are speakers of other Northern Athapaskan languages who also use the term dene for their language; therefore, to avoid any confusion, I will refer to the language as Chipewyan. 1  1  including Chipewyan are shown on the map in Figure 1. Figure 1:  Map of Northern Athapaskan languages (Helm 1981) (Border delineated by crossed lines.)  Chipewyan is estimated to have 12 000 speakers (Cook 1991), giving it the second largest speaker population among Athapaskan languages after Navajo. The majority of the data presented in this thesis was collected through fieldwork sessions with Max Deranger, a native speaker of Chipewyan from Fort McMurray, Alberta, during 1998-1999. Additional fieldwork was conducted in April 1999 with a second speaker from Fort McMurray, Rose Desjarlais. Forms provided by Rose Desjarlais are marked (RD); those not otherwise cited were provided by Max Deranger. Comparisons will be made with previously published data, particularly that of L i (1933, 1946), which was collected in Fort Chipewyan, Alberta, during 1928. Chapter two provides an overview of historical, comparative and theoretical issues which will constitute the essential background information relevant to issues presented in this thesis, starting with a review of the Proto-Athapaskan (PA) linguistic literature. There are several important issues in the historical literature. Leer (1979) gives a reconstruction of the ProtoAthapaskan inventory and hypothesises several major phonological rules occurring in the transition from Pre-Proto-Athapaskan (PPA) to Proto-Athapaskan with reference to  2  correspondences in several of the daughter languages. The consonant inventory of Chipewyan differs significantly from the inventory posited for Proto-Athapaskan. According to Leer (1979), several consonant series became "fronted" in the transition from Proto-Athapaskan to Chipewyan. A second significant difference between the P A inventory and those of the daughter languages concerns the distribution of stem-final consonants. Some languages such as Koyukon have neutralised the contrast between plain and glottalised stops stem-finally, and other languages such as Chipewyan have lost all stops stem-finally. These processes of laryngeal neutralisation and spirantisation are discussed in chapter four. Krauss (1978) presents a mapping of the Athapaskan language area for tone. There is a great deal of variation among tonal properties of modem Athapaskan languages. Krauss makes an important correlation between tone and the feature [constricted glottis] in the post-vocalic segments of the Pre-Proto-Athapaskan stem. Chipewyan is a high-marked tone language, and I discuss a process of tone assimilation in chapter six. In addition to reviewing the historical literature, chapter two presents the consonant inventory of Chipewyan and compares it with that of Proto-Athapaskan, and that of Slave, a related Northern Athapaskan language, with discussion of consonant distribution and variability. Further phonological and acoustic evidence of the Chipewyan consonants is discussed in chapter three. The evidence will provide argumentation for a featural specification of the consonants. Previous hypotheses of featural specifications of consonants in Athapaskan languages is presented, including that of Rice (1994). Rice (1994) examines several synchronic laryngeal processes in Athapaskan, comparing a variety of languages in the family. Rice develops several hypotheses to characterise the specification of laryngeal features, including the Dual Mechanism Hypothesis which I follow for my analysis of laryngeal neutralisation and continuant voicing discussed in chapters four and five respectively. Finally, chapter two describes the vowel inventory of the research dialect of Chipewyan, comparing it to previous inventories documented in the literature. This chapter also includes a brief outline of the Chipewyan morphology. In chapter three, I review the only previous acoustic study done on Chipewyan, that of Hogan (1976). The goal of Hogan's study was to determine some temporal features of ejective consonants and to examine the characterisation of voice onset time. This study is supplemented by my own acoustic findings which focus on the obstruent inventory of Chipewyan and show several interesting discoveries, including voicing of some of the stops traditionally classified as voiceless unaspirated, merger of the interdental stops with the interdental fricatives, merger of the alveolar stops with the palatal stops and clarification of the status of the fricatives [1] and [y]. These findings will make a contribution to our understanding of these issues from the perspective of comparative Athapaskan phonology. The key research question of chapter three concerns the laryngeal properties of the voiceless unaspirated (plain) stops. Some Northern Athapaskan languages have been reported to have voiced, rather than voiceless unaspirated stops. Wave form and spectrogram analysis of the  3  plain stops shows that in the research dialect, the bilabial and dental stops are voiced rather than voiceless unaspirated. The alveolar and palatal stops are shown to be intermediate cases which are not consistently voiced or voiceless throughout the stop duration. The lateral affricate, velar stop and glottal stop are shown to be unambiguously voiceless. The results show that there is a front-back asymmetry among the plain stops, with plain stops near the front of the mouth being voiced, plain stops in the middle being intermediate (i.e. partially voiced) cases, and plain stops near the back of the mouth being voiceless unaspirated. In other words, there is a cline of voicing from front to back, which can be explained by principles of aerodynamics. These findings make an interesting claim regarding the interface of phonetics and phonology. Stops which behave as a class phonologically are shown to have different phonetic realisations, showing that the phonology is not a direct reflection of phonetics. The second interesting finding of the acoustic analysis concerns the interdental series of stops. In the research dialect, the interdental stops (excluding the glottalised interdental stop) are perceived to be fricatives; this is confirmed by the spectrogram analysis. These consonants continue to pattern phonologically with the stops in not undergoing the phonological process of continuant voicing, whereby unvoiced noun stem-initial continuants become voiced when the possessive prefix is added. This constitutes a second example of a mismatch between the phonetics and phonology. Argumentation provided by the acoustic data in chapter three will be used to substantiate hypotheses about laryngeal specification of the obstruents of Chipewyan in a feature geometric model. The remaining chapters examine several phonological processes concerning laryngeal properties, and present them in a current theoretical context, that of Optimality Theory (McCarthy and Prince 1993). Optimality Theory (OT) is a linguistic theory which uses ranked well-formedness constraints to determine an optimal candidate from among any number of possible candidates generated by a given grammar. A n important difference from previous approaches is that instead of taking an input form and applying successive rules which will obtain the eventual grammatical output, Optimality Theory takes a large set of candidate outputs and evaluates them against a hierarchy of constraints which will select the optimal (actual) grammatical output. The set of constraints is determined by Universal Grammar (UG); their ranking is language-particular. In the optimal candidate, lower-ranked constraints may be violated at the expense of satisfying higher-ranked constraints. A n optimal candidate does not have to be perfect; it simply is the one that comes closest to satisfying, or minimally violates, the constraints as ranked. The main principles of Optimality Theory are (McCarthy and Prince 1994:3): a.  Universality. U G provides a set Con of constraints that are universal and  universally present in all grammars. b.  Violability. Constraints are violable; but violation is minimal.  4  c.  Ranking. The constraints of Con are ranked on a language-particular basis; the  notion of minimal violation is defined in terms of this ranking. A grammar is a ranking of the constraint set. d.  Inclusiveness. The constraint hierarchy evaluates a set of candidate analyses  that are admitted by very general considerations of structural well-formedness. e.  Parallelism. Best-satisfaction of the constraint hierarchy is computed over the  whole hierarchy and the whole candidate set. There is no serial derivation. Optimality Theory uses a "Constraint Tableau" to evaluate the candidates. In the following example (McCarthy and Prince 1994:4) shown in (2), constraint A is ranked higher than constraint B. Violation of a constraint is indicated by "*", and "!" indicates a fatal violation. Solid lines in a tableau indicate crucial ranking of candidates, whereas dotted lines indicate noncrucial or indeterminate ranking. The optimal candidate is indicated by "est". (2) Example tableau Candidates  B  A  *  •^k-cand 1 k-cand 2  *!  Chapter four discusses laryngeal neutralisation and spirantisation in Athapaskan languages. Proto-Athapaskan (PA) is posited by Leer (1979) to have had a three-way laryngeal distinction between voiceless unaspirated (plain) stops, voiceless aspirated stops and glottalised stops. A s mentioned above, these distinctions are maintained in all of the daughter languages, at least in stem-initial position. However, stem-final contrasts have been reduced in most languages. P A is posited to have had a two-way contrast between plain and glottalised consonants stemfinally. In the daughter languages, this two-way laryngeal contrast may be maintained (in languages such as Hupa), or the laryngeal distinctions may be neutralised (in languages such as Koyukon). In some languages, such as Chipewyan, a diachronic process of spirantisation has occurred. A l l series of stops in Chipewyan have undergone spirantisation. The spirantisation process is relevant to the discussion of laryngeal neutralisation because while spirantisation involves loss of the manner feature [stop], it also entails loss of the stop's laryngeal features [spread glottis] or [constricted glottis]. Steriade (1997) proposes a cue-based approach to the licensing of laryngeal features. Laryngeal features are typically neutralised in positions where the cues to the relevant contrast are not available, or where additional articulatory measures would be necessary. Laryngeal contrasts are permitted in positions where the scale of perceptibility o f these contrasts is high. The aim of chapter four is to test glottalisation evidence from the Athapaskan language family against Steriade's hypothesis, and uses Optimality Theory constraints developed by Steriade (1997) in the analysis. Chapter five examines the process of continuant voicing. Continuant voicing is a familiar phonological process in Athapaskan languages (Young and Morgan (1980), Cook (1984),  5  Hargus (1988), Rice (1989) and others). The focus environment discussed in this chapter is the stem-initial continuant of nouns. When a possessive prefix is added, a voiceless continuant surfaces as a voiced continuant. Data is presented from two dialects of Chipewyan. Data from the Fort Chipewyan dialect of Chipewyan documented by L i (1933) exhibits the continuant voicing process, and will be examined using Optimality Theory. It is shown that the same set of constraints can account for the continuant voicing process, as well as prevent the process from applying to non-continuants. The second dialect (the dialect of my research) does not exhibit the process, due to the fact that the environment which triggers voicing is not present. This shows evidence of restructuring of the morphosyntactic system in possession-marking and it raises questions concerning the actual underlyingly form of the stem-initial consonant of inalienable nouns. Chapter six presents two examples of tone assimilation in Athapaskan languages in different morphological domains: tone assimilation in verbs in Navajo, and tone assimilation in nouns in Chipewyan. Tone and tonal processes are the synchronic residue of Proto-Athapaskan laryngeal behaviour. In Navajo, certain prefixes are lexically marked for high tone. This tone may spread to a following mora (i.e. syllabic nucleus (vowel or syllabic nasal)) subject to certain constraints. The required constraints will show that both prosodic and morphological structures play an important role in this analysis. Tone will only spread within a bimoraic foot. Feet are aligned to the right edge of the prosodic word. Chapter six provides an explicit account in an Optimality Theory framework of how these constraints interact to account for the basic pattern of tone assimilation while also accounting for examples where tone assimilation does not occur as might be expected. Chipewyan exhibits a process of tone assimilation triggered by the possessive suffix. Possessed nouns in Chipewyan are most often formed by adding a possessive prefix to the noun stem. When the possessive prefix is added, the stem form may remain the same, or may be changed in various ways including addition of a possessive suffix. The possessive suffix also surfaces in some types of compound formation. Addition of the high-toned possessive suffix to some noun stems results in the high tone of the suffix spreading to the preceding vowel in the stem. A similar constraint set is used to analyse the Chipewyan tone assimilation process. Chapter seven summarises the conclusions of each chapter, and discusses issues for further research.  6  Chapter 2 Historical, Comparative and Theoretical Perspectives 2.1  Introduction  This chapter will provide an overview of historical, comparative and theoretical issues which will constitute the essential background information relevant to issues presented in this thesis. Section 2.2 will review the Proto-Athapaskan (PA) linguistic literature. There are several important issues in the historical literature. First, the consonant inventory of Chipewyan differs significantly from the inventory posited for Proto-Athapaskan. According to Leer (1979), several consonant series became "fronted" in the transition from Proto-Athapaskan to Chipewyan. A second significant difference between the P A inventory and those of the daughter languages concerns the distribution of stem-final consonants. Some languages such as Koyukon have neutralised the contrast between plain and glottalised stops stem-finally, and other languages such as Chipewyan have lost all stops stem-finally. These processes of laryngeal neutralisation and spirantisation will be discussed in chapter four. Section 2.3 will present the consonant inventory of Chipewyan and compare it with that of Proto-Athapaskan, and that of Slave, a related Northern Athapaskan language. Consonant distribution and variability will be discussed in section 2.4; further phonological and acoustic evidence of the Chipewyan consonants will be discussed in chapter three. The evidence will provide argumentation for a featural specification of the consonants. Previous hypotheses of featural specifications of consonants in Athapaskan languages will be presented in section 2.5. Section 2.6 will provide a description of the vowel inventory of the research dialect of Chipewyan, comparing it to previous inventories documented in the literature. Finally, this chapter will include a brief outline of the Chipewyan morphology in section 2.7. Section 2.8 will conclude. 2.2 Proto-Athapaskan literature review 2.2.1  Leer (1979)  Leer's (1979) influential work on Proto-Athapaskan (PA) verb stem variation suggests that the great amount of stem variation in Athapaskan languages has developed due to regular phonological processes operating on underlying roots in P A . Leer gives a reconstruction of the Proto-Athapaskan inventory and hypothesises several major phonological rules occurring in the transition from Pre-Proto-Athapaskan (PPA) to Proto-Athapaskan with reference to correspondences in several of the daughter languages. The P A consonant inventory is shown in (1). Note that in the consonant inventories presented in this thesis, the use of "voiced" symbols representing the voiceless unaspirated stops follows traditional Athapaskan usage, i.e. 'd' is phonetically [t], 't' is phonetically [r- ], etc. 1  7  (1)  Proto-Athapaskan consonant inventory (Leer 1979:4)  Obstruents (X)  plain  back velar (uvular)  glottal  (G->  (H)  g  G  ?  k  q  |  q  w  X  X  w  y  Y'  front velar  alveolar  affricate series  (D)  (DZ)  (DL)  (Dz)  (Dz")  (G)  d  dz  X  J  r  t  c  X  C  c  w  i  c  t  6  W  s  f  s  s  z  1  z  z  stops (D) aspirated stops (T) glottalised stops ( T ) voiceless  h  fricatives (S) voiced fricatives (Z) Sonorants (R) front  back  oral  y  w  nasal (N)  y  w  n  With respect to the obstruents, the manner of articulation of stem-initial obstruents according to Leer's reconstruction has not changed much from P A to the daughter languages, with a threeway distinction between voiceless unaspirated or plain stops, voiceless aspirated stops and glottalised stops. However, stem-final contrasts have been reduced in most languages. P A had a two-way contrast between plain and glottalised consonants stem-finally. In the daughter languages, this two-way laryngeal contrast may be maintained (in languages such as Hupa), or the laryngeal distinctions may be neutralised (in languages such as Koyukon). In some languages, such as Chipewyan, a diachronic process of spirantisation has occurred. A spirantisation process which developed during the transition from P P A to P A operated on root-final stops, changing them to fricatives. There were two specific environments for rootfinal spirantisation: when followed by an obstruent-only suffix, and in verb stem-final position where the stem vowel was full . Examples are shown in (2). 3  (2)  P A Spirantisation fLeer 1979:25) When followed by an obstruent suffix: P P A •t'aq-'r 'fly' prog. > P A *t'ax.'t In verb-stem-final position where the stem vowel is full: P P A *t'a-q 'fly' mom. imperf.-opt. > P A *t'a-x.  Proto-Athapaskan full vowels are long except immediately preceding glottal stop or the enclitic *+he-. The enclitic *+he- is a negative enclitic which deletes length from preceding full vowels (Leer 1979:67). Reduced vowels occur in stem syllables only when they precede a consonant other than glottal stop (Leer 1979:5). 3  8  Cases where the stem-final obstruent failed to spirantise (such as the perfective stem) indicated the presence of a vocalic suffix. The spirantisation process operated concomitantly with phonological constriction; if the root-final consonant was glottalised and became spirantised, the stem vowel became constricted. Therefore, the laryngeal feature did not neutralise stemfinally, as we will see happening in some of the daughter languages in chapter four. Addition of the obstruent suffix which caused spirantisation resulted in a consonant cluster which has been reduced in most cases in the daughter languages, including Chipewyan, usually by dropping the second member. Leer suggests that the consonant cluster reduction is dated after P A , and it is specific to groups of daughter languages. A n example of the three stages of development is shown in (3) (Leer 1979:25). (3)  P P A * g 3 c - f > P A * g V s < l > Chip.jee 'hook'  In some languages such as Chipewyan and Dogrib, all stem-final stops/affricates have become fricatives. The processes of laryngeal neutralisation and spirantisation will be discussed further in chapter four. P A verb-stem initial fricatives were thought to be voiceless when immediately preceded by an obstruent but were voiced otherwise, as is the situation in Chipewyan. In P P A , all fricatives were thought to be voiceless, and developed into voiced fricatives in intervocalic environments in P A . Some languages, including Chipewyan, have developed light/heavy stem syllable pairs. Light stems with voiceless stem-final fricatives developed from stems where the original stemfinal fricative was voiceless. Heavy stems with voiced stem-final fricatives developed where the original stem had a vocalic suffix.  4  Modern Athapaskan languages have a synchronic  process of continuant voicing which occurs in environments which are not necessarily intervocalic. Continuant voicing in Chipewyan will be discussed in chapter five. Nasalisation arose from the addition of an obstruent suffix to a nasal-final root. This has happened in languages including Chipewyan, whereas some Athapaskan languages have lost nasalisation completely. P A had stems with both constricted and non-constricted vowels which developed into two opposite types of tone systems: P A constricted vowels become high-toned vowels and P A non-constricted vowels became low-toned vowels in "high-marked" tone languages, of which Chipewyan is an example. In other languages, the same constricted vowels developed into low tone, with non-constricted vowels getting high tone; these systems are "low-marked" tone languages. This is discussed in greater detail by Krauss (1978) which I review in the following section. 2.2.2 Krauss (1978) Krauss presents a mapping of the Athapaskan language area for tone. He notes that tone in This does not imply that there is a weight-sensitive distinction in the behaviour of light/heavy stems.  9  Chipewyan was first mentioned by Petitot (1876) and Legoff (1882). Sapir was among the first to make cross-linguistic comparisons of the tonal differences in Athapaskan languages. L i (1927) in his study of Sarcee was the first to note that a low-toned vowel in open syllables was regularly followed by a glottal stop. In his later (1928) fieldwork with Chipewyan, L i discovers the reverse situation from Sarcee, that high tone occurs before glottal stop, and notes that the high toned-fofms are connected with the post-vocalic glottal stop which has been lost. Krauss (1964) was the first to formally state that P A was probably toneless. There is a great deal of variation among tonal properties of modern Athapaskan languages. The Pacific Coast Athapaskan subgroup has kept much P A glottalisation and shows stress but no tone. Many Alaskan languages (such as Ingalik, Koyukon and Minto) apparently lack phonemic tone in stems which suggests they are also non-tonal. Other languages where stemfinal glottalisation has been lost have developed into either high-marked or low-marked tone languages. According to Krauss, the nature of the development of tone is therefore correlated with the feature [+/- glottal] in the post-vocalic segments of the Pre-Proto-Athapaskan stem. "The [+/5  glottal] feature can easily be suprasegmentalised in a form which is still glottal but not tonal, and which then in some languages (Chipewyan-Hare) becomes high tone, and in others low (Sarcee-Kutchin-Navajo)" (Krauss 1978:10-11). Where the vowel in the P P A stem was followed by a glottal stop, a glottalised consonant or a sequence of resonant-glottal stop, the vowel became constricted in P A . These constricted vowels developed into tone in the daughter languages. Why two opposite tone values developed from the same source is still unresolved. A possible explanation may relate to how the constricted vowels were articulated in different languages; perhaps some languages used raised larynx while others used lowered larynx (Patricia Shaw, p.c). It is an issue which requires further research. Finally, along with the non-tonal and tonal languages, several modem Athapaskan languages such as Carrier and some Alaskan languages have a tonal system most often characterised as "pitch accent". Krauss hypothesised that these are either decaying tone systems or vestiges of an earlier type of tone. Alternatively, perhaps a pitch accent system is one where only limited tone developed from P A . Chipewyan is a high-marked tone language, and I will discuss a process of tone assimilation in chapter six. Having reviewed the historical background, the next section will present the Chipewyan consonant inventory and show how it compares to the Proto-Athapaskan inventory and the inventories of other related Northern Athapaskan languages. 2.3 Northern Athapaskan consonant inventories: Chipewyan vs. Slave Like all Athapaskan languages, Chipewyan has a rich inventory of consonants, which are traditionally grouped into "series" by place of articulation. Each series consists of three types  Note that the term I will be using for [+/- glottal] is [constricted glottis] or [CG].  10  of stops/affricates (voiceless unaspirated, voiceless aspirated and glottalised) two types of fricatives (voiceless and voiced) and a limited number of sonorants. The consonant inventory of Chipewyan is shown in (4) (adapted from L i 1946) . 6  (4)  Chipewyan consonant inventory  Interdental  Alveolar  Lateral  Cor/Dor Palatal  Dorsal Velar  Labio-velar  d  dS  dz  X  J  t t'  c  X  C  g k  c  k  k  w  s  t i  t  d.  te t'e e  g k  s  X  x  w  e.  6  z  V  Y  Y (~[w])  a.  Labial Labial  Coronal Dental  b  b. c.  f.  (m)  r  y  8  Laryngeal Glottal  W  w  ? h(~[w])  w  n  a. voiceless unaspirated stop ("plain" stops)  d. voiceless continuant  b. voiceless aspirated stop  e. voiced continuant  c. glottalised stop  f. sonorant  In the transition from Proto-Athapaskan, several significant changes have occurred in the consonant inventory of languages (such as Chipewyan) which exhibit an interdental series. This can be seen by comparing (4) with the consonant inventory of Proto-Athapaskan, which was shown in (1) above. According to Leer (1979), there has been a systematic "fronting" of consonant series, with the P A * D Z (alveolar) series becoming the interdental series, the P A * D z (palatal) series becoming the alveolar series, the P A front velar series becoming palatal, and the P A back velar or uvular series developing into the velar series. The dental, lateral and glottal series are unchanged from Proto-Athapaskan. The labialised palatal series has merged with the palatal series. In addition, Chipewyan has developed a bilabial stop, and the voiced fricative [z] has been replaced with [y]. The'sonorants have changed considerably, which is discussed in detail in Krauss and Leer (1981), but will not be the focus of this thesis. Examples of vocabulary illustrating each series along with acoustic analysis will be presented in chapter three. In the Northern Athapaskan picture, the consonant inventory of Chipewyan is very similar to other northern languages. One of the most closely related language is Slave, which is characterised by Rice (1989) as a "dialect complex" consisting of Slavey, Bearlake, Hare and Mountain. The consonant inventory of Slave in shown in (5). The consonant inventory follows Li's (1946) phonemicisation but is reorganised according to current Athapaskan practice. [1] is usually classed with the fricatives in Athapaskan languages since it patterns as the voiced alternant of [f] in phonological processes such as continuant voicing, which will be discussed in chapter five. The acoustic and distributional behaviour of [1] in the research dialect will be discussed in section 3.4.3. 7  Li (1946) lists [y] with the fricatives as the voiced alternant of [s] although it also occurs underlyingly. Li (1933:122) notes that "y is pronounced like the Eng. y, without much frication, but it always acts as the voiced correspondent of s ... and seems therefore justifiably listed as such." The acoustic and distributional behaviour of [y] in the research dialect will be discussed in section 3.4.4.  11  Slave consonant inventory (Rice  (5)  Interdental  Alveolar  Lateral  Cor/Dor Palatal  Dorsal Velar  d  d6  dz  X  J  g  t  te t'e  c  X  C  k  t  t  k  s  X  z  Y  Labial Labial  Coronal Dental  a.  b  b.  P  c.  P  1  1  1989:76)  t'  3  3  3  5  C  d.  fl  03  s  e.  vl  63  z  f.  m  1  n  (r) ga. voiceless unaspirated stop ("plain" stops)  Labio-velar  g  w3  k  w2  Laryngeal Glottal  ?  k 2 w  h w  n y 1. Mountam  b. voiceless aspirated stop  2. Bearlake, Hare  c. glottalised stop  3. Slavey  d. voiceless continuant  4 . not found in Hare  e. voiced continuant  5. Bearlake, /w/ in Hare  f. nasal g. sonorant Phonemes occurring in Slave which do not occur in Chipewyan are shown in bold. Also notable is the fact that the interdental series, which occurs in Chipewyan, only occurs in the Slavey dialect of Slave. The next section will discuss consonant distribution and consonant variability. 2 . 4 Consonant distribution and consonant variability In syllable- and word-initial (onset) position, all consonants can occur except for [d6], [Y], [h] and [r]. The interdental affricate [d6] does not occur word-initially, although it occurs steminitially viz.  ?edday 'smoked/dried fish'. (See section 3 . 3 . 3 ) . The velar fricative [Y] can occur ?eyeges 'eggs'. (See section 3 . 4 . 5 ) . The  in onset position anywhere but word-initially viz.  laryngeal fricative [h] never occurs as the initial of a stem syllable (Li  1946:398),  but [h] can  occur in onset position elsewhere. The sonorant [r] cannot occur word-initially or as the initial of a stem-syllable. It does occur in onset position elsewhere, and in coda position. (See section 3 . 5 . ) Syllable-final or coda consonants are limited to the following fricatives and sonorants, shown in  (6):  12  (6)  Syllable-final/coda consonants  [6]  ce6 'duck'  [6]  xa6 'boil' (Li 1933:134)  [s]  sas 'bear'  [z] [f]  der 'crane' (Li 1933:129)  sez 'hiccough' (Li 1933:146)  [1]  del 'blood'  [§] [x]  nates 'dream/vision'  [Y]  xax 'playing cards' coy 'porcupine quill' (Li 1933:149)  [h]  gah 'rabbit'  [n]  dzen 'muskrat/rat'  M  bar 'meat'  The most commonly occurring coda consonants in the research dialect are [n], [h] and [s]. This distribution of coda consonants is much reduced from that exhibited in Proto-Athapaskan. This will be discussed further in chapter four: laryngeal neutralisation and spirantisation. The inventory of the dialect of Chipewyan presented in this thesis differs from the dialect presented in (4) in several ways, which will be the focus of chapter three. The main analytical question to be answered concerns the laryngeal properties of the voiceless unaspirated (plain) stops. Wave form and spectrogram analysis of the plain stops will show that in the research dialect, there is a front-back split among the plain stops, with plain stops near the front of the mouth being voiced, plain stops in the middle being intermediate (i.e. partially voiced) cases, and plain stops near the back of the mouth being voiceless unaspirated. The second interesting variation in the research dialect concerns the interdental series of stops. Acoustic evidence will show that the voiceless unaspirated interdental affricate has merged with the voiced interdental fricative, and the voiceless unaspirated interdental affricate has merged with the voiceless interdental affricate. A n ongoing merger of the alveolar affricate series with the palatal affricate series will also be shown. There are also several notable changes with the fricatives. First, the voiced velar fricative [Y] is often replaced by another phoneme ([y], [w], [g] or [h]). Secondly, the acoustic reality of the phonemes [1] and [y], which have traditionally been classified with the voiced continuants rather than the sonorants, will be discussed. It will be seen that [1] exhibits properties characteristic of fricatives, while [y] exhibits properties more typical of sonorants. Argumentation provided by the acoustic data in chapter three will be used to substantiate a featural specification of the consonants of Chipewyan. To address these issues, a paper by Rice (1994) is particularly relevant, and will be reviewed in the following section. 2.5 Laryngeal features in Athapaskan languages (Rice 1994) A s we have seen, in a typical Athapaskan inventory, there are three laryngeal classes of stops/affricates (voiceless unaspirated, voiceless aspirated and glottalised), two laryngeal  13  classes of fricatives (voiced and voiceless), and one laryngeal class of sonorants (voiced). Rice examines three different hypotheses to determine how the laryngeal oppositions in stops/affricates and fricatives can best be characterised. In the Dual Mechanism Hypothesis, there are two active laryngeal mechanisms, one for stops and one for fricatives. Under the other two hypotheses, only one mechanism is active: voicing in the Voicing Hypothesis, and aspiration ([spread glottis]) in the Aspiration Hypothesis. These hypotheses are shown in (7) (9).  Stops/affricates are represented by t/d and fricatives by s/z, under the tacit assumption that  all place articulations within a particular laryngeally-defined series behave the same. (7)  Dual Mechanism Hypothesis ( D M H ) (Rice 1994:108) (based on Kari 1973; traditional descriptions) d  t  s  voice stop (8)  z  V  spr gl  V  V  V  Voicing Hypothesis (VH) (Rice 1994:108) (based on Stanley 1969, Howren d  (9)  voice  V  stop  V  t  s  1971)  z V  V  Aspiration Hypothesis (AH) (Rice 1994:108) (based on Rice 1988) d spr gl stop  V  t  s  V  V  z  V  These three classifications produce the following feature geometries, shown in (10) - (13).  14  (10)  Dual Mechanism Hypothesis (Rice 1994:115):  Plain stop ROOT  Aspirated Stop ROOT  /  Gltlsed Stop ROOT  \  /  Vis fricative  V d fricative  ROOT  ROOT  \  [stop] Laryngeal  [stop]  [voice]  9  [stop] Laryngeal [constr gl] [spr gl] (11)  Voicing Hypothesis (Rice 1994:115):  Plain stop ROOT /  Aspirated Stop  Gltlsed Stop  ROOT  ROOT  [stop]  [stop] Laryngeal  /  \  [stop] [voice]  Vis fricative  V d fricative  ROOT  ROOT  \ [voice]  [constr gl] (12)  Aspiration Hypothesis (Rice 1994:116):  Plain stop ROOT  Aspirated Stop ROOT /  [stop]  \  [stop] Laryngeal [spr gl]  Gltlsed Stop ROOT /  Vis fricative  V d fricative  ROOT  ROOT  \  [stop] Laryngeal [constr gl]  Laryngeal [spr gl]  Each hypothesis makes different predictions regarding which consonants should pattern together with respect to laryngeal features. Under the Dual Mechanism Hypothesis, plain stops and voiceless fricatives may be expected to share properties because they both lack laryngeal features. The Voicing Hypothesis predicts that plain stops and voiced fricatives should pattern together with respect to the feature [voice]; aspirated stops and voiceless fricatives should pattern together since they both lack the [voice] feature. The Aspiration Hypothesis predicts that aspirated stops and voiceless fricatives should pattern together with respect to the feature [spr gl] as do plain stops and voiced fricatives for lacking this feature. Rice examines several phonological processes to determine which model provides the best classification. The results are summarised in (13).  It is important to note that Rice characterises the feature [voice] as a dependent of the root node rather than of the laryngeal node. This is because: "The type of voicing found in Athapaskan languages is characterised [by Rice and others] as Sonorant or Spontaneous Voice (SV). This feature, labelled [voice] here, defines the class of sonorants, including vowels, underlyingly, and is present in voiced fricatives in Athapaskan languages." (Rice 1994:142) While I adopt Rice's Dual Mechanism Hypothesis to discuss issues in the remaining chapters of this thesis, I prefer to represent the feature [voice] as dependent of the laryngeal node, in parallel with the laryngeal distinctions [SG] and [CG] which characterise the aspirated and glottalised stops respectively. 9  15  (13)  Summary (from Rice 1994:133)  process  language  lexical/post-lexical  support for DMH/VH/AH  d-effect rule cont. voicing  Slave, Navajo  lexical  DMH  Slave, Navajo,  lexical  DMH  Koyukon spirantisation  Koyukon  lexical  DMH  deaffrication  Slave, Navajo, Koyukon  post-lexical  AH/VH  ts --> s deaffrication  Navajo, Koyukon  post-lexical  AH/VH  dz —> z Rice also looks at the distribution of stops and fricatives in the P A consonant inventory, where aspiration is neutralised in final position while voicing is not. The distribution of laryngeal features on stem-final stops and fricatives is asymmetric in the daughter languages, where some languages retain the two-way laryngeal distinction and other languages neutralise this distinction. Both can be accounted for with the Dual Mechanism Hypothesis. The Dual Mechanism Hypothesis is also the best model to explain the first three processes shown in (13). Each process involves either the linking or delinking of a feature, with the resulting representation the one predicted by the Dual Mechanism Hypothesis. While the hypothesis does not account for the last two processes, Rice argues that the hypothesis is not discounted by these examples, since deaffrication is a synchronic rule that applies postlexically. The evidence for this is that fricatives derived from affricates, and underlying fricatives pattern differently with respect to phonological processes. To summarise Rice's argument for postlexical laryngeal features, the crucial point is that redundant features are not present in the underlying representations (as represented by the Dual Mechanism Hypothesis) but are added late in the derivation. Therefore, if the feature [spr gl] is added to voiceless fricatives redundantly, the deaffrication process, delinking of the stop feature, will result in a voiceless fricative, as expected. I will examine Rice's analysis of laryngeal neutralisation and continuant voicing in more detail in chapters four and five respectively. It will be shown that the Dual Mechanism Hypothesis is the one which provides the best representation of the consonants active in these processes. Having outlined the Chipewyan consonant inventory and having discussed the related historical, comparative and theoretical literature, we can turn to the vowel inventory of Chipewyan, presented in the following section. 2.6 Vowel inventory There seems to be a large amount of vowel variability in the various dialects of Chipewyan presented in different sources. Section 2.6.1 will review three proposed inventories, that of Leer (1979), L i (1946) and Cook (1983). Section 2.6.2 will present my proposed vowel  16  inventory for the research dialect and sections 2.6.2-2.6.9 will present examples of each phoneme. 2.6.10 will briefly discuss the phonemic status of tone, and 2.6.11 will discuss long vowels. 2.6.1  Previous analyses  Leer (1979) reconstructs Proto-Athapaskan with four full vowels and three reduced vowels. P A vowels and their correspondences in Chipewyan are shown in (14). (14)  10  P A and Chipewyan vowel inventory (Leer 1979:10-11) Proto-Athapaskan  Chipewyan  *l(0  i(0  *e(-)  e, e a(-)  *a() *u(-) *9  u(-)  *a  a (~ ai/  P A (obstruent) front velar)  *v  o  P A (obstruent) front velar)  3, e (~ a/  (r  uf  P A back velar, ~ i /  P A (obstruent) front velar)  L i (1946) lists a large inventory of vowels, shown in (15). (15)  Vowel inventory (Li 1946:399) Short vowels:  a  s  e  i  0  u  Long vowels:  a-  e-  i-  u-  Nasalised short vowels:  a  ?  u  Nasalised long vowels:  aai  ?•  i v  Diphthongs:  ei  ai  oi  uui  Qi  ui  L i (1946:399) notes that "e varies between [e] and [o]: [e] in open syllables or when followed by t or 1, sometimes s or z; [s] in the diphthong ei or when followed by a voiced final consonant...[and] the diphthongs only occur in syllables followed by a voiceless spirant, except ei which may occur in open syllables." He also states there are a number of "pseudodiphthongs" which result from two vowels fallen together, and long vowels are often due to the coalescence of two vowels. Vowels are also marked by one of two tones: high tone, marked by an acute accent, and low tone which is unmarked. Cook (1983) presents a much simplified vowel inventory, which more closely resembles the Proto-Athapaskan inventory, shown in (16).  1 0  1 1  Note that Leer (1979) bases his Chipewyan inventory on datafromLi (1933). According to Leer (1979:4), Id is phonetically [as], and /a/ is phonetically [o].  17  (16)  Vowel inventory (Cook 1983:416) i  u e  o  o  a The first major difference between the two inventories is the treatment of [e], [e] and [a]. While L i considers [e] and [a] allophones of [e], with [e] considered a separate phoneme, Cook treats schwa and [e] as separate phonemes, with [e] and [e] allophones of [e]. Cook states that [e] occurs in optimally distinct positions such as in stems instead o f affixes or in cited forms rather than in natural texts. Under nonoptimal conditions, [e] is laxed to [e]. The laxing rule is not phonetically conditioned and may be considered to be free variation (Cook 1983:416). Regarding the diphthongs, Cook (1983) points out inconsistencies in Li's data. He suggests that diphthongs need to be distinguished from V C sequences i.e. [ai] vs. [ay]. Cook concludes that most of Li's diphthongs can be derived from V C sequences, or a sequence of two vowels, and the only "true" diphthong is probably [ai]. Cook does make the distinction between the diphthong [ai] and the V - C sequence [ay]. Besides maintaining that there is a clear phonetic difference between the two, Cook points out phonological alternations such as [ai] alternating with [aye], as in  xai/xaye 'winter' and beskai/beskaye 'seagull' (Cook 1983:419-120). Also,  the [y] of [ay] is frequently dropped which does not happen with [ai] since it is phonologically a single segment (Cook 1983:420). With regard to vowel length, Cook (1983) suggests that long vowels are best interpreted as a sequence of two vowels, since that is what most long vowels derive from, as L i (1946) noted. Minimal pairs contrasting short and long vowels could not be found, suggesting that vowel length is not phonemic. Long vowels not derived from a sequence o f two vowels occur in open syllables, particularly in final position or when followed by a voiced fricative.  For  example, [a] in sa 'sun' is considerably longer than [a] in sas 'bear', and [e] in fez 'flour' is considerably longer than [e] in des 'river' (Cook 1983:425). Since this variable length is predictable based on the environment, Cook concludes that it need not be marked. As for nasal vowels, Cook (1983) considers nasalisation distinctive for all tense vowels i.e. all vowels except schwa, although the status of nasal vowels is unstable. 2.6.2 Research dialect vowel inventory The vowel inventory I propose for the research dialect is shown in (17). (17)  Vowel inventory (Gessner 1999) i  u  e (9)  o  e (i)  a  Diphthong: ay Nasalised vowels: j e. e. u Q a  18  While the size of my data set is not large enough to provide examples of minimal pairs for each distinction, the following minimal pairs shown in (18) are found. (18)  Minimal Pairs 'grease'  [i] VS. [1]  %.is 'paper'  tis  [i] vs [u]  cis 'mitts'  cus 'fly'  [i] vs. [a]  si T  sa 'sun, watch'  ci 'wind'  ca 'hat'  [j] vs. [u]  dj 'four'  du 'today'  [j] vs. [a]  tj 'dog'  i a 'many'  dj 'four'  da 'three'  si6e 'my skin'  siSa 'my mouth'  [e] vs. [a] [e] vs. [a]  sine 'my face'  sina 'my eye'  [e] vs. [ay]  t'6e 'sinew'  t'6ay 'bowl'  [e] vs. [ay]  0e 'ice'  Gay 'snow'  [e] vs. [a]  leh 'no'  lah 'work'  beke 'on him/her'  beka 'with or beside him/her'  [u] vs. [a]  yd 'clothes'  ya 'under'  [u] vs. [a]  du 'today'  da 'three'  [o] vs. [a]  co 'big'  ca 'rain'  Examples of each phoneme will be given in the following subsections. 2.6.3  High front vowel [i]  Examples of [i] and [j] are shown in (19). (19)  i  i  si 'I/me'  fj 'dog'  tili 'pail/box/pot'  -cj 'nose'  didzi 'dirty/filthy'  dj 'four'  2.6.4 M i d front vowel [e] Examples of [e] and [ej are shown in (20). (20)  e  S  yake 'window/mirror'  dz? 'gum'  nake 'two'  6? 'ice'  Veyi 'that'  har^s?j Twant'  A s to the status of schwa, I concur with L i (1946) that it is an allophone of [e], although it is difficult to prove conclusively since no minimal pairs contrasting schwa with other vowels can be found. Schwa occurs rarely in my data. Examples of [a] from the research dialect and equivalent forms from other sources are shown in (21).  19  (21)  /  n  /  r  Distribution of schwa Research dialect  Other source  san 'song/music'  san 'song' (Li 1933:146)  nan 'you'  nan 'you' (Li 1933:133)  delzan 'black'  -zan 'to be black' (Li 1933:142)  -cana 'grandmother'  -cun-e 'grandmother' (Li 1933:145)  bar 'meat'  bar 'meat' (Li 1933:127)  Ggoare 'death'  -Siy-e 'death' (Li 1933:139)  na&are 'staying'  na6er 'he is staying' (Li 1946:416)  te6are 'night'  taS-e 'night' (Li 1933:131)  ?ejare 'fear'  -jer 'there is fear' (Li 1933:147)  harajie 'spider'  horadzi 'spider' (Li 1933:143)  har?s?j 'I want' hores?} 'want (it)' lp. imp. (Elford & Elford 1981:176) /_!  /  salah 'five' '2 ?osdt'ale 'fine/okay'  sasulaye 'five' (Li 1946:422)  sasdn 'metal'  cacane 'metal' (Elford & Elford 1981:15)  A s can be seen from the data, the distribution of schwa in the research dialect is consistent with that of L i (1933, 1946) in several forms. However, it also occurs in place of [u], [i], [e], [e], [o] and [a]. Note, importantly, that schwa is never high tone. A l l other vowel qualities except for [i] are contrastive under high tone. Shaw (1998) has postulated that schwa is a non-moraic vowel; as such it would be incapable of bearing tone. The findings here appear to confirm this hypothesis. The second factor to note in the distribution of schwa is that, except for the last three forms, it occurs before [n] and [r]. Other non-high tone vowels which can occur before [n] and [r] are shown in (22). (22)  Vowels / _ n, r [e] M [u] [a]  dzen  'muskrat'  ?ekina8are  'fight/wrestle'  kun  'fire'  det'an  'goose'  Schwa does not contrast in the environment before [n] and [r] with [i], [e] or [o]. It is therefore possible that it may be an allophone of one of these vowels, or perhaps the reduced form of all three. I will tentatively conclude that it is an allophone of [e], as also suggested by L i (1933, 1946). Further research may prove whether or not this is the case.  1 2  In careful speech, 'five' is [solah] therefore schwa reduction here may be due to English influence.  20  2.6.5 M i d front vowel [e] The mid front vowel is by far the most commonly occurring vowel in my data. Examples of [e] and [ej are shown in (23). (23)  e  £  dene 'people'  kqi  dzen 'muskrat/rat'  -sik e 'children'  ?edu 'hot'  ?? 'yes'  'house' w  It is clear that in this dialect, [e] is not an allophone of [e], as suggested by Cook (1983) as can be seen from the following minimal pair:  ya-ke 'window' (sky-?)  vs.  be-ke 'on him/her'. (3p prefix-on)  13  Now we must consider the status of [i]. [i] occurs very rarely in my data, but it is clearly not an allophone of [i] as might be expected, as shown by the following minimal pair, repeated from (18):  ^is 'paper' vs. tis  'grease'.  Examples of [i] from the research dialect and equivalent forms from L i (1933, 1946) are shown in (24). (24)  Distribution of [L] Research dialect  L i (1933. 1946)  si- 'my'  se- 'my' (Li 1946:402)  ni—  ne- 'your' (Li 1946:402)  ne- 'your'  bi- ~ be- 'his/her/its'  be- 'his/her/its' (Li 1946:402)  ni- (adjectival prefix)  ne- (adjectival prefix) (Li 1946:415)  nalts 'urine'  lez 'urine' (Li 1933:150)  bis 'knife'  bes 'knife' (Li 1933:127)  tili 'pail/box/pot'  tali 'kettle/pot' (Li 1933:131)  descm 'bush'  decin 'wood/stick/tree' (Li 1933:148)  -cileh 'younger brother'  -cile 'younger brother' (Li 1933:148)  ?ileh 'no'  ?lile~ Vjhile'no' (Li 1946:422)  A s can be seen from the data in (24), [i] occurs most often in place of [e], although it also occurs in place of [e], [a] and [i], so it may be the result of neutralisation. Note that [i] and [e] often alternate in the second and third person singular possessive prefixes. A n d like schwa, it never seems to occur under high tone. There does not appear to be a consistent allophonic pattern, although there may be some relation with sibilants since si- is always si- (never se-)  and ni- and bi- usually, but not  consistently, precede sibilants.  My consultant says that 'window' derives from 'sky' (cf. yah 'sky') but is unsure of the meaning of Re.  21  I will tentatively conclude that [i] is an allophone of [e], in free variation with [e]. 2.6.6 High back vowel [u] Examples of [u] and [u] are shown in (25).  (25)  u  u  txi 'rope'  mzu 'it's good'  ku 'now'  kun 'fire'  tu 'water' 2.6.7  M i d back vowel [o]  Examples of [o] and [Q] are shown in (26).  (26)  o toh  o 'grass'  kqi  Jole 'ball' -co 'big' 2.6.8 L o w back vowel [a] Examples of [a] and [a] are shown in (27).  (27)  a  a.  gah 'rabbit'  ca 'rain'  sas 'bear'  tabi 'net'  lah 'work'  dac4 'raven/crow'  2.6.9 Diphthong [ay] Examples of [ay] are shown in (28). There is no nasalised counterpart, and in the data collected so far, [ai] only occurs with high tone.  (28)  ay. feay 'bowl' 6ay 'snow' kay 'willow'  2.6.10 Tone Tone is phonemic in Chipewyan, with high (marked) tone and low tone unmarked. The phonemic status of tone is evidenced by the following minimal pairs, shown in (29).  22  (29)  Tone - minimal pairs si-la 'my cousin' (RD)  si-la 'my hand'  si-zi 'my chest' (RD)  si-zi 'my name'  ko6 'clouds'  si-k66 'my neck'  tuwe 'fish'  tuwe 'fish (possessed)'  2.6.11 Long vowels A s to the status of long vowels, I concur with Cook (1983) that long vowels are either the coalescence of two identical vowels, or in other cases, they are predictable by the environment, with vowels lengthened in open syllables, particularly in final position or when followed by a voiced fricative. A n illustrative example can be seen by comparing Figures 48 [ia] 'many' and 49 [lah] 'work' (chapter 3, section 3.4.3). The vowel [a] is much longer in the open syllable in Figure 48 than in the syllable closed by the voiceless fricative [h] in Figure 49. Because this variation is length is predictable, I agree with Cook (1983) that it need not be marked. This concludes the presentation of the vowel inventory. The next section will provide a brief morphological overview of Chipewyan. 2.7 Morphological overview There are three major word classes in Chipewyan referred to by L i (1946): the noun, the verb and the particle. The position adopted in most of the recent Athapaskan literature is that nouns, verbs and postpositions constitute the three major word classes of an Athapaskan language. (See Cook and Rice 1989.) L i (1946) includes postpositions as part of the noun class. Athapaskan morphology is heavily prefixing. Words consist of a stem (root plus suffix, if any) which is generally monosyllabic, and to which several prefixes may be added. Nouns can be inflected for possessor, which may cause additional phonological changes to the noun stem. Particles rarely change form, and include independent pronouns, numerals and adverbials. The Chipewyan verb consists of the verb stem and ten prefix positions; these prefixes serving derivational and inflectional functions as seen in the template shown in (30).  23  (30)  Chipewyan verb template (adapted from L i 1946:410) disjunct  pronominal  prefixes  subj & obj  prefixes  1-2-3-4-  5  7-8-9-10-  -  6  conjunct  14  stem stem  1: incorporated postpositions with their pronominal objects 2: local and adverbial prefixes 3: iterative prefix 4: incorporated noun stems 5: third person pronominal subjects 6: pronominal objects 7: modal prefixes 8: aspectival prefixes 9: pronominal subjects 10: classifiers Note that the verb template presented for Chipewyan does not differ significantly from those posited for other Athapaskan languages. A n interesting fact about the structure of the verb is that consonant distribution is subject to morpheme position in the verb. A l l consonants in the inventory may occur within the disjunct prefixes, but the conjunct prefixes draw from a limited set of the inventory. In the conjunct prefixes, there are no aspirated (with the exception of [t]) or glottalised stops, and no affricates. (Li 1933 quoted i n K a r i 1976:42). The syllable structure is also restricted by the verb morphology. The syllable structure of Chipewyan (and most Athapaskan languages) is considered to be C V ( C ) . In the verb, a coda consonant will only surface stem-finally or in the syllable preceding the verb stem (position 9 or 10). A l l of the other syllables of the verb have a C V - syllable structure. We will see in subsequent chapters how the morphology interacts with the phonology in phonological processes. In chapter four, we will see that laryngeal neutralisation in Koyukon is restricted to a certain morphological domain: the stem-final position. In chapter five, we will see that the continuant voicing process occurs when the possessive prefix morpheme introduces a floating voice feature which causes the stem-initial continuant to voice. In chapter six, we will see two types of tone assimilation. In Navajo, the process of tone assimilation is restricted by both prosodic structure and morphological domain. In Chipewyan, tone assimilation occurs when the high tone of the possessive suffix morpheme spreads to a preceding vowel. This concludes a brief sketch of the important points of Chipewyan morphology. The final section will provide the chapter summary. The template model is the traditional model used to describe the verb in Athapaskan literature. McDonough (1990) has proposed a bipartite model of the verb for Navajo. This proposal will be discussed in chapter six. 1 4  24  2.8 Conclusion This chapter has summarised a variety of historical, comparative and theoretical issues in an attempt to provide the essential background information relevant to issues presented in this thesis. First, a review of the Proto-Athapaskan linguistic literature has provided the historical context in which to situate Chipewyan. Issues raised in the literature become the focus of subsequent chapters. The Chipewyan consonant inventory, introduced in this chapter, has been shown to be different from the posited Proto-Athapaskan inventory in several ways. We will focus on the consonant inventory and the featural specification of the consonants in greater detail in chapter three. Other literature relevant to this topic is that of Rice (1994), which we have reviewed in this chapter. Issues raised by Rice (1994) will also be relevant to the discussion of laryngeal neutralisation in chapter four, and continuant voicing in chapter five. Key points of the historical literature will also provide a starting point for the processes of laryngeal neutralisation and spirantisation in chapter four. The two-way laryngeal contrast stem-finally in Proto-Athapaskan has been neutralised in some of the daughter languages, while other languages have lost all stem-final stops. A s we have seen, loss of stem-final laryngealisation is also closely related to the development of tone in Athapaskan languages, and we will discuss tone assimilation processes in chapter six. To complete the overview, the vowel inventory of the research dialect Chipewyan has been described, comparing it to previous inventories documented in the literature. A s well, we have briefly reviewed the basic structure of Chipewyan morphology, discussing the role played by morphology in the phonological processes discussed in the remaining chapters.  25  Chapter 3 Chipewyan Consonant Inventory: Phonological and Acoustic Evidence 3.1  Introduction  This chapter will focus on the consonant inventory of Chipewyan and will provide new acoustic analysis of the obstruents which will show several interesting discoveries, including voicing of some of the stops traditionally classified as voiceless unaspirated, merger of the interdental stops with the interdental fricatives, merger of the alveolar stops with the palatal stops and clarification of the status of the fricatives [1] and [y]. These findings will make a contribution to our understanding of these issues from the perspective of comparative Athapaskan phonology. The key research question of this chapter concerns the laryngeal properties of the voiceless unaspirated (plain) stops. A s shown in the consonant inventory in chapter two, Chipewyan is a typical Athapaskan language with three different types of stops and affricates: voiceless unaspirated, voiceless aspirated and glottalised. I assume, following Shaw (1991) that stops and affricates are both specified for the feature [stop]. Therefore, stops are marked for manner of articulation whereas fricatives are not. I will refer to the affricates as stops unless it is necessary to distinguish them. Some Northern Athapaskan languages have been reported to have voiced, rather than voiceless unaspirated stops. These voiced stops may occur in all environments, as reported for Tahltan (Shaw 1981-1983, Nater 1989, Bob 1999) or only in certain environments. In Slave for example, the unaspirated, or plain, consonants are reported to be voiceless or weakly voiced, and are often weakly voiced intervocalically (Rice 1989:31). Wave form and spectrogram analysis of the plain stops show that in the research dialect, the bilabial and dental stops are voiced rather than voiceless unaspirated. The alveolar and palatal affricates are shown to be intermediate cases which are not consistently voiced or voiceless throughout the stop duration. The lateral affricate, velar stop and glottal stop are shown to be unambiguously voiceless. The results show that there is a front-back asymmetry among the plain stops, with plain stops near the front of the mouth being voiced, plain stops in the middle being intermediate (i.e. partially voiced) cases, and plain stops near the back of the mouth being voiceless unaspirated. In other words, there is a cline of voicing from front to back, which can be explained by principles of aerodynamics. These findings make an interesting claim regarding the interface of phonetics and phonology. Stops which behave as a class phonologically are shown to have different phonetic realisations, showing that the phonology is in some respects independent of phonetics. Wave forms and spectrograms for the aspirated and glottalised stops were also done, and show that they are indeed aspirated and glottalised respectively in keeping with traditional descriptions. The second interesting fact resulting from the acoustic analysis concerns the interdental series of stops. In the Fort Resolution dialect of Chipewyan, Rice (1978) reports the merger of the interdental stops with the dental stops. In the research dialect, the interdental stops (excluding  26  the glottalised stop) were perceived to be fricatives; this is confirmed by the spectrogram analysis. A more restricted shift reported by Cook (1991) is the merger of two coronal affricate series, with alveolar becoming palatal. This change is evidenced in my data, but this merger appears to be incomplete, as forms with alveolar stops still occur. There are several other differences in the inventory of the dialect of Chipewyan presented in this thesis compared to the dialect documented by L i (1946), presented in chapter.two. First, the voiced velar fricative [y] is often replaced by another phoneme ([y], [w], [g] or [h]) e.g. - yii 'tooth' (Li 1933:135) > ~-wuh documented is [k ], viz. w  or ~ -gu  (RD). Also, the only labiovelar attestation yet  sisik e 'my children'. w  The final issue raised by the acoustic evidence in this chapter concerns the status of the phonemes [1] and [y] which have traditionally been classified with the voiced continuants rather than the sonorants. Acoustic evidence will show [1] to exhibit properties characteristic of fricatives, while [y] exhibits properties more typical of sonorants. Argumentation provided by the acoustic data in this chapter will be used to substantiate a featural specification of the obstruents of Chipewyan. Section 3.2 will review the only previous acoustic study done on Chipewyan, that of Hogan (1976). Section 3.3 will present each series of stop/affricate, section 3.4 will present each series of fricative, and section 3.5 will present the sonorants. Wave forms and spectrograms (contained in the appendix) will be shown for each stop in initial and intervocalic position, and for some of the fricatives. Section 3.6 will conclude. 3.2 Previous acoustic results (Hogan 1976) Hogan's (1976) study appears to be only published study of acoustic data in Chipewyan. The goal of the study was to determine some temporal features of ejective consonants and to examine the characterisation of voice onset time. Glottalic ejectives are produced by glottal closure, raising the larynx to provide a high supraglottal air pressure and a release producing a transient or burst. The study used a sample of 256 words and focused on the stops III (64 tokens), /k7 (56 tokens), and the affricates /t'6/ (12 tokens), lei (84 tokens) and ft/ (40 tokens). Hogan measured five different temporal events in the production of ejectives, which occur in the following sequence: (i) prerelease silence, (ii) the burst, (iii) the silent period between burst and onset of the vowel, (iv) the presence or absence of F2 and F3 transitions, and (v) the fundamental frequency of the onset of the vowel. A l l ejectives exhibit prerelease silence intervocalically. For the intervocalic tokens, the average duration and standard deviation of the prerelease silence were 69.10 and 22.59 msec respectively. The duration of pre-release silence between dental and velar ejectives was not significant, so the increase in volume of the cavity between the glottis and point of closure for the stops is minimal from velar to dental articulation. However, there was a significant  27  difference between the pre-silence durations of ejectives before high and low vowels, which suggests that the dimensions of the pharyngeal cavity which is larger in the production of high front vowels plays an important role in the time required to build up superglottal pressure. Hogan does not mention the relevance of the size of the oral cavity; we will see in section 3.6 that this also plays an important role in the production of ejectives and voiced stops The burst or transient is the result of the air pressure behind the point of closure being released. The measurements of the peak of energy are as follows:  dental ejectives, 3000 Hz; velar  ejectives, from 2000 to 1000 H z depending on whether the stop was before a front or back vowel. The average duration of the transients were: dental ejectives, 25.33 msec (sd 12.07 msec); velar ejectives, 31.25 msec (12.07 sd msec). The affricates begin with an abrupt onset followed by a noise spectrum beginning at: 3000 H z (alveolar affricate) and 6000 Hz (interdental affricate). The noise spectrum for laterally released ejectives had energy in the range from 80-2500 H z , and a peak at 4000 Hz. The average durations are 112.82 msec for Icl and 71.58 msec for/X7. The vowel formants begin abruptly in most cases and only 24% of the tokens showed any evidence of F2 or F3 transitions, which indicates that the vocal tract assumes the configuration for the following vowel in the interval from the release to the onset of phonation. 30% of the vowels begin with a very low fundamental frequency for two to four glottal pulses. The vowel onset resembles creaky voice and is more evident in vowels with low tones. Several different comparisons of the results were made. Mean durations between release and vowel onset varied significantly depending on whether the ejective was a stop or affricate (49.34 msec longer for affricates than for stops), but did not differ significantly between the other groups. Variance based on measurements of the silent interval after the end of the transient or fricative noise was significant between dentals and velars, and between stops and affricates (with the silent interval 37.70 msec longer for the stops than the affricates). Hogan explains that the differences of duration from release to vowel onset between stop and affricates is due to the fact that the air pressure drops much more gradually when the area of constriction is shaped as in the affricates. The reverse difference in the analysis of the post-release silent interval is due to the fact that there has to be a certain amount of time for realisation of the fricative noise. The remainder of the chapter will present new acoustic data from the research dialect of Chipewyan. 3.3 Obstruents: stops and affricates In this section, examples will be given of each stop in initial and intervocalic position. For the acoustic analysis, wave forms and spectrograms were done for two words in each position, with two different tokens of each word. I will generally only show one example illustrative of each stop (see appendix), unless the results are ambiguous or questionable, in which case I will show two examples. The analysis is meant to provide a general description of the acoustic properties of phonemes in the research dialect of Chipewyan; it is not meant to be a scientific  28  survey. A sample including more speakers, as well as more tokens, would be necessary to provide an uncontroversial scientific analysis. The main question to be answered in this section is whether the plain stops are voiced or voiceless unaspirated. There are several indicators of voicing. The main indicators are a periodic wave in the wave form, and a dark "voice bar" on the bottom of the spectrogram. In addition, voiced stops affect the formants of the following vowels. Following a voiced bilabial, there is a rapid increase in all three formant frequencies. Following a voiced [d], there is also an increase in the first formant frequency, but both the second and third formants have a slight fall in frequency. Following a voiced velar stop, there is a narrowing of the distance between the second and third formants. (Ladefoged 1975). 3.3.1  Bilabial stop  Like all Athapaskan languages, Chipewyan has an impoverished series of bilabials with only one stop, [b], traditionally described as voiceless unaspirated. Proto-Athapaskan (PA) is not reconstructed with any bilabial stops; bilabials in the daughter languages, including Chipewyan [b], are thought to be reflexes of the P A sonorant *w (Krauss and Leer 1981). Examples of [b] are provided in (1). (1)  Bilabial stop Initial  Intervocalic  bit'as 'outside'  tabi 'net'  bis 'knife'  tab^le 'tent'  bar 'meat'  si bar 'my stomach'  Waveforms and spectrograms of [b] are shown in Figures 2 and 3 .  15  Figure 2 shows [b] in initial position. Although it is difficult to tell because of the short voice onset time ( V O T ) , the wave form immediately before the vowel appears to be periodic, there is a voice bar at the bottom of the spectrogram, and the formant transition into the following vowel is quite sharp. These are all indications of a voiced stop. This is much clearer in the intervocalic token shown in Figure 3. The wave form is periodic and there is a voice bar, although the voicing diminishes throughout the duration of the stop. It appears that [b] in this dialect of Chipewyan is voiced, rather than voiceless unaspirated. This is a crucial difference, since previous descriptions of Chipewyan have considered the laryngeal properties of [b] to be in parallel with the other voiceless unaspirated stops. However, if [b] did derive from the voiced P A sonorant *w, perhaps it is not unexpected that it would also be voiced in the daughter language. On the other hand, if the shift from sonorant to obstruent were complete, one might expect voicelessness.  Recordings were made using a Marantz tape recorder and the computerised acoustic analysis program Signalyze was used to produce the wave forms and spectrograms. 1 5  29  3.3.2 Dental stops Chipewyan has a full series of dental stops, although there are no dental fricatives. This is identical to the Proto-Athapaskan reconstruction (Leer 1979), with the exception that some instances of Proto-Athapaskan * d become [r] intervocalically, which will be discussed in section 3.5. Examples of the dental stops are provided in (2). (2)  Dental stops Initial  Intervocalic  [d]  dac4 'crow, raven'  nedah 'it's heavy'  [t]  [fj  dene 'people'  sideze 'my sister'  duwe 'difficult'  ?edu 'hot'  tab! 'net'  yati 'language'  tede 'bed'  keti 'eight'  tu 'water'  jietue 'juice'  t'a 'feather'(Li 1933:131)  det'anco 'eagle'  teO 'coal'  Xat'i 'whose'  ture 'soft, slimy' (Li 1933:132)  ?itu 'jacket'  One of the significant phonological changes that has occurred in several dialects of Chipewyan since Li's first documentation, is the so-called t-k shift. (See Haas (1968) on the Yellowknife dialect and Rice (1978) on the Fort Resolution dialect.)  In these dialects, [t] and [k] have  merged, with [k] replacing [t] everywhere. In other words, the reflexes o f P A *t and *k have merged. There is no evidence of this shift in the dialect presented in this thesis. Some examples of the t-k shift (from Haas 1968:166) along with forms from the research dialect are shown in (3). t>k  shift  16  CC  YC  Research dialect  Gloss  taye  kaye  da  'three'  tan  kan  ten  'ice'  tali  kili  tili  'kettle*  seta  seka  sita  'my father'  yatai  yakiye  yati  'language, word'  tSakjr  —  'needle'  t6atjii  Waveforms and spectrograms of the dental stops are shown in Figures 4-9. Figures 4-6 compare the dental stops in initial position, each followed by the same vowel [e]. Figure 4 shows periodicity of the wave form, a voice bar in the spectrogram, and a sharp  YC indicates Yellowknife Chipewyan, spoken in Yellowknife, NWT, and CC refers to the Fort Chipewyan dialect recorded by Li (1933). 1 6  30  transition to the vowel, which indicates that [d], like the bilabial stop [b], is voiced rather than voiceless unaspirated. Because of this result, this makes the argument for sonorance as the historical source of voicing unlikely, since [d] did not derive from a sonorant. Figure 5 shows the voiceless aspirated stop [t]. The release of the aspirated stop is marked by a sharp onset of noise that appears as 'fuzziness' in the upper frequencies, similar to that produced by a fricative, [t] is clearly voiceless since the wave form is almost flat, and there is no voice bar in the spectrogram. The formant transitions are also much less marked compared to Figure 4. Finally, the glottalised stop [£] is shown in Figure 6. The ejective release is shown at around 5200 ms, followed by a slight bit of aspiration. Comparing Figures 5 and 6, the voice onset time of [t] is much longer than that of [t], which is expected of a glottalised consonant. Figures 7-9 compares the dental stops in intervocalic position. Intervocalically, [d] is also voiced, rather than voiceless unaspirated, as can be seen by comparing the wave form of Figure 7 with the wave form of the voiceless aspirated [t] in Figure 8. If the voicing were due to carryover from the preceding vowel, one would also expect to see a similar amount of voicing carryover in the voiceless aspirated token. This is not the case. Figure 9 shows [t] intervocalically; note that the V O T is much shorter than for [(] in initial position (Figure 6). 3.3.3 Interdental stops The interdental series is less common across the Athapaskan family, occurring only in some of the northern languages. Proto-Athapaskan is not reconstructed with this series; it developed from the P A * D Z series (Leer 1979). The plain stop [d6] occurs very rarely. I have no tokens of this phoneme in word-initial position, neither in my data nor in any other source. It is, however, found word-internally.  The aspirated and glottalised stops are more prevalent.  Examples of each are provided in (4). (4)  Interdental stops Initial  [dS]  none  Intervocalic ?ed6ai 'smoked/dried fish' nid6en 'think (lp dual imperfective)' (Elford&Elford 1981:162) danidSen 'think (lp pi. imperfective)' (Elford&Elford 1981:162)  [t0]  [t'0]  t64i 'star'  ?et6en 'caribou'  t6e 'stone/rock'  sitGi 'my head'  t6i?ah 'pillow'  sitGiyah 'my hair'  t'Gai 'bowl'  set'Gi 'my flesh' (Li 1933:141)  t'6e 'sinew'  sit'Gene 'my bone'  t'Ouwe 'milk'  bit'SuwE 'her breast'  Waveforms and spectrograms are shown in Figures 10-16. A s mentioned in section 3.1, Rice (1978) has reported the merger of the interdental stops with the dental stops in the Fort Resolution dialect of Chipewyan, as shown in (5) (Rice 1978:145).  31  (5)  Interdental > dental merger 17 Gloss 'stone, rock' 'star' 'head' eten  ?et6sn  'caribou'  This type of phoneme simplification appears to be happening to some extent in this dialect as well. However, rather than merging with the dental series, the stop is being replaced by its corresponding fricative. A n interesting parallel can be drawn from the accommodations made for English interdental fricatives by native French speakers. While speakers from Quebec tend to replace the interdental fricatives ([6/6]) by alveolar stops ([t/d]), speakers from France typically replace the interdentals with alveolar fricatives ([s/z]). Evidence of this merger can be seen from the two tokens of initial [t6] shown in Figures 10 and 11. There does not appear to be a clear stop release (compare [t] in Figure 5), only the random noise pattern characteristic of fricatives. Figures 12 and 13 show two tokens of the glottalised affricate [t'6]. In this case, both the ejective release and fricative fuzziness can be seen in each token, which shows that it is an affricate, and has not merged with a fricative or the glottalised dental stop. (Compare [(] in Figure 6.) Figure 14 shows [dS] intervocalically. Like the plain stops [b] and [d], [dS] is voiced rather than voiceless unaspirated, as seen by the periodic wave form and the voice bar. Furthermore, it has the random noise pattern of a fricative throughout, which seems to suggest that it has merged with the voiced fricative [6]. Intervocalic [t6] (Figure 15) also appears to be a fricative. Intervocalic [t'6] (Figure 16) is still glottalised, although with little frication. To my own perception, the voiceless aspirated stop sounds like a fricative, particularly in initial position. To summarise, it appears from spectrogram analysis that the voiceless unaspirated stop [d6] has merged with the voiced fricative [6]. The voiceless aspirated stop [t6] has merged with the voiceless fricative [6], while the glottalised stop [t'0] clearly maintains its status as a separate phoneme. This merger is an example of lenition, a common process cross-linguistically. This conclusion raises the question as to whether these fricatives derived from stops will pattern with the other continuants in phonological processes. I hypothesise that they do not pattern with the continuants. If they did, one would expect continuant voicing to occur in stem-initial position of nouns when the possessive prefix is added. (See chapter five for discussion of the continuant voicing process.) A s can be seen from the following two examples, shown in (6), continuant voicing does not occur.  RC indicates Resolution Chipewyan, spoken in Fort Resolution, NWT, and CC refers to the Fort Chipewyan dialect recorded by Li (1933). In Resolution Chipewyan, the voiceless unaspirated stop [d&] does not merge with the voiceless unaspirated [d]. In this dialect, all glottalised stops are lost, so the glottalised interdental [t'6] becomes [d6]. Also note that this dialect exhibits the t-k shift discussed in section 3.3.2, so [t] in this dialect only has [t0] as its source since the reflexes of PA *t have merged with *k (Rice 1978). 1 7  1 8  32  (6)  Noun stem  [-tei] l-QV  Possessed form 'head'  [-tGi-yah] /-Oi-yabV 'hair'  [sL-tSi] /si-ei/ 'my head' [si-t6i-yah] /si-6i-yah/ 'my hair  3.3.4 Alveolar stops The * D Z series is reconstructed in Proto-Athapaskan; for those languages such as Chipewyan where P A * D Z became interdentals, the alveolar D Z series is hypothesised to have developed from the P A * D z (J) series (Leer 1979). A s mentioned in section 3.1, a consonant merger is in progress in this dialect where some alveolar forms have become palatal (DZ < D z (j)), i.e. essentially reversing the historical change. This change is evidenced in my data, as shown in (7). (7)  Alveolar > palatal merger Alveolar  Alveolar > Palatal  dzole 'ball' (Li 1933:144) horadzi (Li 1933:143) ca 'beaver' (Li 1933:144) ?e<tcuzi 'trap' (Haas 1968:175) celt'uy 'tobacco' (Haas 1968:175) -cazE'kidney' (Li 1933:145) -cuz 'elbow' (Li 1933:145) cuze 'blow-fly' (Li 1933:145)  jole 'ball' harajie 'spider' cah 'beaver' ?ecu6i 'trap' celtue 'tobacco, cigarettes' -ceSe 'kidney' 1 9  -cuze 'elbow' cuze 'fly'  Notice that the fricatives in the series are not affected by this merger, although [z] in 'trap' and 'kidney' are replaced by [6]. A n d , the merger is not complete, as forms are still found with alveolar affricates, as shown in (8). (8) [dz]  [c]  [c]  Alveolar stops Initial  Intervocalic  dzen 'muskrat/rat'  ?edzah 'cold'  dzenie 'ratroot'  sidzi 'my heart  dz? 'gum'  sidzage 'my ear' (RD)  c£ ~ 54 'excrement'  stcj my nose  c£kQ? 'outhouse'  stcage ~ stcage 'my ear'  ce6e 'blanket'  sice6e 'my blanket'  ca 'hat'  sicana 'my grandmother'  cane 'wife'  sicane 'my wife'  ceni 'friend'  siciya 'my grandfather'  Comparing forms in (7) and (8) suggests that the change from alveolar to palatalised  1 9  This form is exceptional in that it becomes glottalised as well as palatalised.  33  consonants may be lexically idiosyncratic, rather than phonologically conditioned. The shift is not specific to any one manner of articulation as all three are affected. It is not specific to syllable or morpheme structure, nor is it specific to particular surrounding vowels or tones. Contrasting forms are found in very similar environments, such as -coze > -SeSe 'kidney' but cede 'blanket'. Furthermore, some forms such as 'excrement' and 'ear' are variable. Therefore, I conclude that this is an ongoing merger which appears to be lexically idiosyncratic. Waveforms and spectrograms for the alveolars are shown in Figures 17-23. Figure 17 shows initial [dz]. It is difficult to discern voicing from the wave form. There is a light voice bar but no sudden formant transition. It appears that it is may be voiced, but not voiced throughout, as can be seen by comparing it to Figure 19 which shows that [c] is clearly voiceless and aspirated. Figure 18 shows a second token of initial [dz]. It also looks to be only weakly voiced. Figure 20 shows initial [c] with glottalisation indicated by the dark ejective spike and a brief silent period after the fricative release. Comparing Figures 21 and 22 shows that [dz] is voiced, and [c] voiceless intervocalically. Glottalisation can clearly be seen shortly before 10200 ms in Figure 23. To summarise, while the plain bilabial and dental stops were shown to be voiced, the plain alveolar stop appears to be weakly voiced, or not voiced throughout the stop duration. 3.3.5 Lateral stops The lateral series of consonants in Chipewyan is identical to that reconstructed for ProtoAthapaskan (Leer 1979). The voiceless unaspirated and voiceless aspirated stops occur very rarely in all sources. The glottalised stop is more common. Examples of each stop are shown in (9). (9)  [k]  [X]  [X]  Lateral stops Initial  Intervocalic  A4ne 'mouse'  naAdf 'smile/laugh'  A.ati 'who/how'  hoyahore^i 'there is suspicion' (Li 1933:150)  Xainu 'what'  ?eX|sj 'to where' (Li 1933:150)  Xis 'grease'  ?ey oz hoXoi 'a little brush, a mop used  sekase Xiiye 'my tonsils'  to splash water on the canoe'  (Li 1933:151)  (Li 1933:151)  Xis 'paper' Xoh 'grass' Xu 'rope'  neXah 'your buttocks' •fjXus 'leash' GeXue 'iceline'  w  Waveforms and spectrograms are shown in Figures 24-28. No examples of intervocalic [X] are shown since the one form listed by L i (1933) was not known by my consultant.  34  Figure 24 shows initial [k] to be voiceless unaspirated, as is consistent with previous descriptions. There is no voice bar or periodicity in the wave form. Random noise pattern in the upper frequencies in Figure 25 shows [X] to be aspirated, as well as voiceless. The prominent ejective spike in Figure 26 shows [X] to be glottalised. Intervocalically, in Figure 27, [X] appears to be voiceless, and only slightly aspirated, and Figure 28 shows glottalised  [t].  3.3.6 Palatal stops Since the Proto-Athapaskan palatal series developed into the alveolar series in Chipewyan (PA *Dz 0 )  >  D Z ) as discussed in section 3.3.4, the palatal series is hypothesised to have  developed from the P A front velar series (PA * G > Dz) (Leer 1979).  20  Examples of each stop  are shown in (10). (10)  Palatal stops Initial  [T]  [c]  [£]  Intervocalic  ja 'here'  ?ejere 'cow'  jie 'berry'  hsrajie 'spider'  jole 'ball'  Vejuleh 'mosquito'  ca 'rain'  dac£ 'raven/crow'  cize 'cat'  bsce 'its tail'  cus 'fly'  necileh 'it's small'  celtiie 'tobacco/cigarettes'  nacds 'kiss'  c£k aze 'girl/young woman'  ?ecu6i 'trap'  cis 'mitts'  becazj 'away from him' (Li 1946:404)  w  Waveforms and spectrograms are shown in Figures 29-34. Figures 29 and 30 show initial []]. Both tokens look to be weakly voiced due to the presence of the light voice bar. Intervocalically in Figure 33, [J] looks partially voiced, showing a voicing bar through part of the stop and some periodicity in the wave form. Specifically, the stop portion of the affricate is voiced and the fricative portion is voiceless, [c] and [t] in initial position (Figures 31 and 32) are voiceless aspirated and glottalised respectively, [c] is also voiceless aspirated intervocalically (Figure 34). I do not have any recorded tokens of intervocalic [&]. To summarise, like the plain alveolar stop, the plain palatal stop appears to be an intermediate case with partial voicing through the stop duration. A s mentioned in the introduction, a cautionary note must be made about phonemes such as those in this series, where the evidence does not clearly point to voiced or voicelessness. Further research to collect more tokens, and from more speakers, would be necessary to make a definitive conclusion.  2 0  Some languages do have velars in cognate forms, such as Hupa karj 'rain' cf. Chipewyan ca 'rain'.  35  3.3.7 Velar stops Since the Proto-Athapaskan front velar series developed into the palatal series in Chipewyan (PA * G > Dz), the velar series is hypothesised to have developed from the P A back velar or uvular series (PA * G ' > G) (Leer 1979). Examples of each stop are shown in (11). (11) [g]  [k]  [R]  Velar stops Initial  Intervocalic  gah 'rabbit'  yagole 'butterfly'  gokus 'pig'  yagodene 'angel'  gu 'bug/insect'  tGiyagu 'lice'  kazba 'ptarmigan/partridge'  hokamo 'it tastes good'  kqi  xaike 'winter boots'  'house'  kun 'fire'  yukqi  'clothing store'  kai 'willow'  ?eka 'fat'  kioe 'boat'  ?eki6e 'gun'  ko6 'clouds'  stkoG 'my neck'  Waveforms and spectrograms are shown in Figures 35-41. Figures 35 and 36 show two tokens of initial [g]. Unlike the other non-affricated stops [b] and [d], which were shown to be voiced, [g] is voiceless unaspirated (no voice bar nor periodic wave form). Note however, there is a slight amount of frication, which is consistent with Li's description of [g] often being pronounced with a [y] glide (Li 1946:398). Even intervocalically (Figure 39), [g] is voiceless unaspirated. This is normally expected in an Athapaskan language. There are no surprises with the remaining two stops either, with [k] showing aspiration (Figures 37 and 40) and [k] showing glottalisation (Figures 38 and 41). 3.3.8 Labiovelar stops The labiovelar series is of questionable phonemic status in many Athapaskan languages. In the Slave dialects, for example, labiovelars are rare, as can be seen from the consonant inventory presented in chapter two. While Pre-Proto-Athapaskan is reconstructed with a labialised uvular series * G , these merged with the non-labialised series in Proto-Athapaskan, accompanied by a W  rounding of reduced stem vowels to *v (Leer 1979). While L i (1946) lists the labiovelars as separate phonemes, he notes that they only exist before or after [o], [u], and sometimes before [a] (Li 1946:398). This suggests that their occurrence is predictable from the surrounding vowel context. The only labiovelar occurring in my data is [k ] in the following two forms: w  (12)  sik e  'children'  cek az£  'girl, young woman'  w  w  21  Cf. sezkene 'my children' (Li 1933:141) which my consultant glosses as 'my family' or 'my kids'.  36  Note that neither is followed by a rounded vowel. Waveforms and spectrograms are shown in Figures 42 and 43. Labialisation is the most widely found secondary consonantal articulation and is most common with velar obstruents and uvulars. Labial consonants are accompanied by a low second formant transition and lower second formant in adjacent vowels (Ladefoged and Maddieson 1996). In Figure 42, labialisation can be seen by the second formant transition, which begins at around 900-1000 H z and reaches steady state at just over 1400 Hz. Compare this to other tokens of [e] (Figures 4, 5, 6, 7 and others) where the second formant is between 1400 and 2300 H z with little or no transition from the preceding consonant. Figure 43 also shows a lower second formant and second formant transition. (Compare to [a] in Figures 9 or 19.) In this token, it appears much more like the [w] is part of a diphthong rather than labialisation on the consonant since voicing begins immediately after the release of the stop. A s mentioned in section 3.2, Cook (1986:3) does not recognise the labiovelar series as a separate phonemic series because 'the labial feature is interpreted as an onglide of a diphthong, e.g., [k on] = /kuon/ 'fire'." While this is a reasonable proposal, I have not found any other hw  occurrences of [ue] or [ua] diphthongs in the data. In fact, [ay] is the only attested diphthong, as discussed in section 2.6.9. On the other hand, near-minimal pairs can be found with [k ] and another velar stop [k], as w  shown in (13). (13)  sik e  'children'  sike 'on me'  cek aze  'girl, young woman'  beka 'with or beside him/her'  w  w  I will therefore conclude that [k ] is a labialised consonant, rather than a sequence of [k] w  diphthong, albeit its occurrence is very limited similar to the bilabial nasal [b]. 3.3.9 Glottal stop Glottal stop is reconstructed for Proto-Athapaskan (Leer 1979) and is arguably attested as phonemic in all Athapaskan languages. Some examples are shown in (14).  37  (14) [?]  Glottal stop Initial  Intervocalic  Post-consonantal  ?asi 'something'  si?e 'my uncle' (RD)  hsr^sVj 'I want'  ?eyi 'that'  0i?ar 'waking up'  nis?i 'I see'  ?eriel 'thunder'  t0i?ah 'pillow'  beces?a 'I am facing him' ( E l f o r d & E l f o r d 1981:71)  L i (1933, 1946) does not consider glottal stop to occur word-finally (following a vowel), and I concur with this, whereas some Athapaskan languages, such as Navajo, exhibit glottal stop in this environment. When glottal stop follows a vowel, it is often possible to see the occurrence of creaky voiced phonation at the end of the vowel (Ladefoged and Maddieson 1996:73) but this does appear to be the case in word-final vowels shown in this chapter. Waveforms and spectrograms are shown in Figures 44 and 45. In general, glottalised stops are characterised by a silence between the oral release and the beginning of voicing for a following vowel, and can be accompanied by laryngealisation of the following vowel (Ladefoged and Maddieson 1996:74). Figure 44 shows glottal stop in word-initial position. A small amount of irregular constricted voicing is shown in the wave form. Another example of glottal stop in initial position was shown in Figure 33. Figure 45 shows word-internal glottal stop in a C  V environment.  Cross-linguistically this is a relatively uncommon context. There is a clear silent period between 6800 and 6900 ms, and a brief amount of creaky voicing in the wave form preceding the vowel. A n intervocalic token of glottal stop can be seen in Figure 11. This completes the inventory of stops in Chipewyan. The next section will present the fricatives. 3.4 Fricatives This section will provide examples of the fricatives in Chipewyan. Each series discussed in section 3.3 includes a voiceless and voiced fricative, except for the labial and dental series, and the only laryngeal fricative is the voiceless [h]. Examples of vocabulary illustrating each series will be provided, showing initial, intervocalic and final occurrences. Examples of fricatives occurring pre- and post-consonantly will not be shown since they only occur in limited environments. Fricatives occur post-consonantally in verbs when the fricative is verb-stem initial and preceded by the classifiers [f] or [1] or certain (consonant-final or consonant-only) pronominal subject prefixes (when the classifier is [d] or zero). These classifiers and subject prefixes are the only fricatives that occur pre-consonantly in verbs each surfacing directly before the verb stem. A fricative may also occur pre- or post-consonantly in compound nouns. Wave forms and spectrograms of each fricative confirm that all voiceless fricatives are voiceless, and all voiced fricatives are indeed voiced. Therefore, wave forms and spectrograms will not be shown unless to illustrate features other than voicing.  38  3.4.1  Interdental fricatives  Examples of the interdental fricatives are provided in (15). (15)  Interdental fricatives Initial  Intervocalic  Final  [6]  Gai 'snow'  jeGsre 'careful'  ceG 'duck'  Ge 'ice'  yii-naGe 'next' (Li 1946:403)  VekiG 'gun'  GeG 'flour'  se-naGe 'in front of me' (Li 1946:403)  GeG 'flour'  deSai 'salt'  -yaS 'to be frightened'  naceSe 'bag/backpack/suitcase'  (Li 1933:134)  G^Ssre 'death'  -t'GaS 'to be knock-kneed'  [6]  Se 'skin' Sare 'liver'  22  23  Li(1933:141) A s discussed in section 3.2.3, the interdental voiced/voiceless unaspirated and voiceless aspirated stops appear to be merging with the interdental voiced and voiceless fricatives respectively. Compare Figures 10, 11, 14 and 15 with the true voiceless and voiced interdental fricatives shown in Figures 46 and 47. 3.4.2 Alveolar fricatives Examples of the alveolar fricatives are provided in (16). (16) [s]  [z]  Alveolar fricatives Initial  Intervocalic  Final  sa 'sun/wristwatch'  ?asi 'something'  5tis 'paper'  si 'I/me'  basi 'I want, need'  sas 'bear'  salah 'five'  Vosot'sle 'fine, okay'  cus 'fly'  sizi 'my name'  xaz 'scar' (Li 1933:134)  zi  name ' 24  tuzane 'trout'  cez 'firewood'  Xize 'paper bag'  (Li 1933:144)  3.4.3 Lateral fricatives Examples of the lateral fricatives are provided in (17).  Note that my consultant will give the forms for 'skin', 'liver' and 'name' (see section 3.3.2) in isolation with the voiced fricative in initial position, but he prefers to treat them as inalienably possessed (i.e. with the possessive prefix). Note that when forms listed are only from another source, there are no examples of the phoneme in this environment in my database. See footnote 8. 2 2  2 3  2 4  39  (17) [i]  [1]  Lateral fricatives Final  Initial  Intervocalic  i a 'many'  t'asi  (Elford and Elford 1981:10)  (Li 1933:126)  TJ 'dog'  ceii 'frog'  bet 'sheep'  (Elford and Elford 1981:10)  (Li 1933:127)  tuwe 'fish'  t6ati<fi 'sewing needle'  xrt 'darkness'  (Elford and Elford 1981:17)  (Li 1933:135)  lah 'work'  jole 'ball'  bXgal 'fork'  lidi 'tea'  nalis 'urine'  ?eriel 'thunder'  tulu 'path/road'  -del 'blood'  tobah 'socks'  25  4hslderi 'garbage'  ?et 'beaver dam'  (RD)  In the Athapaskan languages, [1] is treated as a voiced fricative rather than a sonorant since it patterns phonologically like the other fricatives. It alternates with the voiceless fricative [i] in phonological processes such as continuant voicing. (See chapter five.) It is therefore considered to be [+continuanf].  26  A voiced lateral approximant is characterised acoustically by well-defined formant-like resonances (Ladefoged and Maddieson 1996:193) whereas fricatives are characterised by random noise patterns in the higher frequencies. Wave forms and spectrograms of [T] and [1] are shown in Figures 48 and 49. It can be seen that [1] has no distinct formant structure, and there is a small amount o f fuzziness in the upper frequencies characteristic o f a fricative, so the acoustic phonetic evidence does seem to corroborate the phonological evidence for classifying Chipewyan [1] as a fricative rather than a lateral approximant. 3.4.4 (Alveo-)Palatal fricatives The palatal fricative [y] is usually classified as a voiced fricative (Li 1933, 1946; and Rice 1989 for Slave) because it alternates with the alveopalatal fricative [s] in phonological processes such as continuant voicing where [s] becomes [y] in the stem-initial position of nouns when the possessive prefix is added. (See chapter five.) Examples of the alveopalatal fricatives are provided in (18).  [y] does not occur in final position except as part of the diphthong [ay].  Note that 'tea' and 'socks' are both loan words from French i.e. [late] 'the tea' and [leba] 'the socks'. Kaisse (1999) disusses cases where [1] can be considered [-continuant]. However, in languages such as those in the Athapaskan family where obstruent lateral affricates are opposed to obstruent lateral fricatives, the fricatives are [+continuant] (Kaisse 1999). 2 5  2 6  40  (18) [3]  Alveopalatal fricatives Initial  Intervocalic  Final  san 'music/song'  sason 'metal/iron'  nates  s?lai 'hills'  hisu '(you) grab (imperative)'  'dream/vision' nit'us 'hit/punch/  sj0 'wart' (Li 1933:147)  bang/knock' [y]  yah 'sky'  tOiya 'hair'  yati 'language'  ?eyi 'that*  yii 'cloth/clothes'  deneyu 'man'  However, spectrograms of [y] both initially and intervocalically show [y] to be much more approximant-like with clear formant structures, as shown in Figures 50 and 51. Therefore, it seems more accurate to classify [y] with the sonorants rather than with the voiced continuants. 3.4.5 Velar fricatives Examples of the velar fricatives are provided in (19). (19) [x]  [y]  Velar fricatives Initial  Intervocalic  Final  xax 'playing cards'  ?et'axa 'suddenly,  xax 'playing cards'  xay 'winter'  all of a sudden'  xa6 'boil' (Li 1933:134)  (Li 1933:134)  none  ?eyeges 'eggs'  soy 'dampness, moisture'  -caye ~ -cage 'ear'  (Li 1933:143)  ?ey£nete 'hurry the hell up'  toy  'grass' (Li 1933:151)  The voiced velar fricative is not found in any forms in initial position in my data nor in any other source.  In the research dialect, it is often being replaced by another consonant. The  voiced velar fricative is also variable in other Athapaskan languages. For example, in Slave, it may be labialised before a round vowel or pronounced [w] (Rice 1989:32). In Bearlake, it can also be pronounced [y] before a front vowel (Rice 1989:33). Examples of the Chipewyan alternations are shown in (20). (20)  Voiced velar fricative alternations Research dialect  Other source  -yah 'hair'  -ya 'hair' (Li 1933:134)  -wuh 'tooth'  -yu 'tooth' (Li 1933:135)  -gu 'tooth' (RD) -caye ~-cage 'ear'  -dzay-a 'ear' (Li 1933:143)  Xoh 'grass'  toY  'grass' (Li 1933:151)  A s shown in (20), [y] may be replaced by [y], [w], [g] or [h] in the research dialect. The alternations occur in the following contexts: y/ hi  a, w/  u, g/  u, g/V  V (sometimes), and  #. While there do seem to be some patterns emerging, the data set is too small to conclude  41  whether these alternations are phonologically predictable or not. It is clear that [y] is most often retained when it is intervocalic. 3.4.6 Laryngeal fricative Examples of the laryngeal fricative [h] are provided in (21). (21)  Laryngeal fricative Initial  Intervocalic  [h]  haliyij 'everybody/all'  Xaha 'why'  harajie 'spider'  hjheslar 'I am falling asleep' (Li 1946:415)  holca 'greywavie'  hehe6del 'they have started' (Li 1946:416)  I class [h] with the fricatives following traditional descriptions such as L i (1946). Spectrograms of [h] show it to be characteristic of a fricative, with random noise patterns. Often word-final [h] is very difficult to hear, but my consultant has very clear intuitions as to whether or not an [h] is present, which always turn out to be confirmed by spectrograms. Examples of word-final [h] were seen in Figures 36, 49 and 51. This concludes the inventory of fricatives. The next section will present the final group of consonants, the sonorants. 3.5 Sonorants Krauss and Leer (1981) have reconstructed the following sonorants in Proto-Athapaskan: *w, * n , * y and *rj. Stem-initially, P A *w has tended to become occlusive, becoming either a nasal sonorant [m] or an oral obstruent [b]. In Chipewyan, P A *w has become [b]. P A * n has remained [n] in Chipewyan. P A *y has mostly remained [y] in the daughter languages. Stem-finally, *w and * y are usually lost. Pre-Proto-Athapaskan * n is replaced by nasalisation of the vowel before an obstruent suffix. P A * n has mostly merged with * n (Krauss and Leer 1981). The bilabial nasal is of questionable phonemic status in Chipewyan. L i (1946:399) states that it only occurs once as the initial of the stem homa' it stinks', [m] occurs in three different lexical  items in my data: mah 'stink', ?amah 'shame' and sqmba 'money' (cf. L i (1933:144) tsqba 'money'). The distribution of the alveolar nasal [n], is shown in (22). (22) [n]  Sonorant [n] Initial  Intervocalic  Final  nene 'long'  dene 'people'  descen 'bush/tree'  ni 'earth/land'  deniye 'moose'  san 'music/song'  nuh 'island'  nuni 'wolf  kun 'fire/firewood'  42  A s mentioned in section 3.3.2, Chipewyan [r] developed from some instances of P A * d . The articulation of the sonorant [r] is palato-dorsal (like English) rather than L i ' s description of Ixl as a tongue tip trill, [r] is often an allophonic variation of [d] intervocalically, e.g.  desj 'I say'  but ?ekwaresj'l say that, say so' (Li 1946:400). Cook (1986:3) questions the phonemic status of [r] entirely, since "The syllable final dhas become rotacized in all communities except Tadoule Lake, Manitoba, and in intervocalic syllable initial position d rotacizes more often than not in most communities." Upon first glance, it does not seem that [r] is strictly an allophone of [d] since its distribution in my data is not entirely in predictable alternation with [d]. While only [d] can occur in word-initial position, and only [r] can occur in word- or syllable-final position, both occur in similar environments intervocalically. Examples are presented in (23). (23) Distribution of [d] and [r] [d] d a c | 'crow, raven'  word-initial  iri none  didzi 'dirty, filthy' diiwe 'difficult syllable-initial/  nedah 'it's heavy'  harajie 'spider'  intervocalic  yagddene 'angel'  te6are 'night'  ?edii 'hot'  sare 'my older sister'  none  bar 'meat'  syllable-final  deyar 'hard' deryi 'this' However, when examining the contexts of the intervocalic examples more closely, it can be seen that [d] occurs intervocalically when it is stem-initial, while [r] is intramorphemic, as shown in (24) . 27  (24)  ne-dah 'it's heavy'  hara-jie 'spider'  ne- adjectival prefix (Li 1946:415)  hara morpheme gloss unknown  -daS 'to be heavy' (Li 1933:128)  -dzi horadzi 'spider' ( L i 1933:143)  yagd-dene 'angel'  teSare 'night'  ya-gdle 'butterfly' (sky-crawlers)  morpheme breakdown unknown  dene 'person'  cf. teSe 'night' (Elford and Elford 1981:17)  ?e-du 'hot'  sare 'my older sister'  ?e- indef. obj. prefix (Li 1946:416)  s(i)- first person possessive prefix  -du 'hot' ( L i 1933:129)  -are 'elder sister' (Li 1933:126)  Based on these examples, we can hypothesise that [d] is realised as [r] intervocalically when it is not stem-initial. Note further that [d] in coda position is realised as [r]. This concludes the presentation of the consonant inventory.  2 7  This is comparable to a pattern observed in Bantu by Larry Hyman (Douglas Pulleyblank, p.c).  43  3.6 Conclusion The goal of this chapter has been to present the consonant inventory of Chipewyan and accompanying acoustic analysis. Wave form and spectrogram analyses have revealed several interesting points about the consonant inventory o f the research dialect o f Chipewyan distinct from the dialect documented by L i (1933, 1946). First of all, Athapaskan languages are traditionally described as having three types of stops: voiceless unaspirated, voiceless aspirated and glottalised. Analysis of the voiceless unaspirated (plain) stops has shown that specific representatives of voiceless unaspirated stops, the bilabial [b] and the dental [d], are consistently voiced, rather than voiceless unaspirated. The alveolar and palatal plain affricates are intermediate cases which are not consistently voiced or voiceless throughout the stop duration. The remaining plain stops/affricates are voiceless unaspirated. The generalisation which can be drawn from these results is that there is a front-back distinction in the laryngeal property of voicing of the plain stops. How can we account for this generalisation? Kingston (1996) finds a correlation between voicing of stops and place of articulation cross-linguistically. Examination of stops in UPSID (the U C L A Phonological Segment Inventory Database) shows that voiced stops are less common the more posterior the articulation. Furthermore, Kingston finds an asymmetric distribution of voiced and voiceless stops in languages which have a stop at the velar or bilabial place of articulation but not the other. The asymmetric distribution supports a back-to-front implicational hierarchy for voiced stops, i.e. / g / > / d / > /b/ where the notation "x > y" indicates that the presence of the sound x in a language implies the presence of the sound y, but not vice versa (Kingston 1996:49). A similar asymmetry was found to exist between implosives and ejectives, with implosives exhibiting a back to front hierarchy, like the voiced stops, and ejectives exhibiting a front to back hierarchy. Kingston (1996) provides an aerodynamic explanation for this front-back asymmetry which involves the difference in the volume of the oral cavity between the stop closure and the glottis. Both the volume of the oral cavity when the closure is made, and the how much the volume of the oral cavity may be increased during the interval of the closure, are relevant. The further back a stop is articulated, the smaller the initial volume of the oral cavity behind the closure will be. Maintaining vibration of the vocal folds requires that the air continue to flow up from the sub- to the supra-glottal cavity between the vocal folds, which necessitates that the pressure of the air below the glottis must be greater than that above. Having greater volume in the oral cavity initially, or increasing that volume, slows the increase in supra-glottal air pressure, which means that a stop is more likely to remain voiced through a closure with a more forward articulation (Kingston 1996:51). This explanation can account for why the two most anterior stops, the bilabial and dental, are voiced whereas the other plain stops are not. Furthermore, these findings make an interesting claim regarding the interface o f phonetics and phonology. Stops which behave as a class phonologically are shown to have different phonetic realisations, showing that the phonology is in some respects independent of phonetics. Since these phonetically voiced and voiceless unaspirated stops pattern together phonologically, the voicing feature in the stop series is not distinctive phonologically, as it is for the fricative series.  44  A l l voiceless aspirated and glottalised consonants in the research dialect have been shown to be as expected. A second finding involves consonant mergers. In the research dialect, it appears that the traditionally classified voiceless unaspirated interdental affricate [d6] has merged with the voiced interdental fricative [6]. Likewise, the voiceless aspirated interdental affricate [t6] has merged with the voiceless interdental fricative [0]. Whereas these last two mergers involve a change in manner, there is also evidence of affricate consonant mergers involving a change in place: specifically, the alveolar affricates [dz, c, c] are merging with the palatal affricates [j, c, c] but this merger is not complete. The only labiovelar attested is the voiceless aspirated [k ]; it is of limited phonemic status. w  Regarding the fricatives, spectrogram analysis has shown that [1], the voiced correspondent of the voiceless lateral fricative [i], is definitely more like a fricative than an approximant (sonorant), but [y], the voiced correspondent of the voiceless alveopalatal fricative [s], is clearly more approximant-like. This provides a second example of a mismatch between phonetics and phonology. Two phonemes, [s] and [y], with different phonetic realisations (continuant and sonorant respectively), partem together phonologically. The velar fricative [y] in the research dialect is often replaced by other consonants including [y], [w], [g] and [h], in various contexts. It is usually maintained intervocalically. Based on the findings of this chapter, (25) represents the phonetic consonant inventory for the research dialect of Chipewyan. Phonologically, the plain stops (a. and b.) continue to pattern together, with relative voicing in complementary distribution. Changes are highlighted in bold. (25)  a. b.  Chipewyan (research dialect) consonant inventory Labial Labial  Coronal Dental  b  d  c.  t*  d.  {  e.  f.  Interdental  Cor/Dor Palatal  Dorsal Velar  X x  yc c  z  X 1  k k k  s  i  Alveolar  Lateral  dz/c c  1  t'e 6 0  h  c  h  & y s  h  Labiovelar  Laryngeal Glottal  (k**) ?  Y X  (m) n ga. p ain stops: voiced  (w) r e. voiced continuant  b. plain stops: voiceless unaspirated c. voiceless aspirated stop d. glottalised stop  f. voiceless continuant g. sonorant  h  The revisions of the laryngeal properties of these consonants affect their featural specifications. For the processes discussed in the remaining chapters, I will be arguing for revisions to the Dual Mechanism Hypothesis (Rice 1994) which was presented in chapter two. A revised 45  version is shown (26).  The only difference from the version presented in chapter two is the  addition of the category of voiced stops, and the inclusion of the glottalised stops in the table. (26)  Revised Dual Mechanism Hypothesis (adapted from Rice 1994:108) Vd  Vis  Vis  Gltsed  Vd  Vis  stop  unasp  asp  stop  fric  fric  stop  stop  V  [SG]  V  [CG]  V V  voice stop  V  V  V  While the addition of the category of voiced stop reflects the phonetic realisations of [b] and [d], the voiced stops pattern phonologically with the voiceless unaspirated stops. The revised feature geometry for each classification is shown in (27). (27) Revised featural specifications (adapted from Rice 1994:115) V d Stop ROOT /  V i s Unasp Stop ROOT  Aspirated Stop ROOT /  \  [stop] Laryngeal  [stop]  ROOT  ROOT /  \  [stop] Laryngeal  [SG]  [voice] V d Fricative  \  [stop] Laryngeal  Gltlsed Stop  [CG]  Vis Fricative ROOT  Laryngeal [voice] The differences between this specification and the version presented in Rice (1994) include the addition of the voiced stop to represent the voiced bilabial and dental stops present in the research dialect, and the representation of [voice] as a dependent of the laryngeal node rather than a direct dependent of the root node. With these theoretical assumptions in place, we can proceed to examine several laryngeal phonological processes in Chipewyan: laryngeal neutralisation and spirantisation (chapter four), continuant voicing (chapter five) and tone assimilation (chapter six).  46  Chapter 4 Laryngeal Neutralisation and Spirantisation 4.1  Introduction  Steriade (1997) proposes a cue-based approach to the licensing of laryngeal features. Laryngeal features are typically neutralised in positions where the cues to the relevant contrast are not available, or where additional articulatory measures would be necessary. Laryngeal contrasts are permitted in positions where the scale of perceptibility of these contrasts is high. Howe (1998) has shown that Steriade's analysis is not adequate for characterising certain cases of laryngealisation in Wakashan languages. In the South Wakashan language Nuuchahnulth, glottalised sonorants are strictly prevocalic, and in the North Wakashan language Oowekyala, are preglottalised word-initially but postglottalised word-finally (Howe 1998). Steriade's cuebased approach predicts just the opposite. The aim of this chapter is to test glottalisation evidence from the Athapaskan language family against Steriade's hypothesis. A s discussed in chapter two, Proto-Athapaskan (PA) is posited to have had a three-way laryngeal distinction between voiceless unaspirated or plain stops, voiceless aspirated stops and glottalised stops. These distinctions are maintained in the daughter languages. However, stemfinal contrasts have been reduced in most languages. P A is posited to have had a two-way contrast between plain and glottalised consonants stem-finally. In the daughter languages, this two-way laryngeal contrast may be maintained (in languages such as Hupa), or the laryngeal distinctions may be neutralised (in languages such as Koyukon). In some languages, such as Chipewyan, a diachronic process of spirantisation has occurred. A l l series of stops in Chipewyan have undergone spirantisation. The spirantisation process is relevant to the discussion of laryngeal neutralisation because while spirantisation involves loss of the manner feature [stop], it also entails loss of the stop's laryngeal features [spread glottis] or [constricted glottis]. Section 4.2 will review Steriade's (1997) Licensing by Cue hypothesis in some detail. Section 4.3 will present data illustrating the glottalisation patterns in Athapaskan languages. Section 4.4 will present an Optimality Theory (McCarthy and Prince 1993, 1994, Shaw 1994, Steriade 1997) analysis of the languages which maintain laryngeal distinctions, and of the languages which neutralise laryngeal distinctions stem-finally. Section 4.5 will present the case of Chipewyan where stem-final laryngeal contrasts are neither preserved nor neutralised; stemfinal stop consonants are spirantised. Section 4.6 will conclude. 4.2 The Licensing by Cue hypothesis (Steriade 1997) Steriade proposes a cue-based approach to the licensing of laryngeal features. Differing from previous prosodically-based theories that have argued for the licensing of laryngeal features based on syllable position, this hypothesis bases licensing on position of best perceptibility. Laryngeal features are typically neutralised in positions where the cues to the relevant contrast are not available, or where additional articulatory measures would be necessary. Laryngeal contrasts are permitted in positions where the scale of perceptibility of these contrasts is high.  47  Steriade examines a wide variety of languages which provide evidence for this hypothesis. Regarding glottalisation, Steriade notes that there are two classes affected: sonorants and obstruents. A classification relevant for patterns of glottalisation often involves the sonorant/obstruent distinction. The timing of the glottal closure with respect to the edges of the oral constriction tends to be different for the two classes. "The laryngeal constriction is timed to the onset of the oral closure in sonorants, and to its release in obstruents" (Steriade 1997:77). This may be illustrated as shown in (1) and (2). Square brackets indicate onset and offset of gesture. (1)  Preferred timing for oral and glottal constriction in glottalised sonorants (Steriade 1997:77)  (2)  Glottal gestures:  ... adduction.. ][  Oral gestures:  [  contextual  internal  cues  cues constriction  vowel  ]  ][...consonantal sonorant....]  Preferred timing for oral and glottal constriction in glottalised stops (Steriade 1997:78) contextual cues Glottal gestures: Oral gestures:  constriction...][...adduction.. [...obstruent...release...][  vowel  ]  The hypothesis of Licensing by Cue predicts that the glottalised sonorants will be optimally realised when there is a preceding vowel or sonorant (left-hand cue). Glottalised stops will be optimally realised when followed by a vowel or sonorant (right-hand cue) as shown in the scales in (3) and (4). (3)  Perceptibility scale for glottalised sonorants (Steriade 1997:81) [+son]  (4)  >  #  , [-son]  Perceptibility scale for glottalised obstruents (Steriade 1997:81) [+son]  >  #,  [-son]  Sonorants are more likely to neutralise in the absence of a preceding vowel or sonorant, whereas obstruents will be most likely to neutralise in the absence of a following vowel or sonorant. Steriade suggests that the two classes are also differentiated by distinct perceptual realisations: sonorants or preglottalised consonants have "creaky voicing" whereas stops or post-glottalised consonants have explosive or ejective release. She postulates different features for these  48  effects; therefore, glottalised sonorants and glottalised obstruents do not form a natural class. Different ranking of the perceptibility constraints with faithfulness constraints can produce different grammars. As an example, the constraint ranking for Yokuts is shown in (5). In Yokuts, glottalised sonorants are disallowed after consonants and word-initially, i.e. in environments where context clues are lacking. Ejective stops, however, are allowed in all environments. (5)  Constraint ranking for Yokuts (Steriade 1997:82) (Constraints are ranked from top to bottom.) * creak constraints  faithfulness conditions Preserve [ejection]  *creak/{[-son], #}  *eiection constraints  28  ^ejection/  {[-son],#}  *ejection/  V  Preserve [creak] *creak/V  Preserve [ejection] is undominated which ensures that all underlying ejectives are preserved. Preserve [creak] only outranks the bottom creak condition, so the only glottalised sonorants that will surface are those in a post-vocalic environment. Having presented the theoretical background, we can turn to the Athapaskan data. 4.3 Data: glottalisation patterns in Athapaskan Athapaskan languages do not include glottalised sonorants in their inventory, therefore I will focus on the analysis of glottalised stops. Note that in the Athapaskan literature "stem" refers to root plus suffix, and uninflected stems (i.e. stems without any prefixation) are monosyllabic The Athapaskan stem normally takes the form CV(C) (Krauss and Golla 1996). Therefore, stem-initial may also be considered "onset" and stem-final considered "coda". Furthermore, phonological processes that target the stem generally only affect continuants, so the phonological environment which precedes the stem does not have any impact on the possibilities of stem realisation of stops. First I will present the stem-initial distribution. 4.3.1 Stem-initial distribution All modern Athapaskan languages have maintained the three-way distinction between plain, aspirated and glottalised stops in stem-initial position. This is illustrated with examples from Chipewyan, shown in (6). In the examples given, the first consonant of each item is the steminitial consonant. Note that while most of the examples shown are nouns, the same distributional patterns are attested for verbs.  In keeping with the terminology used elsewhere in this thesis, I will use the feature [CG] or [constricted glottis] in place of Steriade's term 'ejection'.  2 8  49  (6)  Stem-initial contrasts Glottalised  Series  Plain  Aspirated  Dental  -del 'blood'  -ta 'father'  -tgze 'back'  Interdental  -doar 'it rattles'  t6e 'stone'  t'6e 'sinew'  Alveolar  dz? 'gum'  ci  Lateral  A4ne 'mouse'  Xis 'grease'  Xis 'paper'  Palatal  ja 'here'  ca 'rain'  cis 'mitts'  Velar  gu 'insect'  kun 'fire'  koe  (Li 1933:140) ca 'hat'  'excrement'  'clouds'  4.3.2 Stem-final distribution: retention of plain and glottalised stops O f the twenty-three Northern Athapaskan languages, stem-final glottalisation has been lost in all languages except Kolchan (Upper Kuskokwim), Tanaina, Ahtna, Tagish, Tsetsaut and Sarcee (Krauss and Golla 1996). To my knowledge, stem-final glottalisation has also been lost in the Apachean subgroup, but has been retained in the Pacific Coast subgroup. Hupa, the only surviving language of the Pacific Coast subgroup (Golla 1977), is representative of the most conservative type of Athapaskan language, that which maintains both plain and glottalised stops stem-finally. Examples are shown in (7), all cited from Golla (1996). (7)  Stem-final contrasts Series  Plain  Glottalised  Dental-  [d] i i d 'smoke' (p. 367)  [ti Wimit' 'my belly' (p. 367)  alveolar  [dz] k^iwididz 'string/rope' (p. 367)  [c] tic 'cane' (p. 384)  Alveopalatal  [T] c i j 'firewood' (p. 367)  [c] die 'valley quail' (p. 367)  Velar  [gy] xo-wiligy 'story' (p.367)  [&]  Back velar  [g' ] yinag' 'upstream' (p. 367)  [q] io-q 'salmon, fish' (p. 367)  Lateral  [k] not in inventory  [t]  w  dinky 'four' (p. 367) citehsde-X 'they went off  (p. 367) 4.3.3 Stem-final neutralisation Some Athapaskan languages neutralise the distinction between plain and glottalised consonants stem-finally, with plain stops surfacing. A representative example is Koyukon, with examples shown in (8).  50  (8)  Stem-final neutralisation Proto-Athapaskan form *nsq *?a-6 *?a-X  Koyukon form nsq (perf. stem) ?oc (perf. stem) ?oX (perf. stem)  29  Gloss 'swallow' (Leer 1979:54) 'few go' (Leer 1979:56) 'chew' (Leer 1979:58)  The third pattern, stem-final spirantisation, where both laryngeal and manner features are neutralised, will be presented in section 4.5. The next section will present an Optimality Theory analysis of the different patterns of glottalisation seen to this point. 4.4 Analysis 4.4.1 Previous analysis Rice (1994) presents the problem of laryngeal neutralisation. The Dual Mechanism Hypothes (Rice 1994), discussed in chapter two, allows for the differences in distribution of stops. As we have seen, some Athapaskan languages with syllable-final stops reflect patterns of distribution of Proto-Athapaskan, while many others are more restrictive in terms of which stops, if any, are allowed syllable-finally. This distribution is shown in (9). (9)  Distribution of stops (Rice 1994:112)  plain aspirated glottalised  stem-initial V V V  stem-final V (V)  The feature geometry showing plain and glottalised stops based on the Dual Mechanism Hypothesis, is shown in (10). (10)  Dual mechanism hypothesis (Rice 1994:115): Plain stop ROOT  I [stop]  Gltlsed Stop ROOT  / \ [stop] Laryngeal [CG]  Laryngeal neutralisation involves delinking of the laryngeal node, with the resulting phoneme plain stop. Leer (1979) notes that while in many orthographies final non-glottalised stops are written as voiceless unaspirated (plain) stops, he has chosen to follow the practice of writing voiceless aspirated stops. This is because of citation of data from languages such as Tanacross and Ingalik which have light:heavy final pairs; where the stem ends in an obstruent, the light final is voiceless and the heavy final is voiced. 2 9  51  4.4.2 Optimality Theory analysis Steriade (1997) argues that laryngeal constriction is timed to the release of oral closure in obstruents. Therefore, the hypothesis of Licensing by Cue predicts that the glottalised stops will be optimally realised when followed by a vowel or sonorant (right-hand cue) as shown in the scale in (11), repeated from section 4.2. (11)  Perceptibility scale for glottalised obstruents (Steriade 1997:81) [+son]  >  #,  [-son]  Obstruents will be most likely to neutralise in the absence of a following vowel or sonorant. This yields the following constraints: (12)  *[CG]/  {[-son], #}:  [CG] is dispreferred before a non-sonorant or word-finally.  (13)  *[CG]/  V : [CG] is dispreferred before a vowel.  The Licensing by Cue hypothesis suggests that the ranking of constraint (12) over constraint (13) is universal. These perceptibility constraints interact with a faithfulness constraint. (14)  Preserve [CG]: Underlying [CG] must be preserved as such.  The first tableau shows the cases of stem-initial obstruents where glottalisation is preserved. Tableau 4.1:  Stem-initial ejection  tee  Chipewyan te6 'coal' ? ] /'charcoal' (Krauss 1964:127)) Preserve [CG] (cf. P A *t'ex *[CG V w  a. de6 b. te0 E^C.  *! *!  tee  Ranking of the faithfulness constraint above the cue constraint ensures that underlying ejectives i.e. consonants specified for the feature [CG] in the input surface in the output, as in the optimal candidate (c). The next tableau shows the case of laryngeal neutralisation illustrated by Koyukon.  52  Tableau 4.2:  Stem-final neutralisation Koyukon -naq (perf. stem) 'swallow' (cf. P A *-nsq) (Leer 1979:54)  -naq  *[CG]/  {[-son], #}  *[CG]/  V  *!  "^a. -naq b. -naq  Preserve [CG]  *!  As can be seen from tableau 4.2, ranking of the second cue constraint above the faithfulness constraint achieves the desired result: laryngeal distinctions are neutralised word-finally, in an environment where the cues for realising glottalisation are not optimal. Tableau 4.3 illustrates the case exemplified by Hupa, where plain and glottalised stops are retained stem-finally. This case poses a challenge for Steriade's (1997) hypothesis; glottalisation should be dispreferred when the appropriate cues are missing. In this case, glottalisation should be dispreferred word-finally because there is not following vowel to optimise perception. However, Steriade states (1997:78): "This is not to say that ...a following vowel is indispensable for [t]: but the contexts of optimal perceptibility will clearly be different for the two segment types." So, while word-final glottalisation is not optimal for perceptibility, is still possible. This is achieved by making the faithfulness constraint the most highly-ranked. Tableau 4.3:  Stem-final ejection HupaWimit' 'my belly' (Golla 1996:367)  30  (cf. P A *-wad? 'belly' (Krauss 1964:126)) Wimit'  Preserve  a. Wimid  *! *!  b. Wimit «®"c. Wimit'  [CG]  *rcGi/ *  ir-son],#}  \*\CGV  1  v  A s shown in tableau 4.3, the grammar for Hupa has the faithfulness constraint ranked the highest, ensuring that any underlying glottalised consonants are preserved whether steminitially or stem-finally. The following section will present the case of spirantisation, as exemplified by Chipewyan.  [W] is described as a voiceless labiovelar semivowel.  53  4.5 4.5.1  Spirantisation Background  As noted in chapter two, a spirantisation process developed during the transition from P P A to P A which operated on root-final stops, changing them to fricatives. There were two specific environments for root-final spirantisation: (i) when followed by an obstruent suffix, and (ii) in verb stem-final (i.e. word-final) position where the stem vowel was full. In Chipewyan and several other Athapaskan languages, spirantisation has occurred to a much greater degree. A s discussed in chapter two, syllable-final consonants in Chipewyan are limited to the fricatives and sonorants shown in (15), repeated from chapter two. (15)  Syllabi e-final/coda consonants m [6] [s] [z] [f]  ce0 'duck' X96 'boil' (Li 1933:134) sas 'bear' sez 'hiccough' (Li 1933:146) der 'crane' (Li 1933:129)  [1] [s] [x]  del 'blood' nates 'dream/vision'  [Y]  coy 'porcupine quill' (Li 1933:149)  M  gah 'rabbit'  xax 'playing cards'  [n]  dzen 'muskrat/rat'  [r]  bar 'meat'  Stops (with the exception of glottal stop; see note 3) in word-final position in ProtoAthapaskan developed into fricatives or sonorants. Examples are shown in (16). (16)  Proto-Athaoaskan  ChiDewvan  Gloss  *T9d  T9r  'smoke' (Krauss 1978:Table4)  *Xa-t *xenc  A,ar  'algae, moss' (Krauss 1978:Table 3)  sj0  'wart' (Krauss 1978:Table7)  *ci-A"  cif  'scab' (Krauss 1978:Table 3)  *ca-X  M  'cradle' (Krauss 1978:Table 3)  *-?a-X  -?al (perf. stem)  'chew' (Leer 1979:58)  -?ai (imp. stem)  'chew' (Li 1933:126)  * te' J  •fez  'flour' (Krauss 1978:Table6)  *-?a-6  -?az (perf. stem)  'few go' (Leer 1979:56)  -?as (imp. stem)  'two persons go' (Li 1933:126)  *se-q  sey  'spit' (Krauss 1978:Table 3)  *-n9q  -nay (perf. stem)  'swallow' (Leer 1979:54)  *-qe?  -ke '  'foot' (Krauss 1978:Table 5a)  t  3  Recall from chapter two that stem-final glottal stop developed into high tone on the preceding vowel Chipewyan. 3 1  54  4.5.2 Analysis The constraints presented so far are not adequate to account for the spirantisation cases, as shown in tableau 4.4. Stem-final glottalisation can be prevented by the appropriate ranking of Preserve [CG], but the desired candidate with the stem-final fricative is no better than the ones where the stem-final consonant has been neutralised. Tableau 4.4:  Spirantisation Chipewyan c i i 'scab' (cf. P A *ci-X) (Krauss 1978)  *rCG]/  cit  {[-son], #}  Preserve [CG]  *[CG]/  V  * *  >sra. ciX «^b. ciX  *!  c. ciX"  *  isrd. c i i  While Steriade's (1997) cue-based approach can account for neutralisation of laryngeal features, it does not address cases where manner features are neutralised. I do not know of any research that has shown whether certain cues are relevant for the realisation of particular manner features. Steriade's analysis has focused exclusively on phonetic cues and has specifically argued that syllable structure is not relevant. However, the post-vocalic spirantisation exhibited in Chipewyan seems to suggest that constraints making reference to syllable structure would best explain the process. Cross-linguistically, spirantisation most typically occurs in a postvocalic context (Kenstowicz 1994:35). For example, in Tiberian Hebrew, oral stops (except for geminates) spirantise postvocalically (McCarthy 1981). A n d , in Dominican Island Carib, an Arawakan language, aspirated /p/ and Ikl spirantise to lil and Ixl, especially following a loud stressed vowel (Taylor 1978). Nisqa'a reduplication provides a third example (Shaw 1987, 1994). Based on the cross-linguistic evidence for spirantisation to occur post-vocalically, I propose that a constraint expressing the limitation of codas to continuants is justified for Chipewyan. This condition on c o d a s (17)  32  can be stated as follows:  C O D A - C O N D : *[STOP]/  ] a : [STOP] is not licensed in coda position.  33  Recall from the Dual Mechanism Hypothesis (Rice 1994) that the three types of stops are the only phonemes marked underlyingly for [STOP]. These phonemes never occur in coda position; this constraint will ensure that is the case. The interaction of these constraints is shown in tableau 4.5.  For further reference on constraints on codas, see McCarthy and Prince (1994) and see Shaw (1994, 1999) for use of this particular constraint. I am using the manner feature [STOP] rather than [-CONT] to be consistent with terminology used elsewhere in this thesis. They may be considered to be equivalent. 3 2  3 3  55  Tableau 4.5:  Spirantisation (revised) Chipewyan cif 'scab' (cf. P A *ci-t) (Krauss 1978)  cit  CODA-COND:  nSTOPV a. ciX b. ciX c. crX"  la  *! *! *  d. c i f  Preserve  *[CG]/ {[-son], #}  *[CG]/  V  [CG]  * * *! *! *  <s°e. cif  The coda condition constraint succeeds in eliminating the first three candidates. Candidate (d), with syllable-final glottalised fricative, is eliminated by the highest-ranked cue constraint. Note that I did not include candidates with syllable-final sonorants; these are possible candidates but additional faithfulness constraints requiring faithfulness to manner of articulation would prevent their emergence as optimal. 4.6 Conclusion Some of the glottalisation evidence from Athapaskan supports Steriade's (1997) analysis. Stem-initially, glottalised stops are retained, as exemplified by Chipewyan. This result is predicted from Steriade's hypothesis: optimal identification of glottalisation contrasts is dependent on the right-hand context such as a following vowel. Conversely, identification of glottalisation is not optimal when the right-hand context is not a vowel, i.e. the end of the word. This is shown by evidence of laryngeal neutralisation in Koyukon. Evidence from Hupa, however, differs. In Hupa, glottalisation contrasts are maintained word-finally; since there is no following vowel, this is not the optimal context for retention of glottalisation. While this case is less common than the previous ones, Steriade's hypothesis can accommodate this through the relative ranking of constraints; the faithfulness condition Preserve [CG] is undominated, and therefore ranked above the cue constraints. In several Athapaskan languages including Chipewyan, no stops are maintained in syllablefinal condition; a process of spirantisation has occurred. Spirantisation involves neutralisation of manner as well as laryngeal features. Steriade's analysis of laryngeal neutralization contexts does not address cases where manner is neutralised. Spirantisation in Chipewyan can be accounted for within Optimality Theory by resorting to a syllable-sensitive coda condition banning stops in syllable-final position.  56  Chapter 5 Continuant Voicing 5.1  Introduction  Continuant voicing is a familiar phonological process in Athapaskan languages (Young and Morgan (1980), Cook (1984), Hargus (1988), Rice (1989) and others). The focus of continuant voicing in this chapter is the stem-initial continuants in nouns. When a possessive prefix is added, a voiceless continuant surfaces as a voiced continuant. Section 5.2 will outline relevant background information concerning the continuant voicing process. Section 5.3 will present data from the Fort Chipewyan dialect of Chipewyan documented by L i (1933) which exhibits the continuant voicing process, and will be analysed using Optimality Theory (McCarthy and Prince 1993). Section 5.4 will present the research dialect of Chipewyan which does not exhibit the process, due to the fact that the environment which triggers voicing is not present. Section 5.5 will conclude. 5.2 Background information 5.2.1  Possessive prefixes  In Athapaskan languages possession is most often indicated by the addition of the possessive prefixes. The Chipewyan prefixes are (Li 1946:402): (1)  se-  'my'  nuhe-  'our'  ne-  'your'  nuhe-  'your (pi)'  be-  'his, her, its'  hube-  'their'  ye-  'his, her, its, their, when the subject of the sentence is also in the third but not  the same person' [i.e. disjoint anaphor] ?e-  'indefinite possessor'  ?ede-  'one's own, my own, your own, his own, etc'  The possessed form of nouns in Chipewyan often differs from the unpossessed form. "The difference usually consists of a change of tone or the addition of a suffix -i or the change of the initial voiceless spirant to the voiced form. But not all nouns undergo such changes and some, like names of body-parts or kinship terms, must always be possessed." (Li 1933:123) The change we are focusing on in this chapter is the voicing of the stem-initial fricative, and examples will be shown in section 5.3. 5.2.2 Target consonants The consonants which occur in stem-initial position and are expected to participate in the voicing process are shown in (2).  57  (2)  Voiceless  e  s  s  i  X  6  z  y  1  Y  continuant Voiced continuant Note that the fricative [h] does not participate in the voicing process. First, [h] does not occur in stem-initial position in Chipewyan. Second, [h] is specified for spread glottis and the feature [SG] does not cooccur with the feature [VOICE]. Finally, [h] does not have a voiced counterpart. 5.2.3 Theoretical as sumptions A s shown in the consonant inventory presented in chapter two, Chipewyan exhibits several different classes of laryngeal features; this chapter is focusing on the laryngeal properties of stem-initial continuants. For characterising the distinction between these classes of laryngeal features, I am following the revised Dual Mechanism Hypothesis (revised from Rice 1994:108), as shown in (3), repeated from section 3.6. The only difference from the version presented in chapter two is the addition of the category of voiced stops. The crucial claim is that the so-called voiceless unaspirated stops/affricates are unmarked in terms of laryngeal specification, in contrast to both voiced [b] and [d], and the voiceless unaspirated stops/affricates, designated [SG]. Note, importantly, that while the voiced stops and the voiceless unaspirated stops/affricates have different phonetic realisations, they pattern together as a class phonologically in the process of continuant voicing. (3)  Revised Dual Mechanism Hypothesis (adapted from Rice 1994:108) Vd  Vis  Vis  Vd  Vis  stop  unasp  asp  fric  fric  stop  stop  V  [SG] voice stop  V V  V  •  V  Under this hypothesis, stops are marked for manner whereas fricatives are not. The continuant voicing process will only affect those consonants unmarked for [stop], by causing a voiceless continuant to become voiced. Addition of a possessive prefix introduces a floating [voice] feature which attaches via a laryngeal node to the root node of a voiceless fricative, producing its voiced alternant. I have chosen to use the above hypotheses regarding feature specification because, as discussed in chapter two, Rice (1994) has shown how the hypothesis can account for several phonological processes in Athapaskan languages including the Athapaskan "d-effects", spirantisation, devoicing in Koyukon and continuant voicing.  58  5.2.4 Previous analyses Some analyses have presented the continuant voicing process as a process of word-initial devoicing, rather than a process of voicing. For example, McDonough (1990:172) presents the word-initial devoicing rule for Navajo shown in (4). (See also Cook (1984) on Sarcee.) (4)  +cont —>  -voi / #  Rice (1989) argues against a devoicing analysis based on evidence from Slave. First, there are nouns with voiced continuants in initial position. Second, when noun stems with a voiceless initial continuant are incorporated into verbs, the continuant remains voiceless. If the continuant were voiced underlyingly, one would expect it to remain voiced when incorporated within the verb. A n example from the Hare and Bearlake dialects of Slave is shown in (5) (Rice 1989:66). (5)  kinashinededa 's/he walks around singing' cf. shj (unpossessed form), -yine (possessed form) 'song'  Finally, in certain types of compounds, the voiceless continuant of the first member of the compound remains voiceless even when possessed, as shown in the Slave forms in (6) (Rice 1989:948). (6)  ieteh 'bread' i e 'flour' + t'eh 'charcoal' -iet'ehe (possessed form of compound) cf. -lehe 'flour' (possessed form of non-compound)  If the initial continuant were underlyingly voiced, one would expect it to be voiced in the possessed form of the compound. Therefore, Rice (1988, 1991, 1994) presents the process as one of voicing rather than devoicing. (See also Hargus (1988) on Sekani.) A stem-initial fricative becomes voiced when preceded by a segment in nouns. Rice posits a morpheme [voice] which indicates inflection. It surfaces in verbs, possessed nouns and inflected postpositions, environments where one might expect an inflectional morpheme, and it is absent in non-possessed nouns and non-inflected postpositions. When the morpheme links to a stem-initial fricative, the fricative surfaces as voiced. Rice argues that the voice feature does not associate to a stem-initial stop because under the (original) Dual Mechanism Hypothesis, voicing is not distinctive for stops, and because of structure preservation, a voiced stop cannot surface since it is not found underlyingly in the lexical phonology.  34  The idea of a [voice] morpheme which indicates all types of inflection is controversial. In my Note that Rice explains the exceptions to voicing, such as the compound structure in (6), by arguing for application of default rules (see Rice 1988) or by superimposing a bisyllabic prosodic structure where only the right member of the bisyllabic foot can be voiced (see Rice 1991). 3 4  59  analysis, I am simply treating [voice] as a floating feature which is associated with the possessive prefix. A n d since under the revised version of the Dual Mechanism Hypothesis, voicing is distinctive for stops, one cannot appeal to structure preservation. Instead, I will use a markedness constraint requiring that segments specified for [stop] cannot be specified for [voice]. The next section will present the continuant voicing process in Chipewyan. 5.3 Continuant voicing 5.3.1  Data  Examples of nouns from the Fort Chipewyan dialect which undergo this process of steminitial continuant voicing are shown in (7), cited from L i (1933). Unpossessed  Possessed  Page  Gloss  Number  Form  Form  eue  se-6u6  my spear  139  sa  se-za  my sun, my watch  141  sa  se-zane  my play, my joy  142  sey  se-zeye  my spittle  142  sel  se-zele  my call  142  sme  se-zme  my summer  143  son  se-yone  my song  146  fez  se-lez-e  my urine  150  tus  se-lus  my spoon  150  fur  se-lure  my scab, my chunks of  150  ice fuwe  se-luwe  my fish  150  xai  se-yay-e  my root  134  xel  se-yel-e  (my) something to be  134  carried on the back -xu  se-yu, se-yii  135  tooth  Arguments made against a devoicing analysis for Slave by Rice (1994) can also apply to Chipewyan. Chipewyan also has voiced continuants in initial position (cf. examples such as 66 'skin', and lah 'work' in chapter three ). Secondly, when noun stems with a voiceless initial 35  continuant are incorporated into verbs, the continuant remains voiceless. If the continuant were voiced underlyingly, one would expect it to remain voiced when incorporated within the verb. A n example is shown in (8).  Recall that [1] is classified as a fricative/continuant in Athapaskan languages. In chapter three, we saw evidence that [1] has acoustic properties of a fricative. 3 5  60  (8)  besadee 'It shines (from afar). (It is shining.)' (Elford & Elford 1981:141) cf. sa 'sun' (unpossessed form) se-za 'my sun' (possessed form)  The following subsection will show how this process can be interpreted under an Optimality Theory analysis. 5.3.2 Optimality Theory analysis Optimality Theory can account for the application of the continuant voicing process to steminitial fricatives, while also accounting for the failure of voicing to apply to stem-initial stops, by appealing to universal markedness constraints rather than simply stipulating to which segments the process can apply. The following constraints will be used in this analysis : 36  Grounding Constraints: The first constraint prevents glottal stop and any of the glottalised stops or affricates from becoming voiced, since glottalised consonants specified for voice do not occur in the consonant inventory. (9)  *[CG]/[voice]: A segment specified for the feature [CG] (constricted glottis) cannot be specified for [voice].  The next constraint prevents the aspirated consonants from becoming voiced, and also prevents [h], which is specified for [SG], from becoming voiced, since there is no voiced counterpart for [h] in the consonant inventory. (10)  *[SG]/[voice]: A segment specified for the feature [SG] (spread glottis) cannot be specified for [voice].  Markedness constraint: In terms of universal markedness theory, voiced stops are considered more marked than voiceless stops. "Languages with only a single series of stops (from the point of view of laryngeal setting) are reported to have voiceless stops" (Ladefoged and Maddieson 1996:53). On a scale of universal sonority, voiceless stops are the least sonorous segments, followed by voiced stops, voiceless continunants, voiced continuants, sonorants and vowels. (Blevins 1995). Studies of child language acquisition have also shown voiced consonants to be more marked, since voiceless unaspirated consonants are acquired before voiced ones (Macken 1995). This constraint will prevent all stops, but crucially the voiceless unaspirated ones, from becoming voiced. References for the development of these constraints include Archangeli and Pulleyblank (1994) (grounded constraints), McCarthy and Prince (1995) (faithfulness constraints), Pulleyblank (1996) (DEP-PATH) and McCarthy and Prince (1993) (alignment constraints). 3 6  61  (11)  *[stop]/[voice]: A consonant specified for the manner feature [stop] cannot be specified for [voice].  Faithfulness Constraints: M A X and D E P constraints prevent the deletion ( M A X ) or insertion (DEP) of features or paths to features, ensuring that the output is as faithful as possible to the input. (12)  M A X [SG]: A [SG] specification in the input must have a correspondent in the output.  (13)  M A X [voice]: A [voice] ([v]) specification in the input must have a correspondent in the output.  (14)  M A X [stop]: A [stop] specification in the input must have a correspondent in the output.  (15)  D E P - P A T H [voice]: A n y output path between [voice] ([v]) and an anchor must have a correspondent path in the input.  A P A R S E constraint ensures that if a [voice] feature is present, it links to an obstruent. (16)  P A R S E [voice]: The feature [voice] ([v]) must associate to an obstruent.  Alignment Constraint: Finally, an alignment constraint ensures that when the feature [voice] is parsed, it aligns with the left edge of the noun stem. (17)  A L I G N ([voice], L E F T , noun stem, L E F T ) : Align the left edge of the feature [voice] ([v]) with the left edge of the stem.  The following tableaux illustrate how these constraints interact to account for the continuant voicing process in Chipewyan. Note that vowels will not be specified for [voice] in the tableaux. Since voicing on vowels is redundant, I assume [voice] is assigned by default.  62  Tableau 5.1:  Chipewyan continuant voicing - stem-initial continuant se6ue  se-Gue  'my spear' (cf. L i 1933:139 0u0 'spear')  *[CG]  *[SG]  MAX  MAX  *[stop]  Align  Parse  DEP-  /[V]  /[V]  [SG]  [v]  /[v]  LEFT  [v]  PATH  [v]  [v]  *  *!  a. s e 0 u 6  [v] b.  *!  sedud  M c. s e 6 u 6 \  / [v]  d. s e 6 u 0  [V] In candidate (a), the [voice] feature has been parsed, but it has not aligned with the left edge of the stem. There is only one violation because the [voice] feature can only be parsed to obstruents, so it is only misaligned by one segment, [6]. Candidate (b), completely faithful to the input, is eliminated because the [voice] feature has not been parsed. Candidate (c) satisfies left alignment, but the feature has also aligned to the right edge of the stem. This incurs two violations of D E P - P A T H [v], for two links to the [voice] feature which were not present in the input. It also looks like a gapped configuration, which is not well-formed. (See Archangeli and Pulleyblank (1994) for a discussion of gapped configurations.) However, since the [voice] feature can only be parsed to obstruents, this candidate is not considered to have a gapped configuration. The optimal candidate (d), incurs one violation of D E P - P A T H [v] for the linking of the [voice] feature but satisfies all the other constraints. Comparison of candidate (b) with the optimal candidate shows crucial ranking between P A R S E and D E P - P A T H . The next tableau shows how the same constraints prevent the voicing process from applying to stops or affricates (non-continuants).  63  Tableau 5.2:  Chipewyan - no continuant voicing - stem-initial stop ?eka0 'something's handle' (cf. L i 1933:137 ka0 'rod, handle')  ? e - k a 0  *[SG]  MAX  MAX  MAX  *[stop]  Align  Parse  DEP-  A [v][SG][stp]  /[V]  [stop]  [SG]  [v]  /[V]  LEFT  [v]  PATH  a. ? e g a 0  *!  h  [v]  *  *  / l'\ [v][SG][stp]  *  *!  b. ? e g a 0 / \ [v][stp] c. ? e x a 6  [v] d. ? £ y a 6  *  *!  *  *!  *  *  [v]  *  *!  e. ? e k a 6  / \ 1 [SG][stp][v]  * f. ? e k a 6 / \ [v]  [SG][stp]  In candidate (a), the [voice] feature has linked to the voiceless aspirated stop, violating one of the grounding constraints and the markedness constraint. Candidate (b) has lost its [SG] specification, violating the grounding constraint M A X [SG] and also violates the markedness constraint prohibiting stops from becoming voiced. Candidate (c) has become a voiceless fricative by losing both underlying features [SG] and [stop], violating M A X [stop] and M A X [SG]. Candidate (d) has also lost both underlying feature specifications but fares better than candidate (c) since the [voice] feature has been parsed. Candidate (e) fails because the [voice] feature, while parsed, is not aligned with the left edge of the stem, which violates A L I G N L E F T by one segment. The optimal candidate is completely faithful to the input. This is the desired result, since the continuant voicing process should not affect stops. The only constraint violated by the optimal candidate is P A R S E . This tableau establishes crucial ranking between A L I G N and P A R S E . (Compare candidate (e) with the optimal candidate.) In the Fort Chipewyan dialect documented by L i (1933), there are no voiced stops. In the research dialect, certain stops are phonetically voiced, although this is not a contrastive distinction within the phonological system. Recall from chapter three that plain stops with more forward articulation, [b] and [d], are phonetically voiced, while plain stops/affricates towards the back of the mouth are voiceless unaspirated.  64  The next tableau presents a hypothetical case, with underlying voiced stops, and shows that the same set of constraints will not affect the underlying specification of the voiced stops. The case is hypothetical because in the research dialect, my consultants use an independent postpositional construction rather than the possessive prefix to indicate possession. If the possessive prefix were used, this would be the expected result. Tableau 5.3: Expected output for possession on stem-initial voiced stop ssbes 'my knife (cf. L i 1933:127 bes 'knife') 1  s  E  -be s  / \  *[CG]  *[SG]  MAX  MAX  *[stop]  Align  Parse  DEP-  /[v]  /[V]  [stop]  [v]  /[V]  LEFT  [v]  PATH  [v] [v][stp] a. s £ - p e s  [v]  [v]  *  *!  [stp]  b. z  E  *  -be s  *  *!  1 /\ [v][v][stp]  *  *  c. s E - b e s  / \ [v]  [v][stp]  Candidate (a) has lost the underlying voice specification on the voiced stop Ibl, becoming a voiceless unaspirated Ibl (phonetically [p]). This is a violation of M A X [stop]. The voice feature has not been parsed. The second candidate retains the underlying specification, incurring one violation of the markedness constraint, and parses [voice], but since it is not parsed to the left edge of the stem, this violates A L I G N - L E F T by one segment. The optimal candidate is completely faithful to the input. It incurs one violation of the markedness constraint and one violation of P A R S E , but is still the best option. This tableau establishes crucial ranking between M A X [voice] and *[stop]/[voice]. Optimality Theory provides a good illustration of the continuant voicing process in dialects of Chipewyan exhibiting the process. The next section will present data from the research dialect which does not show any evidence of continuant voicing. 5.4 Research dialect: no continuant voicing Continuant voicing does not occur in the research dialect because the environment inducing continuant voicing, addition of the possessive prefix, is not present since most noun stems in this dialect use a postpositional construction to indicate possession. Most of the nouns which retain the possessive prefix are inalienably possessed; all of the inalienably possessed fricative-initial forms are voiced, and since they do not occur in isolation, it is impossible to say whether the underlying fricative is voiced or voiceless. The few  65  alienably possessed noun stems in the data which take the possessive prefix do not begin with fricatives. Alienably possessed nouns are used with the postpositional form si-cj possessive prefix si -  3 8  3 7  rather than using the  'my' for the majority of nouns. Some examples are shown in (8).  (18) Use of postposition for possession sicj sa ~ sa sicj  'my wristwatch/I have a wristwatch'  sicj san ~ ssn sicj  'my music/I have some music'  sicj tus ~ i u s sicj  'my spoon/I have a spoon'  39  The conventional form of possession would use the possessive prefix, accompanied by continuant voicing, i.e. si-za, si-ysn, si-lus. Additional examples showing fricative-initial noun stems are presented in (19).  40  Li (1946:403) glosses the postposition -cj as 'from'. As discussed in chapter three, the vowel [E] in the possessive prefixes is generally [i] in this dialect. Both forms in each pair mean both things. Only non-compounded forms are presented, since in other Athapaskan languages, the continuant voicing process affects compounds differently. (See for example Rice 1989 on Slave.) I did however elicit some forms with inalienable nouns in compounds; they were still voiced in the compound cf. se-la-gane 'my finger' (myhand-something skinny), be-6a-t0iya 'his mustache' (his-mouth-hair). 3 7  3 8  3 9  4 0  66  Nouns used with postposition  Nouns used with  (Alienable Nouns)  possessive prefix  'snow'  -la  'cousin' (RD)  6e  'ice'  -la  'hand'  Gita  'door'  -ljye  'daughter' (RD)  Gptai  'cup'  -yaze  'child/baby'  Gen  'star'  -yize  'son' (RD)  6e6ere GeG  'leather' 'flour'  'chest' (RD) 'name'  a  'dog'  -zi -zi -6a  TUS  'spoon'  -6e  fuwe  'fish'  yu  'skin' 'clothes'  fuzane  'trout'  lah labah labadah  'work' 'socks'  hdi sa  'tea'  sas  'bear'  s£mba  'money'  san  'music/song'  sasdn  'metal/iron'  Gai  'mouth'  4 1  'potato' 'sun/wristwatch'  s£lai  'hills'  yah  •sky'  yati  'language'  xai  'bat/club'  xax  '(playing) cards'  xai  'winter'  One speaker defines the difference between the possessive prefix and the postposition as follows: " si- means 'my', and sicjmeans 'I have' or 'I got'." The second speaker says they don't use the possessive prefix with most nouns because "We don't relate the animals and things to be ours, only if you want to tell a story maybe" whereas she describes the inalienable nouns by saying "There's no such thing as it standing by itself. It has to belong to someone." Lack of continuant voicing is not unprecedented in Athapaskan languages O f Slave, a closely related Northern Athapaskan language, Rice (1989:69) writes: "In innovative speech in all dialects, the rule of voicing often does not apply in possessed nouns." A n example is shown in (20) (Rice 1989:69). (20)  fu selu, se?fu, sefri  'spoon' 'my spoon'  The forms for 'socks', 'potato' and 'tea' are loan words from French.  67  However, while the Slave example in (20) shows that continuant voicing may not occur in innovative speech, this is not the case for the inalienably possessed examples from Chipewyan, where forms with the possessive prefix are always voiced. Use of the postpositional form has also been documented. From L i (1946:402): "In Chipewyan there is a distinct tendency to use a periphrastic expression to denote possession for those nouns except [inalienably possessed nouns] i.e., ij secj 'my dog' (< dog me-from)." Richardson (1968) calls this postposition a 'verbal noun' which acts as a transitive verb. Some examples are given in (21) (Richardson 1968:4). (21)  becj  'he has it'  fue becj deneyu becj deneyu fue becj  'he has a fish' 'the man has it' 'the man has a fish'  Furthermore, "Sometimes becj is used as a possessive word instead of a verbal noun. This is usually only used with the one word, i j , although in some dialects it can be used with certain other words" (Richardson 1968:8) but also cf. 'fish' in (21). Finally, postpositions also take the possessive prefixes e.g. [si-ba] 'for me' (lp. possessive prefix-for). One would expect continuant voicing to also occur with continuant-initial postpositions when the possessive prefix is added. However, the list of postpositional stems given in L i (1946) does not list any postpositions with voiceless continuant-initial stems. To summarise, continuant voicing does not occur in the research dialect because the environment inducing continuant voicing, addition of the possessive prefix, is not present in most nouns. I do not want to suggest that continuant voicing never occurs in this dialect, only that there is no clear evidence based on morphophonemic alternations of the process in the data collected so far. 5.5 Conclusion Like other Athapaskan languages, certain dialects of Chipewyan (namely, the Fort Chipewyan dialect documented by L i 1933, 1946) display a phonological process of continuant voicing. A noun stem-initial voiceless continuant becomes voiced when the possessive prefix is added. A n Optimal Theoretic account of the process affecting stem-initial continuants is presented, and it is shown that the same set of constraints will not affect stem-initial stops, whether voiceless or voiced. The research dialect of Chipewyan does not provide clear evidence of a voicing alternation due to the non-occurrence of paradigmatically-related forms in contexts which would crucially attest to the synchronic validity of the process. Most alienable nouns take a postpositional form to indicate possession, and inalienable forms never occur without the possessive prefix, so it is not possible to determine the underlying form. There are two possible options to consider (Patricia A . Shaw, p.c.): (i) restructuring of the lexicon; all inalienable stems have been  68  restructured w i t h a v o i c e d stem-initial continuant.  O r , (ii) i n a l i e n a b l e s t e m s still are v o i c e l e s s in  the l e x i c o n b u t a l l u n d e r g o c o n t i n u a n t v o i c i n g as i n o t h e r d i a l e c t s . O n e w a y o f t e s t i n g the h y p o t h e s e s w o u l d be to f i n d e x a m p l e s o f l o a n w o r d s b e g i n n i n g w i t h a n initial v o i c e l e s s f r i c a t i v e . U n d e r h y p o t h e s i s (i), o n e w o u l d e x p e c t t h e v o i c e l e s s f r i c a t i v e t o r e m a i n v o i c e l e s s u n d e r p o s s e s s i o n . U n d e r h y p o t h e s i s (ii), o n e w o u l d e x p e c t t h e v o i c e l e s s f r i c a t i v e to b e c o m e v o i c e d w h e n t h e n o u n is p o s s e s s e d . T h i s is a n i s s u e f o r f u r t h e r r e s e a r c h .  In either c a s e h o w e v e r , there h a s b e e n a r e s t r u c t u r i n g i n the m o r p h o s y n t a c t i c s y s t e m o f h o w p o s s e s s i o n is m a r k e d . T h e p r e f i x s e t s u b c a t e g o r i s e s f o r i n a l i e n a b l e n o u n r o o t s , w h e r e a s t h e postpositions s u b c a t e g o r i s e f o r alienable n o u n roots. T h i s m a k e s the r e s e a r c h dialect different from  o t h e r d i a l e c t s o f C h i p e w y a n w h e r e t h e m o s t c o m m o n w a y o f m a r k i n g p o s s e s s i o n is b y  u s e o f the p o s s e s s i v e prefix.  69  Chapter 6 Tone Assimilation 6.1  Introduction  This chapter will present two examples of tone assimilation in Athapaskan languages in different morphological domains: tone assimilation in verbs in Navajo, and tone assimilation in nouns in Chipewyan. Tone and tonal processes are the synchronic residue of Proto-Athapaskan laryngeal behaviour. A s discussed in chapter two, there is a wide amount of variation among tonal properties in Athapaskan languages. Not all languages are considered tonal; those that are may differ in tonal polarity: some languages have default high tone while some languages have default low tone. In addition, some languages have been characterised as having a "pitch accent" system rather than a tonal system. The predominant theory of Athapaskan tonogenesis (Krauss 1978) suggests that all tonal variants arose from constriction on vowels. A stem vowel followed by a stem-final glottal stop in Proto-Athapaskan caused the vowel to be constricted (laryngealised). In many of the daughter languages, final glottal stop is no longer present; constriction on the vowel in the proto-language is realised as tone in these daughter languages. However, while in some languages the constriction developed into high tone, in others it developed into low tone. Navajo and Chipewyan are both considered to be "high-marked" tone languages, but they are illustrative of this tone reversal i.e. cognate forms have opposite tonal values in the two languages. A s discussed in chapter two, why two opposite tone values developed from the same source is still unresolved. In Navajo, a language of the Apachean subgroup, certain prefixes are lexically marked for high tone. This tone may spread to a following mora subject to certain constraints. The required constraints will show that both prosodic and morphological structures play an important role in this analysis. Tone will only spread within a bimoraic foot, and feet are aligned with the right edge of the prosodic word. This chapter will use an Optimality Theory (McCarthy and Prince 1993) framework to provide an explicit account of how these constraints interact to account for the basic pattern of tone assimilation while also accounting for examples where tone assimilation does not occur as might be expected. Chipewyan, a Northern Athapaskan language which is the focus of this thesis, exhibits a process of tone assimilation triggered by the possessive suffix. Possessed nouns in Chipewyan are most often formed by adding a possessive prefix to the noun stem. When the possessive prefix is added, the stem form may remain the same, or may be changed any of three ways, including: (i) voicing of the stem-initial fricative, which was discussed in chapter five, (ii) addition of a possessive suffix, or (iii) addition of a high tone. These three types of changes are not mutually exclusive; for example, nouns may undergo continuant voicing and add the possessive suffix. Or, nouns may already have a lexical possessive suffix in which case possession may be marked by the addition of high tone. The possessive suffix also surfaces in some types of compound formation. Addition of the high-toned possessive suffix to some noun stems results in the high tone of the suffix spreading to the preceding vowel in  70  the stem. The possessive suffix is always high-toned unless it has become lexicalised as part of the noun, in which case it may or may not have high tone. This process of tone assimilation will be discussed in the second half of this chapter. Section 6.2 will present background information on Navajo, and section 6.3 will present the tone assimilation process in Navajo. Section 6.4 will outline the Optimality constraints underlying the proposed analysis, and section 6.5 will present the tableaux which motivate the relative ranking of the proposed constraints for Navajo.' Section 6.6 and 6.7 will cover the Chipewyan data and analysis. Section 6.8 concludes. 6.2 Navajo background information This section will present background information which will be essential in the analysis of tone assimilation. Navajo is a language of the Apachean subgroup of Athapaskan spoken in the southwestern United States. Navajo is most often characterised as a high-marked tone language (Kari 1976, Young and Morgan 1992, and others) where low tone is the (unmarked) default.  42  The verb in Navajo, as in all Athapaskan languages, is a complex structure  consisting of a verb stem and several prefixes. The model followed in Kari (1976:40) includes ten prefix classes: (1)  ##AD V+IT+PL#OB J + D E I C + A S P + M O D E + P E R F + S U B J+CL+STEM## 1  2  3  4  5  6  7  8  9  10  A n internal boundary is distinguished between prefixes 3 and 4, with 1-3 considered "disjunct" prefixes and the remaining prefixes "conjunct" prefixes. The conjunctive prefixes are formed from a very limited set of consonants including b, d, n, i , 1, s, dz, sh, j , h, hw, gh and ?  4 3  Note  that all aspirated and glottalised consonants are excluded; therefore most laryngeal distinctions have been neutralised in the conjunct prefixes. The disjunctives can take almost the full range of consonants; which ones are excluded is not stated (Kari 1976:43). The conjunct-disjunct distinction is relevant for restricting the application of a number of phonological processes. The classifier (prefix 10) and verb stem are usually considered to form a separate morphological domain from the rest of the verb-word because as Kari (1976:40) notes, a number of rules apply only before classifier + stem, including tone assimilation. There are four classifier prefixes, [i, d, 1] or 0 (zero) and each verb must have a classifier. The [1] classifier is actually an alternation of the [i] classifier, produced by voicing of [i] under "D-effect" (Young and Morgan 1992:885), but is usually listed as a separate classifier. The zero classifier often marks the intransitive (but not always), [i] is most usually a causativiser-transitiviser and [d] a passiviser. [1] is most frequently a passive of a causative. (Kari 1976:20-21)  de Jong and McDonough (1993) examine whether Navajo is best classified as a tone or pitch accent language. Acoustic evidence points toward it being a tone language. The orthography used by Kari (1976) follows Young and Morgan (1943) and is the system most used by Navajo communities. Most of the orthographic symbols are equivalent to the IPA symbols, except for the following: ts = [ c ] , ts' = [c], j = [J], ch = [c], ch' ='[t], zh = [2], sh = [s], dl = [A.], tl = [X], tl' = [t], gh = [Y], x, h = [x] and ' = [?]. (Kari 1976:13) 4 2  4 3  71  McDonough (1990) proposes an alternative analysis of Navajo verb structure. She presents the verb as a bipartite structure, consisting of a verb stem and an "INFL" stem, a synthetic morph of mode inflected for subject. Each stem has prefixes. The verb is built in two sections which are joined at some point in the derivation. Each stem acts as a separate domain for phonological rules. This bipartite structure is illustrated as follows (McDonough 1990:25): (2)  clitics  #  [Agr af  A s p Mode/s af  stem  [cl-stem] rb Ve  afstem  The minimal word requirement in Navajo necessitates that the verb is bisyllabic (McDonough 1990:137). The verb stem itself is always monosyllabic so it can be distinguished as the last C V ( V ) ( C ) of the verb. Both analyses recognise an important morphological boundary between the classifier-verb stem and the rest of the affixal domain. The next section will present the data. 6.3 Navajo tone assimilation This section will present the description and illustrating data of the process of tone assimilation in Navajo including examples of cases where tone assimilation does not occur as expected. A l l data is from Kari (1976) who presents a derivational analysis of tone assimilation process in Navajo. In Navajo, certain prefixes have inherent high tone, such as /ch'iV 'out', /k'iV 'apart', /bi 7 'against', /na7 'iterative', lol 'optative', and III 'perfective'. (Kari 1976:59) These prefixes trigger tone assimilation, where the prefix assimilates its inherent high tone onto the vowel to its right across one or more intervening consonants. Low-toned prefixes do not trigger tone assimilation. The following examples with iterative prefix /na7 illustrate (Kari 1976:59): (3)  Imperfective  Iterative  hanofchaad  hana'no'fcha'  'card it' 2p. dual  hanishchaad  hana'nfshcha'  'card it'1 p.  'adrfbas  n''di'fbas  'make a trip' 3p.  'azhdiibas  nfzh'dffbas  'make a trip' 4p.  44  However, tone assimilation does not take place if there is more than one C V sequence to the right of the high tone prefix, or if the vowel to the right of the intervening consonant is long, as shown in the following examples (Kari 1976:60): (4)  ch'i'na'nfshche'e'h (lp. iterative) but chTninishche'e'h (lp. imp.) 'drive it out' na'ha'shbiih (1 p. iterative)  but  na'hojrfbiih (4p. iterative) 'build a hogan'  nina'ha'shniih (lp. iterative)  but  nina'hiilniih (lp dual, iterative) 'trade'  ya'dfshtih (lp. imp.)  but  ya'diiltih (lp. dual imp.) 'chatter'  The disjunct prefixes take the shape CV (and rarely CVC) but the na-/na- prefixes may reduce (optionally) to syllabic n-/ri- in certain phonological environments. (Young and Morgan 1992:843) Which segments can function as a tone-bearing unit will be discussed in section 6.4.  72  Note that Kari (1976) does not give morpheme breakdowns for the examples shown in (3) and (4). Additional examples with morpheme breakdowns are shown in (5). Often several other phonological rules intervene between the underlying form and the surface form. Kari (1976) uses several derivational rules in his analysis of the verbs to arrive at the ultimate form. For clarity, I will only list the underlying form, the point in the derivation where tone assimilation takes place, and the surface representation. A l l page numbers refer to Kari (1976).  45  (5) Examples of tone assimilation in Navajo  Underlying Form  Tone Assimilation  Surface Form  Gloss  Pg  ni'na'no'fka'  'herd them back' 2d  135  Trigger: disjunct Target: disjunct na'+na'#ni+oh+i+ka'  na'+na'Sn+o'h+f+ka'  Trigger: disiunct Target: conjunct na#ji+f+kad  na'Sji'+t+kad  n'ji'fkad  'to sew it' 4  113  na+bi'#ni+ghil  na+bi'#ni'+ghil  nabi'ni'ghil  'to push it about' 2  113  ch'i'#ni+sh+d+ne'e'h  ch'i'#ni'+sh+d+ne'e'h  ch'i'ni'sh'ne'e'h  'to crawl out' 1  123  na'#yi+f+ze'e'h  na'#yi'+i+ze'e'h  neTse'e'h  'tan it' 3  133  ha+na'#ghi+i'+sh+d+ziY'  ha+na'#ghi'+sh+d+ziY'  hana'a'sdziT'  'exhale' 1  180  Trigger: conjunct Target: conjunct o'+ni+l+ghaf  o'+ni'+l+gha't  wo'o'lghaf  'eat meat' 2  140  ha#di+o'+ni+l+gha'a'sh  ha#d+o'+ni'+l+ghaVsh  hado'o'lgha'a'sh  'shout' 2  141  gh+i'+ni+cha  gh+i'+ni'+cha  yi'ni'cha  'cry' 2  173  ni+i'+ni+d+na'  n+i'+ni'+d+na'  yi'ni''na'  'arrive crawling' 2  191  si+i'+ni+f+tin  s+i'+ni'+f+tin  si'ni'ftin  'freeze it' 2  197  di+si+i'+ni+l+dzil  d+i'+ni'+l+dzil  di'ni'ldzil  'start to hop' 2  211  Note that there are no examples of tone spreading from a disjunct prefix (with no overt conjunct prefixes) spreading to the verb stem, or tone spreading from a conjunct prefix to the verb stem. The derivational approach presented in Kari (1976) stipulates that tone assimilation cannot apply if there is more than one C V sequence to the right of the high tone prefix, or if the vowel to the right of the intervening consonant is long. Two C V sequences entails two vowels, or two moras, and a long vowel entails two moras. Therefore, the generalisation which can be The number indicates person. Note also the following abbreviations: "d" dual # disjunct boundary + conjunct or formative boundary  73  drawn from this is that tone will not spread beyond one vowel, or one mora. The Optimality Theory analysis will account for this generalisation by means of prosodic and alignment constraints, as we will see in the next section. 6.4 Optimality Theory constraints This section will outline how tone assimilation in Navajo can be described within the O T framework. 6.4.1  Faithfulness constraints  The first family of constraints which will be relevant in this analysis is the family of faithfulness constraints . A M A X constraint will prevent the deletion of any underlying 46  tones. A D E P constraint will prevent a tone from linking to an anchor to which it was not linked underlyingly. (6)  M A X - I O (Tone): Every tone in the input has a correspondent in the output.  (7)  D E P - P A T H (Tone): Any output path between tone and an anchor must have a correspondent path in the input.  A n output obeying all faithfulness constraints will be identical to the input. Change will occur if another constraint dominates a faithfulness constraint. For spreading of tone to occur, a constraint from the A L I G N family will be used, which we will see in section 6.4.3. 6.4.2 Prosodic Constraints  47  First, something must be said about the assumptions of tone association. For tone to be realised directly, it must be attached to a tone-bearing unit (TBU). While some analyses of tone languages have argued for the syllable as the tone-bearing unit, others have argued for the mora. For Navajo, I will argue that the tone-bearing unit is the mora. In the Navajo data given, tone associates to a vowel or else to a syllabic nasal.  48  Vowels and nasals which constitute a  syllable nucleus are moraic, and therefore capable of bearing tone. Tone will spread to a following vowel (one mora). The tone span encompasses two vowels (two moras). Tone will not spread to a long vowel (which is maximally considered to be two moras), or when there is more than one C V prefix following the inherent high tone (two vowels = two moras). If the syllable were the tone-bearing unit, one could not explain the patterns where tone assimilation does not occur as expected. For example, if tone spread to a following syllable, the output forms shown in (8) would be expected. (Data repeated from (4).  Syllable  boundaries are represented by ".") MAX-PATH (Tone) and DEP-PATH (Tone) are adapted from the faithfulness constraints presented in Pulleyblank(1996). Prosodic constraints FT-BINu and PARSEu are taken from McCarthy and Prince (1993). Note that in the data, syllabic nasals only occur in word-initial position before homorganic but nonidentical segments.  4 6  4 7  4 8  74  (8)  Syllable as T B U  Input  Expected Output  Actual Output  a. ni.na'.hash.niih  ni.na'.ha'sh.niih  ni.na'.ha'sh.niih  b. na'.ho.jii.biih  na'.ho'.jif.biih  na'.ho.jii.biih  c. ni.na'.hiil.niih  ni.na'.hi'iT.niih  ni.na'.hiil.niih  In the first example, tone spreads to the following syllable which obtains the correct output. However, in the second example, tone spreading to the following syllable does not result in the correct output. Likewise, in example (c), tone has spread to the following syllable which contains a long vowel; this is also not the correct output. Therefore, weight must be taken into account. The type of constraint needed is a type of bounded constraint that will limit the tone from spreading when the relevant domain for tone spread is greater than one mora. Different types of bounded constraints have been used for restricting tone assimilation in other tone languages. (See Myers (1997) on Bantu). The prosodic device which will describe the two-mora bounding restriction in this analysis is the foot. (9)  F T - B I N p : Feet must be binary under moraic analysis.  I propose that the foot is a trochaic foot. Because the spread of tone is from left to right, a trochaic foot can be seen as spreading tone from the head of the foot to the rest of the foot. Furthermore, tone does not spread to long vowels, as one might expect if the foot were iambic. For determining feet, I will argue that prefix coda consonants in Navajo are non-moraic. Note that coda consonants occurring in prefix syllables come from a restricted set. The fourth person subject pronoun ([z] contracted from ji - Qi-]) may surface as a coda in position 5. (See example (l)d.) Various forms of the subject prefixes including [s, z, s, h] may sometimes surface as codas in position 9. O f the classifier prefixes in position 10, only [r, 1] will ever surface as codas. (Classifier [d] merges with the verb stem in the "D-effect" process.) A l l of these consonants are fricatives, which are classed as obstruents.  49  There are many examples of quantity-sensitive languages where obstruents are considered to be weightless, or non-moraic. Zee (1988) discusses evidence from Lithuanian that supports the hypothesis that only sonorant consonants, and not obstruents, are moraic in coda position. Rising tone accents, where high tone links to the second mora of a syllable, are found on syllables with long vowels or on syllables with short vowels followed by sonorant consonants,  While [1] is usually considered to be a sonorant, in Athapaskan languages it consistently patterns with the fricatives in phonological alternations such as the "D-effect" and fricative stem-initial/stem-final voicing alternations and so is classed with the obstruents. (See discussion of Chipewyan [1] in chapter three.) And in the classifier position, it is simply the voiced counterpart of [+]. (See section 6.2). For these reasons, I will consider it to be an obstruent, not a sonorant.  75  but never on syllables where short vowels are followed by obstruents. A second piece of evidence comes from infinitive verb formation where vowel lengthening is analysed as mapping infinitive stems to a bimoraic template; lengthening does not occur in sonorant-closed stems. Thirdly, long vowels will shorten within a syllable closed by a sonorant, but not an obstruent. Y i p (1995) surveys tone in East Asian languages. In Chinese languages, she notes that sonorant-final syllables may bear any tone, but obstruent-final syllables only bear a subset of (usually level) tones. One possible explanation can be given making the following assumptions (Yip 1995:488): "Contour tones have two tonal root nodes. The mora is the T B U . Only sonorant codas can be moraic." Thus, if obstruent codas are non-moraic, they can not be a tone-bearing unit, so only level tones will be found in obstruent-closed syllables.  50  Shaw (1992) examines reduplication in Nisgha. She provides evidence that resonant consonants are moraic in Nisgha, and thus eligible to satisfy the reduplicative template, whereas obstruents are not. A similar case is found in Korean (Kim 1997) where a reduplicative template copies the rightmost heavy syllable. In coda position, sonorant consonants are considered heavy whereas obstruents are not. In this analysis of Navajo, I will treat prefix coda consonants as non-moraic. The main argument for this is the fact that if prefix coda consonants were moraic, prefix syllables with long vowels and a coda consonant would be trimoraic which is a marked syllable structure. A second prosodic constraint will be required which will ensure that foot-parsing be exhaustive. (10)  P A R S E p : A l l moras must be parsed into higher prosodic structure.  Finally, I will argue that roots are footed independently i.e. every root is a foot, whether or not it satisfies F T - B I N p . Therefore, it is crucially ranked above F T - B I N p . (11)  R O O T = F O O T : Every root is parsed into a foot.  I propose that the unmarked root form in Navajo is C V C and C V : . If, unlike prefix coda consonants, root coda consonants are considered to be moraic, then both of the unmarked root forms in Navajo are bimoraic. The rationale behind treating coda consonants in the root domain as moraic and coda consonants in the prefix domain as non-moraic is that, crosslinguistically, roots sustain greater markedness than affixes. Subminimal C V roots are marked roots; lexically footing them is a reflection of their marked status. Superheavy C V : C roots can satisfy the bimoraic footing requirement if the final coda consonant, at the edge of the domain, is treated as extrametrical. Note however, that Yip (1995) ends up adopting a syllabic analysis over a moraic one for Chinese tone.  76  When every root is footed, the domain of footing for the prefixes begins at the boundary between prefix 10, the classifier prefix, and the verb root/stem. 6.4.3 Alignment Constraints  5 1  Cross-linguistically, there is a tendency for tones to associate from left to right (Pulleyblank 1986). Using a foot-based analysis, tone assimilation can be described using an alignment constraint. (12)  A L I G N - T O N E - R (Tone Right, Foot Right): A n y tone within a foot will align with the right edge of the foot.  Feet are optimally aligned with the right edge of the prosodic word. (13)  A L I G N - F O O T - R (Foot Right, P W d Right): Align the right edge of the foot with the right edge of the prosodic word.  The following section will illustrate with tableaux how the constraints outlined above interact to describe tone assimilation in Navajo.  Alignment constraints A L I G N - T O N E - R and A L I G N - F O O T - R are adapted from McCarthy and Prince (1993). 5 1  77  6.5 Tone assimilation tableaux Tableau 1 will illustrate the basic pattern of tone assimilation in Navajo (Examples in (3), first column of (4) and (5)). Tableau 6.1: H  Tone assimilation:  ninahashniih  MAX10  1 u u u uu u ni.na'.has.n i: h u u n uu{u} a. (ni.)(na.has.)(n i:) h H  /  1 \  li  ^  H  *!  ROOT= FOOT  [ninalm sniih] Rl'trade'  FTBINn  PARSE n  ALIGNFOOT-R  ALIGNTONE-R  DEPPATH  *  **** **  *!  **** **  *!  *** **  *!  **** **  *  *!  **  *  **  P-P-ijJ-}  b. (ni'.)(na'.ha's.)(n i:) h H  1 c. (ni.na'.)(has.)(n i:) h H  1 \  H H li d. (ni.)(na'.ha's.)(n i:) h H  1\ u u u uu{u} e. (ni.na'.ha's.)(n i:) h H  1 M- li li liliili} f. ni.na'.has.(n i:) h *  H  1 u u u uu{u} g. (ni.na'.)has.(n i:) h H  *  **  *  **  *  **  *!  1 U U. u uu{u} h. ni.(na'.has.)(n i:) h H  1  \  w  u uu{u} i. ni.(na'.ha's.)(n i:") h H U  1 \ u u p. uu{u} "^j. ni.(na'.ha's.)(n i:) h  78  *  Candidate (a) is eliminated because the inherent tone on the na- prefix has been deleted, which violates M A X - I O . This candidate also incurs one violation of F T - B I N p and several violations of A L I G N - F O O T . These alignment violations are assessed by moras; the first foot is misaligned from the right edge of the prosodic word by four moras, and the second foot is misaligned by two moras. Foot binarity is violated in candidates (b), (c) and (d) because of a monomoraic foot, and by a trimoraic foot in candidate (e). P A R S E eliminates candidate (f) where only the verb stem is parsed, leaving three moras unparsed. The remaining candidates each incur one violation of P A R S E because one mora in each is left unparsed. Each candidate incurs at least two violations of A L I G N - F O O T because the leftmost foot is not aligned with the right edge of the prosodic word (misaligned by two moras). Candidate (g) incurs another violation of the constraint because the leftmost foot is misaligned from the right edge of the prosodic word by three moras, thus it is eliminated. A L I G N - T O N E succeeds in eliminating candidate (h), where the tone span has not aligned with the right edge of the foot. Finally, candidate (i) fares as well as the optimal candidate on the first six constraints; it falls to D E P P A T H to eliminate this candidate since the tone has linked to three additional anchors; the optimal candidate (j) has only formed one additional link. This tableau establishes crucial ranking between F T - B I N p and P A R S E p , by comparing the optimal candidate with candidates (b), (c), (d) or (e). Comparing candidate (h) with the optimal candidate also establishes crucial ranking between A L I G N - T O N E and D E P - P A T H . So, Tableau 6.1 has illustrated the basic pattern of tone assimilation in Navajo. The next two tableaux illustrate the cases where tone assimilation does not occur as expected (second column of (4)).  The first is the case where there is more than one C V prefix to the right of the high  tone prefix. The second is where the target prefix contains a long vowel, i.e. cases where there is more than one mora, or T B U , in the affixal domain. The proposed analysis of the first type of case is presented in Tableau 6.2 (following page). M A X - I O eliminates candidate (a) where the inherent high tone has been deleted. Foot binarity is violated in candidates (b), (c) and (d). The remaining candidates have one violation of P A R S E except for candidate (e) which incurs three violations and is therefore eliminated. A s in the previous tableau, each candidate incurs two violations of A L I G N - F O O T because the leftmost foot is misaligned by two moras. Candidate (f) is eliminated by this constraint because the leftmost foot is misaligned by three moras. Candidate (g) fails because the tone is not right aligned with the foot. The remaining candidates are eliminated by D E P - P A T H because tone has linked to additional anchors in each. The optimal candidate (j) has one violation of P A R S E , two violations of A L I G N - F O O T , and satisfies A L I G N - T O N E vacuously; because the inherent tone is not contained within a foot, it has nothing to which to align. Because it has not inserted any additional links, D E P - P A T H is also unviolated. Tone assimilation is optimal if it spreads to the right edge of the prefixal domain, and if the tone spread is bounded by the foot containing the inherent tone. If either one of these conditions cannot be satisfied, it is more optimal for tone assimilation to not occur at all. The candidate which is most faithful to the underlying form, candidate (j) is optimal.  79  Tableau 6.2: tone  Blocking of tone assimilation; more than one C V prefix after inherent na'hojiibiih [na^hojrfbiih] R4'build a hogan'; MAX10  H  1 U  ROOT= FOOT  FTESINu  PARSEu  ALIGNFOOT-R  *  **  U U  ALIGNTONER  DEPPATH  na'.ho.jrt.b i: h u u u uu{u} a. na.(ho.ji't.)(b i:) h H  *! *!  *** **  *!  **** **  *!  **  1 \ u u u b. (na'.ho'.)0it.)(b i:) h H  1  *  c. (na'.)(ho.jif .)(b i:) h H  **  1 \ \ u n u uu{u} d. (na'.ho'.Ji''t.)(b i:) h H  1 H- H u e. na'.ho.ji'r.(b i:) h *  H  *  1 \ u u n un{u} f. (na'.ho'.)jit(b i:) h H !  *  **  *  **  *  **  *  **  *!  *  \  u u u uu{u} g. na'.(ho'.jif .)(b i:) h H  1  \ \  u u u un{u} h. na'.(ho'.ji'f.)(b i:) h H | \ \ \ \ u u a uu{uj i. na'.(ho'.ji'f.)(b i:") h H u u u uu{u} "ffj. na'.(ho.jit.)(b i:) h  Tableau 6.3 shows the second type of condition on tone assimilation: words where the target prefix contains a long vowel.  80  Tableau 6.3:  Blocking of tone assimilation; long vowel in target prefix  ninahiilniih [nina'hiiIniih] R l d'trade' MAX-  H  10  1  ROOT= FOOT  FTBINu  PARSEu  ALIGNFOOT-R  **  **  ALIGNTONER  DEPPATH  ni.na'.h i: l.n i: h u n U|i. |xu{u} a. ni.na.(h i: l.)(n i:) h  *!  **  *!  H i  w  c c  **  cc(c)  b. (ni.na'.h i:" l.)(n i:) h H  1  c c cc  cclcl  c. ni.na'.h i : l.(n i:) h H  *l*  **  *i*  ****  1  c c cc  cc{cl  d. ni.na'.(h i : l.)(n i:) h H  1  c c cc  cc(c}  e. (ni.na'.)h i : l.(n i:) h H  1 \  c c c c  cc(c)  f. ni.(na'.hi:')i: l.(n i:) h H  1  **** **  \  c c  cc  1  ccicl  h. (ni.na'.)(h v." l.)(n i:) h H  | cc  \\ cc  *  **** **  \\  cc  *!  cclcl  g. (ni.na'.)(h i : ' l.)(n i:) h H  c c  *  *##  **** **  \\ cclcl  i. (ni.na'.)(h i : " l.)(n i:")h H  **** **  1 u u uu uu{u} (ni.na'.)(h i : l.)(n i:) h  81  * 1 ***  In Tableau 6.3 (preceding page), candidates (a) through (f) are eliminated by the four highestranking constraints. The interesting candidates are (g) through (j). Each incurs six violations of A L I G N - F O O T , four for the leftmost foot and two for the middle foot. Candidate (g) is eliminated because there is only tone on one mora of the long vowel, tone has not aligned to the right of the foot; (h), (i) and (j) all satisfy A L I G N - T O N E . Once again, it falls to D E P - P A T H to eliminate candidates (h) and (i) because more links have been added in these candidates than in the optimal candidate. A s with the other exception, the candidate which is most faithful to the underlying form, candidate (j) is optimal.  Tone assimilation has not occurred when it  cannot spread to the right edge of the affixal domain while remaining within its foot. This tableau establishes crucial ranking between P A R S E p and the alignment constraints by comparing candidates (c), (d), (e) or (f) with the optimal candidate. Finally, returning to the question of whether prefix coda consonants are moraic in Navajo, I will show in the next tableaux that the correct candidate will not be selected if prefix coda consonants are moraic. The symbol '©' indicates the actual output, the candidate which should be the optimal one. Tableau 6.4:  Moraic coda consonants: wrong candidate selected nina ha shniih [nmalia^sniih]  H  MAXIO  1 u u uu uuu ni. na'. h as. n i : h H  ROOT= FOOT  Rl'trade' FTBINu  PARSEn  ALIGNTONER  DEPPATH  **  1 u u uu uu{u} "S'a. (ni.na'.)(has.)(n i:) h H  ALIGNFOOT-R  *!  *  **  *  1 \ ©b. ni.(na'.ha' s.)(n i:) h In this example, (a) emerges as the winning candidate. The correct candidate, (b) fails because it violates foot binarity. However, assuming moraic coda consonants, a second option is available. McDonough (1990) has argued that the coda consonant of the LNFL stem (/s7) can be considered extrametrical, because it is at the edge of the LNFL stem domain. If so, it could be excluded from the foot parse, and the optimal candidate would emerge. However, this explanation would still not be able to account for tone assimilation in fourth person conjugations, as we will see in the next tableau. The fourth person prefix ji- can only refer to human (or personified animal) subjects and is often introduced into the verb in coreference constructions. It is also used as an honorific. (Kari 1976:24). It occurs in prefix position 5. It is often contracted to [z], and usually surfaces as a coda when it is contracted. Because prefix position 5 is not at the edge of any morphological domain, a coda consonant surfacing there cannot be considered extrametrical.  82  The next tableau will present the fourth person iterative form nizhdiibas  [ni'zdi' Tbas] 'make a <  trip', repeated from the data in (3). Kari (1976) does not give the morpheme breakdown of this form, but I presume it to be close to that shown in (14). (14)  [mzdFtbas] na-}i-di-i-bas 2-5-6- 10-stem iterative-4p.-aspectual/thematic prefrx-classifier-make a trip  Tableau 6.5:  Moraic coda [nfzdribas] consonants with extrametricality: wrong candidate selected nizhdiibas 'make a trip' 4 MAX10  H 1  ROOT= FOOT  FTBINu  u u uu uu n V z.d i i.b a s  H 1 n v-  PARSEu  ALIGNFOOT-R  *  ***  *  **  ALIGNTONER  DEPPATH  c{nl CC «®"a. (n i'z.)d i t. (b a s) *!  H 1 u u{u} uu b. n i'.(zd i)+.(b a s)  **  *!  H 1  *  \  u u u{u} HC ©c. ( n i l .di')t.(bas)  H 1  **!*  *  \  uu u{u} uu ©d. n i'z ,d \ t.(b a s) Candidates (c) and (d) are both equivalent to the attested form and should emerge as optimal. However, foot binarity eliminates candidate (c), and P A R S E eliminates candidate (d). Foot binarity also eliminates candidate (b). With the constraints shown, candidate (a), without tone assimilation, emerges as the winning candidate. I conclude that the optimal analysis of tone assimilation treats Navajo prefix coda consonants (obstruents) as non-moraic. The following tableau illustrates that this assumption allows the correct output to be deemed optimal in my analysis.  83  Tableau 6.6:  Non-moraic coda consonants: example with fourth person prefix nizhdiibas  [ni"zdfibas] 'make a trip' 4 MAX-  H  10  1 u u uu uu n i z.d i t.b a s H  ROOT= FOOT  FTBINu  PARSEn  ALIGNFOOT-R  ALIGNTONER  **  *!  DEPPATH  1  1 c c CC a. (n i'z.d i +.) (b a s) H  1  \  **  **l  **  *  \  u u uu b. (n i'z.d i t.) (b a's) H  1  \  C C "S'c. (n i'z.d i t.) (b a s) The three best candidates are shown. Candidate (a) is ruled out by alignment and candidate (b) is eliminated by D E P - P A T H . The correct candidate is shown to be optimal. To summarise the conclusions to this point, we have seen that vowels or syllabic nasals, i.e. nuclei of a syllable, are tone-bearing units in Navajo. The levels of representation are the mora, syllable and foot. Feet are bimoraic, trochaic feet which are constructed from the right edge of the prosodic word. In the process of tone assimilation, tone spreads from left to right within the foot containing the inherent tone, therefore, tone spread is maximally two moras. I have also argued that roots/stems must be independently footed due to their marked status. The remainder of the chapter will present a tone assimilation process in Chipewyan. 6.6 Tonal effects of the Chipewyan possessive suffix Possessed nouns in Chipewyan are most often formed by adding a possessive prefix to the noun stem. When the possessive prefix is added, the stem form may remain the same, or may be changed in some way. In his sketch of the Chipewyan grammar, L i (1946) outlines five changes which the noun stem may undergo when the possessive prefix is added. These changes are shown in (15) (Li 1946:402).  84  (15)  Changes effected by the possessive prefix  Change  Examples  a. Without change of form, except for the  6146 'spear': be-Su6 'his spear'  voicing of the initial fricative.  kj 'a mound, heap': ?e-kj 'beaver lodge' b^n-e 'edge': be-b£n-e 'its edge'  b. With the adding of the possessive  t'6en 'bone': se-t'6en-e  'my bone'  suffix - 8 .  del 'blood': se-del-e 'my blood' c4 'excrement': be-cin-8 'his excrement' sa'watch': se-za 'my watch'  c. When the noun has an original suffix  i u y - 8 'fish': s 8 - t u y - e 'my fish' w  w  -e, the possessed form is formed by  b^n-e'war party': dene-b^n-e 'a war  changing this -e to -e.  party of Indians'  d. Without the suffix -8, but with a shift  t'a 'feather': be-ta 'its feather'  of tone of the stem from low to high.  ke 'moccasins': be-ke 'his moccasins'  e. With the possessive suffix and with a  tu 'water': kun-tu-e 'liquor (<fire-water)'  shift of tone.  dze 'gum': tOe-dzey-e 'tar (<stone-gum)'  This section will examine the last type of change, where addition of the possessive suffix causes a low-toned root to become high. In the second example cited in (e) above, if the possessive suffix -e is added to a high-toned root, the root surfaces with low tone in the compound i.e.  dze  'gum':  tde-dzey-e  'tar (<stone-gum)'. However, this is the only example of  this process of dissimilation cited by L i (1933, 1946). Nor have I found any additional examples in my fieldwork. Also, many forms do not show dissimilation in an environment where one would expect it, such as the following examples, repeated from (c) and (d): c§ 'excrement': and  be-c^n-e  'his excrement';  tu-nie 'shore' (water-land).  b$n-e  'war party':  dsne-b£n-e  'a war party of Indians';  Since there .are many more examples of this type, I will  conclude that the dissimilation pattern cited by L i (1946) is not an active phonological process in the grammar of Chipewyan, and is the exception rather than the rule. Additional examples of the assimilation process are shown in (16): (16)  tu 'water':  kun-tu-e 'liquor (< kun 'fire' - tu 'water')' J ie-tii-e 'juice (< ] ie 'berry' - tu 'water')'  tu 'rope':  ij-tu-e 'leash (<ij 'dog' -fax'rope')' 6^-tu-e 'iceline (< 0e 'ice' - tu 'rope')'  The examples of the tone assimilation process are very limited, since most nouns in Chipewyan do not take the possessive suffix, either in the possessed form, or in compounds. Examples of compounds which do not take the possessive suffix are shown in (17).  85  (17)  Compounds without possessive suffix Noun-Noun compounds Xis-tili  paper box (paper-pail/box)  Xoh-tili  basket (grass-pail/box)  Xu-ke  shoelace (rope-shoe)  cek aze-yu  dress (girl-clothes)  ce6-eyeges  duck eggs (duck-eggs)  bor-kQ? C4-1CQ?  food store/meat market (meat-house) outhouse (excrement-house)  daXah  soap (?-grease)  det'an-kQ?  nest (bird-house) mooseberries (moose-berry)  w  deniye-jie harajie-kQ?  kun-fus ~ fus-kun  web (spider-house) wooden spoon (wood/fire - spoon)  Udi-tili  teapot (tea-pail)  ni-Xis  map (land-paper) bear trap (bear-trap)  sas-?ecu5i sasdn-tili  tin can (metal-pot)  t6i-ci  comb (head-?)  tei-?a  pillow (head-?)  tSi-ya-bis  razor (head-hair-knife)  t6i-ya-gu  lice (head-hair-bugs)  xai-ke  winter boots (winter-boots)  ya-go-dene  angel (sky-crawler (butterfly)-person)  yu-kQ?  clothing/department store (cloth-house)  fu-ciye  fish guts (fish-guts)  fu-dene  fisherman (fish-person)  tu-kQ?  fishhouse (fish-house)  ?et©en-6e ?eka-kun  caribou hide (caribou-skin) candle (fat-fire)  Noun-Adjective co mpounds detan-co  eagle (bird-big)  tu-co  lake (water-big)  t6e-co  mountain (rock-big)  ij-CQ  horse (dog-big)  ij-ce  bitch/female dog (dog-?)  Noun-Verb comDOunds priest (person-he speaks) dene-yati  eee-te  bread/bannock (flour-cooked)  In some cases, it appears that the possessive suffix has become lexicalised, as in the examples shown in (18).  86  (18)  Nouns with lexicalised suffix  52  teltd-e  tobacco/cigarettes  biscan-e  automobile  ciy-e  guts/intestines  daislen-e  devil/evil person  dEn£-susl£n-£  real people/human beings (person-?-sfx)  d£n£-k66-£  chief (person-leader-sfx)  S£iaiy-£  hills (RD)  ta-ba>£ ya-gdl-£  tent (?-to hang and spread out-sfx)  tuw-£  fish  tuzan-£  trout  ki6-e  boat  butterfly (sky-crawlers-sfx)  The next section will present the analysis of tone assimilation from an Optimal Theoretic perspective. 6.7 Tone assimilation analysis Most of the constraints presented in the analysis of tone assimilation in Navajo are relevant in the analysis of Chipewyan, with three exceptions. R O O T = F O O T does not play a role in the Chipewyan assimilation process, at least not with the data collected so far. Along with M A X - I O , I will include M A X - P A T H , which I also assume is highly ranked in Navajo, but it is easier to see its effect in the Chipewyan data. It will prevent a tone from delinking from an anchor to which it was linked underlyingly. (19)  M A X - P A T H (Tone): A n y input path between a tone and an anchor must have a correspondent path in the output.  Finally, the tone alignment constraint is revised as shown in (20). (20)  A L I G N - T O N E - L (Tone Left, Foot Left): Align the tone of the possessive suffix with the left edge of the foot.  We must make the stipulation that it is the tone of the possessive suffix that aligns, rather than just any tone, because there is no other evidence of tone assimilation in the grammar of Chipewyan. The ranking of the constraints does not need to be altered from the ranking at work in Navajo. Tableau 6.7 will illustrate the process of possessive suffix tone assimilation in Chipewyan.  Polysyllabic forms are also polymorphemic, but I am not sure of the morpheme breakdowns for most of the forms listed here. 5 2  87  T a b l e a u 6.7:  C h i p e w y a n t o n e as s i m i l a t i o n  tu H  H  1  'water': MAX-  1  uu u u k ii n-t U - E H  kiin-tri-e  'liquor (<fire-water)' ALIGNFTPARSEn FOOT-R BINu  10  MAXPATH  *  *!  **  *!  **  ALIGNTONEL  DEPPATH  1 uu nn a. (k u'n)(t u E ) H H  1 b.  /  uu nu (kun)(tu£) H  **  H  1  *!  1  uu u u c. (k ii n)(t u E ) H H  1  "3"d.  *  **  *  / i  CC C C (k u n)(t u E )  A s i n the t a b l e a u x f o r N a v a j o , a l i g n m e n t v i o l a t i o n s are a s s e s s e d b y m o r a s . E a c h c a n d i d a t e i n c u r s t w o v i o l a t i o n s o f A L I G N - F O O T b e c a u s e t h e l e f t m o s t f o o t i s n o t a l i g n e d w i t h the r i g h t e d g e o f t h e p r o s o d i c w o r d . I n c a n d i d a t e ( a ) , t h e i n h e r e n t h i g h t o n e o n t h e f i n a l m o r a a n d its c o r r e s p o n d i n g l i n k have been deleted, v i o l a t i n g M A X - I O and M A X - P A T H .  I n candidate (b),  the i n h e r e n t h i g h tone has s h i f t e d l e f t w a r d s f r o m the p o s s e s s i v e s u f f i x to the adjacent root. T h i s is a v i o l a t i o n o f b o t h M A X - P A T H a n d D E P - P A T H .  C a n d i d a t e (c) is c o m p l e t e l y f a i t h f u l  to t h e i n p u t ; h o w e v e r , i t i s e l i m i n a t e d b y A L I G N - T O N E - L , s i n c e t h e t o n e h a s n o t a l i g n e d to t h e left e d g e o f the foot.  T h e o p t i m a l c a n d i d a t e m a i n t a i n s t h e i n h e r e n t h i g h t o n e , a n d the t o n e h a s  spread w i t h i n the foot.  A c o m p a r i s o n o f (c) a n d (d) s h o w s c r u c i a l r a n k i n g b e t w e e n the  alignment constraint A L I G N - T O N E - L and D E P - P A T H .  6.8  Conclusion  T h i s chapter has e x a m i n e d t w o instances o f tone a s s i m i l a t i o n , one i n N a v a j o w h e r e inherent h i g h t o n e s p r e a d s f r o m l e f t to r i g h t i n v e r b s , a n d o n e i n C h i p e w y a n , w h e r e i n h e r e n t h i g h t o n e spreads f r o m r i g h t to left i n n o u n s . F i r s t , tone a s s i m i l a t i o n i n N a v a j o o c c u r s w h e n inherent h i g h tone o n c e r t a i n p r e f i x e s spreads to a f o l l o w i n g C V s y l l a b l e . T o n e a s s i m i l a t i o n d o e s not take p l a c e i f there is m o r e t h a n one C V s y l l a b l e to t h e r i g h t o f t h e h i g h t o n e p r e f i x , o r i f t h e t a r g e t p r e f i x c o n t a i n s a l o n g v o w e l . F u r t h e r , it w i l l n o t s p r e a d i n t o the v e r b stem/root.  T o n e a s s i m i l a t i o n c a n be a n a l y s e d i n  O p t i m a l i t y T h e o r y u s i n g a m o r a i c , foot-based analysis, where tone w i l l spread w i t h i n a binary t r o c h a i c foot, w h e n the r i g h t e d g e o f the foot is a l i g n e d w i t h the r i g h t e d g e o f the p r o s o d i c w o r d . T h e m a x i m a l tone s p r e a d is t w o m o r a s . A c o m p l e t e l y f a i t h f u l c a n d i d a t e w i t h n o tone a s s i m i l a t i o n e m e r g e s w h e n t h e f o o t c o n t a i n i n g the i n h e r e n t t o n e i s n o t a l i g n e d w i t h t h e  88  boundary between the prefixes and the verb stem/root. Finally, the analysis suggests that prefix coda consonants are not moraic in Navajo whereas root coda consonants can be, and the root/stem is independently footed due to its marked status. One significant issue raised by this analysis is that reference to both prosodic and morphological structure is required to explain tone assimilation in Navajo. Furthermore, the morphological division between disjunct and conjunct is superseded by prosodic structure, but the morphological domain between the prefixes and the verb stem is not. The second instance of tone assimilation has been illustrated by Chipewyan. Addition of the possessive suffix -e in Chipewyan may trigger a process of tone assimilation, where high tone will spread to the vowel of the root, although the number of nouns which take the possessive suffix are very limited, so examples of the assimilation process are rare. The process can be accounted for with the same set of constraints used to account for Navajo, with adjustments to the specified arguments of the alignment constraints, but no reranking of the constraints.  89  Chapter 7 Conclusions and Issues for Further Research The goals of this thesis included the following: (i) presentation of data from a previouslyundocumented dialect of Chipewyan to bring new evidence to bear on the cross-linguistic picture of Athapaskan languages; (ii) investigation of the acoustic properties of Chipewyan obstruents, focusing mainly on the stop series; (iii) identification of specific issues related to the phonetics/phonology interface; and (iv) examination of several morphophonemic processes from a cross-linguistic and comparative historical perspective to test the tenets of feature specification, privative features, constraint definition and interaction. Chapter two has summarised a variety of historical, comparative and theoretical issues to serve as the essential background information relevant to issues presented in this thesis, and has outlined diachronic and dialectal variation in the research dialect of Chipewyan. First, a review of the Proto-Athapaskan linguistic literature has provided the historical context in which to situate Chipewyan. Issues raised in the literature have become the focus of subsequent chapters. The Chipewyan consonant inventory, introduced in chapter two, has been shown to be different from the posited Proto-Athapaskan inventory in several ways, discussed in greater detail in chapter three. Other literature introduced in chapter two relevant to the topic of laryngeal feature specification is that of Rice (1994). Issues raised by Rice (1994) have been shown to be relevant to the discussion of laryngeal neutralisation in chapter four, and continuant voicing in chapter five. Key points of the historical literature have provided a starting point for the processes of laryngeal neutralisation and spirantisation in chapter four. The two-way laryngeal contrast stem-finally in Proto-Athapaskan has been neutralised in some of the daughter languages, while other languages have lost all stem-final stops. A s we have seen, loss of stem-final laryngealisation is also closely related to the development of tone in Athapaskan languages; tone assimilation processes have been analysed in chapter six. To complete the overview in chapter two, the vowel inventory of the research dialect Chipewyan has been described, comparing it to previous inventories documented in the literature. A n d , we have reviewed the basic structure of Chipewyan morphology, discussing the role played by morphology in the phonological processes discussed in the remaining chapters. The goals of chapter three have been to present the consonant inventory of Chipewyan and accompanying acoustic analysis, and to investigate the laryngeal properties of voiceless unaspirated stops to substantiate a laryngeal specification of the obstruents of Chipewyan. Wave form and spectrogram analyses have revealed several interesting points about the consonant inventory of the research dialect of Chipewyan distinct from the dialect documented by L i (1933, 1946). First of all, Athapaskan languages are traditionally described as having three types of stops: voiceless unaspirated, voiceless aspirated and glottalised. Analysis of the voiceless unaspirated (plain) stops has shown that specific representatives of voiceless unaspirated stops, the bilabial [b] and the dental [d], are consistently voiced, rather than  90  voiceless unaspirated. The alveolar and palatal plain affricates are intermediate cases which are not consistently voiced or voiceless throughout the stop duration. The remaining plain stops/affricates are voiceless unaspirated. The generalisation which can be drawn from these results is that there is a front-back distinction in the laryngeal property of voicing of the plain stops, with a cline of voicing from the front to the back of the mouth. This front-back asymmetry can be accounted for by an aerodynamic explanation involving the difference in the volume of the oral cavity between the stop closure and the glottis (Kingston 1996). The further back a stop is articulated, the smaller the initial volume of the oral cavity behind the closure will be. Maintaining vibration of the vocal folds requires that the air flow up from the sub- to the supra-glottal cavity between the vocal folds, which necessitates that the pressure o f the air below the glottis must be greater than that above. Having greater volume in the oral cavity initially, or increasing that volume, slows the increase in supra-glottal air pressure, which means that a stop is more likely to remain voiced through a closure with a more forward articulation (Kingston 1996:51). This explanation can account for why the two most anterior stops, the bilabial and dental, are voiced throughout the stop duration whereas the other plain stops are not completely voiced, or not voiced at all. These findings make an interesting claim regarding the interface of phonetics and phonology. Stops which behave as a class phonologically are shown to have two different phonetic realisations, showing that the phonology is not a direct reflection of phonetics; it abstracts beyond phonetic realities to a single phonological class. Since voiced and voiceless stops pattern together, the voicing feature is not distinctive phonologically as it is for the fricatives. A second finding of chapter three involved consonant mergers. In the research dialect, it appears that the traditionally classified voiceless unaspirated interdental affricate [d6] has merged with the voiced interdental fricative [6]. Likewise, the voiceless aspirated interdental affricate [t6] has merged with the voiceless interdental fricative [0]. However, these fricatives continue to pattern with the stops in phonological processes such as continuant voicing. This has controversial implications for the featural specifications: how are consonants which are unmarked for the feature [stop] exempt from patterning with other consonants unmarked for [stop]? This issue is left open to future research. Another consonant merger involves the alveolar affricates [dz, c, c] which are merging with the palatal affricates Q, c, 6] but this merger is not complete. The only labiovelar attested is the voiceless aspirated [k*]. This is of limited phonemic status. These constitute further examples of diachronic change. Regarding the fricatives, spectrogram analysis has shown that [1], the voiced correspondent of the voiceless lateral fricative [T], is definitely more like a fricative than an approximant (sonorant), but [y], the voiced correspondent of the voiceless alveopalatal fricative [s], is clearly more approximant-like. This provides another example of a mismatch between phonetics and phonology. Two phonemes, [s] and [y], with different phonetic realisations (continuant and sonorant respectively), pattern together phonologically.  91  The velar fricative [y] in the research dialect is often replaced by other consonants including [y], [w], [g] and [h], in various contexts. It is usually maintained intervocalically. In chapters four through six, we have seen several phonological processes concerning laryngeal properties examined in a current theoretical context, that of Optimality Theory (McCarthy and Prince 1993). The goal of chapter four has been to test glottalisation evidence from the Athapaskan language family against Steriade's Licensing by Cue hypothesis which governs the realisation of laryngeal contrasts. Some of the glottalisation evidence from Athapaskan supports Steriade's (1997) analysis. In Athapaskan languages, glottalised stops are retained stem-initially, as exemplified by Chipewyan. This result is predicted from Steriade's hypothesis: optimal identification of glottalisation contrasts is dependent on the right-hand context such as a following vowel. Conversely, identification of glottalisation is not optimal when the right-hand context is not a vowel, i.e. the end of the word. This is shown by evidence of laryngeal neutralisation in Koyukon. Evidence from Hupa, however, differs. In Hupa, glottalisation contrasts are maintained word-finally; since there is no following vowel, this is not the optimal context for retention of glottalisation. While this case is less common than the previous ones, Steriade's hypothesis can accommodate this through the relative ranking of constraints; in Hupa the faithfulness condition Preserve [CG] is undominated, and therefore ranked above the cue constraints. In several Athapaskan languages including Chipewyan, no stops are maintained in syllablefinal condition; a process of spirantisation has occurred. Cross-linguistically, spirantisation most typically occurs in a postvocalic context. Spirantisation in Athapaskan involves neutralisation of manner as well as laryngeal features. Steriade's analysis of laryngeal neutralization contexts does not address cases where manner is neutralised. Spirantisation in Chipewyan can be accounted for within Optimality Theory by resorting to a syllable-sensitive coda condition banning stops in syllable-final position. The goal of chapter five has been to investigate the synchronic process of continuant voicing in Chipewyan. In this phonological process, a noun stem-initial voiceless continuant becomes voiced when the possessive prefix is added. Optimality Theory can account for the process' affecting stem-initial continuants, and the same set of constraints will not affect stem-initial stops, whether voiceless or voiced. The research dialect of Chipewyan does not provide clear evidence of a voicing alternation due to the non-occurrence of paradigmatically-related forms in contexts which would crucially attest to the synchronic validity of the process. Most alienable nouns take a postpositional form to indicate possession, and inalienable forms never occur without the possessive prefix, so it is not possible to determine the underlying form. There are two possible options to consider (Patricia A . Shaw, p.c): (i) restructuring of the lexicon; all inalienable stems have been restructured with a voiced stem-initial continuant. Or, (ii) inalienable stems still are voiceless in the lexicon but all undergo continuant voicing as in other dialects. In both cases, there has been a restructuring in the morphosyntactic system of how possession is marked. The prefix set subcategorises for inalienable noun roots, whereas the postpositions subcategorise for alienable  92  noun roots. This makes the research dialect different from other documented dialects of Chipewyan which most commonly use the possessive prefix to indicate possession. Chapter six has examined two instances of tone assimilation in Athapaskan languages in different morphological domains: tone assimilation in Navajo, where inherent high tone spreads from left to right in verbs, and tone assimilation in Chipewyan, where inherent high tone spreads from right to left in nouns. Tone and tonal processes are the synchronic residue of Proto-Athapaskan laryngeal behaviour. In Navajo, tone assimilation occurs when inherent high tone on certain prefixes spreads to a following C V syllable. Tone assimilation does not take place if there is more than one C V syllable to the right of the high tone prefix, or if the target prefix contains a long vowel. It will not spread into the verb stem/root. Tone assimilation can be analysed in Optimality Theory using a moraic, foot-based analysis, where tone will spread within a binary trochaic foot, when the right edge of the foot is aligned with the right edge of the prosodic word. The maximal tone spread is two moras. A completely faithful candidate with no tone assimilation emerges when the foot containing the inherent tone is not aligned with the boundary between the prefixes and the verb stem/root. Finally, the analysis suggests that prefix coda consonants are not moraic in Navajo whereas root coda consonants can be, and the root/stem is independently footed due to its marked status. One significant issue raised by this analysis is that reference to both prosodic and morphological structure is required to explain tone assimilation in Navajo. Furthermore, the morphological division between disjunct and conjunct is superseded by prosodic structure, but the morphological domain between the prefixes and the verb stem is not. In Chipewyan, addition of the possessive suffix -e may trigger a process of tone assimilation, where high tone will spread to the vowel of the root, although the number of nouns which take the possessive suffix are very limited, so examples of the assimilation process are rare. The process can be accounted for with the same set of constraints used to account for Navajo, with adjustments to the specified arguments of the alignment constraints but no re-ranking of the constraints. This thesis has shown significant changes between the research dialect of Chipewyan and previously documented dialects, including both changes in the phoneme inventory, and changes in the phonological process of continuant voicing. One of the main issues these findings raise for future research concerns cross-linguistic comparisons. Is the voicing of voiceless unaspirated stops seen in languages such as Tahltan and Chipewyan a pan-Athapaskan phenomenon, or is it only occurring in certain languages of the family? Furthermore, what are the reasons behind these substantive changes? Possible explanations include natural language change over time, restructuring of the language under pressures of language loss or death, or accelerated language change due to language contact and bilingual influence. Finally, the findings raise several issues regarding the phonetics-phonology interface. The evidence of phonological patterning independent of, and sometimes contrary to, the phonetic realisations of the consonant focuses on important issues for future research.  93  References Archangeli, Diana and Douglas Pulleyblank.  1994.  Grounded Phonology. Cambridge:  M I T Press. Bob, Tanya. 1999. Tahltan morphophonemic processes in Optimality Theory. In M . Caldecott, S. Gessner and E. K i m (eds.)  University of British Columbia Working  Papers in Linguistics 1: 1-18. Bob, Tanya. 1999. Laryngeal Phenomena in Tahltan. M . A . Thesis, University of British Columbia. Blevins, Juliette. 1995. The syllable in phonological theory. In J. Goldsmith (ed.) T i e  Handbook of Phonological Theory. Cambridge: Blackwell Publishers Ltd, 206244. Cook, Eung-Do. 1983. Chipewyan vowels.  International Journal of American  Linguistics 49: 413-427. Cook, Eung-Do. 1984.  A Sarcee Grammar. Vancouver: University of British Columbia  Press. Cook, Eung-Do. 1986. Sound change, morphological change, and paradigm regularity in Chipewyan. M s . , University of Calgary. Cook, Eung-Do. 1991. Linguistic divergence in Fort Chipewyan.  Language in Society  20, 423-440. Cook, Eung-Do. 1997. The phonological status of conjunct prefix vowels. Conference handout, Athapaskan Language Conference, University of Oregon, May 17-18, 1997.  UCLA Working Papers in Phonetics 84: 165-182. Elford, Leon and Marjorie Elford. 1981. Chipewyan Dictionary. Prince Albert: Northern de Jong, K e n and Joyce McDonough. 1993. Tone and tonogenesis in Navajo.  Canadian Evangelical Mission. Gessner, Suzanne. 1998-1999. Chipewyan fieldnotes. Gessner, Suzanne. 1999. Tone assimilation in Navajo. In M . Caldecott, S. Gessner and E. K i m (eds.)  University of British Columbia Working Papers in Linguistics 1:  33-48. Golla, Victor. 1977. A note on Hupa verb stems.  International Journal of American  Linguistics 43: 355-358. Golla, Victor. 1996. Sketch of Hupa, an Athapaskan language. In I. Goddard (ed.)  Handbook of North American Indians V . 17. Washington: Smithsonian Institution, 364-389. Hargus, Sharon. 1988.  The Lexical Phonology of Sekani. New York: Garland Publishing  Co. Hass, Mary.  1968. Notes on a Chipewyan dialect.  International Journal of American  Linguistics 34: 168-175. Helm, June (ed.). 1981.  Handbook of North American Indians V . 6. Washington:  Smithsonian Institution. Hogan, John. 1976. A n analysis of the temporal features of ejective consonants.  Phonetica 33: 275-284. Howe, Darin. 1998. Markedness-as-faithfulness in the interaction of [constricted glottis] and [sonorant]. M s . , University of British Columbia.  94  Kaisse, Ellen. 1999. Are laterals [-continuant]? University of British Columbia Colloquium Series, March 15, 1999. Kari, James. 1976. Navajo  Verb Prefix Phonology. New York: Garland. Generative Grammar. Cambridge: Blackwell  Kenstowicz, Michael. 1994. Phonology in Publishers Inc.  K i m , Eun-Sook. 1998. Korean partial reduplication in Optimality Theory. M s . , University of British Columbia. Kingston, John. 1996. Introduction to phonetic theory. M s . , University of Massachusetts, Amherst. Krauss, Michael. 1964. Pro to-Athapaskan-Eyak and the problem of Na-Dene: the phonology.  International Journal of American Linguistics 30: 118-131.  Krauss, Michael. 1978. Athapaskan tone. M s . , Alaska Native Language Center. Krauss, Michael. 1982. Proto-Athapaskan * k in Chipewyan, 1742-1800: philological evidence.  International Journal of American Linguistics 48: 73-82.  Krauss, Michael and Victor Golla. 1981. Northern Athapaskan languages. In J. Helm  (ed.) Handbook of North American Indians. V . 6. Washington: Smithsonian Institution, 67-85. Krauss, Michael and Jeff Leer. 1981.  Athapaskan, Eyak and Tlingit sonorants.  Fairbanks: Alaska Native Language Center. Ladefoged, Peter. 1975.  A Course in Phonetics. New York: Harcourt Brace Jovanovich,  Inc. Ladefoged, Peter and Ian Maddieson. 1996.  The Sounds of the World's Languages.  Oxford: Blackwell Publishers Ltd. Leer, Jeff. 1979.  Proto-Athapaskan Verb Stem Variation, Part One: Phonology.  Fairbanks: Alaska Native Language Center. L i , Fang-Kuei. 1933. A list of Chipewyan stems.  International Journal of American  Linguistics 7: 122-151.  Linguistic Structures of Native  L i , Fang-Kuei. 1946. Chipewyan. In H. Hoijer (ed.) America. New York: Viking Fund, 398-423.  L i , Fang-Kuei. 1956. A type of noun formation in Athapaskan and Eyak.  International  Journal of American Linguistics 22: 45-48. L i , Fang-Kuei. 1964. A Chipewyan ethnological text.  International Journal of American  Linguistics 30: 132-136. Macken, Marlys. 1995. Phonological acquisition. In J. Goldsmith (ed.) The Handbook of  Phonological Theory. Cambridge: Blackwell Publishers Ltd., 671-698. McCarthy, John. 1981. The representation of consonant length in Hebrew. Linguistic Inquiry 12:322-327. McCarthy , John and Alan Prince. 1993. Generalised alignment. In Geert Booij & Jaap van Marie (eds.) Yearbook of Morphology.  Dordrecht: Kluwer, 79-153.  McCarthy , John and Alan Prince. 1994. The emergence of the unmarked: in Prosodic Morphology.  Optimality  NELS 24, 333-379.  McCarthy , John and Alan Prince. 1995. Faithfulness and reduplicative identity. Papers in Optimality Theory,  University of Massachusetts Occasional Papers 18: 249-  384. McDonough, Joyce. 1990.  Topics in the Phonology and Morphology of Navajo Verbs.  PhD dissertation, University of Massachusetts. Amherst: G L S A .  95  Myers, Scott. 1997. O C P effects in Optimality Theory.  Natural Language and  Linguistic Theory 15:847-892. International Journal  Nater, Hank. 1989. Some comments on the phonology of Tahltan.  of American Linguistics 55: 25-42. Tone in Lexical Phonology. Dordrecht: D. Reidel Publishing  Pulleyblank, Douglas. 1986. Company.  Pulleyblank, Douglas. 1996. Neutral vowels in Optimality Theory: a comparison o f  Canadian Journal of Linguistics 41: 295-347. International Journal of American Linguistics 44:144-145. Yoruba and Wolof.  Rice, Keren. 1978. A note on Fort Resolution Chipewyan.  Rice, Keren. 1988. Continuant voicing in Slave (Northern Athapaskan): the cyclic application of default rules. In M . Hammond and M . Noonan (eds.) Theoretical Morphology. New York: Academic Press, 371-388. Rice, Keren. 1989. A  Grammar of Slave. Berlin: Mouton de Gruyter.  Rice, Keren. 1991. Prosodic constituency in Hare (Athapaskan): evidence for the foot.  Lingua 82: 201-245. Rice, Keren. 1994. Laryngeal features in Athapaskan languages.  Phonology 11: 107-  147. Richardson, Murray.  1968.  Chipewyan Grammar. Northern Canadian Evangelical  Mission, C o l d Lake, Alberta. Shaw, Patricia. 1981-1982. Tahltan fieldnotes. Shaw, Patricia. 1987. Non-conservation of melodic structure in reduplication. In A . Bosch et al. (eds.)  Papers from the Parasession on Autosegmental and Metrical Phonology.  C L S , 291-306. Shaw, Patricia. 1991. Consonant harmony systems: the special status of coronal harmony. In  Phonetics and Phonology, Volume 2, The Special Status of Coronals: Internal and External Evidence. San Diego Academic Press Inc.,  C. Paradis and J.-F. Prunet (ed.) 125-157.  Shaw, Patricia. 1992. Templatic evidence for the syllable nucleus.  Proceedings of NELS 23.  Amherst: G L S A . Shaw, Patricia. 1994. Minimality and markedness. Invited paper presented at the Workshop in Prosodic Morphology, Utrecht, Netherlands. Shaw, Patricia. 1998. O n serial derivation. Paper presented at the Workshop on Structure and Constituency in Languages of the Americas, Regina, Saskatchewan. Shaw, Patricia. 1999. Reduplicative non-identity in Nisqa'a. Paper presented at S S I L A summer meeting, Urbana-Champaign, Illinois. Steriade, Donca. 1997. Phonetics in phonology: the case of laryngeal neutralisation. M s . , UCLA. Taylor, Douglas. 1978. Consonantal patterns in Arawakan.  International Journal of  American Linguistics 44:121 -130. Y i p , Moira. 1995. Tone in East Asian languages. In J. Goldsmith (ed.) The Handbook of  Phonological Theory. Cambridge: Blackwell Publishers Ltd., 476-494. Young, Robert and William Morgan. 1980. The Navajo Language. Albuquerque: University of New Mexico Press. Young, Robert and William Morgan. 1992.  Analytical Lexicon of Navajo. Albuquerque:  University of New Mexico Press.  96  , Draga. 1988.  Sonority Constraints on Prosodic Structure. PhD dissertation, Stanford  University.  97  Appendix Figure 2: Initial [b]; [bis] 'knife'  Figure 3: Intervocalic [b]; [tabi] 'net'  Figure 4 : Initial [d]; dene 'people'  98  Figure 5: Initial [t]; tede 'bed'  4600  4700  4800  4900 T  »«ii«WN  I  -iii i  Figure 6: Initial [fj; teG 'coal'  iv f (  1  Figure 7: Intervocalic [d]; ?edii 'hot'  -200 -2540  2000  21 no  2200  2300  2400  m  Hz  •4100-3200•2300-  0inn, •  •1400•500-  ttt.  99  2500  Figure 8: Intervocalic [t]; yati 'language'  Figure 9: Intervocalic ft]; det'anco  , 100  7200  'eagle'  7300  7400  7500  76Q0  7700  7800  7900  3000  8  Hz •41001  ii  mm-  f  Figure 10: Initial [t0]; t©e 'stone/rock'  -240  7  II  2220 16100  16200  16500  16400  165C :|o  (71  H  lUhhMiii-  4100 •3200•2300••1400-  y  •500 100  16600  16700  16800  Figure 11: Initial [0]; t©i?ah 'pillow'  Figure 12: Initial [t'0]; t'0ay 'bowl'  Figure 13: Initial [t'0]; t'0uwe 'milk'  101  Figure 14: Intervocalic [dS]; ?ed6ay 'smoked/dried fish'  4800  4900  5000  5100  5200  1 I'W^I Hz  it  •4100  ...wi,..., i p)ll  11  • i  Hi'1  f  •3200•1400  "•^•H«M-  ..ijyyyyywi • kl j i ii. -i  L  '•it"  i  Figure 15: Intervocalic [tG]; ?et6en 'caribou'  Hz  •4100-  -2300•1400•500 Figure 16: Intervocalic [t'6]; sit'eene 'my bone'  lldLUujU.ll ,,  LL  i  5300  Figure 17: Initial [dz]; [dzen] 'muskrat/rat'  Figure 18: Initial [dz]; [dz?] 'gum'  Figure 19: Initial [c]; [c£] 'excrement'  10000  103  Figure 20: Initial [c]; [ca] 'hat'  4300  Figure 21: Intervocalic [dz]; [?edzah] 'cold'  Figure 22: Intervocalic [c]; [stci] 'my nose'  104  Figure 23: Intervocalic [c]; [sicane] 'my wife'  11000  Figure 24: Initial [X]; [>ojne] 'mouse'  Figure 25: Initial [X]; [Xis] 'grease'  9100  105  Figure 26: Initial [t]; [tu] 'rope'  1 0900  10900  11000  11100  11200  11300  1 1400  Hz  •41 00•3200•2300•1400•500 Figure 27: Intervocalic [k]; [naXoi] 'smile/laugh'  f-1240528012400  12500  12600  12700  Hz ••41 00•3200 • •2300•1400•500  Figure 28: Intervocalic [t]; [neXah] 'your buttocks'  12800  12900  13000  13100  13200  13  Figure 29: Initial [j]; Qa] 'here'  Figure 30: Initial [}]; [Jole] 'ball'  1  400  500  111--  Figure 31: Initial [c]; [ca] 'rain'  107  -_L_i  .  600  7(  Figure  32: I n i t i a l  [cj;  [c-eltue]  'tobacco/cigarettes'  F i g u r e 33: I n t e r v o c a l i c [j]; [?e]ere] ' c o w '  Figure  34:  Intervocalic  5200  [CJ; [dac4]  'raven/crow'  5300  5400  5500  5600  5700  5S00  59C  Figure 35: Initial [g]; [gokus] 'pig'  Figure 36: Initial [g]; [gah] 'rabbit'  9200  9300  9400  9500  9600  9700  9800  9900  Figure 37: Initial [k]; [kd£] 'house'  12900  13000  15-, 00  15200  13500  15400 15500 -—r  Hz  •4100IJHj  5200 •2300•1400•500 109  !J„j§..  13600  1  Figure 38: Initial [k]; [ko0] 'clouds'  -4120 17300  17400  17500  i1  *  Hz  0  •-4100-•3200 • ••2300-  i  1  •1400•500  Figure 39: Intervocalic [g]; [yagole] 'butterfly' -440 -6200 2000  2100  2300  Hz -4100-  2400  • \  *1  • j  1  I tlilllllljltfr  1  idiiUiikiiiii  m  Figure 40: Intervocalic [k]; [hokamo] 'it tastes good'  1 I .lUill^l'^^^^ *^  -1600  •^•••••'IH^^^ffllrlllf^  - S420 5000  5100  5200  j  Hz •4100  5500  1  li  •2300  •500-  5400  L  •3200  •1400  5300  i  "  f  1  ,..|....  {  J  5600  5700  5800  5900  Figure 41: Intervocalic [k]; [sikdG] 'my neck' -140^  i..  -2580-  3000  8100  3200  8300  8400  8500  8600  8700  Hz  8800  8900  ill  -4100•3200  1  •2300  1 M  j  Iflf  1  •1400 ••500-  Figure 42: Labiovelar [k ]; [sik s] 'children' w  w  ••.....AAN.,WIWI^^  .•i;  -4960  •'• 18500  1 8400  18600  18700  Figure 43: Labiovelar [k ]; [cek aze] 'girl, young woman' w  w  111  18800  18900  19000  Figure 44: Initial [?]; [?asi] 'something' HO -fl 440  13100  13200  13300  13400  13500  13600  13700  13300  Figure 45: Post-consonantal [?]; [har^s?j] 'I want'  Figure 46: Initial [0]; [0e] 'ice' mil 1rclM  w u w ~  17600 ii  17700  17800  llllililll' K I I fi  17900  llll  1 8000 I ... I  Hz •4100-3200•2300-  •500-  —  IBB  "  •1400-  -  t piMttMlttltili i llHUkfcl.1  It.  112  18100  1 8200  1 8300  1:  Figure 47: Initial [6]; [de] 'skin*  -2620 9800  9500  9900  10000  1 01 QCj  Hz  •41 0 0 •3200•2300 • •1400--500-  LL  Figure 48: Initial [i]; [<ia] 'many'  540™ 5240-  1 0800  10700  10600  10500  1 0900  1 1 00  11  Hz  i4-  •4100  ll ii I  •3200 •2300 if-'-  • MM  •500Figure 49: Initial [1]; [lah] 'work'  17000  17100  17200  Hz -4100" •3200 • -2300 • •1400  §i|11pi  m  ........  T  •500  113  17300  17400  17500  17600  17700  Figure 50: Initial [y]; [yake] 'window'  Figure 51: Intervocalic [y]; [tGiyah] 'hair'  14400  1 4500  14600  114  1 4700  1 4800  1 4900  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0089167/manifest

Comment

Related Items